Composition and function of the C1b/C1f region in the ciliary central apparatus

Motile cilia are ultrastructurally complex cell organelles with the ability to actively move. The highly conserved central apparatus of motile 9 × 2 + 2 cilia is composed of two microtubules and several large microtubule-bound projections, including the C1b/C1f supercomplex. The composition and function of C1b/C1f subunits has only recently started to emerge. We show that in the model ciliate Tetrahymena thermophila, C1b/C1f contains several evolutionarily conserved proteins: Spef2A, Cfap69, Cfap246/LRGUK, Adgb/androglobin, and a ciliate-specific protein Tt170/TTHERM_00205170. Deletion of genes encoding either Spef2A or Cfap69 led to a loss of the entire C1b projection and resulted in an abnormal vortex motion of cilia. Loss of either Cfap246 or Adgb caused only minor alterations in ciliary motility. Comparative analyses of wild-type and C1b-deficient mutant ciliomes revealed that the levels of subunits forming the adjacent C2b projection but not C1d projection are greatly reduced, indicating that C1b stabilizes C2b. Moreover, the levels of several IFT and BBS proteins, HSP70, and enzymes that catalyze the final steps of the glycolytic pathway: enolase ENO1 and pyruvate kinase PYK1, are also reduced in the C1b-less mutants.


Results
Identification of the proteins positioned in close proximity to Tetrahymena Spef2 ortholog. The genome of Tetrahymena thermophila encodes two proteins with homology to Spef2/CPC1, here named Spef2A (TTHERM_01142770) and Spef2B (TTHERM_00633390). Both proteins were identified in the Tetrahymena ciliome 22 . However, the N-terminal calponin-homology (CH) domain was predicted only in Spef2A (Figs. S1, S2A). Therefore, we assumed that ~ 200 kDa protein, Spef2A, is a true ortholog of mammalian Spef2. When expressed as C-terminal 2V5 or 3HA fusions under control of the native promoter, Spef2A localized in cilia, along their entire length with exception of the distal tip ( Fig. 1A-A").
To identify potential binding partners of Spef2A, we engineered Tetrahymena cells expressing Spef2A-HA-BirA* under control of the native promoter and performed proximity labeling assays to identify proteins that are biotinylated within ~ 10 nm 23 . Mass spectrometry analysis of the biotinylated proteins purified from cilia isolated either from the wild-type (negative control) or Spef2A-HA-BirA* expressing cells (Fig. S3A) revealed that Spef2A, Cfap69, Cfap246, Cfap174, Tt170, and Adgb, a protein with limited similarity to androglobin and calpain-7, were repeatedly recovered from cilia of Spef2A-HA-BirA*-expressing cells (Tables 1, S2). None of Tetrahymena enolases, HSP70s, or putative FAP39 orthologs were detected among the biotinylated proteins.
Immunofluorescence studies of Tetrahymena cells expressing 2V5 fusions of Cfap69, Cfap246, and Adgb (under their native promoter) showed that like Spef2-2V5, these proteins were enriched in cilia except for the distal tip (Fig. 1).
Cfap246 (TTHERM_00188400) is a 53 kDa, leucine-rich repeats-containing protein, with homology to an N-terminal fragment of Chlamydomonas FAP246, and human leucine-rich repeats and guanylate kinase domaincontaining protein (LRGUK) (Fig. S2C). In contrast to the much larger human ortholog (94 kDa), Tetrahymena Cfap246 protein lacks the guanylate kinase domain within its C-terminal region (Fig. S1).
Androglobin/Adgb (TTHERM_00290850) is a large protein (180 kDa) showing similarity to human androglobin and calpain-7 within the N-terminal region and to androglobin within the short C-terminal fragment (Figs. S1, S2E). Orthologs of Adgb are present in most of the animal lineages 24 .
To verify if identified proteins are indeed located in close proximity to each other, we performed reciprocal BioID experiments by expressing Cfap69, Cfap246, and Adgb as C-terminal HA-BirA* fusions under their native promoters and Adgb with N-terminally positioned BirA*-HA (Figs. S3A,B). As expected Cfap69, Cfap246, Adgb, Spef2A, Cfap174, and Tt170 were enriched among the proteins that were biotinylated in cilia of Cfap69-HA-BirA* and Cfap246-HA-BirA* expressing cells (Tables 1, S3-S5). Although Adgb-HA-BirA* and BirA*-HA-Adgb proteins were targeted to cilia, mass spectrometry failed to detect biotinylated proteins, including Adgb, in experimental samples (even after prolonged 16 h incubation in biotin-enriched medium (Fig. S3A)). Therefore, we attempted to identify potential Adgb interacting proteins using immunoprecipitation. Adgb and Cfap69  Cilia of CFAP69 and SPEF2A deletion mutants lack a C1b/C1f projection and exhibit abnormal rotational motion. Next, we engineered Tetrahymena strains with deletions of CFAP69 or ADGB using the germ-line-based targeting by homologous DNA recombination (CFAP69-KO and ADGB-KO cells) 25,26 and SPEF2A-coDel, ADGB-coDel, and CFAP246-coDel mutants using the co-Deletion method based on the induction of scnRNAs 27 (Fig. S4). The targeted loci were analyzed by PCR to confirm complete loss of the targeted sequence. We could not recover strains homozygous for the deletion of CFAP246 (see Material and methods) and therefore we analyzed knockdown of CFAP246-coDel strains. A force generated by cilia beating propels Tetrahymena cells. The CFAP69-KO and SPEF2A-KO cells on average traveled only approximately 41% and 35%, respectively, of the distance of the wild-type cells ( Fig. 2A,B,D,H). ADGB-KO or ADGB-coDel mutants and those with Cfap246 knockdown (CFAP246-coDel) were less affected and swam at the rate of 81% and 74% of the wild type, respectively ( Fig. 2F-H). Reduced swimming rate of CFAP69-KO and SPEF2A-coDel cells, could be caused either by a reduced number of cilia or changes in their length or altered cilia beating. The immunofluorescence analyses of the wild-type and mutant cells using an anti-α-tubulin antibody showed that the density of cilia in the CFAP69-KO and SPEF2A-coDel mutants appeared normal but cilia were approximately 8% (6.05 ± 0.52 µm) and 11% (5.89 ± 0.48 µm), respectively, shorter than those assembled by wild-type cells (6.56 ± 0.5 µm) ( Fig. 2I-L).
Next, we analyzed the motion of cilia using a high-speed video camera (Fig. 3A, Supplementary Movies 1-5). In the wild-type cells cilia beat with two apparent phases, the power and recovery strokes, taking place in different planes. During the power stroke, a tip of the straight cilium follows a semicircle perpendicular to the cell surface. During the recovery stroke, the cilium bends near the base and moves closer to the cell surface to reach the initial pre-power stroke position (Supplementary Movie 1). In contrast to the wild-type cells, cilia of CFAP69-KO and SPEF2A-coDel mutants exhibited a rotatory motion slightly inclined to the cell surface (Fig. 3A). Moreover, the neighboring mutant cilia frequently collided (Supplementary Movies 2, 3). In the mutants deficient in either Cfap246 or Adgb, the power and recovery strokes were well-defined but the amplitude was slightly reduced and the waveform of the cilium during the recovery phase was slightly altered (Supplementary Movies 4, 5, Fig. 3A). Ectopic expression of HA-Cfap69 in the CFAP69-KO background restored cells swimming rate close to the wild-type level (Fig. 2C). Because the coding region of SPEF2A is large, we could not perform a similar rescue experiment for the SPEF2A-coDel mutants. However, we were able to recover cells with a wildtype motility by replacing the deleted region in the SPEF2A-coDel cells with a 3 kb fragment of the wild-type genomic DNA (Fig. 2E). Thus, we conclude that the abnormal ciliary functions in both mutants were due to the deletions at the targeted loci.
In Tetrahymena cilia, similar as in Chlamydomonas flagella, the CA projections differ in size and shape. Ultrastructural analyses of mutant cilia cross-sections using TEM revealed that the C1b projection was shorter Table 2. Mass spectrometry-based identification of the ciliary proteins co-immunoprecipitated with Cfap69-3HA or Adgb-3HA. www.nature.com/scientificreports/ Observed shortening of cilia in both mutants is statistically significant (p < 0.0001, t-test). www.nature.com/scientificreports/ or missing in CFAP69-KO and SPEF2A-coDel cilia (Figs. 3B-D' ,F, S5). Thus, likely both proteins are crucial for either the assembly or stability of the C1b complex. In some ADGB-coDel cilia, C1b seemed smaller or twisted (Figs. 3E-E' , S5). In contrast, CFAP246-coDel cilia did not have obvious defects. In Tetrahymena the C1f projection is not that apparent in TEM cross-sections as in Chlamydomonas flagella cross-sections. However, because in Chlamydomonas cpc1 mutants both C1b and C1f projections are lost, it is possible that also in Tetrahymena the entire C1b/C1f region is lost and therefore we will further refer to CFAP69-KO and SPEF2A-coDel knockouts as C1b/C1f-less mutants.
The levels of putative C1b/C1f subunits, and IFT and BBS proteins are reduced in cilia assembled by CFAP69 and SPEF2A mutants. The C1b is one of the largest projections of the CA microtubules.
Because C1b (and possibly C1f) are either missing or greatly reduced in the Tetrahymena CFAP69-KO and SPEF2A-coDel mutants, most likely not only the targeted protein but also other components of the C1b/C1f are missing. Moreover, C1b projection may stabilize neighboring projection(s) and thus, loss of C1b could destabilize the C2b (as in Chlamydomonas 13,14 ) and perhaps C1d projection.
To identify proteins that are either missing or reduced in C1b/C1f-less cilia, we compared the protein composition of cilia of the wild type, CFAP69-KO, and SPEF2A-coDel mutants by mass spectrometry (six independent samples for each cell type) (Tables 3, S6). MS/MS failed to detect Cfap69 peptides in the CFAP69-KO cilia and detected only a single Spef2A peptide in six samples of the SPEF2A-coDel cilia, arguing that the obtained mutants lack the targeted proteins. The levels of other putative C1b/C1f supercomplex subunits, Adgb and Cfap246 were strongly reduced (Tables 3, S6). Interestingly, the level of Cfap174 was unaltered.
Interestingly, the levels of the proteins that mediate intraciliary transport were either substantially (IFT-A and BBS proteins) or moderately (IFT-B proteins) reduced in the C1b/C1f-deficient cilia (Table 3). These data correlate with the slightly reduced length of cilia in CFAP69-KO and SPEF2A-coDel mutants.
Level of C1b/C1f proteins is mainly regulated within cytoplasm. Lack or reduced levels of C1b/C1f subunits in CFAP69-KO or SPEF2A-coDel cilia (Fig. 4C) can be explained either by their inability to stably dock to the central microtubules in the absence of Spef2A or Cfap69, or by the reduction of the total amounts of C1b/ C1f subunits in mutant cells. Study in Chlamydomonas showed that the level of mRNAs for proteins that form the same ciliary complex can be co-regulated 29 .
We quantified the levels of SPEF2A, CFAP69, CFAP246, and ADGB mRNAs in the wild-type and CFAP69 and SPEF2A knockouts. qRT-PCR revealed that the levels of CFAP246 and ADGB mRNAs were basically unaffected in mutants (Fig. S6A). Similarly, the levels of CFAP69 mRNA in SPEF2A-coDel and SPEF2A mRNA in CFAP69-KO cells were similar to that of the wild type. As expected, the CFAP69 mRNA was undetectable in CFAP69-KO cells. Surprisingly, a prominent amount of the SPEF2A transcript was present in the SPEF2A-coDel cells, suggesting that a transcript was produced at the locus carrying the deletion. Thus, the expression levels of mRNAs for the individual C1b/C1f components are not affected by the losses of other subunits. Therefore, the reduced levels of the C1b/C1f subunits in C1b/C1f-deficient cilia are either due to reduced protein synthesis, decreased transport into the cilia, or increased protein degradation. Therefore, next we assessed the total levels of C1b/C1f proteins in CFAP69-KO and SPEF2A-coDel cells. Because the antibodies against Tetrahymena CA proteins are not available, we engineered strains that express 2V5-tagged fusions of Cfap69, Spef2A, Cfap246, or Adgb each under the control of the respective native promoters. The transgenes were incorporated into the macronuclear genome in wild-type, CFAP69-KO, or SPEF2A-coDel genetic backgrounds.
The macronuclear genome of Tetrahymena contains ~ 45 copies for each protein-coding gene. Initially, the biolistically introduced transgenes incorporate into one to few loci and the ratio of the transgene to endogenous alleles increases with increasing the selection pressure during cell multiplication (so-called phenotypic assortment, see Materials and Methods). Using qPCR and genomic DNA as a template we confirmed that similar number of transgene copies were assorted in all cell strains (Fig. S6B). A western blots analysis of the total cell extracts revealed that the non-targeted C1b/C1f subunits were undetectable or greatly reduced in the C1b/C1f knockouts (Fig. 4A,B). Thus, in the absence of either Cfap69 or Spef2A, other subunits of the C1b/C1f complex may be more prone to proteolytic degradation.
The longevity of a ciliary protein within the cell body may depend upon the presence of partner proteins that stabilize the complex 22 . Therefore, we overexpressed HA-Cfap69 (using cadmium-inducible promoter) 30 in the SPEF2A-coDel cells expressing either Cfap246-2V5 or Adgb-2V5 (under native promoters). The overproduced HA-Cfap69 was targeted to cilia (Fig. 4E,F). Despite this, the Adgb-2V5 was still undetectable while Cfap246-2V5 was stabilized within the cell body ( Fig. 4D) but not targeted to cilia (Fig. 4C). Thus, likely (1) Cfap69 stabilizes Cfap246 but not Adgb and (2) within C1b/C1f complex Cfap69 likely binds to Cfap246.
The expression of Rsp4/6A-GFP or Rsp4/6C-GFP does not rescue CFAP69-KO and SPEF2A-coDel mutant cell motility. The pf6, an immotile Chlamydomonas mutant lacking the C1a projection 9,13 can partly regain motility when one of the radial spoke proteins (RSP3, RSP4 or RSP6) is extended by a C-terminal epitope tag 16 . RSP3 forms a part of the radial spoke stem while the paralogous RSP4 and RSP6 31 are components of the radial spoke head 32 . Their C-termini are located on or above the radial spoke head upper surface that temporarily comes in contact with the CA projection(s) 16 . In contrast, the artificial expression of RSP4 did not rescue Chlamydomonas cpc1 mutant lacking C1b/C1f complex (Table S1 in 16 ).
Next, we investigated if the elevated level of Rsp4/6A-GFP or Rsp4/6C-GFP affects cells motility. The swimming velocity of the wild-type cells grown for 16-18 h in the medium with cadmium (2.5 µg/ml) was reduces by approximately 15% (WT-Cd, 1103 ± 176, n = 51, WT + Cd, 935 ± 176, n = 53). When Rsp4/6A-GFP or Rsp4/6C-GFP were overexpressed (cells were grown in a culture medium with cadmium, Fig. S8A) the swimming velocity of the otherwise wild-type cells was reduced by 27% (Rsp4/6A-GFP) or 31% (Rsp4/6C-GFP) while overexpression of the radial spoke proteins in CFAP69-KO and SPEF2A-coDel mutants did not change or slightly reduced cells motility (Figs. S8B-J, S9). Thus, similar as in cpc1 mutant, the expression of the Rsp4/6-GFP did not improve mutant cells' motility, suggesting a different mechanism of the signal transduction between radial spokes and C1a or C1b projections.  www.nature.com/scientificreports/ taining the CH domain) in the SPEF2A-coDel background, both under MTT1 promoter. Western blots showed that when overexpressed both truncated proteins were present in cilia (Fig. 5B). Interestingly, expression of HA-Cfap69-M1-R248 partially rescued the slow motility phenotype of the CFAP69-KO cells, on average to the level of 80% of the wild-type (Fig. 5C). Spef2A truncation did not have such a rescuing effect (Fig. 5C). We conclude that disease-related truncated variants of Cfap69 and Spef2A can be targeted to cilia and thus the targeting determinants are present within the remaining N-terminal domains of these proteins. Furthermore, among the variants that cause MMAF, Gln255X Cfap69 could be partially functional while the Arg304X Spef2 is likely severely functionally compromised. www.nature.com/scientificreports/

Discussion
In contrast to outer doublet components, the composition of the CA is poorly characterized. Surprisingly, although the overall CA architecture is similar in cilia/flagella of diverse species 3 , approximately 60% of the identified Chlamydomonas CA proteins lack obvious orthologs in other ciliated eukaryotes including humans 19,20 . These observations raise two questions. First, what is the protein composition of the CA in cilia and flagella assembled by other species and specifically, are there additional conserved components in other lineages? Second, since patterns of cilia/flagella beating vary in different species, do lineage-specific CA components contribute to the regulation of the beating patterns?
We have shown that evolutionarily conserved Spef2A, Cfap69, Cfap246, Adgb, and Tetrahymena-specific Tt170 are likely structural components of the C1b/C1f supercomplex in Tetrahymena and that lack of the entire C1b/C1f complex changes cilium motion from two-phase to rotatory. Besides Tt170 and Adgb, orthologs of other putative C1b/C1f proteins were also suggested to build C1b/C1f structure in Chlamydomonas 13,14,19,20 . Information concerning localization and function of the novel putative C1b/C1f components are limited. Some data link these proteins with cilia/flagella in mammals. In human and mice sperm flagella, CFAP69 is present in the midpiece and its truncation results in either mislocalization or loss of SPEF2 35 . This agrees with our data showing that Spef2A is reduced in the CFAP69-KO cilia. Mutations in CFAP69 cause male infertility 35,37 but the connection between CFAP69 and ultrastructural changes in sperm cells is unclear.
Interestingly, in mice, CFAP69 is also present in the immotile olfactory cilia of the olfactory sensory neurons (OSN) 38 . The OSN cilia can be divided into two segments, proximal containing a CA (9 × 2 + 2) and distal with a decreasing number of microtubules (from 9 × 1 to 4 × 1) 39 . It will be of interest to determine whether in the OSN cilia CFAP69 is present only in the CA-containing proximal fragment or along the entire cilia length, suggesting another intraciliary localization and likely function in these sensory cilia.
Cfap246 orthologs contain LRR domains in their N-termini. The LRR domains form a horseshoe shape which provides a scaffold for protein-protein interactions 40 . In contrast to Tetrahymena Cfap246, Chlamydomonas FAP246/CHLRE_14g618750v5 and the mammalian LRGUK are much larger proteins. LRGUK-1, besides LRR domains, also has a guanylate kinase domain while predicted FAP246/ CHLRE_14g618750v5 contains C-terminal EF-hand domains. Thus, Cfap246 orthologs could mediate protein-protein interactions (LRR) and have lineagespecific functions that involve Ca 2+ signaling and local regulation of GMP, and indirectly, cGMP.
In humans, LRGUK is highly enriched in the trachea, testis (GDS3113/190,191; https:// www. ncbi. nlm. nih. gov/ geopr ofiles) and spermatozoa and in mouse Lrguk-1 localizes to the acrosome and the sperm tail 41,42 . LRGUK-1 mutant mice either lack the sperm tail or assemble a short one 41 . It remains to be determined if LRGUK is indeed a C1b/C1f component in mammals.
In Chlamydomonas, FAP174 binds to an unidentified protein tentatively named AKAP240 (A-kinase anchoring protein) 43 suggested to be a component of the C2 region of the CA 44 . On the other hand, FAP174 co-immunoprecipitates with FAP246 19 . This latter result agrees with our data showing that Cfap174 is positioned near Cfap246. Surprisingly, the level of Cfap174 was unaltered in cilia of Tetrahymena SPEF2A-coDel or CFAP69-KO mutants but slightly reduced in Chlamydomonas cpc1 mutant 20 . Considering that C2b projection is partly or entirely missing in flagella of the Chlamydomonas cpc1 mutant, we speculate that FAP174/Cfap174 is positioned at the very distal end of C1b or C2b projection. This model is supported by following data: (1) Cfap174 coimmunoprecipitates with Adgb but not with Cfap69 which together with Spef2A likely docks the projection to the C1 microtubule, (2) the distal part of C1b but not the entire projection is altered in the Adgb knockdown cells, suggesting that Adgb forms a distal part of C1b, (3) Cfap174 is highly biotinylated in cells expressing Cfap246-HA-BirA* but not in cells expressing Spef2A-HA-BirA*, and thus likely positioned in close proximity to Cfap246. Alternatively, Cfap174 may have multiple axonemal docking sites as it was suggested for Chlamydomonas 20 .
Orthologs of Chlamydomonas FAP42 were found only in unicellular green algae. A detailed inspection of the FAP42 and Adgb sequences shows that both proteins contain a calpain-like domain near their N-termini. Thus, we speculate that the Adgb and FAP42 have similar position within C1b/C1f supercomplex. Importantly, in mammals androglobin is expressed at higher level in cells assembling motile cilia and its expression is coregulated by FOXJ1 24,45 .
In contrast to what has been observed in Chlamydomonas 21 , in Tetrahymena neither enolases nor Hsp70 were detected among proteins associated with C1b/C1f subunits (did not specifically co-immunoprecipitated with C1b/C1f subunits or were biotinylated in cells expressing BirA*-tagged C1b/C1f proteins, Tables S2-5). The genome of Tetrahymena encodes several proteins with similarity to Hsp70 and three enolases. Interestingly, the level of Hsp70 encoded by TTHERM_00105110 and Eno1 was reduced in FAP69-KO and SPEF2A-coDel mutants lacking C1b/C1f, suggesting that structural C1b/C1f proteins could transiently serve as a scaffold for Hsp70 and Eno1 docking or that Hsp70 and Eno1 loosely attach to the C1b/C1f projections. Enolases catalyze the penultimate step of glycolysis i.e., the conversion of 2-phosphoglycerate to phosphoenolpyruvate. Interestingly, in the C1b/C1f-deficient cells also the level of the pyruvate kinase, an enzyme that catalyzes a transfer of the phosphate group from phosphoenolpyruvate to ADP yielding synthesis of ATP, is significantly reduced. Thus, likely as was earlier shown in Chlamydomonas 21 , C1b/C1f function as a scaffold for enzymes that locally regulate ATP level.
The predicted molecular mass of FAP39/CHLRE_02g145100v5, a P-type ATPase, is 127 kDa that is close to the weight of the unidentified protein that co-sediments with CPC1/SPEF2 13 . The Tetrahymena genome encodes several P-type ATPases and three of them were found in the Tetrahymena ciliary proteome (Table S6). One of these three proteins (TTHERM_00522430) was reduced in the C1b/C1f-deficient cilia (Tables S6). However, we did not identify FAP39 orthologs in BioID and co-IP assays. Moreover, both Chlamydomonas FAP39/ CHLRE_02g145100v5 and three ciliary Tetrahymena Cfap39 orthologs likely contain transmembrane domains (as predicted http:// www. cbs. dtu. dk/ servi ces/ TMHMM/) that makes their presence in the CA structure unlikely. www.nature.com/scientificreports/ The bioinformatics search using Tetrahymena and human genome databases did not reveal any orthologs of Chlamydomonas CHLREDRAFT_177061 or CHLREDRAFT_170023. The presence of the WD40 repeats in the latter rendered limited similarity to WD40-containing protein, FAP57. However, the Tetrahymena protein identified with the highest score in the blastp search was not detected as biotinylated in Spef2A-, Cfap69-or Cfap246-HA-BirA* expressing cells (Tables S2-5).
To sum up, we propose that Spef2A, Cfap69, Cfap246, Adgb, and Tt170 form a scaffold of the Tetrahymena C1b/C1f projections, Cfap174 is positioned at the distal end of C1b or C2b, while the enolase Eno1 and Hsp70, are either loosely or transiently attached to C1b/C1f projections.
Tetrahymena SPEF2A-coDel and CFAP69-KO cells assemble cilia that are slightly shorter than that in wildtype cells. Interestingly, the levels of IFT and BBS proteins are reduced in those mutants. The BBSome interacts with IFT and mediate transport of the ciliary membrane proteins 46 . The lower level of intraflagellar transport proteins could account for the reduced length of mutant cilia.
The levels of IFT and BBS proteins were not investigated in Chlamydomonas cpc1 mutant but the length of mutant flagella was similar as in wild types 13,14 . Some Chlamydomonas CA mutants assemble short flagella. However, in the CP-less short flagella assembled by Chlamydomonas pf15 (katanin p80), pf18 or pf19 (katanin p60) mutants, the levels of the IFT and BBS proteins was significantly 19,47 or slightly 20 elevated compared to wild-type flagella, and IFT proteins and BBS4 were trapped in the lumen of the CA-less, detergent extracted axonemes 47 . The amount of the trapped IFT proteins was reduced when the CA re-assembled 47 . Thus, the reduced level of IFT and BBs proteins in C1b/C1f-less Tetrahymena cilia is an unexpected result. It would be interesting to investigate if the lack of other CA projections also affects the level of IFT and BBS proteins and cilia length in Tetrahymena.
Interestingly, in mice, the IFT20 was identified as a Spef2 partner protein in maturing sperm cells 48 . In mammals, the differentiation of spermatids requires an intense transport of cargoes along the manchette microtubules and this process involves intraflagellar transport proteins 49 . However, how Spef2 is related to intraflagellar and IFT-related inter-manchette transport is to be determined.

Materials and methods
Tetrahymena strains and culture. The wild-type CU428.2 and B2086.2 Tetrahymena strains were obtained from the Tetrahymena Stock Center (Cornell University, Ithaca, NY, USA). Wild-type and other motile strains were grown in the SPP (Super Proteose Peptone) medium 50 with the antibiotic-antimycotic mix at 1:100 (Sigma-Aldrich, St-Louis, MO, USA). Mutants with greatly reduced motility were grown in MEPP (Modified Enriched Proteose Peptone) medium, on which cells take up nutrients without using oral cilia 51 , supplemented with the antibiotic-antimycotic mix at 1:30 (Sigma-Aldrich, St-Louis, MO, USA). Cells were grown with moderate shaking (80 rpm) at 30 °C. Cell swimming and cilia beating patterns were analyzed as previously described 52,53 . To induce protein overexpression, cells were cultured in media with 2.5 µg/ml CdCl 2 for 16-18 h.
Expression of tagged proteins, gene knockout, and rescue. All DNA fragments were amplified using Phusion HSII High Fidelity Polymerase (Thermo Fisher Scientific Baltics, Lithuania) and genomic DNA purified from CU428.2. The primers used are listed in Table S1. To express proteins with C-terminal -3HA, -2V5, or -HA-BirA* tags, native loci were modified by DNA homologous recombination using plasmids with fragments of the coding region and the 3'UTR obtained by modifications of pFAP44-3HA-neo4, pFAP44-2V5-pPur, or pFAP44-HA-BirA*-neo4 plasmids 53 . MluI and BamHI restriction endonucleases were used to insert a fragment of the coding region and PstI and XhoI were used to clone a fragment of 3'UTR (Table S1). For genomic biolistic transformation, the targeting plasmids were digested with MluI and XhoI to separate the transgenes from the plasmid backbones.
To express Adgb with an N-terminal BirA*-HA in the native locus, the 5'UTR and the coding region of FAP44 were removed from the Neo2-3HA-FAP44 plasmid 53 and replaced by approximately 1 kb fragments of the 5'UTR and a coding region starting with the ATG codon of the ADGB gene, both amplified by PCR with primers listed in Table S1. The 3HA tag was replaced by BirA*-HA sequence. The resulting transgene carries the MTT1 promoter and the neo2 selectable cassette 54 . The plasmid was digested with SacII and BamHI to separate the transgene from the plasmid backbones prior to biolistic bombardment.
To overexpress proteins as fusions with C-terminal -HA or -GFP tag, the entire coding region was amplified by PCR with the addition of MluI and BamHI sites at 5' and 3' ends, respectively (Table S1) and cloned into pMTT1-GFP 55 or pMTT1-HA 56 plasmids enabling the integration of the transgene into BTU1 locus and overexpression controlled by the Cd-inducible MTT1 promoter 30 . To select transformed Tetrahymena cells, based on paromomycin or blasticidin resistance, a neo2 54 or bsr cassette 57 was inserted between 5'BTU1 and MTT1 promoter. Plasmids were digested with SacII and ApaI restriction endonucleases to separate a transgene from the plasmid backbone.
Approximately 10 µg of the digested plasmid DNA was precipitated onto DNAdel Gold Carrier Particles (Seashell Technology, La Jolla, CA, USA) according to the manufacturer's instructions and used to biolistically transform CU428.2 cells. Positive clones were selected for 3-4 days at 30 °C on SPP supplied with (depending upon the introduced transgene) 100 µg/ml paromomycin (Sigma-Aldrich, St-Louis, MO, USA) (transgenes with neo2 cassette 54 ), 100 µg/ml paromomycin and 1.5 µg/ml CdCl 2 (transgenes with neo4 cassette 58 ), 200 μg/ ml puromycin (BioShop Canada Inc., Canada) and 1.5 µg/ml CdCl 2 (transgenes with pPur cassette 59 ), or 60 μg/ ml blasticidin (BioShop Canada Inc., Canada) (transgenes with bsr cassette 57 ). The positive clones were grown in medium with decreasing concentrations of CdCl 2 (to 0.05-0.1 μg/ml) and either an increasing concentration of paromomycin (up to 1 mg/ml) or blasticidin (up to 100 μg/ml), or a constant concentration of puromycin to promote phenotypic assortment. www.nature.com/scientificreports/ The Tetrahymena knockout cells were obtained either by the germline gene disruption approach 25,26 or by the coDeletion approach 27 . Primers used to amplify fragments of the targeted genes are listed in Table S1. In the case of CFAP246-coDel and ADGB-coDel we were unable to obtain mutant cells with all copies of the targeted gene disrupted in the macronuclear genome (all analyzed clones had gene knockdown). When we attempted to engineer germ-line CFAP246 knockout cells, only few paromomycin-resistant clones were found among transformed exconjugants (all 6MP-sensitive), suggesting a deletion of the introduced transgene during the rearrangement of the genome of new macronuclei. Therefore, in the case of the CFAP246 gene we could investigate only the effect of CFAP246 knockdown. At least two independent clones of the coDel and germ-line knockout strains were obtained.
To rescue Tetrahymena knockout cells, CFAP69-KO cells were transformed with a transgene enabling expression of the HA-Cfap69 from the BTU1 locus, under the control of the MTT1 promoter (cells were grown without CdCl 2 ). Mutants obtained by the coDeletion approach were rescued with the approximately 3 kb fragment of the genomic DNA, encompassing the deleted region and about 1 kb upstream and downstream of the deletion. The rescued cells were selected based on the restored wild-type cell motility.
Quantitative real-time PCR and RT-PCR. A real-time PCR was carried out using genomic DNA or cDNA as a template and a PowerUp SYBR Green Master Mix (Thermo Fisher Scientific Baltics, Vilnius, Lithuania) with Standard cycling protocol according to the manufacturer instruction in StepOnePlus Real-Time PCR System (AB Applied Biosystems, Foster City, CA, USA). A genomic DNA was purified with Tissue DNA Purification Kit (EURX, Gdansk, Poland) using cell culture protocol provided by the manufacturer. Total RNA was isolated with Universal RNA Purification Kit (EURX, Gdansk, Poland) with the on-column DNAse digestion protocol provided by the manufacturer. Approximately 500 ng of purified RNA was subjected to reverse transcription with SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (Thermo Fisher Scientific Baltics, Vilnius, Lithuania) and oligo dT, according to manufacturer instruction. To compare the levels of PCR products between housekeeping and experimental genes, for each gene a standard curve, with the use of know amounts of DNA (either plasmid or PCR product) was generated. For each sample, the initial amount of the genomic DNA or cDNA was estimated using Step One Plus Software.
Morphological and physiological tests. To evaluate cilia length 60 cells were stained with anti-α-tubulin 12G10 antibodies, and confocal images were recorded with a 0.32-µm distance between z-sections. A length of cilia was measured on merges of two to four z-sections using ImageJ bundled with 64-bit Java 1.8.0_172. The cell swimming paths and cilia beating were recorded as described 52,53 . Immunofluorescence and transmission electron microscopy. For in situ protein localization, Tetrahymena cells were fixed and stained on coverslips as previously described 52,61 . The primary antibodies were used at the following final concentrations: monoclonal mouse anti-HA antibodies (Covance, Berkeley, CA, USA) 1:300, monoclonal rabbit anti-HA antibodies (Cell Signaling Technology, Danvers, MA, USA) 1:300, monoclonal rabbit anti-V5 antibodies (Cell Signaling Technology, Danvers, MA, USA) 1:1600, polyclonal rabbit anti-GFP antibodies (Abcam, Cambridge, UK) 1:6000, anti-α-tubulin 12G10 antibodies (Developmental Studies Hybridoma Bank, Iowa University, Iowa City, IA, USA) 1:300, and the secondary antibodies, anti-mouse or anti-rabbit IgG conjugated either with Alexa-488 or Alexa-555 antibodies (Invitrogen, Eugene, OR, USA), all in concentration of 1:300. After washing, the coverslips were mounted in Fluoromount-G (Southern Biotech, Birmingham, AL, USA) and viewed using a Zeiss LSM780 (Carl Zeiss Jena, Germany) or a Leica TCS SP8 (Leica Microsystems, Wetzlar, Germany) confocal microscope.
To analyze ciliary ultrastructure cells were fixed as described 52 and samples were viewed using a JEM 1400 transmission electron microscope (JEOL Co, Tokyo, Japan).
For analyses of the biotinylated ciliary protein, samples were run on the 10% SDS-PAGE gel, transferred onto nitrocellulose membrane, and blocked overnight with 3% BSA in the TBST at 4 °C. Next, the nitrocellulose membranes were incubated for 4 h at RT with the streptavidin-HRP (Thermo Fisher Scientific, Rockford, IL, USA) diluted 1:40 000 in the blocking solution, washed and biotinylated proteins were detected using Westar Supernova kit (Cyanagen, Italy).
Total cilia proteome analyses. Approximately 5 × 10 7 wild-type (control) or mutant cells from logarithmic cell culture were deciliated 62 . Cilia were lysed with 7 M urea, 2% CHAPS, 40 mM Tris-HCl, pH 7.4 and 100 µg of protein per sample were prepared according to FASP protocol with minor modifications 64 . In brief, the protein sample was diluted up to 200 µl with urea buffer (8 M urea in 50 mM ammonium bicarbonate) and incubated for an hour with 50 mM TCEP at 60 °C to reduce cysteine residues. After ultrafiltration onto 10 kDa molecular weight cut-off ultrafiltration units (Vivacon, Sartorius Stedim, Goettingen, Germany), samples were washed twice with urea buffer and incubated with 50 mM IAA for 30 min in dark at RT. Modified proteins were washed three times with 8 M urea buffer followed by three rinses with 50 mM ammonium bicarbonate to remove denaturant. Each time the filters were centrifuged at 14,000xg at for 15 to 30 min until the membrane was dry. Digestion was carried out overnight using trypsin/LysC mix (Promega, Madison, WI, USA) in a 1:25 enzymeto-protein ratio at 37 °C on vortex. Peptides were collected from filters by centrifugation and two additional washes with 0.5 M NaCl and 50 mM ammonium bicarbonate, respectively. Combined eluates were acidified with trifluoroacetic acid (TFA) and vacuum-dried. Peptides were reconstituted in 0.1% TFA with 2% ACN and subjected to LC-MS/MS analysis.
Mass spectrometry. Three µg of peptides from each total cilia proteome sample and one third of proximity labeling assay and immunoprecipitation samples were analysed on an LC-MS system composed of an UPLC chromatograph (nanoAcquity, Waters, Milford, MA, USA) directly coupled to a Q Exactive or Elite mass spectrometer (Thermo Scientific, Rockford, IL, USA). Data acquisition for analysis of total cilia proteome was solely performed on Q Exactive system. Peptides were trapped on C18 pre-column (180 µm × 20 mm ,Waters, Milford, MA, USA) using water containing 0.1% FA as a mobile phase and then transferred to a nanoAcquity BEH C18 column (75 µm × 250 mm, 1.7 µm, Waters, Milford, MA, USA) using acetonitrile gradient (0-35% ACN in 160 min) in the presence of 0.1% formic acid at a flow rate of 250 nl/min. Data acquisition was carried out using a data-dependent method with top 12 precursors selected for MS2 analysis after collisional induced fragmentation (CID) with an NCE of 27. Full MS scans covering the mass range of 300-2000 were acquired at a resolution of 70,000 with a maximum injection time of 60 ms and an automatic gain control (AGC) target value of 1e6. MS2 scans were acquired with a maximum injection time of 60 ms and an AGC target value of 5e5 with an isolation window of 3.0 m/z. Dynamic exclusion was set to 30 s. www.nature.com/scientificreports/ of the peptide and fragment were established separately for individual LC-MS/MS runs 65 , resulting in 5 ppm (for parent) and 0.01 Da (for fragment ions) values. Data were searched with automatic decoy option. The statistical significance of peptide identifications was estimated using a joined target/decoy database search approach. This procedure provided q-value estimates for each peptide spectrum match (PSM) in the dataset. All PSMs with q-values > 0.01 were removed from further analysis. A protein was regarded as confidently identified if at least two of its peptides were found. Proteins identified by a subset of peptides from another protein were excluded from analysis. The mass calibration and data filtering were carried out with MScan software, developed in-house (http:// prote om. ibb. waw. pl/ mscan/).

Qualitative
Quantitative MS data processing. The lists of peptides that matched the acceptance criteria from the LC-MS/MS runs were merged into one common list and overlaid onto 2-D heat maps generated from the LC-MS raw files A more detailed description of the quantitative extraction procedure implemented by our inhouse software is available in 66 . The abundance of each peptide was determined as the height of a 2-D fit to the monoisotopic peak of the tagged isotopic envelope. Quantitative values were next exported into text files, along with peptide/protein identifications, for Diffprot softwere for non-parametric statistical analysis of differential proteomics data 65 Diffprot was run with the following parameters: number of random peptide sets = 10 6 ; clustering of peptide sets-only when 90% identical; normalization by LOWESS. Only proteins with q-value below 0.05 or those present in only one of two compared analytical groups were taken into consideration during further analysis.
Phylogenetic analysis. The orthologs of Spef2, Cfap69, Cfap246, Cfap174, and Adgb were identified in the NCBI database using Blastp search and Chlamydomonas or human proteins as baits. The sequences of Tetrahymena orthologs were obtained from Tetrahymena Genome Database (TGD, http:// cilia te. org). The protein amino acid sequences were aligned using the ClustalX2 program 67 , edited using the SeaView program 68 , and the similar/identical amino acid residues in the multiple sequence alignments were highlighted using the GeneDoc program 69 . The domain analyses were performed using InterPro (https:// www. ebi. ac. uk/ inter pro/) 70,71 .

Data availability
Data generated or analyzed during this study are included in this published article (and its Supplementary Information files).