Tat-dependent bundling pilus of a halophilic archaeon assembles by a strand donation mechanism and facilitates biofilm formation

Significance

Biofilm formation, where many individual cells come together in a dense network, is widely regarded as the default growth mode for many microorganisms and plays an important role in bacterial pathogenesis. Archaea also form biofilms, but the molecular details underlying this process remains understudied. Here, we describe a class of bundling filaments that are important for biofilm formation in a class of archaea living in water nearly saturated with salt.

Abstract

Diverse extracellular filaments present on the surface of archaea mediate multiple key processes, such as motility, adhesion, and biofilm formation. Although several archaeal filament types have been characterized in considerable detail, many remain understudied, particularly those utilizing noncanonical secretion systems. Here, we describe the Tafi bundling pilus that facilitates biofilm formation in the haloarchaeon Natrinema sp. J7-2. Unlike previously characterized archaeal pili, Tafi is secreted via the twin-arginine translocation (Tat) pathway, which transports fully folded proteins across the cytoplasmic membrane. Structural analysis reveals that although Tafi pili assemble via a canonical strand-donation mechanism, the pilin subunit (TafE) adopts a distinct structural topology that sets it apart from the previously characterized Sec-dependent pilins that form bundling pili in archaea. Sequence analyses show that TafE homologs are also present in thermophilic archaea from different phyla, but Tat-signal sequences are exclusive to pilins of halophilic archaea. Nevertheless, we find that Tat signal peptides in haloarchaeal TafE-like pili were exchanged back to the Sec signal peptides on multiple independent occasions. These findings expand our understanding of the diversity and evolution of archaeal extracellular filaments and highlight the Tat pathway as a route for pilus assembly in halophilic archaea.

Extracellular filaments in archaea and bacteria perform important functions and are particularly crucial for survival under extreme environmental conditions (1–4). Archaea, which are well known for inhabiting extreme ecosystems (5–7), exhibit an array of functionally diverse extracellular surface filaments (1). Archaeal flagella, adhesive type IV pili, bundling pili, conjugation pili, cannulae, and cytochrome nanowires have all been characterized at near-atomic resolution (8–14). Several other filament types—such as Ups pili, Mth60 fimbriae, Hami, and Tafi—have been identified, but their structures remain to be characterized (14–20). The biogenesis of these filaments involves distinct secretion and assembly mechanisms. Several filament types, including flagella and adhesive type IV pili, possess a conserved type IV prepilin-like signal sequence, suggesting their assembly proceeds via a type IV pilus-like pathway. In contrast, other filament systems—such as bundling pili, cytochrome nanowires, Mth60 fimbriae, cannulae, and Hami—have a Sec-dependent signal sequence and are secreted through the Sec translocation pathway.

Biofilm formation is widely regarded as the default growth mode for many microorganisms (21), conferring numerous advantages such as localized access to nutrient-rich environments and enhanced resistance to abiotic stresses, viral predation, and antimicrobial agents (22–25). The development of a biofilm involves cellular aggregation mediated by extracellular filaments and is further facilitated by the secretion of extracellular substances, such as exopolysaccharides, DNA and protein/glycoprotein polymers, which enhance cell–cell and cell–surface adhesion (26–28). While many proteinaceous filaments contribute secondarily to biofilm formation, the bundling pili are notable for their importance in the biofilm formation (8, 9, 29, 30). A distinctive feature of bundling pili is their assembly via a strand-donation mechanism, wherein one subunit donates a β-strand to complete a β-sheet in an adjacent subunit, resulting in a unique sinusoidal filament (8, 9). In contrast to conventional type IV pili or flagella, which typically exist as discrete filaments extending from the cell surface, bundling pili assemble into larger, multistranded bundles that display mechanical stability (8, 9, 29). These bundles are implicated in cell–cell and cell–surface attachment, and biofilm development, and processes critical for survival in extreme habitats.

In archaea, the bundling pili have been predominantly observed in thermophiles and hyperthermophiles of the orders Thermoproteales and Thermofilales, both in the phylum Thermoproteota (formerly known as Crenarchaeota) (8). However, no homologs of pilin proteins forming the bundling pili were identified in other archaea. Nevertheless, unknown extracellular filaments were proposed to contribute to adhesion and biofilm formation in haloarchaea (31, 32). Recently, a distinct type of extracellular filament denoted as Tat-dependent fimbriae (Tafi) has been identified in haloarchaeon Haloarcula hispanica (20). Tafi pilus differs from the traditional bundling pili in its secretion mechanism (20). Specifically, Tafi pilins are exported via the twin-arginine translocation (Tat) pathway, which is characterized by its ability to transport fully folded proteins across the cytoplasmic membrane (33–35). This contrasts with the Sec-dependent pathway, which translocates proteins in an unfolded state (34, 36). Mutation of the diagnostic arginine residues (RR) in the Tat signal peptide to lysines abolished Tafi formation, confirming the role of the Tat translocation system in the pilus biogenesis (20). However, the structural details of Tafi organization remained unknown.

Here, we describe that an extremely halophilic archaeon Natrinema sp. J7-2 also produces Tafi pili that tend to bundle, connecting multiple cells. We describe the structural and sequence analyses of the Natrinema Tafi pilus, which has a canonical Tat-dependent signal sequence. Tafi pilin displays a distinct structural topology compared to previously described Sec-dependent archaeal bundling pilins, such as AbpA of Pyrobaculum calidifontis (8) and the thread subunit Saci_0406 of Sulfolobus acidocaldarius (9). We show that homologs of Tafi pilins are also present in diverse thermophilic and hyperthermophilic archaea, where they seem to be secreted via the Sec pathway rather than the Tat system. Finally, we provide insights into the bundling of the Tafi filaments and show that they play an important role in biofilm formation.

Results

Thin Extracellular Filaments of Natrinema Cells Tend to Bundle.

We observed in electron micrographs that negatively stained Natrinema sp. J7-2 (37) cells are interconnected by thin filaments (Fig. 1A), which were named Tafi (see below) (20). Flexible morphology (Fig. 1B) and a narrow width (~4.5 nm) ruled out the possibility that Tafi are flagella or type IV pili, both common surface appendages of archaeal cells. Indeed, type IV pili of Natrinema sp. J7-2 have a diameter of ~10 nm (38, 39). The filaments were sheared off mechanically and analyzed using cryoelectron microscopy (cryo-EM). We observed that the detached filaments formed bundles of varying thickness (Fig. 1 C and D), with the power spectra showing a consistent meridional peak at ~1/(33 Å), indicating a repetitive structural feature at this frequency (Fig. 1E). Two types of extracellular filaments, archaeal bundling pili [ABP (8) and threads (9)] and extracellular cytochrome-containing nanowires (ECNs) (11), are known to have a similarly large axial rise (~33 Å). Thus, in principle, Tafi could be related to either archaeal bundling pili or ECNs. To identify the Tafi pilin and to assess its relationship with other archaeal filaments, we determined the high-resolution structure of Natrinema Tafi using cryo-EM.

Fig. 1.

Extracellular filaments of the Natrinema sp. J7-2. (A and B) Electron micrographs of Natrinema sp. J7-2 cells with extracellular filaments after negative staining by uranyl acetate at different magnification. Filaments forming the bundles are indicated by yellow arrowheads. (C and D) Cryo-EM micrograph of the filaments sheared off from the Natrinema sp. J7-2 cells in 18% salt buffer. Bundles of the filaments are indicated by yellow arrowheads. (E) Average power spectrum generated from the vertically aligned cryo-EM images of the filament bundles showing consistent meridional peak at 1/(~33 Å).

Structure of the Natrinema Tafi Pilus.

Tafi filaments in the isolation buffer containing high salt (~18%) tend to form bundles of varying thickness (Fig. 1 C and D) which prevent high-resolution structural studies. We therefore tried to find conditions which would allow for solving the structure of individual filaments. The bundles could be partly disassembled into constituent individual filaments by 4× diluting the sample in water (Fig. 2A). In contrast to the ABP and threads, both of which feature a prominent sinusoidal appearance (8, 9), the Tafi displayed less of a sinusoidal morphology in cryo-EM micrographs and 2D-class averages (Fig. 2 A and B). We obtained a 4.2 Å resolution cryo-EM map of the Tafi filament by helical reconstruction. The map features the β-strand-rich structure employing a strand-donation assembly mechanism (Fig. 2C) used by bundling pili of hyperthermophilic archaea (8).

Fig. 2.

Cryo-EM of Natrinema sp. J7-2 Tafi-E pilus. (A) Cryo-EM micrograph, (B) 2D-class average, (C) high-resolution map and atomic model of Tafi-E pilus. (D) Structure of TafE pilin (Uniprot id: I7CYY9). (E) Strand-donation between two subunits of Tafi-E pilus with one subunit shown with its surface lipophilicity. Polar and hydrophobic surfaces are represented by green and gold color respectively. (F) Surface charge representation on the structure of TafE pilin. (G) Density likely corresponding to the posttranslational modification of the side chains of the respective residues in TafE pilin.

Although the bundles observed on the surface of Natrinema sp. J7-2 indeed resembled those previously described for ABP and threads, no homologs of the corresponding pilins could be found outside of the phylum Thermoproteota (8, 9). Given that the AbpA protein which forms ABP filaments is structurally related to bacterial TasA pilin, we surmised that a distant homolog of the Tafi pilin may be detectable by structure-based searches. Thus, we searched for structural homologs of AbpA (8) using FoldSeek (40) against the AlphaFold50 database (41). Although no structural homolog of AbpA were found in Natrinema, related proteins were identified in other haloarchaea. We then used the structure of one of these haloarchaeal proteins, namely, from Halobacterium bonnevillei (A0A6B0SEE0), to search the AlphaFold50 database for more divergent homologs in Natrinema species. This search yielded a significant hit to Natrinema sp. J7-2 protein NJ7G_0677 (Uniprot# I7CEQ2; 219 residues) with FoldSeek probability of 0.97, which forms a cluster with NJ7G_2828 (Uniprot# I7CYY9; 215 residues). The two proteins were examined for their fit within the cryo-EM density map, with NJ7G_2828 protein showing a better fit compared to NJ7G_0677. Specifically, Ser66 of NJ7G_2828 fits well in the density, unlike Arg54 of NJ7G_0677, Phe93 of NJ7G_2828 fits better than Val81 of NJ7G_0677, and Val211 of NJ7G_2828 matched better than Tyr213 of NJ7G_0677. Thus, we conclude that NJ7G_2828 is the major pilin of the bundling filaments of Natrinema sp. J7-2 and hereinafter refer to it as TafE (see below for the explanation of naming).

TafE is a β-strand-rich protein with nine β-strands (B, B’, C, D, E, F, G, H, and I) adopting the β-sandwich fold (Fig. 2D). We compared its structure with the available structures of archaeal pilins that assemble into extracellular filaments via a strand-donation mechanism, namely, AbpA, Saci_0406, Hyper2, CanA, and CanX, all from hyperthermophilic archaea (8, 9, 14). DALI structural comparisons showed that TafE is structurally similar to these proteins, with Z-scores ranging from 4.2 to 7.4. The greatest similarity was to AbpA, whereas the lowest was to CanA (Fig. 3A). The structural similarity heatmap, generated by a DALI all-to-all structural comparison, placed TafE at the interface of the two major clusters, archaeal bundling pili (ABP and threads) on one side and cannulae filaments (Hyper2, CanA, and CanX) on the other side (Fig. 3A). Indeed, TafE is an outlier of both of these two clusters in a structural correspondence plot derived from the eigenvectors of the DALI analysis (Fig. 3B). Together, the heatmap and correspondence plot suggest that the TafE structure substantially differs from those of other archaeal extracellular filaments assembling via a strand donation mechanism.

Fig. 3.

Structural comparison of Natrinema Tafi-E pilin with other archaeal extracellular filaments that follow the strand donation mechanism of assembly. (A) Structural similarity heatmap generated using the Z-score calculated by DALI. (B) Structural similarity correspondence plot derived from the eigenvectors of the DALI analysis for all pili. (C) The 2D depiction of the structural topology of all proteins analyzed in (A) and (C). AbpA, Archaeal bundling pilin; Saci, Sulfolobus acidocaldarius thread filament protein; CanA, CanX, and Hyper2, in vitro/in vivo formed archaeal cannulae.

Topology comparisons suggest that although archaeal bundling pilins such as AbpA, Saci_0406, Hyper2, CanA, and CanX share a typical jelly-roll fold composed of two β-sheets with four β-strands each (BIDG and CHEF, as shown in Fig. 3C), TafE shows a variation, with one β-sheet formed by three strands (BID) and the other by five strands (CHGEF) (Figs. 2D and 3C). Additionally, a long loop between strands G and H forms a hairpin-like structure that inserts between the two β-sheets (Fig. 3C). The B′ strand, together with its preceding loop, aligns next to the donor β-strand (B) from a neighboring subunit (Fig. 3C). Strand B from the neighboring subunit sits in the groove between strands D and B’, completing the β-sheet (BID) (Fig. 2 D and E). The residues on strand B are distributed in such a way that one face is hydrophobic and the other polar. The hydrophobic face interacts with the hydrophobic groove within the globular domain of the neighboring subunit, whereas the polar surface faces the solvent, thereby establishing the strand-donation mechanism (Fig. 2E). Together, the two β-sheets, BID (with strand B contributed by a neighboring subunit) and CHGEF, along with the G-H loop and the B′ strand with its preceding loop, form the β-sandwich in the TafE structure (Fig. 3C).

The TafE subunit exhibits a width ranging from ~30 to ~45 Å, depending on the orientation. The helical polymerization of TafE subunits with an axial rise of ~33 Å and twist of −80 ° gives rise to a Tafi filament with the maximum width of ~45 Å. The surface of the pilus is characterized by bumps and grooves, and it has less of a sinusoidal appearance than ABP. As is typical for proteins from haloarchaea (42), TafE is an acidic protein with an overall negative surface charge, resulting in a highly negative surface on the Tafi filament (Fig. 2F). Additional densities next to the sidechains of Asp55, Gln60, Thr120, Thr123, Asn142, Thr145, Asn179, and Glu181 are observed on the filament surface suggesting possible glycosylation or other posttranslational modifications of these residues (Fig. 2G).

Natrinema Pilins Are Secreted by the Tat Pathway to Assemble the Tafi Pilus.

The TafE sequence analysis using SignalP (43) tool revealed a signal peptide associated with the Tat pathway, featuring the Tat motif “RRNVL” starting at position 5 (Fig. 4A). Typically, filaments that assemble via strand-donation in bacteria and archaea possess the Sec-signal peptide for secretion through the Sec pathway (8, 9, 44–47). Recently, however, a new type of haloarchaeal filaments, dubbed Tat-fimbriae (Tafi), was described in H. hispanica (20). A distinguishing feature of Tafi is that their major pilin, TafA, has a Tat signal sequence and the filament was predicted to assemble via a strand-donation mechanism (20). Structural information to validate this prediction for the Tafi filaments of H. hispanica is still lacking.

Fig. 4.

Natrinema tafE gene is part of Tat pathway gene cluster. (A) Sequence of TafE pilin (Uniprot id: I7CYY9) with predicted Tat-motif site highlighted in yellow. (B) Genetic neighborhood of TafE encoding gene (tafE) in Natrinema sp. J7-2 genome and its comparison with that of H. hispanica. Predicted structures of the proteins encoded by respective neighboring genes are shown below in pink in comparison with the Taf proteins of H. hispanica (green) indicating structural similarities between the two.

In H. hispanica, the TafA encoding gene (tafA) is located near six other taf genes (tafB to tafG), which encode proteins proposed to play various roles in the Tafi biogenesis (20). We examined the neighboring genes of tafE in Natrinema sp. J7-2 (genomic coordinates 2,443,480-2,453,057) and compared the predicted structures of the corresponding proteins with the Taf proteins from H. hispanica (Fig. 4B). The comparative analysis revealed that TafB (putative transcription regulator), TafD (signal peptidase), TafE (minor pilin), and TafG (DUF5305 family protein, predicted to assist TafD) of H. hispanica have close structural counterparts in Natrinema sp. J7-2 (Fig. 4B). Nevertheless, there are notable differences between the two Tafi loci. In particular, pairwise comparison with DALI showed that Natrinema TafE (major pilin) exhibits greater structural similarity to Haloarcula TafE, which has been proposed as a minor pilin (Z-score 9.4, RMSD 2.8 Å, alignment length of 126), as compared to the major pilin TafA (Z-score 4.4, RMSD 3.5 Å, alignment length 105; SI Appendix, Fig. S1). Notably, Natrinema does not encode an ortholog of Haloarcula TafA. Therefore, we refer to the major pilin as TafE, following the previously proposed nomenclature (20). Notably, no close homologs of the Haloarcula TafC and TafF, the proposed pilus tip and membrane anchor proteins, respectively, were identified in the Natrinema Tafi locus. However, analysis of the structural models of proteins encoded within the Natrinema Tafi locus suggested functional counterparts for both proteins. In particular, NJ7G_2827 (Uniprot# I7BYJ3) appears to function as the tip protein in the Tafi pili of Natrinema. This protein contains two domains, one that matches a domain of Haloarcula TafC, and another, likely corresponding to the outermost tip of the pilus, that appears to be unique. Two proteins NJ7G_2829 and NJ7G_2831 in Natrinema contain a single predicted membrane-spanning α-helix, followed by a jelly-roll-like domain structurally similar to that of H. hispanica TafF. Thus, we propose that NJ7G_2829 (TafF1) and NJ7G_2831 (TafF2) may serve as the structural equivalent of TafF in the pilus assembly (Fig. 4B). However, it is of note that TafF of H. hispanica contains three C-terminal transmembrane α-helices as opposed to NJ7G_2829 and NJ7G_2831. Finally, the Natrinema Tafi locus contains an additional small gene, NJ7G_2826, not present in the H. hispanica Tafi locus and encoding a largely unstructured protein of unknown function (Uniprot#: I7CK27).

Bundling Mechanism.

Tafi forms remarkably straight bundles of two different morphologies: continuous striations parallel to the bundle axis (type 1) and striations tilted with respect to the bundle axis (type 2), apparent in both the micrographs and 2D averages (Fig. 5 A and B). It is possible that the two different morphologies may represent different views of the same bundle architecture. To explore the structural organization of the bundle, we analyzed the power spectra of both types (Fig. 5 A and B), which revealed a relatively high degree of order. Both spectra show a meridional reflection at ~1/33 Å, suggesting the side-by-side register of filaments. The meridional reflection arises from the rise per subunit in each component filament. If the filaments were not in register, but staggered, this reflection would be off of the meridian. Notably, the equatorial peak in the spectrum from type 1 occurs at ~1/40 Å, indicating ~40 Å spacing in projection between vertically aligned filaments (Fig. 5A). In contrast, type 2 exhibits an equatorial peak at ~1/47 Å (Fig. 5B). Asymmetric reconstruction of the bundle produced a low-resolution map that, while not resolving individual pilin subunits, clearly resolves the filament packing (Fig. 5 C and D). Filaments are arranged in a pseudohexagonal lattice as the spacing between filaments in one direction is ~48 Å and another direction is ~42 Å (Fig. 5). Such packing gives rise to the lattice lines of filaments separated by ~40 Å (Fig. 5D). We generated 2D projections (SI Appendix, Fig. S2) of the map and compared them to experimental 2D averages. Two projections clearly reproduced features observed in experimental data, including both the parallel and tilted striation patterns (Fig. 5D). Previous studies have shown that bacterial TasA pili adopt an antiparallel arrangement, whereas archaeal bundling pili can pack in both a parallel and antiparallel manner (8, 9, 29). Although our current map lacks sufficient resolution to determine filament polarity, the presence of two distinct interfilament spacings (~42 Å and ~48 Å) implies multiple interface types. Given that the purely antiparallel stacking is not possible in hexagonal packing, a mixed parallel and antiparallel packing likely underlies the Tafi bundle architecture (SI Appendix, Fig. S3). To investigate whether the observed spacings (~42 Å and ~48 Å) correspond to these differing orientations, we generated simulated models with filaments aligned in the most compact configurations for both parallel and antiparallel arrangements (SI Appendix, Fig. S3). In the antiparallel configuration, complementary interactions between filament surface features (bumps and grooves) allow for tighter packing, producing a spacing of ~42 Å. Conversely, in the parallel configuration, steric hindrance between aligned bumps results in looser packing with a spacing of ~48 Å. The coexistence of parallel (~48 Å) and antiparallel (~42 Å) packing disrupts ideal hexagonal symmetry, giving rise to a pseudohexagonal lattice in the Tafi bundle (SI Appendix, Fig. S3).

Fig. 5.

Bundling mechanism of Natrinema Tafi pili. Representative cryo-EM micrograph, 2D average, and average power spectrum of the Tafi bundle with morphology type 1 (A) featuring the continuous parallel striations and type 2 (B) displaying the tilted striations. (C) Cryo-EM map of Tafi bundle from different viewpoints. (D) Top view of the map shown in C and its 2D projections. (Scale bar, 50 nm.)

TafE Is Important for Biofilm Formation.

In bacteria, bundling TasA pili play a major role in biofilm formation and maintenance (48–50). However, the role of bundling pili in archaea remains largely unknown. Recently, it has been shown that Natrinema sp. J7-2 forms biofilms. Deletion of the highly expressed type IV pili decreased but did not eliminate the biofilm formation, suggesting the involvement of other factors (39). Thus, we assessed the role of Tafi in Natrinema sp. J7-2 biofilm formation. To this end, the effect of tafE gene deletion on the biofilm formation efficiency of closely related Natrinema sp. CJ7-F (SI Appendix, Table S2) was assessed by staining the corresponding cultures with crystal violet, followed by measurement of absorbance at OD₅₈₀. Deletion of tafE substantially reduced the amount of crystal violet staining, suggesting that Tafi play an important role in biofilm formation (Fig. 6). The contribution of Tafi to biofilm formation was similar to that of the type IV pili (39). To determine whether TafE is constitutively expressed and whether its expression is influenced by type IV pilin—which also contributes to biofilm formation—we conducted transcriptomic analysis of TafE during planktonic growth in Natrinema sp. CJ7 and the Natrinema sp. CJ7 ΔpilA1-4 mutant. TafE was found to be highly expressed in both strains, indicating that its expression is constitutive and independent of the expression of type IV pilins (SI Appendix, Fig. S4).

Fig. 6.

Tafi is important for the biofilm formation of Natrinema sp. J7. The ability of the indicated strains to form biofilm was tested using a standard biofilm assay as described in Methods. Crystal violet staining was used to detect the biofilms. (A) Representative images of the stained biofilms from three independent experiments. (B) Quantitation of biofilm, measured by absorbance at wavelength 580 nm. The Y-axis represents the absorbance (OD₅₈₀). Significance testing of WT and knockout strain was performed using a one-sample t test (**P < 0.005; *P < 0.05; NS, not significant P > 0.05).

Recurrent Switching between Tat and Sec Secretion Pathways among TafE Homologs.

In a previous study, BLASTP searches queried with the TafA sequence of H. hispanica led to the identification of 74 homologs in diverse species of halophilic archaea (20). Although using this approach no TafA homologs were detected outside of Halobacteriales (kingdom Methanobacteriati, former Euryarchaeota), complementary searches with the sequence of TafD uncovered Tafi-like loci in hyperthermophilic archaea of the orders Archaeoglobales and Thermococcales, both within kingdom Methanobacteriati (20). To further explore the distribution of Tafi systems in archaea, we used the TafE sequence of Natrinema sp. J7-2 as a query to iteratively search (3 iterations, E < 1e-5 inclusion threshold) the archaeal RefSeq protein sequence database using PSI-BLAST. The obtained dataset, filtered to include sequences of at least 170 aa and no longer than 350 aa, contained over 1,400 TafE homologs. These were further filtered to 80% identity over 80% of sequence length using mmseqs2 to remove closely related sequences, yielding the final dataset of 830 sequences. As expected, the vast majority of retrieved homologs were from halophilic archaea. Notably, not only TafE but also TafA of H. hispanica was retrieved, consistent with the close structural similarity of the two proteins (SI Appendix, Fig. S5). Besides haloarchaeal proteins, among the identified sequences were homologs from Archaeoglobales and Thermococcales, confirming the previous results obtained using the TafD as a marker protein. Unexpectedly, we also retrieved TafE homologs encoded by thermoacidophilic archaea of the order Sulfolobales (kingdom Thermoproteati, former Crenarchaeota). Although homologs were identified in all genera of Sulfolobales, namely, Acidianus, Metallosphaera, Saccharolobus, Stygiolobus, Sulfodiicoccus, Sulfolobus, Sulfuracidifex, and Sulfurisphaera, their conservation was rather patchy, with only a handful of species per genus encoding the corresponding proteins. The presence/absence varied even in the strains of the same species. Notably, no orthologs of the thread subunit Saci_0406 were retrieved during our search, suggesting that Sulfolobales archaea encode at least two distinct varieties of strand-donation pili. Furthermore, S. acidocaldarius, which produces thread filaments, lacks the gene for TafE, consistent with the experimental evidence showing that threads were the only filaments present on the cell surface of S. acidocaldarius mutants lacking the genes for flagellum, adhesion pili, and UV-inducible pili (9, 51). Unfortunately, transcriptomics data are not available for any of the Sulfolobales species that encode the TafE-like pili. AlphaFold modeling suggested that the Sulfolobales TafE homologs are competent of forming pili via the strand-donation mechanism (SI Appendix, Fig. S6). Clustering of the TafE homologs based on pairwise similarity followed by community detection using convex clustering algorithm implemented in CLANS (52) showed that haloarchaeal sequences formed nine clusters of at least 10 members, with the largest cluster including 501 and the smallest 12 sequences. Sulfolobales sequences (n = 21) formed a separate cluster, whereas Archaeoglobales and Thermococcales sequences (n = 42) clustered together, separately from haloarchaeal or Sulfolobales homologs (Fig. 7A).

Fig. 7.

Diversity of archaeal TafE homologs. (A) Clustering of the TafE homologs based on their sequence similarity. Lines connect nodes (sequences) with P-value ≤ 1e-6. The nodes are colored based on the taxonomic affiliation, with the haloarchaeal representatives colored according to their signal sequences. (B) Maximum likelihood phylogeny of TafE homologs. The tree is rooted with the clade including Sulfolobales sequences. The clade including Haloarchaeal sequences is colored depending on the signal sequence of the corresponding taxa. Branches with Sec signal sequences nested within clades dominated by sequences with Tat signal sequences are indicated with numbered circles. The scale bar represents the number of substitutions per site. The circles at the branches represent the aLRT-SH support values higher than 80%. The color legends for both panels are provided in the figure.

Analysis of the signal sequences of all TafE homologs in our dataset showed that the majority of haloarchaeal sequences (n = 698) contained the Tat signal sequence. In contrast, consistent with the previous results (20), Archaeoglobales and Thermococcales proteins had the Sec-dependent signal peptides and we found that the same was true for the TafE homologs of Sulfolobales. Unexpectedly, a considerable fraction of haloarchaeal sequences (n = 71) also contained the Sec signal peptide followed by the signal peptidase I cleavage site (Dataset S1). In the CLANS network, haloarchaeal sequences with the Tat- and Sec-dependent signal peptides were intermixed (Fig. 7A). To further explore the relationship between the TafE homologs, we calculated a maximum likelihood phylogeny of the corresponding proteins (Fig. 7B). In the tree rooted between the sequences encoded by members of the two major archaeal kingdoms formerly known as Crenarchaeota (Sulfolobales) and Euryarchaeota (the rest of sequences), homologs from Archaeoglobales and Thermococcales were at the base of the euryarchaeal clade. Consistent with the clustering analysis (Fig. 7A), sequences of Archaeoglobales and Thermococcales did not form monophyletic subclades, suggesting frequent horizontal exchange of the tafE genes, likely facilitated by the shared ecological niche. All haloarchaeal homologs formed a well-supported monophyletic clade, as a sister group to the Archaeoglobales/Thermococcales clade. Notably, sequences with the Sec-dependent signal peptide were at the base of the haloarchaeal clade. This topology suggests that Tat-dependent Tafi evolved from the pili with the Sec signal peptide.

Notably, not all haloarchaeal homologs with the Sec signal peptide occupied the basal position. We identified seven cases where sequence homologs (either single sequences or clades of sequences) with Sec signal peptides were nested among proteins with Tat signal peptides. This observation strongly suggests multiple independent reversions to the likely ancestral Sec secretion pathway. Notably, proteins from the largest clade of revertants (number 5 in Fig. 7B) were confidently predicted to form filaments through strand-donation mechanism. In contrast, single revertants 1-4, 6, and 7 (Fig. 7B) were predicted to use strand-donation mechanism for formation of nonfilamentous assemblies or not to form oligomers (SI Appendix, Fig. S6). However, the AlphaFold modeling results should not be considered as evidence of oligomerization or lack thereof, but rather as predictions, which need to be verified experimentally.

Discussion

The distinguishing feature of the Tafi pili, compared to other structurally related pili assembled by strand-donation mechanism, such as ABP and threads, is that their pilins are secreted by a different translocation pathway, the Tat-pathway. The Tat-pathway is widespread in halophilic archaea allowing the export of folded proteins as opposed to the Sec-pathway which translocates the substrate proteins in an unfolded state. In haloarchaea, the TatA and TatC family integral membrane proteins are proposed to form the Tat-pore (53). Natrinema sp. J7-2 proteome contains three Tat family proteins—NJ7G_3513 (Uniprot# I7C065, TatA family, 108 residues), NJ7G_0620 (Uniprot# I7CEK6, TatC family, 322 residues), NJ7G_0619 (Uniprot# I7CPH1, TatC family, residues 774) which are likely to make the Tat-pore for the secretion of TafE and other Tat-substrates. None of these Tat proteins share homology with catalytic usher and chaperone proteins that are found in bacteria implying that though Tafi assembles through the strand donation mechanism, its assembly is likely to be independent of chaperone and usher assistance. Recently, it has been shown using an in vitro system that two archaeal proteins (CanA and Hyper2) can self-assemble into filaments through a strand-donation mechanism without any chaperone/catalyst (14). Therefore, the chaperone/usher independent assembly of Tafi might be similar to that used by CanA and Hyper2.

Comparison of the Taf proteins between H. hispanica and Natrinema sp. J7-2 indicated that several proteins (TafB, TafD, and TafG) are well conserved between the two species of halophilic archaea. In contrast, other proteins within the corresponding loci, namely TafC, TafF1, and TafF2, exhibited greater variability, each sharing only one of their domains, indicating significant divergence. Notably, functions have been assigned to three conserved proteins—TafB, TafD, and TafG. Specifically, TafD, in association with TafG, is proposed to exhibit the signal peptidase activity, whereas TafB is likely involved in regulation of transcription (20). In contrast, the more variable proteins (TafC, TafF1, TafF2, and TafA/E) are proposed to contribute primarily to the structural assembly of the Tafi pilus. This pattern suggests that the evolutionary divergence within the taf gene clusters is largely confined to structural components, while the enzymatic elements remain under stronger evolutionary constraint.

Using the TafE sequence of Natrinema sp. J7-2 as a query, we identified a large number of homologs widespread not only in haloarchaea but also hyperthermophilic archaea from two different kingdoms, including members of the order Sulfolobales, which were previously shown to contain thread filaments that also use the strand-donation assembly mechanism, but are evolutionarily distinct from Tafi. Phylogenetic analysis strongly suggests that Tat-dependent pilins have evolved in haloarchaea from Sec-dependent relatives. Strikingly, however, our analysis suggested multiple relapses from Tat-dependent to Sec-dependent secretion of TafE-like pilins in haloarchaea. Notably, different archaeal species contain variable numbers of tafA/E-like genes in the corresponding loci. TafA and TafE of H. hispanica, albeit divergent, are phylogenetically relatively close, suggesting that they have evolved through gene duplication (Fig. 7B). In the clustering analysis, TafA formed a small cluster suggesting a relatively narrow spread of close TafA-like homologs in archaea. Thus, we hypothesize that TafE-like sequences represent the ancestral variant of the pilins. Under such a scenario, the TafA ancestor would have evolved through duplication of the major TafE-like pilin gene. Subsequently, TafA was selected as the major pilin, with TafE-like homolog assuming the role of a minor pilin.

Structural comparison of the TafE pilin with other archaeal pilins that assemble using a strand-donation mechanism revealed substantial differences from either ABP/thread-like or CanA-like filaments of hyperthermophilic archaea, suggesting that Tafi represent the third major group of archaeal pili assembled using this mechanism. Tafi pilins, while polymerizing via the canonical strand-donation mechanism, adopt a distinct structural fold that deviates from the typical jelly-roll fold. This alternative fold is a variant of the β-sandwich fold, comprising two β-sheets: one with three strands and the other with five strands (Fig. 3C). However, it is of note that TafE exhibits greater structural similarity to bundle-forming pili proteins (AbpA and Saci_0607) than to cannulae-forming proteins (CanA, CanX, and Hyper2) (Fig. 3A). This suggests that, although TafE is an outlier among these groups, it is likely to be more closely related evolutionarily to AbpA and Saci_0607 than to the cannulae-forming proteins.

Our structural analysis provided insights into the bundling of the Tafi filaments. We found that Tafi filaments have a highly negatively charged surface (Fig. 2F), which would typically lead to repulsion between filaments, preventing bundle formation. We observe bundle formation in the 18% salt solution, with the bundles disassembling upon dilution. This observation suggests that cations in the solution may play a bridging role, neutralizing the negative charges on the filaments and facilitating bundle formation in the high-salinity environments. Notably, given the fluctuations in salinity in natural habitats (e.g., due to water evaporation or rainfall), the dissolution of the bundles may help cells cope with low-salt conditions. Collectively, our results provide insights into the diversity, distribution, and evolution of filaments assembled using the strand-donation mechanism in archaea.

Methods

Strains, Culture Conditions, and Transformation.

Different strains of Natrinema sp. J7 were cultured in Halo-2 medium or 18% modified growth medium (MGM) at 45 °C. Halo-2 medium contained 250 g of NaCl, 30 g of MgCl₂·6H₂O, 2.5 g of lactalbumin hydrolysate (Difco Laboratories), and 2 g of Bacto yeast extract (Difco Laboratories) per liter of water. Casamino Acids medium (Hv-Ca) or MGM were prepared as described previously (refs). Hv-Ca contained 144 g of NaCl, 18 g of MgCl₂·6H₂O, 21 g of MgSO₄·7H₂O, 4.2 g of KCl, 5 g of Amicase (Sigma), 0.5 g of CaCl₂, and 30 mL of 1 M Tris-HCl (pH 7.5) per liter of water. 18% MGM contained 144 g of NaCl, 18 g of MgCl₂·6H₂O, 21 g of MgSO₄·7H₂O, 4.2 g of KCl, 5 g of peptone (Difco Laboratories), 3 g of Bacto yeast extract (Difco Laboratories), 0.5 g of CaCl₂, and 30 mL of 1 M Tris-HCl (pH 7.5) per liter of water. Solid and soft agar plates contained 15 g and 5 g Bacto Agar (BD) per liter, respectively. When needed, 5-fluoroorotic acid (5-FOA) was added at a final concentration of 0.04 mg/mL, while mevinolin (Mev) was added at 5 μg/mL in 18% MGM for haloarchaeal cultures. Escherichia coli were grown at 37 °C in Luria–Bertani (LB) medium supplemented with ampicillin (0.1 mg/mL) when necessary. Transformation of haloarchaea including Natrinema sp. J7 was performed using the modified polyethylene glycol (PEG) method as described previously (54, 55). All strains, plasmids, and primers used in this study are listed in SI Appendix, Tables S2–S4.

Sample Preparation.

A preculture of 5 mL was grown to stationary phase at 45 °C under agitation, then diluted into 500 mL culture and grown to exponential phase (OD₆₀₀ ≈ 0.5). Natrinema sp. J7-2 cells were collected by centrifugation (Sorval SLA1500 rotor, 8,000 rpm, 10 min, 20 °C). The cell pellet was resuspended in 20 mL of 18% SW buffer (Hv-YPC medium in the absence of yeast extract, peptone, and casamino acids), and the cell suspension was vortexed for 30 min to shear off the pili. The cells were removed by centrifugation (Sorval SLA1500 rotor, 8,000 rpm, 30 min, 20 °C). The supernatant was collected, and the pili were pelleted by ultracentrifugation (Beckman 60 Ti rotor, 35,000 rpm, 2 h, 4 °C). The supernatant was removed, and the pellet was resuspended in 500 μL of 18% SW buffer.

Negative Staining Electron Microscopy.

For negative-staining transmission electron microscopy, 10 μL of pili preparation were adsorbed onto glow-discharged copper grids with carbon-coated Formvar film and negatively stained with 2.0% (wt/vol) uranyl acetate. The samples were observed under the FEI Tecnai Spirit BioTwin 120 microscope operated at 120 kV.

Cryoelectron Microscopy.

Grid preparation.

A 3 µL of Tafi pilus sample was applied to the glow-discharged lacey carbon grid (Ted Pella, Inc.), excess sample was blotted away by Whatman filter paper #1 to leave the thin film of the sample on the lacey carbon, which was then plunge-frozen in liquid ethane using the Leica EM GP. The frozen sample on grid was stored in liquid nitrogen until cryo-EM imaging.

Screening, data collection, and image preprocessing.

Frozen grids were initially evaluated for ice thickness and pilus distribution using a 200 keV Glacios cryoelectron microscope (Thermo Fisher Scientific). High-resolution cryo-EM data were then acquired from the selected grid using a Titan Krios microscope (Thermo Fisher Scientific), operating at 300 keV and outfitted with a Gatan K3 direct electron detector (Gatan Inc.). Each movie received a total electron dose of approximately 50 e⁻/Ų and the pixel size was 1.08 Å. Raw cryo-EM data were processed using cryoSPARC, where motion correction and contrast transfer function (CTF) estimation were performed using the software’s patch motion correction and patch CTF estimation tools (56).

Helical reconstruction of Tafi filament.

Three-dimensional reconstruction was performed using helical processing workflows within cryoSPARC (46). Segments of Tafi were manually selected from multiple micrographs to produce 2D class averages, which served as templates for automated particle picking using the Template Picker tool. Bad segments were removed through repeated rounds of 2D classification. An averaged power spectrum generated from the Tafi segments was used to determine the initial helical symmetry parameters. The 3D reconstruction was performed using the Helix Refine job in cryoSPARC, applying a rise of 33.36 Å and a twist of −80.26°. The quality of the resulting map was enhanced through multiple refinement steps, including Local CTF refinement and Non-uniform refinement in cryoSPARC, followed by map sharpening using EMReady (57). Cryo-EM data collection and processing statistics are provided in SI Appendix, Table S1.

Asymmetric reconstruction of Tafi bundle.

Asymmetric construction of Tafi bundle was conducted using cryoSPARC. The Tafi bundles were manually boxed to extract the overlapping ~100,000 segments (shift ~50 Å) of bundles (3× binned, pixel size 3.24 Å/px) from micrographs. Bad segments were removed by iterative 2D classifications yielding the stack of 46,264 segments. The initial model was generated by ab-initio reconstruction, which was further refined by homogeneous refinement. The resultant map shows the separated density rods representing the stacked Tafi filaments within the bundle. A mask enclosing nine Tafi filaments was generated in ChimeraX (58), and was used for the local refinement of the map. The resultant map and mask were multiplied to generate the final map, used for the representation.

Model Building.

The TafE pilin (NJ7G_2828, UniProt ID: I7CYY9) structure predicted by AlphaFold served as the starting model for rigid-body fitting into the cryo-EM density map (59). This initial model was further refined through several iterative rounds of model adjustment using COOT (60), followed by automated real-space refinement in Phenix (61). Through this modeling process, a Tafi filament structure comprising three subunits was built, with each subunit encompassing residues 28 to 215. Detailed model statistics can be found in SI Appendix, Table S1.

Construction of tafE Knockout Strain.

Construction of the tafE deletion strain in Natrinema sp. CJ7-F was performed as described previously. Briefly, plasmid for gene deletion, pNBK-del-tafE, containing the upstream and downstream flanking sequence of tafE gene was transformed into CJ7-F. Transformants of single allelic exchange were selected on HV-ca medium. The transformants were subcultured in rich medium MGM for secondary homologous recombination. The cultures were then diluted and plated on MGM plates with 5-FOA (0.25 mg/mL) to select for recombinants. Three independent clones were picked for PCR analysis and sequencing. Clones with successful deletion of the gene were stocked.

Construction of Plasmids for Homologous Recombination-Mediated Gene Deletion.

pNBK-del- tafE used for construction of ΔtafE deletion strains carry upstream and downstream flanking sequences of tafE. Overlap extension PCR was performed using primer pairs deltafE -F1/R1 and deltafE -F2/R2 and the genomic DNA of Natrinema.sp CJ7 as a template to amplify upstream and downstream homologous arms of tafE. The PCR products were cut and ligated into HindIII-AflII digested pNBK-F by 2× MultiF Seamless Assembly Mix (ABclone, Cat# RK21020).

RNA Extraction and qPCR.

To determine the difference in tafE genes expression of Natrinema sp. CJ7 and CJ7△pilA1-4, the strains were incubated to middle exponential phase (OD₆₀₀ = 0.6-0.8) in Halo-2 medium. To isolate RNA, 1 mL of fresh cell culture was placed in an ice-water bath for 5 min, followed by centrifugation at 12,000 rpm and 4 °C for 2 min. The supernatant was discarded, and the cell pellet was resuspended in 1 mL of TRIzol reagent (Invitrogen, CA) through gentle pipetting. Subsequently, 200 μL of prechilled chloroform was added to the mixture, which was inverted vigorously for homogenization and incubated at room temperature for 2 min. The sample was then centrifuged at 12,000 rpm and 4 °C for 30 min to achieve phase separation. The upper aqueous phase containing RNA (450 μL) was transferred to a fresh tube and mixed with an equal volume of precooled isopropanol by inversion. After 10 min of incubation at room temperature, RNA precipitation was performed by centrifugation at 12,000 rpm and 4 °C for 20 min. The resultant pellet was washed twice with 1 mL of ice-cold 70% ethanol through resuspension and centrifugation (12,000 rpm, 4 °C, 10 min). Residual ethanol was removed by air-drying the open tube in a biosafety cabinet for 10 min at room temperature. The purified RNA was dissolved in RNase-free water and quantified. RNA integrity was verified by 1.5% agarose gel electrophoresis and its purity and integrity was further monitored by NanoDrop 2000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). Qualified RNA samples were reverse-transcribed to cDNA using the PrimeScript RT Reagent Kit with gDNA Eraser (TaKaRa) according to the manufacturer’s protocol for subsequent quantitative real-time PCR (qPCR) analysis.

qPCR was performed to detect the tafE gene. rpoB gene located on the host chromosome was used as a reference. The specific primers used for qPCR are listed in SI Appendix, Table S3. The reaction was set up as described previously (62), 20 μL mixtures were prepared containing 5 μL of template, 10 μL of iTaq Universal SYBR green Supermix (Bio-Rad), 1 μL of primer pairs (10 μM), and 4 μL of distilled water. Amplification was performed according to the manufacturer’s instructions. Finally, the qPCR data were analyzed by the 2-ΔCT method (63). Three independent experiments were performed, and error bars indicated the SD. Details of strains, plasmid, and primers used in the study are provided in SI Appendix, Tables S2–S4.

Biofilm Formation Assay.

Biofilm formation of CJ7 and CJ7ΔNJ7G_2828 (tafE deletion strain) was conducted using a crystal violet assay adapted from previous work (64). Overnight cultures in MGM were collected and adjusted to an OD₆₀₀ = 0.1. 2 mL of the diluted cell suspension was added to test tubes and incubated at 45 °C for 7 d. After incubation, the test tubes were fixed at 60 °C for 1 h (the duration can be appropriately extended) and then 2 mL of 0.1% crystal violet staining solution was added and left at room temperature for 30 min. The crystal violet was then discarded, and the tubes were washed twice with deionized water. 2 mL of a 1:1 mixture of 95% ethanol and 30% acetic acid was added to dissolve the stained biofilm fixed in the wells, left at room temperature for 10 to 15 min. Images of the staining were taken and absorbance at OD₅₈₀ was measured to quantify the amount of biofilm.

Sequence Based Phylogenetic Analysis.

Homolog sequences to TafE were searched in the RefSeq protein sequence database using PSI-BLAST (three iterations, E < 1e-5 inclusion threshold). Resultants 1,400 hits were filtered sequentially first by length parameter: 170aa < length < 350 aa, and later identity parameter (<80% identity over 80% of sequence length) using mmseqs2 (65), yielding the dataset of 830 sequences (Dataset S1). Pairwise similarity-based clustering of these sequences was conducted using community detection using convex clustering algorithm implemented in CLANS (52). For phylogenetic analysis, the sequences were aligned using MAFFT web server (66) with the G-INS-1 option. The positions with low information content were removed using trimal with the gap threshold of 0.2 (67). The final alignment contained 223 amino acid sites. Maximum likelihood phylogenetic tree was calculated using IQ-tree v1.6.12 (68). The substitution model best fitting the data was identified with ModelFinder (69) and was WAG+F + R10. The branch support was assessed using SH-like approximate likelihood ratio test (aLRT) with 1,000 replicates. The tree was visualized with iTOL v6 (70).

Structural Analysis and Representation.

All the structure predictions were conducted using AlphaFold server (71). DALI server (http://ekhidna2.biocenter.helsinki.fi/dali) was used for structural comparison (72). All structural figures were prepared using ChimeraX (58).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

Cryo-EM map and model of Tafi pilus are available at the Electron Microscopy Data Bank (EMD-71067) (73) and Protein Data Bank (9P0A) (74), respectively.