Quantitative proteomics reveals insights into the assembly of IFT trains and ciliary assembly Free

Cilia are cellular organelles that protrude from the cell surface. They function in cellular motility as well as signaling (Goetz and Anderson, 2010; Zhou and Roy, 2015). Defects in ciliary structure or function are associated with a cohort of human diseases and/or developmental disorders (Reiter and Leroux, 2017). Cilia are built through a ciliary transport process termed intraflagellar transport (IFT) (Kozminski et al., 1993), which delivers ciliary building blocks for ciliary assembly and maintenance and returns turnover products. The IFT machinery is composed of anterograde motor kinesin-2, retrograde motor IFT dynein and two large protein complexes, IFT-A and IFT-B, that serve as cargo adaptors (Cole et al., 1998; Jordan and Pigino, 2021; Taschner and Lorentzen, 2016). At the ciliary base, the IFT-A and IFT-B complexes together with inactive IFT dynein are assembled into linear arrays termed IFT trains (Jordan et al., 2018; Pigino et al., 2009; van den Hoek et al., 2022; Wingfield et al., 2017). IFT trains undergo anterograde transport towards the ciliary tip, which is powered by kinesin-2 (Kozminski et al., 1995; Snow et al., 2004). At the tip, following remodeling of IFT-A and IFT-B complexes and activation of IFT dynein (Jordan et al., 2018; Pedersen et al., 2006; Toropova et al., 2017), new IFT trains are assembled and undergo retrograde transport towards ciliary base driven by IFT dynein (Pazour et al., 1999; Porter et al., 1999). During anterograde transport, kinesin-2 interacts with IFT-B (Funabashi et al., 2018; Zhu et al., 2021), which further binds to IFT-A and inactive dynein (Jordan et al., 2018). During retrograde transport, active IFT dynein is believed to bind to IFT-A, which further interacts with IFT-B as mutations in IFT-A often cause accumulation of IFT-B at the ciliary tip (Iomini et al., 2009; Zhu et al., 2017a). The return of kinesin-2 occurs through diffusion in Chlamydomonas (Chien et al., 2017; Pedersen et al., 2006), by transport in C. elegans, and by both diffusion and transport in mammalian cells (Luo et al., 2017; Prevo et al., 2015; Williams et al., 2014).

Although the four parts of the IFT machinery act in functional modules to mediate cargo transport (Lechtreck, 2015), the stoichiometric relationship is not well understood. Structural studies of IFT-A, IFT-B and IFT dynein complexes have provided a deeper understanding of the IFT machinery (Hesketh et al., 2022; Meleppattu et al., 2022; Petriman et al., 2022; Toropova et al., 2017); however, these do not provide information on the stoichiometry of the whole IFT machinery. Elucidation of IFT trains by cryo-electron tomography (cryo-ET) is poised to address this question. Currently, the structures of the anterograde IFT trains have been resolved by cryo-ET and the stoichiometry for IFT-B:IFT-A:IFT dynein has been estimated to be 8:4:2 (Jordan et al., 2018). One limitation for cryo-ET technique is that it can only resolve structurally uniform IFT trains. The revealed stoichiometry of IFT machinery might not represent the situation in a whole cilium. In addition, it has proved not possible to determine the stoichiometric relationship of kinesin-2 in the IFT train for unknown reasons. By analyzing Chlamydomonas cilia via non-quantitative mass-spectrometry, a stoichiometry of 8:4:7:6 for IFT-B:IFT-A:IFT dynein:kinesin-2 has been estimated (Lechtreck et al., 2009), which differs from the stoichiometry as revealed by cryo-ET (Jordan et al., 2018). Another fascinating question is how many IFT proteins are present in a cilium and whether that is sufficient to account for the known rate of ciliary assembly. To address these questions, we performed absolute quantification of IFT proteins in Chlamydomonas cilia by using quantification concatamers (QconCATs) combined with liquid chromatography mass spectrometry (LC-MS) (Beynon et al., 2005).

The experimental design for the quantification is outlined in Fig. 1A. To design the IFT QconCAT (hereafter IFT-Q), we selected peptides from Chlamydomonas IFT proteins: two IFT-Q-peptides from each target protein of IFT dynein (Dhc1b and FAP133), kinesin-2 (all three subunits), IFT-A (all six subunits) and IFT-B [12 subunits out of 16 excluding four smaller subunits (IFT20, IFT22, IFT25 and IFT27)] (Fig. 1B). These peptides were selected based on trypsin digestion and detection from cilia by LC-MS in a pilot experiment and other criteria, such as the presence of residues that can be labeled and the absence of cysteine and methionine in the peptide (Beynon et al., 2005; Pratt et al., 2006). C-terminally His-tagged IFT-Q was expressed in Escherichia coli in the presence of [¹³C]arginine and [¹³C]lysine to generate stable isotope-labeled quantification standards. The IFT-Q proteins were purified using nickel affinity chromatography to homogeneity (Fig. 1C). The cilia isolated were free of cytoplasmic contaminations (Fig. 1D). Absolute quantification of the target proteins would be achieved by spiking known amounts of the labeled IFT-Q into isolated cilia followed by LC-MS.

For the initial quantification, we mixed different amounts of IFT-Q with constant amount of ciliary proteins to determine whether the measured heavy:light (¹³C:¹²C) ratio of the target peptides showed a linear relationship at different amounts of the added IFT-Q. A total of 0.02, 0.05, 0.1 and 0.2 µg of IFT-Q was added, respectively, to 50 µg ciliary proteins (ratios of 1:2500, 1:1000, 1:500 and 1:250) followed by LC-MS. The ion chromatograms of heavy (¹³C) Q-peptide and light (¹²C) native peptide pairs were extracted, surface areas quantified and ratios calculated (Table S4). For each IFT protein as examined, the heavy:light ratio from at least one of the target peptides showed a linear relationship at different amounts of the added IFT-Q (Fig. S1), which indicates that our assay is well suited for quantifying the selected IFT proteins.

Previous studies have suggested that constraining the standard-to-analyte ratio to within a 10-fold difference would lead to most accurate quantification (Lawless et al., 2016). By examination of our data, we found that the ratios between the IFT-Q and native peptides in the mixture with 0.02 µg IFT-Q and 50 µg cilia were in the range of 3-fold (Table S4). Thus, to improve the rigor in our quantification, we performed further analysis under this condition for LC-MS with three independent preparations of cilia samples (Table S5). The stoichiometry of the subunits within IFT-B, IFT-A and kinesin-2 are close to 1:1, whereas the stoichiometry of IFT dynein heavy chain Dhc1b and the intermediate chain FAP133 is basically 2:1 (Table 1), which are consistent with structural and/or biochemical studies (Hesketh et al., 2022; Meleppattu et al., 2022; Petriman et al., 2022; Toropova et al., 2017).

The absolute quantification of IFT proteins allows us to learn the quantity of each complex, including IFT-B, IFT-A, kinesin-2 and IFT dynein, per cilium (Table 2). The quantity of each complex is derived from averaging all subunits in each complex (Table 1). To support the reliability of our absolute quantification, we compared our quantification data with other published data and found that it is quite consistent with them. Given that Chlamydomonas has a cilium length of 12 µm, an anterograde speed of IFT of 2 µm/s and IFT frequency of 1.22 events/s on average as derived from previous literatures (see Table S1), there will be 7.32 IFT trains in anterograde motion at a given time. As the anterograde IFT trains would have half of the total IFT-B in a cilium [i.e. 378 (756/2)], each IFT train would have ∼52 IFT-B molecules on average (378 IFT-B/7.32 IFT trains). IFT-B repeats every 6 nm along the long axis of the IFT train (Jordan et al., 2018). Thus, the length of the IFT train on average would be 310 nm long (52 IFT-B×6 nm/IFT-B), which is very close to the average train size of 312 nm as determined in cryo-electron tomography (cryo-ET) (Jordan et al., 2018).

Next, we asked whether the calculated amount of IFT proteins in a cilium is sufficient to sustain rapid ciliary growth. Chlamydomonas cilia regenerate with decreasing kinetics, such that regeneration is slower at increasing ciliary length. The fastest growth rate of cilia was found to be 400 nm/min, which occurs at a ciliary length of ∼3 µm (Rosenbaum et al., 1969). For this rapid growth, delivery of 11,650 tubulin dimers per minute to the ciliary tip is required for microtubule polymerization [233 tubulin dimers for an 8 nm long 9+2 cilium, 233 dimers×(400 nm/min/8 nm per tubulin dimer)=11,650 dimers/min]. In a full-length cilium, IFT-B is delivered at a rate of 3806 per minute (1.22 event/s×52 IFT-B per train×60 s=3806). Because axonemal tubulin dimers are delivered by the IFT81 and IFT74 subunits of IFT-B in 1:1 stoichiometry (Bhogaraju et al., 2013; Hao et al., 2011; Kubo et al., 2016). Thus, 3806 tubulin dimers can be delivered per minute in a full-length cilium, which could not meet the requirement of tubulin delivery for rapid ciliary assembly, in which a delivery rate of 11,650 tubulin dimers per minute is required. Thus, how is it possible to explain the rapid ciliary growth based on tubulin delivery and IFT? It has been shown previously that the amount of IFT proteins in a cilium is roughly consistent regardless of ciliary length during ciliary assembly (Marshall et al., 2005). To make this happen, the delivery rate of IFT proteins in a shorter assembling cilium would be expected to be proportionally increased, which is consistent with the increased train size at the early stages of ciliary assembly (Engel et al., 2009; Vannuccini et al., 2016). For an assembling cilium of 3 µm long, the delivery rate of IFT proteins would need to be four times of that of a full-length cilium of 12-µm long. Thus, 15,224 IFT-B or tubulin dimers could be delivered per minute (4×3806 per minute), which is ∼1.3-fold more than the 11,650 tubulin dimers required for rapid ciliary growth. Thus, the amount of IFT machinery in a cilium is sufficient to sustain rapid ciliary assembly.

For the delivery of tubulins into cilia by IFT, our above reasoning is based on IFT81 and IFT74 being the sole tubulin carriers in IFT. Additional IFT-B subunits (IFT54, IFT57 and IFT38) contain potential tubulin-binding calponin homology domains (CH domains) (Bhogaraju et al., 2014; Taschner et al., 2016), but only the CH domain of IFT54 has been verified for its microtubule- or tubulin-binding activity (Ling and Goeddel, 2000). However, deletion of the CH domain of IFT54 does not affect the rate of ciliary growth (Zhu et al., 2017b). We propose that IFT81 and IFT74 contain the only domain for binding tubulins during IFT. Taken together, our quantification data show that the IFT machinery in a cilium is sufficient and highly efficient for sustaining rapid ciliary growth.

Based on our quantification data, the stoichiometry of IFT-B:IFT-A:IFT dynein:kinesin-2 in a cilium was ∼6:4:2:3 (Table 2). A previous proteomic analysis of cilia calculated a stoichiometry of 8:4:7:6, which was simply based on the mean number of unique IFT protein peptides detected per 10 kDa (Lechtreck et al., 2009). In studies of in situ IFT trains, a stoichiometry of 8:4:2 for IFT-B:IFT-A:IFT dynein has been estimated (Jordan et al., 2018). The 2:1 stoichiometry for IFT-A:IFT dynein is consistent with our data (Fig. 2A). IFT-B has a periodicity of 6 nm whereas that for IFT dynein is 18 nm (Jordan et al., 2018). In addition, subtomogram averaging of IFT trains also gave a ratio of 3:1 for IFT-B:IFT dynein (van den Hoek et al., 2022). These results are consistent with our analysis for the stoichiometric ratio of IFT-B:IFT dynein. Thus, we propose that the stoichiometric ratio of IFT-B:IFT dynein is 3:1 (Fig. 2A).

Interestingly, there is a discrepancy in the ratio of IFT-B to IFT-A (2:1 versus 1.5:1) between the cryo-ET data and our analysis. Although we cannot exclude the possibility that our assay has a bias, our results raise a possibility that not every IFT-A in a train registers with two IFT-Bs. First, this is consistent with the 11 and 6 nm periodicities for IFT-A and IFT-B respectively (Jordan et al., 2018). Second, a recent structural study of anterograde IFT trains showed a disparity in the interaction between IFT-A and IFT-B (Lacey et al., 2023). In class A, IFT-A and IFT-B were in sync where two IFT-B register with one IFT-A. While in class B, the interaction between one IFT-A and one IFT-B exists. Thus, one IFT-A might register with one or two IFT-B complexes. The combinations of different IFT-A and IFT-B interactions might result in a ratio for IFT-B:IFT-A in a cilium of between 1 and 2. Cryo-ET would only resolve uniform structures; IFT trains with non-uniform interactions between IFT-A and IFT-B might be averaged out during cryo-ET analysis and hence not be resolved. Because IFT trains are assembled at the ciliary base, it is possible that incompletely assembled trains might accidentally enter cilia (van den Hoek et al., 2022). Thus, in addition to the anterograde IFT trains identified by the current cryo-ET technology, trains with other conformations might exist.

Next we asked whether the plastic interaction between IFT-A and IFT-B could be affected by protein compactness during IFT train assembly. During IFT train assembly at the ciliary base, IFT-A and IFT-dynein are sequentially recruited and assembled with IFT-B (van den Hoek et al., 2022; Wingfield et al., 2017). We reasoned that binding of IFT dynein to IFT-B might influence the interaction between IFT-A and IFT-B in the anterograde IFT train. To this end, we performed quantification of IFT proteins of cilia from IFT dynein heavy chain DHC1b mutants. A dhc1b mutant in Chlamydomonas was generated by gene editing using CRISPR/Cas9 (Fig. S2). Consistent with the results of previous studies (Pazour et al., 1999; Porter et al., 1999), the mutant formed stumpy cilia. As expected through LC-MS analysis, DHC1b was not detected in dhc1 mutant cilia, and IFT-A and IFT-B in mutant cilia were accumulated compared to their levels in wild-type cilia (Tables S2 and S6). Kinesin-2 also increased (∼7-fold), which might indicate that rapid diffusion of kinesin-2 is hindered by the crowded IFT complexes in the mutant cilia. Interestingly, the stoichiometry of IFT-B:IFT-A in the mutant cilia was ∼2:1 (Table 2; Table S2), which contrasts with the 1.5:1 stoichiometry in wild-type cilia. Therefore, we conclude that the stoichiometric interaction of IFT-B and IFT-A is plastic and can be influenced by IFT dynein binding to IFT-B (Fig. 2B). However, this cannot explain the estimated 2:1 ratio for IFT-B:IFT-A in the typical anterograde IFT trains identified by cryo-ET where IFT dynein is present. The result might suggest that other unknown factors are also be involved in IFT-A and IFT-B interactions. Loss of IFT dynein induces ciliary accumulation of IFT complexes. It is intriguing to know whether IFT complexes can form intact trains or exist in dissociated conformation, which could be addressed in the future by cryo-ET.

IFT-B directly interacts with kinesin-2 to undergo anterograde transport (Funabashi et al., 2018; Zhu et al., 2021). Kinesin-2 has been shown to diffuse back after IFT turnaround at the ciliary tip in Chlamydomonas (Chien et al., 2017; Engel et al., 2009). Thus, the amount of kinesin-2 in a cilium includes kinesin-2 moving on anterograde IFT trains as well as that in diffusion. Diffusion of kinesin-2 instead of return with retrograde trains would lead to a ∼4-fold increase in kinesin-2 inside a cilium compared to its normal level (Chien et al., 2017). Thus, the amount of kinesin-2 on the anterograde trains would be about one-eighth of total kinesin-2 inside a cilium. Because the molecules of kinesin-2 per cilium was determined to be ∼350 (Table 2; Table S4), thus, ∼44 kinesin-2 are on anterograde IFT trains in a cilium. Given that half of the IFT-B in a cilium are in anterograde transport (i.e. 378 IFT-B complexes) (Table 2), one kinesin-2 would drive 8.59 IFT-B complexes (∼8 IFT-B complexes) (Fig. 2A).

In summary, we have quantified the exact amounts of IFT proteins inside Chlamydomonas cilia by using quantification concatamers. We have determined the absolute amount of IFT proteins per cilium and the stoichiometry for IFT protein complexes and motors. The absolute quantification of IFT proteins from whole cilia complements previous cryo-ET results and has allowed us to gain a deeper understanding of the assembly of the IFT trains. The average anterograde train size based on the numbers of IFT-B per cilium is consistent with that estimated via cryo-ET analysis (Jordan et al., 2018). Furthermore, the amount of IFT-B per cilium is only ∼23% in excess of that needed to sustain rapid ciliary growth at the initial phase of ciliary assembly, suggesting that the IFT machinery is very suited for the needs. The 1:1 stoichiometry within IFT-B, IFT-A and kinesin-2, and 2:1 for dynein heavy chain to intermediate chain in IFT dynein are consistent with results from structural and biochemical studies. The 3:1 stoichiometry for IFT-B to IFT dynein and 2:1 for IFT-A to IFT dynein are consistent with the cryo-ET data (Jordan et al., 2018; van den Hoek et al., 2022). These results suggest that our quantification assay is faithful. However, the stoichiometry of 3:2 for IFT-B:IFT-A is different from the estimated 2:1 stoichiometry in the anterograde IFT trains as resolved by cryo-ET (Jordan et al., 2018). One possible explanation is that only uniform trains can be resolved structurally. Our results indicate that trains with other conformations might exist. It has been shown that the interaction between IFT-A and IFT-B might have plasticity in that one IFT-A could register with one or two IFT-B complexes (Lacey et al., 2023). The different interactions might be affected by protein compactness during IFT train assembly because we have shown that the stoichiometry for IFT-A:IFT-B is influenced by the presence or absence of IFT dynein. Finally, our work also allows us to determine how many IFT-Bs can be carried by one kinesin-2. Taken together, our data provide a basis for deeper understanding of the IFT machinery and cilia assembly. The stoichiometry for IFT complexes and motors that we have obtained is based on isotope labeling and protein quantification using LC-MS. Errors might come from incomplete protein digestion and different efficiency in detecting the target peptides. The amounts of IFT43 and IFT122 are relatively low compared to other IFT-A subunits, which could reflect a limitation of the method or indicate the presence of incompletely assembled IFT-A complexes in cilia. To reveal interaction patterns among and within complexes of IFT machinery, higher resolution in situ structures of the of all IFT trains present in a cilium are required.

The Chlamydomonas reinhardtii wild-type strain 21gr (mt+; CC-1690) was obtained from the Chlamydomonas Resource Center (University of Minnesota, St. Paul, MN, USA). The dhc1b-sg3 mutant was generated by CRISPR/Cas9 from the wild-type strain following a previous procedure (Wang et al., 2022). Cells were cultured on 1.5% agar M plates or in liquid M medium at 23°C with aeration under a 14-h-light and 10-h-dark cycle as described previously (Sager and Granick, 1953). For transformation, cells were grown in liquid Tris-acetate-phosphate medium under continuous light.

The cDNA sequence for the IFT-QconCAT protein (IFT-Q) was chemically synthesized with codon optimization for gene expression in E. coli followed by cloning into pET-28a using a seamless cloning and assembly kit (Taihe Biotech., China). For stable isotope-labeling of the IFT-Q protein during expression in E. coli rosetta (DE3) (CWBIO, China), cells were cultured in SILAC DMEM (Thermo Fisher Scientific, USA) supplemented with 10 mM [¹³C]L-lysine and [¹³C]L-Arginine (Thermo Fisher Scientific), and 100 µg/ml kanamycin and chloramphenicol (Solarbo, China). Cells were lysed by sonication in lysis buffer (50 mM Tris-HCl, 100 mM NaCl, pH 8.0). The inclusion bodies that contained IFT-Q was washed in lysis buffer containing 1% Triton X-100 followed by resuspension in buffer A (50 mM Tris-HCl, 300 mM NaCl, 8 M urea, pH 8.0). IFT-Q was purified on a Ni-NAT column (Smart Lifesciences, China) and stored after concentration in buffer S (50 mM Tris-HCl, 1% Sodium deoxycholate, pH 8.5) at −20°C (Pratt et al., 2006).

Cilia were isolated after deciliation by pH shock (Craige et al., 2013; Wang et al., 2013). Briefly, cilia detached from the cell body by pH shock were purified by zonal centrifugation on a sucrose density gradient. Isolated cilia were resuspended in buffer S followed by frozen in liquid nitrogen and stored at −80°C.

SDS-PAGE and immunoblotting analysis were essentially performed as described previously (Piao et al., 2009). Cells were dissolved in HMDEK buffer (10 mM Hepes pH 7.2, 5 mM MgSO4, 1 mM DTT, 0.5 mM EDTA, 25 mM KCl) containing protease inhibitor cocktail (complete-mini EDTA free, Roche), 20 µM MG132 and 65 µM MG101 (Selleck, China), boiled in 1× SDS sample buffer for 10 min, separated on SDS-PAGE, transferred to PVDF membranes and probed with the indicated antibodies. Rabbit polyclonal antibody against DHC1b was made by immunizing a bacterial expressed protein fragment (amino acids 700–991) and affinity purified (ABclonal, China). The information for the other primary antibodies is presented in Table S3. Original images of immunoblots can be found in Fig. S3.

Cells were fixed with 1% glutaraldehyde and imaged with a differential interference contrast (DIC) microscope (Zeiss Axio Observer Z1) equipped with a CCD camera (QuantEM512SC, Photometric, USA) using a 100× objective.

The protein concentration was determined by using BCA protein assay kit (Tiangen, China). Ciliary samples with 50 μg of protein were mixed with known amounts of IFT-Q (0.02, 0.05, 0.1 or 0.2 μg). The samples were co-digested with sequencing-grade trypsin and endoproteinase Lys-C (Sigma-Aldrich, USA) followed by salt removal by SDB-RPS and separation by a 120 min gradient elution in a nano-HPLC system, which directly interfaced with a Thermo Orbitrap Fusion Lumos mass spectrometer. The high-pressure LC (HPLC)-targeted peptides were extracted and analyzed using Protein Discoverer 2.3 (Thermo Fisher Scientific).

All the data were independently verified for two or more times. Data were presented as mean±s.d.

We thank discussions with Dr Gaia Pigino during the course of this work.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.262152.reviewer-comments.pdf