Evolutionary trajectory for nuclear functions of ciliary transport complex proteins

SUMMARY

Cilia and the nucleus were two defining features of the last eukaryotic common ancestor. In early eukaryotic evolution, these structures evolved through the diversification of a common membrane-coating ancestor, the protocoatomer. While in cilia, the descendants of this protein complex evolved into parts of the intraflagellar transport complexes and BBSome, the nucleus gained its selectivity by recruiting protocoatomer-like proteins to the nuclear envelope to form the selective nuclear pore complexes. Recent studies show a growing number of proteins shared between the proteomes of the respective organelles, and it is currently unknown how ciliary transport proteins could acquire nuclear functions and vice versa. The nuclear functions of ciliary proteins are still observable today and remain relevant for the understanding of the disease mechanisms behind ciliopathies. In this work, we review the evolutionary history of cilia and nucleus and their respective defining proteins and integrate current knowledge into theories for early eukaryotic evolution. We postulate a scenario where both compartments co-evolved and that fits current models of eukaryotic evolution, explaining how ciliary proteins and nucleoporins acquired their dual functions.

INTRODUCTION

The last eukaryotic common ancestor (LECA) was a remarkably complex organism as inferred from phylogenetic reconstruction. The consensus is that it was a flagellated (1, 2), sexually dividing cell (3, 4) unable to phagocytose (5), with mitochondria (5), an endoplasmic reticulum (ER) (6), and the namesake of this kingdom of life, a nucleus (7–9). It is still hotly debated in which order the previously mentioned organelles arose and if they (especially the nucleus) came to be autogenously (10, 11) or via a symbiotic relationship between the Asgardarchaeal eukaryotic ancestor and an engulfed α-proteobacterium (12) or a myxobacterium and euryarchaeote (13). However, LECA has become more and more tangible through better modeling and phylogenetic inference in recent years.

To accommodate this plethora of membrane-bounded organelles and ensure their function, the means to facilitate directional and unambiguous transport must have already been established. The paralogous expansion of ancestral coat protein families likely shaped the organellar identity [organellar paralogy model (11, 14–19)], thereby giving rise to the complexity we observe in many lineages today. This provided a platform for lineage-specific expansion or reduction of the trafficking scaffold [e.g., rhoptry-specific RONs in Toxoplasma gondii (20) or caveolin and flotillin in metazoa for clathrin-independent endocytosis (21)].

One of the defining features of LECA was the cilium (22). The eukaryotic cilium is a highly specialized signaling “organelle” rich in receptors and able to move and facilitate locomotion (23) (Fig. 1). The term “organelle” has been used unanimously in the field but is misleading in that cilia are not compartmentalized through a continuous membrane, but rather extend from the plasma membrane into the extracellular space. They are shielded by a specialized compartment at the base that allows sorting and regulated entry and exit of ciliary molecules (24). While in extant eukaryotes, we see many different forms of cilia, whether they are immotile primary cilia specific to Metazoa or the abundance of motile cilia (more often called flagella) in almost all eukaryotes, they all trace back to the cilium (or cilia) found in LECA. The LECA cilium likely fulfilled the same functions as cilia we see today and utilized the same molecular machinery for assembly, disassembly, and maintenance as extant eukaryotes, namely the intraflagellar transport system (IFT). The IFT machinery consists of three distinct complexes (IFT-A, IFT-B, and the BBSome), all of which trace back to a common ancestral complex (25). This “proto-IFT” is itself derived from a hypothetical protocoatomer complex that gave rise to paralogous membrane-coating families that shaped the organelles found in modern eukaryotes. Although lineage-specific adaptations occurred during the evolution of the complexes involved in IFT [e.g., cytosolic assembly and disassembly of cilia in Batrachochytrium dendrobatidis and Plasmodium falciparum (26, 27) which lack the BBSome], these complexes are among the most conserved proteins derived from the protocoatomer. While the complexes are mostly known for their involvement in cilia function and dysfunction, studies show that single proteins, but not the entire complexes, play a role in different processes such as cell cycle progression and cytokinesis (28–31), establishment of the immune synapse (32, 33), transcription factor translocation (34, 35), and DNA damage signaling (36–38), with the latter two drawing a direct connection between cilia and nucleus [also reviewed in reference (39)]. It has indeed been shown that several IFT and BBSome proteins can localize to the nucleus (34, 40–42) and that they fulfill potential roles in gene regulation (34, 40), although the reason as to why they do this and how common this is among eukaryotic clades remains unclear.

Fig 1.

Overview of a eukaryotic cilium and intraflagellar transport machinery. Cargo is transported from the proximal basal body along the ciliary axoneme to the tip, where cargo is released, the IFT train is re-structured, and sent back toward the basal body.

In this review, we aim to provide an insight into the proteomic makeup of LECA from a membrane coat point of view, how the membrane-bounded organelles arose, and how they had the opportunity to interact during eukaryotic evolution. Ultimately, we establish a possible evolutionary trajectory for how ciliary proteins acquired nuclear functions, and if currently observed nuclear functions could be ancestrally conserved from LECA.

PROTEOMIC MAKEUP OF LECA ENDOMEMBRANE SORTING SYSTEMS

The ability to deform membranes is not exclusive to eukaryotes and can be observed in multiple cases of prokaryotic life. Bending membranes is essential for cell division regardless of kingdom. Some prokaryotes even form intricate membrane invaginations that need to be stabilized by proteins to maintain their structure. In fact, bacterial organisms with stacked membrane invaginations were first considered as ancestors of eukaryotes, as these structures resemble endomembrane organelles (43) and, in some cases, even engulf parts of the genetic information (44). These are, however, analogous structures that arose independently in prokaryotes and do not represent ancestors of eukaryotic endomembrane systems (45). Phylogenetic reconstruction of LECA paints the picture of a surprisingly complex cell equipped with the hallmark organelles of extant eukaryotes and whose proteomic diversity is potentially on par with modern eukaryotes (22). To face the challenges that come from hosting multiple organelles, LECA had an established set of distinct trafficking pathways to and from organelles and the plasma membrane that provided specialized cargo delivery and directionality of transport (Fig. 2). Since there are already excellent studies and reviews encompassing the plethora of complexes that were likely established in LECA together with their regulatory protein families and tethering complexes (14, 21, 46–49), we focus here on the evolution of the ones most closely involved with the eukaryotic cilium: the type I, type II, nucleoporin, and IFT family of membrane coats.

Fig 2.

Overview of protocoatomer-derived vesicle coats and their field of action. While parts of the IFT/BBSome are derived from protocoatomer, they are not vesicle coats sensu stricto. They do, however, associate with membranes for cargo delivery.

Protocoatomer-derived vesicle coats

Membrane coats differentiate vesicles and proteins depending on their origin and destination and were an essential tool for sorting the highly complex endomembrane machinery in LECA. By the time different eukaryotic lineages expanded, LECA most likely had a complex system of endomembrane organelles and subcellular compartments established that warranted sorting of cargo (14, 21, 22, 47, 50). Specific families stand out by their connection through an ancestral protein complex: the type I and type II coats, the nucleoporins (Nups) building the nuclear pore complex (NPC), and the ciliary trafficking proteins of IFT and the BBSome, all of which shape the identity of their respective compartments/organelles. They all trace back to the aforementioned “protocoatomer” protein complex that was likely established before LECA (25, 51–54). Although they diverge substantially in sequence, their α-solenoid/β-propeller structure is highly conserved and hints at their shared evolutionary history.

The type I family of coats consists of coatomer protein I (COPI), five adaptins (AP-1–5), and clathrin. COPI coats vesicles in retrograde Golgi-to-ER and intra-Golgi transport (55–58), with additional roles in endosomal transport, lipid homeostasis, mRNA transport, and nuclear envelope (NE) breakdown (59–74). Given their versatility in different trafficking pathways, it is likely they were present early in eukaryotic evolution and adapted to multiple functions. AP-1 and AP-2, in concert with clathrin, build vesicle coats that destine their cargo from the plasma membrane or the trans-Golgi network to early/recycling endosomes. The remaining AP-3–5 are involved in the sorting of vesicles for the late endosome/lysosome or have currently undefined functions (75, 76). Clathrin coats vesicles budding from the plasma membrane and vesicles destined for the trans-Golgi network and endosomes (77).

Type II family coats are COPII and HOPS (homotypic fusion and vacuole protein sorting)/CORVET (class C core vacuole-endosome tethering) proteins (50). COPII proteins facilitate the anterograde transport of vesicles from the ER toward Golgi, while the HOPS/CORVET coats are also involved in early to late endosomal sorting and late endosome to lysosome/vacuole transport.

The Nups that build the NPC share structural features with and are related to both type I and type II coats (78, 79). They form distinct complexes of eightfold symmetry in the nuclear envelope that allow selective diffusion of proteins and molecules through the NPC. Through duplication, Nups diversified before LECA, which led to the formation of subcomplexes within the NPC: the cytosolic face with filaments attached to a ring, which itself is connected and anchored via inner rings and membrane rings to the nucleosolic complex with nuclear rings and the nuclear basket (80).

The last group of protocoatomer-derived membrane coats present in LECA does not actually coat a membrane. The IFT/BBSome complexes share a common ancestor that is related to COPIε (25) and form multiprotein complexes that are needed for ciliary trafficking. They link motor proteins to their cargo and aid in the entry and exit of ciliary cargo. In contrast to other protocoatomer family proteins, IFT/BBSome proteins share considerable sequence identity across different eukaryotes, which reflects their highly specialized function in ciliary trafficking. Recent studies could show that they not only localize to the cilium or the ciliary base but also to the nucleus of some species (34, 35, 40–42). Nuclear functions could be shown for some of the IFT/BBSome proteins, but not all. This raises the question of how it could be that ciliary proteins are able to enter the nucleus and whether this phenomenon can be traced back to the evolutionary history of the nucleus and cilia. For this, we will highlight the possible order of emergence for each of the two organelles and infer potential co-evolution scenarios that could explain the duality of ciliary proteins functioning in the nucleus of eukaryotic cells.

CILIA

Across the eukaryotic Tree of Life, cilia can be found in all major lineages but with frequent secondary losses (1, 81, 82). The purpose of cilia is largely homologous, serving both sensory and locomotive functions (83, 84). Cilia are anchored to the cell via the basal body, a barrel-like structure made from nine symmetrically arranged microtubule (MT) triplets that serve as a nucleation point for the ciliary axoneme (Fig. 1). The basal body itself is highly conserved as is evident from the phylogenetic reconstruction of its components (82). The axoneme that is the center scaffold of the extending cilium is itself a structure made of MT doublets with an additional central pair in the case of motile cilia. As the structure and function of cilia and the basal body are conserved, so are the proteomes, with crucial components essentially identical in most ciliated species (85, 86). The A- and B-tubules of each axonemal doublet are the “highway” for all ciliary trafficking that is needed to assemble, disassemble, and maintain the cilium in its architectural and functional integrity. On these tracks, highly conserved, specialized intraflagellar transport complexes, IFT-A, IFT-B, and the BBSome, are moved by kinesin-2 family (anterograde) or dynein 1b/cytoplasmic dynein 2 family (retrograde) motor proteins that connect to ciliary cargo proteins (e.g., transmembrane receptors) via the BBSome functioning as an adaptor complex [reviewed in references (87, 88)]. This highly conserved machinery is likely derived from a diversification of the “protocoatomer” during the shaping of the organellar landscape in the transition from first to last eukaryotic common ancestor (25). How did the building blocks for cilia arise before LECA? Their high degree of conservation calls for an early adoption of the ciliary machinery that is directly linked to the possible nuclear functions of ciliary proteins that can be appreciated today.

Origin of eukaryotic cilia

Three major schools of thought about the emergence of cilia have emerged since the 1960s, but only two of them persist to this day. One of the first hypotheses was formulated by Lynn Margulis [later Sagan (89)], alongside her theory of sequential endosymbiosis that tried to explain the origin of other organelles like mitochondria. She hypothesized that the cilium is (like the mitochondrion) the result of an endosymbiotic event between an archaebacterium and, in this case, a bacterial spirochaete. The theory is, however, poorly supported by phylogenetic relations of eukaryotic ciliary proteins and spirochaete bacterial proteins and was therefore abandoned. Another theory postulated a viral origin, where the virus contained information about the precursors of tubulin and where the capsid served as a template for the basal body/ciliary axoneme (90). In this scenario, the cell would have supplied the motors and IFT precursors to the budding virus, resulting in an elongating axoneme. However, there is also little supporting data for this theory. A third hypothesis was first introduced in 1974 and describes an autogenous origin of both cilia and basal body from a primitive microtubule organizing center (91). This hypothesis is favored by many researchers as it does not necessitate a fortuitous union of two prokaryotic cells where one of them would lose its outer membrane and complete genetic fingerprint (as would be the case in Sagan’s theory) and also allows a gradual evolution of the cytoskeletal and MT network from prokaryotic precursors rather than a sudden genetic addition of a viral entity.

If we assume the theory of an autogenous origin for cilia and basal bodies to be true, there is a scenario where the primitive microtubular network co-evolved with ancestors of ciliary trafficking proteins and motor proteins. Initially, the IFT precursors and motors could have facilitated directed transport of receptors and vesicles to and from the plasma membrane [as was already hypothesized in reference (92)] and by paralogous expansion, begin to focus on one specific route, potentially with specific cargo affinity for receptors and receptor-based signaling components. The membrane at the destination would then be enriched in signaling molecules, providing a significant advantage for sensory perception over random receptor distribution over the whole cell surface, up to the point where receptors would inhibit each other sterically at the membrane patch subdomain (93). Consequentially, once this membrane patch is oversaturated, selective pressure would favor a cell surface architecture that further allows enhancement of signal capture and processing (94). Cilia provide this advantage: mathematical modeling of ciliary ligand capture rates concludes that an immotile cilium has the same capture rate constant as a membrane patch 3.8 times the ciliary surface area (95). In shear flow, the cilia even enhance the capture rate sixfold.

In addition to the enhanced reception of environmental cues (and prey capture), an emerging cilium provides the benefit of motility: with a growing network of microtubules and associated motors, the primitive eukaryote would have been able to move along surfaces or change direction in suspension according to a physical or chemical gradient. Although gliding is usually associated with actomyosin-dependent changes in the cytoskeleton, cilia of Helkesimastix (96) and Chlamydomonas (97–99) have been found to provide a comparable form of mobility. Recent studies identified a basal eukaryotic group that currently comprises exclusively marine heterotrophic biflagellates (100, 101). This group contains morphologically almost indiscernible species although they are only distantly related, suggesting that this morphostasis represents an ancestral eukaryotic bauplan that diversified into the recognized supergroups (102–106). Whether the signaling, locomotive, or feeding functions of cilia came first would require querying deeply branching eukaryotic proteomes. But whichever scenario is true, this theory provides a plausible evolutionary scenario for the emergence of a directed transport facilitated by diversifying families of trafficking proteins and molecular motors early in eukaryotic evolution. It could also help explain how the cilium improved the survival chances of early eukaryote ancestors, where even incremental elongations of the ciliary axoneme would provide an advantage compared to non-ciliated cells.

Given that there are many evolutionary advantages for ciliated organisms concerning signal transduction, signal capture, communication, locomotion, and feeding, it is necessary to trace back the components that make cilia possible to their evolutionary origins to substantiate the basis of the theory. Here, we concentrate on the main scaffolding protein of cilia, tubulin, and the IFT complexes, IFT-A, IFT-B, and the BBSome, the latter three still without confident placement in their evolutionary context. However, recent advances in phylogenetic reconstructions of the complexes allow for a more refined view of the evolution of cilia.

Evolution of tubulin and MT-dependent trafficking

There is indeed evidence of a distant relative to tubulin in prokaryotes. The prokaryotic FtsZ gene codes for a protein needed for chromosome segregation (107) that is still needed for plastid division in plants and algae (108, 109). Its loose sequence similarity led to it being hypothesized as the ancestor of eukaryotic tubulin (110). Physiologically, the structure of the protein and its dynamics are conserved: polymerization of both FtsZ and tubulin results in the longitudinal extension of fibers, with dynamical assembly and disassembly. However, a crucial difference between the two is the ability to interact with other proteins: prokaryotic FtsZ does not show any interaction with accessory factors unlike eukaryotic tubulin (111). In eukaryotes, tubulin is subjected to many post-translational modifications and interacts with a plethora of other cytosolic proteins and cofactors (112). With regards to cilia, one group of these clearly stands out: molecular motor proteins, such as kinesin and dynein, that are needed for transport not only along the MT network within the cell but also along the axoneme of cilia and flagella. From studies on the evolutionary relationships of different dynein subclasses, the above theory of a polarized membrane patch prior to the establishment of the cilium is plausible as IFT-related dyneins were the first family to diverge from cytoplasmic dyneins and, therefore, theoretically able to transport IFT precursors before the ciliary dyneins had evolved (113). Cilia-specific dyneins further provide the means for motile cilia. We can infer from phylogenetic reconstructions of different subclasses of dyneins that LECA had multiple paralogous dynein subclasses, meaning that they were already expanding pre-LECA (85).

The establishment of a sophisticated ciliary transport system, however, not only requires motors and tracks but also a means to attach cargoes onto the motors. Eukaryotic cells implement different IFT complexes to facilitate ciliary cargo delivery to their destination. The evolutionary history of IFT complexes, IFT-A, IFT-B, and the BBSome, does not only shed light on the transition of first eukaryotic common ancestor (FECA) to last eukaryotic common ancestor but also how parts of these subcomplexes may have acquired additional functions in an increasingly complex pre-eukaryotic cell.

Evolution of IFT

Intraflagellar transport is the process of cargo delivery from the base of the cilium through a selective barrier along the axoneme to the tip and back down and out of the cilium again. Eukaryotic cells established three main complexes to facilitate this movement: IFT-A, IFT-B, and the BBSome (87, 88, 114, 115). IFT-A and IFT-B directly interact with kinesin-2 family (anterograde) and cytoplasmic dynein 2 family (retrograde) proteins for the locomotion of pre-assembled IFT trains. These trains are assembled at the ciliary base and loaded with various cargo proteins, such as G-protein-coupled receptors (116–118), motility-related outer dynein arm proteins (119–121), αβ-tubulin heterodimers (122, 123) and the BBSome (124). The BBSome itself is an adaptor for membrane-anchored receptors that need to be transported into the cilium (125) but is also required for the regulated export of cargo from the ciliary membrane (126).

The IFT complexes are also structurally related; however, confidently placing the root between type I coats, type II coats, and the IFT/BBSome remains a challenge to this day. Unlike proteins from the other two families, IFT/BBSome proteins share astonishing sequence homology across vastly different eukaryotes together with the structural conservation that allows them to be confidently traced back to LECA. The high degree of conservation indicates, however, that the proteins were more functionally restricted and had less opportunity to diversify and expand as other coating complexes that retained structure but not sequence throughout their evolution.

For the complexes themselves, the most likely scenario of emergence was postulated some years ago (25). Here, the protocoatomer was the ancestral complex that gave rise to an IFT precursor related to COP I ε. The first subcomplex would have been IFT-B, parts of which expanded in a paralogous way and gave rise to BBSome components BBS4 and BBS8. The latest addition to this model would be IFT-A. Interestingly, parts of each complex bear a resemblance to different COP I subunits. Protein domains found in COP I α and β′ can be found in IFT-A (IFT122, IFT140, WDR19, and WDR35) and IFT-B (IFT80 and IFT172), domains aligning with parts of COP I ε in IFT-A (TTC21), IFT-B (IFT88, TTC26, and TTC30), and the BBSome (BBS4 and BBS8) (25). This mingling of different COP I family protein domains could be a hint of a parallel expansion of both COP I and IFT/BBSome proteins, where duplications in COP I subunits could have led to their inclusion into the developing IFT complexes. Some candidate members of the Lokiarchaea, a closely related clade of Eukarya, bear the genetic information for at least one COP I-like protein (127), meaning that the foundation for this class was already laid before the endosymbiosis event between archaea and bacteria. This would explain how the COP I family proteins could expand quickly and be present in a diverse range of trafficking proteins and derivatives. Although the expansion of COP I derivatives themselves likely only happened after the acquisition of the endosymbiotic bacterium (128), the same model predicts an increase in gene duplications for membrane trafficking and cytoskeletal components before endosymbiosis. This argues for a rather complex archaeal cell that could have provided membrane trafficking routes to diversify quickly into specialised transport systems of ER, GA, and cilium.

At this point, it is worth mentioning that different theories address the acquisition of endomembrane organelles, such as the ER, GA, and nucleus, and the proteins that shape them. As trafficking between these compartments and the cilium is inherently linked (if the protocoatomer hypothesis holds true), the emergence of ciliary proteins directly correlates with the establishment of structurally defined organelles or compartments.

To further investigate how ciliary proteins acquired nuclear functions, we need to dissect the architecture and evolutionary history of the nucleus, nuclear pore complexes, and nucleocytoplasmic transport adapters.

EUKARYOTIC NUCLEUS

The nucleus is the namesake of the domain Eukarya (from Greek εὖ “well, good,” and κάρυον “nut, kernel”) and the container of genetic information in an eukaryotic cell. The nucleus is a membrane-bounded organelle adjacent to the ER and its membrane is, in fact, continuous with the ER’s membrane. It sequesters DNA from the cytoplasm, thereby restricting access of soluble macromolecules and allowing precise spatiotemporal regulation of gene activation and repression, transcription, and translation. Its appearance separates the prokaryotes (bacteria and archaea) from eukaryotes and is considered one of the major transitions in life (129, 130). The nucleoplasm is accessible via highly selective NPCs, multi-protein channels in the MDa range with an eightfold symmetry that regulate the import and export of mRNAs and proteins. Phylogenetic reconstruction is not simple as sequence similarity between species is generally low, but comparative genomics revealed that LECA most likely had a fully functional NPC (9).

Architecture and evolution of nuclear pore complexes

Despite its omnipresence in eukaryotes, the proteomic makeup and overall architecture of the nucleus, and specifically of the NPCs, are subjected to quite some variability (8). Throughout the eukaryotic Tree of Life, it has undergone considerable lineage-specific adaptations (131) that reflect the adaptability of the system to multiple biological niches. NPC architecture is the same between eukaryotes (Fig. 3): the nucleoporins forming the NPC build a core scaffold of eightfold-symmetrical linked outer and inner rings on the cytosolic and nucleosolic face of the pore; the nucleoplasmic face is directly integrated into the nuclear lamina. The NPC extends into both directions on the cytosolic face with cytoplasmic filaments and the nuclear face with proteins forming the nuclear basket. The central channel is decorated with phenylalanine-glycine (FG) repeats that obstruct the free diffusion of molecules exceeding certain sizes (132–135) by forming a brush-like structure. While this structure is present in principle in all eukaryotes, the Nups themselves are only well conserved on a structural level (136–148). They are all composed of the same α-solenoid/β-propeller structure found in descendants of the hypothetical protocoatomer, which are likely relatives of archaeal predecessors (49). It is highly likely that Nups were present in LECA (9) and that they expanded before the radiation of eukaryotic lineages (78, 138, 146, 147, 149). As they have little sequence conservation (7, 8, 46), evolutionary pressure was not on the exact amino acid composition of the Nups but more on their ability to form the same complex in different ways. Nups are related to other membrane-coating proteins from both the COP I and COP II families (147), as well as adaptin-like proteins (138, 146, 150, 151). Interestingly, relatives of COP I- as well as Adaptin-domain-containing protein-coding genes could be identified in archaea (127).

Fig 3.

Structure of a nuclear pore complex. NPCs have eightfold radial symmetry. The NPC is anchored to the NE membrane by transmembrane nucleoporins (membrane ring). Membrane, inner, and outer rings associate with each other to form a barrel-like structure with a central pore. Cargoes enter the nuclear pore at the docking site and are transported by electro- and hydrostatic interactions with phenylalanine-glycine residues. FG residues convey selective transport through the nuclear pore. Cargo complexes leave the pore through the nuclear basket. Transport into the cytoplasm is analogous, with cargoes exiting from the export platforms. Based on reference (49).

As previously discussed, the expansion of this protein family led to the diverse membrane coats that shape organellar identity and assemble into some of the most intricate multiprotein complexes of eukaryotic cells. Likewise, the association of early Nup precursor proteins with membranes could likely lead to primitive repetitive structures resembling proto-NPCs in the (possibly incomplete) NE. Recent studies have shown that the currently closest sister clade within Heimdallarchaea, the Hodarchaeales, has extensive sets of genes coding for proteins involved in information processing, cytoskeletal components, a diverse trafficking machinery, and parts of the endosomal sorting system of eukaryotes (152). The apparent complexity of pre-eukaryotic cells is further elevated after eukaryogenesis, where duplication events likely shaped the full organellar landscape (128). Through duplications, the NPC gained considerable diversity in composition that retained interaction sites with other NPC subunits. Although the forming NE and NPC were initially likely to be “leaky” in the sense that they did not fully sequester DNA and its interacting proteins from the cytosol, there are obvious advantages of an incomplete NE and NPC: even purely physical obstacles like a non-selective membrane with rather large fenestrations would slow DNA-interacting molecules’ diffusion, limit gene activation and repression, and ultimately shape signaling cascades and cell cycle stages. Furthermore, the spatiotemporal separation of transcription and translation allowed for regulatory elements like introns to be inserted into the genome, giving cells the ability to produce many transcripts from a single gene. The permeability of NPCs for gene regulatory components is an evolutionarily selectable feature, favoring more complex and functionally fine-tuned structures that allow precise gene regulation. Higher selectivity would pressure the evolving cell into co-evolution of proteins that are desired in close proximity to DNA for gene regulatory (e.g., methylation and transcriptional activators or repressors) or stability (e.g., damage repair and histones) purposes and their unhindered entry into the restricted compartment. After the establishment of a full NE, the only point of entry for proteins into the nucleus would be the pores formed by proto-NPCs, which need not be the sophisticated, highly selective complexes that were most likely present in LECA. Even non-selective pores formed by membrane-deforming proto-Nups would convey size selection simply by their architecture: depending on the size of the molecule that wants to traverse the pore [where larger molecules diffuse more slowly than small ones (153)] and the diameter of the pore itself, the probability of a molecule not only encountering but also passing through a non-selective pore becomes smaller as molecule size increases and pore size decreases (154–157). Extant NPCs do, however, implement further measures to ensure that no unwanted molecules enter the nuclear space. The actual pore itself is decorated with FG moieties (136, 138) that stem from the so-called FG-Nups. These FG-Nups are highly dynamic, intrinsically disordered proteins with large hydrodynamic radii and effectively decrease the size of the pore further (158). They prevent random passage through the nuclear pore by inter- and intramolecular electrostatic interactions (159, 160), resulting in high entropy within the NPC. Establishing such an intricate structure as the NPC to prevent free nuclear localization, however, comes with the need to also import and export proteins at the right times. Part of this regulation is conveyed by FG-Nups themselves; the selective import and export of nuclear-bound molecules are, however, also enhanced by adapter proteins, the karyopherins.

Karyopherins

Karyopherins (Kaps) are a large family of proteins involved in nuclear import and export pathways. The phylogenetic reconstruction showed that LECA probably had an extensive repertoire of Kaps (52), consistent with the presence of a rather complex NPC and, therefore, also a sophisticated nuclear import and export system (52). They also form mostly α-solenoid/β-propeller structures, with the dominant protein domains being HEAT and ARM repeats (54, 161, 162), and are related to Nups and other membrane coats via their likely protocoatomer origin. Depending on their function, they are also conveniently called importins and exportins. Both protein families are transporters for nuclear proteins. Through hydrophobic interactions of Kaps with the FG residues within the NPCs, the Kap-cargo complex is able to overcome the high entropy that would normally impede free diffusion of the cargo proteins (163). Although diffusion through FG repeats is possible, the presence of specialized transporters with higher affinity to cargo binding sites in NPCs effectively blocks unspecific passage (157, 164, 165).

The import and export pathways do not only depend on Kaps but they also use the RanGTP/RanGDP gradient across the NE to further guarantee the correct direction of transport. Importins in their cargo-bound state translocate through the NPC (166) and release their cargo in the nucleus before binding nucleoplasmic RanGTP and being exported again. In the cytoplasm, the hydrolysis of RanGTP to RanGDP releases importin to be used in another cycle. For proteins to be exported from the nucleus, exportin associates with its cargo and nuclear RanGTP to form the export complex, only to be released in the cytoplasm through the hydrolysis of RanGTP to RanGDP. This gradient is crucial for nucleocytoplasmic transport (167).

Cargo is recognized by nuclear localization or export signals (NLS and NES, respectively) within their amino acid sequences and are bound by the respective importin or exportin. As Kaps have been established before the radiation of eukaryotic phyla, so must have been the sequences marking proteins as nuclear cargo. Canonical sequences seem to be functional across different eukaryotic lineages, so they most likely also originated before LECA. To understand how proteins can function in the nucleus, it is worth exploring how nuclear signal sequences came to be in the first place.

Evolution of nuclear signal sequences NLS and NES

NLS and NES are amino acid sequences that confer nuclear import or export capability to a protein. There is a range of sequences and biophysical properties that define classes of NLS and NES, and those sequence motifs are highly conserved throughout eukaryotes. Classical NLSs are either monopartite or bipartite targeting signals comprising basic amino acids that associate with Kaps for nuclear import. Monopartite NLSs contain mostly lysine and/or arginine residues and loosely follow the consensus motif K-(K/R)-X-(K/R) first identified in the SV40 large T antigen (168–171). Bipartite NLSs are separated by a spacer of variable length, but otherwise share the same biochemical properties. The first description of bipartite NLSs was in the protein nucleoplasmin, with their consensus sequence being (K/R)(K/R)X_10–12(K/R)_3/5 (172–175). Other motifs have been identified since, such as the monopartite acidic M9 motif (also known as PY-NLS, containing proline and tryptophan) from the mRNA-binding protein hnRNPA1 (176–180). Although all are either recognized directly by importin β or sequentially by importin α and then importin β and different motifs convey different transport rates (181), all motifs are sufficient to facilitate nuclear import.

NESs function analogously to NLSs but facilitate export from rather than import into the nucleus. They are composed of hydrophobic stretches of amino acids usually rich in leucines, following the consensus (L/I/V/F/M)-X_2–3-(L/I/V/F/M)-X_2–3-(L/I/V/F/M)-X-(L/I/V/F/M) (182–185). Nuclear export is facilitated by the association of the Kap exportin-1 (Xpo1; also called chromosome region maintenance 1, CRM1) (186–189), and no other major protein export pathways have been described to date.

The high degree of functional conservation of NLS and NES argues for an ancient origin. Interestingly, structural evidence for a prokaryotic origin comes from ribosome analyses. Prokaryotic ribosomal proteins contain water and magnesium ions that give the ribosome structural integrity. In human ribosomes, water and magnesium are replaced by moieties that maintain the same contacts to rRNA and functionally also serve as NLS (190). As ribosomal proteins comprise about half of the most conserved proteins throughout all domains of life (191), future proteomic analyses into deeply branching eukaryotes, archaea, and bacteria could uncover more similarities and strengthen this theory. Recent work could point to a similar direction for the origin of NLS: NLS-like amino acid sequences can also be identified in the DNA-binding regions of prokaryotic proteins (192). Indeed, in extant eukaryotes, NLSs often co-localize with DNA- or RNA-binding motifs (192–194). It would make sense that the first proteins to contain NLSs in the prokaryote-to-eukaryote transition would be the proteins already associated with DNA, as their DNA-associated functions most likely would also be needed in the evolving pre-LECA cell. Purifying selection could then have sorted out obsolete proteins. This would sufficiently explain how a primitive nuclear proteome could arise in the first place.

CILIARY PROTEINS IN THE NUCLEUS

How can the evolutionary insights into organelle acquisition and formation help us formulate a hypothesis for the nuclear functions of ciliary proteins? There are clear links between the two compartments: both nucleus and cilia are connected to the cytosol via aqueous pores and are not completely bounded by a lipid membrane, leaving them accessible to macromolecules. Free diffusion is possible for smaller molecules in both cases, but a similar size cutoff exists for both (195). Parts of the NPC could be identified at the ciliary base (196, 197), suggesting a role for NPC proteins in the formation and function of a hypothetical ciliary pore complex (CPC) that is currently not well understood. Import and export in cilia and nucleus are driven by a RanGTP/GDP gradient, where cilia and nucleus are both high in RanGTP and low in RanGDP (198, 199). Some protein cargoes destined for the cilium get imported by importin β2 along this gradient in the same fashion as nuclear cargo in the nucleus (200). The ciliary KAP3 (kinesin-associated protein 3) uses its ARM repeats for both ciliary and nuclear localization, and a RanGTP gradient controls its translocation from the cytosol to either nucleus or cilia of Chlamydomonas reinhardtii (201). There is some evidence that even suggests a ciliary localization sequence akin to the NLS that recognizes ciliary cargo for import (195, 196), although their presence alone does not suffice the import criteria.

Oftentimes, cilia sense environmental cues that result in the release of proteins from the ciliary space into the cytoplasm to intricately regulate gene expression through modulation of signal transduction cascades and transfer of transcription factors to the nucleus (202, 203). One such case is the GLI transcription factor responsible for functional hedgehog signaling: it is sequestered in the cilium by binding to a homodimer of KIF7 that electrostatically and sterically resembles a DNA double strand (204). Other cases include kinesin-associated proteins actively shuttling between the cilia and nucleus (201, 205). More recent work has shed light on IFT machinery localizing to both cilia and the nucleus (41, 42), and some could identify gene regulatory functions (34, 35, 40) or effects on cell homeostasis (206) of single IFT components, which underlines their non-ciliary functions and might hint at cases of co-option.

Both nucleus and cilia have likely been present since LECA (11, 25, 48, 82, 92, 207, 208) and were major factors in what discerns eukaryotes from prokaryotes (129, 130). While their shared common ancestor, the hypothetical protocoatomer (25, 51–54), is still elusive, it most likely originated after the endosymbiosis between bacteria and archaea. It is a matter of debate at which point the cilium was established after this point, raising an important question for the non-ciliary roles of ciliary trafficking proteins: were non-ciliary functions ancestral, or did ciliary proteins “slip” into new functions by parallel inventions of diversifying eukaryotic organelles and radiating gene families that established organellar identity? In the following, we will summarize prevailing theories as to how endomembrane organelles were formed and put them into relation to the appearance of the cilium and, thus, ciliary proteins in the nucleus.

Outside-in vs inside-out vs mitochondrial origin of endomembrane organelles

Multiple scenarios for the origin of endomembrane organelles such as NE, ER, and GA have been put forward over the last years. Three main theories paint a picture of the timing, topology, and consequences of mitochondrial endosymbiosis, formation of the nuclear envelope, and subsequent evolution of eukaryotic organelles: the outside-in (O-I) and inside-out (I-O) models, as well as the mitochondria-derived organelles (Fig. 4). As these models have been discussed in much detail and have received attention in other recent publications (209–211), we will shortly describe each, but mainly focus on their implication for the formation of cilia.

Fig 4.

(A) Outside-in, (B) inside-out, and (C) endosymbiont-borne models for the generation of endomembrane organelles. (A) Crypts formed by plasma membrane invaginations gradually form a network inside the archaeal cell, increasing the surface and eventually encasing the genetic information. An epibiotic symbiont could be incorporated into the host’s cell body by a larger crypt. Note that numerous membrane fusion events would be needed for membrane scission from the plasma membrane and formation of a separate NE. (B) Protrusions from the archaeal host reach into the extracellular space, making contact with other archaea and possibly symbiotic bacteria [see references (127, 212)]. The tips could then enlarge, engulfing epibiotic bacteria, until contacts with other protrusions would lead to membrane fusion. This would result in lamellar membrane-bounded inclusions in the cytoplasm that would later become the NE and ER. (C) Before the formation of endomembrane organelles, the future endosymbiont is incorporated into the host’s cytoplasm. Constant shedding of outer membrane vesicles (OMVs) from the symbiont would then gradually replace the host’s own plasma membrane. In the cytoplasm, these OMVs could produce GA- and ER-like organelles by fusion. Based on references (12, 209–211).

The O-I model (Fig. 4A) posits that the NE is formed independently from an endosymbiont by invaginations of the outer archaeal membrane. These structures would form similarly to phago- or endocytic cups in eukaryotes and extend around the DNA until more or less the entire genetic information is encapsulated. Multiple fusion (between independent invaginations) and scission events (from the plasma membrane) would be needed to facilitate the formation of a nuclear barrier in this manner. The benefits of an incomplete barrier are, however, present as described above, so could theoretically be under selective pressure, favoring the formation of increasingly selective pores until a full NE is formed. An epibiotic symbiont could enter the cell in the same way, eventually being fully enclosed in an outer membrane “vesicle.” This model requires multiple proteins that positively and negatively curve membranes, as were present in archaea closely related to eukaryotes (152). As membrane-bending proteins are likely the ancestors of both NPC and ciliary trafficking complexes and an increased surface area of the outer cell boundaries would provide an evolutionary advantage, this model would allow for a simultaneous co-evolution of the cilia and nucleus. In this scenario, it is, however equally feasible to have a selective NE with NPCs before the acquisition of a cilium.

The I-O model (Fig. 4B) proposes that an archaeon formed protrusions projecting from the cell body that eventually expanded and engulfed epibiotic bacteria, laying the foundation for the enslavement of the mitochondrion. Tendril-like projections have been observed in species of archaea that contact either other archaea (127, 213) or epibiotic bacteria (212). The positive curvature needed at the base of the projections could be facilitated by COP II-like proteins that have been reported in archaea (214). These membrane projections have been shown to potentially contain ribosomes and long filaments, akin to actin in eukaryotic filopodia (127). Interestingly, in extant eukaryotes, both ribosomes (215) and actin (216–220) can still be found regulating cilia dynamics. It is conceivable that the archaeal actin homologs were replaced with tubulin (or its prokaryotic relative FtsZ) after endosymbiosis and lateral gene transfer took place to yield eukaryotic cilia. During the transition from archaeon to LECA, the protrusions would eventually expand and fuse to form a “new” outer cell perimeter, leaving internal cristae that could (via fusions and extensions) form ER, GA, and the NE. The model explains the presence of “half-pores” that could differentiate into full NPCs later on and that are reminiscent of NPC constituents still found at the base of eukaryotic cilia (196).

The third model proposes the endosymbiosis event to be a prerequisite to the formation of organelles in the developing LECA (Fig. 4C). Similar to the O-I and I-O models, an epibiotic bacterium is incorporated into the host’s cytoplasm. However, this event triggers the de novo synthesis of endomembrane organelles rather than happening concomitantly. Through the secretion and insertion of bacterial outer membrane vesicles (OMVs), the host membrane consisting of archaeal isoprene ethers is replaced with bacterial fatty acid esters (12) (that are also found in eukaryotes); further secretion of OMVs into the host’s cytoplasm then forms primitive organelles through vesicle fusions. These would then differentiate into the organellar landscape found in extant eukaryotes after lateral gene transfer and expansion of the COP I-like proteins already present in archaea (152). While this theory does not comprehensively answer the question of how the endosymbiont facilitated its entry into the host, it explains how the host’s membrane could be replaced to resemble an eukaryotic plasma membrane. It is also possible that after either the O-I or I-O mechanism occurred, the endosymbiont behaved in the way described above.

What are the implications of each model for the formation of cilia and ciliary transport proteins in the nucleus? Given the presence of COP I-like and adaptin-domain-containing protein-coding genes in archaeal relatives of eukaryotes, the genetic basis for the invention of cilia is present in any model. Under the premises of the O-I model, the formation of the NE and ER would be the first structures to evolve from plasma membrane invaginations and scissions, thus predating the development of cilia or a cilia-specific proteome. This would be congruent with data from van Dam et al. (25) that suggest parts of the IFT complexes and BBSome are descendants of a COP I subunit. The COP I-related IFT complexes could then differentiate later, as could COP II family proteins, allowing a co-evolution scenario for NPC and IFT. COP I-like proteins could also potentially coat the invaginations themselves: in eukaryotes, COP I-coated vesicles are found in retrograde trafficking from GA to ER. If the coats persisted into the forming future NE, these would be the foundation for crude NPCs. The implication is that the pores formed under the O-I model would have been permeable for longer as the closure of membranes would be an extra step in the formation of the NE, meaning a more gradual evolution of NPCs and potentially more time for expansion of COP I-like proteins into IFT and COP II family proteins. The I-O model on the other hand means that the protrusions from the cell body would have been gated by COP I-like proteins in the first place, already resembling a sort of “proto-pore,” which would then have selective permeability during the evolution of NE and NPC. In this scenario, the proteins decorating the protrusion rim (which may or may not be COP I-like proteins) would have a higher selective pressure to form permeability barriers and convey selectivity compared to the O-I model as the DNA would be restricted more. It is therefore probably more likely that cilia formed later than the NPC under these circumstances and that ciliary proteins were not initially localized in the nucleus but later gained access through directed evolution toward nuclear accessibility. Under any model, cilia would only have been able to evolve after a lateral gene transfer from endosymbiont to host, as the ability to construct a microtubule-like scaffold likely came from endosymbiotic FtsZ genes. This leaves the third model where organelles are derived from endosymbiont-borne OMVs with the most opportunity for co-evolution of NPC and ciliary trafficking complexes: if endosymbiosis occurred before the establishment of any internal organelles, the cilium could form due to the FtsZ genes from the endosymbiont, the archaeal COP I-like and adaptin-domain-containing proteins could expand due to increasing energy availability, and ciliary signaling cascades could be selected for nuclear accessibility while NPCs were forming. However, apart from OMV shedding in prokaryotes (12) and the energetic advantage of an early endosymbiont (221), this theory is the most hypothetical and hard to prove without archaeological evidence.

CONCLUDING REMARKS

Shedding light onto the evolution of organelles also moonlights potential intersections and platforms that are needed for the co-evolution of protein subsets of each organelle. Their shared evolutionary history allows us to infer potential trajectories for co-evolution. Several points speak for a co-evolution of the nucleus and cilia and, by extension, for potentially ancestral nuclear roles of ciliary proteins. As mentioned above, several proteins have functions in both compartments. But the similarities do not end there. Both the nucleus and the cilium use RanGTP/RanGDP gradients and importins to facilitate translocation from the cytosol to the respective compartment, in the case of the cilium, only for specific proteins (198, 200, 222, 223). A CPC was proposed some time ago that acts akin to an NPC and even might recognize CLS similar to NLS (195, 196). This CPC is structurally distinct from the NPC as it is not constituted of all canonical NPC functional subunits and is not anchored to the membrane via transmembrane Nups (196) but might convey import selectivity and anchoring via association with other ciliary gating complexes such as the NPHP/MKS complex (224, 225). Nucleoporins have been identified at the ciliary base, orchestrating ciliary import in a manner akin to their role at nuclear pores. Nup98 (226), Nup205 (197), Nup62 (196), Nup93, Nup98, and Nup188 (227, 228), as well as Nup85 (226), are all situated at the base of cilia facilitating ciliary cargo sorting and playing pivotal roles in determining the left-right axis in mammals (227, 229, 230). While the NPC is tethered to the nuclear envelope through interactions between membrane ring Nups and other NPC constituents, cilia present just one surface with a lipid bilayer that could potentially serve as an anchor for the CPC. Nevertheless, the NPHP module responsible for cargo sorting occupies the membrane-facing ciliary space within the transition zone (231, 232) and has demonstrated interactions with Nups (229). Whether Nups are also affixed to the axonemal microtubules, forming a complete NPC/CPC, or if they line the inner perimeter of the cilium in a “brush border” arrangement remains currently undisclosed, necessitating further comprehensive interactomic investigations to fully elucidate the spatial arrangement of Nups at the ciliary base.

Taken together, the evolution of single organelles and their proteomes allows us to form a hypothesis about how the nuclear functions of ciliary trafficking proteins came to be: if cilia were born from a polarized membrane patch with signaling receptors that provided enhanced mobility, it is possible that effector molecules of these receptors regulated a change in gene expression in response to environmental cues, meaning the signaling was directly from the proto-cilium to DNA. Low-complexity, non-expanded signaling pathways might have been in place before sequential/combinatorial regulation, meaning proteins from the proto-cilium also affected gene regulation directly and needed to associate with DNA. For this, “ciliary” proteins would also have needed DNA-binding motifs (that were also present before cilia) that could eventually evolve into NLS, as has been proposed (192–194). Through this, ciliary proteins would have a preexisting propensity to enter the nucleus before the NE and NPC were fully in place. The obvious evolutionary advantage in pre-eukaryotes would be that no additional layers of regulation or de novo invention of nuclear effectors were needed due to the recycling of existing mechanisms like nuclear import. This would also explain why some of the Nups could be re-purposed to function in ciliary import sorting. In this scenario, “ciliary” proteins would have dual functions until cilia were fully established after NPCs were functionally selective barriers. With the addition of layers for signaling cascade regulation, some of the ciliary proteins would likely have lost the need to translocate to the nucleus directly and instead remained ciliary (Fig. 5). However, several theories could be applied to explain the formation of the NE, NPC, and other organelles, all of which bring their own advantages and drawbacks. To fully answer the question in which order organelles and compartments evolved, it is crucial to gain an understanding of the relationship between the derivatives of the protocoatomer. The uncertain placement of the acquisition of BBSome/IFT complexes relative to the acquisition of other organelles makes it hard to untangle the possible scenarios of how ciliary proteins could have evolved nuclear functions. Phylogenetic reconstruction of earlier diverging eukaryotic clades will bring clarity to how NPCs looked and functioned in LECA and will also help resolve the relationship between type I, type II, and IFT/BBSome coats. This will help determine the exact sequence of organelle evolution and, by extension, shed light on eukaryogenesis itself by clarifying how basic biological functions arose de novo in the developing pre-eukaryotic cell. The field of early eukaryogenesis is moving at a rapid pace due to increasingly sensitive and sensible reconstruction approaches through phylogenomics and phylotranscriptomics, and growing data availability and computing power will certainly lead to more ground-breaking discoveries in the next few years that will ultimately culminate in a clearer picture of what FECA and LECA may have been.

Fig 5.

Sequence of events leading to the ER, GA, cilium, and NE. In the post-FECA cell, the initial protocoatomer likely facilitated directed delivery from the plasma membrane to a proto-ER and vice versa. DNA was freely accessible due to a lack of NE structure. With the enrichment of receptors on a specific patch of the plasma membrane, the cell developed a primitive membrane protrusion, the proto-cilium, that further enhanced signaling capabilities. Pre-existing signaling cascades were inherited, novel pathways established, and regulated via intermediate regulatory levels. With the paralogous expansion of the protocoatomer, the cell was able to differentiate between inward and outward transport and to establish an intermediate compartment for sorting, the proto-GA. The type I coats used for inward transport could also be co-opted by the developing proto-cilium. After further diversification of type I and type II coats, the NE could form with primitive pores, limiting free diffusion of “ciliary” signaling effector proteins into the forming NE. However, some of the ciliary proteins still could enter (either due to the size or presence of advantageous protein structures), while others relied on intermediate messengers that could. At the time the NE with NPCs was fully formed, some signaling pathways must have been lost. Some pathways made use of other proteins that could enter, and others retained their nuclear functions fully.