The physiological cargo adaptor of kinesin-2 functions as an evolutionary conserved lockpick

Significance

The use of the homodimeric kinesin-2 in ciliogenesis has so far been demonstrated conclusively in the Caenorhabditis elegans model only. In this work, we uncover an activation mechanism of the homodimeric kinesin-2 that is shared between the multicellular animal C. elegans and the ciliate Tetrahymena thermophila. We identify two strictly conserved cargo adaptors that are necessary and sufficient to allosterically activate the respective motors. Notably, adaptors from the distantly related unicellular and mammalian models function as a ‘lockpick’ to fully unleash the activity of the worm kinesin-2 motor, suggesting that this activation mechanism is ancient and likely applies to more organisms given the conservation of the adaptors. Therefore, homodimeric kinesin-2 may play more fundamental roles in ciliogenesis than previously recognized.

Abstract

Specific recognition of cellular cargo and efficient transport to its correct intracellular destination is an infrastructural challenge faced by most eukaryotic cells. This remarkable deed is accomplished by processive motor proteins that are subject to robust regulatory mechanisms. The first level of regulation entails the ability of the motor to suppress its own activity. This autoinhibition is eventually relieved by specific cargo binding. To better understand the role of the cargo during motor activation, we dissected the activation mechanism of the ciliary homodimeric kinesin-2 from Caenorhabditis elegans by its physiological cargo. In functional reconstitution assays, we identified two cargo adaptor proteins that together are necessary and sufficient to allosterically activate the autoinhibited motor. Surprisingly, the orthologous adaptor proteins from the unicellular green algae Chlamydomonas reinhardtii also fully activated the kinesin-2 from worm, even though C. reinhardtii itself lacks a homodimeric kinesin-2 motor. The latter suggested that a motor activation mechanism similar to the C. elegans model existed already well before metazoans evolved, and prompted us to scrutinize predicted homodimeric kinesin-2 orthologs in other evolutionarily distant eukaryotes. We show that the ciliate Tetrahymena thermophila not only possesses a homodimeric kinesin-2 but that it also shares the same allosteric activation mechanism that we delineated in the C. elegans model. Our results point to a much more fundamental role of homodimeric kinesin-2 in intraflagellar transport (IFT) than previously thought and warrant further scrutiny of distantly related organisms toward a comprehensive picture of the IFT process and its evolution.

Almost all eukaryotic cells deploy numerous different myosins, kinesins, and dyneins to transport an astonishingly diverse set of intracellular cargo on the actin and microtubule cytoskeleton. These motor proteins appear to know when to be active or inactive, and which particular cargo to transport among many. Unmasking the molecular mechanisms of how motor proteins recognize their designated cargo, and how they are switched on and off at the right time and place is therefore a prerequisite to understand the intricacies of intracellular transport.

One particularly strict regulation of motor behavior is observed during the intraflagellar transport (IFT) that is essential for the construction of virtually all eukaryotic cilia or flagella (1–3). IFT is driven by the oppositely directed kinesin-2 and dynein-2 motors that are activated reciprocally between the ciliary base and the tip, respectively (4). While all ciliated eukaryotes studied so far deploy a dynein-2 for the retrograde transport of the IFT trains from the ciliary tip to the base, the use of kinesin-2 motors for the transport from the ciliary base to the tip differs considerably between species. For example, the green algae Chlamydomonas reinhardtii builds its flagella with the heterotrimeric CrFLA8/10/KAP kinesin-II motor (SI Appendix, Fig. S1, Top, see Table S1A for an overview of kinesin-2 motors and their oligomerization properties) (5–8). The multicellular animal Caenorhabditis elegans deploys a homodimeric CeOSM-3 kinesin-2, in addition to the heterotrimeric kinesin-II, to build its cilia (SI Appendix, Fig. S1, Middle) (9). Curiously, even though the mouse ortholog of the CeOSM-3 motor, the MmKIF17 kinesin-2, is also found in the cilium, it does not function as an IFT motor (10–13). Instead, MmKIF17 behaves as an inactive passenger on the IFT trains (SI Appendix, Fig. S1, Bottom). Molecular mechanisms that give rise to these species-specific differences in kinesin-2 regulation remain unknown.

Despite their divergent behavior in vivo, however, it is clear that the respective kinesin-2 orthologs from C. elegans (CeOSM-3) and mouse (MmKIF17) are both autoinhibited, i.e., they can suppress their own adenosine triphosphatase (ATPase) activity and switch themselves off (14, 15). Such self-regulatory mechanisms in fact apply to several myosins and dyneins as well (4, 16–18). Autoinhibition is thought to prevent futile adenosine triphosphate (ATP) hydrolysis if the motor is not bound to any intracellular cargo and cargo-binding eventually relieves the autoinhibition to enable active transport of cargo by the motor protein (17, 19). In most cases, recruitment of a motor to its designated cargo is mediated by an adaptor protein (20–23). In the case of the CeOSM-3 kinesin-2, it is the evolutionary conserved adaptor protein DYF-1/IFT70 (CeIFT70 hereafter, see SI Appendix, Table S1B for the species-specific nomenclature and the respective orthologs of the adaptors used in this study) that specifically recruits and activates the autoinhibited motor for directional transport in vivo and in vitro (Fig. 1A) (24, 25). The latter thus provokes the question why the mammalian MmKIF17 kinesin-2 remains inactive despite the presence of MmIFT70 in the cilium (SI Appendix, Fig. S1, Bottom and Table S1B) (11, 26). However, the knowledge of homodimeric kinesin-2 motors is limited as they have been mechanistically characterized in the C. elegans and mouse models only (14, 15, 25).

Fig. 1.

Activation mechanism of CeOSM-3. (A) To power its IFT trains toward the ciliary tip, C. elegans deploys a heterotrimeric kinesin-II and a homodimeric CeOSM-3 kinesin-2 motor (C. elegans components are color-coded in red hereafter). At the ciliary base, the autoinhibited CeOSM-3 must be recruited to the IFT trains and activated for anterograde transport. Here, the CeIFT70 adaptor plays a key role in the specific recruitment and activation of the CeOSM-3 kinesin-2. (B) Proposed models of autoinhibition and activation of the CeOSM-3 motor. The distal C-terminal tail (553-699) folds onto the N-terminal head domains (1-349) to inhibit the motoŕs enzymatic activity. This autoinhibitory folding is mediated by the conserved glycine (G444) that is found roughly in the middle of the coiled-coil stalks (350-552) (see SI Appendix, Fig. S2). (i) The autoinhibition can be relieved by mutating the flexible glycine into a glutamic acid residue (CeOSM-3^G444E). Preventing folding of the tail onto the head domains results in a constitutively activated motor with a velocity of ∼1.1 to 1.2 µm/s (15, 25). (ii) The same activation is achieved when the autoinhibited CeOSM-3 motor binds to its CeIFT70 adaptor subunit (25). (iii) The CeIFT70-mediated incorporation of the autoinhibited CeOSM-3 into its cargo complex (CeIFT70/52/46/88), significantly increases the motor’s speed to ∼1.7 µm/s (25).

While it is clear that specific motor recruitment to a given cargo and motor activation represent key steps of regulated intracellular transport, the molecular details of motor-cargo interaction and how such interaction gives rise to the activation of the motor protein remain largely unknown. Scarcity of reconstituted motor-cargo complexes is arguably one of the major obstacles on the way toward a molecular picture of how cells regulate the transport of intracellular cargo. Here we turned to the first reconstituted kinesin-2-cargo complex to dissect the cargo-mediated motor activation mechanism as a physiologically relevant example.

Results and Discussion

The Distal C Terminus is Involved in the Autoinhibition of CeOSM-3.

Kinesins dimerize with their coiled-coil domain (or stalk) of varying lengths followed by a random coil domain (or tail domain) at the distal C terminus (Fig. 1B). Both, the stalk and the tail domains have been implicated in the autoinhibition of kinesin motors (14, 27, 28). Autoinhibition is thought to take place through intramolecular folding that puts the N-terminal catalytic head domain in direct contact to the distal tail domains (Fig. 1B) (27, 29, 30). Consistent with this notion, preventing intramolecular folding through mutation of the flexible glycine residue through a point mutation (CeOSM-3^G444E) sufficed to partially activate the autoinhibited CeOSM-3 kinesin-2 from C. elegans (15), and the motor could now move with ∼1.1 to 1.2 µm/s on surface-attached microtubules (Fig. 1B, i) (15, 25). A similar result was obtained by binding the autoinhibited CeOSM-3 to its specific CeIFT70 adaptor protein (Fig. 1B, ii) (25). Of note, when the CeOSM-3/CeIFT70 complex was incorporated into its physiologically relevant cargo (CeIFT70/52/46/88), the motor’s speed further increased to ∼1.7 µm/s in vitro (Fig. 1B, iii) (25). This full activation in fact best recapitulates the in vivo velocities of the CeOSM-3 motor in the absence of the heterotrimeric kinesin-II that shares the work-load with the homodimeric CeOSM-3 in the middle segment of axonemes in C. elegans (Fig. 1A) (31).

To delineate the mechanism of full activation of the autoinhibited CeOSM-3 kinesin-2, we first removed its stalk and tail domains and asked whether the motor domains alone move as fast as the physiological CeOSM-3-cargo complex (Fig. 1B, iii) (25). To create a double-headed motor without any potential inhibitory stalk and/or tail domains, we forced dimerization of the enzymatic motor domains (CeOSM-3^1-349) with the so-called leucine zipper GCN4 (Fig. 2A, Right) (see SI Appendix for the protein sequences of all constructs used in this study) (32). We have additionally introduced a C-terminal GFP-tag to follow the movement of the truncated motor on surface-attached microtubules in a Total Internal Reflection Fluorescence (TIRF) microscope (Fig. 2A, Right). Indeed, the artificially dimerized motor reproduced previously observed full speed of the physiological CeOSM-3-cargo complex (Fig. 2A, Middle vs. Right and SI Appendix, Movie S1) (25).

Fig. 2.

The C-terminal tail domain is required for the autoinhibition of the CeOSM-3 kinesin-2. (A) The binding of the IFT subunit CeIFT70 activates the full-length motor with a velocity of 1.3 ± 0.3 µm/s (n = 210). CeIFT70/52/46/88 fully activates the full-length motor [v = 1.7 ± 0.6 µm/s, n = 187, both replotted from ref. (25) for direct comparison]. Removal of the entire stalk and tail domains also fully activates the truncated CeOSM-3^1-349 motor that is dimerized with an artificial coiled-coil from GCN4 (v = 1.7 ± 0.5 µm/s, n = 237) (see SI Appendix, Table S2 for all run length values that correspond to the respective velocity plots). (B) The full-length CeOSM-3 motor does not display any significant activity in microtubule-activated ATPase assay (dark blue diamond), while the truncated, GCN4-dimerized CeOSM-3^1-349 is active (light blue circle). The presence of the GCN4-dimerized tail domain (CeOSM-3^553-699, green circle), but not the stalk domains (CeOSM-3^350-444 and CeOSM-3^445-552, red circle) inhibits the activity of the truncated CeOSM-3^1-349 motor. (C) The fluorophore-labeled CeOSM-3^1-349 motor was titrated with the tail domain CeOSM-3^553-699 (green), and the respective CeOSM-3^350-444 (red) and CeOSM-3^445-552 stalk domains (blue) in MST assays. While the stalk domains failed to bind to the motor altogether (blue and red), the distal tail domain displayed clear binding to the truncated CeOSM-3^1-349 motor (green). The velocity data points in A were fitted to a Gaussian distribution (± width of distribution). The data points in C were fitted using Hill fitting. Error bars in the ATPase and MST assays represent SD; all data obtained from three independent protein purifications each.

Next, we created truncations of the C-terminal stalk and tail domains to delineate the domains that are required for the inhibition of the catalytic motor domains. To this end, we monitored the respective activities of the full-length CeOSM-3 and the truncated, fully active CeOSM-3^1-349 motor in microtubule-activated ATPase assays. The full-length CeOSM-3 remained inactive (Fig. 2B, dark blue diamonds), as seen previously (15). Removal of the entire stalk and tail domains significantly increased the activity of the CeOSM-3^1-349 motor (Fig. 2B, dark blue diamonds vs. light blue circles).

To demarcate the domains responsible for the autoinhibition of the full-length CeOSM-3, we next asked if the addition of the respective stalk and tail domains in trans would suffice to suppress the activity of the truncated CeOSM-3^1-349 motor. The presence of the stalk domains (CeOSM-3^350-444 and CeOSM-3^445-552) did not alter the activity of the CeOSM-3^1-349 motor (Fig. 2B, red vs. light blue circles). The tail domain (CeOSM-3^553-699), in contrast, was sufficient to suppress the activity of the truncated CeOSM-3^1-349 motor in trans (Fig. 2B, light blue vs. green circles).

To demonstrate the direct interaction between the motor and the tail domains, we conducted a series of microscale thermophoresis (MST) assays (33). While we observed no binding between the motor and the stalk domains (Fig. 2C, blue and red circles), the distal tail domain clearly displayed binding to the truncated CeOSM-3^1-349 motor (Fig. 2C, green circles). We note however that a lack of binding in trans does not fully exclude the possibility of interaction between the head and stalk domains as the binding strength may be too weak to observe in the MST assays. Previous work in fact supports a direct interaction between the head and stalk domains in other kinesin motors, including the mammalian MmKIF17 kinesin-2 ortholog of the CeOSM-3 motor (14, 28).

We conclude that the motor domain alone is capable to walk as fast as the fully activated CeOSM-3. The CeIFT70 adaptor protein in conjunction with other cargo proteins relieves the autoinhibition of the full-length CeOSM-3 motor imposed at least by the C-terminal tail domain by sequestering it via a direct interaction. We next turned to the CeIFT70 adaptor protein to delineate its role in the full activation of the CeOSM-3 kinesin-2 from C. elegans.

The CeIFT70 Requires the Presence of Another Coadaptor to Fully Activate the CeOSM-3 Kinesin-2.

As detailed above, the full activation of the autoinhibited CeOSM-3 kinesin-2 is observed only in the context of the physiological cargo complex that is formed by four different subunits (CeIFT70/52/88/46) (Fig. 2A, Left vs. Middle) (25). Since the CeIFT70 adaptor protein alone does not fully activate its motor, it is conceivable that the CeIFT70 adaptor needs to interact with at least one additional subunit of the cargo complex to fully activate the CeOSM-3 motor. Consistent with this notion, the CeIFT70 ortholog from C. reinhardtii (CrIFT70) was shown to wrap around a short, proline-rich stretch of the conserved CrIFT52 protein in a previous high-resolution crystal structure (34). We therefore asked whether this rather unusual interaction between the CrIFT70 and the CrIFT52 proteins plays any direct role in the full activation of the CeOSM-3 motor. To test this possibility, we labeled the respective C. elegans orthologs (CeIFT70 and CeIFT52) with different fluorophores, along with the CeIFT88 subunit of the physiological cargo complex as a control (Fig. 3 A and B, Left). Dual tracking of the differentially labeled adaptor subunits ensured that the data are obtained exclusively from the motors each bound to both, CeIFT70/52 or CeIFT70/88 complexes, respectively.

Fig. 3.

CeIFT70 and CeIFT52 subunits are necessary and sufficient for the full activation of the CeOSM-3 motor. (A vs. B) CeIFT70^SNAP647 fully activates the CeOSM-3 motor only in the presence of CeIFT52^GFP (B, v = 1.8 ± 0.2 µm/s, n = 91) but not in the presence of the CeIFT88^GFP subunit of the physiological cargo complex (A, v = 1.2 ± 0.2 µm/s, N = 117). Velocity data were fitted to a Gaussian (± width of distribution) distribution (see SI Appendix, Table S3A for two-sample t tests of the velocities). Data were obtained from three independent protein preparations each.

In the absence of any adaptor protein, CeOSM-3 remained autoinhibited and was incapable of directional movement (SI Appendix, Movie S2, Top Left) as expected from previous work (15, 25). The presence of the respective adaptor complexes CeIFT70/52 and CeIFT70/88 activated the CeOSM-3 motor (SI Appendix, Movie S2, Top Right and Bottom Left).

However, presence of the CeIFT88 subunit, in addition to CeIFT70, failed to fully activate the CeOSM-3, and the motor moved at a comparatively slow speed of ∼1.2 µm/s (Fig. 3A), which is similar to what has been observed by the CeIFT70-mediated activation (Fig. 3A vs. 2A, Left) (25). In contrast, the presence of the CeIFT52 subunit, in addition to CeIFT70, was necessary and sufficient to reproduce the previously observed full activation of the autoinhibited CeOSM-3 by its physiological complex (Fig. 3B vs. 2A, Middle) (25). Consistent with the C-terminal random coil being involved in the autoinhibition of the CeOSM-3 motor (Fig. 2 B and C), the CeIFT70/52 adaptor complex coprecipitated with the random coil domain and not with the coiled-coil stalks or the motor domains (SI Appendix, Fig. S3).

We next tested if CeIFT70/52 could stimulate a mutated version of the motor, CeOSM-3^G444E (Fig. 1B, i). This mutation is known to activate the motor by preventing autoinhibitory folding and allows it to walk with ∼1.1 to 1.2 µm/s (15, 25). Binding of CeIFT70 alone did not suffice to stimulate the velocity of CeOSM-3^G444E, however, CeIFT70/52 stimulated also this version of the motor to move with 1.7 ± 0.3 µm/s as expected (SI Appendix, Fig. S4 A vs. B). In contrast, neither CeIFT70/88 nor CeIFT70/46 could further stimulate the partially activated CeOSM-3^G444E (SI Appendix, Fig. S4 C and D), underscoring the significance of the complex formation between the CeIFT70 and the CeIFT52 subunits in the full activation of the CeOSM-3 motor.

Importantly, the activating potentials of CeIFT70/52 and the G444E mutation were not additive to each other because velocities did not exceed 1.7 µm/s for the CeIFT70/52-bound CeOSM-3^G444E motor. We therefore suggest that both CeIFT70/52 and the G444E mutation influence steps in the same pathway toward relieve of autoinhibition. Lack of additivity also rules out a simple model in which the motor exists in only two states, fully inhibited or fully active, and in which intermediate velocities manifest themselves by a rapid switch between these states. Instead, at least three states must exist to explain the data, a fully inhibited state, a structural intermediate and a fully active state (SI Appendix, Fig. S4E). In the simplest version of this three-state model, the structural intermediate moves at intermediate velocities of ∼1.2 µm/s, but scenarios in which it walks with much lower velocity are theoretically possible. The three-state model also implies that CeOSM-3 is autoinhibited by at least two different mechanisms. Above, we provide evidence for tail-domain mediated autoinhibition of the motor (Fig. 2). The nature of the second autoinhibitory mechanism remains to be explored in more detail in future studies.

Molecular Determinants of Allosteric Activation by the IFT70/52 Cargo Adaptor.

To better understand the specific role of the IFT70/52 adaptor complex, we next created mutants to delineate the regions in the respective subunits that are required for the allosteric activation of the CeOSM-3 motor. To this end, we turned to the high-resolution crystal structure of the orthologous IFT70/52 adaptor complex from C. reinhardtii where CrIFT70 was shown to wrap around a short and conserved proline-rich stretch in the CrIFT52 subunit (34). Notably, removal of this corresponding region in the CeIFT52 subunit (CeIFT52^Δ322-376) abolished the full activation of the CeOSM-3 motor, while the short peptide containing the conserved proline-rich region alone (CeIFT52^322-380) was sufficient for full activation (SI Appendix, Figs. S5A and S6). These results underscore the functional relevance of the evolutionary conserved proline-rich region and suggests that the peculiar way how the CrIFT70 subunit wraps around this stretch in the CrIFT52 protein from the C. reinhardtii model (34) might also take place in the respective orthologs from C. elegans.

Intrigued by this functional interdependence between the CeIFT70 and CeIFT52 subunits of the physiological cargo complex (Fig. 1B, iii), we next turned to the CeIFT70 adaptor protein to unmask the determinants that contribute to the full activation of the autoinhibited CeOSM-3 motor in the presence of the CeIFT52 subunit (Fig. 3). To this end, we focused on a particularly conspicuous repetitive tyrosine motif in the N terminus of the CeIFT70 adaptor protein that is conserved between distantly related eukaryotes (SI Appendix, Fig. S7) (34, 35). The aromatic rings of these repetitive tyrosines were found to tightly stack against the highly conserved proline-rings in the aforementioned proline-rich stretch of the CrIFT52 subunit as seen in the high-resolution structure from the C. reinhardtii model (Fig. 4A) (34). Because this stacking was proposed to stabilize the predominantly hydrophobic interface between the CrIFT70 and CrIFT52 orthologs from C. reinhardtii (34), we asked whether interfering with this conserved interface in the CeIFT70 and CeIFT52 orthologs from C. elegans has any functional consequences for the full activation of the CeOSM-3 motor.

Fig. 4.

The evolutionary conserved tyrosines in CeIFT70 are key to the full activation of the CeOSM-3 kinesin-2. (A) A conserved tyrosine motif Y55,57,58,59 of CrIFT70 forms an exceptional stacking arrangement with the proline rings of P357,358 in CrIFT52 in the high-resolution crystal structure from Taschner et al. (34). (B) Sequence alignment of the repetitive tyrosines (see also SI Appendix, Fig. S7) in the wild-type CeIFT70 (top lane) with the CeIFT70^Ala mutant Y51/53/54/55A (green) and CeIFT70^Phe mutant Y51/53/54/55F (orange), respectively. (C) The full activation of the CeOSM-3 motor by the wild-type CeIFT70/52 adaptor complex replotted from Fig. 3B for direct comparison. (D) Mutating the wild-type tyrosines into alanines in CeIFT70^Ala,SNAP647 prevents the full activation of the CeOSM-3 motor (v = 1.3 ± 0.2 µm/s, n = 145). (E) The full activation of CeOSM-3 is rescued by replacing the alanines with phenylalanines in CeIFT70^Phe,SNAP647 (v = 1.7 ± 0.2 µm/s, n = 109). Velocity data were fitted to a Gaussian (± width of distribution) distribution (see SI Appendix, Table S3A for the corresponding two-sample t tests). Data were obtained from three independent protein preparations each.

To test this hypothesis, we mutated four tyrosines into alanines in the CeIFT70 subunit (Fig. 4B). The alanine mutations in the CeIFT70 adaptor abolished the CeIFT70/52 mediated full activation, and the CeOSM-3 moved at decreased speeds that were consistent with the CeIFT70-mediated activation of the motor (Fig. 4 D vs. C and SI Appendix, Fig. S5B and Movie S3, Left) (25). We rescued the full activation of CeOSM-3, however, by mutating the tyrosines into phenylalanines instead of alanines (Fig. 4 E vs. D and SI Appendix, Fig. S5B and Movie S3, Right). Phenylalanine is in fact structurally similar to tyrosine, except that it lacks the hydroxyl group on the aromatic ring. Based on the capability of phenylalanines to fully substitute the function of wild type tyrosines, we propose that the hydrophobic stacking between the tyrosines in CeIFT70 and prolines in CeIFT52, as seen in the C. reinhardtii proteins (34), is necessary and sufficient for the full activation of the CeOSM-3 motor. Consistently, the N-terminally truncated CeIFT70b^Δ1-59 that lacks the conserved tyrosine motif failed to fully activate CeOSM-3 (SI Appendix, Table S1B and Figs. S8 and S5C).

Yet, the presence of such potentially functional interaction between the CrIFT70 and CrIFT52 subunits from the green algae C. reinhardtii is puzzling given that it solely deploys a heterotrimeric kinesin-2 motor that in turn is recruited to the IFT trains via the Kinesin Associated Protein (KAP) (SI Appendix, Fig. S1, Top) (6–8). In contrast, a homodimeric kinesin-2 is deployed in the mouse model. However, the mouse motor appears to be inactive inside the cilium despite the presence of the IFT70/52 adaptor complex (SI Appendix, Fig. S1, Bottom). We next turned to the C. reinhardtii and mouse models to assess the functional relationship between the kinesin-2 orthologs and the IFT70/52 cargo adaptor, respectively.

The Adaptor Function, and not the Kinesin-2 Motor, is Kept Conserved.

To our surprise, not only did the respective CrIFT70/52 and MmIFT70/52 adaptor complexes from C. reinhardtii and mouse activate the CeOSM-3 motor from C. elegans, but the activation levels were indistinguishable from the worm CeIFT70/52 adaptor complex (Fig. 5 and SI Appendix, Figs. S5 D and E, S9, and S10 and Movie S4). This remarkable functional equivalency suggests that a functional on-switch for the homodimeric kinesin-2 motor from the multicellular C. elegans may have developed early and kept conserved throughout the evolution. Given that the adaptor complex from mouse was fully functional, we next dissected the behavior of the mouse kinesin-2 to assess whether the same activation adaptor-mediated activation mechanism also applies to this particular motor.

Fig. 5.

C. reinhardtii and mouse adaptors phenocopy the function of the C. elegans adaptors CeIFT70 and CeIFT70/52 in vitro. The autoinhibited CeOSM-3 motor is activated by the IFT70 and IFT70/52 adaptors from the distantly related C. reinhardtii and mouse (see SI Appendix, Fig. S9 for the respective velocity distributions and values). The respective CrIFT70 and MmIFT70 adaptors along with the CrIFT70/52 and MmIFT70/52 adaptor complexes reproduce the behavior of the worm CeIFT70 and CeIFT70/52 proteins in functional reconstitution assays (see SI Appendix, Fig. S10 for the functional equivalency of the serial orthologs of MmIFT70).

As seen previously with the respective kinesin-2 motors from C. elegans and mouse (SI Appendix, Movies S2, Top Left and S5, Left) (14, 15), the MmKIF17 kinesin-2 did not display any directional movement. It only diffused along the surface-attached microtubules (SI Appendix, Movie S5, Left). This behavior is consistent with the full-length MmKIF17 motor being autoinhibited (14). If true, disengagement of the inhibitory C terminus either by binding to an artificial cargo or simply by its C-terminal truncation should activate the MmKIF17 motor. Indeed, surface attachment of the full-length MmKIF17 via its C terminus as well as the C-terminal truncation MmKIF17^1-747 activated the respective motors to levels that were consistent with previous in vivo studies (SI Appendix, Movies S5, Right, S6, and Fig. S11 A and B) (14, 36, 37). These results demonstrate that the recombinantly expressed MmKIF17 construct is functional in our reconstitution assays and suggests that the diffusive behavior of the full-length motor results from the C terminus-mediated autoinhibition, consistent with previous results (14).

Next, we asked whether the autoinhibited MmKIF17 interacts with MmIFT70, and if so, whether this interaction relieves the autoinhibition of the motor to allow directional transport. To this end, we first assessed if MmKIF17 motors colocalize efficiently with the MmIFT70 adaptor protein (SI Appendix, Table S1B). In particular, we tested the full-length MmKIF17 along with the truncated MmKIF17^1-747 motor that lacked its C terminus as a control for C-terminal adaptor binding (SI Appendix, Fig. S11C). MmKIF17 with or without its C terminus did not efficiently colocalize with the MmIFT70 protein and its serial orthologs MmIFT70A2 and MmIFT70B (SI Appendix, Table S1B), suggesting that the motor-adaptor protein pair has lost their ability to interact with each other in the course of evolution (SI Appendix, Fig. S11C, lanes I to IV), nor did MmKIF17 interact with the CeIFT70 (SI Appendix, Fig. S11C, lane V). In functional transport assays, the MmKIF17 motor failed to display any colocalized movement with the MmIFT70 adaptor and consequently remained autoinhibited (SI Appendix, Movie S7, Left). In stark contrast, the CeOSM-3 motor displayed a significant colocalization with the distantly related mouse MmIFT70 adaptor (SI Appendix, Fig. S11C, lane VI) as expected from the ability of the mouse adaptor to activate the CeOSM-3 motor from C. elegans (Fig. 5 and SI Appendix, Figs. S9 C and D and S10).

Based on our findings from functional reconstitution assays, we therefore propose that loss of activation of the mouse kinesin-2 motor by its adaptor arose from the diversification of the kinesin-2 motor and not the cargo adaptor. Specifically, MmKIF17 motor evolved to lose its ability to interact with the MmIFT70 adaptor for activation, and consequently behaves as an inactive passenger of the IFT trains in the mammalian cilium (SI Appendix, Fig. S1, Bottom). However, at this point, it cannot be fully excluded that MmKIF17 employs other activation mechanisms during IFT in a cell type-specific manner.

Intrigued by the striking conservation of the IFT70/52 adaptor function (Fig. 5), we turned to the KAP that is known to function as a universal adaptor for the heterotrimeric kinesin-2 motors (3, 38). In particular, we asked if the functional compatibility that we observed between the homodimeric kinesin-2 and the IFT70/52 adaptor extends to this paralogous class of kinesin-2 motors as well. To this end, we tested if the heterodimeric CeKLP11/20, which powers IFT together with the homodimeric kinesin-2 in C. elegans (SI Appendix, Fig. S1, Middle), retains the ability to interact with the orthologous KAP proteins from C. reinhardtii and mouse, respectively. Whereas CeKAP bound CeKLP11/20 as expected from previous in vivo and in vitro work (9, 39, 40), the KAP adaptors from C. reinhardtii and mouse did not (SI Appendix, Fig. S12A).

Functionally, KAP binding to the heterodimeric kinesin-2 motors also leads to varying effects. CeKAP did not kinetically activate the CeKLP11/20 motor in single molecule assays (SI Appendix, Fig. S12B), in line with results obtained for the corresponding mouse ortholog, heterodimeric MmKIF3A/3B kinesin-2 (41). In contrast, the KAP adaptor from C. reinhardtii fully activated the heterodimeric CrFLA8/10 motor as we demonstrated recently in functional reconstitution studies (8).

In summary, we discovered that IFT70/52-adaptor binding and the ensuing activation of homodimeric kinesin-2 motors is much more conserved than KAP-adaptor binding and its effects on heterodimeric kinesin-2 motors. These findings demonstrate that the recruitment and activation mechanism of kinesin-2 motors are complex and must be clarified experimentally in a case-to-case basis.

A Functional Cargo-Motor Interface has been Established Early in the Evolution.

C. elegans is so far the only model organism where the use of a homodimeric kinesin-2 during the IFT process has been demonstrated conclusively (SI Appendix, Fig. S1) (3, 9, 40). Given the presence of a functional cargo adaptor already in the green algae C. reinhardtii, we considered the possibility that IFT70/52-regulated kinesin-2 transport predates metazoan evolution. In fact, phylogenetic analyses predict the existence of a number of kinesin-2 motors in several unicellular organisms (3, 42). Yet, whether the predicted proteins assemble into heterotrimeric or homodimeric kinesin-2 motors cannot be determined by sequence analysis, and instead requires recombinant reconstitution or tissue purification (38, 39, 43–45). The unicellular ciliate Tetrahymena thermophila represents a particularly interesting case as it harbors an unusually high number of six kinesin-2 motors in its genome (3, 42). While three kinesin-2 motors (i.e., Kin5, KIN1, and KIN2) have been implicated in IFT in previous in vivo studies (46–48), the oligomerization and regulation mechanism of kinesin-2 motors in T. thermophila remain unknown. The TtKin5, in particular, was suggested to be orthologous to the CeOSM-3, and consistent with its proposed function as a ciliary kinesin-2, TtKin5 was up-regulated after deciliation and localized along the ciliary axoneme (46). Of note, the orthologous IFT70 adaptor from this organism also contains the conserved tyrosine motif that is essential for the allosteric activation of the CeOSM-3 motor (SI Appendix, Fig. S7), suggesting that it may function as a cargo adaptor for the TtKin5 motor. Further supporting this notion, T. thermophila only assembled short axonemes in the absence of ift70 function (49), resembling the phenotype seen in C. elegans where the absence of ift70 also leads to short cilia (24). We therefore turned to our reconstitution assays to directly contrast the activation mechanisms of kinesin-2 orthologs between the distantly related T. thermophila and C. elegans models.

Recombinantly expressed full-length TtKin5 was homodimeric as judged from size-exclusion chromatography coupled to multiple-angle light scattering (SEC-MALS) analysis (SI Appendix, Fig. S13A), and moved microtubules in gliding filament assays with a velocity of ∼2 µm/s (SI Appendix, Fig. S13B and Movie S8). Consistent with TtKin5 forming a functional homodimeric kinesin-2, it displayed processive motility on surface-attached microtubules (SI Appendix, Movie S9, Top Left). In contrast to CeOSM-3, however, it does not rely on other proteins to become active. Nevertheless, the TtKin5 motor moved substantially slower on surface attached microtubules (∼1.4 µm/s; SI Appendix, Fig. S13C) when compared to its microtubule gliding velocity (∼2 µm/s, SI Appendix, Fig. S13B), hinting at potential autoinhibition that was relieved by surface attachment in the filament gliding assay as illustrated in SI Appendix, Fig. S13B, Left.

If the surface attachment represents the full activation of TtKin5, and if the TtIFT70/52 functions as a cargo adaptor of this kinesin-2 from T. thermophila, then a trimeric complex of TtKin5 with TtIFT70/52 would be expected to form and move at a velocity of ∼2 µm/s. This is what we could observe in the presence of TtIFT70/52, but not the TtIFT70 subunit alone (Fig. 6 A vs. B and SI Appendix, Fig. S5F and Movie S9, Top Right vs. Bottom Left). We note the remarkable functional equivalency between the CeIFT70/52-mediated activation of the CeOSM-3 and the TtIFT70/52-mediated activation of the TtKin5 kinesin-2 motors (Fig. 6 A/B vs. C/D). Even the magnitude of the stimulatory effect is similar.

Fig. 6.

The ciliate T. thermophila and the metazoan C. elegans share a common mechanism of allosteric motor activation. (A vs. B) The presence of the TtIFT70 and TtIFT52 subunits is necessary and sufficient to fully activate the homodimeric TtKin5 kinesin-2 from T. thermophila (A, v = 1.5 ± 0.4 µm/s, n = 250; B, v = 2.0 ± 0.3 µm/s, n = 226), just like the CeOSM-3 kinesin-2 from C. elegans (C vs. D, replotted from Figs. 2A and 3B to facilitate direct comparison to the T. thermophila model). Velocity data were fitted to a Gaussian (± width of distribution) distribution (see SI Appendix, Table S3C for two-sample t tests of the velocities). Data were obtained from three independent protein preparations each.

Together with previous in vivo work (46), results from our reconstitution studies make a strong case that the ciliate T. thermophila model deploys at least one homodimeric kinesin-2 during IFT. However, T. thermophila uses two more kinesin-2 motors (KIN1 and KIN2) for the IFT process that seem to work redundantly, i.e., only the knockout of both motors leads to almost complete loss of cilia in vivo (48). Given that T. thermophila is predicted to have at least three more kinesin-2 motors (3), it is likely that the ciliary deployment of the heterotrimeric and/or homodimeric kinesin-2 motors is redundant in this unicellular eukaryote, conceptually similar to what has been observed in the C. elegans (3, 9).

The fact that kinesin-2 motors from a ciliate (TtKin5) and from a multicellular animal (CeOSM-3) share the same activation mechanism suggests that homodimeric kinesin-2 deployment and IFT70/52-mediated regulation has been an early evolutionary invention to power IFT. Mouse MmKIF17 kinesin-2 represents a remarkable outlier to this strictly conserved machinery, despite being a close ortholog of CeOSM-3 (42). In other words, relatively recent changes occurred in the mouse MmKIF17 kinesin-2 that disrupted the canonical interface between the motor and its adaptor (SI Appendix, Fig. S11C and Movie S7, Left). Evolutionary relatedness of the kinesin-2 motor domains therefore cannot fully predict their activation mechanisms, and as concluded above, must be determined experimentally on a case-to-case basis (SI Appendix, Fig. S14).

Taken together, our results warrant a closer look at the homodimeric kinesin-2 that is considered as an accessory motor of the IFT process in multicellular animals. It will be important to explore further whether the common allosteric activation mechanism of the kinesin-2 motor, which we show is shared between the ciliate T. thermophila and the multicellular animal C. elegans, also applies to other organisms. We predict that the role of the homodimeric kinesin-2 motor may be underappreciated and that it plays a more fundamental role in the IFT process than previously recognized.

Data Materials, and Software Availability

Microscopy movies that are used for the analyses are included in the supplementary data. The corresponding original movie files are available from the corresponding author upon request. Datasets generated during and/or analyzed during the current study, along with all study-specific reagents are available from the corresponding author upon request. All other study data are included in the article and/or supporting information.