Phenome-based approach identifies RIC1-linked Mendelian syndrome through zebrafish models, biobank associations, and clinical studies

Abstract

Discovery of genotype-phenotype relationships remains a major challenge in clinical medicine. Here, we combined three sources of phenotypic data to uncover a novel mechanism for rare and common diseases resulting from collagen secretion deficits. Using zebrafish genetic screen, we identified the ric1 gene to be essential for skeletal biology. Using a gene-based phenome-wide association study (PheWAS) in the EHR-linked BioVU biobank, we show that reduced genetically determined expression of RIC1 is associated with musculoskeletal and dental conditions. Whole exome sequencing (WES) identified individuals homozygous-by-descent for a rare variant in RIC1, and, through a guided clinical re-evaluation, they were discovered to share signs with the BioVU-associated phenome. We named this novel Mendelian syndrome CATIFA (Cleft lip, cAtaract, Tooth abnormality, Intellectual disability, Facial dysmorphism, ADHD), and revealed further disease mechanisms. This gene-based PheWAS-guided approach can accelerate the discovery of clinically relevant disease phenome and associated biological mechanisms.

Introduction

Biomedical research seeks to understand the mechanisms underlying normal biology and what goes awry in disease, with an aim to predict and prevent disease occurrence and to deliver personalized, precision health care¹. Despite decades of studies in animal models of individual gene knockouts, Mendelian disease positional cloning, and Genome-Wide Association Studies (GWAS) to unravel complex diseases, progress towards this goal has been slow. The question is whether these three, typically independent, approaches can be integrated to accelerate discovery of common and rare disease mechanisms and unknown biology.

Collagens are essential components of extracellular matrices and their disrupted secretion has been linked in patients and animal models to general skeletal and craniofacial dysmorphology^2-5. Large extracellular matrix (ECM) cargos, such as procollagen, are synthesized in the endoplasmic reticulum (ER), post-translationally modified in the Golgi and trafficked out to the ECM⁶.

Zebrafish have been instrumental in identifying genes acting at each of these steps^7-9. Variants in proteins at ER-to-Golgi secretory pathway, including the coat proteins II (COPII) SEC23A and SEC24D^10-13 and procollagen modifying enzymes^14,15 cause skeletal and craniofacial defects in zebrafish. SEC23A mutations lead to cranio-lenticulo-sutural-dysplasia (CLSD), a disease characterized by craniofacial and skeletal defects¹⁶. These studies established zebrafish as a powerful tool to study procollagen transport and model skeletal conditions. Though ER-to-Golgi transport of procollagen is relatively well-studied, how procollagen is transported from the Golgi to plasma membrane ^17,18 and the medical phenome¹⁹ resulting from dysfunction of this process remains a long-standing knowledge gap.

We show here that Ric1 and its binding partner, Rgp1, are required to activate Rab6a for procollagen transport through the Trans Golgi Network (TGN) and skeletal development in zebrafish models. We investigated human phenotypes associated with the genetically reduced expression of RIC1 in phenome-linked DNA biobanks. Clinical re-evaluation of subjects previously found to be homozygous-by-descent for a RIC1 variant and zebrafish knockouts validated common-disease phenome in these subjects, including abnormal tooth development and attention deficit hyperactivity disorder (ADHD). This discovery allowed us to describe a novel genetic syndrome, now termed CATIFA, and establish the mechanistic continuum between individual signs of a Mendelian disease and complex traits.

Results:

RIC1 is required for normal skeletogenesis

In search for novel components of the procollagen secretory pathway with essential roles in skeletal biology, we characterized the zebrafish craniofacial round (rnd ^{m641, m713, m715}) mutations, which were identified in a large-scale chemical mutagenesis screen²⁰. We genetically mapped and positionally cloned the round locus to identify mutations in the ric1 gene (KIAA1432, ENSDARG00000063362 in Zv9), (Fig. 1a; Extended Data Fig. 1a-c). By direct sequencing of cDNAs from the three independent rnd alleles, we identified a missense mutation in a conserved residue (R882C) in rnd^m641; a nonsense mutation (Q64X) in rnd^m713; and loss of splice donor site between exons 2 and 3, resulting in a frameshift mutation (A84fs*17) in rnd^m715 (Fig. 1a). Ric1 is a highly evolutionarily conserved protein sharing 71% identity from zebrafish to human. The yeast and human homologs of Ric1 protein and its binding partner Rgp1 were shown to act as a guanine nucleotide exchange factor (GEF) for Rab6 GTPase^21,22. However, the role of Ric1-Rgp1-Rab6a in the context of vertebrate development and physiology has not been established.

Figure 1. Ric1 is required for craniofacial skeleton development and shape.

a, Zebrafish Ric1 protein is highly conserved with 81% similarity to human RIC1 (Clustal Omega, EMBL-EBI). Positional cloning identified mutations in round alleles (rnd ^m641, rnd ^m713, rnd ^m715; see arrows), Rab6 interacting region (blue bar, also Extended Data Fig. 1). b, Live images of ric1^−/− mutant larvae show micrognathia (arrowhead), small head and shortened trunk (dashed line to arrow). c, Box-whisker plot for quantification of head size index (head to total body length ratio) as a measure of microcephaly. Mann-Whitney U-test (two-tailed) was used for statistical comparison; 95% confidence interval (CI). Boxes show 25^th to 75^th percentiles and medians. Whiskers are min to max values. d, Alcian blue and Alizarin red staining of the ric1^−/− mutant shows shorter, malformed Meckel’s cartilage (arrowheads) and shorter, kinked pectoral fins (arrows). All three alleles exhibit indistinguishable phenotypes and fail to complement one another in genetic complementation assays. e, Live images of ric1^−/− mutant embryos (5 dpf, days post-fertilization) overexpressing human RIC1 mRNA, show rescue of jaw protrusion (arrowheads) and elongation of the body length (arrows). Quantification of the rescue experiments, head in f, and body length in g. Statistical comparison by two-tailed Mann-Whitney U-test, CI = 95%, n=3 independent animals for ric1^−/− and WT; n=7 animals for hRIC1 rescue group. h, Hyosymplectic cartilage (HC) dysmorphology shown in double transgenic Tg(Col2a1a:caax-EGFP; Col2a1a:H2A-mCherry) zebrafish that marks plasma membrane (green) and nucleus (red), white bar marks the symplectic arm (SA); maximum intensity projections of live z-stack images by confocal microscopy in WT and ric1^−/− embryos (n=3 independent animals per group). i, 3D rendered structures of HC shapes highlight the malformed and short cartilage shape in ric1^−/−. j, Magnified views of SA chondrocytes. 3D rendered volumes are overlaid onto maximum intensity projection views. k, 3D rendering underscores the dysmorphic shape and reduced volume of ric1^−/− chondrocytes. l, Volume measurements of 5 cells per genotype by Imaris 8, color-coded to match the cells in k. Lines indicate mean and standard deviations in f, g and l.

The zebrafish ric1^−/− mutations present with flattened head, lack of jaw protrusion, kinked pectoral fins and short head/body length index at 5 days post-fertilization (dpf) (Fig. 1b). The microcephaly-like phenotype was corroborated by smaller head size indices compared to age-matched wild-type (WT) siblings (Fig. 1c). Glycosaminoglycan (GAG) staining of sulfated and carboxylated glycoproteins by Alcian blue and Alizarin red revealed that craniofacial cartilage elements and newly formed ossification centers are present in ric1 mutant embryos, but they are malformed and smaller than WT controls (Fig. 1d, Extended Data Fig. 1d,e). Human RIC1 (hRIC1) overexpression by mRNA was sufficient, in genetic replacement experiments, to restore jaw protrusion (Fig. 1e), head and body length in ric1^−/− larvae (Fig. 1f,g), corroborating functional conservation of zebrafish and human RIC1 genes in skeletal development.

To characterize the craniofacial architecture, we examined fluorescently marked transgenic animals by live imaging (Fig. 1h). We analyzed the organ shape of the jaw-supporting hyosymplectic cartilage (HC), an easily identifiable key-like element accessible for in vivo imaging. We imaged the entire HC using confocal microscopy (Fig. 1h), and created 3D reconstructions (Fig. 1i), which showed irregularities including thicker and misshapen arm and body of the HC, also a collapsed foramen, an anatomical passage point (Supplementary video 1,2). High magnification imaging of the single-cell layered HC arm (Fig. 1j,k) revealed smaller, constricted mutant chondrocytes, especially around the midline, and loss of the characteristic stack-of-coins shape (Supplementary video 3,4). Volume measurements of five HC chondrocytes at the corresponding positions of age-matched zebrafish showed reduced volume ~185 μm³ in mutants, compared to average WT volume of ~350 μm³ (Fig. 1l). We conclude that Ric1 function is required for normal cell shape that further conveys organ shape and, ultimately, the overall craniofacial morphology.

BioVU analyses revealed craniodental and skeletal phenotypes associated with genetically predicted reduced RIC1 expression

To compare the zebrafish ric1^−/− phenotype to the corresponding human common-disease phenome, we examined the association of genetically predicted reduced expression of RIC1 in the BioVU Electronic Health Records (EHR) and DNA biobank^23-25. For this analysis, we applied PrediXcan²⁶ which estimates genetically regulated gene expression (GReX) from the publicly available GTEx^27,28 project data as a reference panel. We tested the association of GReX with phecodes²⁵ in ~10,000 subjects of European ancestries in BioVU to determine gene-level associations to disease traits (Fig. 2a). The RIC1 prediction model includes single-nucleotide polymorphisms (SNPs) in the region (within 1 Mbp of start and stop sites) of the gene (Supplementary Table 1). In a traditional PheWAS, one SNP is used to test association; here, we used a variation of the method, which we called “gene-based PheWAS”, because multiple SNPs are used around the gene-of-interest. We focus here on all traits (Phecodes) nominally significantly associated with reduced GReX of RIC1, including skeletal, respiratory (asthma), sensory organs (eye, e.g., corneal opacity), nervous (ADHD), digestive (diverticulosis), and cardiovascular (stricture of artery) (Fig. 2b,c and Supplementary Table 2).

Figure 2. Human common-disease phenome significantly associated with genetically determined reduced RIC1 expression in BioVU biobank and CATIFA patient phenome.

a, Design of gene-based PheWAS study conducted in BioVU biobank. b, Summary graph of traits (phecodes) significantly associated with predicted reduced expression of RIC1 in patients from BioVU biobank, as analyzed by PrediXcan algorithm; complete dataset and sample numbers are in the Supplementary Table 2. Traits are categorized into systems (y-axis), and significance is displayed on x-axis. Significance is tested by logistic regression analysis (two-sided). Multiple testing adjustment is done using Bonferroni correction. c, Individual skeletal phecodes from panel b are listed on y-axis. ‘Number of cases / total cases analyzed’ is indicated as insets within bars. x-axis shows significance levels (logistic regression, two-sided, Bonferroni-corrected). d, SNPs in the RIC1 gene are highly significantly associated, in the UK Biobank, with RIC1 phenome observed in the BioVU biobank. Representative traits and the p-value for the most significant SNP in RIC1 plotted; complete list of SNP-level and gene-level associations can be found in Supplementary Table 4. e, Pedigree of a large multiplex, consanguineous family with affected individuals shaded. Standard pedigree symbols used, genotypes are listed for tested individuals. Het: Heterozygous, Homo: Homozygous. f, Photographs of patients with CATIFA syndrome. Written consent for the use of photographs was obtained from the parents of affected individuals. g, Human Phenotype Ontology heat map of patients’ common clinical features. Table of complete clinical description for each subject is presented in Supplementary Table 5. h, Linkage analysis and LOD score value of CATIFA syndrome-linked RIC1 c3794G>C variant on chromosome 9 (arrow). i, Schematic of RIC1 protein marking the R1265 variant site within the Rab6 binding region. j, Translated RIC1 protein sequence is highly conserved among vertebrates; human, mouse and zebrafish, including R1265 residue (arrowhead).

We subsequently sought evidence for replication of the BioVU-associated phenome through parallel analyses on GReX of RIC1 and on SNPs at the gene in the UK Biobank²⁹ (UKB), and other publicly available GWAS data (Figure 2d, Supplementary Table 3 and 4). Asthma is significantly associated with reduced GReX of RIC1 in UKB after correction for the number tissues and phenotypes, and in data from other independent GWAS (Fig. 2d, Extended Data Fig. 2a-c, Supplementary Table 3 and 4). Association studies in phenome available in UKB reveal SNPs at RIC1 are highly significantly associated with at least one phenotype in each of the following phenotypic categories: skeleton/joint, respiratory, eye, nervous, digestive, cardiovascular and blood, neoplasm and infection (Fig. 2d, Supplementary Table 4).

The most notable skeletal BioVU phecodes included ‘fracture of unspecified bones’, ‘acquired deformities of limbs’, and ‘disorders of tooth development’, specifically, ‘tooth eruption and development’ (Fig. 2c). The PheWAS findings drew our attention to zebrafish dentation, easily visible in older larvae at 7 dpf. Alizarin red staining (calcified tissue) showed the presence of pharyngeal teeth and dermal bones such as cleithrum (pectoral girdle) in wild types, whereas ric1^−/− larvae lacked pharyngeal teeth, and had smaller calcification domains that failed to elongate (Extended Data Fig. 1d,e). These findings confirmed a highly evolutionarily conserved function of RIC1 in skeletal biology and tooth development. They also revealed parallels between zebrafish and the phenotypic spectrum in human subjects with reduced GReX of RIC1.

CATIFA syndrome is characterized by cleft lip, cataract, tooth abnormality, intellectual disability, facial dysmorphism, and ADHD

Prior whole exome sequencing (WES) of a pediatric cataract cohort identified a missense mutation in the RIC1 gene³⁰; patients with a variant in RIC1 were initially characterized as having craniofacial anomalies and developmental delay, in addition to pediatric cataracts. With limited evaluation of these patients for other organ systems and mounting evidence from the PheWAS and the animal model data supporting a more complex phenotype of RIC1 deficiency, we pursued clinical re-evaluation of the patients guided by the medical phenome associated with the reduced GReX of RIC1 in BioVU.

We identified 8 persons in two extended, consanguineous families (Fig. 2e, Extended Data Fig. 2d and Supplementary Table 5) homozygous for the missense mutation in RIC1 and having a range of shared signs and symptoms. All pediatric patients share distinct craniofacial dysmorphology including elongated face, short broad upturned nose with anteverted nares, and long philtrum. Five of the eight children had cleft lip/palate, and all had tooth eruption and/or alignment deficits with extensive caries (Fig. 2f,g). Neurological evaluation revealed global developmental delay and intellectual disability, manifesting with motor, speech, and cognitive deficits ranging from mild to severe. All males show behavioral abnormalities with mild to severe ADHD. Sensory organ examination revealed visual impairment due to cataract, strabismus, and poor visual tracking (Fig. 2f,g and Supplementary Table 5). Other symptoms included sleep disturbance (5/8), bronchial asthma (5/8), small ears (5/8), and clumsy, hypotonic gait (8/8). We collectively refer to these features as CATIFA syndrome (Cleft lip, cAtaract, Tooth abnormality, Intellectual disability, Facial dysmorphism, ADHD). The clinical data of CATIFA patients closely matched the human phenome predicted by PheWAS in BioVU, corroborating the range of medical signs and symptoms associated with RIC1. It was indeed the PheWAS prediction that helped identify the core CATIFA syndrome characteristics.

Using genetic linkage mapping, we had initially linked the CATIFA phenome to a mutation on Chromosome 9 with a supporting LOD score of 3.44 (Fig. 2h). Genotyping confirmed a single nucleotide substitution (c.3794C>G), predicted to cause a missense mutation R1265P in the RIC1 protein (Fig. 2i,j and Extended Data Fig. 2e) that fully segregated with the phenotype in each of the two families in a strictly autosomal recessive fashion. We sequenced cDNA obtained from patient fibroblasts and discovered two types of transcripts. The first is a missense mutation R1265P followed by correct splicing of intron 24. The second transcript carries the R1265P mutation and fails to splice due to intron retention, leading to a stop codon at position 1266, which likely undergoes nonsense-mediated decay (NMD).

Ric1 is an essential component of the procollagen secretory pathway

We sought to identify biological function of Ric1 in vertebrates and the mechanisms underlying CATIFA syndrome. Since Ric1 was implicated in cargo traffic²¹, and mutations described here affect highly secretory chondrocytes, we began by examining cellular localization of procollagen, the predominant secreted cargo in skeletal tissues. By antibody labeling, we found that ric1-deficient zebrafish chondrocytes retain procollagen II (Col2) intracellularly in large inclusions, while WT cells secrete it across the plasma membrane to the ECM (Fig. 3a,b). Quantification showed that intracellular collagen normally occupies <5% of the cytosolic area, whereas in ric1^−/− chondrocytes it takes up ~10 times more surface area (Fig. 3c). We found similar results in fibroblast-like, notochord sheath cells (part of the axial hydrostatic skeleton of vertebrate embryos), accompanied by malformation of notochord basement membrane (BM) (Fig. 3d, Extended Data Fig. 3a,b). Procollagen retention in sheath cells and craniofacial chondrocytes suggests that secretion defect is not cell-type specific, but may be cargo-selective. Thus, we analyzed other cargos and found that they are normally trafficked (Extended Data Fig. 3c-e). We conclude that Ric1 is selectively required for procollagen trafficking.

Figure 3. Ric1 modulates procollagen transport.

a, Experimental design for immunohistochemistry (IHC) analysis of chondrocytes in the head. b, Representative images of Tg(Col2a1a:caax-EGFP) transgenic zebrafish stained with Col2 antibody (14 μm-thick cryosections) in WT and ric1^−/− cartilage, (Collagen 2, magenta; EGFP, green). In WT Col2 signal is in extracellular space, outside caax-EGFP boundaries (dashed line), whereas in ric1^−/− collagen accumulates in chondrocytes (inside caax-EGFP boundaries, arrows). c, d, Quantification of the percentage of cytoplasmic area occupied by Col2 signal in chondrocytes (c), and notochord sheath cells (d). For chondrocytes: N=3 embryos per group, n=30 cells per group. Notochord sheath cells: N=4 embryos per group, n=21 cells for WT, n=35 cells for ric1^−/−. Data in c, d were analyzed with Mann-Whitney U test, two-tailed, CI= 95%. Mean and SD values are indicated with bars. e, Gene expression levels by RNA-seq from live, FACS-sorted chondrocytes, columns indicate mean values, bars show SD. RPKM: reads per kilobase of transcript per million mapped reads. 2 independent samples analyzed per genotype (WT and ric1^−/−); 100 animals were included in each sample. f, TEM images of 4 dpf ric1^−/− craniofacial chondrocytes show large vacuolar structures (arrowhead), f’ higher magnification of boxed area, and f” further zoom in of the blue box showing striated ultrastructure of intracellular collagen fibrils. Control WT images in Extended Data Fig. 6. g, Progression of matrix crosslinking in WT cartilage ECM from 3 dpf to 4 dpf. Note the ECM paucity and deficit in crosslinking at 4 dpf in ric1^−/− ECM. TEM imaging was repeated with three independent ric1^−/− animals at 3 dpf, one at 4 dpf, one WT sibling at 3 dpf and one at 4 dpf. 5 cells were imaged per animal. ECM: extracellular matrix, N: nucleus.

To test whether the effect on collagen secretion is cell-autonomous, we injected low concentrations of guide RNA (gRNA) to generate ric1-mutant cell clones (ric1^gRNA) in WT background using CRISPR/Cas9 system (Extended Data Fig. 4a,b). Ric1-deficient clones recapitulated the round phenotype and accumulated intracellular collagen, whereas neighboring WT cells secreted collagen normally. These data suggest that Ric1 is cell-autonomously required for collagen transport.

RNA-seq with FACS-sorted zebrafish chondrocytes show that ric1 expression at 3 dpf is higher than 4 dpf which could be explained by increased collagen transport at the onset of chondrocyte differentiation, while expression of other collagen transport-related genes remains at similar levels (Fig. 3e) . Quantitative PCR (qPCR) and RNA-seq showed low, but steady ric1 expression throughout developmental stages³¹ (Extended Data Fig. 5a,b,c). Collectively, these results suggest that ric1’s potent chondrocyte-specific functions drive collagen secretion even at basal expression levels (i.e., RPKM range: 0.67–1.35). Round mutations have reduced ric1 expression, potentially due to nonsense-mediated decay (Extended Data Fig. 5d). Similarly, fibroblasts from the CATIFA patient have reduced levels of RIC1 transcript (<50% of control cells) that might result from degradation of non-spliced transcripts (Extended Data Fig. 5e).

We examined subcellular organization of the collagen inclusions in chondrocytes by transmission electron microscopy (TEM) and found ric1^−/− cells contained vesicles of various sizes, while WT cells were devoid of such vesicular structures (Fig. 3f; Extended Data Fig. 6a-g). High magnification images revealed striated material in the core of the large vesicles, that are reminiscent of collagen fibrils normally formed in ECM, as procollagen is processed by enzymes that cleave N- and C-termini (Fig. 3f-f”). Cartilage matrix analysis by TEM showed reduced ECM cross-linking in ric1^−/− at 3 dpf, which appeared absent by by 4 dpf (Fig. 3g). Thus, TEM data corroborate intracellular retention of collagen in ric1^−/− cells and compromised ECM ultrastructure.

Ric1-depletion leads to musculoskeletal and behavioral defects

Common and rare disease phenome associated with RIC1 included gait disturbances and musculoskeletal deficits. Tendon and muscle attachment sites are matrix-rich structures, containing high levels of thrombospondin-4 (TSP4)³² and collagen-I (Col1)³³. We examined, by whole-mount immunofluorescence (WMIF) and TEM analyses, the quality of the muscles and tendons. Confocal images showed shortened and misaligned ric1^−/− cranial muscles (Fig. 4a, Extended Data Fig. 7a), and diffused pattern of TSP4 staining in mutants (Fig. 4a, insets). In WT tenocytes of the tendon posterior to the sternohyoideus (sh, marked by DAPI), collagen-I (Col1) shows discrete punctate pattern while ric1^−/− tenocytes accumulate large clumps in perinuclear regions (Fig. 4b), indicating secretion defects.

Figure 4. Ric1 depletion leads to musculoskeletal, brain and locomotion defects.

a, Whole-mount immunostaining of WT and ric1^−/− mutant larvae for myosin (MF20, muscle) and thrombospondin-4 (TSP4) for tendon area. Ventral view of head muscles, inset: magnified view of boxed tendon region; arrowheads point to TSP4+ domains in the Meckel’s (left) and palatoquadrate (right) elements. n=4 animals for each group. Abbreviations for muscles, ih: interhyoideus, imp: intermandibularis posterior, ima: intermandibularis anterior, sh: sternohyoideus. b, Cryo-section and immunostaining of tendon area along sh muscle for collagen-I (Col1, red), phalloidin (muscle, green) and DAPI nuclear counterstain (blue). Tenocyte nuclei are marked by DAPI, arrows to the Col1 punctae. n=3 animals for each group c, Locomotion tracking of zebrafish embryos at 4 dpf stage. 16 WT and 16 ric1^−/− embryos were tracked and plotted. Each color line indicates a separate animal (individual tracks in Extended Data Fig. 7). d, e, Graphs showing mean velocity of total travel (d) and maximum velocity reached during recorded travel (e) for each embryo. Mean and SD bars are displayed. Two-tailed Mann-Whitney U-test, CI = 95%, 16 WT and 16 ric1^−/− larvae were analyzed. f, Whole mounted zebrafish embryos immunostained for HuC/D (pan-neuronal marker) and WGA (wheat germ agglutinin) at 5 dpf, maximum intensity projection. Forebrain and cerebellum structures are demarcated by dashed lines. Quantification is in Extended Data Fig. 8. 8 independent larvae per each group were analyzed. Arrowheads point to eminentia granularis; arrows to lobus caudalis cerebelli.

Tendons are critical for muscle fibers to attach, and thereby contract³⁴. It was previously shown that muscle fiber attachment defects lead to fiber retraction³⁵. We analyzed muscle fiber attachment quality by WMIF and observed wavy, retracted muscle fibers in intermandibularis posterior and in trunk muscles (Extended Data Fig. 7a-c). TEM images corroborated discontinuity of myofibrils, revealing lack of periodic myofilament patterning and alignment (Extended Data Fig. 7b).

Zebrafish larvae rely on muscles to explore the surrounding by locomotor behavior patterns consisting of bouts of swimming, turning behavior, and motionless resting episodes³⁶. Swimming episodes are divided into: burst swim with fast sharp-angle turns and slow swim with slower routine, wide-angle motions³⁷. Musculoskeletal tissues are critical to execute locomotor behaviors³⁸. To functionally assess motility defects, we recorded live locomotion of 16 WT and 16 ric1^−/− zebrafish larvae. We traced tracks of each embryo’s travel for 14 seconds (Fig. 4c, Extended Data Fig. 7d-f). 13/16 WT and 5/16 ric1^−/− larvae moved from their initial position during the recording. Mean velocity of WT larvae was 4.33 mm/s while that of ric1^−/− mutants was 0.87 mm/s (Fig. 4d). Thus, ric1-mutants exhibit fewer exploratory behaviors and are significantly slower (Fig. 4e), than WT counterparts. Collectively, these results show that ric1-deficiency leads to structural and functional musculoskeletal defects.

To gain insight into neurological CATIFA symptoms, we analyzed zebrafish brain structures. In zebrafish, the forebrain controls social orienting behavior³⁹, which is reflective of social engagement and attention, typically affected in ADHD. We analyzed forebrain and cerebellum by WMIF using HuC/D pan-neuronal marker. (Fig. 4f). The data show reduced forebrain size in ric1^−/− mutants. We also examined the cerebellum, a brain region functioning in sensorimotor control. The cerebellum controls sensorimotor functions, is implicated in emotional and cognitive tasks, and was linked to intellectual disability, ADHD, and cognitive dysfunction⁴⁰. Microscopy analysis revealed smaller cerebellar area in ric1-mutants and a remarkable deficit in neurons at the posterior margin of the cerebellum (Fig. 4f), which regulate motor coordination in response to vestibular information⁴¹. This phenotype is consistent with the clumsy walk of CATIFA subjects. To assess the overall patterning of cerebellar ECM, we examined WGA (wheat germ agglutinin marking N-glycosylated proteins) staining in ric1^−/− larvae. We found noticeable thinning of WGA labeling in rostral cerebellum and optic tectum (Extended Data Fig. 8a, b), suggesting defects in brain matrix formation. Overall, neuronal phenotypes included significantly smaller forebrain and cerebellum, and altered axonal projection (Extended Data Fig. 8c-e), suggesting that loss of ric1 leads to morphological and functional deficits in the nervous system.

Of note, we have not observed detectable defects in ric1^−/− zebrafish lens at stages analyzed (Extended Data Fig. 8f-g), although 7/8 CATIFA patients had cataract.

Ric1-Rgp1 GEF complex regulates procollagen transport by activating Rab6a

Prior in vitro studies showed that Ric1-Rgp1 proteins interact and activate Rab6a by facilitating GTP loading (Fig. 5a). This model predicts that depletion of Rgp1 would result in a deficit similar to the ric1 loss-of-function phenotype, and that overexpression (OE) of constitutively active Rab6a would bypass the Ric1-Rgp1 requirement and rescue procollagen traffic in ric1^−/− mutant embryos.

Figure 5. Ric1-Rgp1 GEF complex regulates procollagen transport via Rab6a activation.

a, Model of Rab6a activation by the Ric1-Rgp1 GEF complex to regulate collagen transport, and constitutively active Rab6a bypassing GEF requirement for collagen transport. b, IF labeling with Col2 antibody and wheat germ agglutinin (WGA) in 3 dpf zebrafish cartilage shows disrupted cell shape and tissue organization in ric1-mutants and rgp1^gRNA (rgp1-guide RNA for CRISPR-Cas9 genome editing, Extended Data Fig. 9) injected WT embryos. Yellow arrows point to intracellular collagen accumulation, orange arrowhead points to extracellular Col2. This experiment was repeated with similar results, 3 independent animals per each group. c, Experimental design for mosaic overexpression (OE) of EGFP-Rab6a (WT form or Q72L constitutively active mutant) fusion protein in zebrafish notochord sheath cells. d, Col2 and EGFP co-immunostaining of sheath cells over-expressing constitutively active Rab6a (Q72L) mutant and wild type forms (bottom panel). Yellow arrows point to intracellular accumulations; orange arrowheads mark intracellular collagen in WT and rescued cells. e, Quantification of intracellular collagen accumulation within the cytoplasm, represented as percentage of Col2-stained area over cell surface by measuring areas from maximum intensity projections using ImageJ ‘Measure’ tool. OE of constitutively active Rab6a (Q72L) rescues intracellular collagen accumulation to WT levels (CI=95%, two-tailed Mann-Whitney U-test); whereas WT Rab6a does not. Mean and SD values are indicated with bars.

To test the first prediction, by injecting a low concentration of rgp1-targeting gRNA, we generated CRISPR-induced mosaic mutants containing clones of rgp1-mutant cells in WT background (referred to as rgp1^gRNA hereafter) (Extended Data Fig. 9a-c). Mosaic rgp1^gRNA embryos resembled ric1^−/− mutants in their skeletal phenotypes (Extended Data Fig. 9d-g). Col2 antibody and WGA (marking glucosaminoglycans in cartilage ECM) staining showed that rgp1^−/− clones contained collagen accumulation similar to ric1^−/− chondrocytes (Fig. 5b). These results support the hypothesis that Rgp1 and Ric1 act together to regulate procollagen secretion and skeletal morphogenesis.

To test the second prediction that Ric1-Rgp1 GEF activation of Rab6a regulates procollagen secretion, we overexpressed constitutively-active, GTP-bound, Rab6a (Q72L)⁴² or WT Rab6a in both WT and ric1^−/− embryos (Fig. 5c). Specifically, we tagged the N-terminus of zebrafish Rab6a with EGFP⁴³ and mosaically overexpressed it under col2a1a promoter using Tol2kit⁴⁴ method (Fig. 5d). IF analysis of mosaic, transgenic embryos revealed that ric1^−/− cells overexpressing constitutively-active Rab6a (Q72L) cleared collagen accumulation to WT levels in a cell-autonomous fashion but mere OE of Rab6a was not sufficient to rescue collagen transport (Fig. 5e), suggesting Rab6a activation is required for collagen secretion.

The GEF activity of RIC1 on Rab6 was shown in vitro²¹ in mammalian cells; however, procollagen traffic through post-Golgi Rab6-positive compartments has not been previously implicated in collagen secretion in vivo. The essentiality of this step in collagen secretion is revealed for the first time by the zebrafish mutations and CATIFA syndrome.

Pathogenic RIC1 variant leads to collagen accumulation in patient fibroblasts

To test pathogenicity of the RIC1 R1265P variant and its role in collagen transport, we analyzed patient-derived fibroblasts. By TEM, we found distended vesicular compartments, similar to those observed in ric1^−/− zebrafish mutants, which were notably absent from control skin fibroblasts (Fig. 6a). Consistently, we found significantly expanded Col1 staining throughout the patient fibroblasts as compared to control cells by IF. (Fig. 6b,c, Extended Data Fig. 10a,b).

Figure 6. Pathogenic RIC1 variant leads to collagen accumulation in CATIFA fibroblasts.

a, TEM images of control BJ fibroblasts and RIC1^{R1265P/R1265P} patient’s dermal fibroblasts show Golgi complex stacks (arrowheads). Arrow points to enlarged Golgi lumen and post-Golgi structures in patient fibroblasts. 5 cells of each group were imaged by TEM. b, Co-localization of Collagen 1 (Col1) and the TGN marker p230 (arrow) by IF in RIC1^{R1265P/R1265P} cells. c, d, Quantification of the intracellular Col1 content (c), and Col1 co-localization with p230 (d) based on images in b. (Student’s t-test, two-tailed, CI= 95%). Mean value (center line) and SD bars are indicated in ‘c’; center line indicates the median and whiskers show min to max values in ‘d’. e, Experimental design for mosaic OE of human RIC1 (hRIC1) gene in zebrafish. EGFP is linked to hRIC1 via self-dissociating viral 2A (v2A) peptide marking hRIC1 expressing cells. f, Antibody labeling for Col2 and EGFP in notochord sheath cells shows collagen deposits (arrows). hRIC1 overexpressing cell (demarcated by dashed line) has fewer deposits (arrowheads). g, Mosaic OE of CATIFA-linked R1265P variant of hRIC1 only partially clears collagen deposits (arrows) while neighboring cells continue to retain Col2 and serve as ric1^−/− controls. Similar observations were made in 3 cells for each group in f, g. h, Experimental design for hRIC1 mRNA OE and analysis of craniofacial morphology rescue. i, Phenotypic scoring results of protruding jaw morphology in hRIC1 mRNA overexpressing larvae and controls. n: number of animals analyzed. j, Dissected Alcian blue preparations of ethmoid plate and trabeculae (dorsal head skeleton) of hRIC1 mRNA injected larvae and controls show mosaic rescue of ric1^−/− cells (arrows) with characteristic clear cytoplasm as in the WT skeletal element. The R1265P variant rescue more closely resembles cell morphology in the round mutants; 3 independent animals of WT and ric1−/−. k, Phenome comparison table for SEC23A gene, associated with cranio-lenticulo-sutural dysplasia (CLSD) in animal models, gene-based PheWAS analysis in BioVU and Mendelian disease patients. l, Phenome comparison table for FBN1 (fibrillin-1) gene, associated with Marfan syndrome. AV: atrioventricular, EKG: electrocardiogram. m, Integrated approach for discovery of gene function and disease mechanism using the animal model, common-disease phenome in a biobank and monogenic, rare Mendelian disease phenome of RIC1/CATIFA syndrome.

Rab6a is known to act at various steps of secretory and endocytic pathways^45-47. To identify the intracellular compartment where RAB6A GEF (RIC1-RGP1) is required for procollagen transport, we tested co-localization of Col1 with compartment-specific markers. TGN markers p230 and Golgin-97 partially co-localized with Col1 staining (Fig. 6b,d, Extended Data Fig. 10a,b). Thus, procollagen transport is blocked at the TGN, placing the RIC1-RGP1 GEF activity in the anterograde secretory pathway for procollagen secretion.

We tested whether RIC1 function is evolutionarily conserved in regulating procollagen transport. In a genetic replacement experiment, we mosaically overexpressed human RIC1 (hRIC1) fused to EGFP by a self-hydrolyzable viral v2A tag⁴⁸ in zebrafish. Cleavage of the v2A connector detaches hRIC1 from EGFP, thereby labeling transgenic cells with diffused intracellular EGFP (Fig. 6e). We found that OE of the WT hRIC1 rescued intracellular collagen accumulation in rnd^−/− cells in a cell-autonomous manner, while having no effect on collagen expression in WT cells (Fig. 6f, Extended Data Fig. 10c,d). We introduced the R1265P variant into hRIC1-v2A-EGFP and assayed for collagen deposits in zebrafish ric1^−/− cells. Mosaic OE of the R1265P variant failed to fully restore collagen secretion, unlike the WT hRIC1 construct (Fig. 6g). The WT hRIC1 was sufficient to rescue jaw protrusion defects in rind^−/− (Fig. 1e). To test whether the CATIFA variant can restore jaw morphology in vivo, we overexpressed either WT or R1265P (Fig. 6h) in embryos produced by ric1^+/− heterozygous pairs. Over 90% of the larvae injected with WT construct show fully or partially protruding jaw morphology, compared to 75% of a control non-injected clutch (Mendelian ratio for recessive mutations), while the R1265P construct did not rescue (Fig. 6i). Alcian blue preparation of ethmoid plate shows clear chondrocyte cytoplasm and rich ECM around rescued cells, and partially restored cartilage shape, which is not seen in R1265P variant (Fig. 6j). These in vivo experiments confirm the pathogenicity of the variant and an essential role of the R1265 residue in ECM secretion and cartilage morphogenesis. Taken together, RIC1 depletion likely leads to RIC1-associated CATIFA syndrome through disruption of procollagen secretion.

Continuum of common and Mendelian phenome in other matrix-related conditions

We analyzed two additional Mendelian conditions to test generalizability of our approach: CLSD and Marfan syndrome, each with extensively-described patient phenome and well-characterized animal models. The BioVU medical phenome associated with reduced GReX of the CLSD-linked SEC23A gene successfully predicted 9/10 patient symptoms¹⁶ (Fig. 6k, Supplementary Table 6). Seven of these phenotypes have been recapitulated in sec23a^−/− zebrafish model¹³ (2/3 phenotypes were not observed in zebrafish, i.e., foot deformities and hemangioma of skin, likely because of species-specific differences). For the FBN1-linked Marfan syndrome, characterized by cardiovascular abnormalities in patients^49,50, and mouse models^51,52, we observed 7/8 phenotypes associated to reduced GReX of FBN1 in BioVU, and 6/8 of the phenotypes including cardiomyopathy, hernia and heart failure in published mouse models (Fig. 6l, Supplementary Table 7). These tests support generalizability of the approach presented here.

Discussion

Collagen constitutes one third of the dry body mass in vertebrates and is essential for the structure and function of a wide-range of organs^53,54. In disease states such as fibrosis and wound healing, collagen is efficiently secreted to ECM to enclose the injury site. Despite such abundance and clinical importance, the exact mechanism by which procollagen moves through a cell is not well-understood. Here, for the first time, we show evidence that the Ric1-Rgp1-Rab6a module in the TGN is an essential transit point for the collagen traffic en route to the ECM.

Disrupted collagen secretion creates a bottleneck, resulting in multi-organ pathologies (e.g., craniofacial dysmorphology, intellectual disability). Despite the strong agreement between the common-disease phenotypes we observed in relation to RIC1 reduced expression and the CATIFA phenotype, we acknowledge that the overlap is incomplete. This might be attributed to the limited patient numbers in BioVU or to their young age relative to some phenotypes, which are known to be age-related e.g., intestinal diverticulosis. Furthermore, ADHD and intellectual disability in CATIFA patients might be due to compromised neural ECM composition and remodeling, which are implicated in synaptic plasticity, neuronal guidance, and maintenance of normal neurological functions^40,55. Thus, a detailed RIC1-linked phenome offers a rich phenotype dataset associated with post-Golgi transport deficits of collagen, providing a reference panel for clinicians.

Genotype-phenotype relationships are hardly linear in patients; thus, to describe the full phenome associated with a single gene remains a standing challenge in genetic medicine. Hence, an approach integrating linear genotype-phenotype relationships and population-level variation may provide a more complete assessment of the phenome linked to a single gene.

Model systems allow investigation of molecular mechanisms, but it is hard to assess physiological relevance in humans. In omics-driven association studies, one can assess polygenic traits, and generate hypotheses, but it is hard to know what is physiologically important in patients^56,57. The challenge of explaining disease mechanism is particularly acute in the common-disease space, but also remains prominent in Mendelian diseases.

Here, we describe a novel approach integrating use of animal models, human common-disease biobanks, and rare Mendelian disease to accelerate and refine the discovery process of disease mechanisms (Fig. 6m). BioVU phenome associated with reduced GReX of RIC1 informed phenome investigations in both the zebrafish model and the human patients homozygous for RIC1 variant. Molecular mechanisms characterized in zebrafish helped explain the phenome observed in CATIFA patients and BioVU subjects with reduced GReX for RIC1. Integration of different sources of data facilitated moving from proximal mechanisms such as reduced gene expression, to cellular mechanisms such as the trafficking of procollagen, to phenotypic mechanisms for gait deficits and swimming behavior resulting from ECM impairment.

Better ability to diagnose the complete disease phenome improves the likelihood of understanding the disease mechanisms. However, clinical evaluation is often limited to the most severe and obvious symptoms patients manifest, thereby hindering the discovery of the full phenotypic spectrum. Initial phenotyping of zebrafish models and re-iteration of gene-associated phenotypes in a biobank led us to “guided clinical re-evaluation” of patients. In our example, craniofacial dysmorphology in zebrafish ric1^−/− mutants and in the BioVU dataset guided examination of skeletal signs in CATIFA patients. In turn, developmental tooth deficits discovered in the biobank prompted teeth re-examination in zebrafish, and led to inclusion as a core CATIFA sign.

Detailed knowledge of the disease mechanisms and spectrum of the associated phenome will enhance opportunities for patient care. For example, preferential use of non-invasive diagnostic methods due to predicted fragility of BM in various organ systems could be prioritized. The discovery of the genetic defects affecting ECM production, and specifically intracellular trafficking, opens new possibilities for molecular diagnosis in pre- and postnatal care, and later in life. The gene-associated phenome in biobank patients also gives a preview of potential later-onset phenotypes (e.g., Supplementary Table 2) and better prepare clinicians and families for planning future care of patients. As genotype-linked health records grow and quantitative genetics tools^58,59 advance in the upcoming years, this approach will advance personalized medicine practices.

Methods

Fish Maintenance and Breeding

Zebrafish were raised under standard laboratory conditions at 28.5°C as previously described⁶⁰. All experiments were conducted according to the guidelines established by the Institutional Animal Care and Use Committee at Vanderbilt University Medical Center (M1700020-00). round alleles m641, m713, m715 were isolated and described in the MGH zebrafish mutant screen^7,20 and were kept in the AB genetic background for phenotypic analysis. Embryos were staged and fixed at indicated stages, i.e. hours post-fertilization (hpf) and days post-fertilization (dpf).

Genetic Mapping and Positional Cloning of Zebrafish Mutations

The round locus was mapped in an F2 intercross using bulked segregant analysis and mapping. DNA samples were genotyped by PCR using SSLP (Simple Sequence Length Polymorphism) markers evenly spaced across the zebrafish genome⁶¹. The DNA lesions in mapped rnd^m641, rnd^m713 and rnd^m715 mutations were identified by Sanger sequencing of genomic DNA from homozygous wild-type F2 animals, heterozygous F2 animals and homozygous mutant F2 animals¹⁰. Transcripts were confirmed by cDNA sequencing.

Homology analysis presented in Fig. 1 was conducted via Clustal Omega (EMBL-EBI) using zebrafish Ric1 (Zv9 assembly) and human RIC1 primary sequences (Uniprot: Q4ADV7).

Participants of the CATIFA Studies and Genetic Analysis

All patients, siblings and parents were clinically evaluated and enrolled in an IRB-approved research protocol with written informed consent (KFSHRC RAC# 2070023). Written photo consent was obtained from the parents for all individuals shown in Fig. 2. Venous blood was collected from all participants in EDTA tubes for DNA extraction. DNA samples were genotyped on the Axiom SNP chip platform according to manufacturer’s instructions (Affymetrix, Santa Clara, CA, USA). SNP genotypes were utilized to perform multi-point linkage analysis with the Genehunter program (version 2.1r5) within the EasyLinkage plus (V5.08) software package⁶². Linkage analysis of both families confirmed a single significant linkage peak (LOD 3.44) on Chr9 spanning RIC1.

CRISPR/Cas9 Genome Editing of the ric1 and rgp1 Loci

CRISPR/Cas9 target sites within the zebrafish ric1 gene were identified using the CHOPCHOP web tool (ENSDART00000082377 transcript of KIAA1432 gene, ric1, on Zv8 genome assembly was used as target template). The following site within exon 25 (out 34 total exons) was targeted: 5’- GGTGCGTAATCTGGGTGAGCAGG-3’. Likewise, the following CRISPR/Cas9 target site within rgp1 gene (GRCz10 assembly) was targeted in this study; 5’-GGTGGTGGCATCTATGGCACGGG-3’. A cloning-free method to generate sgRNA templates was performed as previously described⁶³. Guide RNAs (gRNAs) were synthesized with MEGAshortscript™ T7 transcription kit (ThermoFisher Scientific).

To generate mutations with the CRISPR/Cas9 system, a mixture of 500pg purified Cas9 protein (PNA Bio Inc, # CP01) and 180 pg gRNA was injected into one-cell stage embryos. Injected embryos were grown to 3-5 dpf stages for phenotypic analysis, histology and imaging. Mutations generated in injected embryos were detected via direct sequencing of the region surrounding the target site. PCR-amplified products were cloned into pGEM-T Easy vector (Promega) by T-A cloning and sequenced using SP6 primers in order to individually detect various mutations created by CRISPR/Cas9 system, which are listed in extended data figures (Extended Data Fig. 4b for ric1, and Extended Data Fig. 9b, for rgp1).

BioVU

BioVU is Vanderbilt University Medical Center’s DNA biobank, with DNA on >250,000 subjects linked to electronic health records (EHR) going back an average of 10-15 years, and >25 years in some subjects ²³. Although BioVU originally recruited subjects through an opt-out procedure, subjects have been prospectively consented since 2014. For our study, we utilized whole-genome genotype data in ~10,000 participants of European ancestry. Genotyping had been performed using several genotyping platforms and SNP-level imputation was subsequently conducted using MaCH⁶⁴ on the Michigan Imputation Server in small batches. To determine genomic ancestry, we utilized Principal Components Analysis⁶⁵. We included genetic variants with minor allele frequency (MAF) >1% and imputation quality score >0.80 in downstream analyses.

Comprehensive PrediXcan Analyses and Gene-based PheWAS

To evaluate the role RIC1 may play in the etiology of a disease trait, we utilized the PrediXcan²⁶ approach to build SNP-based prediction models. We estimated the genetically regulated expression (GReX) in the ten thousand BioVU subjects⁶⁶ using the GTEx resource (v6p)^27,67 as a reference transcriptome panel (containing paired tissue-specific transcriptome and genotype data), to identify associated phenotypes⁶⁸ despite the lack of directly measured gene expression data. Specifically, from the weights ${\widehat{\beta }}_j$ derived from the gene expression imputation model for RIC1 (consisting of 21 SNPs; see Supplementary Table 1) and the number of effect alleles X_ij at the variant j, we estimated the genetically determined component of gene expression:

$${\widehat{G}}_i=\sum _j{X}_{ij}{\widehat{\beta }}_j$$

in the BioVU subjects (see Supplementary Table 2). Association between the estimated genetic component of gene expression (i.e., GReX) and a disease trait was determined using logistic regression. PrediXcan proposes a causal direction of effect between GReX and trait (because the germline genetic profile is unlikely to be altered by the trait). We use phecodes²⁵, which aggregate ICD9 or ICD10 codes into a maximum of ~1900 phecodes, in testing the association of the genetically determined expression of RIC1 to the medical phenome. To maximize the quality of the phenome information, we required at least 2 ICD9 or ICD10 codes on different days to instantiate a phecode for a chronic disease diagnosis. Phenome-wide significance of this gene-based test is based on Bonferroni correction of the number of phenotypes examined (~1100 phecodes have at least 20 cases in the 10,000 subjects analyzed and were tested in our analyses), or p< 4.5 × 10⁻⁵, although this is likely to be conservative because of the correlations among phenotypes. The actual sample size for an analysis of a specific phenotype depends on the phenotype (ranged from 4,074 to 9,021 for in RIC1 dataset).

In brief, Phenome-Wide Association (PheWAS) Study we conducted tested GReX against ~1100 phenome codes²⁵. In a traditional PheWAS⁶⁶, one SNP is used to test association; however, we used a variation of the method which we called “gene-based PheWAS” because we used multiple SNPs from the local (cis) regulatory region spanning ~ 1Mb of a genome around the gene of interest.

UK Biobank and GWAS Analyses

To provide independent support for the phenome associated with reduced RIC1 expression observed in BioVU, we conducted association analyses on GReX of RIC1 in the UK Biobank and other public GWAS (summary statistics) datasets and, in parallel, performed analyses of SNP associations (Neale mega-GWAS) in the RIC1 gene locus in the UK Biobank. For the GReX analyses, we applied models using the tissues (Extended Data Fig. 2a) in which we have high-quality prediction performance of the genetic component of RIC1 expression. For the analysis of SNP associations, we considered the most significant SNP in the gene; this can be viewed as Tippett’s (non-parametric) gene-based combination method that rejects low values of the minimum of SNP p-values⁶⁹. We mapped the BioVU phecodes for the core phenotypes of CATIFA as well as the phenotypes reported as nominally significantly associated in BioVU (Fig. 2d, Extended Data Fig. 2b,c and Supplementary Table 3) to the UK Biobank phenotype (ICD10) codes.

Genetic Manipulations in Zebrafish

mRNA overexpression and phenotypic rescue:

The human RIC1 (hRIC1) cDNA clone was obtained from Dharmacon GE Lifesciences (Clone ID 9020606). The missing 5’ region (including start codon) of RIC1 was cloned from a human BJ skin fibroblast (ATCC® CRL-2522™) cDNA. Two fragments were ligated using common AflII restriction site to obtain full length RIC1 coding sequence. Full-length human RIC1 coding sequence was sub-cloned into pCS2+ vector. Patient variant (R1265P) was introduced with site-directed mutagenesis. These vectors were linearized and used for in vitro transcription to synthesize capped mRNA with the mMESSAGE mMACHINE SP6 transcription kit (ThermoFisher). 1 ng mRNA (either WT hRIC1 or R1265P patient variant) per embryo was micro-injected into 1-cell stage zebrafish embryos. Embryos were raised to 5-7 dpf and scored for craniofacial morphology (with live imaging) and cartilage shape (with Alcian blue staining). Double-blinded, two independent investigators scored the rescue of craniofacial morphology at 5 dpf accounting for the level of jaw protrusion beyond the eye plane. For cartilage shape analysis, Alcian blue stained cartilage elements from mRNA injected embryos and non-injected controls were dissected with fine-tipped forceps and wet-mounted in 90% glycerol, imaged by an AxioImager.Z1 equipped with a 20X/0.80 Plan Apochromat objective and HRc camera (Zeiss).

Transgenic constructs:

Transgenic constructs for overexpression experiments (OE) were generated with the Gateway cloning-based Tol2kit method⁴⁴. Full length RIC1 was PCR-amplified with attB1(forward) and attB2 (reverse) recombineering sites and cloned into pDON221. The obtained clone was utilized as a middle entry vector in the LR cloning step of the Multistep Gateway recombineering system, in combination with the p5E-1.7kb Col2a1a promoter⁷⁰, p3E-v2A-EGFP vector (gifted by Dr. J.T. Gamse), pDestTol2pA2 and incubated with Gateway® LR clonase® II enzyme mix. The resulting destination clone was used for mosaic overexpression experiments. v2A-EGFP tag was used to label transgenic cells with cytoplasmic EGFP signal that forms upon self-cleavage of v2A-EGFP from the fusion polypeptide⁴⁸. The R1265P RIC1 variant identified in CATIFA patients was introduced into the destination vector via Q5® site-directed mutagenesis kit (NEB). Zebrafish rab6a was cloned from 4 dpf wild type (AB) cDNA into the vector pEGFP-C1. Next, the EGFP-Rab6a fusion was sub-cloned into the pDONR221 vector via BP cloning, after adding attB1 and attB2 sites with PCR amplification. The obtained clone was utilized as a middle entry vector in the LR cloning as described above. p5E-1.7kb Col2a1a promoter⁷⁰, p3E-ployA, pDestTol2pA2 were used to build rab6a destination vector. The Q72L mutation (constitutively active Rab6a mutant⁴²) was generated on the destination vector via the Q5® site-directed mutagenesis kit (New England Biolabs). Primers used are listed in Supplementary Table 8.

Microinjection of Tol2kit clones for mosaic overexpression.

Zebrafish embryos at 1-cell stage were injected with a combination of 50 pg medaka transposase mRNA⁷¹ and 10-20 pg of destination vector DNA as described previously⁷², and grown to indicated stages in embryo medium. Mosaic clones were detected by EGFP signal.

Histology, Transmission Electron Microscopy (TEM) and Tissue Staining:

Histology:

Histological sections were prepared with JB-4 embedding kit as previously described⁷³. 4-8 μm thick sections were stained with 1% toluidine blue.

Transmission electron microscopy & quantification of vesicles in zebrafish tissues:

Samples were processed for TEM as previously described¹⁰. Briefly, 3 and 4 dpf embryos were fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate by incubating at room temperature for 1 hour first, then overnight at 4°C, followed by TEM sample processing. 70-nm sections were collected on a Leica Ultracut Microtome. Cartilage, notochord and muscle tissues were imaged on a Phillips CM-12 Transmission Electron Microscope provided by the VUMC Cell Imaging Shared Resource. Intracellular vesicles were classified according to electron density and morphology. Width and height of vesicles were measured using ImageJ’s ‘Measure’ tool.

Cartilage and bone staining:

5 or 7 dpf embryos were stained for bone (Alizarin red) and cartilage (Alcian blue) as described previously⁷⁴. 0.02% Alcian blue and 60 mM MgCl₂ in 70% ethanol, 0.005% alizarin red was used as a double staining solution. Stained embryos were imaged with Zeiss Axioimager.Z1 scope equipped with an Axiocam HRc camera. Number of analyzed embryos for each experiment were indicated in corresponding figure panels.

Live Imaging and Immunofluorescence

Live imaging of zebrafish embryos and morphometric analysis:

5 dpf embryos were anesthetized in Tricaine (Sigma), mounted in 3% methylcellulose (Sigma) and imaged using a Zeiss Stemi 2000-C with HRc camera. Using dorsal views in ImageJ, body and head lengths were measured from jaw protrusion to caudal fin fold and to the end of parachordals, respectively. Head size index was calculated by the following formula: head length / body length.

3D rendering of live hyosymplectic cartilage shape and chondrocyte volume:

3 dpf Tg(Col2a1a:caax-EGFP; Col2a1a:H2A-mCherry) double transgenic embryos were lightly anesthetized and mounted in 1.2% low-melting agarose in glass-bottom dishes. Imaging was conducted via LSM880 confocal microscope with 20X/0.80 Plan Apochromat and 40X/1.1 LD C-Apochromat water-immersion objective. Confocal z-stacks of entire hyosymplectic cartilage (~80 μm-thick) and individual chondrocytes (~25μm-thick) were analyzed in Imaris 8.0 (Bitplane) for 3D rendering using the surface tool (shape analysis) and statistics tab (volume analysis). Movies were generated using Imaris’ movie tool to help with the 3D view.

Immunofluorescence (IF) and fluorescence microscopy:

14-μm thick cryo-sections of zebrafish tissues were obtained by Leica CM-1900 cryostat. IF on cryo-sections was performed as previously described⁷². The primary antibodies used were anti-GFP (Vanderbilt Antibody and Protein Resource), anti-collagen type-II (Rockland, #600-401-104), anti-Fibronectin (Sigma, #F3648), anti-β-catenin (Sigma, #C7207), anti-Matrilin (gifted by Dr. E. Kremmer), anti-Col1 (abcam, ab34710), anti-myosin (Developmental Studies Hybridoma Bank, #MF20). 4’,6-diamidino-2-phenylindole (DAPI, Molecular Probes) was used as a nuclear counterstain. Fluorescently labeled secondary antibodies used were either Alexa Fluor 488- & 555- (Life Technologies) or DyLight 550-conjugates (Thermo Scientific). Proteoglycan staining was performed using Alexa Fluor-555 conjugated wheat germ agglutinin (WGA, Life Technologies); muscle staining on sections was with Alexa Fluor 488-conjugated phalloidin (ThermoFisher, #A12379). Fluorescence imaging was carried out using AxioImager Z1 equipped with an Apotome (Zeiss) and an EC Plan Neofluar 100x/1.30 Oil objective.

Analysis of collagen accumulation in zebrafish cells:

Images of collagen IF were analyzed in ImageJ. For collagen accumulation in chondrocytes, the intracellular area of each cell (demarcated by caax-EGFP signal), the collagen area (Col2 signal) and the nuclear area (DAPI) were measured. Percent collagen cytoplasmic area per cell was calculated by the following formula:

$$[\mathrm{Collagen}\mathit{area}∕(\mathrm{Intracellular}\mathrm{area}-\mathrm{Nuclear}\mathrm{area})]{}^{\ast }100$$

For collagen accumulation in notochord sheath cells, maximum intensity projections were utilized to span entire cell boundaries. The region around visible full cells in the maximum intensity projection views were drawn by tracing caax-EGFP signal and sub-selection images were produced. The total area of sub-selection was measured as total intracellular area. The total collagen area was measured by tracing the Col2 signal in individual cells and adding them up. Percent collagen area per cell was calculated by the following formula:

$$[(\mathrm{Total}\mathrm{collagen}\mathit{area}∕\mathrm{Total}\mathrm{intracellular}\mathrm{area})∕\mathrm{number}\mathrm{of}\mathrm{cells}\mathrm{in}\mathrm{the}\mathrm{region}]{}^{\ast }100.$$

Whole-mount IF:

Zebrafish embryos at the specified stages were fixed in Prefer Fixative (Anatech) for 48 hours, rinsed in PBS-0.1% tween 20, post-fixed in 4% PFA for 10 minutes, and then permeabilized in 0.5% triton X-100 in PBS. Next, embryos were bleached with 1.5% H₂O₂ / 1% KOH, then rinsed, blocked in 10% goat serum, 1% DMSO, 1% BSA and incubated in primary antibody (1:250 dilution) overnight at 4°C. Secondary antibody incubation (1:300 dilution) with fluorophore-conjugated antibodies was followed at 4°C overnight. After rinsing in PBS-0.1% tween 20, DAPI staining was performed (1:5000 dilution), and then embryos were post-fixed in 4%PFA for 20 minutes, afterwards, cleared and stored in 90% glycerol. Stained embryos were imaged on a Nikon Spinning Disk Confocal Microscope. Images are presented as maximum intensity z-projections generated by Nikon Elements software. Primary antibodies used are: α-myosin (Developmental Studies Hybridoma Bank #MF20), α-thrombospondin-4 (abcam #ab211143), α-acetylated tubulin (Sigma #T7451), α-HuC/D (ThermoFisher #A-21271). Secondary antibodies used are: anti-Mouse-Alexa Fluor 488 (ThermoFisher #A-11001), anti-Mouse-Alexa Fluor 555 (ThermoFisher #A28180), anti-Rabbit Alexa Fluor 488 (ThermoFisher #A-11008) and anti-Rabbit Alexa Fluor 555 (ThermoFisher #A-21428). Alexa Fluor 488-conjugated Wheat Germ Agglutinin (WGA), ThermoFisher #W11261) was performed at the secondary antibody incubation step with 1:300 dilution in blocking solution. DAPI counterstaining was performed after secondary antibody step, by incubating for 20 minutes in 1:5000 diluted DAPI in PBS-0.1% tween 20.

Brain Anatomy Measurements:

Whole mount zebrafish embryos stained with HuC/D and WGA were imaged on Nikon Spinning Disk Confocal Microscope, with Plan Apo Lambda 10X objective, 0.45NA. Maximum intensity z-projections (150-200 μm depth) were created in Nikon Elements software and analyzed in ImageJ using Freehand selection and measurement tools. Several embryos were processed and statistically analyzed with two-tailed Mann-Whitney U-test (CI = 95%).

Lens imaging:

Lens imaging was performed as previously described ⁷⁵. Briefly, whole intact eyes were dissected from live, anesthetized fish larvae and examined under DIC illumination on Zeiss AxioImager.Z1 equipped with 20X/0.80 Plan Apochromat objective and Flash4.0 camera (Hamamatsu). For transmitted light imaging, lens was further dissected from eye and image was acquired by AxioCam HRc (Zeiss) camera on AxioImager.Z1equipped with 20X/0.80 Plan Apochromat objective.

Human cell culture, TEM and IF analysis

Culture conditions:

Dermal fibroblasts from CATIFA patient (15DG2428) was isolated from skin biopsy and cultured in MEM (GIBCO) supplied with bovine serum at 37°C with 5% CO₂. BJ skin fibroblasts (ATCC® CRL-2522™) were used as the control for all experiments described further.

TEM:

Cells grown in 10 cm² dishes to confluency were fixed in 2.5% glutaraldehyde and in 0.1 M sodium cacodylate by incubating at room temperature for 1 hour first, then overnight at 4°C, and were then further processed for TEM imaging in Vanderbilt Cell Imaging Shared Resource core.

IF:

Prior to collagen-I staining, cells were treated with ascorbate for 1 hour at 37°C with 5% CO₂ as described previously⁷⁶. Cells were fixed in 4%PFA at room temperature for 20 minutes, permeabilized in 1% triton X-100 in PBS, then blocked in 1% BSA. Primary antibodies used were anti-Collagen-I (abcam, ab34710), anti-p230 (BD Biosciences, 61120) and anti-Golgin-97 (ThermoFisher, A-21270). Anti-mouse and anti-rabbit secondary antibodies conjugated to Alexa Fluor 488- & 555 (Life Technologies) were used for fluorescence imaging. DAPI (Molecular Probes) was used as the nuclear counterstain. Slides were imaged with the LSM880 confocal microscope using 20X/0.80 Plan Apochromat and 63x/1.40 Oil objective. Confocal images were acquired under the same settings for control and patient’s fibroblasts.

Image analysis of cultured cells:

Fluorescence intensity of collagen-I signal per cell and total cellular area were measured in ImageJ using the ‘Measure’ tool. Saturated pixels (255 value in dynamic range spectrum) and background-level low pixels (1-10 values) were eliminated from intensity datasets. Then, total intensity per cell was divided by cellular area and used as a factor of intracellular collagen-I content. Colocalization analysis of Collagen-I and the p230 (TGN marker) was performed using the Coloc2 tool in ImageJ. Pearson’s r-values were recorded from multiple cells and compared using the Student’s t-test.

Quantitative Transcript Analyses

Quantitative real-time PCR (qRT-PCR) was performed as described previously¹⁰. Using TRIzol reagent (ThermoFisher Scientific), total RNA was extracted from 12-15 zebrafish embryos for each sample at the specified developmental stage. The TRIzol method was performed to extract total RNA from cultured fibroblasts as well, following the manufacturer’s manual. 500 ng of total RNA was used as a template to make cDNA through reverse transcription using M-MLV reverse transcriptase (Promega) and poly-T primer. Each qRT-PCR reaction was performed with 20 ng of total cDNA, SYBR Green Real-Time PCR Master Mix and 2 μM of each primer. Primers used are listed in Supplementary Table 8. qRT-PCR reactions were run on CFX96 (Bio-Rad) system. Data were analyzed with the -ΔΔCt method, Student’s t-test (CI = 95%). Each sample was independently obtained from three independent zebrafish clutches. A similar approach was taken for qPCR for cultured human fibroblasts

Fluorescence-associated cell sorting (FACS):

GFP+ and GFP− cells were sorted from Tg(Col2a1a:caax-eGFP)⁷⁰ transgenic zebrafish embryos at 3 or 4 dpf stages with FACS Aria II Flow cytometer (BD) as described by Unlu et al (unpublished). Total RNA was isolated from corresponding cell population and used for RNA sequencing. Approximately 100 embryos were used to generate each sample.

cDNA library preparation and next generation RNA sequencing:

cDNA libraries were processed for RNA-seq by the Vanderbilt Technologies for Advanced Genomics (VANTAGE) core facility. 30 million 50-bp single end reads were collected per sample. Two independent samples for each experimental group were included as replicates.

Sequence alignment and differential expression analysis:

Reads were aligned to the zebrafish transcriptome (GRCz10 release 80) and genome (GRCz10) with Tophat2⁷⁷ default parameters. Differentially expressed genes were called with DESeq2⁷⁸ in Bioconductor⁷⁹.

Locomotor Behavior Analysis

2 WT and 2 ric1^−/− mutant zebrafish embryos at 4 dpf stage were transferred into a 27-cm glass bottom dish filled with embryo medium, up to the level of glass-dish barrier. Locomotion behavior of the embryos was recorded for 14-seconds, 0.07 sec / frame with an HRc camera mounted on Stemi-2000C stereoscope (Zeiss) at 0.65X magnification. Movie recording was performed by ZEN2 software (blue edition) with 1×1 binning, 1 ms exposure, B/W setting, at 1388 × 1040-pixel resolution. Movies were opened in ImageJ and analyzed with a MTrackJ plug-in. The position of each embryo was manually tracked frame by frame. Total displacement of embryos was obtained with MTrackJ’s Measure tool. The total displacement / total time (14 s) formula was used to calculate mean velocity. Distance to previous position (D2P) is also recorded by MTrackJ for each single displacement event. Maximum D2P value (MaxD2P) of each embryo was used to calculate ‘maximum velocity reached during the travel’ by the following formula: MaxD2P/ time between frames (0.07 s). A total of 8 videos were generated, accounting for 16 WT and 16 mutant embryos. Initial (x, y) coordinates (generated by MTrackJ, Measure Tool, Points sheet) of each embryo was taken as (0, 0) origin point and all following coordinates were adjusted with reference to the origin, creating relative coordinates values. Relative (x, y) coordinates were plotted in Prism 7 (GraphPad) to display individual tracks. Swim behavior patterns of embryos were scored by analyzing individual frames and categorized as described previously³⁷.

Statistics

For head size indices, collagen accumulation by IF and qPCR data, significance was assessed by non-parametric, two-tailed, Mann-Whitney U-test for comparison of two groups. Statistical analysis was performed with Prism 7.0 (GraphPad). Confidence interval (CI) is 95%. Mean and standard deviation bars are indicated on graphs. The data represented as box-whisker plots show the range between 25^th and 75^th percentile as boxes, minimum and maximum values as whiskers with median at the center. Collagen content and colocalization in human fibroblasts were statistically analyzed by two-tailed Student’s t-test using Prism 7.0 (GraphPad). All zebrafish assays were performed on three or more embryos unless otherwise specified; number of embryos analyzed for each assay was noted in figure panels as ‘n’.

Data Availability

Summary statistics from SNP-level and gene-level (PrediXcan) associations are available in Supplementary Tables. The protected data for the GTEx project (for example, genotype and RNA-sequence data) are available via access request to dbGaP accession number phs000424.v6.p1. All requests for raw (e.g., genotype and phenotype) data and materials are reviewed by Vanderbilt University Medical Center to determine whether the request is subject to any intellectual property or confidentiality obligations. For example, patient-related data not included in the paper may be subject to patient confidentiality. Any such data and materials that can be shared will be released via a Material Transfer Agreement. All gene, transcript and protein sequences can be found at NCBI and Ensembl databases (accession numbers are provided in Methods section).

Extended Data

Extended Data Fig. 1. Genetic analysis of ric1 mutations and dental phenotypes.

a, Genetic linkage analysis and positional cloning mapped the rnd mutation to zebrafish chromosome 21. Microsatellite markers and number of recombination events (recs, blue) in critical region that contains 4 protein-coding genes (orange). b, Sequence conservation analysis (using Clustal Omega, EMBL-EBI) of human and zebrafish Ric1 proteins exhibit 81% similarity, and 71% identity. Positions of mutations detected with direct sequencing of rnd alleles (m641, m713, m715) are shown with arrows. Multiple sequence alignment shows highly conserved R882 residue (red) of the Rab6-interacting region across vertebrate species analyzed, i.e. zebrafish, rat, mouse and human. c, Electropherograms of direct sequencing results from genomic DNA of homozygous WT (+/+), heterozygous (+/−) and mutant (−/−) embryos for all three alleles, shading highlights mutation sites. 10 independent embryos of each genotype were analyzed and all revealed identical results as presented here. d, Ventral view of the teeth (arrows) stained by Alcian blue (cartilage) and Alizarin red (ossification), n=6 animals per group, e, lateral view of the pharyngeal teeth (white arrows) on 7^th pharyngeal arch, n=6 animals per group. Orange arrows point to lack of cleithrum bone elongation in ric1^−/− mutant embryo, 7 dpf; a: anterior, p: posterior.

Extended Data Fig. 2. Genetic analysis of RIC1 in biobanks and additional CATIFA patients.

a, Correlation of imputation (R²) values for RIC1 expression across tissues. Sample sizes are indicated as ‘n’ next to each corresponding tissue on the graph. b, Q-Q plot showing the distribution of all association p-values with asthma in UK biobank. RIC1 association to asthma is statistically significant with p=7×10⁻⁹. Significance threshold is shown with black line set to p=1.5×10⁻⁵. c, Association of genes 4Mb around RIC1 locus with asthma. Note that association of RIC1 reduced expression with asthma is the most significant within this region. d, Pedigree of the additional family contributing to this study (modified from Patel et al., 2017³⁰). Standard pedigree symbols used; affected individuals are shaded, arrow points to patient 15DG2428 whose skin biopsy fibroblasts were sequenced (e), and analyzed by cellular and molecular methods. The mutation site is boxed, and results in Arg to Pro substitution in the protein sequence.

Extended Data Fig. 3. Analysis of matrix components in sheath cells and notochord basement membrane (BM).

a, Representative images of Tg(Col2a1a:caax-EGFP) transgenic zebrafish stained with Col2 antibody in WT and ric1^−/− notochord sections (Collagen-2, magenta; EGFP, green). Note accumulation of intracellular Col2 in mutant sheath cells (arrowheads). 3 animals per group were analyzed and similar results were confirmed. b, TEM images of notochord tissue at 3 dpf. Magnified views of sheath cells and extracellular sheath / basement membrane (BM) are displayed on the right panels. Arrowheads point to intracellular inclusions. ER: endoplasmic reticulum, N: nucleus. Size bars = 500 nm. 3 cells per genotype were imaged and similar observations were confirmed. Collagen II fails to be transported out of Ric1-deficient chondrocytes, while other ECM cargos are secreted normally. c, Experimental design for immunohistochemistry (IHC) analysis with antibodies against Col2, matrilin, β-catenin and fibronectin epitopes on cryosections. d, Col2 (arrows) and Matrilin (arrowheads) co-immunostaining in 3 dpf chondrocytes. 3 animals per group and 5 cells per animal were analyzed and similar results were confirmed. e, β-catenin (arrows) and Fibronectin (arrowheads) co-immunostaining at 60 hours post fertilization (hpf) stage. DAPI (blue) marks nuclei. 3 animals per group and 5 cells per animal were analyzed and similar results were confirmed.

Extended Data Fig. 4. CRISPR/Cas9 genome editing-mediated depletion of ric1 recapitulates collagen secretion defects in a cell-autonomous manner.

a, Co-immunostaining for collagen type-II (Col2) and WGA-labeled glycosylated matrix proteins at 3 dpf zebrafish cartilage. Arrowheads in ric1^gRNA mosaic mutant (ric1^−/− chondrocyte clones in the WT embryo injected with gRNA targeting ric1) point to collagen accumulations. Arrows indicate secreted, extracellular Col2. DAPI (blue) marks nuclei. This experiment was repeated with similar results using 4 animals per each genotype. b, Representative mutations detected in ric1^gRNA mosaic mutants by direct sequencing. PAM: Protospacer adjacent motif.

Extended Data Fig. 5. Ric1 expression in zebrafish and RIC1^{R1265P/R1265P} patient fibroblasts.

a, Temporal expression profile of ric1 across zebrafish developmental stages by PCR. 3 independent experiments were run and presented here. Each experiment contained pooled RNA from 10 embryos of the indicated genotype and stage. Center line indicates the mean; bars show standard error of the mean (SEM). b, Temporal expression profile of ric1 by RNA-seq from whole embryos (data from Expression Atlas, White et al., 2017³¹). c, Gene expression analysis of ric1 in sorted, live chondrocytes and non-chondrocytic cells by RNA-seq. R/FPKM: reads/ fragments per kilobase of transcript in million reads. d, QPCR analysis of ric1 in whole zebrafish embryos at 5 dpf stage, replicated with two primer sets. 3 independent experiments were run and presented here. Each experiment contained pooled RNA from 10 embryos of the indicated genotype and stage. Center line indicates the mean; bars show SD. e, Expression analysis of RIC1 in CATIFA patient’s fibroblasts (RIC1^{R1265P/R1265P}) by qPCR. Two independent primer sets were used for each assay. QPCR data in d and e were normalized to β-actin or GAPDH. Mean expression levels relative to control (WT) are reported in center lines, SD bars are indicated. Two-tailed Mann-Whitney U-test, CI = 95%.

Extended Data Fig. 6. Constricted cell shapes of ric1^−/− chondrocytes at 3 dpf and normal ultrastructure of WT cells by TEM at 3 and 4 dpf.

a, TEM image of WT craniofacial cartilage showing normal chondrocyte shapes. b, A representative WT chondrocyte at 3 dpf b’, Higher magnification reveals vesicular structures associated with normal secretion (arrows). c, A representative TEM image of WT chondrocytes at 4 dpf. c’, Magnified view of boxed area is shown on the right, arrows point to normal vesicular compartments. Intracellular vesicles accumulate in ric1^−/− chondrocytes. d, ric1^−/− craniofacial cartilage chondrocytes contain large inclusions (arrowheads). e, Representative examples of chondrocytes constricted at the midline (arrows), green overlays on b & e are drawn after image acquisition to help with cell shape demarcation. f, TEM micrograph of a representative 3 dpf ric1^−/− craniofacial chondrocyte displaying 3 types of vesicles, classified based on their electron densities: low density (arrows), medium density (double headed arrow) and vacuolar/high density (arrowheads). f’, Magnified view of the boxed region from f. g, Size distribution and average number of vesicles in 3 and 4 dpf ric1^−/− chondrocytes. (20 cells at 3dpf; 10 cells at 4dpf stage). Vesicle diameters were measured both in parallel (W: width) and perpendicular (H: height) dimension to image plane; and plotted as length (μm) values. ECM: extracellular matrix, ER: endoplasmic reticulum, M: mitochondrion, N: nucleus. Mean and SD values are indicated with bars. Three ric1^−/− embryos at 3 dpf and one at 4dpf were imaged with similar results. One representative WT of each stage was imaged with similar results across 10 cells imaged per embryo at each stage analyzed.

Extended Data Fig. 7. Muscle attachment defects in ric1^−/− zebrafish.

a, Whole-mount zebrafish embryos immunostained for myosin. Ventral head views are displayed. Arrows point to retracted fibers, in the form of bright punctate structures. ih: interhyoideus, imp: intermandibularis posterior, ima: intermandibularis anterior. b, Transmission electron micrographs of cranial muscles. Note non-continuous muscle fibers in ric1^−/− (arrow). M: M-line, Z: Z-line. Right panels show magnified views of boxed regions in the left panels (2 independent mutant examples, one representative WT). c, Schematic of tendon matrix and consequent muscle fiber attachment defects. d, Quantification of swim behavior patterns, categorized as either burst (fast, sharp-angle turn) or slow (slower routine, wide-angle motion) swim pattern (Budick & O’Malley, 2000). Mean and SEM are shown as bars. Each data point represents a single embryo, n=16 independent embryos per genotype. Locomotion tracks of individual embryos, e, WT embryos, f, ric1^−/− mutant embryos.

Extended Data Fig. 8. Analysis of brain and eye phenotypes in ric1^−/− zebrafish.

a, Whole mounted zebrafish embryos immunostained for HuC/D (pan-neuronal marker) and WGA (marking N-glycosylated proteins in neural ECM) at 5 dpf (Higher objective view of the same embryos in Fig. 4f). b, Digital zoom-in views showing sparser matrix in optic tectum and cerebellum regions of ric1^−/−. c, Forebrain (n=8 independent animals per genotype) and d, Cerebellum area measurements (n=7 independent animals per genotype) performed by ImageJ, Measure Tool using whole mount immunostained images. Arbitrary units (a.u.) were plotted. Bars indicate the mean, lines are SD. Two-tailed Mann-Whitney U-test, CI = 95%. e, Acetylated tubulin whole-mount immunostaining that labels axonal projections. Right panel shows magnified views. Notable poor bundle structure of axonal projections in ric1^−/− (arrows) and sparse axon tracks in cerebellum (arrowheads). f, Images of dissected whole zebrafish eyes under DIC illumination. g, Transmitted light illumination images of dissected zebrafish lenses. Note comparable smooth lens surface of WT and ric1^−/− in both panels. n: number of embryos analyzed.

Extended Data Fig. 9. Rgp1-depletion recapitulates ric1^−/− mutant phenotype.

a, Schematic for genomic structure of rgp1 gene. Exons are shown as boxes, protein coding in blue and untranslated in gray, introns as red lines. Orange arrow marks guide RNA target site within exon 2. b, Deleterious mutations detected in mosaic rgp1^CRISPR embryos by direct sequencing. c, Proposed model for Ric1-Rgp1 dependent activation of Rab6a on consequent collagen secretion, which modulates cartilage development. d, Live images of 5 dpf WT, ric1^−/− (rnd^m713) and rgp1^CRISPR (mosaic mutant embryo generated with CRISPR/Cas9 genome editing). Blue arrows on dorsal views point to body length differences. e, Alcian blue staining of craniofacial head skeletons at 5 dpf, black arrowheads point to Meckel’s cartilage protrusion and pink arrowheads to ceratohyal cartilage element. f, Graph for ceratohyal length measurements shows a range of lengths, likely due to mosaic nature of rgp1^CRISPR mutants. Both left and right elements are plotted. N: number of embryos, n: number of cartilage elements analyzed (Student’s t-test, two-tailed, CI= 95%). Central line shows mean; bars indicate SD. g, Histological analysis of zebrafish cartilage at 3 dpf with Toluidine blue staining. Arrows point to the secreted ECM surrounding WT chondrocytes, arrowheads to diffused matrix staining around mutant cells. Clone-1 in rgp1^CRISPR mosaic mutant is inferred to be composed of un-edited WT cells, while clone-2 is inferred to be of rgp1-mutant chondrocytes due to phenotypic resemblance to ric1^−/− cells in the middle panel. This experiment was repeated with similar results from 3 independent animals from each indicated genotype.

Extended Data Fig. 10. Collagen-I accumulates in TGN-associated compartments in RIC1^{R1265P/R1265P} patient’s fibroblasts.

a, Representative images (of 3 experiments of independently grown cell cultures) for co-IF of BJ fibroblasts (control) and 15DG2428 (RIC1^{R1265P/R1265P} patient’s dermal fibroblasts) using antibodies against collagen-I and p230 (TGN marker). Confocal images at low magnification (20x/0.80 Plan-Apochromat, WD=0.55mm) are presented. b, Representative low-magnification images (of 3 experiments of independently grown cell cultures) for co-IF with Col1 and alternative TGN marker Golgin-97 taken under the same conditions. Arrows point to TGN-associated Col1 signal. Overexpression of human RIC1 restores collagen secretion in ric1-deficient zebrafish chondrocytes. c, Experimental design to mosaically overexpress human RIC1 (hRIC1) gene in zebrafish. EGFP is linked to hRIC1 via self-cleavable viral 2A (v2A) peptide. d, Co-immunostaining of Col2 and EGFP in cartilage. Green cells express the construct and neighboring cells act as endogenous WT or mutant controls. Arrows point to collagen accumulation; arrowheads show secreted, extracellular collagen. Representative of 3 cells per group, with similar results.

Supplementary Material

SupVid_1

Supplementary video 1 3D rendered structures of hyosymplectic cartilage of 3 dpf WT embryo

SupVid_2

Supplementary video 2 3D rendered structures of hyosymplectic cartilage of 3 dpf ric1−/− embryo

SupVid_3

Supplementary video 3 3D rendered structures of symplectic arm chondrocytes of 3 dpf WT embryo

SupVid_4

Supplementary video 4 3D rendered structures of symplectic arm chondrocytes of 3 dpf ric1−/− embryo