The Meckel syndrome- associated protein MKS1 functionally interacts with components of the BBSome and IFT complexes to mediate ciliary trafficking and hedgehog signaling

The importance of primary cilia in human health is underscored by the link between ciliary dysfunction and a group of primarily recessive genetic disorders with overlapping clinical features, now known as ciliopathies. Many of the proteins encoded by ciliopathy-associated genes are components of a handful of multi-protein complexes important for the transport of cargo to the basal body and/or into the cilium. A key question is whether different complexes cooperate in cilia formation, and whether they participate in cilium assembly in conjunction with intraflagellar transport (IFT) proteins. To examine how ciliopathy protein complexes might function together, we have analyzed double mutants of an allele of the Meckel syndrome (MKS) complex protein MKS1 and the BBSome protein BBS4. We find that Mks1; Bbs4 double mutant mouse embryos exhibit exacerbated defects in Hedgehog (Hh) dependent patterning compared to either single mutant, and die by E14.5. Cells from double mutant embryos exhibit a defect in the trafficking of ARL13B, a ciliary membrane protein, resulting in disrupted ciliary structure and signaling. We also examined the relationship between the MKS complex and IFT proteins by analyzing double mutant between Mks1 and a hypomorphic allele of the IFTB component Ift172. Despite each single mutant surviving until around birth, Mks1; Ift172avc1 double mutants die at mid-gestation, and exhibit a dramatic failure of cilia formation. We also find that Mks1 interacts genetically with an allele of Dync2h1, the IFT retrograde motor. Thus, we have demonstrated that the MKS transition zone complex cooperates with the BBSome to mediate trafficking of specific trans-membrane receptors to the cilium. Moreover, the genetic interaction of Mks1 with components of IFT machinery suggests that the transition zone complex facilitates IFT to promote cilium assembly and structure.

Introduction the MKS and NPHP modules at the ciliary transition zone, hypomorphic alleles of components of the IFT complexes are also associated with ciliopathies. Every component of the IFT-A complex has been found to be mutated in human ciliopathies [35], as have a number of IFT-B complex components [36,37].
Given that ciliopathies have both overlapping and unique features, dissecting how the proteins and protein complexes work together can help us to understand ciliogenesis as well as the molecular basis of these genetic disorders. We set out to genetically test whether components of two ciliopathy complexes, the BBSome and MKS complex, function together to mediate cilia formation and whether either complex cooperates with IFT in cilium assembly. We show that MKS1 cooperates with BBS4 to mediate trafficking of subset of membrane proteins to the primary cilium. Moreover, we demonstrate that MKS1 functions together with the IFT machinery to mediate ciliogenesis, while BBS4 apparently does not.

Ethics statement
The use and care of mice as described in this study was approved by the Institutional Animal Care and Use Committees of Memorial Sloan Kettering Cancer Center (approval number 02-06-013) and Duke University (approval number A246- . Euthanasia for the purpose of harvesting embryos was performed by cervical dislocation, and all animal studies were performed in compliance with internationally accepted standards.

Mouse strains
A previously described Bbs4 genetrap allele [38], in which a trapping vector was inserted into intron 1, was used to generate Bbs4 -/embryos. We also made use of the following ENUinduced alleles in our genetic analysis: Mks1 krc [39], Ift172 avc1 [40], and Dync2h1 mmi [41]. Genotyping for all alleles was performed as previously reported. Details of the numbers of embryos of each genotype obtained for all crosses is presented in S1 Table. Cell culture and immunostaining MEFs were isolated from Embryonic Day (E)10.5 embryos and prepared and maintained as described [42]. Cells were shifted from 10% to 0.5% FBS the day after plating and maintained in low serum media for 48 hours to induce ciliogenesis. Cells were grown on coverslips and fixed in 4% Paraformaldehyde (PFA) in Phosphate Buffered Saline (PBS) for 5 minutes at room temperature followed by methanol for 5 minutes at -20C and washed extensively in PBS + 0.1% Triton X-100 (PBT). Fixed cells were placed in blocking solution (PBT + 5% FBS + 1% bovine serum albumin) for 30 minutes. Cells were then incubated with primary antibodies diluted in blocking solution overnight at 4˚C. Cells were then washed PBT and incubated with Alexa-coupled secondary antibodies and DAPI in blocking solution for 30 minutes at room temperature and affixed to slides for microscopy.
Cross sections of embryos tissues were obtained by fixing embryos in 2% PFA overnight at 4C. Embryos were then cryoprotected in 30% sucrose in PBS followed by embedding in OCT and freezing on dry ice. Transverse sections were cut at a thickness of 12μm onto slides, dried, washed in PBT + 1% serum, and incubated with primary antibodies as described below.

Fluorescence microscopy
Cilia images were obtained using a DeltaVision image restoration microscope (Applied Precision/Olympus) equipped with CoolSnap QE cooled CCD camera (Photometrics). An Olympus 100×/1.40 NA, UPLS Apo oil immersion objective was used. Z-stacks were taken at 0.20μm intervals. Images were deconvolved using the SoftWoRx software (Applied Precision/Del-taVision), and corrected for chromatic aberrations.
Confocal microscopy was performed using an upright Leica SP8 upright laser scanning microscope. Images were taken with a 63X water objective and 1X zoom. Extended views of the confocal datasets were processed using the Volocity software package (Improvision) or ImageJ.
The percentage of ciliated cells was measured as the proportion of cells with either ARL13b or acetylated α-Tubulin as indicated localized adjacent to a γ-Tubulin positive centrosome in 4 randomly selected fields per slide, averaged across 3 slides per genotype. Fields were imaged using a Zeiss Axioplan2 epifluorescence microscope. Cells were counted using ImageJ and data was analyzed with GraphPad Prism 6 software, using a one-way ANOVA with a Tukey-Kramer post-hoc test for significance.
Scanning electron microscopy E10.5 embryos were dissected in PBS at RT and immediately fixed in 2.5% glutaraldehyde and 2% PFA in 0.1M sodium cacodylate buffer, pH 7.4 (Electron Microscopy Sciences) overnight at 4˚C. The neural tube of the embryo at the level of the forelimbs was then separated from surrounding tissue and divided into two lateral halves. This tissue was then stored in 0.1M sodium cacodylate buffer overnight, followed by dehydration, drying and coating as previously described [4]. Scanning electron micrograph images were taken on a Zeiss SUPRA 25 FESEM.

qPCR
The MEFs derived from embryos of each genotype indicated were allowed to reach confluence and serum-starved from for 48 hours. The cells were treated with SAG (100nM) or DMSO for the last 24 hours of serum starvation. Each genotype and condition was plated in triplicate. The RNA was extracted using RNeasy Mini Kit (Qiagen) following manufacturer's instructions, with DNase treatment. Following RNA extraction, cDNA was produced with the iScript cDNA Synthesis Kit (BioRad). RT-qPCR was prepared with Taqman probes for Gli (Thermo, Mm00494654_ml) and housekeeping gene GAPDH (Thermo, Mm99999915_g1). Three technical replicates were performed for each biological replicate, using 9.8 ng of cDNA each, and an accompanying water control. The RT-qPCR was performed as a 96-well standard, using Applied Biosystems 7900HT Real-Time PCR System. Thermal cycling conditions are specified in Taqman Gene Expression Assays Protocol.

BBSome and MKS complex components genetically interact
BBS and MKS syndromes display phenotypic overlap in human patients [17]. To examine whether MKS1 and the MKS complex have roles in ciliary trafficking that may parallel those of other ciliopathy complexes, we tested whether Mks1 interacts genetically with the BBSome component Bbs4. We generated Mks1; Bbs4 double mutant embryos, using the previously described ENU-induced Mks1 krc allele [39] and a Bbs4 genetrap allele (here referred to as Bbs4 -/-) [38]. The Mks1 krc allele harbors a point mutation within a splice donor site that causes an insertion and an in-frame stop codon. Due to the lack of any WT splicing and early truncation of the protein caused by this mutation, Mks krc is thought to be a null allele of Mks1 [39]. We confirmed that another component of the MKS complex, TMEM67, fails to localize to the transition zone of Mks1 krc/krc cells, consistent with reports indicating that the individual components of the MKS complex are required for its stability [31] (S1 Fig). At E13.5, Bbs4 -/embryos were indistinguishable from wild-type siblings, as were Mks1 krc mutant embryos, with the exception of pre-axial polydactyly occurring in about 25% of these embryos (n = 33, 8 exhibited polydactyly) due to disrupted Hh signaling in the limb [39] (Fig 1A-1C). By contrast, Bbs4 -/-; Mks1 krc/krc double mutant embryos were never recovered after E14.5 and displayed microopthalmia and pronounced edema ( Fig 1D). Moreover, while pre-axial polydactyly is seen in a fraction of Mks1 krc/krc mutants as a duplication of digit 1 on at least one limb [39] (Fig  1G), 100% of double mutant embryos examined at E13.5 or E14.5 exhibited a duplication of digit 1 on all four limbs (n = 18; Fig 1D and 1H; summarized in Table 1).
A well-characterized role for non-motile primary cilia in the developing embryo is in mediating Hh signaling; perturbations to primary cilia structure or function lead to disruptions of Hh-dependent patterning in a variety of tissues, including the developing limbs and spinal cord [4][5][6]45]. To determine whether the enhanced severity of the double mutant embryos compared with Mks1 krc/krc or Bbs4 -/single mutants were due to exacerbated defects in Hh signaling, we examined ventral neural patterning in Bbs4 -/-; Mks1 krc , and double mutant embryos. V3 interneuron progenitors, labeled by NKX2.2, depend on high levels of Sonic Hedgehog (SHH) from the notochord and floorplate for their specification. Motor neuron progenitors (labeled by OLI G2) depend on intermediate levels of SHH signal [47]. We examined both cell types at the forelimb level within the neural tube as a read-out of Hh pathway activity in double mutants and Bbs4 and Mks1 single mutants at E10.5. Bbs4 -/mutant embryos exhibit normal neural patterning ( Fig 1J). Previous studies examining the phenotypes of Mks1 mutant embryos report complex neural patterning phenotypes wherein cell types that require intermediate levels of SHH for their patterning are slightly expanded, with the ventral most cells including V3 interneurons slightly reduced in number in more caudal regions of the neural tube [39,48]. We similarly observed only a slight expansion of OLIG2+ motor neuron progenitors at the forelimb level in Mks1 single mutants ( Fig 1K). By contrast, in double mutants NKX2.2+ cells are dramatically reduced and motor neurons have shifted ventrally to span the ventral midline ( Fig 1L). Taken together, these phenotypes suggest that removal of Bbs4 in a Mks1 mutant background exacerbates the Hhdependent patterning defects observed in Mks1 krc/krc embryos. We also tested the responsiveness of cells derived from Bbs4 -/-; Mks1 krc/krc double mutant embryos compared with WT and single mutant cells by performing qPCR for Gli1 levels in response to addition of SMO agonist (SAG), a potent activator of Hh signaling (S2 Fig). We find that both Bbs4 -/and Mks1 krc/krc single mutant fibroblasts exhibited an increase in Gli1 relative to WT cells. Consistent with the enhanced phenotypes seen in double mutant embryos, Bbs4 -/-; Mks1 krc/krc cells exhibit a dampened response to SAG as measured by Gli1 transcript levels, despite the increase in Gli1 in each single mutant.
While Mks1 krc/krc embryos do form cilia in most cell types, these cilia are sparse and shorter than wild-type [39]. To determine whether the enhanced severity of the Hh-related phenotypes in double mutant embryos is due to a greater disruption in ciliogenesis we examined cilia formation in double mutant embryos compared with single Bbs4 -/or Mks1 krc/krc mutants and wild-type sibling embryos.
Mouse embryonic fibroblasts (MEFs) were derived from embryos of each genotype and immunostained for acetylated α-Tubulin, which labels the stable microtubules of the ciliary axoneme. We found that primary ciliogenesis is unaffected in Bbs4 -/mutant cells compared with heterozygous controls (Fig 2A, 2B and 2M), while cilia are noticeably sparse in Mks1 krc/krc mutant cells (Fig 2C and 2M), consistent with previous studies [38,49]. Bbs4 -/-; Mks1 krc/krc double mutant cells had comparable numbers of cilia to those observed in Mks1 krc/krc single mutants (Fig 2C, 2D and 2M). SEM performed on embryonic neural tubes at E10.5 also In contrast, immunostaining with an antibody against the ciliary membrane protein ARL13B in both embryonic tissues and MEFs revealed a dramatic defect in Bbs4 -/-; Mks1 krc/krc double mutants. In both double heterozygous and Bbs4 -/-MEFs, over half of the cells formed an ARL13B+ cilium (56.3% and 56.1%; n>200 for each condition; Fig 2E, 2F and 2N), this number was reduced to a mean of 23.6% in Mks1 krc/krc cells (n = 420 cells, Fig 2G and 2N). Strikingly, even though Bbs4 -/cells form cilia at a rate nearly identical to that of heterozygotes, Bbs4 -/-; Mks1 krc/krc double mutant cells had a more severe defect in ARL13B trafficking than  Fig 2H and 2N). In support of our findings from MEFs, similar defects in ARL13B are observed upon immunostaining of the embryonic neural tube at E10.5 (Fig 2I-2L). Therefore, while cells of Bbs4 -/-; Mks1 krc/krc double mutants have ciliated cells at similar frequencies to those of Mks1 krc/krc mutants, the loss of Bbs4 in Mks1 krc/krc mutants results in additional defect in trafficking of the ciliary membrane protein, ARL13B. Based on this finding, we hypothesized that the BBSome and MKS complexes cooperate in the trafficking of specific ciliary membrane proteins.

BBS4 and MKS1 regulate transport of a subset of ciliary membrane proteins
To confirm that the reduced proportion of ARL13B+ cells in Bbs4 -/-; Mks1 krc/krc double mutants relative to the overall numbers of ciliated cells counted is due specifically to a failure of ARL13B to localize to cilia that are present, we performed double labeling for ARL13B and acetylated α-Tubulin. While in cells from double heterozygous, Bbs4 -/-, and Mks1 krc/krc embryos, each axoneme recognized by acetylated α-Tubulin was also positive for ARL13B ( Fig  3A-3C), in Bbs4 -/-; Mks1 krc/krc double mutants, nearly all cilia labeled by acetylated α-Tubulin lacked ARL13B (Fig 3D). This is consistent with our data quantifying the percentage of cells with each ciliary marker and shows that Bbs4 -/-; Mks1 krc/krc have exacerbated defects in the trafficking of this ciliary membrane protein.
To build upon this finding, we also examined a variety of different proteins that localize to the ciliary membrane to mediate cell signaling. These include INPP5E, a lipid phosphatase that helps to regulate the trafficking of other membrane proteins into and out of the cilium [50,51] as well as SMO and GPR161, G-protein coupled receptors (GPCRs) integral to the Hh pathway [50][51][52].
INPP5E binds to ARL13B, and this interaction has been shown to facilitate the trafficking of INPP5E into cilia [53]. Given this association, and our finding that ARL13B trafficking defects were further enhanced in Bbs4 -/-; Mks1 krc/krc double mutants compared with Bbs4 -/or Mks1 krc/krc , we examined INPP5E localization to cilia in these mutants (Fig 3E-3H, S4 Fig). In double heterozygous cells, we found that a mean of 64.5% of cilia were positive for INPP5E (n = 219). In Bbs4 -/cells, this proportion was significantly reduced (p<0.0001) 37.4% of cilia were INPP5E positive (n = 88). However, we found that there was no detectable INPP5E in either Mks1 krc/krc mutant cilia (n = 33 cells, Fig 3G), or in the cilia of in Bbs4 -/-; Mks1 krc/krc double mutant cells (n = 25 cells, Fig 3H), consistent with a requirement for MKS1 in the trafficking of INPP5E to cilia.
SMO is a GPCR-related protein that functions as an upstream activator of Hh signaling and localizes to cilia. SMO becomes enriched in cilia upon stimulation of the pathway with either SHH or the SMO agonist, SAG [54]. Consistent with previous reports [43,54], we find that SMO becomes more highly enriched in the ciliary membrane in double heterozygous cells upon treatment with SAG (-SAG: 38.1% cilia SMO+, n = 49; +SAG: 84.5% cilia SMO+, n = 40; Fig 3I and 3J). In contrast, in Bbs4 -/cells SMO is enriched in the ciliary membrane in the absence of SAG stimulation (-SAG: 64.1% cilia SMO+, n = 38; +SAG: 83.8% cilia SMO+, n = 65; Fig 3K and 3L), similar to reports for other BBSome mutants [55], while in Mks1 krc/krc cells, SMO accumulation in the cilium is dampened (-SAG: 12.7% cilia SMO+, n = 44; +SAG: 32.2% cilia SMO+, n = 38; Fig 3M and 3N). Accumulation of SMO in cilia upon stimulation with SAG is restored in Bbs4 -/-; Mks1 krc/krc double mutant MEFs, (-SAG: 19.4% cilia SMO+, n = 26; +SAG: 82.4% cilia SMO+, n = 38; Fig 3O and 3P), indicating that loss of Bbs4 is sufficient to overcome the dampening of SMO accumulation in response to SAG that we observe in the Mks1 krc/krc mutant cells. The graph quantifying this data can be seen in S4 Fig. We also examined trafficking of GPR161 in response to Hh pathway activation in each of the mutants. GPR161 is a GPCR that acts as a negative regulator of the Hh pathway. Consistent with its role in suppressing Hh signaling, GPR161 is present in cilia in the absence of pathway stimulation, and is lost from cilia when cells are treated with SHH or SAG to activate the pathway [52]. We observed this pattern of localization in double heterozygous cells (-SAG: 70.8% cilia GPR161+, n = 68; +SAG: 35.8% cilia GPR161+, n = 52; Fig 3Q and 3R, graph of quantification S4 Fig), while in Bbs4 -/mutant cells, GPR161 is detected in cilia even when the cells are exposed to SAG (-SAG: 78.9% cilia GPR161+, n = 72; +SAG: 75.9% cilia GPR161+, n = 54; Fig  3S and 3T). In Mks1 krc/krc single mutant cells, GPR161 was detected in cilia only at low frequency in the absence of pathway activation, and was entirely lost from cilia in cells treated with SAG (-SAG: 8.0% cilia GPR161+, n = 30; +SAG: 0% cilia GPR161+, n = 48; Fig 3U and  3V), suggesting that MKS1 is essential for trafficking of GPR161 into cilia. Surprisingly, in Bbs4 -/-; Mks1 krc/krc double mutant cells, GPR161 was seen at higher frequencies in cilia . Cells were stained for γ-Tubulin (red), acetylated α-Tubulin (green), or SMO (magenta). Two representative images for each condition are included. (Q-X) MEFs of each genotype were treated as described for (I-P) and immunostained for γ-Tubulin (red), acetylated α-Tubulin (green), or GPR161. Due to the low intensity of GPR161 relative to acetylated α-Tubulin, each of these channels is shown separately next to the corresponding merged channel image. Two representative images for each condition are included. (Y-F') MEFs of each genotype were treated as described in I-X and immunostained for α-Tubulin (red), acetylated α-Tubulin (green), or GLI2 (magenta). Two representative images for each condition are included. Scale bar, 5μm.
https://doi.org/10.1371/journal.pone.0173399.g003 compared to Mks1 krc/krc single mutant cells, and the frequency of GPR161 localization to cilia did not change in the double mutant cells upon addition of SAG (-SAG: 21.4% cilia GPR161+, n = 34; +SAG: 29.0% cilia GPR161+, n = 55; Fig 3W and 3X). Together, our SMO and GPR161 data indicate that the enhanced phenotype of Bbs4 -/-; Mks1 krc/krc double mutants is the consequence of combined trafficking defects arising from the dual loss of MKS and BBsome function. This indicates that these complexes act in parallel to regulate ciliary membrane composition.
Bbs4; Mks1 double mutants exhibit disrupted trafficking of the Hh pathway effector GLI2 To test more directly how the disrupted ciliary trafficking observed in the double mutant embryos causes the exacerbated Hh-related phenotypes observed in these embryos, we examined the trafficking of GLI2, the major positive effector of the Hh pathway [54]. Consistent with previous reports, we find that GLI2 becomes enriched at the ciliary tip in response to treatment with SAG in double heterozygous MEFs (-SAG 34.5% cilia GLI2+, n = 79; +SAG 74.9% cilia GLI2+, n = 119, Fig 3Y and 3Z; quantification of GLI2 data in S4 Fig). In Bbs4 -/cells, a high proportion of both unstimulated as well as stimulated cells have GLI2 at the ciliary tip (-SAG 80.6% cilia GLI2+, n = 78; +SAG 90.5% cilia GLI2+, n = 52; Fig 3A' and 3B'). In contrast, we found that in Mks1 krc/krc mutant cells, GLI2 was no longer confined to the ciliary tip but was present throughout the distal portion of the axoneme in nearly half of cilia examined (48.9%, S4 Fig), and like Bbs4 -/cells, GLI2 was seen at fairly high levels in cilia in the absence of SAG (Fig 3C'), and ciliary GLI2 did not increase upon addition of SAG (Fig 3D'; -SAG 88.2% cilia GLI2+, n = 34; +SAG 79.0% cilia GLI2+, n = 36). In Bbs4 -/-;Mks1 krc/krc double mutant MEFs, mis-localization of GLI2 within the axoneme is evident (Fig 3E' and 3F', S4  Fig), however the response of the double mutant cells to stimulation with SAG was dampened in these cells, and proportion of GLI2+ cilia was comparable to that double heterozygous unstimulated cells, in either the presence or absence of SAG (-SAG 38.1% cilia GLI2+, n = 35; +SAG 45.5% cilia GLI2+, n = 44; S4 Fig). We propose that the combination of perturbed ciliary membrane composition and disrupted trafficking of GLI2 along the axoneme causes the exacerbated Hh-related phenotypes observed in double mutant embryos, and the loss of Hh responsiveness seen in double mutant fibroblasts.

MKS1 cooperates with IFT to mediate cilium assembly
Through our examination of the trafficking of Hh components in response to pathway activation in Mks1 krc/krc , Bbs4 -/-, and double mutants, we identified a previously unreported requirement for MKS1 in localizing GLI2 to the ciliary tip. We found that GLI2 is present throughout the distal portion of the axoneme in these mutants (Fig 3C' and 3D'), reminiscent of other mutants in which retrograde IFT trafficking or ciliary tip structure is disrupted [46,56]. We therefore tested whether the MKS transition zone complex cooperates with IFT by assessing whether MKS1 genetically interacts with components of the trafficking machinery.
Similarly, the patterning of ventral neural progenitors is severely perturbed in Mks1 krc/krc ; Ift172 avc1/avc1 double mutants. Cells of the floorplate, labeled by FOXA2, as well as V3 interneuron progenitors, labeled by NKX2.2 depend on high levels of SHH for their specification and patterning. These cell types are patterned normally in both Mks1 krc/krc and Ift172 avc1/avc1 single mutants at this stage (Fig 4E-4G). In contrast, both of these cell types are entirely absent in double mutant embryos (Fig 4H). Motor neuron progenitors, labeled by OLIG2, and differentiated motor neurons, labeled by ISL1, are specified by intermediate levels of SHH. These are dramatically reduced in number in the double mutant embryos, and are shifted ventrally, spanning the midline of the neural tube (Fig 4I-4L). This dramatic enhancement of the severity of the Hh-dependent patterning phenotypes in the double mutant embryos suggests that MKS1 cooperates with the IFT-B complex to facilitate cilia formation. We also examined Hhresponsiveness in MEFs derived from embryos of each genotype by performing qPCR to assess Gli1 expression in cells in the presence or absence of SAG, as described above for our analysis of Mks1 krc/krc ;Bbs4 -/cells. We observed that Ift172 avc1/avc1 single mutants activate some expression of Gli1 in response to SAG, but double mutant cells fail to respond to SAG (S2 Fig). To directly examine whether MKS1 and IFT172 function together to mediate cilium assembly, we derived MEFs from embryos of each genotype and examined the percentage of ciliated cells upon serum withdrawal, using antibodies against ARL13B as well as acetylated α-Tubulin (Fig 5). Cells derived from Ift172 avc1/avc1 single mutants had comparable numbers of ciliated cells to double heterozygous MEFs (Double heterozygous: 81.1% ciliated with ARL13b, n = 116, 62.5% ciliated with acetylated α-Tubulin, n = 76; Ift172 avc1/avc1 : 75.3% ciliated with ARL13b, n = 100, 61.4% ciliated with acetylated α-Tubulin, n-67; Fig 5A, 5B, 5E, 5F, 5I and 5J), while Mks1 krc/krc MEFs exhibited reduced cilia frequencies (35.5% ciliated with ARL13b, n = 230, 37.1% ciliated with acetylated α-Tubulin, n = 85; Fig 5C, 5G, 5I and 5J). Consistent with the enhanced phenotypes related to reduced Hh signaling that we observed in double mutant embryos, we find that cells derived from Mks1 krc/krc ; Ift172 avc1/avc1 mutants have very few cilia labeled with either ARL13B (7.2%, n = 114; Fig 5D and 5I) or acetylated α-Tubulin (8.6%, n = 69; Fig 5H and 5J). We confirmed that the double mutants almost entirely lack cilia by SEM on the embryonic neural tube (S3 Fig). Notably, this failure of ciliogenesis contrasts with the defect we observed in the Bbs4 -/-;Mks1 krc/krc double mutant cells, which show a dramatic reduction in ciliary ARL13B localization, but not in overall cilia frequency, compared with Mks1 krc/krc MEFs. This points to cooperation between IFT and the MKS complex in initiating ciliogenesis and/or in maintaining the cilium.
To more precisely define the defect in cilium initiation in Mks1 krc/krc ; Ift172 avc1/avc1 mutant cells, we examined the recruitment of IFT proteins by determining the percentage of cells with IFT88 or IFT81 localized to the basal body or ciliary axoneme (Fig 6). For IFT88, we found that the percentage of cells with IFT88 present at the basal body in the double mutant cells (36.5%, n = 76) was significantly lower than the proportion observed in either double heterozygous control MEFs (87.8%, n = 70) or Mks1 krc/krc (72.1%, n = 81) and Ift172 avc1/avc1 (79.9%, n = 186) single mutant cells (Fig 6A-6D and 6M). IFT81 was still detected at one of the two centrosomes of Mks1 krc/krc ; Ift172 avc1/avc1 double mutant cells, as well as in the small number of axonemes present in these cells, at similar proportions to both single mutants (Fig 6E-6H and  6N). These results implicate MKS1 and the MKS complex in cooperating with IFT proteins to recruit or retain specific components of the IFT complex at the basal body.
While IFT88 is often seen throughout the ciliary axoneme, as well as concentrating at the ciliary base and tip, IFT81 is typically strongly enriched at the ciliary tip in double heterzygous cells (96.4% of cilia IFT81+, 20.5% with non-tip localized staining, n = 75) [56,59], with localization overlapping that of GLI2 [56]. In Ift172 avc1/avc1 cells, localization of IFT81 was slightly reduced relative to double heterozygous cells (70.9%, n = 186), and the percentage of cells with non-tip IFT81 was slightly elevated (34.6%), primarily due to increased localization at the ciliary base in these cells (Fig 6F and 6P). Similar to our observations for GLI2, we found that accumulation of IFT81 at the ciliary tip failed in Mks1 krc/krc mutants (69.1% of cilia IFT81+, 64.2% with non-tip localization, n = 99, Fig 6G and 6P). The same mis-localization of IFT81 was seen in Mks1 krc/krc ; Ift172 avc1/avc1 double mutants, with 100% cilia in these cells (n = 57, 67.7% cilia IFT81+) exhibiting non-restricted localization of IFT81 (Fig 6H and 6P). IFT81 is a marker for the ciliary tip compartment, which is comprised of the + ends of the axonemal microtubules. The atypical kinesin KIF7 is required for the structure and integrity of the tip compartment [56]. Since the defects in IFT81 ciliary localization are reminiscent of ciliary defects seen in Kif7 mutant embryos, we examined localization of KIF7 in cells from Mks1 krc/krc ; Ift172 avc1/avc1 mutants as well as both single mutants (Fig 6I and 6L). We found the percentage of ciliated cells with KIF7 present in the axoneme was significantly lower in double mutants compared with other genotypes (Double heterozygous: 77.4% cilia KIF7+, n = 58; Mks1 krc/krc : 67.0% cilia KIF7+, n = 59; Ift172 avc1/avc1 : 63.9% cilia KIF7+, n = 26; Double mutant: 29.7% cilia KIF7+, n = 45; Fig 6O). The distribution of KIF7 within the axoneme followed a similar pattern to IFT81, consistent with a role for MKS1 and the MKS complex in the cilium tip structure. In WT and Ift172 avc1/avc1 mutant cells, KIF7 is predominantly found at the ciliary tip (16.2% and 35% of cells had non-tip localized KIF7, respectively), consistent with previous findings (Fig 6I,  6J and 6Q) [41]. However in Mks1 krc/krc cells, KIF7 was broadly localized throughout the distal axoneme in the majority of cells (69.2% non-tip localized KIF7; Fig 6K and 6Q). The number of double mutant cells with KIF7+ cilia was very small (Fig 6L and 6Q; n = 13), yet KIF7 was still localized in a similar pattern to that exhibited in Mks1 krc/krc mutant cells (53.3% non-tip localized KIF7).
To further assess the phenotype of Dync2h1 mmi/mmi ;Mks1 krc/krc embryos, we examined Hhdependent neural patterning. Dync2h1 mutant embryos exhibit severe reductions of the most ventral cell types of the neural tube, but have some motor neurons [5]. For this reason, we examined the specification and patterning of both progenitors labeled by OLIG2 (Fig 7E-7H), and differentiated motor neurons labeled by HB9-GFP (Fig 7I-7L) in Dync2h1 mmi/mmi ;Mks1 krc/krc embryos as well as each single mutant. As we showed for previous experiments, both motor neuron progenitors and differentiated motor neurons in Mks1 krc/krc single mutant embryos were similar to those of double heterozygotes (Fig 7E, 7F, 7I and 7J). In contrast, Dync2h1 mmi/mmi mutant embryos had reduced numbers of OLIG2+ progenitors (Fig 7G), and differentiated motor neurons span the ventral midline, consistent with the loss of more ventral cell types ( Fig  7K). Double mutants exhibit a more severe defect, with a more dramatic reduction in the number of motor neuron progenitors (Fig 7H), and a lack of HB9-GFP+ motor neurons (Fig 7L).
In addition to genetically testing whether MKS1 cooperates with the IFT complex to promote ciliary trafficking, we also examined whether a component of the BBSome genetically interacts with IFT172 by generating Bbs4 -/-; Ift172 avc1/avc1 double mutants. BBSome proteins have been previously implicated in IFT: BBSome proteins traffic through in cilia in Chlamydomonas as IFT cargo [61], and BBS7 and 8 are required for normal IFT in C. elegans [62]. However, in contrast to our findings with Mks1 krc/krc ; Ift172 avc1/avc1 double mutant embryos, we did not observe any enhancement of phenotype of these double mutants, relative to Ift172 avc1/avc1 single mutants (S5 Fig).

Discussion
In this study, we show genetic interactions between MKS1 and the BBSome component BBS4, as well as between MKS1 and IFT machinery. Our data suggest that the BBSome and MKS complex components work both together and also have opposing roles in regulating the trafficking of proteins to the ciliary membrane. We show that BBSome and MKS complex components cooperate in targeting of ARL13B to the ciliary membrane, but have opposing effects on SMO localization, and to a lesser extent GPR161 in response to Hh pathway signaling: loss of Bbs4 on and Mks1 mutant background partially restores SMO and GPR161 localization. Another membrane protein, INPP5E, depends primarily on MKS1 for localization to the cilium, similar to results reported for other components of the MKS complex [33].These varied membrane trafficking defects in the double mutants point to these cells having a disrupted complement of ciliary membrane proteins, which is likely an important contributor to the exacerbated defects observed in the Bbs4 -/-; Mks1 krc/krc double mutant embryos (Summarized in Fig 8A-8C, and Table 1).
Our data are in broad agreement with recent work showing genetic interactions between other components of the MKS complex and the BBSome [33], however differences in our findings also uncover some important additional aspects of how these two complexes function together in ciliogenesis. For example, Yee et al. show that double mutants between Bbs1 and the MKS complex component Tctn1 form cilia at a substantially lower frequency than Tctn1 -/single mutants [33]. In contrast, our Bbs4 -/-; Mks1 krc/krc mutants form acetylated α-Tubulin positive cilia at nearly the same frequency as Mks1 krc/krc single mutants, however the cilia in the double mutants lack ARL13B. This difference is possibly due to the specific alleles used. For example, evidence suggests that BBS1 may play a more essential role in the stability of the BBSome complex than does BBS4 [63]. Nevertheless, perhaps by perturbing BBSome function more weakly, our data reveal that the MKS complex and BBSome cooperate to mediate trafficking of specific membrane proteins, such as ARL13B, to the cilium. Our data showing partial restoration of trafficking of some membrane proteins, such as SMO and GPR161, but not others upon impairment of BBSome function indicates a complex relationship between the MKS proteins and BBSome in ciliary membrane trafficking.
We identify strong genetic interactions between Mks1 and components of the IFT machinery: in double mutants between Mks1 and either a hypomorphic allele of Ift172, or Dync2h1 mmi we observe not only a reduction in ciliary protein trafficking, but also a failure of cilium assembly. This suggests that in addition to the previously described roles for the MKS complex in mediating entry of specific membrane proteins into cilia, MKS1 and the MKS complex cooperate with IFT to mediate the assembly of the axoneme. While the precise mechanisms by which MKS1 mediates IFT remain to be defined, localization or retention of IFT88 to the mother centriole is disrupted in Mks1 krc/krc ; Ift172 avc1/avc1 mutant cells, suggesting that MKS transition zone complex proteins may cooperate with IFT proteins to promote the recruitment or stable retention of IFT complexes at the basal body. This in turn contributes defects in cilium initiation in double mutant cells, and results in the dramatic exacerbation of Hh-related patterning defects seen in these embryos (For summary see Fig 8D, Table 1).
We note that recruitment of IFT88 to the basal body was disrupted by the combined impairment of MKS1 and IFT172, while IFT81 recruitment in the double mutants did not differ from either of the single mutants (Fig 6). A biochemical study of IFT-B proteins found that IFT81 and IFT88 are components of separate IFTB sub-complexes, bridged by a shared interaction with another IFT-B protein, IFT52 [64]. This could explain why these two IFT-B particle components are differentially affected in the Mks1 krc/krc ; Ift172 avc1/avc1 mutant cells.
MKS1 and the MKS complex may also directly or indirectly affect IFT trafficking, and/or the structure of the ciliary tip compartment. For example, we find that GLI2, IFT81 and KIF7 localization in Mks1 krc/krc single mutants is abnormal, with these proteins failing to enrich at the ciliary tip, and instead becoming localized to either a broader portion of the distal axoneme, or diffused throughout the entire axoneme, suggesting that MKS1 may be important for the structure of the ciliary tip. Mks1 krc/krc ; Ift172 avc1/avc1 double mutants also exhibit defects in IFT81 and KIF7 localization as well as a general reduction in the proportion of cilia that are positive for KIF7. Thus, our model is that MKS proteins at the transition zone cooperate with IFT to mediate cilium initiation, through recruiting or maintaining a subset of IFT proteins at the basal body (for example, IFT88). In double mutants, those cilia that are present also have defects in ciliary trafficking and structure, due to a direct or indirect role for MKS1 in the structure of the cilia tip compartment. The double mutants also reveal a cooperative role for MKS and IFT in recruitment of KIF7, a kinesin important for axoneme structure, to the cilium.
A connection between a transition zone complex and IFT was also recently found in C. elegans through the genetic interaction of nphp4, which encodes a component of the NPHP WT and double heterozygous cells. Red outline depicts the normal ciliary membrane integrity (based on ARL13b and INPP5E localization) SMO is present in the ciliary membrane, GPR161 is reduced within the ciliary membrane due to pathway activation. GLI2 and KIF7 are enriched at the ciliary tip. IFT components are enriched at the ciliary tip and the transition zone near the basal body (shown in purple). (B) Mks1 krc/krc cells exhibit defects in the localization of ciliary membranes (dashed red line)-ARL13b is reduced and INPP5E is absent. SMO is reduced, and GLI2, KIF7, and IFT81 are no longer restricted to the ciliary tip. (C) Mks1 krc/krc cells that also lack BBS4 exhibit additional defects in the trafficking of ciliary membrane proteins. ARL13B is absent, however SMO is restored in the cilia by the loss of BBS4. GLI2 is not restricted to the ciliary tip, as with Mks1 krc/krc single mutant cells, however a lower percentage of cells have ciliary GLI2. (D) Mks1 krc/krc cells that also have impaired IFT function due to a homozygous hypomorphic mutation in Ift172 have severe ciliogenesis defects. Less than 10% of Mks1 krc/krc ;Ift172 avc1/avc1 cells are ciliated. Those cilia that are present are shorter than normal. IFT88 is absent from the transition zone or the cilium, while IFT81 and KIF7 are no longer restricted to the tip, as with Mks1 krc/krc single mutant cells.
https://doi.org/10.1371/journal.pone.0173399.g008 transition zone complex, and osm-3 [22]. Osm3 encodes a kinesin important for IFT in the distal segment of the cilium [65], the vertebrate homolog of which is known as Kif17 [66]. Osm-3 did not genetically interact with MKS complex components in this study, however another study revealed that some of the functions of the MKS and NPHP transition zone complexes may have shifted in the course of evolution [33].
Our data on the relationship between the BBSome component BBS4 and IFT contrasts with another report, which showed that double mutants between Bbs7 and a hypomorphic allele of Ift88 (Ift88 orpk ) exhibited enhanced Hh-related patterning defects and embryonic lethality [55]. The reason for this difference in findings could be due to differences in the roles of the distinct protein components of the BBSome with respect to IFT function. For example, BBS7 is unique in that, in addition its role as a core component of the BBSome, it also physically interacts with the BBS chaperonin complex [67], a group of BBS-associated proteins that help to assemble the BBSome [68]. The synergistic phenotypic interaction between Bbs7 and Ift88 could therefore be due to this additional role of BBS7.
Taken together, our findings and those of others show that the separate protein complexes important for cilia structure and trafficking cooperate together to mediate specific aspects of cilia formation. The differences between our findings and those of other recent works highlight the fact that the distinct protein components of these complexes also have unique roles within the complex and could therefore have different relationships with other ciliary protein modules. Further exploration of this area will continue to reveal new information about how ciliogenesis is regulated as well as enhancing our understanding of the genetic complexity seen in ciliopathies. Expression was analyzed with three biological replicates for each experimental condition, with three technical replicates for each biological replicate. The graphs depict the average relative Gli2 levels (normalized to GAPDH for each replicate), error bars are standard deviation. (TIF) S3 Fig. SEM images of cilia from mutant embryos. (A-H) Embryos of the indicated genotypes were collected at E10.5, and the neural tubes were dissected out and fixed. Neural tubes were then opened and imaged en face by SEM. The regions imaged correspond to the ventral portion of the neural tube, anterior to the forelimbs. Scale bar, 1μm. (I) Quantification of the percentage of cells that are ciliated within the neural tube for each genotype. E10.5 SEM images of the neural tube were imaged as in A-H. A minimum of 2 fields were imaged for each of 3 embryos for each genotype at the forelimb level. In each field, cells with clear boundaries were counted and scored for whether or not they had a cilium. The percentage of ciliated cells for each field was recorded. Bars represent the mean of all fields examined for each genotype, and error bars are standard deviation. Genotypes were compared using ANOVA with a Tukey-Kramer multiple-comparison correction. Similar to our data based on percentage of ciliated cells in MEFs: we find that cilia frequency in Bbs4 -/-, Ift172 avc1/avc1 , and Dync2h1 mmi/mmi single mutants is comparable to WT embryos. Mks1 krc/krc ;Bbs4 -/have similar numbers of cilia to Mks1 krc/krc single mutants (p>0.99), however both Mks1 krc/krc ; Dync2h1 mmi/mmi and Mks1 krc/krc ;Ift172 avc1/avc1 double mutants have significantly fewer cilia than Mks1 krc/krc single mutants (p<0.0001, and p = 0.0002, respectively). (TIF) S4 Fig. Quantification of membrane protein localization. Related to Fig 3. (A-D) Each graph shows the percentage of cilia positive for the indicated marker. Bars represent the mean percent positive cilia for each genotype, error bars represent standard deviation. (E) The graph represents the percentage of GLI2+ cilia for each genotype (upon SAG treatment) in which GLI2 was not restricted to the ciliary tip (extending more than 1/3 of the length of the cilium from the tip, or seen in a location of the cilium other than the tip such as the base). Bars represent the mean percent of GLI2+ cilia with non-tip GLI2, error bars represent standard deviation.