Forkhead transcription factor FKH-8 is a master regulator of primary cilia in C. elegans

Cilia, either motile or non-motile (a.k.a primary or sensory), are complex evolutionary conserved eukaryotic structures composed of hundreds of proteins required for their assembly, structure and function that are collectively known as the ciliome. Ciliome mutations underlie a group of pleiotropic genetic diseases known as ciliopathies. Proper cilium function requires the tight coregulation of ciliome gene transcription, which is only fragmentarily understood. RFX transcription factors (TF) have an evolutionarily conserved role in the direct activation of ciliome genes both in motile and non-motile cilia cell types. In vertebrates, FoxJ1 and FoxN4 Forkhead (FKH) TFs work with RFX in the direct activation of ciliome genes, exclusively in motile cilia cell-types. No additional TFs have been described to act together with RFX in primary cilia cell-types in any organism. Here we describe FKH-8, a FKH TF, as master regulator of the primary ciliome in Caenorhabditis elegans. fkh-8 is expressed in all ciliated neurons in C. elegans, binds the regulatory regions of ciliome genes, regulates ciliome gene expression, cilium morphology and a wide range of behaviours mediated by sensory cilia. Importantly, we find FKH-8 function can be replaced by mouse FOXJ1 and FOXN4 but not by members of other mouse FKH subfamilies. In conclusion, our results show that RFX and FKH TF families act as master regulators of ciliogenesis also in sensory ciliated cell types and suggest that this regulatory logic could be an ancient trait predating functional cilia sub-specialization.

Eukaryotic cilia are complex and highly organized organelles defined as 48 specialized membrane protrusions formed from a stereotyped assembly of 49 microtubules. Cilia are composed of hundreds of proteins, required for their 50 assembly, structure and function, which are collectively known as the ciliome 51 ( Figure 1A). Cilia can be classified into motile or non-motile based on their 52 function and structure: motile cilia are responsible for propelling cells or 53 generating fluid flow while non-motile (a.k.a primary or sensory) cilia function as 54 cellular antennae to sense extracellular stimuli (Choksi et al., 2014). Cilia 55 appeared early in eukaryotic evolution and it is thought that ancient cilia 56 displayed mixed motile and sensory functions (Mitchell, 2017). In multicellular 57 invertebrates, primary and motor cilia are restricted to specific cell types. In 58 contrast, in vertebrates, primary cilia are present almost in every cell, including 59 neurons, while motile cilia are present only in specialized cell types. 60 Most ciliome components are shared between motile and primary cilia and are 61 referred as "core" ciliome ( Figure 1A). In addition, motile cilia usually contain 62 specialised axonemal dyneins and other motile-specific components while 63 membrane of sensory cilia is decorated with receptors that trigger downstream 64 signalling cascades when they are activated by small molecules, mechanical 65 perturbations, or radiation.

Persistent enhancer activity of ciliome genes in daf-19/RFX mutants 128
The activity of enhancers for ciliome genes bearing X-boxes is dramatically 129 reduced in daf-19(m86) null mutants. However, for several ciliome fluorescent 130 reporters, some residual activity has been anecdotally reported (Burghoorn et   As expected, daf-19(m86); daf-12 (sa204) double mutants show a dramatic 153 decrease in the number of neurons positive for each reporter (Figure 1C and 154 Figure S2). Importantly, all reporters except tmem-107, mks-1 and osm-5, 155 which correspond to the shortest constructs, show persistent expression in 156 some neurons (Figure 1C, Figure S2 and Supplementary File 1). We 157 hypothesised that these short constructs might lack binding sites for additional 158 TFs working with DAF-19. Indeed, we find that shorter versions of xbx-1 and 159 peli-1 reporter constructs are more affected by daf-19 mutation than 160 corresponding longer constructs, consistent with shorter sequences lacking 161 additional regulatory information ( Figure 1C). 162 Unexpectedly, we find that daf-12 itself has a small but significant effect on the 163 expression of several reporters (namely, che-13, ift-20, osm-1, mks-1 and 164 tmem-107) (Supplementary File 1), suggesting a possible role for this nuclear 165 hormone receptor (NHR) TF in ciliome expression. Altogether our data strongly 166 suggests that additional TF or TFs act together with DAF-19 to directly activate 167 core ciliome gene expression. 168 169

Identification of FKH-8 as candidate regulator of ciliome gene expression 171
We reasoned that similar to daf-19, additional regulators of cilia gene 172 expression could act broadly on many genes coding for ciliome components 173 and in many different ciliated neuron types. Thus, to identify these putative 174 candidates, we combined three strategies: cis-regulatory analysis of the ciliome 175 genes, TF expression enrichment in the sensory ciliated system and TF binding 176 to putative regulatory regions of the ciliome genes. 177 We built a manually curated list of 163 cilium effector genes (See materials and 178 methods and Supplementary File 2). This list can be divided in four categories: 179 1) 73 "core components" present in all types of cilia and thus expressed by all 180 ciliated neurons in C. elegans. Core components include IFT particles, kinesins, 181 dyneins, BBSome complex, etc; 2) 68 "Subtype specific" genes, that code for 182 channels or receptors located in cilia that are expressed in a neuron type 183 specific manner, providing neuron type specific functions; 3) 13 "Broad 184 expression" genes, specifically expressed within the ciliated system but not 185 A) Schema for a sensory cilium. Cilia components (ciliome) can be divided into core and subtype specific categories. Genes whose reporters are analysed in panel C are indicated by their function. B) Left lateral view of C. elegans hermaphrodite ciliated system. Sixty ciliated neurons from 25 different classes are distributed in 5 distinct anatomical regions. C) Ciliome core components show persistent expression in double daf-12(sa204), daf-19(m86) null mutants. Each dot represents the total number of reporter-positive neurons in a single animal. Mean and standard deviation are represented. The mean number of remaining reporter-positive neurons in double daf-12, daf-19 null mutants is indicated. See Supplementary File 1 for raw data, daf-12(sa204) single mutant scorings and sample sizes and Supplementary figure 2 for additional reporter scorings. D) DAF-19/RFX motifs (X-box) are enriched in regulatory sequences of core and broadly expressed ciliome genes. See Supplementary figure 3 for additional enriched motifs. E) Motifs enriched in regulatory sequences of ciliome genes containing X-box sites, potential TF binding to these motifs is unknown. F) sc-RNA-seq data analysis (Cao et al., 2017) identifies 10 TFs specifically enriched in ciliated sensory neurons. These TFs belong to FKH, ZF, NHR and HD families. See Supplementary figure 3 for detailed description of TF expression in each ciliated neuron type. G) ChIP-seq data analysis of 259 available TFs shows that FKH-8 ranks first in direct binding to genes of the ciliome list. See Supplementary figure 3 for core ciliome or subtype specific binding. H) Correlation of total number of peaks and ciliome-list genes bound by TFs shows FKH-8 behaves as an outlier, demonstrating high binding to ciliome genes is not merely due to the high number of FKH-8 binding-events. associated with a well-defined core cilia functions and 4) 9 "Male" genes that 186 code for genes with male-specific cilia functions (Supplementary File 2). 187 De novo motif enrichment analysis using the promoters of these ciliome genes 188 identified previously known RFX consensus binding sites (X-box motif). In 189 agreement with published results, X-box motifs are preferentially associated to 190 "Core" and "Broadly expressed" ciliome genes ( Figure 1D) ( Matrices for TFs. In contrast to the X-box, which is highly specific, TF binding 199 motifs (TFBM) for many TF families are often short and degenerate, thus they 200 appear widely in the genome and provide low information content. This feature 201 might explain the failure to find enriched motifs for additional TFs in ciliome 202 gene regulatory regions. 203 As an alternative to motif enrichment analysis, we turned to TF expression 204 enrichment. We hypothesized that TFs acting broadly on sensory cilia gene suggesting it could be a good candidate to work together with DAF-19. 214 Finally, we interrogated 446 published ChIP-seq datasets, corresponding to 259 215 different TFs (including FKH-8 but not DAF-19), for nearby binding to the 216 ciliome gene list (Supplementary File 2). We find FKH-8 behaves very 217 differently from the rest of TFs with at least one FKH-8 binding peak associated 218 to 49% of the genes on the ciliome gene list (Figure 1G-H). FKH-8 binds both 219 core components and subtype specific ciliome genes ( Figure S3), although, 220 similar to X-box motifs, FKH-8 binding is significantly more prevalent for core 221 ciliome genes (75% compared to 22% binding to subtype genes). Thus, both 222 sc-RNAseq and ChIPseq data analysis pinpoint FKH-8 as a good candidate TF 223 to directly control ciliome gene expression.  (Figure 2D and S4). In addition, there is a 237 high gene expression correlation for the 73 core ciliome genes and daf-19 or 238 fkh-8 expression but not with other TFs (Figure S4). Thus, our analysis shows 239 that FKH-8 is expressed almost exclusively in the whole ciliated sensory system 240 and its developmental expression correlates with core ciliome gene expression.  suggesting it could act as FKH-8 primary binding motif ( Figure 2F). 258 We noticed that eight out of the twelve functional X-boxes present in the core 259 ciliome reporters analysed in Figure 1C overlap with FKH-8 ChIP-seq peaks 260 ( Figure S1). Thus, we next looked for DAF-19 biding motif enrichment in FKH-8 261 bound regions. 21% of FKH-8 peaks contain at least one match for the DAF-19 262 position weight matrix ( Figure 2G). Importantly, predicted X-boxes are 263 preferentially found also at central locations, suggesting they could be in close 264 proximity to FKH-8 bound sites ( Figure 2G). DAF-19 binding sites are less 265 significantly or not significantly enriched in ChIP-seq datasets for other FKH TFs 266 ( Figure S4), suggesting specific co-binding of DAF-19 and FKH-8. 267

fkh-8 mutants show defects in ciliome reporter gene expression 269
The only available fkh-8 mutation, tm292, is a deletion downstream the FKH 270 DNA binding domain, suggesting it could act as a hypomorphic allele ( Figure  271 2A). Thus, we built fkh-8(vlc43), a null deletion allele that removes the whole 272 fkh-8 locus (Figure 2A). We selected eight reporters for six genes that code for 273 core cilia components and that overlap with FKH-8 ChIP-seq peaks ( Figure S1) 274 and analysed their expression both in fkh-8(tm292) and fkh-8(vlc43) mutants. In summary, our cis regulatory and fkh-8 mutant analyses unravel a cell 303 autonomous role for FKH-8 in the regulation of ciliome gene expression. 304 A) Dorso-ventral images from young adult heads expressing different core ciliome gene reporters in wild type and fkh-8(vlc43) null mutant animals. Significant expression defects in 5 distinct anatomical regions for both tm292 and vlc43 alleles are summarized in the right panel. Wild type reporter expression "++" indicates more than 50% of ciliated neurons in the region express the reporter, whereas "+" indicates expression in less than 50% of the neurons. n.e: not expressed. Statistically significant expression defects appear in red. Expression defect below 50% of wild type values are indicated as "+/-" whereas losses higher than 50% are depicted as "-". Enhanced phenotype in vlc43 mutants compared to tm292 is marked with an asterisk.

FKH-8 is required for correct cilia morphology 306
Mutations in several ciliome core component, including osm-5 and xbx-1, 307 whose reporters are affected in fkh-8 mutants, show cilium morphology defects 308 Next, we directly analysed cilium morphology labelling specific subpopulations 315 of ciliated neurons (Figure 4). Cilium length in CEP and AWB neurons is 316 significantly reduced in fkh-8(vlc43) mutants compared to controls, while ADF 317 neuron cilium length is significantly increased (Figure 4). In addition, fkh-8 318 mutants display arborization defects in AWA cilia, while AWC cilia showed no 319 major defects (Figure 4). 320 Thus, FKH-8 is necessary to regulate correct cilium length and morphology in 321 diverse types of ciliated neurons.

fkh-8 mutants display defects in a wide range of cilia mediated behaviours 334
In C. elegans cilia are necessary to mediate sensory functions (Bargmann, 335 1993); thus, we interrogated fkh-8 mutants with a battery of sensory paradigms. FKH-8 is also required for correct dauer entry, which is mediated by ASJ ciliated 358 neuron (Bargmann and Horvitz, 1991).  Figure S7). Thus, our data suggest that, despite its extensive binding 399 to daf-19 locus, FKH-8 does not seem to be required for daf-19 expression, at 400 least in the subpopulation of ciliated neurons directly assayed. 401 Figure 5. FKH-8 is required for the correct display of several sensory mediated behaviours. A) Mutations in fkh-8 significantly impair appropriate backward response to nose touch, revealing functionality defects for the ASH, FLP and/or OLQ ciliated neurons. This phenotype is stronger in fkh-8(vlc43) null mutants than in the hypomorphic tm292 allele. B) Decrease in locomotory rate upon re-entering a bacterial lawn is unaffected in fkh-8 mutants. C) fkh-8 null mutants significantly fail to prevent dauer entry. Pheromones induce dauer in fkh-8 mutants, albeit less efficiently than in controls. D to F) Lack of fkh-8 significantly impairs olfaction-mediated behaviours towards compounds sensed by ciliated AWB and AWC neurons. Diacetyl response, mediated by AWA, is affected but not to a significant level due to high variability in the response. G to I) Attractive chemotaxis towards NaCl is unaffected in fkh-8 mutant animals. Avoidance behaviour towards toxic SDS and copper anions is significantly impaired. Mean and standard deviation are represented in all graphs. See Supplementary figure 6 for quantification of non-cilia mediated behaviours and Supplementary file 3 for raw data and samples' sizes.
Next, we assessed FKH-8::GFP fosmid expression in daf-19(m86); daf-402 12(sa204) double null mutants. In contrast to the pan-ciliated neuron expression 403 pattern seen in wild type, FKH-8::GFP is expressed pan-neuronally in daf-404 19(m86); daf-12(sa204) double mutants (Figure 6B) -1, osm-1 and xbx-1 reporters (Figure 6C, D). Of note, these reporters 423 still show some vestigial expression in the triple mutant ( Figure 6D). We    Importantly, vertebrate sensory ciliogenesis is unaffected in FoxJ1 loss of 511 function mutants (Choksi et al., 2014); thus, FoxJ1 role as a master regulator of 512 ciliogenesis is restricted to motile ciliary cell types. In Xenopus, FoxN4 binds 513 similar genomic regions to FoxJ1 and it is also required for motile ciliome gene 514 expression (Campbell et al., 2016). We find both FOXJ1 and FOXN4, but not 515 FOXI1, which has not been described to be involved in ciliogenesis, are able to 516 functionally substitute FKH-8. This data suggests that specific FKH subfamilies 517 might have an inherent capacity to act as direct ciliome regulators, 518 independently of being expressed in motile or sensory cilia cell types. bind similar genomic regions is not limited to metazoans and it is also present in 546 fungi. In Schizosaccharomyces pombe, which lacks cilia and ciliome genes, 547 Fkh2 FKH TF and Sak1 RFX TF bind the same regulatory regions to control cell 548 cycle gene expression (Garg et al., 2015), suggesting that the joint actions for 549 these TFs could be present before the split of fungi and metazoans. 550 Alternatively, RFX and FKH TFs might have an inherent ability to cooperate that 551 could explain convergent evolution of these TFs in ciliome regulation both in 552 sensory and motile cilia cell types (Sorrells et al., 2018). 553 In light of these data, we hypothesize that RFX and FKH role as co-regulators 554 of ciliome gene expression could precede the emergence of cilia division of 555 labor and the specialization of motile and sensory cilium in different cell types 556 ( Figure 7D).

C. elegans strains and genetics 583
C. elegans culture and genetics were performed as previously described 584 (Brenner, 1974).

Generation of C. elegans transgenic lines 606
Fluorescent reporters for ciliome genes were generated through fusion PCR 607 (Hobert, 2002). To facilitate identification and scoring of reporter-expressing 608 cells, GFP was tagged to the cell's nucleus employing a modified sequence of 609 the classical SV40 large T antigen nuclear localizing signal (NLS) (Kalderon et 610 al., 1984). Regulatory sequences were amplified with custom oligonucleotides 611 from N2 genomic DNA preparations. An independent PCR was used to amplify 612 the 2xNLS::GFP::unc-54 3'UTR fragment from an NLS version of the pPD95·75 613 plasmid (pNF400). Successfully fused PCR products were purified using the 614 Simple-array transgenic lines were generated by intragonadal microinjection 636 into strains of the appropriated genotype. The injection mix was composed by 637 50 ng/µL of a given purified fusion PCR or a rescuing plasmid plus 100 ng/µL of 638 the pFR4 plasmid, rol-6(su1006), as a co-marker (Mello et al., 1991). 639

Generation of C. elegans mutations 640
Whole deletion of the fkh-8 locus was performed through a co-CRISPR strategy 641 All micrographs presented in this paper were acquired with a TCS-SP8 Leica 748 Microsystems confocal microscope using a 63X objective and appropriate 749 zooming conditions. Raw data and statistics for all scorings performed for this 750 work are gathered in Supplementary File 1. 751

Behavioural assays 752
Unless otherwise stated, all mechano-and chemosensory assays were 753 performed over small-scale synchronized populations of young adult 754

hermaphrodites. 755
Nose touch tests were performed as previously described (Kaplan and Horvitz, 756 1993). Ten minutes before the assay, young adult hermaphrodites were 757 transferred to non-seeded NGM agar plates and nose touch responses were 758 Both gentle and harsh touch mechanosensory tests were performed as 763 previously described (Chalfie et al., 1985). Briefly, gentle touch assays were 764 performed by alternatively stroking the animal just behind the pharynx and just 765 before the anus with an eyebrow hair attached to a pipette tip for a total amount 766 of 10 strokes (Hobert et al., 1999). Harsh touch assays were also performed by 767 stroking the worms across the posterior half of their bodies in a top-down 768 manner with a platinum wire. Each worm was tested five times with a 2 minutes 769 interval between each trial (Li et al., 2011). 770 For all aforementioned mechanosensory assays, escape responses of trailed 771 animals were recorded and a population response index (RI) was calculated for 772 every replica as: RI = total number of escape responses / total amount of 773 strokes 774 Chemotaxis towards diacetyl, 2-heptanone, NaCl and 2-nonanone were 775 performed over 3 times freshly washed worms with 1 mL of filtered, autoclaved 776 CTX solution, aspirating the supernatant to a final volume of approximately 100 777 µL. 2 µL of this worm-containing solution with no less than 25 animals were 778 placed at the proper place of the assay plates. During the assays, worms were 779 allowed to freely crawl across the plates for 60 minutes at room temperature 780 and then stored at 4 °C until the next day when worms' positions were scored 781 and behavioural indexes were calculated. 782 With few modifications, volatile diacetyl attraction assay was performed as 783 described by (Margie et al., 2013). A four-quadrant paradigm drawn at the base 784 of non-seeded NGM agar plates was used, adding a 1 cm circular central area 785 that worms had to trespass to be scored. Stock diacetyl (Sigma-Aldrich, 786 #803528) test solution was prepared as a 0.5% V/V mix in absolute ethanol 787 (Scharlau, #ET00101000). Absolute ethanol was used as control solution. 788 Immediately after the worms were plated, 2 µL of a mix combining equal 789 volumes of diacetyl stock solution and sodium azide 1M were pipetted onto the 790 2 test sites (T) of the agar plate. Same procedure was then performed for the 2 791 control sites (C). Chemotaxis index (CI) was then calculated as: CI = (worms in 792 (T1 + T2) -worms in (C1 + C2)) / total scored worms 793 Chemotaxis assay towards 2-heptanone was performed as previously reported 794 Avoidance responses to water-soluble compounds were evaluated using the 825 drop test as previously described (Hilliard et al., 2002). Following (Hilliard et al.,826 2004) with few modifications, well-fed synchronized young adult 827 hermaphrodites were washed three times with M13 buffer. 5 animals were then 828 placed on unseeded NGM agar plates and allowed to rest for 10 minutes. Two 829 test solutions were assayed: 0.1% W/V sodium dodecyl sulfate (SDS) (Sigma, 830 #L3771-100G) and 0.1 mM CuSO4 pentahydrate (Merck, #1027901000), both 831 dissolved in the M13 buffer that acted as control solution. Each animal was 832 tested first with 4 single drops of the control solution and then with 4 single 833 drops of the testing solution, allowing for 2 minutes of recovery between each 834 stimulus. Avoidance response was scored within 4 seconds after substance 835 delivery. Population avoidance index (AI) per genotype and replica was 836 calculated as: AI = number of responses / total amount of drops. 837 Dauer induction was performed using filtered liquid culture obtained from wild 838 type worms grown at 7 worms/µl for 4 days. Briefly, 300µl of pheromone 839 containing extracts or control extracts (culture media alone) were added to 840 60mm OP50-seeded NGM plates. After drying, 10 gravid worms were added 841 and allowed to lay eggs for 18 hours and then removed from the plates. 72h Unless otherwise stated, same two-tailed t-test procedure was followed in the 875 assessment of statistical significance in behavioural experiment. Behavioural 876 responses were ultimately analysed through the corresponding indexes ranging 877 from 0 to 1 (or to -1 to 0 when avoidance responses were assayed). For each 878 type of assay, a population-based mean index was calculated per replica and a 879 final response index was then obtained as the mean of all replicas' means. Prior 880 to hypothesis testing, the Shapiro-Wilk test (Shapiro and Martin, 1965) was 881 used to address for the normality of these final indexes. For the assessment of statistical significance in rescue experiments, data was 892 categorically classified as 'on' or 'off' and the significance of the association was 893 examined using the two-tailed Fisher's exact test. No further multiple testing 894 correction was performed, as fkh-8 null mutants were exclusively compared to 895 wild type worms whereas each rescued line was exclusively compared against 896 the fkh-8 null mutants. 897