Introduction

Motile cilia and homologous flagella are highly evolutionarily conserved, hair-like cell protrusions, supported by a microtubular skeleton, the axoneme, composed of nine peripheral doublets and two central singlets (9 x 2 + 2). The central microtubules, C1 and C2, together with attached complexes, the projections, form the central apparatus (CA), the structure believed to initiate signals required for planar cilia beating. Transient interactions between CA projections and outer doublet-docked radial spokes (RSs) enable transmission of regulatory signals to outer doublet complexes, including the nexin-dynein regulatory complex (N-DRC) and the motor protein-containing outer and inner dynein arms (ODAs and IDAs) (Grossman-Haham et al., 2021; Oda et al., 2014; Smith & Yang, 2004).

Radial spokes are T-shaped, approximately 40-nm long complexes with three morphologically distinct regions: an elongated elastic stalk docked to the axoneme, an orthogonal head, and a neck connecting the stalk and the head. Within the axonemal repeat unit, radial spokes are arranged as triplets (RS1, RS2, RS3) within which the distance between the neighboring spokes is specific to each spoke pair, enabling their identification (Goodenough & Heuser, 1985). The architectures of RS1, RS2, and RS3 are not identical (Barber et al., 2012; Lin et al., 2012, 2014; Pigino et al., 2011; Yoke et al., 2020). The most striking example of the RS morphological diversity is a Chlamydomonas short, knob-like RS3, unfit to directly interact with the CA (Lin et al., 2012; Pigino et al., 2011).

Pioneering studies using the Chlamydomonas pf14 mutant carrying a mutation in the RSP3 gene (Williams et al., 1989) and lacking full-length spokes, RS1 and RS2, led to the identification of radial spoke subunits, RSP1-RSP23 (Yang et al., 2001, 2006). The list of Chlamydomonas RSPs was recently extended and most of the RSPs were assigned to RS1 and/or RS2 spokes (Gui et al., 2021). The RS1 protein composition was also resolved for the murine sperm axoneme (Zhang et al., 2022). However, despite the extensive research conducted using diverse ciliated species, including Chlamydomonas, Tetrahymena (Urbanska et al., 2015; Vasudevan et al., 2015), Ciona intestinalis (Padma et al., 2003; Satouh et al., 2005; Satouh & Inaba, 2009), and mice (Guan et al., 2010; Zhang et al., 2021, 2022), the RS protein composition is still not fully solved, especially that of RS3. The knob-like RS3 is present in the pf14 Chlamydomonas mutant (Lin et al., 2012; Pigino et al., 2011) and thus, its protein composition was not resolved in earlier studies. In organisms where RS3 is a full-length spoke, its architecture differs from that of RS1 and RS2 (Chen et al., 2023; Lin et al., 2012; Pigino et al., 2011). To date, CFAP61 and CFAP251 are the only known RS3 components, building a part of the stalk and an arch-like structure at the RS3 base, respectively (Urbanska et al., 2015). This raises a question about the protein composition of this spoke.

The identification of RSPs and RSP-interacting proteins is crucial to understand the mechanism(s) of signal transduction from the CA to dynein arms. Here, we used the existing Tetrahymena knockout mutants with RS(s) defects (CFAP61-KO (Urbanska et al., 2015), CFAP206-KO (Vasudevan et al., 2015), and CFAP91-KO) (Bicka et al., 2022), and newly engineered mutants lacking RSP3 paralogs to decipher the protein composition of each RS in the triplet by label-free and tandem mass tag (TMT) quantitative mass spectrometry of wild-type and mutant ciliomes. Next, using co-immunoprecipitations and BioID, we verified that the identified proteins are either the subunits of RS complexes or are present in their vicinity. These proteomic approaches combined with cryo-electron tomography (cryo-ET) and subtomogram averaging studies of the ultrastructural alterations in the RS mutants allowed us to assign with high probability the substantial number of known RSPs and newly identified candidate proteins to either RS1, RS2, or RS3. Importantly, among the newly identified proteins, we found not only structural but also enzymatic proteins, including serine-threonine kinases, GMP recycling-related enzymes, adenylate kinases that dock to specific RSs, and an arginine kinase that locally controls the ATP level. We also noted a destabilization of some single-headed inner dynein arms in the analyzed RSP mutants.

As a result, we propose a model of the protein composition of Tetrahymena RSs showing both ubiquitous and RS-specific RSPs, new candidates of RS structural proteins, and RS- associated enzymes. The presence of spoke-specific RSPs agrees with structural differences among the individual RSs detected using cryo-ET (Lin et al., 2012; Pigino et al., 2011). Moreover, we found that in Tetrahymena, there are sub-types of RS1 and RS2 spokes, which further increases the heterogeneity of radial spokes. Together, obtained data contribute to a better understanding of the molecular mechanisms regulating planar cilia motion.

Results

Deletion of different RSP3 paralogs causes heterogeneous phenotypic and ultrastructural defects

The RSP3 dimer is the main component of RSs (Wirschell et al., 2008). It stretches from the RS base (N-terminal end) to the RS head (C-terminal end) and directly interacts with numerous RSPs (Gui et al., 2021; Oda et al., 2014). Comparative analyses of flagellar proteins purified from the wild-type and pf14, an RS1-RS2 spokes-less Chlamydomonas mutant (Yang et al., 2006) carrying a mutation in the RSP3 gene (Williams et al., 1989) and cryo-EM studies combined with proteomics, and molecular modeling (Gui et al., 2021) yielded a model of RS1- RS2 protein composition in this green alga. We used a similar approach to decipher the protein composition of the Tetrahymena RSs. In contrast to Chlamydomonas, which has a single RSP3 gene, the Tetrahymena genome encodes three RSP3 proteins: RSP3A, RSP3B, and RSP3C (Figure 1A). RSP3C is unusual in having an N-terminal ARF (ADP-ribosylation factor) domain and a C-terminal region enriched in glutamic acids. Such unusual RSP3 orthologs were also found in the proteomes of Paramecium, Ichthyophthirius, and Pseudocohnilembus belonging to Oligohymenophorea.

Figure 1.

Tetrahymena RSP3 proteins expressed as C-terminal 3xHA-tagged fusions under the control of native transcriptional promoters localize uniformly along the entire cilia length (Figure 1B-E). To explore whether the distribution of the three RSP3 paralogs differs between RS1, RS2, and RS3, we constructed knockout strains lacking single paralogs (Figure 2-figure supplement 1) using a germ-line approach (Cassidy-Hanley et al., 1997; Dave et al., 2009). Compared to wild- type cells, RSP3 knockout cells swam significantly more slowly (Figure 2A-E). The loss of RSP3B had the most profound effect and reduced the swimming rate to ∼12% of the wild-type rate while deletions of either RSP3A or RSP3C were less damaging (70% and 56% of the wild- type rate, respectively).

Figure 2.

Tetrahymena cells are propelled by the forces generated by cilia beating as metachronal waves. The cilium beating cycle has two phases, the power and recovery strokes that occur in different planes in relation to the cell surface, perpendicular and parallel, respectively. During the power stroke the cilium is straight while during the recovery stroke it bends, and the bend position shifts from the cilium base to the tip as the recovery stroke progresses (Figure 2F) (Bazan et al., 2021; Bicka et al., 2022). Deletion of RSP3A did not apparently affect the ciliary waveform, beat amplitude (Figure 2F), and cilia metachrony (Figure 2-figure supplement 2). In contrast, cilia lacking RSP3B were beating in a less coordinated manner, with inconsistencies in the waveform and amplitude between neighboring cilia and between subsequent beating cycles performed by the same cilium (Figure 2F). Moreover, in the majority of RSP3B-KO cilia their distal part remained bent during the power stroke (Figure 2F). A delayed straightening of the distal end during the power stroke was also observed in cells lacking RSP3C. However, in contrast to RSP3B-KO mutant, in RSP3C knockout, cilia beat with metachrony and beating amplitude was only slightly reduced (Figure 2F and Figure 2-figure supplement 2).

Compared to cilia in wild-type cells, which beat with a frequency of 38.6+/-4.7 Hz, in RSP3A and RSP3C knockouts, the ciliary beating frequency was mildly reduced (on average 31.2+/-8.2 Hz and 31.9+/-5.7 Hz, respectively). In contrast, the loss of RSP3B reduced ciliary beating frequency nearly by half (25.6+/-5.8 Hz) (Figure 2G-H).

The differences in the RSP3 mutant phenotypes suggest that lack of the particular RSP3 paralogs likely results in different ultrastructural alterations. Indeed, cryo-ET and subtomogram averaging analyses of RSP knockout cilia revealed paralog-specific ultrastructural defects (Figure 3, Figure 3-figure supplement 1, and Figure 3-figure supplement 2). The RSP3A-KO mutant cilia frequently (∼35% of the 96-nm axonemal units) lacked either the entire RS1 (∼30%, n=639 units) or RS1 except for the RS1 base (∼67%, n=1417) (Figure 3 and Figure 3-figure supplement 1). The RSP3B-KO cilia lacked the entire RS2 structure in all analyzed 96-nm axonemal repeats. Moreover, 77.5% of the axonemal repeats (n=1622) showed RS1 spoke defects and lacked either the entire RS1 (26%) or, more frequently, its distal fragment (51.5%) (Figure 3 and Figure 3-figure supplement 1). Surprisingly, RS2 was also impaired in the majority of RSP3C-KO axonemal repeats, where it was either missing (20%) or only the RS2 base was well visible (62%). Thus, the RS2 head visible in the average image (Figure 3A) is due to the presence of intact RS2 spokes in 18% of the analyzed axonemal repeats (Figure 3 and Figure 3-figure supplement 1). Interestingly, in all RSP3 mutants, the RS3 structure was intact, strongly suggesting that neither of the RSP3 paralogs is a Tetrahymena RS3 component.

Figure 3.

Taken together, the cryo-ET data suggest that in Tetrahymena there are at least two types of RS1 spokes: (i) RS1 type I containing RSP3A (spokes missing in RSP3A-KO cilia), making up approximately one-third of ciliary RS1 and (ii) RS1 type II, supported by RSP3B (missing in RSP3B-KO cilia), constituting approximately two-thirds of cilia RS1, and at least two types of RS2: (i) RS2 type I containing only RSP3B (3B-3B homodimer, intact RS2 spokes (18%) in RSP3C-KO cilia) and (ii) RS2 type II, having RSP3C in the stalk, likely as a heterodimer with RSP3B. However, we cannot exclude that some RS1 spokes can contain the RSP3A/RSP3B heterodimer or that RSP3A and RSP3B could partly substitute one another in the RS1 structure. Moreover, we cannot rule out that there is a minor fraction of RS2 containing the RSP3C homodimer. We must also keep in mind that in Tetrahymena the radial spokes’ heads are connected and therefore the radial spokes may stabilize each other. Thus, some RS1 spokes could be destabilized in the RSP3B-KO mutant (secondary effect), as a moderate reduction in the level of RSP3A shown in global proteome comparisons (see below) coincided with the RSP3B deletion.

Identification of Tetrahymena radial spoke subunits by comparative mass spectrometry analyses of wild-type and RSP3 knockout ciliomes

To identify RS1, RS2, and RS3-specific components, we isolated cilia from wild-type cells, RSP3 knockouts, and previously described Tetrahymena RSP knockouts: CFAP206-KO, missing either the entire RS2 or only its base, rarely accompanied by RS3 defects (Vasudevan et al., 2015), CFAP61-KO, lacking a part of the RS3 stalk (Urbanska et al., 2015), and CFAP91-KO frequently missing the RS3 and a part of the RS2 base (Bicka et al., 2022), and performed comparative proteomic mass spectrometry analyses using label-free and TMT 10-plex isobaric mass tagging approaches. The experiments were repeated three times to generate independent sets of data (at least 75% of proteins were present in all three replicas). Principal component analysis of the radial spokes mutant ciliomes using Ingenuity Pathway Analysis showed that the RSP3A-KO ciliome is most similar to the wild-type ciliome (Figure 4A). The changes in the protein levels were analyzed using Perseus software (v. 2.0.3) (Tyanova et al., 2016).

Figure 4.

Approximately one-thousand ciliary proteins from wild-type cells and RSP mutants were analyzed in LFQ experiments, while nearly four times more proteins were detected using the TMT approach. To determine the significant differences in the abundance of ciliary proteins between wild-type and knockout samples, we performed a t-test (p <0.05) (Figure 4-Source Data 1’, Figure 4-Source Data 2’) and identified a number of proteins differentially abundant (with a ±1.5-fold change) in knock-out mutants compared to the wild-type control. The obtained data are graphically represented as Volcano plots -log (P value) vs. log₂ (fold change of mutant/WT, Figure 4A) and Venn diagrams (Figure 5). The analyses of the levels of the outer dynein arms, two-headed, inner dynein arm IDAI1/f, tether/tetherhead complex, nexin-dynein regulatory complex, and central apparatus components showed that these structures were generally unaffected (Figure 4-Source Data 1’, Figure 4-Source Data 2’).

Figure 5.

To support and validate the global proteome comparisons, we performed co-IP and BioID assays followed by mass spectrometry analyses to identify ciliary proteins that either directly or indirectly interact with well-evolutionarily conserved, HA-tagged RSPs (Figure 6-Source Data 1’), or are biotinylated in cilia of cells expressing RSPs fused with mutated biotin ligase, BirA* (Figure 6-figure supplement 1, Figure 6-Source Data 2’). Next, we collated all proteomic data with bioinformatics analyses and cryo-ET ultrastructural studies to assign known RSPs and newly identified proteins to RS1, RS2, and RS3 spokes (Figure 6, Figure 6-figure supplement 3, and Table 1).

Figure 6.

Table 1

Localization of the RSP3 paralogs

Changes in the RSPs levels in the analyzed mutants are presented in Figure 4-Source Data 1’ and Figure 4-Source Data 2’ and summarized in Figure 6, Figure 6-figure supplement 3, and Table 1. Global mass spectrometry analyses of the RSP3A-KO, RSP3B-KO, and RSP3C-KO axonemes confirmed that the protein encoded by the targeted gene was completely eliminated. Moreover, the knockout of RSP3A or RSP3C did not affect the levels of other RSP3 paralogs while in RSP3B-KO cilia the level of RSP3A was mildly reduced. The knockout of CFAP61 has no effect on the level of RSP3 paralogs while in CFAP91-KO and CFAP206-KO mutants, the level of RSP3B and RSP3C were reduced, which agrees with the lack of RS2 spokes or part of the RS2 base/stalk in these mutants (Bicka et al., 2022; Vasudevan et al., 2015). Thus, the comparative mass spectrometry analyses of wild-type and mutant ciliomes support our hypothesis that there are two types of RS1 spokes (with either RSP3A-RSP3A or RSP3B-RSP3B dimers). The reduced RSP3A level in RSP3B-KO mutant cilia suggests either that in the absence of RS2, some of the RSP3A-containing RS1 spokes are less stable or that some of RS1 spokes contain the RSP3A-RSP3B heterodimer.

Based on the cryo-EM studies, we also hypothesized that the RS2 spokes contain either the RSP3B-RSP3B homodimer or the RSP3B-RSP3C heterodimer (not excluding, however, the possibility of the existence of the RSP3C-RSP3C homodimer). The comparative proteomic analyses revealed that the level of RSP3B did not change in RSP3C-KO mutant cilia, suggesting that in the absence of RSP3C, the RSP3B from the RSP3B-RSP3C heterodimer is still attached to the axoneme without maintaining the T-shape RS2 structure (Figure 3-figure supplement 1). However, the unaffected level of RSP3C (and RSP3C-associated RSPs, see Figure 6-figure supplement 3) in RSP3B-KO mutant cilia missing all RS2 spokes is surprising. To sum up, we found that RSP3B is a dominant Tetrahymena RSP3 paralog, roughly present in two-thirds of the RS1 spokes and likely in all RS2 spokes.

The RSP3 dimer interacts with other RSPs, forming together the RS head, neck, and stalk (Gui et al., 2021; Pigino et al., 2011; Yang et al., 2006). The bioinformatics analyses revealed that the Tetrahymena genome encodes more than one ortholog of several RSPs (Table 1 and Figure 6-figure supplement 3). Combined comparative ciliomes analyses, co-IP, and BioID assays allowed us to assign the RSPs to individual RSs (Figure 6 and Tables 1 and 2). Of note, the level of proteins that are the subunits of both RS1 and RS2 spokes in the RSP3A-KO mutant can be below the significance threshold (a ±1.5-fold change), as only one-third of RS1 contains the RSP3A paralog. Such proteins were assigned to RS1 based on the BioID data.

Table 2

The components of the Tetrahymena radial spoke stalk

RSP8 and RSP14 are ARM-like (armadillo-like) motif-containing proteins. The level of RSP8 was significantly reduced in mutants with RS2 defects (RSP3B-KO, RSP3C-KO, and CFAP206-KO) but not in RSP3A-KO cilia, which lack a substantial portion of RS1 spokes. In contrast, the level of RSP14 was reduced in mutants with RS1 defects, RSP3A-KO (RS1 type I) and RSP3B-KO (RS1 type II), but was unchanged in RSP3C-KO or CFAP206-KO mutants with only RS2 defects. The reduced level of RSP8 in mutants lacking RS2 and RSP14 in mutants missing RS1 agrees with the BioID assays showing the biotinylation of RSP8 in cells expressing RSP3B, RSP3C, or CFAP206 BirA* fusions and biotinylation of RSP14 in cells expressing either RSP3A-HA-BirA* or RSP3B-HA-BirA*. Thus, RSP14 and RSP8 are RS1 and RS2 proteins, respectively, which agrees with recent studies in Chlamydomonas (Gui et al., 2021) and murine sperm axonemes (Zhang et al., 2022).

RSP7 and RSP11 form a heterodimer near RSP14 or RSP8 (Gui et al., 2021). The Chlamydomonas RSP7, in addition to the RIIa domain (dimerization-anchoring domain of cAMP-dependent PK regulatory subunit) at its N-terminus, also has several EF hand motifs in the middle and C-terminal regions. A search of the Tetrahymena proteome using Chlamydomonas RSP7 as a query revealed two proteins with a weak similarity to Chlamydomonas RSP7, RSP7A, the 39 kDa protein containing N-terminal RIIa and C-terminal calmodulin-binding motif (IQ, isoleucine, glutamine) and 66 kDa RSP7B with an AKA28 (A-kinase anchoring protein) domain and two EF hands. The level of RSP7A was significantly reduced only in RSP3B-KO cells, but RSP7A was biotinylated in cells expressing either of the RSP3 paralogs as BirA* fusions. In contrast, the level of RSP7B was reduced in cilia lacking CFAP91. Accordingly, RSP7B was poorly biotinylated in cells expressing RSP3 paralogs as BirA* fusions but detected by a higher number of peptides among proteins biotinylated in CFAP61-HA-BirA* cilia, leading to the conclusion that this is likely an RS3-associated protein. In Tetrahymena, RSP11 is a small (7.9 kDa) basic protein (pI = 9) with a predicted RIIa domain and limited similarity to Chlamydomonas RSP11. The level of RSP11 was reduced in all RSP3 mutants.

RSP15/LRRC34 (leucine-reach repeat-containing) protein was previously described as an RS2-specific protein in Chlamydomonas (Gui et al., 2021). Indeed, the Tetrahymena RSP15 ortholog was nearly completely eliminated or significantly reduced in mutants with RS2 defects, in agreement with its RS2 localization.

In early radial spoke studies (Yang et al., 2001, 2006), the name RSP20 was assigned to calmodulin and RSP22 to LC8 (cytoplasmic dynein light chain 2). In Chlamydomonas flagella, calmodulin binds to FAP253 between RS1 and RS1-docked IDAs (Gui et al., 2021). In Tetrahymena, the level of calmodulin RSP20/CaM1 was reduced in mutants with RS2 defects (RSP3B-KO and RSP3C-KO) and was biotinylated in cells expressing RSP3B-HA-BirA*, suggesting an association of CaM1 with RS2 and perhaps RSP3B-containing RS1. On the other hand, the RS1 base containing CFAP253 is preserved in RSP3A-KO mutants, and thus, we cannot exclude the possibility that CaM1 is also an RS1 type I protein. Interestingly, the so-called calmodulin 7-2 was moderately reduced in cells with RSP3A or RSP3C deletion, suggesting that this calmodulin-type protein is the RS1 type I (RSP3A-containing) and RS2 type II (RSP3C- containing) component.

In Chlamydomonas flagella, RSP22/LC8 is associated with RS1 (four homodimers) and RS2 (two homodimers) (Gui et al., 2021). Chlamydomonas RSP22/LC8 is most similar to Tetrahymena LC8, a protein reduced in RSP3B-KO cilia. Based on BioID assays, this protein is likely positioned in the vicinity of all RSP3 orthologs, CFAP206, and CFAP91, suggesting an association with both RS1 and RS2. It was recently proposed that RS stalks dock to the LC8 multimers (Gui et al., 2021). Thus, it is possible that some RSP22/LC8 multimers remain associated with the axoneme in Tetrahymena RSP3A and RSP3C mutants. Indeed, the subtomogram average of Tetrahymena RSP3A-KO and RSP3C-KO cilia revealed that the spoke base remained associated with tubule A (Figure 3 and Figure 3-figure supplement 1), which explains why the level of LC8 was not reduced in these mutants.

Recent studies have led to the identification of additional RSPs. FAP207, a MORN motif- containing protein, localizes at the RS1 and RS2 base, and the IQ motif-containing FAP253, orthologous to Ciona CMUB116 (Zhu et al., 2017) and murine IQUB (Zhang et al., 2022), is an RS1 adaptor protein (Gui et al., 2021; Zhang et al., 2022). Tetrahymena CFAP253 and Chlamydomonas FAP253 are smaller than their mammalian orthologs and lack the ubiquitin homolog domain predicted in IQUB. The level of Tetrahymena CFAP253 was unaltered in all studied mutants, but CFAP253 was biotinylated in cells expressing RSP3A or RSP3B fused with BirA*. Thus, CFAP253 is likely an RS1 adaptor also in Tetrahymena cilia. The Tetrahymena genome also encodes one ortholog of FAP207 (Table 1 and Figure 6-figure supplement 3). The level of Tetrahymena CFAP207 was significantly reduced in cells with RS2 defects (except for RSP3C-KO cilia, where the RS2 base is present) but not in RSP3A-KO cilia (the RS1 base is also maintained in RSP3A-KO cilia).

The components of the Tetrahymena radial spoke head

The RS head consists of two identical lobes connected by an RSP16 dimer. In Chlamydomonas, each lobe is composed of RSP4 and RSP6 (paralogous proteins), RSP5, RSP9, the MORN motif-containing RSP1 and RSP10, a fragment of the neck protein, RSP2, and a C- terminus of RSP3 (Gui et al., 2021; Pigino et al., 2011; Yang et al., 2006). The Tetrahymena genome encodes three paralogs of RSP4/RSP6: RSP4A, RSP4B, and RSP4C, and single orthologs of RSP1, RSP9, and RSP10. The identification of RSP1 and RSP10 orthologs was possible only by a combination of bioinformatics and proteomic data, as the Tetrahymena genome encodes 129 proteins containing MORN motifs (Habicht et al., 2015). Only two of those proteins, named here RSP1 and RSP10, were identified among proteins that co- immunoprecipitated with HA-tagged RSP4 and were biotinylated in BioID assays. A blastp search with Chlamydomonas RSP1 or RSP10 as a query revealed that Tetrahymena RSP1 and RSP10 are more similar to Chlamydomonas RSP10 than RSP1.

The levels of RSP1, RSP9, and RSP10 were significantly reduced only in RSP3B-KO lacking all RS2 and two-thirds of RS1. The levels of several other MORN domain proteins, similar to RSP1/10, were also altered in the ciliomes of the analyzed mutants, but none of them was detected in either co-IP or BioID assays (Figure 6-figure supplement 3). Thus, it is unlikely that these other MORN domain-containing proteins are RS head components.

The loss of RSP3B coincided with the reduced levels of RSP4A and RSP4C, while the level of RSP4B was reduced in cilia of RSP3C and CFAP91 knockout mutants. Moreover, RSP4A and RSP4C were biotinylated in RSP3A-HA-BirA*- and RSP3B-HA-BirA*-expressing cells, while RSP4B was biotinylated in RSP4A-HA-BirA*-expressing cells. We propose that the head of RSs containing either RSP3A or RSP3B as a stalk component is composed of RSP:1, 4A, 4C, 9, and 10, while RS2 spokes containing RSP3C likely have RSP4B instead of RSP4C in the head. These conclusions must be considered with caution, as in Tetrahymena, the RS heads are connected (Lin et al., 2012; Pigino et al., 2011), and thus, a subunit of one spoke expressed as a BirA* ligase fusion could biotinylate proteins not only within the same spoke head but also of the adjacent spoke.

A search of the predicted Tetrahymena proteome with Chlamydomonas RSP5 aldo-keto reductase as a query failed to identify an ortholog. Among proteins annotated as aldo-keto reductases in the Tetrahymena Genome Database (TGD), thirteen were found in our Tetrahymena ciliomes, and four were significantly reduced in the RSP3 knockouts. However, these proteins did not co-precipitate with RSPs and were not biotinylated in cilia of Tetrahymena cells expressing RSP-BirA* fusions (Table 1 and Figure 6-figure supplement 3). The human genome also lacks RSP5. It is likely that RSP5 is a Chlamydomonas and related algae-specific protein.

The components of the Tetrahymena radial spoke neck

In Chlamydomonas, the V-shaped radial spoke neck is composed of RSP2, RSP12, RSP16, RSP23/FAP67, FAP198, and FAP385 (Gui et al., 2021). Tetrahymena has single orthologs of RSP2 and RSP23, two paralogs of RSP16, (RSP16A and RSP16B), CFAP198, (CFAP198A and CFAP198B), and RSP12 (RSP12A and RSP12B), but lacks FAP385.

Tetrahymena and Chlamydomonas RSP2 and RSP23/FAP67B proteins are similar only within the N-terminal regions containing DPY-30 and NDK (nucleoside diphosphate kinase) domains, respectively. RSP2 is likely a component of RS1 and RS2, as its level was reduced in both RSP3A and RSP3B knockouts (Tables 1 and 2). The level of RSP23/CFAP67B was not significantly altered in the analyzed mutants, but it was identified in co-IP and BioID assays. In addition to RSP23/CFAP67B, the Tetrahymena genome encodes CFAP67A (TTHERM_00266490), with a high similarity to Chlamydomonas RSP23. CFAP67A was detected neither in co-IP nor BioID assays. Based on recent single-particle cryo-EM and molecular modeling studies, FAP67 (likely CFAP67A) is an A-tubule MIP protein (Kubo et al., 2023). Thus, we propose that RSP23/CFAP67B is, similar to in Chlamydomonas, an RS1 and RS2 component.

The level of RSP16A was significantly reduced in cells with deletion of RSP3A or RSP3B, while the RSP16B level was diminished in cells with deletion of RSP3C, CFAP206, and CFAP91. The presence of RSP16A and 16B in RS1 and RS2 is also supported by BioID data.

Chlamydomonas FAP198 docks near RSP14/RSP8 stalk proteins (Gui et al., 2021). In Tetrahymena, the level of CFAP198A was significantly reduced in RSP3A and RSP3B knockouts, while CFAP198B was reduced in cells with RSP3C deletion, suggesting that CFAP198A is a component of RS1 and RSP3B-containing RS2, while CFAP198B is a subunit of RSP3C- containing RS2. Importantly, both CFAP198A and B were identified in co-IP (CFAP198A) and BioID assays (both paralogs).

The Tetrahymena genome encodes 15 proteins similar to RSP12, the peptidyl-prolyl cis- trans isomerase. Among the seven that were identified in the Tetrahymena ciliome, two are RSP12 candidates. TTHERM_01018330 (RSP12A) could be an RS2 component, as its level was reduced in cilia with RS2 defects (RSP3B-KO, RSP3C-KO, CFAP206-KO). RSP12A was also reduced in CFAP91-KO cells; however, it is not clear whether the detected reduction is due to RS2 elimination or RS3 defects. The level of the isomerase encoded by TTHERM_00466090 (putative RSP12B) was reduced in cells lacking RSP3A or RSP3B. Both proteins were detected in BioID assays by a very limited number of peptides.

New radial spoke candidate proteins

The global comparative proteomic analyses led to the identification of several proteins, the candidate RSPs or proteins positioned in the RS vicinity (Table 1 and Figure 6-figure supplement 3). The levels of two tetratricopeptide repeat (TPR)-containing proteins, TtTPR1 and TtTPR2, were reduced in mutants lacking RS2, TtTPR1 in RSP3C knockout and TtTPR2 in RSP3B knockout. TtTPR1 was found among biotinylated proteins in the BioID assays, while TtTPR2 appeared once in a co-IP. Kelch repeat-containing proteins are involved in various biological processes (Gupta & Beggs, 2014). The level of Kelch repeat-containing protein 1 (Kelch1) was diminished in the CFAP91-KO mutant. In agreement, the BioID and co-IP data also suggest that Kelch1 is an RS3 protein. Kelch2 was reduced in RSP3B and CFAP206 knockouts, suggesting its association with RS2 spoke.

The levels of two leucine-rich repeat-containing (LRRC) proteins similar to LRRC23, LRRC23-like 1 and LRRC23-like 2, were reduced in CFAP91-KO cilia but unchanged in RSP3 mutants, suggesting their association with RS3 spoke. The level of LRRC23-like 2 was also reduced in the CFAP61-KO mutant lacking only a fragment of the RS3 stalk but having a neck and a head. Thus, LRRC23-like 2 could be a component of the RS3 stalk. It was recently reported that LRRC23 is a subunit of the RS3 head in human and mouse sperm (Hwang et al., 2023). Thus, LRRC23-like 1 is likely a conserved component of RS3. The level of another LRRC protein, TTHERM_01084360, was reduced in cilia lacking RSP3B, CFAP206 or CFAP91, suggesting its presence in the RS2 vicinity. Interestingly, a protein with similarity to human LRRC45 (TTHERM_01220370) was found in both RSP precipitates and among biotinylated proteins, but its level in RSP mutant cilia was similar to that in the wild type, suggesting that LRRC45 is not a component of the RSs but is positioned in the RS vicinity.

We also found that the level of TTHERM_00557860 protein with the coiled-coil domain, orthologous to human CCDC104/FAP36/BARTL1, recently shown to be an Arl3-binding protein in the primary cilia (Lokaj et al., 2015), was also altered in RS mutants. The putative CCDC104 ortholog was poorly biotinylated in BioID assays, and its role or precise localization in cilia requires further analyses.

Radial spoke-associated proteins with enzymatic activity

ATP homeostasis

The ultrastructural defects in RS mutant cilia are accompanied by changes in the levels of a number of enzymes (Table 1 and Figure 6-figure supplement 3), suggesting that RSs serve as enzyme docking sites. Dynein arm motor activity requires a constant supply of ATP and its uniform distribution within the entire cilium. ATP is both delivered to cilia from the cell body by diffusion and produced within cilia (Villar et al., 2017). The levels of several enzymes involved in ATP homeostasis were altered in radial spoke mutants.

Enolase 1 (ENO1) and pyruvate kinase 1 (PYK1) are glycolytic enzymes that catalyze two final steps resulting in ATP synthesis. The levels of both enzymes were reduced in RSP3B and CFAP61 knock-outs (ENO1 also in CFAP91-KO). Intriguingly, the reduction of the ENO1 and PYK1 levels was also reported in the Tetrahymena central apparatus mutants, with the deletion of either SPEF2A or CFAP69 genes encoding subunits of the C1b projection (Joachimiak et al., 2021).

Adenylate kinases and phosphagen kinases can serve as ATP-shuttles and locally control ATP levels. Phosphagen kinases (creatine kinase in vertebrates, arginine kinase in Tetrahymena) catalyze the reversible transfer of the phosphate group from ATP to a guanidine substrate. The Tetrahymena genome encodes two arginine kinases, AK1 and AK2 (Michibata et al., 2014; Uda et al., 2006). The levels of ∼40 kDa AK1 and ∼80 kDa AK2 were reduced in RS mutants except for RSP3A-KO. ENO1, PYP1, AK1, and AK2 were not identified in co-IP or BioID assays.

Adenylate kinases reversibly catalyze the ATP+AMP↔ADP reaction and thus locally control the ATP level and maintain the homeostasis of adenine nucleotides (Dzeja & Terzic, 2009). The levels of at least six adenylate kinases (types 1, 7, 8, and 9) were changed in the cilia of Tetrahymena RS mutants. Importantly, adenylate kinases were also found among proteins that either co-immunoprecipitated with RSPs and/or were biotinylated in cilia of cells expressing RSPs fused with mutated BirA* ligase. Thus, adenylate kinases are likely radial spoke-associated proteins, while changes in the levels of arginine kinase 2, ENO1 and PYK1 may reflect more general changes in cilia physiology.

The level of AK8B adenylate kinase was dramatically reduced in RSP3A-KO cilia, while kinases AK7B and AK9 (and to a lesser extent, AK7A) were nearly completely eliminated in cilia of the CFAP91-KO mutant but unaltered in RSP3 mutants. Thus, RS1 is likely the main docking site for AK8B, while AK7A, AK7B, and AK9 preferentially associate with the RS3 spoke or its vicinity. AK1 and AK8A are likely associated with RSP3B-containing spokes, as their levels were significantly reduced in cilia of RSP3B-KO but not RSP3A-KO or RSP3C-KO mutants. With the exception of AK7B, AK8B, and AK9, which were either completely missing or substantially reduced in the analyzed mutants, the levels of other adenylated kinases were only moderately reduced, suggesting that in addition to radial spokes, AK1, AK7A, and AK8A likely associate with other ciliary structures.

cGMP-related enzymes

Of the two Tetrahymena guanylate kinases, GK1 is present in the Tetrahymena ciliome. Guanylate kinase catalyzes the transfer of phosphate from ATP to GMP, producing GDP and ADP. The phosphorylation of GDP to GTP is catalyzed by nucleoside-diphosphate kinases (NDK), and GTP can be converted to cGMP by guanylate cyclase, while 3’5’-cyclic nucleotide phosphodiesterase (PDE) hydrolyzes cGMP (Muthaiyan Shanmugam et al., 2018; Wyatt, 2015; Wyatt et al., 1998). The levels of GK1 and PDE (similar to AK9) were dramatically reduced in CFAP91-KO cilia (PDE also in CFAP61-KO) but not in RSP3 mutants. Moreover, GK1 co- immunoprecipitated with an RS3 base protein, CFAP251. Thus, GK1 and PDE are likely RS3- associated enzymes, and cGMP could play a role in signal transduction in the vicinity of RS3. The RSP23 protein contains NDK domains (Patel-King et al., 2004); however, its level did not significantly change in the analyzed mutants. Interestingly, we identified another protein containing an NDK domain (TTHERM_000160999) whose level was reduced in cilia lacking RSP3B or RSP3C.

Early analyses of ciliary beating using another ciliate, Paramecium, as a model showed that cyclic nucleotide monophosphate-driven chemical signaling controls ciliary beating (Bonini & Nelson, 1988; Hemmersbach et al., 1998; Noguchi et al., 2004). Thus, the immobility of the CFAP91- KO mutants might be caused not only by the lack of RS3 and some IDAs (Bicka et al., 2022) but also by the perturbation of guanidine nucleotide homeostasis.

Serine/threonine protein kinases

Comparative global proteomic analyses also revealed that the levels of three serine/threonine kinases, casein kinase 1 (CK1), calcium/calmodulin-dependent protein kinase, and cAMP-dependent protein kinase (PKA), were altered in radial spoke mutants. In Chlamydomonas, CK1 docks to outer doublets and regulates the phosphorylation of IC138, the IDAI1/f intermediate chain (Yang & Sale, 2000). In Tetrahymena, the level of CK1 was reduced in CFAP91-KO and CFAP61-KO cilia, suggesting its association with RS3 spoke. The presence of CK1 in the RS3 vicinity was confirmed by co-IP and BioID studies; however, it is unlikely that RSs are the sole CK1 docking site.

The RSP3 protein has an AKAP domain (A-kinase anchoring domain) (Gaillard et al., 2001) and thus locally regulates PKA localization and activity by restricting access to substrates. In Chlamydomonas, disruption of the RSP3 AKAP domain affects the activity of cAMP- dependent protein kinase (PKA) and causes abnormal flagellar motility (Gaillard et al., 2006). PKA functions as a tetramer composed of two regulatory and two catalytic subunits (Scott & Pawson, 2009). The levels of two PKA catalytic subunits, TTHERM_00433420 and TTHERM_00658860, and a regulatory subunit (TTHERM_00623090) were reduced in RS2 mutants.

The level of calcium/calmodulin-dependent protein kinase was apparently diminished in mutants with RS2 defects and in CFAP91-KO cilia, suggesting a role of RS2 and perhaps RS3 in calcium/calmodulin-dependent signaling.

Interestingly, at least some RSPs are present in cilia in more than one isoform (2- dimensional analyses of the HA-tagged RSPs, Figure 6-figure supplement 2 and) (Bicka et al., 2022)). It is tempting to speculate that some RSPs undergo phosphorylation and that the phosphorylation status controls other protein binding and/or signal transduction.

Radial spoke defect-related instability of inner dynein arms

Single-headed inner dynein arms (IDA) are docked in pairs at the radial spoke base, dynein a and b at RS1, c and e at RS2, and d and g at RS3. They are composed of IDA type- specific dynein heavy chain (DYH), actin, and centrin (dyneins b, e, g) or light chain protein, p28 (dyneins a, c, d) (Yamamoto et al., 2021). The comparative global proteomic analyses revealed that the deletion of genes encoding the components of the RS stalk also affected the levels of some IDAs (Table 2). The level of DYH9 was reduced in RSP3A-KO cells, while the levels of DYH10, 12, and 25 were decreased in the RSP3B-KO mutant, which coincides with the lack of dynein c density in the 96-nm subtomogram averaged map of RSP3B-KO cilia. The levels of DYH10, 12, and 25 were also reduced in CFAP206-KO, specifically missing dynein c (Vasudevan et al., 2015), and in CFAP91-KO mutants (Bicka et al., 2022), further supporting the assumption that these dynein heavy chains are the components of dynein c. The most frequent ultrastructural defect in Tetrahymena CFAP91-KO mutant cilia is a lack of RS3. As we previously reported, in addition to DYH10, 12, and 25, the levels of DYH16, 22, and 24 are moderately reduced in CFAP91-KO cilia (Bicka et al., 2022), suggesting that these dynein heavy chains are components of IDAd or IDAg.

Tetrahymena has three orthologs of the single-headed IDA light chain protein p28: p28A, p28B, and p28C (Subramanian et al., 2016). The level of p28B was reduced in cells lacking RSP3B and CFAP206, which agrees with p28B being a dynein c subunit (Yamamoto et al., 2021). In CFAP91-KO cilia, not only was the level of p28B reduced (Bicka et al., 2022) but also that of p28A, suggesting that p28A is a main component of dynein d (Yamamoto et al., 2021). The above data provide further experimental evidence regarding the protein composition of single-headed IDAs and show a heterogeneity of the protein composition of IDAs of a particular type (three different dynein heavy chains can be a component of IDAc). This raises a question about the biological significance of such diversity. Interestingly, in both RS2 mutants (RSP3B- KO and CFAP206-KO), only dynein c is missing (Vasudevan et al., 2015) and this work), suggesting differences in the docking of dynein c and dynein e at the RS2 base.

Discussion

Radial spokes are important links in the signal transduction chain from the central apparatus to the dynein arms. Here, using bioinformatics, proteomic and cryo-ET approaches, we were able to define a large part of the protein composition of the individual RSs in the ciliate Tetrahymena (Figure 6). In contrast to other studied organisms assembling motile cilia or flagella, the genome of Tetrahymena encodes more than one ortholog of several RSP genes (RSP3, RSP4/6, RSP7, RSP12, RSP16, CFAP198). This finding agrees with the Tetrahymena genome analyses revealing duplication of some genes encoding proteins involved in sensing or structural complexity (Eisen et al., 2006).

Immunofluorescence analyses revealed that Tetrahymena RSP3 and RSP4/6 paralogs are uniformly distributed along the entire cilium length. However, they are components of RSs of a specific type. Strikingly, in Tetrahymena, the differences in protein composition are not only between RS1, RS2, and RS3 spokes, as was shown earlier in Chlamydomonas (Gui et al., 2021) and mouse sperm axonemes (Zhang et al., 2022) but also within the group of RS1 and RS2 spokes. Tetrahymena likely has at least two types of RS1 (type I containing RSP3A and type II having RSP3B homodimers) and likely two types of RS2 spokes (type I containing RSP3B homodimer and type II with RSP3B-RSP3C heterodimer). At the moment, we cannot exclude the possibility that the RSP3A and RSP3B proteins also form heterodimers and that RSP3C can form homodimers. Interestingly, deletion of CFAP206 has a stronger impact on the level of RSP3C than RSP3B, while the level of CFAP206 is unaltered in RSP3C-KO but significantly reduced in RSP3B-KO, suggesting that RSP3C docks to CFAP206, which in turn binds to RSP3B. We have also found that some RSPs associate with a specific RSP3 paralog, further increasing RS diversity and suggesting co-evolution of some RSPs. Thus, the Tetrahymena cilium is a mosaic of RS1 and RS2 spokes composed of different RSP3 paralogs. At present, the significance of RS1 and RS2 heterogeneity is unknown.

The RS head transiently interacts with CA projections. These interactions can be mechano-chemical (Oda et al., 2014) and electrostatic in nature (Grossman-Haham et al., 2021). The very C-terminal end of RSP3C, compared to the RSP3A and RSP3B proteins, is enriched in glutamic acid residues that increase the RS head surface negative charge. If the RSP3C C- terminal end, similar to the Chlamydomonas RSP3 tail, is a part of the radial spoke head, its presence increases the negative charge of the head surface and thus might strengthen the electrostatic interactions with CA projections. Thus, the incorporation of RSP3C into the RS2 structure could locally strengthen the interactions between the RS and the CA projection(s). Interestingly, the N-terminal part of RSP3C is also unusual in having an ARF domain, which could play a role in signaling.

The analyses of the proteins biotinylated in cells expressing RSPs-BirA* fusions revealed the presence of several CA proteins (Figure 6-Source Data 2’), suggesting close proximity between the projections and the RSs. In particular, numerous peptides were those of two proteins, CFAP46 and CFAP54, the components of the C1d projection, which is surprising, as C1d is not the longest projection. This unexpected result calls for further studies.

The presented proteomic data clearly indicate that the lack of specific RSP3 protein causes not only destabilization of certain RSs but also affects the stability of specific IDAs, the receivers and effectors of transduced regulatory signals and perhaps other minor structures/linkers/proteins positioned in the RS vicinity. In cells with knocked out RSP3B, only one of two IDAs docked at the RS2 base, IDAc, is missing. These data strongly suggest that the docking of two IDAs, IDAc and IDAe, at the RS2 base is different. Moreover, comparative proteomic analyses of wild-type and RSP3B-KO cilia enabled the identification of dynein heavy chains that are components of IDAc.

The global comparative analyses of wild-type and RS mutant ciliomes led to another interesting discovery – the identification of several enzymes that likely dock to specific RSs. Among them are (i) adenylate kinases that together with arginine kinase locally regulate ATP homeostasis, (ii) enzymes involved in GMP recycling, and (iii) serine-threonine kinases that could play a role in the control/regulation of RS-mediated signal transduction. Thus, the phenotypic outcome of the deletion of genes encoding RS structural proteins is not simply a consequence of the radial spokes ultrastructural alterations but also of the alterations of the IDAs, linkers, and changes of the level of nucleotide, Ca2+/calmodulin, and protein phosphorylation- mediated signaling as well as the accessibility of the ATP-stored energy.

Materials and Methods

Cell growth conditions and phenotypic analyses

Tetrahymena cells were grown to the mid-log phase with shaking (80-110 rpm) at 30 °C. Wild- type cells and motile mutants were cultured in a standard SPP medium (Gorovsky et al., 1975), while mutants with major ciliary defects were grown in a rich MEPP medium (Orias & Rasmussen, 1976), both supplied with an antibiotic-antimycotic mix (Sigma-Aldrich, St. Louis, MO, USA) at 1:100 (SPP) or 1:50 (MEPP). The measurements of the cell swimming rate and the analyses of cilia beating (amplitude, waveform and frequency) were described in detail (Bicka et al., 2022). Briefly, for swimming rate analyses, cells at a density of 10⁴ cells/mL were viewed and recorded at room temperature using a Zeiss Discovery V8 Stereo microscope (Zeiss, Oberkochen, Germany) equipped with a Zeiss Plans 10_ FWD 81 mm objective, and an Axiocam 506 camera and ZEN2 (blue edition) software. The length of the trajectories was measured using ImageJ software. To analyze cilia beating, cells at a density of 10⁵ cells/mL were cultured at room temperature for 3 hours, and beating cilia were recorded using a Phantom Miro C110 high-speed camera (Vision Research, Wayne, NJ, USA) mounted on an AXIO Imager M2 microscope (Zeiss, Germany) with either a 40 x oil immersion lens (analyses of cilia beating frequency) or a 63x oil immersion lens (numerical aperture 1.4, analyses of ciliary waveform). Videos were recorded at 900 frames/s and analyzed using ImageJ as described (Bicka et al., 2022).

Knock-ins and knock-outs

Engineering and phenotypic analyses of Tetrahymena mutants with deleted CFAP61 (Urbanska et al., 2015), CFAP206 (Vasudevan et al., 2015), or CFAP91 (Bicka et al., 2022) were described. The RSP3 genes’ fragments used to obtain knock-out and knock-in transgenes were amplified by PCR with the addition of restriction sites using high-fidelity polymerase and the appropriate primers listed in Table S1. To delete fragments of the RSP3 genes, approximately 1.2-1.5 kb fragments positioned upstream and downstream of the targeted gene fragment were amplified and cloned subsequently on both sites of the neo4 resistance cassette (Mochizuki, 2008). Approximately 60 µg of plasmid was digested with ApaI and SacII to separate the transgene from the plasmid backbone, precipitated onto gold particles and used to transform conjugating Tetrahymena cells. Transformed cells were processed as described (Cassidy-Hanley et al., 1997; Dave et al., 2009). The deletion of the targeted fragments was confirmed by PCR.

To express RSP proteins as fusions with C-terminal either -3HA or –HA-BirA* tag under the control of a respective gene promoter, approximately 1 kb fragments of the open reading frame without the stop codon and 1 kb fragment of the 3’UTR were amplified and cloned into CFAP44-3HA-neo4 or CFAP44-HA-BirA*-neo4 plasmids, respectively (Urbanska et al., 2018) to replace CFAP44 gene fragments. On average, 10-15 µg of plasmid was linearized with MluI and XhoI, precipitated onto gold, and used to transform CU428 cells.

Immunofluorescence

To localize HA-tagged RSPs, cells from the overnight culture were fixed on coverslips with a mix of 0.25% Triton-X-100 and 4% PFA in a PHEM buffer. After drying and blocking, cells were stained overnight with a mix of primary antibodies, monoclonal anti-HA antibody (BioLegend, San Diego, CA, USA) diluted 1:300 and 6-11 B1 (Sigma-Aldrich), an anti- acetylated α-tubulin antibody (1:2000) at 4 °C. After washing with PBS, samples were stained for 1.5 h at RT with a mix of the secondary antibodies, anti-mouse or anti-rabbit IgG, conjugated with either Alexa-488 or Alexa-555 (Invitrogen, Eugene, OR, USA) diluted 1:300. Coverslips were mounted in Fluoromount-G (Southern Biotech., Birmingham, AL, USA). Cells were recorded using either a Zeiss LSM780 (Carl Zeiss Jena, Germany) or a Leica TCS SP8 (Leica Microsystems, Wetzlar, Germany) confocal microscope.

Cryo-ET preparation

The axonemes for cryo-ET were cross-linked with glutaraldehyde (final concentration 0.15%) for 40 min on ice and quenched with 35 mM Tris pH 7.5. The axoneme solution at 3.6 mg/mL was mixed with 5 (Cytodiagnostics) or 10 (Aurion) nm gold beads in a 1:1 ratio for a final axoneme concentration of 1.8 mg/mL. Inside the Vitrobot Mk IV (Thermo Fisher) chamber, 4 μl of crosslinked axoneme sample was applied to negatively glow discharged (10 mA, 10 s) C-Flat Holey thick carbon grids (Electron Microscopy Services #CFT312-100). The sample was incubated at 23 °C and 100% humidity for 45 seconds on the grid, followed by 8 seconds of blotting with force 0 and plunge frozen in liquid ethane.

Cryo-ET acquisition and reconstruction

Tilt series were collected using the dose-symmetric scheme from -60 to 60 degrees with an increment of 3 degrees at 2.12 Å per pixel using a Titan Krios equipped with Gatan K3 and BioQuantum energy filter. The acquisition was performed using SerialEM (Mastronarde, 2005). The defocus for each tilt series ranges from -2.5 to -6 μm. The total dose for each tilt series is 120 to 160 e- per Å2. For each view, a movie of 10-13 frames was collected. Motion correction of each view was performed with Alignframes (Mastronarde & Held, 2017). Tomograms were reconstructed using IMOD (Kremer et al., 1996).

Subtomogram Averaging

CTF estimation for each tilt series was performed with WARP (Tegunov & Cramer, 2019). The doublet microtubule for subtomogram averaging was picked using IMOD by tracing the line along the microtubules (Kremer et al., 1996). Subtomogram averaging of the 4-times binned 96 nm repeating unit of WT and mutant strains was performed using the “axoneme align” program (Bui & Ishikawa, 2013). The subtomogram coordinates and alignment parameters were converted to Relion 4.0 for local refinement and classification (Kimanius et al., 2021). The resolutions for the 96-nm repeating unit of the axoneme of wild type, RSP3A-KO, RSP3B-KO, and RSP3C-KO are 18, 20, 22, and 17 Å, respectively. Axonemal repeats from RSP3A and RSP3B mutants showed strong heterogeneity in the occurrence of RS1 and RS2. To analyze the heterogeneity of the mutant strains, three-dimensional (3D) classification without alignment was performed in Relion by masking RS2 and RS1. The axoneme of the RSP3C mutant strain showed heterogeneity in the RS2 head and neck region. Therefore, unsupervised 3D-classification by masking the head and neck region was performed with three classes.

For visualization, tomograms were CTF deconvolved and missing wedge corrected using IsoNet (Liu et al., 2022). The UCSF ChimeraX package was used for the visualization of subtomogram averages, surface rendering, segmentation, and fitting (Pettersen et al., 2021).

Deciliation and Western Blot

All buffers used in biochemical studies were supplied with protease inhibitors (cOmplete, EDTA- free protease inhibitor cocktail, Roche Diagnostics GmbH, Mannheim, Germany). To purify cilia, cells from the overnight, mid-log phase culture were collected, rinsed with Tris-HCl buffer, pH 7.4, resuspended in a deciliation buffer (10 mM Tris-HCl, pH 7.4, 10 mM CaCl₂, 50 mM sucrose), and deciliated by a pH shock (Wloga et al., 2008). Deciliation was monitored under a microscope. Samples containing intact non-moving cells were spun down twice at 1680 x g for 5 min to remove cell bodies, and cilia were collected by centrifugation at 23,300 x g for 30 min. Next, cilia were suspended in 10 mM Tris-HCl buffer (pH 7.4) with protease inhibitors, and the protein concentration was estimated using Pierce™ BCA Protein Assay Kit (Thermo Scientific, Bartlesville, OK, USA).

For Western blot analyses, 30 µg of ciliary proteins were loaded on the gel. After transfer to nitrocellulose, HA-tagged fusion proteins were detected using monoclonal anti-HA antibodies diluted 1:2000, followed by washing (4 x 10 min, TBST) with an incubation with HRP- conjugated secondary antibodies, 1: 10 000 in 5% milk/TBSP (Jackson ImmunoResearch, West Grove, PA, USA). HA-tagged proteins were detected using a Westar Supernova kit (Cyanagen, Italy). Two-dimensional electrophoresis was performed as described (Bicka et al., 2022).

Co-immunoprecipitation and proximity labeling (BioID) assays

For co-immunoprecipitation assay, 200 ml of cells (3 x 10⁵ cells/ml) expressing -3HA tagged radial spoke proteins under the control of a native promoter and wild-type cells (control), all at mid-log phase, were spun down and washed with 10 mM Tris-HCl buffer (pH 7.4). After deciliation, collected cilia were suspended in 10 mM Tris-HCl buffer (pH 7.4) and combined with an equal volume of 2% NP-40 and 1.2 M NaCl in 80 mM Tris-HCl buffer, pH 7.5, with protease inhibitors. After a 15 min incubation on ice, axonemes were pelleted at 20,000 × g and treated with 0.5 M KI, 30 mM NaCl, 5 mM MgSO4, 0.5 M EDTA, 1 mM dithiothreitol, and 10 mM HEPES, pH 7.5. After 30 min on ice, the axonemes were centrifuged (16,000 × g for 15 min at 4 °C), and the collected supernatant was diluted with 50 mM Tris–HCl, pH 7.4 (1:9). The axonemal proteins were incubated overnight with anti-HA magnetic beads with bound anti-HA antibodies (Thermo Fisher Scientific, Waltham, MA) at 4 °C. The bead-bound proteins were analyzed by SDS‒PAGE and silver staining and identified by mass spectrometry.

The proximity-labeling assay (Roux et al., 2012) was performed as described in detail (Joachimiak et al., 2021). Briefly, 200 ml of cells (3 x 10⁵ cells/ml) expressing RSP-HA-BirA* and wild-type cells (control) at mid-log phase were spun down and suspended in 200 mL of 10 mM Tris-HCl buffer (pH 7.4). After approximately 18 h, Tris buffer was supplied with biotin (50 µM final concentration), and the cells were cultured for 4 h at 30 °C. After deciliation, collected cilia were suspended in axoneme stabilization buffer (20 mM potassium acetate, 5 mM MgSO4, 0.5 mM EDTA, 20 mM HEPES, pH 7.5 with protease inhibitors) supplied with 0.2% NP-40, and after 5 min, the axonemes were spun down (30 min at 23,300x g at 4 °C) and suspended in lysis buffer (50 mM Tris–HCl, pH 7.4, 0.4% SDS, 0.5 M NaCl, 1 mM DTT with protease inhibitors). After an hour, samples were spun down (8000x g at 4 °C), and the supernatant was diluted with 50 mM Tris–HCl, pH 7.4 (1:3). The biotinylated proteins were incubated overnight with streptavidin-coupled Dynabeads (Dynabeads M-280 Streptavidin, Thermo Fisher Scientific, Waltham, MA) at 4 °C. The bead-bound biotinylated proteins were analyzed by Western blotting using HRP-streptavidin diluted 1:40 000 in 3% BSA/TBST (Thermo Fisher Scientific, Rockford, IL, USA) and identified by mass spectrometry.

Differential Quantitative Cilia Proteome Analyses

Sample Preparation

Cilia purified from 5x10⁷ cells were resuspended in axoneme stabilization buffer containing 0.2% NP-40 and incubated for 5 min on ice to remove the ciliary membrane. Axonemes were lysed for 1 hour at RT with lysis buffer (1% SDS, 50 mM Tris–HCl, pH 7.4), and protein concentrations were determined by a Pierce™ BCA Protein Assay Kit (Thermo Scientific, Bartlesville, OK, USA). Three hundred micrograms of protein was precipitated using a ReadyPrep 2-D Cleanup Kit (Bio-Rad Laboratories, USA) at −20 °C overnight. Next, samples were centrifuged at 20,000× g for 30 min at 4 °C (5804/5804 R Centrifuge, Eppendorf, USA), and supernatants were discarded. Protein pellets were dissolved in 0.1% RapiGest and 500 mmol/L TEAB and then incubated at 850 rpm for 45 min at 37 °C (Eppendorf Comfort Thermomixer, Eppendorf, USA). Proteins were digested by trypsin (Trypsin Gold, Mass Spectrometry Grade, Promega; protein:enzyme (w/w) ratio – 100:1, at 37 °C for 16 h) according to a standard protein digestion protocol including reduction (by 1,4-dithiothreitol) and alkylation (by iodoacetamide). The digestion reaction was stopped by the addition of 55% trifluoroacetic acid (final concentration of 5%), samples were centrifuged at 20,000× g for 30 min at 4 °C to precipitate RapiGest, and pellets were discarded. Supernatants containing the obtained peptides were purified using Pierce™ Peptide Desalting Spin Columns (Thermo Scientific, Bartlesville, OK, USA) according to the manufacturer’s protocol, dried in a vacuum concentrator at RT (SpeedVac Concentrator Plus, Eppendorf, USA) and stored at -80 °C for further analysis.

LC‒MS/MS analysis of labeled peptides (TMT analysis)

Desalted peptides were dissolved in 100 µl of 100 mmol/l TEAB solution, and peptide concentrations were determined using the Pierce™ Quantitative Fluorescent Peptide Assay (Thermo Scientific, USA). Next, a TMT labeling reaction was performed according to the procedure provided by the manufacturer. Briefly, a volume of sample containing 30 µg of peptides was labeled with a corresponding tandem mass tag. The labeling reaction was carried out for 1 h at room temperature and quenched with 5% hydroxylamine. Additionally, a test sample was prepared to check the efficiency of labeling. Labeled peptides were purified and fractionated using liquid chromatography at high pH. Separation was carried out for 26 minutes at a flow rate of 0.8 ml/min using a UPLC system (Acquity UPLC Class H system, Waters). The mobile phases consisted of water (A), acetonitrile (B) and 100 mmol/L ammonia solution (C). The percentage of phase C was kept constant at 10% throughout the separation. Fractions were collected every 1 minute, starting from the second minute of the run. The peptide elution was monitored spectrophotometrically at 214 nm. Twenty-four fractions were collected and combined to obtain 12 measurement samples. Samples were dried in a vacuum concentrator at room temperature. Peptides were resuspended in 100 μl of 5% acetonitrile and 0.1% formic acid.

Peptides were analyzed on the Evosep One system (Evosep Biosystems, Odense, Denmark) coupled to an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific, USA) according to (Bekker-Jensen et al., 2020). One µg of peptides was loaded onto Evotips C18 trap columns (Evosep Biosystems, Odense, Denmark) according to the manufacturer’s protocol with some modifications. Chromatographic separation of peptides was carried out using a mobile phase flow rate of 500 nl/min in gradient elution mode for 44 min on an EV1106 analytical column (Dr. Maisch C18 AQ, particle size 1.9 µm, 150 µm x 150 mm, Evosep Biosystems, Odense, Denmark). The following gradient elution was used: 0 min – 1% B, 120 min – 35% B, 121 min – 95% B, 124 min – 1% B, 127 min – 95% B, 130 min – 1% B. The eluted peptides were ionized in the positive ion mode in the nano-ESI source with a capillary voltage of 2.1 kV and the temperature of transfer capillary 275 °C. Survey scans from 300 m/z to 1700 m/z were acquired by an Orbitrap mass analyzer at a resolving power of 60 000. The resolving power in the MS2 spectrum measurement mode was 30 000 with the TurboTMT function set to TMT Reagents. HCD-MS/MS spectra (normalized collision energy of 30%) were generated for 25 multiply charged precursor ions from each survey scan. A precursor fit filter was applied to reduce peptide co-fragmentation. Dynamic exclusion was set to 20 s, and the precursor ion intensity threshold was set to 5x10³.

LC-MS/MS analysis of non-labeled peptides (label-free analysis)

Desalted peptides were resuspended in 100 μl of 5% acetonitrile and 0.1% formic acid. Next, peptide samples were analyzed on the Evosep One system (Evosep Biosystems, Odense, Denmark) coupled to an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific, USA) according to (Bekker-Jensen et al., 2020) with some modifications. Chromatographic separation and mass spectrometry detection were conducted as described in a previous article (Bicka et al., 2022).

Data analysis

MS data were analyzed with FragPipe (v. 17.1) (Nesvilab, University of Michigan, Ann Arbor, MI) with MSFragger (v. 3.4) (Kong et al., 2017) and Philosopher (v. 4.2.1) (da Veiga Leprevost et al., 2020). ProteoWizard’s MSConvert (v. 3.0.1908) (Palo Alto, CA) (Chambers et al., 2012) was used to convert the raw MS data to mzML format. The Tetrahymena thermophila UniProt database (canonical and isoform sequences; 27,027 entries) was searched using the following search parameters: (i) digestion enzyme-trypsin/P, up to two missed cleavage sites were allowed, (ii) precursor and fragment ion mass tolerance ±10.0 ppm and ±20 ppm, respectively, (iii) fixed modifications: carbamidomethyl (C), (iv) variable modifications: oxidation (M), deamidation (N), (Q). For TMT samples, additional fixed modification was used (v) TMT modification of lysine (+229.16293) and variable modification (vi) TMT modification of N-termini of protein and peptide (+229.16293). Proteins and peptides were identified using the target-decoy approach with a reversed database. The peptide mass range was set from 500 Da to 5 000 Da. The results were processed with FDR (false discovery rate) set to 1% at the PSM, peptide and protein levels. Quantitative analysis was performed using IonQuant (label-free quantitation) and TMT Integrator (TMT-based quantitation). Statistical analysis was performed with Perseus (v. 2.0.3) (Max Planck Institute of Biochemistry, Martinsried, Germany). In label-free quantitation, missing values were replaced based on quantile regression imputation of left-censored data (QRILC) (Rudolph & Cox, 2019). Student T-test was used for statistical analysis. Proteins were considered to be differentially expressed if the difference in abundance was statistically significant (FDR adjusted p-value < 0.05) and the fold change was equal to or higher than 1.5.

Reagent availability

All plasmids and Tetrahymena mutants are available on request. LFQ and TMT mass spectrometry data are provided as Figure 4-Source Data 1’ and Figure 4-Source Data 2’. The list of all protein identified in co-IP and BioID experiments are provided as Figure 6- Source Data 1’ and Figure 6-Source Data 2’. Nucleotide sequences of the primers used to amplified PCR fragments are provided in Table S1.

Acknowledgements

We thank Dr. Jacek Gaertig for critical reading of the manuscript. The immunofluorescence confocal imaging was performed in the Laboratory of Imaging Tissue Structure and Function, Nencki Institute of Experimental Biology, PAS. The mass spectrometry analyses were done in cooperation with the Mass Spectrometry Laboratory, Institute of Biochemistry and Biophysics, PAS, Warsaw, Poland. The authors have no competing financial interests to declare.

Funding

This work was supported by:

• the National Science Centre, Poland Grants, OPUS13 2017/25/B/NZ3/01609 to Dorota Wloga

• the National Science Centre, Poland Grants, OPUS15 2018/29/B/NZ3/02443 to Ewa Joachimiak

• Natural Sciences and Engineering Research Council of Canada (RGPIN-2022-04774) to Khanh Huy Bui

• Project No. POWR.03.02.00-00-I007/16-00 implemented under the Operational Program Knowledge Education Development 2014–2020 co-financed from the European Social Fund Project NCBiR No. PBS3/A8/36/2015.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Competing interest

The authors declare that no competing interests exist.

Author contributions

Marta Bicka, Investigation, Validation, Formal analysis, Visualization Corbin Black, Investigation,

Avrin Ghanaeian, Formal analysis, Investigation, Writing – review & editing, Visualization Ewa Joachimiak, Methodology, Investigation, Validation, Formal analysis, Visualization, Funding acquisition, Writing—Review and Editing

Anna Osinka, Investigation, Visualization Sumita Majhi, Investigation,

Anna Konopka, Methodology, Validation, Resources, Supervision

Ewa Bulska, Supervision, Resources, funding acquisition, writing - review & editing

Khanh Huy Bui, Investigation, Supervision, Resources, Funding acquisition, Writing – review & editing

Dorota Wloga, Conceptualization, Investigation, Validation, Visualization, Resources, Formal analysis, Supervision, Funding acquisition, Project administration, Writing – original draft, Writing – review and editing

Supplementary Tables

Table S1. List of primers used in this study. The nucleotide sequences recognized by the restriction endonucleases are in bold.

Source data

Figure 2-Source data1’. The analyses of the swimming rate of wild-type and RSP3 mutant cells Figure 2-Source data2’. The analyses of the cilia beating frequency in wild-type and RSP3 mutant cells.

Figure 4-Source data1’. LFQ-based identification of proteins with altered (reduced or elevated) levels in radial spoke mutants.

Figure 4-Source data2’. Tandem Mass Tag (TMT)-based identification of proteins with altered (reduced or elevated) levels in radial spoke mutants.

Figure 6-Source data1’. Mass spectrometry analysis of the ciliary proteins that co- immunoprecipitated with RSP3A-3HA, RSP4A-3HA, RSP4B-3HA, or RSP4C-3HA, all expressed at native levels. The radial spoke proteins are highlighted: RS head (dark blue), RS neck (yellow), RS stalk (green and light green), former CSC complex components and CFAP206 (violet), new RS components identified by Gui et al., 2021 (light brown), new putative RS components (pink), adenylate kinases (dark yellow), other enzymes (orange), and central apparatus components (cyan).

Figure 6-Source data2’. Mass spectrometry analysis of proteins biotinylated in cells expressing RSP3A, RSP3B, RSP3C, RSP4A, RSP4B, RSP4C, CFAP61, CFAP91, or CFAP206 with a C-terminal –HA-BirA* tag under the control of the respective native promoters. The radial spoke proteins are highlighted: RS head (dark blue), RS neck (yellow), RS stalk (green and light green), former CSC complex components and CFAP206 (violet), new RS components identified by Gui et al., 2021 (light brown), new putative RS components (pink), adenylate kinases (dark yellow), other enzymes (orange), central apparatus components (cyan).