A comprehensive history of motility and Archaellation in Archaea

ABSTRACT

Each of the three Domains of life, Eukarya, Bacteria and Archaea, have swimming structures that were all originally called flagella, despite the fact that none were evolutionarily related to either of the other two. Surprisingly, this was true even in the two prokaryotic Domains of Bacteria and Archaea. Beginning in the 1980s, evidence gradually accumulated that convincingly demonstrated that the motility organelle in Archaea was unrelated to that found in Bacteria, but surprisingly shared significant similarities to type IV pili. This information culminated in the proposal, in 2012, that the ‘archaeal flagellum’ be assigned a new name, the archaellum. In this review, we provide a historical overview on archaella and motility research in Archaea, beginning with the first simple observations of motile extreme halophilic archaea a century ago up to state-of-the-art cryo-tomography of the archaellum motor complex and filament observed today. In addition to structural and biochemical data which revealed the archaellum to be a type IV pilus-like structure repurposed as a rotating nanomachine (Beeby et al. 2020), we also review the initial discoveries and subsequent advances using a wide variety of approaches to reveal: complex regulatory events that lead to the assembly of the archaellum filaments (archaellation); the roles of the various archaellum proteins; key post-translational modifications of the archaellum structural subunits; evolutionary relationships; functions of archaella other than motility and the biotechnological potential of this fascinating structure. The progress made in understanding the structure and assembly of the archaellum is highlighted by comparing early models to what is known today.

INTRODUCTION

In 1977, Carl Woese and George Fox published a landmark paper where they described, based on analysis of rRNA sequence characterization, that all living systems belonged to three lines of evolutionary descent, which were named in this early paper the urkingdoms of Urkaryotes, Eubacteria and Archaebacteria (Woese and Fox 1977). At this time, the only members of archaebacteria were a very limited number of methanogens but they would be soon joined by extreme halophiles, Sulfolobus and Thermoplasma (Woese, Magrum and Fox 1978; Magrum, Luehrsen and Woese 1978; Fox et al. 1980). By 1990, Woese et al. (Woese, Kandler and Wheelis 1990) had revised the urkingdoms of life to Domains, renamed the three groupings to their still currently used names Eukarya, Bacteria and Archaea and produced the universal phylogenetic tree based upon rRNA sequence comparisons that would go on to grace most microbiology textbooks.

When archaea were shown to be motile and have tail-like appendages that superficially resembled the flagella of bacteria, it was generally assumed that the archaeal ‘flagella’ were very similar to the bacterial organelles, although perhaps with modifications that allowed for increased stability since archaea were found in those early days almost exclusively in extreme environments of high salt, low pH and/or high temperatures. However, subsequent studies from many laboratories have clearly shown that all three domains have their own evolutionary distinct version of ‘flagella’, commonly now referred to as cilia in eukaryotes, flagella in bacteria and archaella in archaea (Khan and Scholey 2018; Beeby et al. 2020). Archaella are distributed widely throughout the domain Archaea with individual species having a distinct number and location of archaella on their cell surface (Table 1 and Fig. 1). Although a rotary organelle, the archaellum was shown to be evolutionarily related to type IV pili and completely unrelated to flagella, and early studies revealed that archaea did not have homologs of flagellum protein genes in their genomes (Faguy and Jarrell 1999). Many review articles have covered the similarities and differences of archaella, flagella and type IV pili (sometimes including cilia) in great detail (Table 2) (Ng, Chaban and Jarrell 2006; Jarrell and McBride 2008; Pohlschroder et al. 2011; Khan and Scholey 2018; Albers and Jarrell 2018; Denise, Abby and Rocha 2019; Beeby et al. 2020). We now know a considerable amount about archaella, the sole motility system identified in Archaea. Roles for most of the individual archaellum-related proteins are known: the filament proteins, the stator, and the motor proteins as well as archaea-specific proteins that connect the bacterial-like chemotaxis system of the archaea to the archaellum. Much is known also about the post-translational modifications of archaellins, especially the importance of N-linked glycosylation, and increasingly more about regulation of the arl (formerly fla) gene cluster in different model organisms, archaellum structure, anchoring and assembly, the various functions of archaella, and crosstalk to other components of the cell such as pili and biofilm formation.

Table 1.

Archaella distribution on selected archaea from a variety of habitats.

Organism	Habitat	Archaella Distribution
Methanogens
Methanocaldococcus vulcanius	Hydrothermal vent	Three tufts
Methanococcus maripaludis	Salt marsh	Large bundle
Methanospirillum hungatei	Anaerobic digestor	Polar tufts
Methanococcoides burtonii	Polar regions	Monotrichous
Methanonatronarchaeum thermophilum	Hypersaline alkaline lakes	1–5 peritrichous
Extreme halophiles
Halobacterium salinarum	Salted fish and hides	Mono- or bi-polar bundles
Haloarcula quadrata	Brine pool	One or more archaella at one or more sites
Haloferax volcanii	Dead Sea	Mono- or bi-polar
Natronomonas pharaonis	Soda lakes	Polar tuft
Natrialba magadii	Soda lake	Polar tuft
Thermoacidophiles
Thermoplasma volcanium	Solfataras, hot springs	Polar monotrichous
Sulfolobus solfataricus	Solfataras, hot springs	One or more
Metallosphaera sedula	Thermal pond	One or more
Hyperthermophiles
Pyrococcus furiosus	Hydrothermal vent	Monopolar polytrichous
Thermococcus kodakaerensis	Hydrothermal vent	Polar tufts
Archaeoglobus fulgidus	Hydrothermal vent	Monopolar polytrichous
Desulfurococcus fermentens	Freshwater hot springs	Monotrichous
‘Geogemma barossii’	Hydrothermal vent	Lophotrichous
Ammonia-oxidizing Thaumarchaea
Nitrosopumilus adriaticus	Adriatic Sea	Occasionally archaellated
Nitrososphaera viennensis	Garden soil	One or more
Ca Nitrosotenuis uzonensis	Thermal spring	One or two polar/subpolar
Nanoarchaea
Ca Nanoclepta minutus	Hot spring	Lophotrichous bundle

Figure 1.

Electron micrographs highlighting the variety in archaella number and cellular location in selected archaea. S. acidocaldarius, from Jarrell and Albers (2019) Archaellum In: Schmidt, T.M. (ed.) Encyclopedia of Microbiology, 4^th Edition vol 1 pp253-261. UK: Elsevier, with permission. Mc. maripaludis Courtesy of S.-I. Aizawa. Ca. Nanoclepta minutus, Courtesy of A. Reysenbach. A. fulgidus From Jarrell et al. 2007. Flagellation and chemotaxis In: Cavicchioli, R. (Ed.) Archaea: molecular and cellular biology pp 385–410, with permission. Original picture courtesy of R. Rachel. Hfx. volcanii From Li et al. (2019). Mbio 10: e00377-19, with permission. P. furiosus From Jarrell et al. 2007. Flagellation and chemotaxis In: Cavicchioli, R. (Ed.) Archaea: molecular and cellular biology pp 385–410, with permission. Original picture courtesy of R. Rachel. T. acidophilum From Black et al. (1979) J. Bacteriol. 137:456–460, with permission. Ca. Nitrosotenuis uzonensis. From Lebedeva et al. 2013. PLoS One 8: e80835 Picture supplied by R. Hatzenpichler. H. salinarum From Alam and Oesterhelt (1984). J. Mol. Biol. 176:459–475 with permission.

Table 2.

Comparison of archaella, flagella and type IV pili.

Trait	Flagella	Archaella	Type IV Pili
Major structural proteins	Usually single flagellin, though many exceptions	Usually multiple archaellins, with some exceptions	Usually single major pilin, multiple minor pilins
Sequence similarity among structural proteins	No sequence similarities of flagellins to type IV pilins or archaellins	N-terminal sequence similarities of archaellins to type IV pilins	N-terminal sequence similarities of type IV pilins to archaellins
Signal peptides on structural proteins	Flagellins lack signal peptides	Class III signal peptides on archaellins: removed by FlaK/PibD (homolog of PilD)	Class III signal peptides on pilins: removed by PilD
Glycosylation of structural proteins	Flagellins may be glycoproteins: if so O-linked glycans	Archaellins typically glycoproteins: N-linked glycans, rarely O-linked	Pilins may be glycoproteins: if so O-linked glycans
Anchoring structure	Defined basal body anchoring structure with rings and rod	Knob-like anchoring structure	No observed anchoring structure
Filament diameter	18-22 nm	10–14 nm	5–7 nm
Internal channel in filament	Filament has internal channel allowing passage of flagellins	Channel not detected	Channel too narrow to allow passage of subunits
Filament growth	Growth of filament at distal end	Growth of filament at base	Growth of filament at base
Motility	Swimming motility	Swimming motility	Surface motility (twitching)
Nature of filament movement to generate motility	Rotating filament capable of switching direction of rotation	Rotating filament capable of switching direction of rotation	Movement by extension and retraction of filament at base
Energy for movement	Ion gradients (H+ or Na+)	ATP hydrolysis	ATP hydrolysis
Associated chemotaxis system	Associated chemotaxis system	Usually associated bacterial-like chemotaxis system	Sometimes associated chemotaxis system
Shared gene homologues	No shared genes with archaella or type IV pili system	Genes for assembly shared with type IV pili systems	Genes for assembly shared with archaella systems

In this contribution, we present an historical account of the work from many laboratories on a variety of often difficult to grow archaea, with rudimentary genetic manipulation options, that have led to our current understanding of this incredible nanomachine now known as the archaellum. While the term archaellum was only introduced in 2012 (Jarrell and Albers 2012), we shall use this term throughout the review to refer to the archaeal motility structure that was previously described as the archaeal flagellum. Similarly, we have, throughout the text, used the gene abbreviation arl (archaellum) as recommended by Pohlschroder et al. (Pohlschröder et al. 2018) to replace fla (flagellum) for archaella genes and proteins to complete the transition from archaeal flagellum to archaellum. In the recommended change from fla to arl, all of the fla genes and proteins retain their initial letter assignments (i.e. FlaA or flaA is now referred to as ArlA or arlA). We have commented only briefly on taxis in archaea and mainly only when it relates directly to archaellation. The long list of publications on chemotaxis and phototaxis in archaea, focused on studies on Halobacterium salinarum, are mainly from the laboratories of Oesterhelt, Spudich, Engelhard, Hazelbauer, Ordal and Stoeckenius but beyond the scope of this review (for a recent review of archaeal chemotaxis see Quax, Albers and Pfeiffer 2018a).

Early studies before the concept of Archaea

Key Findings: Descriptions of motility, archaellated cells and archaella location

The early studies on motility and archaellation in Archaea date back a century. Long before the concept of Archaea, scientists studied microorganisms that would later be classified as Archaea. These studies only dealt with extreme halophiles: other members of the archaea would not be isolated in pure culture until decades later.

Halophilic archaea have been studied for over a hundred years as the causative agent in the red discoloration and spoilage of salted hides, meat and fish. The first accurate description of organisms thought to be members of the Halobacteriaceae was likely the work of Klebahn (Klebahn 1919) (see (DasSarma, Klebahn and Klebahn 2010) for English translation) who described red, extremely halophilic, rod-shaped organisms (referred to as Bacillus halobius ruber) that lysed in distilled water, but Klebahn (Klebahn 1919) failed to observe any archaella and concluded that there were none. Shortly thereafter, Harrison and Kennedy (Harrison and Kennedy 1922) isolated strains from salted fish and named them Pseudomonas salinaria, later Serratia salinaria. As with Klebahn's strains, these strains have not survived, but they are considered likely similar to later isolates of Halobacterium salinarum (Tindall 1992; Oren 2006) [species initially assigned the epithets Halobacterium salinarium, Halobacterium cutirubrum and Halobacterium halobium are now all classified as Halobacterium salinarum; see (Ventosa and Oren 1996)]. Harrison and Kennedy (Harrison and Kennedy 1922) described the motility of these cells as ‘sluggish’ and the cells were archaellated with one archaellum at each end of the rod-shaped cells. H. salinarum was the first and is, to this day, one of the best studied archaea in terms of archaellation and chemotaxis.

Lochhead (Lochhead 1934) isolated a number of halophilic species from salted hides. Two of these isolates, Serratia salinaria (later H. salinarum) and Serratia cutirubra (later H. cutirubrum) were designated as the type species before further analysis resulted in H. cutirubrum being reclassified as H. salinarum (Ventosa and Oren 1996). Lochhead (Lochhead 1934) reported both of his ‘Serratia’ species to be motile in early stages of growth. S. salinaria possessed a single polar archaellum or frequently two archaella, one at each end, and the archaella were reported to be up to 10 µm long while S. cutirubra possessed a single polar archaellum.

The extreme oxygen sensitivity of methanogens hampered initial attempts to obtain these organisms in pure culture. Indeed, the initial studies of methane-producing organisms were done using isolates that were not pure cultures. The earliest methanogen cultures believed to be pure were Methanobacterium formicicum and Methanosarcina barkeri both of which were nonarchaellated (Schnellen 1947) while the earlier highly enriched strains of Methanosarcina methanica, Methanococcus mazei, Methanobacterium sohngenii and Methanobacterium omelianskii, none obtained free of contaminants, were all described as ‘permanently immotile’ (Barker 1936).

1950s-1970s

Key findings: Archaella/motility observed on methanogens and the wall-less Thermoplasma, first electron micrographs of cells with archaella

Extreme halophiles

In the 1950s–1970s era, a limited number of studies reported archaella or motility on a small number of archaea available in pure culture. Spruit and Pijper (Spruit and Pijper 1952) isolated an obligately halophilic pink organism from solar salt that they described as closely related to the halophilic organisms (assumed to be H. salinarum strains) of Klebahn (Klebahn 1919), Harrison and Kennedy (Harrison and Kennedy 1922) and Lochhead (Lochhead 1934). Interestingly, they failed to stain archaella in their organism although they observed a thin thread originating from the internal contents of the cells upon archaella staining which they thought may have been mistaken as archaella in earlier reports. They state that for these halophiles it is therefore questionable whether archaella ‘have ever been demonstrated and the question needs further investigation’.

The first electron micrographs of cells with archaella appear to be in a study by Houwink (Houwink 1956) on H. salinarum, seemingly providing incontrovertible proof of the existence of archaella in this extreme halophile. He observed motility, although restricted to axial rotation and oscillation at the poles. In agreement with early studies (Harrison and Kennedy 1922; Lochhead 1934), Houwink (Houwink 1956) observed polar location of archaella, although in tufts of 5–10 and not the single archaellum reported initially. Cells were reported to have monopolar archaellation during log phase while they were bipolarly archaellated in stationary phase. No diameters of the archaella were given in this early work.

In unpublished work reported in Dundas (Dundas 1977), Torsvik reported that in addition to the archaella lophotrichously attached to cells (i.e. at one or both poles of the cells), H. salinarum also had large amounts of free archaella unattached to cells which sometimes aggregated into thick spiral bundles that were visible in the light microscope (the so-called ‘super flagella’ of later studies). While attempts to isolate archaella by the traditional method of shearing from the cells failed, free archaella and archaella bundles were successfully isolated using CsCl gradients (Dundas 1977). Archaella bundles dissociated into individual archaella upon exposure to 1 M NaCl. Interestingly, disintegration of the halophile archaella did not occur at acid pH unlike observations with bacterial flagella leading Dundas (Dundas 1977) to state presciently that the ‘flagella proteins from Halobacterium salinarium thus seem to be distinct from the flagellin of other bacteria’.

Methanogens

Methanococcus vannielii (Stadtman and Barker 1951) appears to be the first pure culture of a methanogen to be described as ‘actively motile’ although at that time the presence of archaella was not determined. Indeed, archaella were not reported for this organism until Jones et al. (Jones, Bowers and Stadtman 1977) published electron micrographs showing cells possessing polar tufts of archaella. Balch et al. (Balch et al. 1979) published a list of all the known pure cultures of methanogens which included the motile cultures Methanospirillum hungatei, Methanobacterium mobilis (now Methanomicrobium mobile), Methanococcus voltae, Methanoculleus marisnigri, Methanogenium cariaci as well as Mc. vannielii.

The first study to focus on methanogen surface appendages was from Doddema et al (Doddema, Derksen and Vogels 1979) who reported archaella on cells of Methanobacterium (now Methanobrevibacter) ruminantium PS1 and Methanobacterium AZ (now Methanobrevibacter arboriphilus AZ) but not M. ruminantium M1 or Methanosarcina barkeri, Methanobacterium MOH (M. bryantii), Methanobacterium formicicum or Methanobacterium thermoautotrophicum (now Methanothermobacter thermoautotrophicus). Despite the paper containing an electron micrograph of M. ruminantium PS1 with one apparent archaellum, these findings of archaellated Methanobrevibacter are difficult to reconcile with the formal description of Methanobrevibacter species as nonmotile (Miller 2015).

Thermoacidophiles

Novel archaeal thermoacidophiles were first isolated in the 1970s. Darland et al. (Darland et al. 1970) first isolated Thermoplasma acidophila (now Thermoplasma acidophilum) from a coal refuse pile. No reference to motility or archaellation was given in the initial description of the organism but almost a decade later, Black et al. (Black et al. 1979) showed the cells were motile and possessed a single straight or irregularly curved archaellum about 9 µm long and with a diameter of 9–20 nm. This was an exciting finding as Thermoplasma species lack a cell wall and, in bacteria, motility is not observed in protoplasts or spheroplasts of flagellated cells despite the fact that they retain their flagella (Black et al. 1979). Sulfolobus species were not isolated until 1972 when Brock et al. (Brock et al. 1972) isolated Sulfolobus acidocaldarius from an acidic hot spring in Yellowstone National Park. No mention of motility or appendages that could be archaella were present in the original description. Remarkably, it wasn't until a comprehensive study of five Sulfolobus strains including S. solfataricus (now Saccharolobus solfataricus), Sulfolobus isolate B12 (now Saccharolobus shibatae) and S. acidocaldarius in 1989 that both motility and archaellation were first reported (Grogan 1989).

1980s

Key Findings: First directed studies on archaella reveal unusual features, first archaellin genes cloned, discovery of archaellins as glycoproteins, archaella composed of multiple archaellins appear to be the norm, archaella first observed on hyperthermophiles and Sulfolobales

Extreme halophiles: Isolation of archaella, glycosylation of archaellins, cloning of archaellin genes

The 1980s marked the decade in which studies that focused on archaella began with the first targeted studies conducted on H. salinarum. The initial work by Alam and Oesterhelt (Alam and Oesterhelt 1984) already pointed to unusual properties of archaella compared to flagella. They isolated highly motile strains of H. salinarum which most efficiently produced large bundles of loose archaella, then called ‘super-flagella’, consisting of hundreds of individual archaella. For the first time, they determined the handedness of archaella, showing the halobacterial archaella to be right-handed (a rarity for flagella). The paper also presented the length and helical parameters of halobacterial archaella (wavelength, amplitude and pitch) and determined the swimming speed of H. salinarum to be about 2 µm/s, or about 10 times slower than E. coli. Using tethered cells, they showed for the first time that archaella were rotating structures. Motility of monopolar archaellated cells consisted of smooth runs, a stop and then runs in the opposite or same direction. They demonstrated that archaella propelled cells forward or backwards when the filaments rotated clockwise (CW) and counter clockwise (CCW), respectively. Examination of purified ‘super-flagella’ by SDS-PAGE revealed three major bands each consisting of a series of sub-bands, a heterogeneity that they correctly surmised might be due to differing modifications of these proteins. In a separate contemporaneous study, Alam et al. (Alam et al. 1984) examined archaella and motility in the unusual square organism, (later called Haloarcula quadrata (Oren et al. 1999)) and reported several of the same unusual traits found in H. salinarum, including ‘super-flagella’ and the right-handed helicity of the filaments. However, the archaella of the square organism appeared to be composed of only a single protein species.

Alam and Oesterhelt (Alam and Oesterhelt 1984) made an insightful comment concerning the mechanical anchoring of archaella in H. salinarum where, like in many archaea, there is a cell wall that lacks murein and consists only of S-layer. They believed from their negatively stained samples that the archaella might extend through the cytoplasmic membrane into a cytoplasmic structure that might serve as a mechanical stabilization of the motor. This has proven likely to be the case, at least in euryarchaea, where what may be the same basic structure has subsequently been described in several different species.

In a subsequent paper, Alam and Oesterhelt (Alam and Oesterhelt 1987) were able to develop a purification of H. salinarum archaella by a dissociation and reconstitution method. Archaella preparations in low NaCl concentrations dissociated into their molecular components. Dialysis of this solution against 4 M NaCl led to reconstituted ‘super-flagella’, and SDS-PAGE analysis of this material showed the same three bands previously reported, indicating that all three archaellins were integral to the structure. This remains the only example of reconstitution of archaella and later attempts to reassemble Natrialba magadii archaella from dissociated archaellins were unsuccessful (Fedorov et al. 1995). This paper also showed, by dark field microscopy, polymorphic transitions from normal to curly, ring and straight forms following treatments at different pH and temperature.

The striking band pattern obtained after staining purified H. salinarum archaella analyzed on SDS-PAGE gels was also obtained by Wieland et al. (Wieland, Paul and Sumper 1985) who were able to prove that these sulfated glycoproteins were indeed the archaellins and that they were modified with the same sulfated oligosaccharides as the S-layer protein. These were the first glycosylated ‘flagellins’ known, another early clue to the unusual nature of archaella. The heterogeneity around each band of intensity observed on SDS-PAGE was predicted to be the result of differences in the number of attached sulfated glycans. Interestingly, they observed that the archaellins from the ‘super–flagella’ over-producing M-175 strain were all shifted lower in mass than the corresponding bands from wildtype cells. At the time, it was suggested that the M-175 strain had lost one or more glycosylation sites on its archaellin, but with the subsequent work of Gerl and Sumper (Gerl and Sumper 1988; Gerl, Deutzmann and Sumper 1989) identifying 5 archaellins, all of which are present in the final archaellum structure, this seems unlikely as it would mean that all five archaellin genes would have to undergo a loss of glycosylation sites. More likely is that the M-175 strain had a mutation in a gene involved in glycan synthesis (later on such genes would be designated as agl genes: (Chaban et al. 2006; Eichler, Jarrell and Albers 2013) resulting in the synthesis of a truncated glycan. The archaellins modified with this truncated glycan would all then migrate as smaller proteins. Wieland et al. (Wieland, Paul and Sumper 1985) reasoned that if glycosylation of archaellins was necessary for proper incorporation of the archaella into the cell wall then a glycosylation-deficient strain could have imperfectly anchored archaella that might explain the ‘super-flagella’ overproduction observed in the M-175 strain. A further important discovery that had an impact on the thinking of how glycosylation and archaellation assembly take place occurred when evidence pointed to glycosylation of proteins in H. salinarum occurring on the external side of the cell membrane (Sumper 1987). This external site of glycosylation led Gerl and Sumper (Gerl and Sumper 1988) to thoughtfully state that ‘aggregation to a functional flagellum is likely to occur by a mechanism different from that proposed for the assembly of eubacterial flagella’.

Gerl and Sumper (Gerl and Sumper 1988) cloned and sequenced the archaellin genes of H. salinarum, the first time this had been done for any archaeon. While the SDS-PAGE gel profiles had suggested the presence of 3 archaellins, they found a total of 5 archaellin genes, two in tandem at one locus (arlA1 and arlA2) while three other archaellins (arlB1, arlB2 and arlB3) were found in tandem at a second locus. All five archaellins were similar in length (193–196 amino acids long), with a long stretch of identical or near identical sequences interrupted by three short regions of hypervariability, unique to each archaellin. All five archaellins had predicted molecular masses lower than the bands observed in SDS-PAGE of archaella and all five had three potential N-glycosylation sites. Five archaellins, all modified with various degrees and sites of glycosylation, could readily explain the observed archaella SDS-PAGE pattern. Critically, Gerl and Sumper (Gerl and Sumper 1988) were unable to obtain N-terminal sequence data for the archaellins and thus could not identify any N-terminal processing of the proteins. They observed a hydrophobic stretch of amino acids in the N-terminus of the archaellins that they hypothesized may represent a contact site for archaellin-archaellin subunit interaction. A search of a protein data base revealed no significant sequence similarity between the halobacterial archaellins and other proteins including several available bacterial flagellins, pointing at this early stage of archaella study the uniqueness of the archaella structural proteins. Later, all five archaellins were identified in purified archaella samples (Gerl, Deutzmann and Sumper 1989).

Methanogens: Isolation of archaella from multiple methanogens

In the late 1980s, different laboratories were becoming involved in archaellation in methanogens. Kalmokoff et al. (Kalmokoff, Jarrell and Koval 1988) isolated archaella from Mc. voltae by shearing them from cells as well as using phase separation of membranes in the detergent Triton X-114 and observed two major bands after SDS-PAGE analysis. Electron microscopic examination of archaella isolated using the Triton X-114, but not by shearing, revealed the presence of an apparent anchoring structure unlike that of flagella; there was no distinct basal body with rings but rather only a small knob, also observed in other methanococcal species, immediately preceded by a slight thickening of the archaellum in the area corresponding to a hook in flagella (Fig. 2C). Archaella were also thinner than typical flagella, at only 13 nm diameter. Polar membrane-like structures often located in close proximity to archaella insertion sites were observed in thin sections of Mc. voltae (Koval and Jarrell 1987) and were postulated to be involved in anchoring the archaella, as suggested first for H. salinarum (Alam and Oesterhelt 1984). Cruden et al. (Cruden, Sparling and Markovetz 1989) used extraction with Triton X-100 to isolate archaella from Methanothermococcus thermolithotrophicus, and also observed polar membrane-like areas underneath areas of archaella insertion. Archaella were reported to be 13 nm in diameter and ended with a hook of variable length that could be discerned by a different staining pattern than the filament and also by its difference in thickness. These authors also observed a notch on one side at the junction of the curved region and the filament for Mtc. thermolithotrophicus archaella. It was reported that the intact archaella ended in a Gram positive-like basal body with 2 rings although many of the electron micrographs seemed to show a knob similar to that observed on Mc. voltae archaella. The archaella preparations consisted of at least 2 proteins by SDS-PAGE analysis. They also reported that detergent-extracted archaella of Msp. hungatei were composed of two proteins and appeared to be a little thinner than those of Mtc. thermolithotrophicus, at about 10 nm in diameter with a slightly thicker hook.

Figure 2.

Electron micrographs of archaella and protofilaments. (A) Archaella preparation from P. furiosus. Scale bar = 100 nm. From Nather et al. 2006. J. Bacteriol. 188:6915–6923, with permission. (B) Archaella isolated from Msp. hungatei by shearing. Scale bar = 100 nm. From Southam et al. 1990. J. Bacteriol. 172:3221–3228, with permission. (C) Archaella isolated from Mc. voltae by extraction with detergent OP-10. Inset shows curved hook-like regions ending in an anchoring knob. Archaella diameter is 12 nm. Courtesy of. S.I. Aizawa. (D)Natrialba magadii archaella in a buffer at pH 7.0 with 0.04% NaCl. The archaella (thicker filaments) are approx. 10 nm while the thinner protofilaments are 3–5 nm. Picture courtesy of Mikhail Pyatibratov.

Archaellated hyperthermophiles

Hyperthermophiles, organisms with an optimum temperature for growth of 80°C or above, were not isolated until the pioneering work of Stetter and Zillig beginning in the early 1980s (Stetter 2006). Remarkably, archaella were found on many hyperthermophiles, including ones growing near or even above 100°C, such as the impressively archaellated Pyrococcus furiosus (Fiala and Stetter 1986), a stark indication of the remarkable stability of these motility organelles. The first hyperthermophile to be isolated was Methanothermus fervidus and it was initially reported to be motile by bipolar polytrichous archaellation (Stetter et al. 1981). However, the filaments appear too thin to be archaella, in keeping with the lack of arl genes in the genome and the very recent update of the formal description of the genus Methanothermus which now lists cells as nonmotile (Akinyemi, Shao and Whitman 2021). The first hyperthermophilic archaeon shown to be motile, archaellated and with the arl gene cluster was Methanocaldococcus jannaschii (Jones et al. 1983; Bult et al. 1996).

Sulfolobaceae

Grogan (Grogan 1989) was the first to show that members of the Sulfolobaceae family, including S. solfataricus and S. acidocaldarius were motile and had archaella, long after their initial isolation (Brock et al. 1972). He also showed counterclockwise rotation of tethered cells, and therefore clockwise, nonreversing rotation of archaella. While motility was clearly established, ‘a classical form of bacterial chemotaxis was not observed; no directional response to carbon source gradients could be established’. It would be shown much later that Sulfolobus and Saccharolobus species, in fact, lack chemotaxis genes (Briegel et al. 2015). Grogan (Grogan 1989) also isolated filaments from S. shibatae and showed they were comprised of a single diffuse band of about 33 kDa which stained as a glycoprotein.

1990s

Key findings: archaellin genes shown to be essential for archaellation; isolation of archaella from diverse methanogens; archaea genomes have no homologues to genes involved in flagellation; evidence glycosylation is needed for archaella formation; archaellins are related to type IV pilins; archaellins made with signal peptides; first pre-archaellin peptidase studies; model for type IV pilus-like assembly proposed; first review of archaella published

Methanogens

Studies of euryarchaeotes, specifically halophiles and methanogens, dominated the work on archaella throughout the 1990s. Kalmokoff et al. (Kalmokoff, Karnauchow and Jarrell 1990) reported the N-terminal sequences of archaellins from Mc. voltae and Msp. hungatei were highly conserved to those of the H. salinarum archaellins deduced from the cloned genes obtained earlier (Gerl and Sumper 1988). The fascinating result was that the N-terminal sequences of the methanogen archaellins did not start with a methionine and aligned with the halophile sequences starting at position 13. This immediately suggested that archaellins were made as preproteins with short signal peptides, intriguing because flagellins are never synthesized as preproteins (Macnab 2004). This finding, coupled with the observation that the halophile archaellins are not glycosylated until they reached the cell surface (Sumper 1987), provided strong evidence from two distinct avenues that archaella might be assembled differently than flagella. If archaellins were synthesized with a signal peptide, it suggested archaellins crossed the cytoplasmic membrane where the signal peptide was removed and they were then glycosylated at the cell surface and finally assembled into a filament. It was also observed that the thin diameter of archaella may even preclude the existence of a hollow interior large enough to accommodate the passage of archaellin subunits.

Kalmokoff et al. (Kalmokoff and Jarrell 1991) reported the cloning and sequencing of the archaellins from a second archaeon, Mc. voltae. Four archaellin genes were found, despite the fact that the archaella preparations appeared to have only 2 major bands upon SDS-PAGE analysis. Northern blots indicated all four archaellin genes were transcribed; arlA by itself immediately upstream of the three other archaellin genes, arlB1, arlB2 and arlB3. All proteins were related to each other in sequence, with approximately the N-terminal 50 amino acids being highly conserved and, as with the halobacterial archaellins, the N-termini were very hydrophobic. Multiple transcripts were detected, all originating upstream of arlB1 with the most abundant one encompassing arlB1 and arlB2. Larger transcripts extended well beyond the sequence obtained downstream of arlB3 in the study and included at least one additional, non-archaellin gene, that was hypothesized to be involved in archaellin function or biosynthesis. A key finding was the observation that the start of the gene did not match with the already reported protein N-terminal sequence. There was an 11 or 12 N-terminal amino acid stretch predicted from the gene sequence which appeared to be a signal sequence cleaved from the proteins before they were incorporated into the filament. This was the first time a comparison was made of the genes and the actual filament proteins of the same organism to demonstrate a signal peptide.

Southam et al. (Southam et al. 1990) reported on the characterization of archaella from Msp. hungatei, including their cellular insertion. The ultrastructure of Msp. hungatei is complicated: individual cells are surrounded by a cell wall and then by a sheath with the cell ending with a spacer plug at each end (Beveridge, Sprott and Whippey 1991). Cells can grow in chains with internal cells of the chain separated by cell spacers about 400 nm in length. Archaella were never observed to exit through the sheath but always at the end through the spacer plugs which has pores large enough to accommodate the archaellum diameter. Cells were observed to swim at about 8 µm/s and not to reverse the direction of swimming. In log phase, cells were present as single cells or in short filaments which had archaella at the ends, while cells in stationary phase had much longer chains of cells with limited motility and few, if any, archaella. Archaella were about 10 nm diameter. SDS-PAGE examination of archaella isolated by shearing of cells revealed a pattern of two bands of close molecular masses (24 and 25 kDa) similar to that reported previously (Cruden, Sparling and Markovetz 1989). Sheared archaella lacked a discernible hook or anchoring structure (Fig. 2A, B). However, when the strains were inadvertently grown in distilled water mixed with tap water, a third higher molecular mass smear occurred with an easily observed ladder pattern within the smear. Glycoprotein stains were positive for all the archaellin bands and chemical deglycosylation resulted in a marked decrease in apparent mass. The effect of the medium composition on archaellation in Msp. hungatei was examined further by Faguy et al. (Faguy, Koval and Jarrell 1993) who showed that at low calcium levels and at suboptimal temperatures the cells were nonarchaellated, although archaellins were still synthesised. This study also presented an early attempt to synchronize the production of archaella in archaea by manipulation of the growth medium. Msp. hungatei cells were grown in low calcium containing medium to produce long nonarchaellated cells, and they were then sub-cultured into medium of increased calcium, resulting in short archaellated cells after about 2.5 h. During the transition from long to short cells, archaella were only observed on the short filaments. Extraction of Msp. hungatei archaella from spheroplasted cells using CHAPS or deoxycholate revealed a knob-like anchoring structure not observed in sheared archaella samples (Faguy, Koval and Jarrell 1994a), as noted previously for methanococci, but no hook. The archaella filaments were sensitive to a wide variety of detergents, but stable up to 80°C and over a pH range of 4–10. Lack of a genetic system has limited the study of these unusual archaea.

Archaella were subsequently isolated from a number of other methanogens including Mc. vannielii and M. marisnigri and the component archaellins were shown to possess the same highly conserved N-terminus as found for Msp. hungatei, Mc. voltae and H. salinarum archaellins (Kalmokoff, Koval and Jarrell 1992). Archaella preparations from these organisms, as well as from Methanothermus fervidus, Methanococcus maripaludis, Methanococcus deltae (later classified as Mc. maripaludis strain ∆RC) and Mcc. jannaschii were all composed of multiple major bands, presumed to be archaellins, upon examination by SDS-PAGE. The archaellin bands from several species stained positively as glycoproteins. Several did not however, including those of Mc. voltae and Mc. maripaludis, even though it was later shown by mass spectrometry analysis that they were glycoproteins (Voisin et al. 2005; Kelly et al. 2009).

An early study showing the importance of N-glycosylation for archaellation was published by Bayley et al. (Bayley, Kalmokoff and Jarrell 1993). When Mc. maripaludis strain ∆RC was passaged in increasing amounts of bacitracin, it led to the appearance of only lower molecular mass, hypoglycosylated archaellins detected by western blots, with the wildtype archaellin mass returning if these cells were subsequently transferred without bacitracin. Bacitracin is known to inhibit N-glycosylation if the precursor glycan employs a dolichol di-phosphate carrier but not if the carrier is dolichol monophosphate, since bacitracin forms complexes only with dolichol di-phosphate. EM analysis revealed a lack of archaella in cells grown in high levels of bacitracin, suggesting that a bacitracin-sensitive dolichol-diphosphate carrier was responsible for the glycosylation of the archaellins. This is unlike the observation in H. salinarum where the oligosaccharide precursor of archaellins is linked to a dolichol monophosphate and hence bacitracin insensitive (Sumper 1987). Although earlier studies with the H. salinarum super-archaella producing strain suggested that glycosylation affects proper archaella insertion into the cell envelope, the experiments with Mc. maripaludis ∆RC suggested for the first time that at least some amount of N-glycosylation was required for archaella filament formation itself.

Patel et al. (Patel et al. 1993) developed a method to form and regenerate protoplasts of Mc. voltae that were lacking the S-layer. EMs depicted heavily archaellated protoplasts but whether the protoplasts were actually motile was not reported. Using this discovery, as well as other critical advances made in the early development of genetic tools for methanogens (Gernhardt et al. 1990), Jarrell et al (Jarrell et al. 1996b) succeeded in isolating the first mutants in archaella formation in any archaeon, in Mc. voltae. They transformed wildtype cells with an insertional vector, incapable of self-replication, carrying an internal fragment of arlA and found transformants in which the vector has inserted into and disrupted arlA or arlB2. Insertional inactivation of arlA left the cells as archaellated as the wildtype, although they swarmed less in soft agar plates, but insertion into arlB2 resulted in nonmotile, nonarchaellated cells, providing definitive proof of the involvement of these genes in archaellation.

The genome sequence of the first archaeon, the archaellated Mcc. jannaschii, showed multiple archaellin genes and the same downstream genes as seen in Mc. voltae (Bult et al. 1996). Genome sequences of the archaellated Archaeoglobus fulgidus (Klenk et al. 1997) and Pyrococcus horikoshii (Kawarabayasi et al. 1998) quickly followed and they also possessed homologues to many or all of the Mc. voltae arl cluster genes. The genomes of these three archaea lacked homologs to any bacterial genes involved in flagellation (Faguy and Jarrell 1999). Curiously, both A. fulgidus and P. horikoshii (and subsequently many other archaellated species) were shown to have a bacterial-like chemotaxis gene set which was presumed to interact with the novel archaeal motility apparatus to modulate direction of rotation, but how this occurred did not start to be deciphered until much later (Schlesner et al. 2009; Alatyrev et al. 2010).

By the late 1990’s, archaellin gene sequences were available from a very limited number of archaea. In all cases, multiple archaellin genes were found in tandem with very well conserved 5’ ends. This latter observation led to the use of PCR to amplify the 3 archaellin genes from Mc. vannielii (Bayley et al. 1998). N-terminal sequencing of two protein bands obtained from SDS-PAGE of purified archaella indicated they were ArlB1 and ArlB2; ArlB3 could not be detected in purified filaments obtained by shearing. Subsequent work on Mc. voltae and Mc. maripaludis indicated that ArlB3 in those organisms was located in the cell proximal part of the archaellum and likely formed the curved region (Bardy et al. 2002; Chaban et al. 2007). This region would be expected to be missing or at least grossly underrepresented in archaella samples obtained by shearing and may be a facile explanation for the failure to detect ArlB3.

A group headed by Oleg Fedorov began to study archaella of methanogens and many halophilic archaea in the 1990s. Kostyukova et al. (Kostyukova et al. 1992) analyzed purified archaella of Mtc. thermolithotrophicus by SDS-PAGE and observed three major bands, two of which stained weakly with Alcian blue suggesting small amounts of attached glycan. It would be many years before the protein bands would be correlated to specific archaellin genes and the structures of the archaellin N-linked glycans determined (Kelly et al. 2020).

The hydrophobic N-termini of archaellins can form strong intersubunit bonds that can lead to nonspecific aggregation of archaellin monomers. It was suggested that cytoplasmic chaperones may be needed to prevent aggregation of newly synthesized archaellins (Kostyukova et al. 1994). Chaperones may be needed to efficiently deliver archaellin subunits to the assembly point on the cytoplasmic membrane and cytoplasmic ATP-dependent archaellin binding proteins have been reported from Euryarchaeota (Polosina et al. 1998). Chaperones for many flagella proteins have been identified (Khanra et al. 2016; Ribardo et al. 2019) but this avenue of research has not been explored to any extent for archaella.

Extreme halophiles and haloalkaliphiles

Work on halophile archaella and motility in the 1990s continued to focus on H. salinarum but new studies appeared on haloalkalophiles as well. Marwan et al. (Marwan, Alam and Oesterhelt 1991) observed that H. salinarum swam faster by clockwise than by counterclockwise rotation of the archaella bundle. They concluded that the majority, if not all, of the individual archaellar motors of a cell rotate in the same direction at any given time. Reported here, and more thoroughly later (Kupper et al. 1994), was that individual archaella filaments ended in a ‘basal body’ or at least a knob indicating that each archaellum was driven by its own motor. Using atomic force microscopy, Jaschke et al. (Jaschke, Butt and Wolf 1994) confirmed the presence of individual motors for each H. salinarum archaellum.

Kupper et al. (Kupper et al. 1994) showed that archaella of H. salinarum were inserted into what they called a polar cap, although they could not determine if the cap was part of the wall or membrane or located below the membrane in the cytoplasm (Fig. 3). This would be the focus of later work, especially in multiple publications from the group headed by Antonina Metlina, but investigations of thin sections of many euryarchaea revealed the presence of a cytoplasmic structure, referred to by many different names (disk-like lamellar structure (DLS), polar membrane and polar cap), located below the entry point of archaella (Koval and Jarrell 1987; Cruden, Sparling and Markovetz 1989; Gongadze et al. 1993; Briegel et al. 2017; Daum et al. 2017) which may be a further anchoring place for the archaella. The disk-like lamellar structure is best studied in H. salinarum (Speranskii et al. 1996). Here, the complex structure located directly under the insertion of polar archaella was found to be an electron dense line that runs parallel to the CM, about 20 nm from the CM, and extends up to 300 nm in length (Fig. 3). According to Speranskii et al. (Speranskii et al. 1996), not a single bundle of archaella was observed without the DLS at the base. At high magnification, the DLS consisted of a number of layers with two electron dense layers separated by a colorless gap, similar to a lipid bilayer membrane. In lysed cells, the DLS was often observed to still be attached to an archaella bundle, indicating a strong link between the two.

Figure 3.

Archaella and the polar cap/discoid lamellar structure cellular anchor. (A) Archaellated polar caps of H. salinarum as obtained by gel filtration. Scale bar = 1 µm. From Kupper et al. 1994. J. Bacteriol. 176: 5184–5187, with permission. (B)H. salinarum ghost thin section showing discoid lamellar structure (DLS; thick arrows) and polar organelles (thin arrows). Scale bar = 1 µm. From Metlina 2004. Biochemistry (Mosc) 69:1203–1212, with permission. (C) Electron cryo-tomographic slice of P. furiosus. Arrows indicate polar cap; Arc, archaella; MC, motor complex; SL, S-layer; CM, cytoplasmic membrane. Scale bar = 200 nm. From Daum et al. 2017. eLife 6: e27470.

Tarasov et al. (Tarasov et al. 1995) studied the stability of H. salinarum archaella by scanning microcalorimetry, circular dichroism and electron microscopy. Melting of archaella in a microcalorimeter in the presence of 20% NaCl demonstrated the incredible thermostability of the structures as they did not denature even at temperatures up to 130°C. In the presence of decreasing amounts of NaCl, archaella were shown to be denatured, but nevertheless archaellar structures were still present by EM. Archaella preparations in high salt were stable to protease treatment but less so in 1% NaCl. Pyatibratov et al. (Pyatibratov et al. 1996) reported that the archaella morphology of the haloalkaliphilic archaeon Natronobacterium pharaonis (now Natronomonas pharaonis) became disordered below 10% NaCl, and in the absence of this salt no filaments could be observed. Archaella were resistant to protease activity except at <10% NaCl levels. Remarkably, after protease treatment, EM analysis still revealed filament structure although the diameter was much reduced from 12.2 nm to 5–6 nm. The reduced thickness was explained by the protease treatment removal of external parts of the archaellin molecules (presumed to be the C-termini) without affecting the domains involved in subunit-subunit interaction that form the filament core (presumed to be the conserved hydrophobic N-termini).

Fedorov et al. (Fedorov et al. 1994) showed that archaella of the haloalkaliphilic Natronobacterium magadii (now Natrialba magadii) were composed of 4 major protein bands when analyzed by SDS-PAGE. With decreasing NaCl concentrations, the 10 nm archaella filaments were found to dissociate into thinner 3–5 nm filaments, called protofilaments (Fig. 2D). Similar-sized filaments were also observed upon longterm storage of T. acidophilum archaella in medium at low pH, although these thin filaments were not observed on whole cells (Faguy et al. 1996). It was proposed that the archaella were formed from multiple protofilaments with the connections between protofilaments being less stable than the connections between the subunits of the protofilaments themselves (Fedorov et al. 1994). Attempts to reconstitute archaella from dissociated archaellins were unsuccessful.

Archaella from several halophiles and haloalkaliphiles, including H. salinarum, Haloferax volcanii (formally classified as nonmotile at the time) Halobacterium saccharovorum (later Halorubrum saccharovorum) and N. pharaonis showed the now common features of multiple (2–5) archaellins that stained presumptively as glycoproteins although the staining of N. pharaonis archaellins was poor (Serganova et al. 1995).

Sulfolobus and Thermoplasma

SDS-PAGE analysis of purified archaella from species living in thermoacidophilic habitats, namely S. shibatae and T. volcanium, revealed a single protein band (41 kDa) for T. volcanium and two bands of similar mass (31 and 33 kDa) for S. shibatae, all of which stained as glycoproteins (Faguy et al. 1996). N-terminal sequencing of the T. volcanium protein band revealed a sequence identical to that of Mc. voltae archaellin at 18 of the N-terminal 20 amino acid positions. The two archaellins of S. shibatae had an identical N-terminal sequence for all 23 amino acids obtained and these were identical at 12/23 positions to archaellin of Msp. hungatei. Since it was later determined that Saccharolobus species possess only a single archaellin gene (Albers and Jarrell 2015a) and previous analysis of isolated archaella of S. shibatae revealed only a single diffuse glycoprotein band of approximately 33 kDa (Grogan 1989), the two bands of similar molecular mass and identical N-terminal sequence most likely represents two variants of the same gene product modified slightly differently by glycosylation.

A different kind of study on archaeal motility was presented by Lewus and Ford (Lewus and Ford 1999) who reported on the temperature-sensitive motility of S. acidocaldarius. They observed the swimming speed to be relatively slow (7–15 µm/s) compared to mesophilic and hyperthermophilic bacteria tested with motility observed from 45°C to above 80°C. Both swimming speeds and run length times were increased as the temperature increased, suggestive of a survival mechanism that allowed cells to migrate from lethal hot spots in their natural environments to areas of lower temperature. The authors emphasized that the conditions they used for study meant they were not studying thermotaxis, as cells were not exposed to gradients of temperatures but rather incubated at different constant temperatures.

Thermococcus

The first directed studies of archaellin genes from a non-methanogen or halophile were published by Nagahisa et al. (Nagahisa et al. 1999) who reported 5 tandem, co-transcribed archaellin genes in Pyrococcus kodakaraensis (later Thermococcus kodakaraensis). All five archaellins had the conserved N-terminus already reported for other archaellins but the size of the proteins encoded varied much more than that observed previously, with T. kodakaraensis ArlB4 being 219 amino acids long but ArlB2 being a whopping 580 amino acids and the others falling in the 258–294 amino acid range. Northern blotting revealed several transcripts all originating upstream of arlB1, with the longest one extending from arlB1 through all the arl accessary genes hypothesized at the time.

Pre-archaellin peptidase

By the mid-1990s, a number of N-terminal sequences of archaellins were available and it was evident that archaellin sequences were not related to flagellins. Faguy et al. (Faguy et al. 1994b) reported sequence similarity between archaellins and members of the type IV pilin-transport superfamily at their N-terminus. Members of this family have unusual signal peptides that are processed by a specific signal peptidase called the prepilin peptidase, named after its role in processing type IV pilins. The members of the type IV pilin-transport superfamily typically have a group of conserved associated genes usually located in the immediate vicinity. At this time, no neighboring genes had yet be reported for archaellins but Faguy et al. (Faguy et al. 1994b) suggested they were likely present and, if so, it would offer a novel method of assembly of a prokaryotic ‘flagellum’ using type IV pilus assembly as a model, i.e. processing of the structural components by a specific signal peptidase and assembly of the new subunits at the base of the structure.

In 1999, it was hypothesized that not only archaellins but also sugar binding proteins in Sulfolobus and potentially a limited number of other secreted substrates in archaea were processed by the same system (Albers, Konings and Driessen 1999). This was based on the presence of demonstrated (for the glucose-binding protein) or predicted short signal peptides ending in a glycine followed by a hydrophobic stretch of amino acids as observed initially in archaellins. However, based on the conserved amino acids surrounding the cleavage site of archaellin sequences available at the time that were lacking in most of the substrates presented by Albers et al. (Albers, Konings and Driessen 1999), Jarrell et al. (Jarrell, Correia and Thomas 1999) argued that there was likely to be an archaellin-specific processing mechanism and they believed that ‘the preflagellin peptidase will prove to be a more or less dedicated enzyme for the processing of the preflagellins and perhaps a limited number of related proteins’. As subsequent publications proved, a dedicated pre-archaellin peptidase was found in Mc. maripaludis (initially designated FlaK, now proposed to be ArlK) but a corresponding type IV prepilin peptidase-like enzyme (PibD) with a wider range of substrates was found in Sulfolobus species and in Hfx. volcanii.

As a step towards identifying the archaeal homolog of the type IV prepilin peptidase (TFPP) that was responsible for cleaving the signal peptides from archaellins, overexpression of Mc. voltae pre-archaellin was successfully achieved in both E. coli and Pseudomonas aeruginosa (Bayley and Jarrell 1999). Processing of pre-archaellins expressed in the membranes of E. coli was shown when membranes of Mc. voltae were added as source of the peptidase in an in vitro assay. This was the first study of archaellin processing by a TFPP and work on archaeal TFPP enzymes intensified in the next decade.

The first archaella review

In 1996, the first review of archaella was published and it presented, for the first time, a formal model for archaellum assembly using a type IV pilus assembly model (Jarrell, Bayley and Kostyukova 1996a). In this model, archaellins are made as preproteins with type IV pilin-like signal peptides. The preproteins are delivered by chaperones to the cytoplasmic membrane where the signal peptides are cleaved by a then unidentified pre-archaellin peptidase, homologous to the bacterial TFPP. Subunits are glycosylated before being incorporated into the growing filament at the base, as in pili systems but unlike flagella. Initial support for this model followed with the report of sequence downstream of arlB3 in Mc. voltae. Initially, 7 genes believed to be co-transcribed with arlB1-arlB3 were found and designated flaC, D, E, F, G, H and flaI (Bayley and Jarrell 1998), and now termed arlC, D, E, F, G, H, I (Pohlschröder et al. 2018). Later arlJ was included as a crucial conserved part of the arl gene cluster. Both arlI and arlJ were later shown to be homologs of genes crucial for type IV pili assembly (Bayley and Jarrell 1998; Peabody et al. 2003).

2000s

Key findings: Sulfolobales become model organisms for archaellation studies; processing of archaellins by signal peptide removal and N-glycosylation; the pre-archaellin peptidase identified; arl gene cluster transcription studies; evolutionary interest of the arl gene cluster; arl gene mutant analysis; distribution of the multiple archaellins in the archaellum; archaella structure; rotation of archaella is driven by ATP hydrolysis

From the beginning of the new millennium, it became much more common to see researchers using different model organisms to address similar conceptual questions concerning the archaellum, including generating arl gene mutants, archaellin processing and attempts to localize and assign roles to the multiple archaellins within the structure. Sulfolabales species became the first crenarchaote model organisms for archaella research. Comparative studies in euryarchaeotes and crenarchaeotes were instrumental in defining what features are common and what features are unique in archaella from different organisms.

Pre-archaellin peptidase

As the new decade began, work on the archaellin signal peptide and its processing intensified. Importantly, the gene encoding the prepilin peptidase homolog was identified in methanogens and Sulfolobus. Bardy and Jarrell (Bardy and Jarrell 2002) identified the gene encoding the pre-archaellin peptidase in Mc. maripaludis and named it flaK (now arlK). The placement of arlK or its equivalent TFPP in the genome is unusual. It lies outside of the arl gene cluster in all but the rarest cases, such as in Mcc. jannaschii where it directly follows arlJ. ArlK homologues were found in all motile archaea with the initial exceptions of S. acidocaldarius, T. volcanium and A. pernix (Desmond, Brochier-Armanet and Gribaldo 2007). Later analysis using improved sensitive methodologies, however, did find an arlK family homolog in these organisms as well (Makarova, Koonin and Albers 2016). ArlK expressed in E. coli localized to the membrane which could then be used in in vitro assays to confirm the pre-archaellin peptidase activity of ArlK, the first non-archaellin archaella-associated gene product to be assigned a specific role in archaellation. Insertional inactivation of arlK in Mc. voltae resulted in non-motile and non archaellated cells, for the first time showing that removal of the signal peptides from archaellins was an essential step in assembly of the filament (Bardy and Jarrell 2003). Site-directed mutagenesis demonstrated conclusively that the two aspartic acid residues of ArlK, conserved with ones required for activity in type IV prepilin peptidases, were necessary for cleavage of the pre-archaellin signal peptide.

At about the same time, Albers et al. (Albers, Szabo and Driessen 2003) identified a S. solfataricus prepilin-like signal peptidase by searching the Cluster of Orthologous Groups of proteins (COG) database (Tatusov et al. 2000), COG 1989. They showed that when this protein, designated PibD, was expressed in E. coli it could cleave both pre-archaellin and a truncated version of a sugar binding protein preGlcS, hypothesised to assembled into a pilus-like structure called the bindosome at the cell surface. Szabo et al. (Szabo, Albers and Driessen 2006) showed that PibD is an aspartic acid protease and that D23 and D80 were critical for activity, as is true of the equivalent residues in ArlK and TFPPs. Interestingly, while mutants with deletion/inactivation of arlK were obtained in methanogens, attempts to delete pibD in S. solfataricus were unsuccessful (Albers and Pohlschroder 2009) even though subsequently it was possible to delete pibD in S. acidocaldarius (Henche et al. 2014).

Meanwhile, the conditions of an effective in vitro pre-archaellin processing assay using methanococcal membranes as a source of the peptidase (Bayley and Jarrell 1999) were optimized and, importantly, the smaller molecular mass product obtained after archaellin processing was shown directly by N-terminal sequencing to be same as the mature Mc. voltae ArlB2 protein found in filaments (Correia and Jarrell 2000).

Alignments of the limited number of pre-archaellin sequences available revealed a number of conserved positions relative to the signal peptide cleavage site, such as a +3 glycine, -1 glycine and the -2, -3 positions which were almost always held by charged amino acids (Fig. 4) (Thomas, Chao and Jarrell 2001a; Bardy, Eichler and Jarrell 2003). Several changes of the -1 Gly in Mc. voltae ArlB2 resulted in no or poor cleavage of the archaellin (Thomas et al. 2001), with similar effects observed for changes of other conserved positions. In S. solfataricus, numerous site directed changes to the preGlcS protein at or near the cleavage site were also shown to have deleterious effects on the cleavage reaction (Albers, Szabo and Driessen 2003). Ng et al. (Ng et al. 2009) examined the minimal signal peptide length required for cleavage by ArlK and PibD since ArlK only cleaved archaellins, with signal peptides of 11–12 amino acids, while predicted substrates for PibD included ones with signal peptides as short as 3 amino acids. ArlB2 substrates modified to have truncated signal peptides truncated from 11 to 3 were expressed in E. coli. ArlK was able to process ArlB2 with signal peptides as short as 5 amino acids, while PibD was able to process signal peptides as short as 3 amino acids, again highlighting differences in the archaellin-specific ArlK and the more promiscuous PibD.

Figure 4.

N-terminal alignment of selected pre-archaellins from several species with signal peptide cleavage site indicated. The accession number of each archaellin is provided next to the species name. In red, the conserved charged residues (acidic or basic) that immediately precede the cleavage site, which takes place as indicated after the conserved glycine, in orange. The +3 conserved glycine is indicated in light pink. Highlighted in blue, the hydrophobic stretch of amino acids that form the core of the archaellar filament.

Szabo et al. (Szabo et al. 2007a) showed diverse archaeal proteins in S. solfataricus were possible substrates of PibD. They developed a program called FlaFind (http://signalfind.org/flafind.html) and used it to identify over 300 proteins in 22 archaeal species with class III pilin-like signal peptides, many of which were archaellins. In Mc. maripaludis, three FlaFind genes encoding proteins with a domain of unknown function DUF361 (later shown to encode type IV pilins) were found in a predicted operon with a gene encoding a novel class of prepilin peptidase which was termed EppA. EppA specifically processes these pilins while FlaK specifically processes archaellins. The presence of two TFPPs is a clear distinction from the PibD-containing archaea where one enzyme cleaves all substrates with class III signal peptides.

Arl gene cluster

Since the flagellum is a complex structure requiring over 50 proteins for its structure, assembly and regulation (Macnab 2004), it was somewhat surprising that virtually all the genes needed for assembly and structure of the archaellum appear to be present in a single cluster of typically <12 genes. While this gene cluster is often referred to as the arl operon (Fig. 5), in fact, for many archaea the co-transcription of all the genes has not been shown and for others, archaellin genes are found in the opposite orientation of the other arl genes and cannot be co-transcribed with them. Most archaellated members of the two known phyla of the time, Euryarchaeota and Crenarchaeota, were shown to have one of two highly conserved versions of the arl cluster, originally designated as FlaI and FlaII, but now here referred to as Arl1 and Arl2 clusters (Desmond, Brochier-Armanet and Gribaldo 2007).

Figure 5.

The arl gene clusters in a variety of Archaea.

Arl1 clusters are widely found throughout the Euryarchaeota including Thermococcales, Methanococcales, Halobacteriales and Thermoplasmatales. Some variations are found but, in general, Arl1 clusters begin with several archaellin genes (usually 2–5) followed by the arl accessory genes arlC, arlD, arlE, arlF, arlG, arlH, arlI and arlJ. Some archaea, like Halobacteriales, have non-archaellum genes, often chemotaxis genes such as ones encoding methyl-accepting chemotaxis proteins (MCP), located within the arl gene group (Fig. 5). Here, archaellin genes are often adjacent to, but in the opposite orientation to, the arl-associated genes and so archaellins and the arl–associated genes are not co-transcribed. A commonality in other species (e.g. P. furiosus, Hfx. volcanii) is that arlC, arlD and arlE may not be found as separate genes but instead as fusions in various combinations that result in the ‘fused’ proteins ArlCE, ArlDE or ArlCDE (Makarova, Koonin and Albers 2016).

Arl2 clusters are found in Crenarchaeota as well as certain Euryarchaeota. The main features of Arl2 clusters that distinguish them from Arl1 clusters are that Arl2 clusters lack arlC, arlD and arlE but have arlX in their place (or genes distantly related to arlD/E in the case of the Euryarchaeotes with an Arl2 cluster) and usually the order of arlF and arlG is reversed. The best studied Arl2 cluster is in S. acidocaldarius where the order is a single archaellin gene, arlB, followed by arlX, arlG, arlF, arlH, arlI, arlJ, with the pre-archaellin peptidase equivalent, pibD, located outside the cluster (Lassak et al. 2012b). To date, deletion of any arl gene, with the exception sometimes of one of the multiple archaellin genes, leads to non-archaellation of the resulting mutant. Remarkably, this minimal set of arl genes in Arl2 systems, which in S. acidocaldarius consists of only 7 genes plus pibD, appears to be all that is needed for archaellum structure, assembly, anchoring and rotation.

Homologues of arl genes are found in all archaea that have been described as motile, with rare exceptions. No arl genes have been found in the genomes of Pyrobaculum aerophilum or Methanopyrus kandleri, despite both having been described as motile with a monopolar bundle or polar tufts of archaella, respectively (Kurr et al. 1991; Volkl et al. 1993; Slesarev et al. 2002). These discrepancies have yet to be completely sorted out and will likely require isolation of these structures to identify their protein components. Conversely, Methanosarcina species have been described as non-motile despite apparently coding for Arl and chemotaxis proteins (Galagan et al. 2002; Deppenmeier et al. 2002). Under certain growth conditions, maybe in conjunction with cell morphology changes, these cells could possibly become motile but this has never been observed.

Arl associated proteins localize to the membrane

The arl gene clusters of Mc. maripaludis, Mtc. thermolithotrophicus and Mcc. jannaschii are very similar to that found in Mc. voltae with multiple archaellin genes followed by arl accessory genes arlCDEFGHIJ, and Northern blots detected mRNAs encoding at least some of the archaellins and sometimes some of the arl associated genes (Thomas and Jarrell 2001). However, none of the detected transcripts was long enough to encode all the arl genes. ArlC, ArlD, ArlE, ArlH and ArlI were all localized to the membrane fraction even though several of the proteins do not have predicted membrane spanning domains suggesting their localization to a membrane-anchored complex. Western blots conducted with purified intact archaella did not detect the presence of any Arl-associated proteins in the actual archaella structure. Interestingly, western blots detected two signals for ArlD, one at 52 kDa and one at 15 kDa. The ArlD sequence of Mc. voltae has an in-frame methionine with an upstream ribosome-binding site (RBS) that could lead to the observed 15 kDa signal. In a Mcc. jannaschii proteome study several Arl proteins were detected, but for ArlD only the truncated protein was identified. This suggested that the smaller protein is also made in this hyperthermophile (Mukhopadhyay, Johnson and Wolfe 2000). Thus, ArlD likely encodes 2 proteins. Thomas and Jarrell (Thomas and Jarrell 2001) also noted the similarity of ArlG at the N-terminus to archaellins and the presence of a Walker Box A in ArlH (essential for ATP binding) and Walker Box A and Box B (essential for ATP hydrolysis) in ArlI, the latter known already to have homology to the ATPase PilT in type IV pili systems. ArlJ, predicted to have 7–9 membrane spanning domains, was hypothesized to form a secretory complex with ArlH and ArlI to export archaellins. Later, ArlJ was shown to be homologous to a TFP system membrane protein (PilC) that interacts with the ATPases PilT/B (Peabody et al. 2003).

arl gene inactivation studies in Methanococcus, Halobacterium and Sulfolobus

By 2000, the existence of an arl gene cluster that encoded archaellins as well as a number of genes likely involved in archaellum structure and/or assembly was apparent. At this time, the only mutants in arl genes were ones isolated in Mc. voltae by insertional inactivation that showed that the arl gene cluster was involved in archaella formation, but not the involvement of individual genes within the cluster (Jarrell et al. 1996b). Thomas et al. (Thomas, Pawson and Jarrell 2001b) showed that insertional inactivation of arlH, with polar effects on arlI and arlJ, led to nonmotile, non-archaellated cells showing at least one of the genes of this cluster was essential for archaella formation. Subsequently, Thomas et al. (Thomas et al. 2002) isolated non polar mutants in arlI and arlJ, both of which were nonmotile and non-archaellated, linking each of these gene products to a critical role in archaellation. At the same time, Patenge et al (Patenge et al. 2001) showed that the arl gene cluster was responsible for archaella biogenesis in H. salinarum. They created the first in-frame deletion of an arl gene in arlI and this strain was nonmotile and non-archaellated. Szabo et al. (Szabo et al. 2007b) were able to provide evidence for the essential nature of arl genes in the motility of crenarchaeotes for the first time when they demonstrated that a mutant carrying a disruption in arlJ in S. solfataricus was non-motile and non-archaellated. Using new genetic techniques developed in Mc. maripaludis (Moore and Leigh 2005), Chaban et al. (Chaban et al. 2007) reported the successful in-frame deletion for all 3 archaellin genes as well as arlC, arlF, arlG, arlH and arlI in this methanogen. No attempt was made to delete arlJ, and deletions of arlD and arlE were unsuccessful. All mutants obtained were found to be non-archaellated, with the exception of arlB3. This mutant made archaella and was motile but with an unusual swimming pattern of tight circles, and the isolated archaella lacked the curved hook-like region found on wild-type archaella, thus implicating ArlB3 as the major component of this structure, consistent with data previously presented in Mc. voltae (Bardy et al. 2002). This study provided the first evidence that ArlC, ArlF, ArlG and ArlH were all necessary for archaellum formation.

Distribution and Role of Multiple Archaellins

In the early 2000s, studies in methanogens and halophiles began to address the common occurrence of multiple archaellins in purified archaella. Less commonly, some archaea like Sulfolobus species have only a single archaellin and that archaellin forms the filament. For those archaea with multiple archaellins, no consensus for the distribution or role of multiple archaellins has emerged and instead a variety of findings have been presented in the organisms studied.

In Mc. voltae, the major components of the filaments were shown to be ArlB1 and ArlB2, while early studies failed to detect either ArlA or ArlB3. Bardy et al. (Bardy et al. 2002) were able to detect both ArlA and ArlB3 in purified archaella using Western blots developed with specific antibodies raised to the unique regions of each of these archaellins. They employed archaella samples prepared in three ways. First, detergent-extracted cells released so-called intact archaella with many archaella filaments having a curved region superficially resembling a flagellum hook and ending in a knob, presumably representing the cell proximal region ending in an attachment knob (Fig. 2C). The hooks were of a more variable length than seen in flagella, a feature shared with Mtc. thermolithotrophicus archaella (Cruden, Sparling and Markovetz 1989). Second, archaella were isolated by shearing intact cells. The sheared archaella rarely had the hook-like region or the knob, suggesting that these were left behind attached to the cells. A third sample was obtained by first shearing the cells and then extracting the resulting cells with detergent. This resulted in the isolation of short archaella pieces (designated archaella stubs) enriched for the cell proximal region containing hooks and knobs. Comparison of archaella stubs with sheared archaella showed an increased presence of ArlB3 in the stubs, suggesting that this archaellin might be forming the cell proximal hook-like region. ArlA was detected in all three archaella samples suggesting this minor component may be located throughout the filament. These findings were supported by a later MS analysis of purified Mc. voltae archaella (Voisin et al. 2005), where good coverage of ArlB1 and ArlB2 was obtained when sheared samples were used but good coverage of ArlB3 could only be obtained using intact archaella. Very poor coverage of ArlA was obtained in either case, also consistent with its previously reported low abundance in the filament (Bardy et al. 2002).

Hooks have rarely been observed outside of methanococcal archaella, although the thermoacidophilic euryarchaote, Aciduliprofundum boonei, also had archaella with ‘very elegant hooks’ (Reysenbach and Flores 2008). Archaella isolated from Sulfolobales and Hfx. volcanii do not have hook-like regions and at least in some instances, such as Sulfolobus species, these archaea only have a single archaellin gene, making it impossible to have one archaellin form the filament and another form a hook.

Tarasov et al. (Tarasov et al. 2000) examined the role of the multiple archaellins in H. salinarum. In this organism, the A archaellin genes are located in tandem (arlA1, arlA2) while the B archaellin genes (arlB1-B3) are located elsewhere on the chromosome, also in tandem. All five archaellins were found in purified archaella with ArlA1 and ArlA2 forming the major part of the filament (Tarasov et al. 2000; Syutkin et al. 2012). Disruptions of the arlA cluster, or the arlB cluster, or arlA2 only, all resulted in abnormal archaella structure, location and/or function, indicating both ArlA and ArlB archaellins are needed for proper functioning archaella. Cells with any of the three insertions into archaellin genes were much less motile than wild-type cells. The authors concluded that the ArlA archaellins form the bulk of the filament while the ArlB archaellins may have a specialized role near the basal body. Later, short curved filaments were observed in H. salinarum archaella composed of only ArlB when the arlA gene cluster was deleted. If only the arlB1 or arlB3 archaellin genes were present, no curved structures were observed on the cell surface suggesting initially that ArlB2 formed a hook-like region (Beznosov, Pyatibratov and Fedorov 2007) although a later study would refute this (Beznosov et al. 2013).

The haloalkaliphile N. magadii has four archaellin genes, arlB1-B4, with arlB3 having a large insertion in the middle (Serganova et al. 2002). They argued that this central part of ArlB3 may be a result of a horizontal transfer of a part of a flagellin gene from E. coli of the EPEC 1 group. The four archaellin genes were shown to be co-transcribed, likely with downstream genes. Detailed analysis of the composition of the archaella using Western Blots, N-terminal sequencing and Immuno EM led to the hypothesis that the archaellum filament is composed of several longitudinal rows of protofilaments each consisting of a single archaellin (Fedorov et al. 1994; Serganova et al. 2002; Pyatibratov et al. 2002).

Pyatibratov et al. (Pyatibratov et al. 2008) reported that in Haloarcula marismortui two different phenotypes were found, where archaella were composed of primarily either ArlB or ArlA2 archaellins. Interestingly, arlA2 is located on a plasmid while arlB and the typical arl accessory genes are found on the main chromosome. The filaments composed of ArlB or ArlA2 archaellins were of different diameters and both were much thicker than typical archaella. Later, it was shown that these two genes are ecoparalogs (Syutkin et al. 2014b).

Importance of N-linked glycosylation of archaellins for archaellation and motility

In 2005, the study of glycosylation of archaellins begun by Sumper, Wieland and colleagues on H. salinarum in the 1980’s was taken up in methanogens, beginning with Mc. voltae (Voisin et al. 2005). In this methanogen, all possible N-glycosylation sequons (Asn-X-Ser/Thr, where X is any amino acid except Pro) in the 4 archaellins covered in the sample were modified with the same trisaccharide with the linking sugar being GlcNAc. Shortly thereafter, Chaban et al (Chaban et al. 2006) identified the first genes in any archaea shown to be involved in glycosylation. These genes were designated agl genes (archaeal glycosylation) and were identified in Mc. voltae as the oligosaccharyltransferase aglB and the glycosyltransferase aglA, the latter involved in the transfer of the final sugar of the trisaccharide. In a later study, using a different version of Mc. voltae strain PS, the glycan modification of the archaellins was identified as a tetrasaccharide with the first three sugars being the same as reported in the initial study, but with an additional fourth residue of 220 or 262 Da (Chaban et al. 2009). This suggested that the wildtype version of the glycan is likely to be the tetrasaccharide and the trisaccharide may have been the result of an unknown mutation that developed upon long time passage of the PS strain in the Jarrell laboratory that affected the biosynthesis or attachment of the final sugar.

Mutants with insertional inactivation of agl genes had archaellins with reduced apparent masses compared to wild-type cells; in addition, aglA mutants were less motile with fewer archaella than wild-type cells and aglB mutants were non-motile and non-archaellated. This study provided strong genetic evidence for a key role of archaellin glycosylation in archaella filament formation and function, findings that would subsequently be mirrored in other archaea.

Kelly et al. (Kelly et al. 2009) reported the structure of a unique tetrasaccharide N-linked glycan to multiple positions of the three Mc. maripaludis archaellins. Subsequent papers identified many of the genes involved in the pathways for biosynthesis, assembly and attachment of the glycan (VanDyke et al. 2009; Jones et al. 2012; Ding et al. 2013; Jarrell et al. 2014; Siu et al. 2015). As with Mc. voltae, mutations in agl genes that resulted in truncated glycans of less than two of the four sugars led to non-motile, non-archaellated cells, while glycans of three or two sugars led to cells that remained archaellated but demonstrably less motile than wild-type cells (VanDyke et al. 2009).

Internal sites of archaella anchoring

Work on the DLS anchoring of archaella continued in the Metlina laboratory. Kireev et al. (Kireev et al. 2006) confirmed that the archaella of H. salinarum were attached to DLS or polar caps as Kupper et al (Kupper et al. 1994) reported and showed this was the case for N. pharaonis and N. magadii as well. Further EM studies on H. salinarum (Speranskii, Novikova and Metlina 2008) revealed bulging areas in the DLS that corresponded to archaella insertion sites. Evidence provided by examining ‘ghosts’ obtained by mild lysis of cells to remove the internal cytoplasm showed that the DLS are firmly bound to the CM with archaella still attached (Fig. 3). Metlina (Metlina 2004) has presented evidence for two different structures, the DLS associated with the archaella and another well-ordered structure located under the cell membrane similar to the ‘polar organelle’ described much earlier in bacteria (Murray and Birch-Andersen 1963) (Fig. 3).

Rotation of archaella depends on ATP not ion gradients

A key finding outlining another fundamental difference between archaella and flagella was the report by Streif et al. (Streif et al. 2008) that rotation of the archaella of H. salinarum depended on ATP and not ion gradients as in flagella. H. salinarum can generate ATP by bacteriorhodopsin (BR)-mediated photosynthesis, by respiration or by fermentation of arginine. When cells are incubated under anaerobic conditions without any energy sources, the cells eventually become non-motile. Motility was shown to be gradually restored when the ATP pool of the cell was also restored. If a proton gradient was established but the ATPase was inhibited, motility could not be restored. However, motility could be restored even under conditions of zero membrane potential, providing that arginine fermentation took place. Taken together, these data showed that motility, and implicitly archaella rotation, was ATP-dependent.

Other roles for archaella besides swimming

Future studies would reveal a number of additional roles for archaella besides swimming but Nather et al. (Nather et al. 2006) were the first to show an additional function for the archaella of P. furiosus. They observed by a variety of electron microscopic techniques that the archaella of P. furiosus could form cable like aggregates of many archaella that formed cell to cell connections during stationary phase (Fig. 6). In addition, archaella were used to attach P. furiosus cells to a variety of solid surfaces including sand grains taken from the original habitat of the organism. Later, it was shown that the cell-to-cell contacts mediated by archaella could be to cells of other species like Methanopyrus kandleri resulting in an archaeal bi-species biofilm at 95°C (Schopf et al. 2008).

Figure 6.

Attachment of cells to surfaces by archaella and cell-to-cell connections by archaella bundles. (A) Scanning electron micrograph showing attachment of Mcc. villosus cells to a surface and to other cells via bundles of archaella. Bar = 1 µm. From Jarrell et al. 2013. Life 24:86–117. Original picture courtesy of Gerhard Wanner, University of Munich, Germany. (B)P. furiosus grown on carbon-coated gold grids. Scale bar = 2 µm. From Nather et al. 2006. J. Bacteriol. 188:6915–6923, with permission. (C)Mc. maripaludis interacting with each other and the underlying surface via archaella bundles. Scale bar = 1 µm. From Jarrell and Albers 2019. Archaellum In: Schmidt T.M. (ed) Encyclopedia of Microbiology 4^th Edition Vol 1:253–261, with permission.

First structural models for archaella show uniqueness and lack of central filament channel

Until the work of the Trachtenberg group in the early 2000s, virtually nothing was known about archaella structure and symmetry. Initially their work focused on the filaments of H. salinarum R1M1. The major findings were that the archaellum differed significantly from flagella in symmetry and structure and was closer to that of type IV pili. The filament diameter was approximately 10 nm and there was a ∼3.3 subunits/turn as reported for some type IV pili structures, compared to ∼5.5/turn for plain flagella (Cohen-Krausz and Trachtenberg 2002). Of great significance was the finding, obtained by examination of a three-dimensional density map obtained from cryo-negative-stained filaments, that the core of the filament did not contain a central channel that could allow passage of subunits from the cytoplasm to the filament tip. This was an important piece of evidence in support of the type IV pilus model of archaellum assembly proposed earlier (Jarrell, Bayley and Kostyukova 1996a) and spelled the end of any possibility that archaella were assembled in a flagellum-like fashion. These findings were extended using the archaella from H. salinarum M175 which produces super-archaella bundles which can be converted easily to relatively straight filaments which are more favorable for high resolution analysis (Trachtenberg, Galkin and Egelman 2005). A three start filament core was presented surrounded by three larger globular domains. The Iterative Helical Real Space Reconstruction (IHRSR) map of the archaella was a very good fit with the atomic structure of the PAK type IV pilin of P. aeruginosa. The archaellum was described as ‘a bacterial propeller with a pilus-like structure’ (Trachtenberg and Cohen-Krausz 2006). A third study focused on the crenarchaeote S. shibatae (Cohen-Krausz and Trachtenberg 2008). S. shibatae filaments were thicker at 13–14 nm diameter and naturally straight with striations regularly spaced at 5.4 nm intervals; a similar large diameter (14.5 nm) was reported for archaella of S. solfataricus (Szabo et al. 2007b). As also reported for S. solfataricus (Szabo et al. 2007b), the filaments of S. shibatae possessed a right-handed helicity. Despite the great phylogenetic distance between S. shibatae and H. salinarum, the archaella structures shared general characteristics that made them distinct from flagella samples leading to the prediction that the basic symmetry would be conserved for all archaella.

Archaella formation is not constitutive and regulated by growth conditions

While an understanding of the molecular basis of the control of archaellation would not start to unfold until the next decade, studies in the 2000s clearly demonstrated that archaellation was not constitutive.

The first paper to show regulation of arl cluster genes in any archaeon was in Mcc. jannaschii (Mukhopadhyay, Johnson and Wolfe 2000) where the controlling factor was, unusually, H₂ partial pressure. When Mcc. jannaschii was grown in a reactor under conditions of hydrogen excess, the cells were generally devoid of archaella and proteomic analysis revealed very low or even undetectable levels of four Arl proteins, including a major archaellin. Conversely, the cells started to synthesize archaella when hydrogen became limited.

Shortly thereafter, a proteomics study revealed differences in the abundance and apparent modifications of ArlB1 and ArlB2 in Mcc. jannaschii in response to both growth conditions and growth phase, although no electron micrographs of the cells under the various conditions were reported (Giometti et al. 2001). 2-DE analysis revealed that Mcc. jannaschii could express multiple forms of both ArlB1 and ArlB2 differing in mass and pI with the higher molecular mass forms favored at normal H₂ levels and the lower form at low H₂ levels. The production of archaellins, albeit different forms varying in mass, under conditions of high or low H₂ concentration, differed from the earlier findings where only very low amounts of archaellin were detected under conditions of excess H₂ (Mukhopadhyay, Johnson and Wolfe 2000). The lower MW forms of ArlB1 and ArlB2 were barely or not detected in exponential phase but significantly expressed in stationary phase and only minimal amounts of any ArlB1 and ArlB2 forms were observed when cells were grown in limiting amounts of ammonium. The differences in archaellin mass observed under different growth conditions might represent archaellins modified by different levels of glycosylation, or perhaps even fully glycosylated vs non-glycosylated forms.

Leigh's group studied the global response of Mc. maripaludis to leucine, phosphate, nitrogen and H₂ limitations using quantitative proteomics, RT-PCR and microarray analysis (Xia et al. 2006; Hendrickson et al. 2008; Xia et al. 2009). In these global response studies, cells were not examined by EM for the presence of archaella, however. Nitrogen limitation negatively affected archaellins, possibly to allow the N₂-fixing Mc. maripaludis to conserve energy under demanding growth conditions. Archaellins were positively affected under phosphate and H₂ limitation, the latter finding consistent with results in Mcc. jannaschii. Transcripts for arl genes were also shown to decrease in abundance under conditions of leucine limitation.

Szabo et al. (Szabo et al. 2007b) reported a strong induction of arlB in stationary phase compared to log phase in S. solfataricus grown in rich medium but arlB induction could not be correlated with increased archaellation due to technical limitations. They also showed there was strong induction of arlB under unfavourable nutritional conditions, namely in minimal medium or when cells were grown on sugars instead of the preferred peptides.

A rare example of arl gene transcription at this time was the report by Tarasov et al. (Tarasov et al. 2000) that the transcription of the arlB gene cluster was linked to that of the arlA gene cluster in H. salinarum. If the arlA gene cluster is disrupted or even if only arlA2 is disrupted, then transcription of the arlB genes is inhibited, but not vice versa, suggesting transcription of arlA occurs first and these archaellins are incorporated first.

More similarities of archaella to Type IV pili and Type II secretion systems

Comparisons of type IV pili, type II secretion and archaellum systems led to the identification of several common features: major and minor structural components (pilins, archaellins), a prepilin peptidase-like enzyme to process the structural preproteins, plus a homologous ATPase and a multi-spanning transmembrane component (Table 2). The ATPase ArlI of the archaella system was first shown to be homologous to ATPases of type IV pili systems/type II secretion systems by Bayley and Jarrell (Bayley and Jarrell 1998). Figurski's group expanded this finding by examining the phylogeny of secretion ATPases and found that the ArlI group grouped with bacterial tadA-like enzymes first reported in Actinobacillus actinomycetemcomitans (now Aggregatibacter actinomycetemcomitans) (Kachlany et al. 2000; Planet et al. 2001). Peabody et al. (Peabody et al. 2003) showed that the archaeal system had both the ATPase homologues of Type IV pili and Type II secretion systems as well as the transmembrane component, which in the archaella system is ArlJ. In their phylogeny tree, ArlI was not specifically grouped with TadA. ArlJ is significantly larger than TM components of the Type IV pili and Type II secretion systems with 7–9 TM domains while the bacterial counterprts have only 4. Albers and Driessen (Albers and Driessen 2005) showed that ArlI associated with the membrane of S. solfataricus and demonstrated for the first time that it possessed cation-dependent ATPase activity.

Linking the chemotaxis system to the archaellum begins

Chemotaxis systems are for the most part beyond the scope of this review. Interested readers are referred to other reviews (Quax, Albers and Pfeiffer 2018a; Szurmant and Ordal 2004) and to an analysis of the protein-protein interactions of the taxis transduction system (Schlesner et al. 2012). One of the more interesting observations of the archaella system is the presence of a chemotaxis system that has the hallmarks of the bacterial system and which was likely transferred into archaea via horizontal gene transfer (Wuichet and Zhulin 2010; Briegel et al. 2015). Almost all of the data on taxis in archaea comes from studies on H. salinarum which have revealed that the archaeal signal transduction cascade and the specific Che proteins function in archaea as they do in bacteria. The key part is that CheA relays an environmental signal to CheY, phosphorylating this protein. Phosphorylated CheY, as in bacteria, can then change the rotation of archaella resulting in a change in swimming direction. Given the fundamental differences in the archaellum and flagellum including the lack in archaea of the flagellum switch component FliM to which bacterial CheY-P binds, it was an intriguing question how the chemotaxis and archaellum systems interacted. Using protein-protein interaction methodologies in H. salinarum, Schlesner et al. (Schlesner et al. 2009) showed that CheY interacted with two proteins called CheF1 and CheF2. CheF1 and CheF2 additionally were shown to interact with the arl proteins ArlCE and ArlD. These archaeal-specific proteins thus were shown to be essential in relaying chemotaxis signals to the archaella.

It should also be noted that while chemotaxis systems are found in archaellated Euryarchaea species as well as members of the Thaumarchaea, they are not found in archaellated members of Crenarchaeota, possibly suggesting an undiscovered sensory system in these archaea or the use of archaellation as simply a dispersal mechanism without the need for a chemotaxis system.

2010s

Key findings: Archaeal motility structure renamed the archaellum; regulation of arl transcription; atomic models of archaella; roles for most Arl proteins; Haloferax volcanii becomes an important model organism for archaella studies; first reports of archaella in Nanoarchaeota and Thaumarchaeota; high swimming speeds recorded; more functions for archaella; archaella and chemotaxis system interactions; arl gene cluster studies in more archaea; more studies on archaellin N-linked glycosylation; biotechnology studies on archaella

The name change: archaeal flagella become archaella

By 2012, the evidence was overwhelming that each of the three domains had a distinct organelle responsible for swimming motility, all called the ‘flagellum’. Even between the two prokaryotic domains, Bacteria and Archaea, the structures were unrelated evolutionarily with abundant evidence instead pointing to a specific relationship of archaella and type IV pili (Table 2). We felt that a new name was justified to distinguish the archaeal structure (Jarrell and Albers 2012). There were previous attempts to denote bacterial and eukaryotic motility structures with distinct names, perhaps most notably by Lynn Margulis (Margulis 1980). She argued that the term flagella was ambiguous, referring to both the bacterial and totally unrelated eukaryotic organelle and suggested the use of the term undulipodia, meaning ‘little waving feet’ to refer to the eukaryotic flagella while retaining the flagella epithet for the bacterial structures. The suggested changes were not universally adopted possibly because undulipodia didn't convey what flagella really did (Corliss 1980) or perhaps because the term was a mouthful. Even Margulis (Margulis 1980) had to admit ‘the adjective ‘undulipodiated’ and the noun ‘undulipodiates" for flagellated and flagellates respectively, may not be euphonious to all ears’. The term flagellum became even more ambiguous with the discovery of the unique status of the archaeal organelle. Now, the data demonstrated clearly that there were three completely unrelated structures all called flagella. When we proposed ‘archaellum’ as a new name for the archaeal organelle, we preferred that the new term could be used easily, rather than being derived from an ancient language that would accurately describe it (Jarrell and Albers 2012). Critically, the proposed term ‘archaellum’ could be used like flagellum; subunits would be archaellins and cells would be archaellated. Not surprisingly, there was initial support and criticism for the idea (Eichler 2012; Wirth 2012) but it is now a widely, if not yet universally, accepted term both within the archaeal field and among bacterial researchers. More recently, it was recommended by Pohlschroder et al. (Pohlschroder et al. 2018) that the gene designations should be changed from fla to arl (archaellum) to complete and unify the transition. The scientific world is transitioning to the use of three distinct terms to recognize the three distinct motility organelles in the three Domains of life; cilia for Eucarya; flagella for Bacteria; archaella for Archaea (Khan and Scholey 2018; Beeby et al. 2020).

Deciphering how the archaeal ‘bacterial-like’ chemotaxis system interacts with the archaellum continued

Alatyrev et al. (Alatyrev et al. 2010) searched for proteins that could connect the chemotaxis system of archaea to the archaellum and they discovered independently of Schlesner et al. (Schlesner et al. 2009) the same protein (CheF1) but they named the same gene as cheM. They showed interactions of CheM to immobilized His-tagged CheY in both H. salinarum and in P. horikoshii. According to the STRING database on known and predicted inter-protein interactions, they found the most probable partners of CheM to be CheY and CheC (a phosphatase for CheY-P). Schlesner et al. (Schlesner et al. 2012) subsequently found CheC2 (one of three CheC proteins) interacted with CheF1 and CheF2 which are adaptors connecting the chemotaxis system to the archaella. This might indicate that dephosphorylation of CheY-P happens at the archaellum switch in H. salinarum, as shown for B. subtilis CheY-P.

Despite its close similarity to bacterial CheY, archaeal CheY interacts with the archaeal adaptor protein CheF and not the bacterial switch protein FliM, a homologue of which is not found in archaea. How this is accomplished was studied by Quax et al. (Quax et al. 2018b). They showed that both the Mc. maripaludis CheY version and bacterial CheY have the same overall structure and mechanism of Mg²⁺-dependent phosphorylation. In contrast to bacterial CheY, however, the methanococcal CheY has a negatively charged region in the N-terminus of helix 4–important for the CheY/FliM interaction in bacteria—and this charge difference may be crucial for the interaction with CheF. The CheY-CheF interaction is dependent on the phosphorylation of CheY and is much stronger in a mutant of CheY that cannot be dephosphorylated. The structure of CheY from P. horikoshii has also been published. The phosphorylated residue and residues involved in transducing the phosphorylation signal to a biologically relevant output superimpose well, but the negatively charged region observed in Mc. maripaludis CheY is, in P. horikoshii, more similar to the bacterial homologue (Paithankar et al. 2019).

First reports of archaella in new phyla

The new decade of the 2010s saw the first reports of archaella in the phyla Nanoarchaeota and Thaumarchaeota. While the first nanoarchaeote isolated, Nanoarchaeum equitans from marine habitat (Huber et al. 2002), is non archaellated and its genome does not include genes for archaella (Podar et al. 2013), Nanoarchaeota have been more recently isolated from terrestrial thermophilic habitats (Candidatus Nanopusillus acidilobi) with genomes that contain arl genes (although not a traditional arl gene cluster (St John et al. 2019)). The confirmed expression of archaellum proteins indicated a likely ability to synthesize archaella under certain conditions, perhaps upon detachment of the cells (Wurch et al. 2016). St. John et al. (St John et al. 2019) demonstrated archaella on detached cells of Candidatus Nanoclepta minutus, which also has a complement of archaellum genes.

Nitrososphaera viennensis (Steiglmeier et al. 2014) is an archaellated ammonia-oxidizing member of the phylum Thaumarchaea (Brochier-Armanet et al. 2008). Archaella have also been observed on certain members of the genus Nitrosopumilis and the genomes of others have a seemingly complete arl gene cluster, implying that other members of this phylum are also archaellated (Bayer et al. 2016; Qin et al. 2017).

Hfx. volcanii becomes a model organism for archaellation

Although Serganova et al. had initially reported the isolation of archaella in 1995 ((Serganova et al. 1995), it was the multiple publications of the Pohlschroder group that established Hfx. volcanii as another model organism with a well-developed genetic manipulation system for study of archaellation, despite this organism being originally described as nonmotile (Mullakhanbhai and Larsen 1975). They formulated a defined medium in which the cells were routinely and clearly motile in swarm agar (Tripepi, Imam and Pohlschröder 2010). They identified an arl gene region in Hfx. volcanii that had two archaellin genes, separated by a single divergently transcribed gene, followed by the typical arl accessory genes arlC/E, F, G, H, I and J. While Hfx. volcanii is a euryarchaeote, it processes all substrates bearing class III signal peptides by a single PibD-like enzyme like the crenarchaeote S. solfataricus, indicating that the processing strategy of the euryarchaeote Methanococcus species, which uses a separate enzymes for archaellins and pilins, may be the rarity. The pibD mutant was, as expected, non-motile. It was also nonadherent, implicating pili as surface adhesion structures, since mutants deleted for archaellins, while nonmotile, had no adhesion defect. Deletion of aglB also rendered the cells non-motile, highlighting N-glycosylation as a critical modification of archaellins for formation of archaella, as seen in Mc. voltae and Mc. maripaludis.

Tripepi et al. (Tripepi et al. 2012) showed by MS analysis that purified Hfx. volcanii archaella were composed mainly of ArlA1, with only a minor amount of ArlA2, with both staining as glycoproteins. Deletion of only arlA1 led to non-motile cells. The archaellin N-linked glycan was determined to be the same pentasaccharide as reported on the S-layer protein (Calo, Kaminski and Eichler 2010; Kandiba et al. 2016). Mutant cells synthesizing non-glycosylated archaellins, as in the aglB mutant, or archaellins modified with only a single or the first two sugars of the glycan could not assemble functional archaella, but cells making archaellins with the first 3 sugars could. Cells lacking aglB can still make pili and adhere to surfaces and form microcolonies in the absence of archaella.

Another study focused on the role of the minor archaellin ArlA2 (Tripepi et al. 2013). A small subset (2%) of cells that are deleted for arlA1 and produce only the minor archaellin ArlA2 produce archaella, although cells are nonmotile (Note that later Esquivel and Pohlschroder (Esquivel and Pohlschroder 2014) reported that ArlA2 alone is sufficient to make functional archaella). Very interestingly, deletion of arlA2 alone results in cells still producing ArlA1 and these cells are hypermotile with longer and more abundant archaella and expression of arlA2 in trans in an arlA2 deletion mutant does not reduce the hypermotility phenotype. The authors conclude that ArlA2 has both a structural and a regulatory role and this regulatory role may depend on the relative abundance of ArlA2 to ArlA1. A hypermotility phenotype resulting from the deletion of an archaellin gene has not been observed in other archaea.

The Pohlschroder group showed a connection between archaeosortase A and motility. ArtA is involved in the proteolytic processing and anchoring of the S-layer glycoprotein (Abdul-Halim et al. 2020). An artA deletion strain had multiple phenotypes including a severe motility defect (Abdul Halim et al. 2013). Although the reason for the motility impairment in the artA mutant is unknown, one possibility is that defects in the S-layer interfere with proper anchoring of the archaellum motor.

Other possible genes involved in motility of Hfx. volcanii were revealed following a screen of a Hfx. volcanii transposon library for nonmotile mutants (Legerme et al. 2016). For many of the transposons, it was not formally shown whether the site of the insertion was actually the gene that caused the defect but for ∆hvo_2876, encoding a hypothetical protein, deletion of the gene resulted in nonmotile and non-adhering cells that had much reduced levels of both archaella and pili.

Analyses of archaella and arl gene clusters in other archaea

Pyrococcus furiosus and an unusual arl gene cluster

The composition of purified archaella from P. furiosus, when analyzed by SDS-PAGE, revealed a single glycoprotein band but its N-terminal sequence did not match exactly any of the multiple archaellins available in the database. Ultimately, Näther-Schindler et al. (Näther-Schindler et al. 2014) determined that a piece of the original genome sequence was missing in this region and that another archaellin gene encoding a protein that matched their N-terminal sequence was on this missing piece. This archaellin gene was upstream of two others, arlB1 and arlB2, and was thus designated as arlB0. All three archaellins could be detected in the single protein band obtained in SDS-PAGE but a later atomic model of the P. furiosus archaellum revealed that a density map of the archaellum could only accommodate ArlB0, indicating that the major part of the filament is composed of ArlB0 (Daum et al. 2017). The arl gene cluster of P. furiosus consisted of genes very similar to other euryarchaea with three archaellins followed by arlC D F G H I J. There was no arlE. Transcription of the arl genes was limited in early exponential phase. Single transcripts for arlB0 and for arlJ were found but no single transcript spanning arlB0 to the end of arlJ was found. Instead, numerous co-transcripts of various arl genes were identified by northern blots and RT-PCR analyses. The arl gene cluster is unusual in that it begins with two genes upstream of the first archaellin arlB0, which can form a co-transcript with arlB0. The first gene is a likely transcriptional activator while the second is a postulated methyltransferase. Evidence was presented later that the transcriptional activator is a homolog of EarA, the transcription factor necessary for transcription of the arl cluster in Mc. maripaludis (Ding et al. 2017a). The methyltransferase was named fam for flagella associated methyltransferase with unknown specificity, and although it was postulated to act on archaellins there is no data as yet to support this.

Lewis et al. (Lewis et al. 2015) reported on a ‘lab strain’ of P. furiosus that had, over time, accumulated mutations that resulted in growth characteristics that were distinct from the wildtype culture, including several frameshifts in arl genes. This lab strain was not archaellated, and several observations suggested that archaella might be involved in other processes such as cell aggregation and biofilm formation.

Deletion of arl genes in S. acidocaldarius

Lassak et al. (Lassak et al. 2012b) deleted each of the seven genes of the arl gene cluster in S. acidocaldarius and showed by electron microscopy that all were essential for archaellation including, for the first time, arlX. Complementation with the wildtype copy of the deleted gene in trans restored archaellation to wildtype levels except for arlB and arlG, where complementation resulted in short, incomplete archaella.

Halorubrum lacusprofundi

H. lacusprofundi ACAM34 is a psychrotrophic haloarchaeon with a complete arl gene cluster containing a single archaellin gene arlB (Syutkin et al. 2012). Despite formally being described as nonmotile and non-archaellated, demonstrably motile cells were enriched for and archaella were isolated. Archaella were composed of a single major glycoprotein of 50 kDa that was identified as the product of the one archaellin gene. This is a rare case of an extreme halophile where archaella are made from a single archaellin.

More recently, Pyatibratov et al (Pyatibratov et al. 2020) reported that H. lacusprofundi strains could have one archaellin gene (arlB2) as in H. lacusprofundi ACAM 34 or two archaellin genes as in H. lacusprofundi DL18 (arlB1 and arlB2). Both strains produced archaella but DL18 was more motile. Each of the two arlB genes were expressed, either singularly or together, in a Hfx. volcanii strain that had its two archaellin genes deleted. Cells expressing ArlB1 or both archaellins were more heavily archaellated than cells expressing only ArlB2, for which many cells were non-archaellated. In addition, cells expressing arlB2 had only a small percentage of motile cells (5%) compared to cells expressing arlB1 (30%) or cells expressing both archaellins (60%).

The DL18 archaella were found to be always heteropolymeric, composed of both archaellins. This is interesting since all the existing atomic models have shown archaella are composed of only one type of archaellin even in cells with multiple archaellin genes, even though all the multiple archaellins are typically found in purified archaella samples. The two-component archaella of strain DL18 are more stable and much more resistant to lower salinity than the single-component archaella of strain ACAM34, suggesting a cooperative and close interaction between the two archaellins in the filaments. Archaella with both archaellins allow for adaptation to a wider variety of external conditions. The authors conclude that ArlB1 and ArlB2 ‘form a stable heterodimer that then assembles into the archaella.’

Haloarcula marismortui: archaellins as ecoparalogs

In H. marismortui, there are two archaellin genes, arlB and arlA2. Unusually, neither has a canonical N-glycosylation sequon although both archaellins stained weakly as glycoproteins (Syutkin et al. 2014b). Two previously isolated strains of H. marismortui produced archaella with either only ArlB or with ArlA2 and only a minor amount of ArlB (Pyatibratov et al. 2008).. The ArlB strain lacks the plasmid that has the arlA2 gene so its archaella can only be composed of ArlB. Both the ArlB and ArlA2 strains are motile with functional archaella. The ArlA2 strain swam better under more extreme conditions (higher temperature and lower salinity), while the ArlB strain functioned better as the salinity increased. It was proposed that ArlA2 and ArlB archaellins replace each other under different environmental growth conditions and hence they were called ecoparlogs, i.e. paralogous genes that have different environmental specialization. This was the first example in archaea where one archaellin could replace another when grown under different environmental conditions.

N-Glycosylation is necessary for proper archaella assembly and function but is N-glycosylation of archaellins actually necessary?

It was now well established in several different archaea that mutations that led to defects in the N-glycosylation pathways led to defects in archaellum assembly or function. What was not clear was whether these archaella effects were due to defects in N-glycosylation of archaellins directly, or to indirect effects due to the effects on functions of other, unrelated proteins that were N-glycosylated. For example, S-layers of archaea are widely modified by N-glycosylation and other proteins encoded in the arl gene cluster have been shown to contain potential or proven N-glycosylation sites (e.g. ArlG, ArlF, ArlJ and ArlX in Sulfolobus acidocaldarius: (Meyer, Birich and Albers 2015; Tsai et al. 2020)). A fortuitous spontaneous mutation was observed in Mc. maripaludis ArlB1 where one glycosylation site was removed; however this mutant was still archaellated, indicating that not all glycosylation sites were essential for archaellation in this methanogen (Jones et al. 2012). Studies to specifically address the requirement of archaellin glycosylation in archaellation were conducted on three model organisms with, surprisingly, three different results. Tripepi et al. (Tripepi et al. 2012), working on Hfx. volcanii, provided evidence that three potential sites of N-glycosylation in archaellin ArlA1, out of six total sites, were occupied by a pentasaccharide found also on the S-layer protein. Peptides containing the three other potential sites were not detected in this study although a fourth occupied site was later identified (Esquivel et al. 2016). Interestingly, each of the three modified sites were shown to be critical for motility as complementation of an arlA1 deletion with arlA1 in which any one of the three sites were mutated to lack the canonical N-glycosylation sequon, did not rescue the swimming defect and there was no evidence of archaella. A second study, on the single archaellin of S. acidocaldarius, told a completely different story (Meyer, Birich and Albers 2015). In ArlB, there are six potential sites of N-glycosylation and all are occupied. ArlB deletion strains complemented with a mutant ArlB version missing all six glycosylation sites were motile although with their motility reduced by 40%, indicating that a functional archaellum could be assembled with completely non-glycosylated archaellins. Since it was shown previously that cells carrying a deletion in agl16 (encoding a thermostable glycosyltransferase) resulting in archaellins modified with a glycan missing only the terminal glucose, had severe motility defects, it may be that other glycoproteins need to have a full glycan to function and this affects the assembly/function of the archaellum. These may include archaellum accessory proteins. Finally, a third paper to look at the requirement of archaellin glycosylation examined the effect of removal of known glycosylation sites in the ArlB2 archaellin of Mc. maripaludis (Ding et al. 2015). ArlB2 has 5 potential glycosylation sites but the first site is unoccupied. The four occupied sites were changed to address whether any one site or a combination of sites were needed for archaella assembly and function, leading to the conclusion that ArlB2 had to be glycosylated at at least some site for archaellation and motility but no specific site was essential. Interestingly, when the arlB2 mutant of Mc. maripaludis S2 was complemented with an archaellin from Mc. maripaludis ∆RC that had 95% amino acid sequence identity but two extra glycosylation sites, the complemented cells were more motile in swarm plates than cells complemented with the S2 version of arlB2.

More glycosylation studies on archaella

Zaretsky et al. (Zaretsky et al. 2019) showed that an N-linked tetrasaccharide composed of a hexose and 3 hexuronic acids, decorated the 5 archaellins of H. salinarum. The structure differed from previously reported results (Wieland, Paul and Sumper 1985), in that no evidence for sulfation of the hexuronic acids was found. Analysis of an aglB mutant revealed that these cells were impaired for motility and were non-archaellated, as in other archaea. Reduced transcription of archaellin genes was observed in the aglB mutant, suggesting that N-glycosylation may be important for archaellin gene transcription.

In Mtc. thermolithotrophicus, the arl gene cluster contains four archaellin genes (arlB1, arlB2, arlB3 and arlB4), and all contain multiple N-linked glycosylation sequons (Kelly et al. 2020). Analysis of sheared, purified archaella by SDS-PAGE revealed the presence of two or three major protein bands, depending on the length of time the samples were boiled prior to loading on the gel, as previously reported (Kostyukova et al. 1992). The protein bands were identified as ArlB1 and ArlB3, with no peptides unique to ArlB2 or ArlB4 detected. Both ArlB1 and ArlB3 were modified with a unique branched N-linked glycan composed of seven sugars. Failure to detect ArlB2 or ArlB4 may be due to either their low abundance in archaella or because they occupy a cell proximal location and mainly remain attached to the cells following shearing.

Functions of archaella other than motility

To date, the archaellum is the only known motility structure in archaea but the archaellum is known to have other functions besides motility.

Bellack et al. (Bellack et al. 2011) observed that on gold grids, cells of the newly isolated hyperthermophilic methanogen Methanocaldococcus villosus were found enmeshed in a dense network of archaella. Up to half of the cocci were connected to each other by archaella cables (Fig. 6), showing that archaella can be used for motility as well as adherence to other cells and abiotic surfaces as shown earlier for P. furiosus (Nather et al. 2006). Co-cultures of P. furiosus and Mcc. villosus formed large aggregates where the different cells interacted by their archaella. Such interaction was observed both on different solid surfaces and when cells were grown in liquid medium (Weiner et al. 2012). This interaction was suggested to enhance H₂ exchange.

Jarrell et al. (Jarrell et al. 2011) used mutants incapable of making pili, archaella or both appendages to show that both archaella and type IV pili are required for efficient attachment of Mc. maripaludis to various surfaces. Attachment was via archaella which left the cells as thick cables before unwinding at the surface and these bundles also connected cells (Fig. 6). It is unclear why cells possessing archaella—but not pili—cannot attach, but it was surmised that initial attachment to surfaces may be mediated by pili, and only then can archaella make a more permanent attachment. The involvement of archaella in surface attachment was not found in H. salinarum or Hfx. volcanii (Tripepi, Imam and Pohlschröder 2010; Losensky et al. 2015).

Unlike many studied archaea that have numerous archaella per cell, wildtype Sulfolobales species typically have only one or very few per cell. The role for archaella and different pili in attachment of two different species to several abiotic surfaces varied. Wild-type S. solfataricus cells were able to attach to a variety of abiotic surfaces but neither non-archaellated mutants or mutants unable to synthesize UV inducible (Ups) pili could do so, implicating both surface structures as necessary for adherence (Zolghadr et al. 2010). In S. acidocaldarius, three different T4P-like structures that could play a role in adhesion are produced: Aap pili, Ups pili and archaella. However, at least in adhesion to glass, Aap pili promote the attachment to surfaces, whereas Ups pili and archaella only displayed a minor role in surface adhesion (Henche et al. 2012a).

A novel role for archaella was postulated when it was demonstrated by conductive atomic force microscopy that the archaellum of Msp. hungatei is electrically conductive (Walker et al. 2019), possibly aiding in surface attachment or even signalling between cells. Electrically conductive appendages have been suggested to aid extracellular electron transfer within syntrophic communities.

Swimming speeds

There are few reports of swimming speeds of archaea, but early studies reported swimming speeds of extreme halophiles to be much slower than E. coli (2 µm/s for H. salinarum; (Alam and Oesterhelt 1984). Herzog and Wirth (Herzog and Wirth 2012) studied swimming speeds of many selected archaea and reported a huge variation from very slow for H. salinarum (3 µm/s) to very fast for Mcc. jannaschii (380 µm/s). In terms of body lengths per second, Mcc. villosus was 470 body lengths/s at maximum speed, the highest for any organism on Earth. They also observed that archaea, especially hyperthermophiles, exhibited two kinds of swimming which they called relocate and seek. In the relocate mode, cells swim rapidly in straight lines and tumbles were not observed. This kind of swimming would allow cells to move rapidly to another environment from an unfavorable one. The seek type of swimming observed was a much slower zig zag swimming that occurred if cells swam close to a surface. This type of swimming would allow cells to stay in a locally restricted environment sensed to be favorable and allow for attachment to a surface. The two types of swimming, coupled with archaella being used either for swimming or attachment, may allow cells to stay in their optimum native environment. It has been suggested that the switch from fast swimming mode to the much slower zig-zag scanning mode of swimming may be because all archaella rotate for the fast swimming mode, while only some rotate in the slower zigzag mode. Consistent with this is the observation that cells can attach to very small particles and move those around; here some archaella would be involved in attachment to the solid while others would be still involved in swimming (Wirth, Luckner and Wanner 2018).

Mora et al (Mora et al. 2014) reported the temperature ranges for swimming and swimming speeds for 15 species of Thermococcus as well as Mcc. villosus. In the case of Mcc. villosus, the cells swam at the highest speed at 140°C possibly to try to leave a lethal temperature zone; cells kept at this extreme temperature died within 2–3 min. Studies with Thermococcus stetteri provided the first evidence of thermotaxis in Archaea.

Shahapure et al. (Shahapure et al. 2014) studied swimming in S. acidocaldarius at different temperatures. By observing cells tethered via an archaellum to a glass capillary in a thermomicroscope, the authors could rule out that the movement was due to an extension‐retraction mechanism as observed for bacterial type IV pili (Mattick 2002). For the majority of tethered cells (70%), archaellum rotation occurred continuously in counterclockwise direction while 30% of the population exhibited either spontaneous switching or clockwise rotation, in a stochastic process. At 75°C, swimming speed was measured at an average velocity of 13.5 ± 3.5 μm/s with over half the cells observed to be motile but this dropped to 7.12 μm/s at 50°C, where less than 1% of the cells were motile. It was proposed that the swimming speed response to temperature may influence cells in nature with a bias towards surface colonization at lower temperatures while favouring fast movement away from surfaces at higher temperatures.

Growth conditions and archaella production

Differential expression of the archaellum has been found since 1993, when it was published that the archaellation of Msp. hungatei is dependent on calcium levels and temperature (Faguy, Koval and Jarrell 1993). More recently, several additional studies examined the correlation of archaellation with growth conditions in various archaea. Ding et al (Ding et al. 2016a) found that growth temperature but not variations in medium composition affected archaellation in Mc. maripaludis. Cells grown between 22–35°C were well archaellated and had similar amounts of ArlB2 when examined by western blot, with ArlB2 levels decreasing with increasing temperature up to 42°C, when ArlB2 synthesis ceased. qRT-PCR experiments were in agreement with the western blots, and cells at 42°C were largely non-archaellated when examined by EM.

While Mcc. jannaschii had been shown to produce archaella in a H₂-dependent manner (Mukhopadhyay, Johnson and Wolfe 2000), Topçuoğlu et al. (Topçuoğlu et al. 2019) showed, using RNA-Seq, that arl gene expression does not change under high or low pH₂, two seemingly contradictory results that may be explained by the use of chemostat cultures in one study and a batch reactor in the other.

Tripepi et al. (Tripepi, Imam and Pohlschröder 2010) showed that Hfx. volcanii was motile in a complex medium but not motile in a defined medium (CDM) where arl cluster genes were expressed at only low levels. They reported a new defined medium (CA) in which cells were reproducibly motile but the key ingredients of CA vs CDM media that allowed for motility were not identified. Furthermore, Hfx. volcanii cells are specifically motile when they are rod-shaped in the early log-phase; afterwards, they adopt an immotile round-shaped morphology (Duggin et al. 2015; Li et al. 2019). Kinosita et al. (Kinosita et al. 2016) found that when H. salinarum was grown in casamino acid medium, the proportion of swimming cells is higher (85% swimming cells) than when cells are grown in the usual peptone based medium.

Motility of certain Thermococcus species, namely T. guaymasensis could also depend on the growth media (Mora et al. 2014). Some species produced archaella in some media and not in others, and were motile only in those media where archaella were produced.

Regulation of archaellation

Sulfolobus

Although most of the initial research on the archaellum focused euryarchaeotes, archaellum regulation is better understood in crenarchaeotes, with most of the research been performed in S. acidocaldarius where a host of positive and negative regulators have been identified (Table 3).

Table 3.

Archaellum regulators identified in S. acidocaldarius.

Regulator	Effect of deletion on motility	Transcription and translation levels	Archaella and swimming speeds	PTMs	Effects on parlB or parlX	Effects on Arl protein levels
ArnA	Hypermotile	ArnA levels not increased upon starvation.	KO shows more archaellation	Phosphorylated by ArnC	Repression of parlB under inducing and non-inducing conditions	ArnA knock-out has higher ArlB and ArlX levels under starvation conditions, compared to the wild-type.
ArnB	Hypermotile	arnB upregulated under nutrient depletion. ArnB levels do not increase upon starvation.	KO shows more archaellation	Phosphorylated by ArnC and ArnD	Repression of parlB under inducing and non-inducing conditions (note: ArnB may have more impact on ArlB translation than transcription)	ArnB knock-out has higher ArlB and ArlX levels under starvation conditions, compared to the wild-type.
ArnC	Impaired motility	Transcripts and protein increase in starvation	KO has same number of archaella but fewer cells archaellated	Auto-phosphorylates	KO has no effects under starvation or in rich medium	In KO, ArlB accumulates under starvation conditions slightly earlier than in the wild-type
ArnD	Hypermotile	Transcripts slightly upregulated upon starvation	KO has same number of archaella but more cells archaellated	Auto-phosphorylates	KO increases expression of arlB under rich medium.	In KO, ArlB detected when cells were grown in rich medium and ArlB accumulates under starvation conditions earlier than in wild-type
		No obvious change in protein levels			KO has no effect on arlB mRNA levels under starvation
ArnS	Impaired motility	Transcription induced by starvation conditions.	KO has fewer archaellated cells, slower swimming. Archaella slightly shorter.	Auto-phosphorylates.	KO has no effects on arl proteins under nutrient-rich conditions and arlB transcripts still accumulate upon induction, although induction of arlB is delayed.	Deletion has no effects of arl gene transcription under nutrient-rich conditions.
Abfr1	Non-motile			Found phosphorylated in a strain lacking the phosphatase PTP.	In KO, transcripts of both operons are downregulated	abfr1 deletion strain does not accumulate ArlB under nutrient-rich or starvation conditions
					Binds parlB, but binds also other promoters
ArnR	Low motility	Increased transcription during starvation		Phosphorylated by ArnC	In KO, transcripts of both operons are downregulated under starvation conditions. ArnR exerts its function only on parlB.	ArlB and ArlX accumulate under starvation conditions in arnR KO, but much less than in wild-type. Overproduction of ArnR results in ArlB production in absence of stress.
ArnR1	Low motility	No change (low level transcription both in rich medium and under starvation)		Phosphorylated by ArnC	None, but arnR1 knock-out in a arnR deletion mutant results in no induction of arlB.	ArlB and ArlX accumulate under starvation conditions in arnR KO, but less than in wild-type.Overproduction of ArnR1 results in ArlB production in absence of stress.
PP2A	Hypermotile	No change between rich and starvation medium			Deletion of pp2a causes upregulation of all arl genes	Deletion of pp2a causes ArlB accumulation under nutrient-rich conditions

The arl gene cluster of S. acidocaldarius was found to be transcribed as two independent transcriptional units: one controlled by a promoter located upstream of arlB, and the other by a promoter located upstream of the adjacent downstream gene, arlX (Lassak et al. 2012b). The latter promoter was found to be constitutively expressed, while the arlB promoter was found to be activated by starvation conditions, specifically nitrogen starvation. Moreover, it was suggested that the terminator of the first transcriptional unit (i.e. downstream of arlB) was weak, resulting in read-through of the archaellar accessory genes upon induction of arlB. The authors suggested that the organisation of the arl gene cluster in two distinct transcriptional units allows for precise control over the production of proteins and assembly of the archaellum, which are both energy-intensive processes. By having the promoter upstream of arlX constitutively active but on a low level, S. acidocaldarius is able to quickly assemble archaella upon induction of arlB, and further scaling up the production of archaella-associated proteins under circumstances where motility is advantageous.

The first hint for the identity of at least one potential regulator came from studies by Duan and He (Duan and He 2011) who showed that the Forkhead (FHA) domain-containing protein ST0829 of Sulfolobus tokodaii interacted specifically with the arlX promoter. ST0829 had previously been shown to be the substrate of a eukaryote-like Ser/Thr kinase (ST1565), and interaction assays between ST0829 and DNA were shown to be dependent on the phosphorylation status of ST0829. With this information, it was possible to devise an initial model of a regulatory network, in which ST1565 conveys environmental signals to ST0829 by phosphorylating it, therefore modulating its binding affinity to the arlX promotor.

Shortly after the studies of Duan and He (Duan and He 2011), the homologue of ST0829 in S.acidocaldarius was shown to indeed impact archaella synthesis (Reimann et al. 2012). The homologue of ST0829 was named ArnA (archaellum regulatory network), located in a cluster with a von Willebrand A (vWa) domain-containing protein, ArnB. No Euryarchaeotes seem to possess this cluster (Ding et al. 2017a). ArnA and ArnB were later confirmed to interact in vivo (Ye et al. 2020). Both ArnA and ArnB can be phosphorylated: ArnA by the kinase ArnC, and ArnB by both ArnC and ArnD, and both proteins can be dephosphorylated by the Ser/Thr phosphatase PPP2A. ArnA and ArnB were identified as archaellum repressors, since deletion of either or both proteins leads to higher synthesis of ArlB and ArlX under starvation, which translates to hyperarchaellation and faster swimming speed. Strangely, though, the effect of arnA and arnB deletion was modest on a transcriptional level, as neither transcriptional unit driven by the arlB or arlX promoter was as upregulated as would be expected from the swimming phenotype of those strains. This observation hinted at a possible post-transcriptional regulatory role of ArnA and ArnB. Recently, the structure of ArnA and of the close homologue of ArnB, Saci_1209, were solved (Hoffmann et al. 2019). ArnA and ArnB interacted in vivo under nutrient-rich conditions, and this interaction was lost under starvation conditions in a time-dependent manner.

Soon after the documentation of the two repressors in S. acidocaldarius, two activators, named ArnR and ArnR1, where identified by comparative genome map analysis, as the genes encoding both proteins flank the arl gene cluster in S. acidocaldarius (Lassak et al. 2013). ArnR and ArnR1 are paralogues specific to S. acidocaldarius, and both are predicted to have an amino-terminal helix-turn-helix (HTH) domain and a similar C-terminal transmembrane (TM) helix. Between the C- and the N-termini, each protein has a putative HAMP and sensor domain, which differ more between ArnR and ArnR1 than their HTH and TM domains, suggesting that these proteins bind to the same DNA sequence but in response to different stimuli. Strains lacking arnR and arnR1 show less ArlB and ArlX upon starvation, and this effect is more pronounced for the ΔarnR than for the ΔarnR1 strain; a double knock-out strain, however, shows a synergistic effect of these two regulators, as no ArlB nor ArlX are detected. The effect of ArlB and ArlX depletion, as expected, resulted in a reduction or loss of motility. Contrary to the repressors ArnA and ArnB, which possibly exert their regulatory effects on a post-transcriptional level, ArnR has a strong transcriptional effect: expression levels of both arlB and arlX do not increase as drastically upon starvation in the ΔarnR strain as they do in the wild-type. This effect is not observed for the arnR1 deletion, but deletion of both paralogues resulted in no induction of either archaellar promotor, suggesting that both ArnR and ArnR1 act at the transcriptional level, although it cannot be excluded that ArnR1 has also a post-transcriptional regulatory role. ArnR and ArnR1 were identified as activators of the arlB promoter, with higher expression levels of accessory arl proteins due to the read-through effect described before (Lassak et al. 2012b). The binding sites for ArnR/ArnR1 are inverted repeats upstream of the transcription factor B recognition element (BRE) of arlB.

In order to understand the environmental signal sensed by these two activators, the conditions that result in archaella induction in S. acidocaldarius were re-evaluated. Previously, it was reported that the absence of a nitrogen source in the form of tryptone was the trigger for archaella biosynthesis (Lassak et al. 2012b). However, more detailed studies in minimal media led to the conclusion that expression of the arl gene cluster was responding to a low cellular energetic status.

This study by Lassak et al. (Lassak et al. 2013) was the first to identify a transcriptional activator of the arl gene cluster and hinted already at how complex the regulatory network of the archaellum is in S. acidocaldarius. For example: arnR is upregulated 160-fold after 1 h of starvation and necessarily precedes the accumulation of ArlB. Since a positive feedback loop of arnR expression was excluded, the conclusion is that another, yet unidentified hierarchically higher transcriptional factor must be involved in archaellum regulation. Moreover, ArnR and ArnR1 were found to form multimers and to be phosphorylated by ArnC, indicating that there is yet another level of regulation involving these two proteins (Bischof, Haurat and Albers 2019), which does not seem to involve ArnA or ArnB. Furthermore, organisms such as Methallosphaera cuprina code for ArnA but not for ArnR, suggesting that these two proteins may correspond to independent regulatory networks.

Soon after the identification of the activators ArnR/ArnR1, another activator of archaella expression was identified in S. acidocaldarius, during a study exploring the regulation of biofilm formation (Orell et al. 2013). One homolog of the transcription factor Lrs14, renamed Abfr1 (Archaea Biofilm Regulator 1) was shown to bind and activate the promotor of arlB, although the DNA binding of this protein is not very specific. Transcripts of arlB and arlX were found to be downregulated when abfr1 is deleted, resulting in decreased ArlB and ArlX levels and motility loss. This protein was later found in a phosphoproteome study of S. acidocaldarius to be phosphorylated, with phosphomimicking mutants shown in vitro to bind DNA with lower affinity. Besides regulating arlB and arlX, abfr1 also regulates its own activation (Reimann et al. 2013; Li et al. 2017).

At around the same time of the characterisation of known arl repressors and activators, the central role of phosphorylation on the regulation of the archaellum became ever more apparent (Reimann et al. 2013). S. acidocaldarius encodes two predicted phosphatases: Saci_PTP and Saci_PP2A. Saci_PTP is a dual-specific phosphatase for phosphorylated Ser/Thr and Tyr, while Saci_PP2A is specific for phosphorylated Ser/Thr residues. In order to evaluate how phosphorylation impacts the cell, whole-transcriptome analyses were performed in the hyperphosphorylated strains Δsaci_ptp and Δsaci_pp2a, in comparison with the wild-type strain. Several proteins associated with the archaellum or with archaella regulation were identified in this phosphoproteome study, including ArlJ, ArnB, Abfr1, ArnR1, ArnD, and ArnC. While no up- or downregulation of archaella-associated genes was identified for Δsaci_ptp, in the Δsaci_pp2a strain arnR, arnR1, arlX, arlG, and arlI were all found to be upregulated. These results suggested that cells lacking Saci_PP2A would be hypermotile, a result that was confirmed in swimming assays. Moreover, ArlB accumulation in Δsaci_pp2a takes place even in rich medium, with starvation further increasing the accumulation of both ArlB and ArlX. Inhibition of PP2A seems therefore to be part of the cell signalling pathway that leads to the starvation response, which includes archaella biosynthesis. A recent report from the Albers group (Ye et al. 2020) identified proteins interacting with PP2A in S. acidocaldarius cells under starvation conditions. ArnA and ArnB were detected, alongside an ATP/GTP binding protein (Saci_1281) and a universal stress response protein (Saci_0887). The interaction between ArnA and ArnB was shown to be stronger when these proteins are hyperphosphorylated, and to be disrupted by PP2A, curiously independently of the phosphatase activity of the latter. PP2A expression levels do not change under starvation conditions, indicating that despite the strong phenotype upon pp2a deletion, other factors—probably kinase proteins—are important to determine whether the arl gene cluster is expressed or not.

During the last decade, the role of the kinases that had been shown to phosphorylate ArnA and ArnB was also revisited (Hoffmann et al. 2016), as well as that of the the starvation-induced kinase ArnS (Haurat et al. 2017). These two new studies provide some insight not only into the regulatory network of the archaellum in S. acidocaldarius, but also into the sheer complexity of this network, as attested to by the results obtained in both studies. ArnC, ArnD, and ArnS are all kinases that can auto-phosphorylate. ArnC and ArnD can both phosphorylate ArnB, the negative regulator of the archaellum: ArnC and ArnD both phosphorylate ArnB residue T344, and ArnC can further phosphorylate ArnB in T280, T343, and T353. The similar mechanism of ArnC and ArnD towards ArnB would suggest that deletion of either protein would result in the same phenotype but this is not the case: while the ΔarnC strain shows motility defects in swimming assays, ΔarnD is hypermotile. To complicate matters further, ΔarnCΔarnD is also hypermotile, and complementation with either arnC or arnD partially recovers the wild-type phenotype. Both single-deletion strains harbour archaella and both strains show one or two archaella per cell. The only difference between the two strains, as assessed by TEM, is that the number of cells that are archaellated is slightly reduced in ΔarnC and slightly increased in ΔarnD, compared to the parental strain Accumulation of ArnC takes place under starvation conditions, while arnD transcription and translation is constitutive. However, the deletion strain ΔarnD shows arlB transcription and translation even during growth in rich medium, while neither deletion strain differs from the wild-type regarding arl expression in starvation. It is possible that under nutrient-rich conditions ArnD is active and phosphorylates one or more factors (potentially including ArnB), repressing arlB expression. In the absence of ArnD, phosphorylation does not take place, and consequently neither does the repression of arlB. Curiously, though, hyper-, instead of hypophosphorylation seems to be the cell state during starvation, as observed in a phosphoproteome study of S. acidocaldarius strains with either of the two phosphatases deleted (Reimann et al. 2013). Thus, it remains unknown how ArnC and ArnD exert regulation over motility, though the apparent contradictory phenotypes of the knock-out strains seem to point at a fine-tuning of the concentration of both kinases which, depending on the side to which the balance tips, leads to either activation or repression of motility—transcriptionally or otherwise. ArnS is equally mysterious. The expression of this kinase takes place in two steps: at the beginning of starvation arnS is only slightly upregulated, while at a later stage a full 46-fold increase relative to the beginning of starvation is achieved. Deletion of arnS results in a lower number of archaellated cells and in lower swimming speeds, although archaellated cells show the same number of filaments as the wild-type. It was therefore suggested that ArnS regulates the archaellum by a post-translational mechanism directly in the archaellum motor complex. Nevertheless, a mathematical model developed with the aim of understanding the complex regulatory network suggested as well that ArnS may simultaneously inhibit the transcription of arnR and promote the translation of arnR mRNA, hence having also—albeit indirectly—a regulatory role in the transcription of archaella genes.

There is also cross regulation of pili and archaella in S. acidocaldarius (Lassak, Ghosh and Albers 2012a; Henche et al. 2012b). When genes required for the synthesis of Aap pili are deleted, cells produce more archaella. While this was true for all aap genes, the effect was most noticeable for an aapF deletion where a hyper-archaellated phenotype was observed. Cells deleted for the aap genes also showed increased swarming in soft gelrite (a thermostable agar substitute) plates. However, this crossregulation of the two appendages does not appear to extend to other Sulfolobales (Rowland et al. 2020).

The archaella regulation network in S. acidocaldarius has proven to be more complex than was previously thought, and unravelling the regulatory network in S. acidocaldarius might not provide general guidelines for understanding archaella regulation in other Crenarchaeotes, let alone other archaea: of the 26 crenarchaeal genomes available in 2013, 11 of them were shown to code for the archaellum. In this limited subset of genomes, the repressors ArnA and ArnB and the activators ArnR, ArnR1, and Abfr1 were not universally found. Moreover, ArnC was found only in the order Sulfolobales, and ArnD seems to be more specific still, since it was only identified in S. acidocaldarius (Hoffmann et al. 2016).

Mc. maripaludis and EarA

In contrast to the series of transcription factors that influence regulation of transcription of arl genes in crenarchaeotes, only a single transcription factor has been identified in euryarchaeotes. Deletions in most arl genes or in many agl genes leads to non-archaellated cells in Methanococcus species even though initially the arl gene cluster is still transcribed. However, upon continued transfer in the laboratory, such mutants cease to produce detectable ArlB2, indicating that transcription of the arl genes has ceased. It was reasoned that a second mutation in a transcriptional activator for the arl gene cluster had occurred in these cells that presumably results in an energetic saving under conditions where archaella is not synthesised. A whole genome comparison between two versions of an aglB mutant, one making ArlB2 and one not, led to the identification of a single nucleotide deletion in the gene encoding a conserved hypothetical archaeal protein (MMP1718) with a winged HTH DNA-binding domain that belongs to a family of transcriptional activators (Ding et al. 2016b). When a deletion of mmp1718 was done in wild-type cells, transcription of the arl gene cluster was negligible by qRT-PCR and the cells were non-archaellated. ArlB2 expression and archaellation was complemented when mmp1718 was provided in trans and MMP1718 was renamed EarA, euryarchaeal archaellum regulator. Electrophoretic mobility shift assays and isothermal titration calorimetry results demonstrated that EarA binds specifically to short consensus sequences immediately upstream of the arl promoter.

Later, a mutant of the earA deletion strain in which transcription and translation of the arl gene cluster, archaellation and motility were all restored was isolated, despite the continued absence of EarA (Ding et al. 2017b). Analysis of the arl gene cluster promoter of this spontaneous mutant revealed a deletion of three adenines within a string of seven adenines in the BRE portion of the promoter. When the 3 adenine deletion in BRE was reconstructed in a stock culture of the earA mutant, very similar phenotypes to that of the spontaneous mutant were found. The deletion of the three adenines resulted in the BRE now more closely resembling the BRE sequence in promoters known to have high basal transcription levels, suggesting EarA may help to recruit transcription factor B to the weak arl gene cluster BRE.

Homologues of earA are widespread in euryarchaeotes with the exception of extreme halophiles (Ding et al. 2017a) and the gene for EarA is always found adjacent to a gene encoding a predicted methyltransferase whose role in archaellation, if any, is unknown. In many archaea, earA homologues are found immediately upstream or downstream of arl gene clusters and/or che genes or between arl and che gene clusters but this is not so in Mc. maripaludis.

H. marismortui

Initial studies identified two archaellins in H. marismortui, arlB found next to other arl genes on chromosome I and arlA2 located on the pNG100 plasmid (Syutkin, Pyatibratov and Fedorov 2014a; Syutkin et al. 2012). Archaella of wildtype cells are composed mainly of ArlA2 archaellin with a minor amount of ArlB, while archaella of a ∆pNG100 strain has archaella that can only be composed of ArlB. Both archaellins can form functional archaella alone. Syutkin et al. (Syutkin et al. 2019) demonstrated the salt-dependent regulation of the two archaellins in H. marismortui. Transcription and translation of arlA2 in wild-type cells were optimal in conditions of low (20%) salinity and reduced at higher salinity (30%). This was very different to what was observed with arlB in the ArlB strain mutant where the plasmid encoding arlA2 is lost. Here, transcription of arlB and ArlB synthesis is similar for all tested salinity levels. While ArlB archaellins could be detected in all 3 salinities tested (20%, 25% and 30%), archaella in the ArlB strain were most abundant at the highest salinity, suggesting that while archaellins are present intracellularly in all cases, they are only assembled into archaella at certain salinities, indicating regulation at the level of archaellin secretion, distinct from transcriptional or translational regulation. Interestingly, Faguy et al (Faguy, Koval and Jarrell 1993) reported a similar observation in Msp. hungatei grown at different temperatures.

Hfx. volcanii

Although EarA homologues have not been found in extreme halophiles, a gene encoding a protein containing a HTH domain with similar length to EarA was identified in the Hfx. volcanii genome located between archaellum and chemotaxis genes. Homologues of this protein were identified in other halophiles, suggesting a potential role for this protein in transcriptional regulation of the archaellum in this branch of Euryarchaeota (Ding et al. 2017a). In addition, a transposon screen identified hvo_0246 (a putative ArsR family regulator protein-encoding gene) as possibly being involved in the regulation of archaella expression, as two independent disruptions of this gene abolished motility (Legerme et al. 2016).

A novel post-translational regulation of archaella-driven motility by pilins in Hfx. volcanii was reported by Esquivel and Pohlschroder (Esquivel and Pohlschroder 2014). Hfx. volcanii has a set of 6 adhesion pilins, all distinct, but with an absolutely conserved hydrophobic domain. Mutants lacking the 6 pilin genes exhibit a severe motility defect with the majority of the cells lacking archaella, phenotypes which can be rescued by the expression of any one of the 6 pilins supplied in trans. Deletion of genes that are involved in the biosynthesis of pili, but which are not pilin genes, do not affect motility indicating that it is the presence of pilins, and not assembled pili, that affected the motility regulation. Further work showed the regulatory effect of pilins involved only the conserved hydrophobic domain. The authors present a model that hypothesizes that the hydrophobic region of pilins sequesters an unknown protein that directly or indirectly inhibits archaellar motility. During planktonic growth, when cells make archaella, the cells also express pilins which are only slowly incorporated into pili. The unincorporated pilins are free to sequester the motility inhibitor, meaning that archaella are readily made. Upon adhesion to a surface, pilins are more rapidly incorporated into pili leaving less free pilins and thus releasing the previously sequestered inhibitor of motility which now interferes with archaella biosynthesis or stability. Such a model could lead to a rapid response to changing environmental conditions and allow a rapid switch from motile planktonic cells to sessile biofilm cells.

Studies on the signal peptide processing of archaellins continues

The interest in various aspects of archaeal TFPPs continued throughout the last decade. ArlK and PibD are members of the GXGD protease family that also includes type IV prepilin peptidases, signal peptide peptidases and presenilin, a protein that is mutated in Alzheimer disease (Hu et al. 2011). The first member of the GXGD family to be crystallized was ArlK from Mc. maripaludis which was solved at 3.6 Å resolution (Hu et al. 2011). The crystal structure indicated that the enzyme must undergo a conformational change to bring the two essential aspartic acid residues (D18 and D79), located in TM helices α1 and α4 (in the conserved GXGD motif), into close proximity for catalysis. Crosslinking experiments that prevented movement of D18 towards D79 led to an inactive enzyme.

Henche et al. (Henche et al. 2014) studied the amino acids key to catalysis in PibD of S. acidocaldarius. They created site directed mutant forms of PibD at the catalytic aspartic acid residues, D21 and D78, which both led to an inactive enzyme, as well as in the other amino acids in the GxGD motif (actually GGAD in S. acidocaldarius) i.e. G75, G76 and A77. A G75E PibD variant was inactive while a G76E mutation had no effect on catalysis. If A77 was changed to a nonpolar amino acid, the resulting enzyme was still active while A75E was inactive. From this mutant data, as well as in silico analysis of other class III signal peptidases, they proposed redefining the class III peptidases/TFPP/prearchaellin peptidases as GxHyD rather than GxGD enzymes.

More recent studies (Kuo et al. 2015) showed that ArlK actually has a dual role, acting as a pre-archaellin peptidase but also as an ion channel, and thus an example of a chanzyme. Purified ArlK reconstituted into a lipid bilayer conducted Na⁺ with a slope conductance of 85 pS. In a D79N mutant where one essential aspartic acid residue is mutated, the catalytic activity is lost but the protein still acted as a Na⁺ channel and also now was Ca²⁺ permeable. Thus, a single mutation affected both catalytic activity and the ion selectivity of ArlK, linking the catalytic and ion transport activities of this protein. By screening a small molecule library, Coburger et al. (Coburger et al. 2016) indentified the first inhibitor of ArlK, which could also prove to be potentially useful in studies of presenilin.

Nair et al. (Nair and Jarrell 2015) showed there is a difference in the order of the two post-translational modifications, signal peptide removal and N-glycosylation, in archaellins and pilins in Mc. maripaludis even though both are type IV pilin-like proteins. Mc. maripaludis has a separate enzyme to remove signal peptides from each of the two substrates: ArlK processes archaellins and EppA processes pilins. Both pilins and archaellins are modified by an AglB-dependent N-glycosylation. Using mutants deleted for arlK or aglB or both genes, it was shown that archaellins can be processed by ArlK even if the archaellins are not glycosylated and N-glycosylation can still occur on archaellins that still have their signal peptide. Which of these two events happens first in the cell is not known and perhaps they occur simultaneously. Using mutants in aglB and eppA, it was shown that EppA can remove signal peptides from non-glycosylated pilins, but pilins, unlike archaellins, are not glycosylated until they have had their signal peptides cleaved.

Most recently, Jarrell (Jarrell 2020) investigated the importance of the conserved +3 glycine position of archaellins in the cleavage reaction carried out by FlaK. Analyses of known archaellin sequences available in Genbank showed that the +3 position (i.e. three amino acids after the cleavage site to remove the signal peptide), was virtually invariant and when the +3 glycine of the Mc. voltae archaellin was changed to valine, it was found to not be processed by FlaK (Thomas, Chao and Jarrell 2001a). For Mc. maripaludisFlaB2, mutation of the +3 glycine to any of 10 amino acid residues still allowed for archaellin processing by ArlK in in vitro cleavage assays. Thus, the reason for the strict conservation of glycine at this position remains unresolved.

The type IV filament (TFF) superfamily evolutionary history

Denise et al (Denise, Abby and Rocha 2019) examined the evolutionary history of the TFF superfamily using phylogenetic analysis of all available complete bacterial and archaeal genomes. This comprehensive analysis extends recent work which focused on archaella structures only (Desmond, Brochier-Armanet and Gribaldo 2007; Makarova, Koonin and Albers 2016). This superfamily includes structures in both Bacteria and Archaea which diversified throughout evolution to become involved in such varied functions as swimming, twitching, DNA uptake, protein secretion and adhesion. The superfamily includes archaella, Type II secretion systems, bacterial and archaeal type IV pili, the mannose-sensitive hemagglutinin pilus, Tad pili, and competence pili. Of these, the archaellum is the only one that rotates the filament for propulsion. The core components of these diverse systems include an assembly ATPase (ArlI for archaella), an integral cytoplasmic platform protein (ArlJ), filament proteins with a class III signal peptide (archaellins) and a prepilin peptidase (ArlK) that cleaves the filament protein signal peptide, although the peptidase can be co-opted from sister systems within the cell. These core proteins are usually complemented by a few additional proteins (for example a secretin to allow passage of the structure through the peptidoglycan layer in bacteria) needed for the function of the respective machine. The researchers found that of the common components, the ATPase, integral membrane protein and major filament protein were all phylogenetic markers for the evolution of the superfamily. Identifying TFFs in searches of all available complete bacterial and archaeal genomes, the phylogeny of the systems indicates they were likely present in the last universal common ancestor. This ancestral system possessed the basic architecture of current systems and would assemble filaments proteins on top of the platform protein embedded in the cytoplasmic membrane using the assembly ATPase. From this ancestor, two lineages, a bacterial and an archaeal one, then diversified to provide the plethora of functions accredited to the superfamily. This diversification involved multiple gene duplications, gene fissions and deletions, horizontal gene transfer, genetic rearrangements and addition of new components to a relatively small number of core homologous components.

Archaella as virus attachment sites

A plethora of unusual viruses, including many new virus families, have been identified and studied in archaea (Krupovic et al. 2018)but identification of their initial cell surface receptor is comparatively rare (Porter, Russ and Dyall-Smith 2007; Rowland et al. 2020). Since many archaea are surrounded by an S-layer which forms their only wall component, the possibilities for cell surface candidates as the initial attachment site for viruses is limited as common bacteriophage receptors like cell wall proteins and lipopolysaccharide are less common or simply not found in archaea. A metaproteomics study of haloarchaea in Deep Lake Antarctica linked variations in S-layer proteins, archaellins and other surface proteins to evasion of virus infection (Tschitschko et al. 2015). Two H. lacusprofundi strains differing in the number of archaellin genes were identified: the ACAM34 strain with only one archaellin gene (arlB2), and strain R1S1 with two tandem archaellin genes (arlB1 and arlB2). As was speculated for the archaellin ecoparalogs in H. marismortui (Pyatibratov et al. 2008), if H. lacusprofundi R1S1 could switch between archaellins to make archaella, it could reduce infections by viruses that bind to specific structural features of archaella (Tschitschko et al. 2018).

Biotechnology applications of archaella

There are numerous reports on the use of the biological supramolecular structures for nanotechnology, mainly pili, flagella and viruses. These structures are used as a scaffold for binding particular ligands, for example by insertion of specific peptides—known to bind the ligand of choice—within the monomer that forms the supramolecular structure. While flagella have been used, a case has been made for the advantages of archaella since they are often more stable to extremes and are often composed of multiple archaellins which can then be separately tagged with epitopes that bind different ligands, making for polyfunctional material (Beznosov, Pyatibratov and Fedorov 2018). In addition, some archaella are known already to be adhesive which may prove beneficial. Since N-linked glycans are believed to be located on the exposed surface of archaellins, initial studies inserted epitopes like the the FLAG epitope between two closely located N-glycosylation sites on H. salinarum archaellins and confirmed the epitope was surface exposed since it was detected by anti-FLAG antibodies (Beznosov, Pyatibratov and Fedorov 2009). The FLAG modified archaella contained 5 negatively charged Asp residues and bound positively charged metal ions. Modified archaella of H. salinarum mineralized by cobalt oxide and applied to a conductive nickel grid support have been tested as a material for the negative electrode of a lithium-ion battery. The reversible capacity of such nanostructured samples exceeded 400 mAh/g and their stability in cycling increased significantly when archaella were fragmented by sonication prior to mineralization (Beznosov et al. 2011). Later, iron oxide was used with the FLAG tagged archaella to obtain nanostructured iron oxide composites for use as a lithium ion battery anode and obtained increased electrical capacity (Beznosov et al. 2015). Given the solved spatial structures of archaella from different archaea, it should be possible to now more accurately predict probable sites for inserting peptides that would be surface exposed. Peptides that bind different ligands can also be inserted, such as LKAHLPPSRLPS for binding colloidal gold (Beznosov et al. 2013). The work in this area is nicely summarized in a recent review by that group (Beznosov, Pyatibratov and Fedorov 2018).

Walker et al. (Walker et al. 2019) showed that the archaella of Msp. hungatei are electrically conductive. Until that report, electrically conducive protein filaments (e-pili) were only known for certain bacteria, most extensively studied in Geobacter species (Lovley and Holmes 2020). One advantage of Msp. hungatei archaella is that the structure was previously reported (Poweleit et al. 2016) but a major disadvantage is the lack of a genetic system for manipulating these organism. Using the published structure as a guide, Walker et al. (Walker et al. 2019) identified a core of tightly packed phenylalanines that may be the route of the electrical conductance. These studies may have uses in the design of so-called green electronic materials.

Deciphering roles for the Arl proteins

Until the 2010s roles for most the Arl proteins were undetermined, save for the archaellins and the pre-archaellin peptidase ArlK/PibD. However, in the last decade, led mostly by efforts from the Albers laboratory, roles for each of the Arl accessory proteins have been shown and enormous headway has been made in the elucidation of the entire archaellum structure (Table 4).

Table 4.

Functions of Arl Proteins in Archaellation.

Arl Protein	Role or potential role	Interaction partners
Archaellin	Major filament proteins
ArlCDE	Putative motor switch	CheF, Polar cap (?)
		ArlCDE localization depends on ArlH
ArlX	Hypothesised motor complex scaffold	ArlH and ArlI
	Forms rings in vitro with variable diameter
	Only in Crenarchaeota, replaces ArlCDE
ArlF	Potential stator	S-layer proteins and ArlG
ArlG	Potential stator, forms filaments that likely extend across the pseudo-periplasm	ArlF
ArlH	ATP-binding protein hypothesized to modulate the activity of ArlI and provide the switch from assembly to rotation mode	ArlI, ArlX (in Crenarchaeota), ArlCDE (in Euryarchaeota)
ArlI	ATPase, energises assembly and rotation of archaella filament.	ArlH and, in Crenarchaeota, ArlX
	ATP-dependent hexamerization	Hypothesized interaction with ArlJ via the N-termini
ArlJ	Transmembrane protein, likely platform protein for assembly of export apparatus	Unknown, assumed to interact with ArlI in the export complex
ArlK/PibD	Pre-archaellin peptidase, removes signal peptides from archaellins

ArlI, The dual-function ATPase

ArlI, a homologue of the assembly ATPase of type IV pili systems, has been shown to have ATPase activity in S. acidocaldarius and S. solfataricus (Lassak et al. 2012b; Albers and Driessen 2005) and is the only Arl protein known to possess such activity. It was thus believed early on to be the protein involved in assembly of archaellins into the filament. Archaellum rotation is also dependent on ATP hydrolysis and it was shown that assembly of the archaellum and its rotation were both driven by ArlI (Reindl et al. 2013). Interestingly, the two activities could be separated as the first 29 amino acids of ArlI are required for motility but not assembly. The same study determined the structure of ArlI resolved to a resolution of 2.0 Å in the presence of Mg²⁺ and ATP, corresponding to the first atomic structure of an archaellum-associated protein.

The monomer of ArlI has a bi-lobed structure, with a flexible and variable N-domain (NTD), and a C-terminal region (CTD) with a conserved ATPase domain containing Walker A and Walker B motifs (Reindl et al. 2013). ArlI forms a hexamer in the presence of ATP, with the ATP molecules binding at the interface of neighbouring protomers (Ghosh et al. 2011; Reindl et al. 2013). The hexamer has a crown-like shape, with the rigid CTDs forming the ring of the crown and the flexible NTDs forming its points (Fig. 7).

Figure 7.

The structure of four of the archaellum motor complex proteins have already been solved, in the case of ArlH for both a crenarchaeote and a euryarchaeote. Top:S. acidocaldarius ArlH (PDB: 4YDS; Chaudhury et al. 2015) is predicted to form a hexameric complex, but experimental evidence for this prediction is lacking. Walker A and B motifs are represented, respectively, in pink and in blue. Mg²⁺ is represented as a yellow sphere and the bound ATP is represented as sticks. Middle: the hexameric ArlI (PDB: 4IHQ; Reindl et al. 2013) from S. acidocaldarius is shown to the right, with one of the monomers shown to the left with ADP bound. The ATPase-domain containing C-terminal is represented in light orange. The ADP molecule is bound to the monomer in a pocket with the Walker A and B motifs (represented in pink and blue, respectively). A Mg²⁺ cation is also present in the ATP-binding pocket. The variable N-terminus is represented in orange, with the exception of a triple helix bundle shown in cyan and the first 29 residues shown in purple. The triple helix bundle localizes ArlI to the membrane and the first 29 residues are essential for rotation, but not assembly, of the archaellum. Bottom: the heterocomplex of ArlF/G (PDB: 5TUG; Tsai et al. 2020) and the ArlF (PDB: 4P94; Banerjee et al. 2015) and ArlG (PDB: 5TUH; Tsai et al. 2020) monomers, also derived from S. acidocaldarius. The complex is a heterotetramer of two ArlF and two ArlG promoters. The residues Tyr68 and Ile86 are essential for dimer formation and also for motility, suggesting that the former is essential for the latter. See the Current Models for the proposed relative location of each of the proteins in the motor complex.

The ADP-bound ArlI has two open active sites and four closed, buried active sites (Reindl et al. 2013). ATP-binding assays with the fluorescent analogue MANT-ATP in ArlI from P. furiosus have shown that saturation is achieved when the concentration of MANT-ATP reaches 1/3 of the concentration of PfArlI, suggesting that this is a conserved feature of ArlI proteins (Chaudhury, van der Does and Albers 2018). Based on structural analyses, it has been proposed that during cycles of ATP binding, hydrolysis and release, the hexameric ArlI changes conformation resulting in a movement that is suitable for both the insertion of archaellins into the growing filament and rotation of the assembled filament (Reindl et al. 2013).

Although the mechanism of archaellar filament rotation remains unknown, it has been hypothesised that ArlI interacts with ArlJ (Ghosh et al. 2011). Despite the current gap in knowledge regarding the structure of ArlJ and its homologues, bioinformatical predictions suggest the presence of positively charged cytosolic loops in ArlJ (Ghosh et al. 2011), which possibly interact with the mostly negatively charged N-termini of ArlI, and therefore ArlJ may relay the ATP-hydrolysis-induced conformational movements of ArlI to the cell exterior. Biophysical studies of archaella rotation in H. salinarum (Kinosita et al. 2016; Iwata et al. 2019; Nishizaka, Masaike and Nakane 2019) have also uncovered the torque observed for the H. salinarum archaellum cannot be explained by the hydrolysis of 6 ATP molecules per rotation, but only by more than 6 hydrolysed ATP molecules per revolution. A possible explanation proposed for this apparent discrepancy is that there is a symmetry mismatch between ArlI and ArlJ. It was further suggested that the archaellum motor rotation achieves an efficiency of 100% if an ArlI hexamer is interacting with an ArlJ dimer (Iwata et al. 2019). Determination of the structure of ArlJ and the archaellum motor complex will determine if this is the case.

ArlJ, likely assembly platform protein

ArlJ was recognised early as one of the few archaellum proteins with bacterial homologues, specifically PilC in type IV pilus systems and GspF in type two secretion systems (Peabody et al. 2003), leading to speculation that ArlJ might be the platform protein where the archaellum motor assembles (Szabo et al. 2007b). While deletion analysis has confirmed the essential role for ArlJ in archaellation in euryarchaeotes and crenarchaeotes (Szabo et al. 2007b; Lassak et al. 2012b; Thomas et al. 2002) biochemical and structural studies on ArlJ have lagged behind studies on other Arl proteins, probably because this is the only integral membrane protein encoded in the arl gene cluster. While ArlJ has been suggested to form dimers, biochemical evidence for the oligomeric state of ArlJ in the archaellum motor is still lacking. The hypothesis that PilC rotates, thus allowing for pilin insertion (Chang et al. 2016), is an interesting observation as it may provide the link to how a rotating motor (the archaellum) was repurposed from a non-rotating molecular machine.

ArlH as a potential modulator of the ATPase ArlI

Initially, the only Arl proteins thought to have bacterial homologues were ArlI and ArlJ (Ng, Chaban and Jarrell 2006), while ArlH was thought to be specific for archaea; more recent evidence, gathered in particular after the structure determination of ArlH from S. acidocaldarius (Chaudhury et al. 2015) and Mcc. jannaschii (Meshcheryakov and Wolf 2016), has revealed that the closest structural homologue of ArlH is the bacterial protein KaiC which belongs to the RecA superfamily of ATPases (Pattanayek et al. 2004; Chaudhury et al. 2015; Meshcheryakov and Wolf 2016; Johnson et al. 2017). Proteins of the KaiC family are widespread in archaea, where they may be important hubs of regulatory networks in other cellular processes besides motility, although experimental evidence for this hypothesis is missing (Makarova, Galperin and Koonin 2017). ArlH has a motif similar to that of ATPases which include Walker A and Walker B boxes although the Walker B motif is incomplete—possibly explaining why ArlH does not exhibit ATPase activity (Ghosh et al. 2011; Chaudhury et al. 2015; Meshcheryakov and Wolf 2016; Seraphim and Houry 2020).

The instability of ArlX in the absence of ArlH in S. acidocaldarius initially suggested that these two proteins interacted with each other (Lassak et al. 2012b; Banerjee et al. 2012), a hypothesis later confirmed when a truncated version of ArlX (ArlXc; see below) was found by microscale thermophoresis (MST) to interact with ArlH (Lassak et al. 2012b; Banerjee et al. 2013). The same study confirmed the predicted interaction between ArlI and ArlH. Cryo-electron microscopy studies of ArlH and ArlX later showed that ArlH interacts with ArlXc by localising inside rings formed by the latter (Chaudhury et al. 2015).

The structure of S. acidocaldarius ArlH was solved by X-ray diffraction to a resolution of 2.3 Å in the presence of Mg²⁺ and ATP (Chaudhury et al. 2015). ArlH was found to have a RecA-fold consisting on a large, twisted central β-sheet surrounded by α-helices (Fig. 7). The ATP binding pocket is located in an exposed and conserved surface, suggestive of an oligomerisation interface between ArlH protomers as observed for other RecA-family proteins and, in particular, KaiC (Pattanayek et al. 2004; Chaudhury et al. 2015). Nucleotide-binding by ArlH was found to be essential for archaella assembly: in vivo studies in S. acidocaldarius showed that mutations in the Walker A or Walker B motifs (both rendering ArlH unable to bind ATP) result in archaellation loss. It was further observed in in vitro studies with the more stable P. furiosus proteins that ArlH mutant proteins that do not bind ATP also do not interact with ArlI. It was therefore proposed that the interaction between ArlI and ArlH is strictly required for archaellum assembly, although the outcome of such interaction remains mysterious. Perhaps the switch between ArlI energizing assembly of archaellins into the archaellum and its role in rotating an assembled archaellum is regulated by ArlH, with the nature of the switch possibly being a time-dependent autophosphorylation mechanism given the homology of ArlI and KaiC (Daum and Gold 2018). Meshcheryakov and Wolf (Meshcheryakov and Wolf 2016) solved the structure of ArlH from the euryarchaeote Mcc. jannaschii. The structures of ArlH from S. acidocaldarius and Mcc. jannaschii share a similar architecture, which is expected given the centrality of this protein in the archaellum motor. Paradoxically, however, surface plasmon resonance studies between ArlI and ArlH from Mcc. jannaschii suggest that higher concentrations of ATP reduce the affinity between ArlI and ArlH (Meshcheryakov and Wolf 2016).

Since ATPase assays have consistently failed to show ATPase activity in ArlH, the current hypothesis is that ArlH has a modulatory role, perhaps towards ArlI. In vitro studies of P. furiosus ArlH and ArlI showing that the ATPase activity of ArlI increases by about 80% in the presence of ArlH are consistent with such a role (Chaudhury, van der Does and Albers 2018). Sub-tomogram averaging of the motor complex of P. furiosus determined by cryo-EM suggested that ArlH is localised there (Daum et al. 2017), consistent with fluorescence microscopy studies in Hfx. volcanii, which also suggested that ArlH is localized to the archaellum motor (Li et al. 2019).

ArlF and ArlG, archaellum stator

Both ArlG and ArlF, two more Arl proteins shown to be essential for archaellation (Lassak et al. 2012b; Chaban et al. 2007) have archaellin domains, but they lack the class III signal peptide that is present in bona fide archaellins. To date, the only known structures of ArlG and ArlF homologues are truncated variants of S. acidocaldarius proteins (Banerjee et al. 2015; Tsai et al. 2020).

The truncated version of ArlF crystallised as a dimer, and size exclusion chromatography and small-angle X-ray scattering (SAXS) data show that ArlF also exists as a dimer in solution. A point mutation in one of a series of hydrophobic amino acids identified in the putative dimer interface, I86K, resulted in monomeric ArlF. The dimerization of ArlF is required for motility, as a variant of arlF carrying this mutation cannot complement a ΔarlF strain (Banerjee et al. 2015). The 1.65 Å crystal structure shows that the archaellin domain of ArlF has a β-sandwich fold with 8 anti-parallel β-strands in two sheets, and a β-barrel fold which imparts great stability to the protein (Fig. 7). This last feature is particularly relevant given that ArlF localizes to the periplasm, as inferred from the interaction between ArlF and S-layer proteins of S. acidocaldarius, and therefore it is subjected to the harsh environmental conditions where this thermoacidophile lives.

The structure of ArlG was determined by X-ray diffraction 5 years after ArlF (Tsai et al. 2020). ArlG was also purified as a soluble, truncated version containing only the archaellin domain (sArlG). The structure of sArlG was solved to a resolution of 1.93 Å (Fig. 7). The structures of sArlG and sArlF are rather similar, and reconstitution studies at pH 3 led to heterocomplex formation whose structure was solved to a resolution of 2.47 Å. Structural data and SAXS data indicated that the most likely oligomeric form of the complex is a heterotetramer (Fig. 7). Mutation studies guided by structural knowledge of the ArlG/ArlF tetramer showed that when the interaction between ArlF and ArlG is disrupted by the mutation ArlF^I86K or ArlG^Y68K, archaella cannot be assembled. In these studies, either ArlG or ArlF were overexpressed in S. acidocaldarius, and curiously it was found that ArlF overexpression results in archaellation loss. The similarities of ArlF and ArlG with true archaellins suggested that ArlG and/or ArlF could form filaments. This was shown to be the case for the soluble domain of ArlG from P. furiosus (PfsArlG), but the soluble domain of ArlF from P. furiosus (PfsArlF) did not show such a tendency. More interesting still, negatively stained micrographs showed that while PfsArlG alone forms filaments, in the presence of PfsArlF knob-like structures are formed instead (Tsai et al. 2020).

Direct evidence for the periplasmic localisation of these proteins was achieved by ectopically expressing an HA (hemagglutinin)-tagged variant of either ArlF or ArlG in S. acidocaldarius, followed by immunoblot analysis of the membrane fraction and the soluble fraction (cytosol and pseudoperiplasm) using anti-HA antibodies (Tsai et al. 2020). These experiments resulted in two conclusions: (i) ArlF and ArlG are found in both the soluble and membrane fractions; and (ii) there are two protein variants of ArlG and two variants of ArlF. In the case of ArlG, there is a full-length variant which localises to the cell membrane, and a cleaved variant which can be found in the soluble fractions. Cleavage of ArlG is PibD-independent, and no archaella were detected in strains where cleavage of ArlG cannot take place. ArlF, on the other hand, localised mostly to the membrane, and had a higher and a lower molecular weight variant. The higher molecular weight variant was found to be due to glycosylation at N92.

The current model posits that ArlG and ArlF are transported to the membrane, where they remain until they are required for archaella assembly. When ArlG is processed, it interacts with ArlF in a heterotetrameric form and then by forming a filament structure it pushes ArlF out of the membrane until it interacts with the S-layer which stops self assembly of the ArlG filament. When this structure is completed, ArlG and ArlF provide structural stability to the motor complex by anchoring the archaellum motor to the cell surface. Besides, ArlG is in a position where it also functions as a stator: ArlG proteins are inserted in the membrane, where they provide a platform for the assembly of the remaining ArlG filament and they can interact with ArlJ via their transmembrane domains. As a potential rotating element of the motor complex, ArlJ can then rotate against the transmembrane domains of ArlG. Recent observations in S. islandicus are consistent with this interpretation. Here, the S-layer exists in a stalk and cap structure with SlaA forming the cap and SlaB forming the stalk. Zhang et al. (Zhang et al. 2019) created strains deleted for slaA, slaB and remarkably in both S-layer genes. Even in the double mutant where there was no S-layer present, the cells had archaella, but the mutants were nonmotile suggesting anchoring archaella to the S-layer is required for motility (Zhang et al. 2019). If the role of ArlF and ArlG is to stabilize archaella by connecting to the S-layer (akin to Mot/Pom proteins attaching flagella to peptidoglycan), then in the S-layer deficient mutant, archaella assemble but the ArlF/G complex would not be able to connect the archaella with the S-layer in order for torque to occur.

ArlX, essential but unknown function

ArlX is an unusual Arl protein, found only in Crenarchaeotes, that is predicted to have regions that are distantly homologous to domains of methyl-accepting proteins (Ghosh and Albers 2011). It is located within the arl locus of S. acidocaldarius and shown to be essential for the archaellum assembly (Lassak et al. 2012b).

ArlX is a monotopic membrane protein, and is predicted to have a transmembrane α-helix in the N-terminus, with the soluble domain rich in helices and containing a triple coil (Banerjee et al. 2012). The largest construct of the protein that was successfully purified was ArlXc, corresponding to the soluble domain of ArlX (i.e. deletion of the first 37 amino acid residues). Purification of this construct by size exclusion chromatography showed that the oligomeric species formed by ArlXc is >600 kDa, and analysis by electron microscopy showed that this structure is ring-shaped. The observed rings had varying diameter estimated to be between 26 and 38 nm. Possibly the variance in observed diameters is an artefact from in vitro experiments, and the diameter of ArlX might be stabilised by the presence of any other of the proteins of the motor complex, particularly ArlI and ArlH which interact with ArlXc (Banerjee et al. 2013).

Prior studies had shown that ArlB, ArlH, ArlI and ArlJ all seemed essential for the stabilisation of ArlX in vivo (Lassak, Ghosh and Albers 2012a). Pull-down experiments demonstrated that ArlXc interacted via its C-terminus with ArlI (Banerjee et al. 2012), and since ArlI is a cytosolic protein, this experiment provided evidence for the first time that the soluble domain of ArlX was located in the cytoplasm. ArlI was later shown to interact with the C-terminus of ArlX by both its C- and N-termini (Banerjee et al. 2013).

The interaction between ArlX and another protein of the motor complex was visualised for the first time in 2015 (Chaudhury et al. 2015). Cryo-electron microscopy measurements showed that monomers of ArlH localised inside the ring formed by ArlX. Due to the diversity of the ring diameter exhibited by ArlXc, it was not possible to conclude how many monomers of ArlH interact with ArlX in vivo. Nevertheless, class sorting suggests that an ArlXc ring with a 20-fold symmetry harbours 9–10 ArlH monomers (Chaudhury et al. 2015).

ArlCDE

The genes arlC, arlD, and arlE are divergent genes thought to be replaced by arlX in Crenarchaeota or by an unknown gene in Thaumarchaeota (Desmond, Brochier-Armanet and Gribaldo 2007; Makarova, Koonin and Albers 2016). These genes are thought to be an adaptation to the bacterial-like chemotaxis machinery in some archaea (Albers and Jarrell 2015b; Albers and Jarrell 2018). The observation that ArlCDE can exist either as individual proteins or as fusion proteins in different permutations in different archaea led to the hypothesis that the gene products interact to form a complex in the archaellum motor (Ng, Chaban and Jarrell 2006; Makarova, Koonin and Albers 2016).

Early studies had shown that ArlC and ArlD localized to the cell membrane in Mc. voltae but were not detected in purified archaella (Thomas and Jarrell 2001) and that ArlCE and ArlD in H. salinarum interacted directly with proteins of the CheF family, which represent archaeal-specific connectors between the chemotaxis machinery and the archaellum motor (Quax, Albers and Pfeiffer 2018a; Schlesner et al. 2009). This suggested that ArlCE and ArlD are not only components of the archaellum motor, but that they are possibly involved in motor switching during chemotaxis. Recent fluorescence microscopy localisation experiments in Hfx. volcanii associated ArlD to the archaellum motor, by showing that ArlD localised to the cell poles where the archaellum is also located (Li et al. 2019). ArlCE was shown to co-localise at the cell poles with ArlD, consistent with an interaction between ArlCDE as had been originally proposed. It was further observed that the localisation of ArlCE and ArlD to the membrane is interdependent, as neither protein localises to the cell pole without the other, leading to the suggestion that ArlCE and ArlD form a pre-complex before localising to the archaellum motor complex. Interestingly, the polar localisation of ArlCDE was shown to be reduced in the absence of ArlH, for the first time indicating that ArlCDE interact with other components of the motor complex (Li et al. 2020). The lack of atomic structures of ArlCDE makes it difficult to ascribe densities observed in the cryo-electron microscopy subtomogram averaging that have been obtained for P. furiosus and T. kodakarensis to either of these proteins. Nevertheless, a cytosolic density seen in the archaellum motor complex of P. furiosus has been suggested to correspond to ArlCDE. This density is located below a density which fits well with the known structures of ArlI and ArlH. Strangely, however, such density has not been observed for T. kodakarensis, although it is likely that the ultrastructure of the archaellum motor is similar for both organisms (Briegel et al. 2017; Daum et al. 2017).

Atomic models of archaella filaments

The first low-resolution 3D reconstructions of archaella filaments date from the beginning of the 2000s and were all led by efforts of the Trachtenberg laboratory (Cohen-Krausz and Trachtenberg 2002; Trachtenberg, Galkin and Egelman 2005; Trachtenberg and Cohen-Krausz 2006; Cohen-Krausz and Trachtenberg 2008). These seminal studies were performed with transmission and/or scanning electron microscopy and it was not possible to discern secondary structures of archaellins. However, the advent of cryo-electron microscopy has made it possible to achieve near-atomic resolution of archaella.

Archaellar filaments of Msp. hungatei, measured to be, on average, 10 µm in length and 10 nm in diameter, were reconstructed to an averaged resolution of 3.4 Å by Poweleit et al. (Poweleit et al. 2016). The axial rise and the rotation per subunit were calculated to be, respectively, 5.32 Å and 108.02°. These measurements led to an estimate of ∼61 500 archaellin monomers per filament. Although Msp. hungatei has three archaellin genes, only ArlB3 was detected in the cryo-EM structure and in MS analysis of the major band (36 kDa smear) observed upon SDS-PAGE analysis of purified archaella. However, that analysis also showed two lower mass archaellin bands of approximately 25 kDa as found earlier (Cruden, Sparling and Markovetz 1989; Southam et al. 1990). Typical of archaellins, ArlB3 is a glycoprotein but surprisingly O-glycosylation seems to be the prevalent PTM in Msp. hungatei ArlB3. Ser59, Ser61, Ser65, Thr100, Ser104 and Thr108 were predicted to be O-glycosylated, while N-glycosylation was predicted to occur at Asn159 (although at an atypical N—glycosylation site), while Gln134 was also predicted to be modified, although it was not possible to define the chemical nature of the modification. The cryo-EM map and Edman degradation showed that the mature archaellin lacks its first 6 residues, which constitute the signal peptide of ArlB3.

The ArlB3 monomer has a tadpole-shape, with a long, hydrophobic α-helical N-terminal domain (NTD) and a C-terminal domain (CTD) that forms an eight-stranded anti-parallel β-barrel (Fig. 8). In the filament, the NTD forms the core of the filament, while the CTD constitutes the outside of the filament. In the filament, each archaellin interacts with eight other monomers. This extended intermolecular network is thought to underlie the stability of the Msp. hungatei filament. The core of the filament is not hollow, confirming previous findings on other archaella.

Figure 8.

The Msp. hungatei archaellum filament (PDB: 5TFY; Poweleit et al. 2016). A. Representation of 10 ArlB3 chains composing a filament. Each chain is coloured from N-terminus to C-terminus (blue to red). The highly hydrophobic N-terminus of the archaellins are located in the interior of the filament, while the β-sheets of the variable C-terminal domain decorate the external side of the filament. Note the absence of an internal channel. To the right, an inset showing a single archaellin monomer, showing the long, N-terminal α-helix and the globular CTD composed of eight anti-parallel β-sheets. B. A complex network of intermolecular interactions between the archaellins are responsible for the stability of the filament. Shown is a string of phenylalanine residues that runs parallel to the axis of the filament which are partly responsible for the stability of the filament, and possibly for the recently reported electrical conductivity of this archaellum. This feature is not conserved in the filaments of P. furiosus or Mc. maripaludis, whose core consists instead of the side chain of hydrophobic residues, and therefore they are not likely to conduct electrons. Otherwise, the three filaments have a similar architecture.

After the characterisation of the Msp. hungatei filament, cryo-EM was used to calculate not only the density map of the archaellum filament but also the archaellum motor complex of P. furiosus, to a resolution of 4.2 Å (Daum et al. 2017). The diameter and helical parameters of the P. furiosus filament are similar to those measured for Msp. hungatei. The ultrastructure of the archaellar filament from P. furiosus follows a similar architecture to that described for Msp. hungatei, namely a core formed by bundled α-helices, and a highly hydrophilic exterior formed by β-sheets. Again, as with Msp. hungatei, only one of the three archaellins of P. furiousus, ArlB0 with its 5 amino acid signal peptide removed, appears to be present in the filament. The remaining archaellins might possibly form filaments on their own or be part of the proximal and/or distal region of the filament mainly composed by ArlB0. The NTD of ArlB0 consists of a conserved α-helix, and the CTD of β-sheets that form a twisted β-barrel. Each archaellin interacts by hydrophobic interactions via its NTD with 6 neighbour archaellins; in addition, the more hydrophilic C-terminus of the α-helix may establish interactions by hydrogen bonds or other electrostatic interactions with neighboring α-helices and β-strands.

The amphipathic character of the archaellins was also suggested to be the driving force for the spontaneous assembly of archaellins into filaments. A striking difference between the archaellum of P. furiosus and that of Msp. hungatei is that the former seems to be homogenously N-glycosylated throughout its surface. The glycosylation sites were inferred from extra-densities in the surface of the filament that cannot be attributed to amino acid side chains; moreover, these densities could be mapped to asparagines located in N-glycosylation sequons: N67, N90, N98, N122 and N128. Contrary to the archaellins of Msp. hungatei, no O-glycosylation was detected on the archaellins of P. furiosus.

Two years after the studies on the filament from P. furiosus, an archaellin was for the first time seen with atomic resolution (Meshcheryakov et al. 2019). A truncated variant of the archaellin ArlB1 from Mcc. jannaschii (MjArlB1_Δ1–38), only with its globular domain, was crystallised and its structure solved to a resolution of 1.5 Å. As seen in the archaellins from Msp. hungatei and P. furiosus, the MjArlB1 also contains a CTD rich in β-sheets that organise in a β-barrel. However, in this structure it was also possible to identify a β-hairpin that binds a metal ion. The authors determined the divalent cation to be Ca²⁺, although they did not exclude the possibility that in its native environment, the place of Ca²⁺ may be taken by some other metal. This structure was then used to refine a cryo-EM map of the archaellar filament of Mc. maripaludis to a resolution of 4 Å. The sequences of each of the three archaellins of Mc. maripaludis were tested, but only the sequence of ArlB1 fit to the cryo-EM map, similar to what was observed for P. furiosus and Msp. hungatei. All three archaellins of Mc. maripaludis had previously be shown to be N-glycosylated at multiple positions (Kelly et al. 2009). The helical parameters are very similar to those in Msp. hungatei and P. furiosus. The core of the filament was shown to be composed of α-helical bundles, with the exterior decorated by the β-barrels of the archaellins’ CTDs. The archaellins of Mc. maripaludis also bind a metal ion but, instead of Ca²⁺, Mg²⁺ seems the most likely candidate. This metal ion does not seem to be involved in intermolecular interactions; instead, it seems to be required to stabilise the conformation of the globular domain, which in part favours the polimerized state of the filament, since in the presence of chelating agents, the filament depolimerizes. The mature protein is missing its 12 amino acid signal peptide.

Overall, the deduced structures of the three archaella filaments show a common theme: the archaellins are tadpole-shaped and they form a helical filament whose core is made of a bundle of α-helices, with the twisted β-barrels of the archaellins pointing to the outside of the filament (Fig. 8). This helical filament has an axial rise of around 5 Å with a rotation of approximately 108°. When looked along the axis of the filament, archaella have a 10-fold symmetry. There is no significant inner lumen in the filament. All filaments are composed of glycosylated archaellins, but some may be richer in N-glycosylation and others in O-glycosylation.

Assembly model

The first assembly model for the archaellum was proposed in 1996, when close to nothing was known about which proteins were needed to form the archaellum, apart from the archaellins. Nevertheless, at that time some archaellins had already been cloned and the evidence that archaellins were synthesised as pre-proteins processed by a type IV prepilin-like signal peptidase suggested an assembly model similar to the bacterial T4P (Jarrell, Bayley and Kostyukova 1996a) (Fig. 9).

Figure 9.

The evolution of models for archaellum structure and assembly over the last 25 years. 1990s The first model was proposed in 1996 based mainly on work in Mc. voltae. Then, only the archaellins, synthesized as preproteins, had been unequivocally shown to be required for archaellation. Due to the similarities of archaellins and type IV pilins, the archaellum was suggested to be assembled in a manner similar to T4P. Minor archaellins were proposed to have a role either in the termination or initiation of filament formation, or both. 2000s By the end of the 2000s,  a more complex model for archaellum assembly in Methanococcus was proposed. The structure and some of the components in the assembly of the N-linked glycan attached to archaellins were identified and this information was incorporated into assembly models. The main filament proteins were ArlB1 and B2, with ArlB3 forming the cell proximal hook-like region and ArlA found in low abundance throughout the filament. The pre-archaellin peptidase was identified as ArlK. The arl accessory genes were known, but not their function. ArlHIJ were proposed to form an export complex, and ArlF and ArlG were suggested to anchor the archaellum to the underlying polar cap. Current Models In current models of the archaellum, somewhat different versions are required for Crenarchaeotes and Euryarchaeotes, owing to the presence of ArlX only in the former and the presence of a polar cap and ArlCDE only in the latter. The major advances in the current models compared to earlier ones lie in the elucidation of the roles and locations of most of the Arl accessory proteins.

By the end of the 2000s, when the existence of the arl gene cluster was established, a new revised model was proposed (Thomas, Bardy and Jarrell 2001c; Ng, Chaban and Jarrell 2006; Ng et al. 2008). However, there was still little knowledge regarding the function of the accessory proteins. ArlI and ArlJ were recognised as homologues to proteins found also in T4P, and it was postulated that they powered the assembly and served as a platform for the assembly/motor complex, respectively. A set of che genes in several archaea had already been identified at this time, but it remained unknown how a bacterial-like chemotaxis system interacted with the archaellum motor. By the end of the 2000s, the similarities between archaella and T4P, and the differences between archaella and flagella were already well established.

Today, the model for archaella assembly—which is still based in part on some speculation—is considerably different from and more complete than its predecessor (Fig. 9). Several major factors have contributed to this: the elucidation of the role of most of the arl accessory proteins and the knowledge gained from archaellum atomic structures and the partial structures of motor complexes obtained by cryo-EM for two Euryarchaeotes: P. furiosus and T. kodakarensis (Briegel et al. 2017; Daum et al. 2017). In fact, there are two variants of the model: one corresponding to Crenarchaeotes and other to Euryarchaeotes. This is necessitated by the observation of a polar cap or DLS only in Euryarchaeotes and the presence of arlX in Crenarchaeotes instead of arlCDE found in Euryarchaeotes.

Integrating the knowledge recently obtained on the Arl accessory proteins, a broad assembly model that applies to both phyla can be proposed. A complex formed by several accessory proteins self-assembles into the export complex (which is at a later stage also the motor complex). Within this complex, ArlJ is the platform protein inserted within the membrane. Interacting with the positively charged cytosolic loops of ArlJ is ArlI, on the cytoplasmic side of the cell membrane. ATP hydrolysis by the hexameric ArlI results in conformational changes that allow for the insertion of archaellins. ArlH interacts with ArlI, possibly modulating its activity (e.g. by increasing the ATP hydrolysis rate). ArlG and ArlF—hypothesised to be part of the stator—are theoretically not required for the assembly of the filament; nevertheless, in the absence of the either gene cells are non-archaellated. Perhaps these two proteins need to be the first archaellin-like proteins to be exported, possibly defining the channel within which the archaellins will assemble into the filament. It is possible that ArlF is first exported, immediately followed by ArlG, which forms a filament that grows from the base. When ArlF interacts with the S-layer, the growth of the ArlG filament stops, allowing for the assembly of the archaellins into the archaellum filament. At this point it is necessary to establish the difference between the crenarchaeal and euryarchaeal motor complex. In the Crenarchaeal motor complex, a ring formed by ArlX forms in the membrane, surrounding the motor complex. It remains an open question whether the ArlX ring forms first, thus indicating where the remaining proteins of the motor complex should assemble, or whether it forms after the other accessory proteins are already establishing the secretion/motor complex. In Euryarchaeotes, ArlCDE form a cytosolic ring around ArlI and ArlH. Possibly ArlCDE interact with the polar cap, effectively anchoring the archaellum in a way that is required to allow for its rotation at a later point. Furthermore, the archaeal-specific CheF protein interacts with ArlCDE, therefore coupling the chemotaxis system to the archaellum motor complex.

Archaellin synthesis and export is predicted to be similar in all archaellated archaea: pre-archaellins are synthesised in the cytosol and localised to the membrane, where they are glycosylated and processed by the PibD/ArlK signal peptidase, prior to being inserted at the base of the growing filament. The order of the two PTMs is not known and each can occur in the absence of the other.

Of particular relevance for the development of a motor complex model is the observation of such motors in situ. Briegel et al. (Briegel et al. 2017) observed the motor complex of T. kodakarensis KOD1 focusing on the polar cap, which was suggested—as it had been suggested previously—to anchor ‘the archaellar basal body in part to provide leverage for rotation’. The polar cap, whose protein composition is unknown, has a conical shape with a blunt end. The diameter of the cone base ranges between 220 nm and 525 nm, and the narrow end of the polar cap is close to, but not touching, the cell membrane. This has also been proposed as a polar organising center, concentrating archaella in a single bundle, and/or concentrating signalling molecules. Evidence for the latter possibility is that polar cones are associated with chemosensory arrays.

The cryo-tomogram also allowed the observation of a partial motor complex. A cytosolic ring was observed right underneath the cell membrane, which is hypothesised to be ArlI by comparison with the PilB/T of T4P. Two further disks located below ArlI, in the cytosol, are hypothesised to be ArlH and ArlCDE, but this conclusion was reached based on those proteins being located in the cytoplasm. It was not possible to observe any density that could be ascribed to the membrane protein ArlJ or the pseudo-periplasmic ArlF/G.

In the same year, cryo-electron tomography was also used to observe the motor complex of P. furiosus archaella (Daum et al. 2017) which also co-localised with a polar cap (Fig. 3). The polar cap of P. furiosus is located 35 nm away from the basal complex of the archaellum, a bit closer that what has been measured for T. kodakarensis (44 nm). The diameter of the base of the polar cap was measured to be within the range of 200–600 nm. The polar cap of P. furiosus was also found to be associated with a hexameric protein array, leading the authors to suggest that the polar cap functions not only as an anchoring site for the archaellum, but also to recruit proteins necessary for motility. While the identity of the hexagonal protein array could not be determined, it was suggested that it could correspond to an unknown P. furiosus-specific type of chemoreceptor.

As with T. kodakarensis, clear densities could only be observed for the cytosolic portions of the motor complex, preventing once more the identification of ArlJ, ArlF and ArlG. A cytosolic ring density could be fitted well with the hexamer of S. acidocaldarius ArlI; in this case, the structure fits best when the N-terminus of ArlI point towards the membrane, which is in accordance with the model suggesting that a triple-helix-bundle of ArlI interacts with ArlJ. Lying below ArlI, the authors could fit the structure of a hypothetical ArlH hexamer, suggesting that ArlH is present in the mature complex. A further ring-shaped density was observed around the densities ascribed to ArlI and ArlH and, with the same rationale used for the T. kodakarensis motor complex, these densities are hypothesised to be ArlC and ArlDE.

CONCLUDING REMARKS

The research summarized in this review illustrates what a remarkable motility structure the archaellum is. Employing a model adapted from type IV pili systems and utilizing a small fraction of the number of proteins needed for a functional flagellum, archaea have nevertheless assembled a motility structure that is efficient, extremely stable to external conditions and capable of multiple functions other than swimming. The archaellum system is inducible and can be connected to a chemotaxis pathway to aid in cells migrating to optimal environmental niches. The archaellum can be co-regulated with type IV pili synthesis and biofilm formation to optimize resource expenditure and survival in natural habitats. As highlighted by the incredible swimming speeds that can be reached by various hyperthermophiles, it is apparent that the archaellum is a very effective motility structure.

Much has been discovered since the first electron micrographs of archaella. The novelty of the archaellum as a domain-specific swimming organelle related to type IV pili, with structural components synthesized initially as pre-proteins cleaved by TFPP and typically modified with N-linked glycans is well established. The presence of easily identified homologues to type IV pilus systems, mixed with archaeal-specific components, results in a novel structure for swimming. We now know that archaella are widespread in phylogenetically diverse archaea from a wide variety of often extreme habitats, which proteins are required for archaellation, the function of most of those Arl proteins, and some of the regulatory events that lead to archaellation or its repression in a limited number of species.

However, major areas of research remain to be explored. Genetic systems are only available for a limited number of model organisms and what we know about archaella in a few archaea is not likely to be true of all archaea, as witnessed by the different regulatory schemes of various Sulfolobales and Methanococci. Different transcriptional and post-transcriptional controls have been hinted at in several species and continued discovery of transcriptional activators and repressors for the arl genes will no doubt be forthcoming in different species. New twists and adaptations can be expected as archaella are studied in additional species such as members of the Nanoarchaeota, Thaumarchaeota and other proposed newly discovered phyla. Virtually nothing is known about the order of assembly of the various archaellum components and advances in genetic systems and technology may be needed to decipher this pathway. Little is known of the composition and possible function of the polar cap/DLS in archaella anchoring, how it is connected to the archaellum and why is seems limited to only euryarchaeotes. We can expect advances in the use of archaella as scaffolds for biotechnology that have only been reported in limited potential applications thus far. Similarly, it is likely that additional biophysical, biochemical and structural studies will elucidate interactions and stoichiometries of the motor components and the generation of torque. We can also anticipate more studies connecting the chemosensory system and archaellum rotation and additional and even more detailed atomic models of the entire apparatus including motor and anchoring sites. While we can reflect on the significant knowledge accumulated to date on various aspects of the archaellum, the enhanced interest in all aspects of archaellum research in the last decade bodes well for a more robust understanding of this unique apparatus.

ACKNOWLEDGEMENTS

KFJ would like to thank the continuous financial support of the Natural Sciences and Engineering Research Council of Canada for this research during the length of his career. He would also like to acknowledge the significant contributions, generosity and friendships of former graduate students, collaborators and sabbatical leave hosts who made his research career so enjoyable. JNdSM and S-VA were supported by the Collaborative Research Centre SFB1381 funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 403222702 – SFB 1381. JNdSM was supported in part by the Excellence Initiative of the German Research Foundation (GSC-4, Spemann Graduate School) and in part by the Ministry for Science, Research and Arts of the State of Baden-Wuerttemberg. The article processing charge was funded by the Baden-Württemberg Ministry of Science, Research and Art and the University of Freiburg in the funding programme Open Access Publishing.

Conflicts of interest

None declared.