The G protein-coupled receptor Gpr161 regulates forelimb formation, limb patterning and skeletal morphogenesis in a primary cilium-dependent manner

Limb development is an orchestrated process involving initiation of the limb bud from the lateral plate mesoderm (Gros and Tabin, 2014; Duboc and Logan, 2011), patterning of the limb bud (Zeller et al., 2009), and skeletal morphogenesis (Kronenberg, 2003). Sonic hedgehog (Shh) and Indian hedgehog (Ihh) are crucial factors in limb and skeletal development, but they have distinct functions in these processes. Shh expression starts in the posterior forelimb bud from embryonic day (E) 9.5 (Platt et al., 1997; Lewis et al., 2001) and continues until E12 to regulate limb bud patterning (Zeller et al., 2009). In contrast, Ihh is secreted from pre-hypertrophic and hypertrophic chondrocytes starting at E11.5 (Bitgood and McMahon, 1995) and regulates endochondral bone formation (Long et al., 2004; St-Jacques et al., 1999; Lanske et al., 1996; Vortkamp et al., 1996). Ihh also inhibits early stages of osteoblast differentiation during intramembranous bone formation (Abzhanov et al., 2007). Shh- and Ihh-mediated activation in limb and skeletal morphogenesis has been studied extensively (Mak et al., 2006; Kobayashi et al., 2005; Butterfield et al., 2009). However, the role of suppression of these morphogenetic pathways beyond the periods of expression of Shh/Ihh is underappreciated.

The primary cilium is a microtubule-based dynamic cellular appendage that mediates extracellular signaling particularly with respect to vertebrate Shh signaling (Goetz and Anderson, 2010). Activation of the Shh pathway by formation of the Gli transcriptional activator (GliA) and basal repression of the pathway by Gli transcriptional repressor (GliR) are both dependent on the primary cilium (Goetz and Anderson, 2010). Binding of Shh to patched 1 (Ptch1) triggers removal of Ptch1 from cilia, and promotes smoothened (Smo) enrichment in cilia, which mediates GliA formation (Corbit et al., 2005; Rohatgi et al., 2007). In contrast, the basal repression machinery of Shh signaling involves protein kinase A (PKA)-mediated GliR formation in a cilia-dependent manner (Mukhopadhyay and Rohatgi, 2014).

We recently described that the cilia-localized orphan G protein-coupled receptor (GPCR) Gpr161 functions as a negative regulator of Shh signaling during early neural tube development in mice (Mukhopadhyay et al., 2013). Gpr161 knockout results in increased Shh signaling (ventralization) throughout the rostrocaudal extent of the neural tube, without disrupting cilia. Gpr161 determines Gli3R formation possibly via constitutive cAMP signaling. Another negative regulator, suppressor of fused (Sufu), restrains Gli3 in cytoplasm and promotes Gli3R processing (Humke et al., 2010) in a cilia-independent step (Jia et al., 2009). Importantly, lack of Gpr161-, PKA- and Sufu-dependent basal suppression cause high Shh signaling during mouse neural tube development (Tuson et al., 2011; Mukhopadhyay et al., 2013; Svärd et al., 2006), similar to Ptch1 knockout, which results in activation of the canonical Smo-dependent arm of the pathway (Goodrich et al., 1997). Thus, both transcriptional activation and basal repression mechanisms are crucial for regulation of high Shh signaling.

In addition to Gpr161, we and other groups have recently described other important factors in the basal repression machinery of Shh signaling. Typically, mutants that disrupt cilia, such as those affecting the intraflagellar transport-B (IFT-B) complex, cause low Shh signaling in the neural tube (Goetz and Anderson, 2010; Huangfu et al., 2003). Paradoxically, mutations in the IFT-A complex subunits, despite having bulbous ciliary tips caused by defective retrograde IFT (Liem et al., 2012; Ocbina et al., 2011; Qin et al., 2011; Tran et al., 2008), result in increased Shh signaling in the neural tube. Interestingly, mutations in the tubby-like protein 3 gene (Tulp3) phenocopy mutations in IFT-A subunits and Gpr161 by causing increased Shh signaling in the caudal neural tube (Norman et al., 2009; Patterson et al., 2009). The pre-ciliary function of the IFT-A core complex in binding and ciliary trafficking of Tulp3, an adapter in gating of ciliary GPCRs including Gpr161, explains the high Shh signaling observed in IFT-A mutants, despite having abnormal cilia (Badgandi et al., 2017; Mukhopadhyay et al., 2013, 2010). Thus, IFT-A-regulated trafficking of Tulp3 and Gpr161 regulates basal suppression of Shh signaling.

During limb and skeletal development, mesenchymal cells of the limb bud, perichondrial cells, chondrocytes, osteoblasts and osteocytes are ciliated (Donnelly et al., 2008; Farnum and Wilsman, 2011; Haycraft et al., 2007; Malone et al., 2007; Xiao et al., 2006; Wilsman et al., 1980). A subset of diseases caused by primary cilia/centrosome defects (ciliopathies) are associated with skeletal phenotypes, and classified as skel­etal ciliopathies, highlighting the role of cilia in limb and skeletal development. These diseases include Sensenbrenner syndrome, Jeune syndrome or asphyxiating thoracic dystrophy (ATD), and the short rib-polydactyly group (SRPs) (Huber and Cormier-Daire, 2012). Particularly, mutations in the IFT-A complex have been associated predominantly with Sensenbrenner syndrome, a syndrome with craniofacial and ectodermal abnormalities that are distinct from other skeletal ciliopathies (Lin et al., 2013). Tissue-specific deletion of Sufu in mice increases Shh signaling in limb bud and cranial mesenchyme, affecting limb patterning (Zhulyn et al., 2014) and preventing intramembranous bone formation in skull (Li et al., 2017), respectively. However, the role of cilia in repression of Shh signaling and in causing distinct phenotypic characteristics of skeletal ciliopathies is not well understood.

Here, we show that Gpr161 regulates forelimb formation and Shh-dependent limb bud patterning, as well as endochondral and intramembranous bone formation in a cilia-dependent manner. Our results demonstrate an unexpected and crucial role played by the cilia in basally repressing the hedgehog pathway in these developmental processes, even at times when the activating morphogen is absent.

To study the role of Gpr161 during different stages of limb and skeletal development, we generated a mouse allele floxed on both sides of exon 4 (Gpr161^f/f) (Fig. S1A). We determined that embryos homozygous for global deletion using this allele (Gpr161^−/−) were lethal by E10.5, had craniofacial defects (Fig. 1A,A′), neural tube ventralization (Fig. S1B), increased Shh signaling, and decreased Gli3R levels in E9.5 whole embryo lysates (Fig. S1C), similar to the previous null allele targeting exon 3 (Mukhopadhyay et al., 2013). Thus, the present allele is a null allele, and the conditional form of the allele can be used to study tissue-specific phenotypes.

Gpr161 is broadly expressed in the limb buds (Mukhopadhyay et al., 2013), and is localized to primary cilia of forelimb bud mesenchymal cells (Fig. S1D). Interestingly, by E10.25, Gpr161^−/− embryos exhibited complete lack of forelimb buds, despite development of hindlimb buds (Fig. 1A,A′, Fig. S1E). Histological analysis confirmed the absence and presence of forelimb and hindlimb buds, respectively (Fig. 1B,B′, Movies 1-4). Expression of the fibroblast growth factor family gene Fgf10 is initially restricted in the lateral plate mesoderm destined to become forelimb (prospective forelimb field), and observed later in the distal mesenchyme of the forelimb bud (Ohuchi et al., 1997; Agarwal et al., 2003; Rallis et al., 2003). At E9.25, Fgf10 was not expressed in the prospective forelimb field of Gpr161^−/− embryos, but was expressed in forelimb buds of littermate controls (Fig. 1C). The T-box transcription factor Tbx5 is expressed in the prospective forelimb field prior to Fgf10, and is later expressed in the developing forelimb bud (Agarwal et al., 2003; Rallis et al., 2003; Sekine et al., 1999). At ∼E9.5, although forelimb buds were absent in Gpr161^−/− embryos, Tbx5 was expressed in the prospective forelimb field (Fig. 1D), unlike Fgf10 (Fig. 1C), and was expressed in forelimb buds of littermate controls (Fig. 1D). In contrast, Hoxd13 expression was evident in the hindlimb buds at E10.25, although its expression was broad and diffuse in contrast to tight posterior expression in littermate controls (Fig. 1E). Thus, knockout of Gpr161 results in lack of forelimb, despite establishment of a Tbx5-expressing prospective forelimb field.

To test the role of Gpr161 during limb bud development, we conditionally deleted Gpr161 using Prx1-Cre. Expression of Prx1-Cre is initiated in the prospective forelimb field (Hasson et al., 2007) following Tbx5 (Nishimoto et al., 2015; Minguillon et al., 2012), with later expression in both forelimb and hindlimb mesenchyme (Logan et al., 2002). Unlike Gpr161^−/−, conditional deletion of Gpr161 in forelimb fields using Prx1-Cre (Prx1-Cre; Gpr161^f/f, and hereafter designated as ‘Gpr161 cko’) did not cause defective limb formation, presumably resulting from later depletion of Gpr161 in the prospective limb field compared with the knockout.

Prior to Shh expression in the posterior forelimb bud at E9.5, mutual antagonism between Gli3R and the basic helix-loop-helix transcription factor dHand (Hand2) prepatterns the forelimb mesenchyme to cause posterior expression of bona fide Shh pathway targets and of 5′ Hoxd genes such as Hoxd13 (Te Welscher et al., 2002a). Interestingly, although Gli3 transcript levels were unchanged (Fig. S2A), consistent with a lack of Gli3R protein activity, Gli1, Ptch1 and Hoxd13 expression was expanded throughout the Gpr161 cko forelimb buds, in contrast to restricted posterior expression in littermate controls at ∼E9.5 (Fig. 2A-C′). Thus, Gpr161 knockouts exhibit prematurely expanded Shh signaling.

As Gpr161 cko demonstrated prematurely high signaling prior to Shh expression, we tested for pathway activity and limb bud patterning in Gpr161 cko after Shh expression at E9.5. By E10.5, Gpr161 cko embryos demonstrated (1) ectopic Shh expression in anterior forelimb field (Fig. 2D), (2) increased expression of Gli1 and Ptch1 in both anterior and posterior forelimb fields, as opposed to mainly posterior field expression in control littermates (Fig. 2D), (3) increased Gli1 and Ptch1 transcripts in the forelimbs compared with control littermates despite partial (∼50%) knockdown of Gpr161 transcripts (Fig. 2E), suggestive of increased Shh signaling, and (4) extended expression of the bone morphogenetic protein antagonist Grem1 and Hoxd13 into anterior limb fields (Fig. 2D) (Te Welscher et al., 2002b; Zeller et al., 2009). Patterning of the hindlimbs in Gpr161 cko was not affected until E10.5, but hindlimbs exhibited Hoxd13 expansion anteriorly by E10.75 (Fig. S2B). Similar to Gpr161 cko, the Prx1-Cre; Ptch1 conditional knockout embryos also exhibit increased Shh signaling, although forelimbs are severely stunted, and hindlimbs manifest patterning defects (Butterfield et al., 2009; Zhulyn et al., 2014). Thus, limb-specific deletion of Gpr161 resulted in prematurely expanded Shh signaling- and ectopic Shh-dependent patterning defects.

Coincident with the increased Shh signaling and lack of Gli3R that persisted into E12.5 (Fig. 3A,B), E13.5 Gpr161 cko embryos had an increased number of mesenchymal and chondrogenic condensations, resulting in an increased number of digit fields (Fig. 3C). During later embryonic development, we detected syndactyly, extra middle digits (metacarpal/metatarsal number 6-7) with lack of the first digit, bifurcated middle metacarpals/metatarsals and bifurcated/trifurcated phalanges (Fig. 3D). Other mutants with increased Shh signaling such as Prx1-Cre; Sufu ^f/− and Prx1-Cre; Ptch1^f/f also show increased digit fields (Zhulyn et al., 2014; Butterfield et al., 2009), although the later manifestations were not characterized due to late gestational lethality. Thus, increased Shh signaling in Gpr161 cko results in polysyndactylous phenotypes.

During later embryonic development in Gpr161 cko mutants, the limb long bones were severely shortened (Fig. 4A,B). As apparent from skeletal staining with Alizarin Red (stains mineralized cartilage/bone) and Alcian Blue (stains unmineralized cartilage), there was no mineralization in the forelimb long bones, even at the end of gestation (Fig. 4A,B, Fig. S3B) or shortly after birth, beyond which point the embryos did not survive. However, mineralization in the digits in the Gpr161 cko forelimbs was unaffected (Fig. 4B). The hindlimb long bones, in particular tibia and fibula, were shortened or bent, respectively, with the femur being relatively less affected (Fig. 4B). The relative strength of phenotypes in forelimb long bones with respect to hindlimb long bones is probably reflective of earlier and nearly complete deletion of Gpr161 in forelimbs with respect to hindlimbs, based on earlier Prx1-Cre expression in forelimbs (Logan et al., 2002) (Fig. S3A). In addition, fusion of ribs and sternum in the ventral midline in Gpr161 cko was absent (Fig. 4A, Fig. S3C). Thus, Gpr161 cko forearm endochondral bones lacked mineralization.

In contrast to endochondral bone formation, intramembranous bones arise directly by mesenchymal differentiation compacted into sheets, and does not require a cartilage mold (Kronenberg, 2003; Abzhanov et al., 2007). Cranial vault (calvarium) and facial bones arise directly from deep layers of the dermis via intramembranous ossification (Abzhanov et al., 2007). In the mouse, the calvarial bones posterior to the coronal suture (posterior skull), except a part of the interparietal bones, are derived from cranial mesoderm, whereas frontal bone and facial bones are derived from the cranial neural crest cells (Chai and Maxson, 2006). Interestingly, coincident with Prx1-Cre expression in cranial mesenchyme (Fig. S3A) (Logan et al., 2002; Goodnough et al., 2012), there was a complete lack of posterior skull mineralization in Gpr161 cko (Fig. 4A). However, most of the frontal, facial bones and mesoderm-derived endochondral bones in the base of the skull, such as basioccipital and basisphenoid, were not affected (Fig. 4A, Fig. S3D). Most of the scapula also ossifies by intramembranous ossification, and in Gpr161 cko scapular ossification was mostly lacking (Fig. 4B). Thus, Gpr161 cko are lacking in intramembranous bone formation.

We further investigated the reasons for the lack of mineralization in the forelimb long bones in Gpr161 cko. We detected a complete lack of trabecular long bone and bone collar formation in forelimb long bones using von Kossa staining, and noticed concomitant accumulation of cartilage using Safranin O staining at E17.5 (Fig. 5A). During endochondral bone formation, the periarticular/round chondrocytes mature into columnar chondrocytes, which further differentiate into prehypertrophic and hypertrophic chondrocytes (Kronenberg, 2003). In Gpr161 cko embryos, we noted that round chondrocytes were predominantly present throughout the forelimb long bones, and there was an almost complete lack of columnar and hypertrophic chondrocytes at E17.5 (Fig. 5A, Fig. S4A). To confirm further the steps where chondrocyte maturation is blocked, we quantified proliferation by measuring incorporation after an acute pulse of bromodeoxyuridine (BrdU). BrdU incorporation in periarticular/round chondrocytes has been shown to be lower than that in columnar chondrocytes, with no BrdU incorporation in the differentiating hypertrophic chondrocytes (Kobayashi et al., 2005, 2002). In the Gpr161 cko, BrdU-positive round chondrocytes persisted throughout the forelimb long bones (Fig. 5B), lacked the typical anatomy of columnar chondrocytes (Fig. 5A), and had BrdU incorporation rates lower than columnar chondrocytes from littermate controls (Fig. 5B). Cyclin D1 and p130 (Rbl2), one of the Rb proteins, are expressed complementary to each other in proliferating and hypertrophic chondrocytes, respectively (Fig. S4B) (Yang et al., 2003; Long et al., 2001). In line with the persistence of proliferating chondrocytes throughout the forearm long bones in Gpr161 cko, cyclin D1 was present in these chondrocytes, along with an absence of p130 (Fig. S4B). Compared with forearms, ossification in hindlimb long bones was less affected or not affected, with lack of von Kossa staining only in tibia (Fig. S4C). Thus, lack of Gpr161 prevents chondrocyte maturation beyond the periarticular/round chondrocyte stage.

We further tested the failure of maturation of chondrocytes, with respect to Ihh signaling, and osteoblast differentiation in the Gpr161 cko using radioisotopic in situ hybridization for detecting expression of relevant transcripts. Columnar chondrocytes differentiate into prehypertrophic and hypertrophic chondrocytes that express and secrete Ihh and collagen X (ColX; Col10). Ihh increases Gli1 and Ptch1 levels in proliferating chondrocytes and in adjacent perichondrium. Ihh also results in production of parathyroid hormone-like peptide (PTHrP) in periarticular cartilage, which prevents differentiation of columnar to prehypertrophic chondrocytes in a negative-feedback loop (Lanske et al., 1996; Vortkamp et al., 1996) (Fig. S5A).

Coincident with a lack of pre-hypertrophic and hypertrophic chondrocytes, there was an almost-complete lack of Ihh and ColX transcripts at E17.5 in the forearm long bones in Gpr161 cko versus littermate controls (Fig. 6A). Both Gli1 and Ptch1 transcripts were reduced in the proliferating chondrocytes and perichondrium at E17.5, consistent with lack of Ihh signaling (Fig. 6A). In addition, Ihh expression was lacking at E14.5, and expression of PTHrP (Pthlh) transcripts was also reduced (Fig. 6B, Fig. S5C). We further tested expression of the osteoblast progenitor marker Runx2 and the early osteoblast marker Osx (Sp7) in the trabecular bone and bone collar. Both transcripts were completely absent in the forearm long bones in Gpr161 cko (Fig. 6A). Interestingly, the forelimb autopods showed retention of mineralization and bone ossification (Fig. 4B, Fig. 5A), chondrocyte maturation, Ihh signaling, and osteoblastogenesis (Fig. S5B), despite similar knockdown of Gpr161 transcripts compared with the forearm, reflecting regional specificity in bone ossification programs in the forelimb. As seen with mineralization and bone ossification (Fig. 4B, Fig. S4C), the hindlimb long bones were less affected or not affected with respect to expression of Ihh, Ihh target genes, and osteoblast maturation (Fig. S5D). Thus, forearm long bones in Gpr161 cko show lack of Ihh signaling, along with an absence of Ihh-secreting hypertrophic chondrocytes. Simultaneously, there is a complete lack of osteoblastogenesis in trabecular bone and bone collar.

We confirmed the complete lack of ossification in the posterior calvarium by von Kossa staining in E17.5 Gpr161 cko embryos, but noted no compensatory accumulation of cartilage using Safranin O (Fig. 7A). In addition, using in situ hybridization we determined that Runx2 and Osx transcripts were completely missing in the calvarium of Gpr161 cko embryos, corresponding to the lack of ossification (Fig. 7B). Similar to the Runx2 knockout (Komori et al., 1997), which also lack intramembranous bone formation in calvarium, there was barely any other layer of cells between the skin and meninges. The skin layers, including the dermis overlying the missing cranium, were also dramatically thinner (Fig. 7A,B). Thus, osteoblastogenesis is completely blocked during intramembranous bone morphogenesis with no compensatory chondrogenesis in Gpr161 conditional mutants.

As Gpr161 localizes to the primary cilia, we tested the role of cilia in phenotypes resulting from Gpr161 deletion by combining with deficiencies in the IFT-B complex protein lft88 that disrupt cilia. Conditional knockout of Ift88 results in a smaller growth plate during endochondral bone formation, without affecting ossification and intramembranous bone formation (Haycraft et al., 2007; Song et al., 2007). We generated single conditional knockouts of Prx1-Cre; Ift88^f/f­ (Ift88 cko) and double conditional knockouts of Prx1-Cre; Ift88 ^f/f­; Gpr161^f/f­ (Gpr161; Ift88 dko) (Fig. 8A). Primary cilia were completely missing in chondrocytes and surrounding perichondrium in Ift88 cko and Gpr161; Ift88 dko, unlike littermate controls, which were ciliated (Fig. S6A).

Strikingly, Ift88 was genetically epistatic to Gpr161 in skeletal morphogenesis and limb patterning as described below. First, the endochondral and intramembranous bone mineralization phenotypes in Gpr161 cko are cilia dependent. Gpr161 cko showed lack of mineralization of forelimb long bones, absent posterior skull, lack of sternum and no rib fusion (Fig. 4). Remarkably, the Gpr161; Ift88 dko rescued mineralization in the forearm long bones, and rib fusion in the ventral midline (Fig. 8A,B). Unlike Gpr161 cko, which lack ossification, we detected restoration of endochondral bone formation in forelimb long bones using von Kossa staining (Fig. 8C), and concomitant generation of columnar and hypertrophic chondrocytes using Safranin O staining in the Gpr161; Ift88 dko (Fig. 8C, Fig. S6A). Thus, concomitant Ift88 deficiency prevents endochondral ossification phenotypes arising in the Gpr161 cko. Interestingly, the proliferating chondrocytes in Gpr161 cko were ciliated, irrespective of BrdU labeling and cyclin D1 expression (Fig. S6B,C). Therefore, chondrocyte proliferation in Gpr161 cko might undergo a cilia-dependent pathway. Second, intramembranous bone formation in cranium and scapula was partially restored in Gpr161; Ift88 dko. Particularly, frontal bones and most of the parietal bones were formed, with regions in the posterolateral parietal bones still lacking mineralization, but with no compensatory chondrogenesis (Fig. 8A,B). Thus, concomitant Ift88 deficiency prevents intramembranous ossification defects arising in the Gpr161 cko. Third, lack of cilia in Gpr161; Ift88 dko embryos prevented syndactyly, bifurcated metacarpals/metatarsals and bifurcated/trifurcated phalanges (Fig. 8C), as observed in Gpr161 cko autopods (Fig. 3D). Instead, Gpr161; Ift88 dko autopods had phenotypes resembling Ift88 cko, such as preaxial polydactyly in feet (Fig. 8C) (Haycraft et al., 2007). Thus, patterning defects in Gpr161 cko limbs are cilia dependent.

The importance of basal suppression of Shh pathway, its interaction with Ihh signaling, and the role of cilia-dependent signaling during limb and skeletal morphogenesis is not well understood. Here, using detailed phenotypic analysis of germline and conditional Gpr161 knockouts, we uncover multiple crucial steps that are regulated by Gpr161. First, we show that forelimbs are not formed in Gpr161 knockouts, despite establishment of Tbx5-expressing prospective limb fields. Second, we show that limb-specific deletion of Gpr161 caused premature expansion of Shh signaling resulting from lack of Gli3R activity. Gpr161 deletion also caused ectopic Shh expression and increased pathway activity. Defective limb bud patterning from increased Shh signaling caused polysyndactyly affecting middle digits. Third, we demonstrate that endochondral bone formation (both bone collar and trabecular bone) in forearm was severely affected upon limb-specific Gpr161 deletion. Proliferating round/periarticular-like chondrocytes failed to differentiate into columnar chondrocytes and accumulated in forearms, along with a corresponding absence of Ihh signaling. Thus, Gpr161 inhibits periarticular chondrocyte proliferation. Fourth, we show that ossification in posterior skull and scapula were disrupted in Gpr161 conditional knockouts, suggesting that Gpr161 promotes osteoblastogenesis during intramembranous bone formation. Finally, we demonstrate that defects in limb patterning, endochondral and intramembranous skeletal morphogenesis were suppressed in the absence of cilia, indicating that the pathways affected upon Gpr161 deletion are cilia-dependent (Fig. 9).

Limb bud formation is orchestrated through the following stages (Duboc and Logan, 2011): (1) an induction stage during which axial cues and a combinatorial Hox code confers limb-forming potential to the lateral plate mesoderm; (2) an initiation stage during which Tbx5 or Tbx4 are expressed in the presumptive forelimb- or hindlimb-forming areas, respectively, followed by Fgf10; and (3) an outgrowth stage when a stable positive-feedback loop between mesenchymal Fgf10 and ectodermal Fgf8 is established. Prx1-Cre; Tbx5^f/f has been shown to lack forelimbs (Rallis et al., 2003), whereas Fgf10 is required for both forelimb and hindlimb development (Sekine et al., 1999; Ohuchi et al., 1997). Although Fgf10 has been proposed to be a direct transcriptional target of Tbx5 (Agarwal et al., 2003; Ng et al., 2002), other inputs, including retinoic acid, also regulate Fgf10 expression (Nishimoto et al., 2015).

The specificity of Gpr161 knockouts in preventing formation of forelimbs, but not of hindlimbs, rules out indirect effects affecting forelimb formation. As Tbx5 is expressed in the prospective forelimb field in Gpr161 knockouts, unlike Fgf10, Gpr161 is likely to function in forelimb initiation or outgrowth at a step that facilitates the function of Tbx5, upstream of Fgf10. The most parsimonious model would be that Gpr161 functions by promoting Gli3R generation and preventing premature Shh signaling in the prospective forelimb field (Fig. 9A). Alternatively, Gpr161 might be affecting other cellular pathways in preventing forelimb bud formation (Feigin et al., 2014).

The role of restricting premature Shh signaling in the prospective forelimb field during forelimb initiation has not been formally tested. First, premature activation of Shh pathway by deleting Ptch1 using Prx1-Cre, which expresses after Tbx5 (Nishimoto et al., 2015; Minguillon et al., 2012), results only in forelimb outgrowth defects stemming from a lack of Gli3R-dependent specification of anterior progenitors (Butterfield et al., 2009; Zhulyn et al., 2014). Second, knockouts of other negative regulators of the Shh pathway, such as Ptch1 and Sufu, arrest by E9-9.5 (Goodrich et al., 1997; Svärd et al., 2006). The embryonic lethality of Gpr161 knockouts at E10.5, at an age after limb buds are established, allows a window of opportunity for looking into the role of suppression of premature Shh signaling in forelimb bud formation. However, unlike Gpr161 knockouts, Gli3 knockout mice possess forelimbs (Litingtung et al., 2002). Thus, a combination of a lack of Gli3R formation and increased GliA generation might be necessary for preventing forelimb formation (Fig. 9A).

Gpr161 cko forearm long bones were composed of proliferating chondrocytes similar to periarticular/round chondrocytes, as determined by BrdU incorporation and morphology. This could result from a direct role of Gpr161 in preventing periarticular/round chondrocyte proliferation. Alternatively, lack of Ihh formation in Gpr161 cko could prevent differentiation into columnar chondrocytes (Kobayashi et al., 2002, 2005). Gli3, particularly the Gli3R form, has been proposed to function as a repressor of Ihh-mediated differentiation of periarticular into columnar chondrocytes, rather than in earlier chondrocyte proliferation (Koziel et al., 2005). Of note, Gli3; Ihh double knockouts rescue the lack of columnar chondrocyte formation in Ihh knockouts (Koziel et al., 2005). Gpr161 loss prevents both Gli3R and Ihh formation, but the phenotype in Gpr161 cko is unlike that of Gli3; Ihh double knockouts. Thus, Gpr161 functions as a rheostat in preventing proliferation of periarticular chondrocytes, upstream of Ihh-mediated and Gli3R-suppressed differentiation of periarticular to columnar chondrocytes (Fig. 9C).

Gpr161 cko forearm long bones showed a complete lack of Runx2-expressing osteoblast progenitors and osteoblastogenesis. Ihh determines osteoblast differentiation during bone collar formation in endochondral bone (St-Jacques et al., 1999; Long et al., 2004). However, absence of bone collar formation is more severe in Gpr161 cko forelimbs compared with Ihh knockouts or conditional/mosaic knockouts of Smo, the activator of the hedgehog pathway (Long et al., 2004). The persistence of periarticular/round chondrocyte proliferation along with lack of Ihh signaling might explain the severity of lack of bone ossification in the Gpr161 cko. Less severe bone collar defects in Col2a1-cre; Ptch1^f/f mutants (Mak et al., 2006) with respect to Gpr161 cko mutants could result from inefficient knockout in perichondrium or developing periosteum in the former.

Lack of skull ossification in Gpr161 cko is similar to that observed in Runx2 and Osx knockouts (Otto et al., 1997; Komori et al., 1997; Nakashima et al., 2002) and Prx1-Cre; Runx2 conditional knockouts (Takarada et al., 2016). In Gpr161 cko, Runx2 and Osx expression in the presumptive calvarium was inhibited. Therefore, the promotion of osteoblastogenesis by Gpr161 might initiate in calvarial mesenchymal progenitors that precede the appearance of Runx2-expressing pre-osteoblasts (Abzhanov et al., 2007). The skin layers, including the dermis overlying the missing cranium, were also dramatically thinner in Gpr161 cko. Wnt ligands released from the epidermis regulate both dermis condensation and cranial intramembranous bone specification (Goodnough et al., 2012; Chen et al., 2012). Mutants with conditional deletion of β-catenin lack intramembranous bone, but have compensatory chondrogenesis (Hill et al., 2005; Day et al., 2005; Tran et al., 2010). Rather, the simultaneous lack of intramembranous bone morphogenesis and compensatory cartilage formation in posterior skull in Gpr161 cko is similar to overactivation of β-catenin in cranial mesenchyme (Goodnough et al., 2012). Future work should focus on understanding the link between Gpr161 and β-catenin signaling in regulating intramembranous bone formation, compensatory chondrogenesis, and cranial dermis development.

The sustained proliferation and accumulation of periarticular/round chondrocytes seen in Gpr161 cko was suppressed in the absence of cilia, indicating that the chondrocyte proliferation step is likely to be cilia dependent. Hedgehog signaling targets such as Gli1 and Ptch1 were not upregulated in the periarticular chondrocytes. Periarticular chondrocytes possess cilia embedded in the ciliary pocket and are surrounded by the cartilaginous extracellular matrix, raising the possibility that certain unknown, possibly mechanosensory, stimuli might regulate chondrocyte proliferation (Malone et al., 2007; Xiao et al., 2006) (Fig. 9C).

The lack of intramembranous bone formation in Gpr161 cko was suppressed in the absence of cilia. Thus, Gpr161 blocks a cilia-generated signaling pathway inhibiting osteoblastogenesis during intramembranous bone formation. Premature Ihh/Shh signaling could be the cilia-gener­ated pathway that inhibits intramembranous osteoblastogenesis (Fig. 9D). First, conditional knockouts of negative regulators of the Shh pathway, Ptch1 (Mak et al., 2006), Sufu (Li et al., 2017) and Gpr161, show lack of skull formation. Deletion of Sufu in the cranial neural crest activates Gli1 levels in cranial mesenchyme (Li et al., 2017). Second, simultaneous deletion of Gli2 restores calvarial bone formation in the Sufu conditional knockouts (Li et al., 2017). Finally, Ihh and Shh are expressed in the osteogenic front of the developing intramembranous bone (Kim et al., 1998; Lenton et al., 2011), and, by direct binding to Ptch1 expressed in pre-osteoblasts and chondrocyte-like osteoblasts, prevent further osteoblastogenesis (Abzhanov et al., 2007).

Overall, studying Gpr161 mutant phenotypes provides the molecular, subcellular and cellular resolution required for understanding cilia-dependent processes in limb bud formation, chondrocyte proliferation and intramembranous osteoblastogenesis.

Targeting of the fourth exon of Gpr161 (NM_001081126.1) by homologous recombination in mouse ESCs of the C57BL/6 strain was carried out by EUCOMM. The ESCs were injected into host embryos of the C57BL/6 albino strain by the transgenic core (Dr Robert Hammer's laboratory, UT Southwestern Medical Center, Dallas, TX, USA). The mutant germline allele was crossed with germline Flp-O (Jackson Laboratory, stock #012930) for deleting the FRT-LacZ-Neo-FRT cassette to generate the exon 4 floxed allele (Fig. S1A). This floxed line was crossed with Prx1-Cre (Logan et al., 2002; Jax strain #005584). The targeted recombination results in deletion of most of the fourth exon, except its initial 461 bp. This results in truncation of Gpr161 after its initial 153 amino acids (NP_001074595.1). Simultaneously, crossing with CAG-Cre recombinase line (Sakai and Miyazaki, 1997), in which Cre is expressed ubiquitously, generated the Gpr161 knockout allele (Fig. S1A). Genotyping of Gpr161 alleles was performed using primers in the deleted fourth exon (P1; 5′ CAAGATGGATTCGCAGTAGCTTGG), flanking the 3′ end of the deleted exon (P2; 5′ ATGGGGTACACCATTGGATACAGG), and in the Neo cassette (P3, 5′ CAACGGGTTCTTCTGTTAGTCC). Wild-type, floxed and knockout bands were 816, 965 and 485 bp, respectively (Fig. S1A). Double mutant analysis was performed using Ift88 conditional allele (Haycraft et al., 2007; Jax strain #022409). Yolk sac DNAs were used for genotyping embryos. Noon of the day on which a vaginal plug was found was considered E0.5. All the animals in the study were handled according to protocols approved by the UT Southwestern Institutional Animal Care and Use Committee, and the mouse colonies were maintained in a barrier facility at UT Southwestern, in agreement with the State of Texas legal and ethical standards of animal care.

The affinity purified polyclonal antibody against Gpr161 was described previously (Pal et al., 2016) (1:200). Other commercial antibodies and reagents are described in supplementary Materials and Methods.

Primary cell culture, reverse transcription and quantitative PCR were performed according to standard protocols and are described in supplementary Materials and Methods. Embryos were processed for Gli1/3 immunoblotting as described previously (Wen et al., 2010).

Antisense riboprobes were made using the following templates: Ptch1, Gli1, Shh, Ihh (from Andrew McMahon's lab, University of Southern California, CA, USA; and from Deanna Grant, Andrew Peterson's lab, Genentech, South San Francisco, CA, USA), Tbx5 (from Virginia Papaioannou's lab, Columbia University, NY, USA), Sox9, Col2a1 (from Steven Vokes lab, UT Austin), Fgf10, Hoxd13 (from Xin Sun lab, University of Wisconsin, Madison) (Sun et al., 2002), Osx, ColX (from Rhonda Bassel-Duby, Eric Olson Lab, UT Southwestern Medical Center, Dallas), Runx2 (from Yingzi Yang lab, Harvard School of Dental Medicine), and PTHrP (from Henry Kronenberg lab, Massachusetts General Hospital). Whole-mount in situ hybridization using digoxigenin-labeled probes was performed on embryos using standard protocols. Images were acquired using a Leica stereomicroscope (M165 C) with digital camera (DFC500) or Zeiss stereomicroscope (Discovery.V12) and AxioCam MRc.

Radiolabeled sense and antisense probes were generated by Sp6, T3 or T7 RNA polymerases and ³⁵S-UTP (>1000 Ci/mmol; NEG039H, PerkinElmer LAS Canada) using linearized cDNA templates by in vitro transcription using the Maxiscript kit (AM1324 M, Life Technologies). Radioisotopic in situ hybridization was performed as previously described (Shelton et al., 2000). Briefly, 5-µm-thick sections were deparaffinized, permeabilized and acetylated prior to hybridization at 70°C with riboprobes diluted in a mixture containing 50% formamide, 0.75 M NaCl, 20 mM Tris-HCl, pH 8.0, 5 mM EDTA, 10 mM NaPO₄, 14% dextran sulfate, 1× Denhardt's, and 0.5 mg/ml tRNA. Following hybridization, the sections were rinsed with increasing stringency washes, subjected to RNAse A (2 µg/ml, 30 min at 37°C) and dehydrated prior to dipping in K.5 nuclear emulsion gel (AGP9281; Ilford, UK). Autoradiographic exposure was conducted for 21 days to 35 days. Photographic development was carried out with D-19 Developer Substitute and Kodak Fixer (26920-4; 26942, Ted Pella). Sections were counterstained with Hematoxylin, dehydrated with ethanol, cleared with xylene, and cover slipped with synthetic mounting media (SP15, Fisher Chemical). Radioisotopic in situ hybridizations were analyzed using darkfield and brightfield microscopy. Sense (control) riboprobes established the level of background signal. Review and photography of all radioisotopic in situ hybridizations were carried out on a Leica DM2000 photomicroscope equipped with brightfield, and incident-angle darkfield illumination. Photomicrography was achieved using this microscope and an Optronics Microfire digital CCD color camera using PictureFrame 3.0 acquisition software (Optronics). The resulting ISH silver-grain signal was imaged with camera settings to produce near binary intensity and contrast. The ISH signal was pseudocolored red, and then overlaid to their concomitantly imaged brightfield image using Adobe Photoshop CS4 (Adobe Systems).

Skeletal preparations were made by a slight modification of the Alcian Blue/Alizarin Red staining procedure described by Kessel et al. (1990). Specimens were fixed in 99% ethanol for 24 h (embryos older than E15 were first de-skinned and eviscerated), and then kept in acetone for another 24 h. Incubation in staining solution (1 volume of 0.3% Alcian Blue in 70% ethanol, 1 volume of 0.1% Alizarin Red S in 96% ethanol, 1 volume of absolute acetic acid, and 17 volumes of 70% ethanol) was performed for 2-3 days at 37°C. Samples were rinsed in water and kept in 1% potassium hydroxide/20% glycerol at 37°C overnight, with additional incubation at room temperature until complete clearing. For long-term storage, specimens were transferred into 50%, 80% and finally 100% glycerol. Images were acquired using a Leica stereomicroscope (M165 C) with digital camera (DFC500).

Von Kossa stain for calcification of mineralized cartilage/bone and Safranin O stain for cartilage were performed according to standard methods (Sheehan and Hrapchak, 1980; Bancroft and Stevens, 1990). In brief, von Kossa slides were deparaffinized, impregnated with 5% silver nitrate, developed with 5% sodium thiosulfate, and then counterstained with Nuclear Fast Red. Safranin O slides were deparaffinized, stained with Weigert's iron Hematoxylin, differentiated in acid-alcohol, counterstained with 0.2% Fast Green, de-stained with 1% acetic acid, and cartilaginous mucopolysaccharides colorized with 0.1% Safranin O before final differentiation with 95% ethanol. Following final differentiations and washes, von Kossa and Safranin O slides were dehydrated, cleared, and coverslipped with synthetic mounting media.

Dams were injected intraperitoneally with 25 mg/kg BrdU and embryos collected 3 h post-injection. Embryos or limbs for histology were fixed in 20 volumes of freshly prepared 4% paraformaldehyde/PBS pH 7.4 (PFA). Harvested embryos were paraffin processed following PFA fixation. Following 2 N hydrochloric acid denaturation (BrdU) or pH 6.0 citra-based heat antigen-retrieval, serial sections were quenched of autofluorescence with 100 mM glycine and blocked against endogenous mouse IgG and secondary antibody host-serum affinity by utilizing commercially available blocking reagents (Vector Mouse on Mouse Kit, BMK-2202). Sections were incubated overnight at 4°C with primary antibody (anti-BrdU, 1:25 or other antibodies). For BrdU staining, subsequent biotin/streptavidin-fluorescein detection of bound primary was conducted the following day according to MOM kit instructions. Nuclei were counter stained with propidium iodide (5 µg/ml) prior to coverslipping with Vectashield (Vector Laboratories). Immunofluorescence of cultured cells and embryo cryosections was performed according to standard protocols after fixation in 4% PFA. The coverslips or cryosections were mounted using Fluoromount-G (Southern Biotech). Images were acquired on a Zeiss AxioImager.Z1 microscope, a PCO Edge sCMOS camera (BioVision Technologies), and PlanApochromat objectives (10×/0.45, 40×/1.3 oil, 63×/1.4 oil), controlled using Micromanager software at room temperature. Between 8 and 20 z sections at 0.5-0.8 µm intervals were acquired. Maximal projections from images of stacks were exported from ImageJ/Fiji using a custom written macro (Marcel Mettlen, Schmid lab, UT Southwestern Medical Center, and available upon request) using similar parameters (image intensity and contrast) for image files from the same experiment. Stereo images of embryos were taken on the Zeiss SteREO Discovery V.12 microscope using the 0.63× lens with AxioVision software and LEICA S8AP0 with LAS V4.8 software. Scanning electron microscopy was performed according to standard protocols and is described in supplementary Materials and Methods.

Statistical analyses were performed using Student's t-test for comparing two groups or Tukey's post-hoc multiple comparison tests between all possible pairs using GraphPad Prism.

We thank UT Southwestern's transgenic, molecular pathology, and electron microscopy cores for transgenic mouse generation, tissue processing and microscopy facilities. We thank the mouse animal care facility for animal care. We acknowledge gifts of reagents from Tamara Caspary, Andrew McMahon, Deanna Grant, Andrew Peterson, Virginia Papaioannou, Steven Vokes, Xin Sun, Rhonda Bassel-Duby, Eric Olson, Yingzi Yang and Henry Kronenberg. We thank Sandra Schmid, Issei Shimada, Fred Grinnell and anonymous reviewers for comments on the manuscript, and Jacek Topczewski and Peter Michaely for discussions.