Zebrafish were raised under standard laboratory conditions at 28.5°C with a constant photoperiod (14 h light: 10 h dark) as previously described (18).

CRISPR/Cas9 target sites within the zebrafish rgp1 gene were identified using the CHOPCHOP web tool (51). sgRNA templates were generated as previously described (52). gRNAs were synthesized with the MEGAshortscript T7 transcription kit (AM1354, ThermoFisher Scientific; Waltham, MA).

To generate mutations with CRISPR/Cas9 system, a mixture of 500 pg purified Cas9 protein (CP01, PNA Bio Inc.; Newbury Park, CA) and 150 pg gRNA was injected into one-cell stage WT AB embryos. Injected embryos were grown to 5 dpf for phenotypic analysis and to identify founders carrying mutant alleles in the germline. Founders were then outcrossed to transgenic or WT lines to generate stable lines carrying the listed mutations ( Figure 1B ). In silico protein translation was based on cDNA sequences obtained from mutant larvae ( Figure 1C ).

Larvae were anesthetized at 5 dpf in 0.15 mg/mL Tricaine (TRS1, Pentair; Cary, NC) and mounted in 3% methylcellulose on a bridge slide. Body lengths were measured from jaw protrusion to caudal fin fold from dorsal views in FIJI software (53). Head lengths were measured from jaw protrusion to the pectoral fin attachment point.

Larvae at 5 dpf were fixed in 2% PFA and stained overnight in 0.02% Alcian Blue (A5268, Sigma; St. Louis, MO) and 60 mM MgCl₂ in 70% EtOH as previously described (54). Larvae were then bleached (1.5% H₂O₂, 1% KOH), washed in 0.25% KOH/25% glycerol, and cleared in 0.25% KOH/50% glycerol.

RNA was isolated from at least 30 3 dpf larvae from independent mating crosses of fish. Larvae were homogenized in TRIzol Reagent (15596018, ThermoFisher Scientific; Waltham, MA), and RNA extraction was performed as previously described (55). RNA (1 µg) obtained from samples was treated with DNase I before inactivation by EDTA chelation. Samples were mixed with an 18-nucleotide poly-T primer and allowed to anneal before adding dNTPs and M-MLV Reverse Transcriptase (M1701, Promega; Madison, WI).

rgp1 cDNA was isolated from total WT zebrafish cDNA by PCR amplification and purification (28104, Qiagen; Hilden, Germany). The cDNA was inserted into pGEM-T using TA subcloning (A1360, Promega; Madison, WI). Following restriction enzyme digestion, the insert was moved into pCS2+ and linearized by NotI digestion. Synthesized capped mRNA transcripts were produced via in vitro transcription with mMESSAGE mMACHINE SP6 transcription kit (AM1340, ThermoFisher Scientific; Waltham, MA). 400 pg of mRNA diluted in 0.3X Danieau Buffer was micro-injected into single cell-stage embryos. Embryos were grown to 4 dpf for subsequent visualization and live imaging. An observer blinded to injection content scored larvae for the extent of the protruding jaw phenotype.

Transgenic constructs for overexpression of zRgp1, zRab6a, and hRAB8A were generated as previously described using the Gateway cloning-based Tol2kit method (56). Full-length constructs for pME were PCR-amplified with attB1 and attB2 sequences and cloned into pDONR221 using BP clonase II (11789020, ThermoFisher Scientific; Waltham, MA). p3E constructs were PCR-amplified with attB2R and attB3 sites and cloned into pDONR p2R-p3 using BP clonase II. A Multisite Gateway Cloning LR reaction was performed using the p5E-1.7-kb Col2a1a promoter, pME vectors, p3E vectors, pDEST-Tol2pA2, and LR clonase II enzyme mix (11791020, ThermoFisher Scientific; Waltham, MA) (57, 58). The resulting destination clones were used for mosaic overexpression experiments.

Zebrafish embryos at the one-cell stage were injected with 50-100 pg medaka transposase mRNA and 10-100 pg recombined pDEST plasmid. Larvae were collected at 3 dpf for sample processing and imaging.

Site-directed mutagenesis was performed on plasmids using the Q5 Site-Directed Mutagenesis Kit (E0554S, NEB; Ipswich, MA) as previously described with the following modifications (59). Template DNA was combined with sequence-specific SDM primers generated using NEBaseChanger online software ( Table 1 ). Following Q5 PCR amplification, template DNA was degraded, and the newly synthesized product was ligated using the included Kinase/Ligase/DpnI enzyme mix. The final plasmid product was transformed into DH5α E. coli for subsequent plasmid isolation.

Larvae at 3 dpf were fixed overnight in 4% PFA, washed, and moved to 30% sucrose overnight. The following day, larvae were mounted in OCT embedding medium (6769006, ThermoFisher Scientific; Waltham, MA), frozen, cryosectioned, and transferred onto Superfrost slides (12-550-15, ThermoFisher Scientific; Waltham, MA) as previously described (60). Slides were dried on a heat block, rehydrated in PBS, and mounted onto a Sequenza staining rack. Sections were incubated with Proteinase K for antigen retrieval and permeabilized in PBS+0.5% Triton X-100, then blocked in 2% BSA and 2% NGS. Samples were stained with primary antibodies ( Table 2 ) diluted in blocking solution and incubated overnight at 4°C. The following day, samples were rinsed and incubated in secondary antibodies ( Table 2 ), rinsed again, and incubated in DAPI before being mounted using Prolong Gold Antifade Agent (P36930, ThermoFisher Scientific; Waltham, MA). Fluorescent imaging was carried out using an AxioImager Z1 equipped with an Apotome and an EC Plan Neofluar 100X/1.30 Oil objective and ORCA-Flash4.0 Digital CMOS Camera. Co-localization analysis of CHP and LBPA was performed using FIJI’s Coloc2 function. Cell boundaries were used to highlight a region of interest, which was then used to make a mask. Pearson’s R values were collected from cells across WT and rgp1^-/- larvae.

Images of collagen IF were analyzed in FIJI. For collagen accumulation in chondrocytes, the intracellular area of each cell (demarcated by caax-eGFP signal), collagen area (Col2 signal), and nuclear area (DAPI) were measured. The following formula was used to calculate the percent intracellular collagen area per cell (18):

Larvae were fixed at 3 dpf in 2.5% glutaraldehyde in 0.1 M sodium cacodylate by incubating at room temperature for 1 hour first, then overnight at 4°C, as previously described (18). After rinsing in 0.1 M sodium cacodylate, samples were post-fixed with 1% osmium tetroxide in 0.1 M sodium cacodylate for 1 hour. After additional rinsing, specimens were sequentially dehydrated in increasing concentrations of ethanol and then propylene oxide, infiltrated with resin step-wise, and then embedded in resin for 48 hours at 60°C. Coronal sections of 10 µm were collected until craniofacial chondrocytes were visible under low magnification microscopy based on their location in the jaw, characteristic stacking arrangement, and surrounding ECM. 70-nm sections were collected on TEM grids following sectioning on a Leica Ultracut Microtome and analyzed on a Phillips CM-12 Transmission Electron Microscope provided by the Vanderbilt Cell Imaging Shared Resource. Images of the hyosymplectic element were acquired from tiled images and stitching across single sections. Individual ultrastructural elements were identified by morphological characteristics and were pseudo colored.

Live vesicle movement in WT and rgp1^-/- transgenic larvae was performed on a Zeiss LSM880 confocal microscope. Embryos obtained from gRNA-generated stable mutant lines crossed to mosaic (Col2a1a:caax-eGFP) adults were injected with a Tol2 construct expressing Col2a1a:mCherry-zRab6a for transient, transposon-mediated DNA integration. Zebrafish larvae at 3 dpf that were grown in PTU were selected for imaging if they expressed caax-eGFP and expressed substantial mCherry-zRab6a in the hyosymplectic cartilage. Larvae were anesthetized in 0.15 mg/mL Tricaine, embedded in 1.2% low-melt agarose (A20070, RPI; Mount Prospect, IL) on glass-bottom confocal dishes, and overlaid with embryo media containing 0.4 mg/mL Tricaine. For time-lapse imaging, larvae were imaged under a 40X/1.1 LD C-Apochromat water-immersion objective. Frame intervals of 300 ms were collected over the course of 300 frames for each sample with 300 ms delay between acquisitions. Individual frames were deconvolved using Airyscan (Zeiss; Jena, Germany). Supplemental Movies 1 , 2 are sped up to 100 frames/second for display purposes.

Confocal time-lapse images from live transgenic larvae were analyzed in Imaris 9.7.2 (Oxford Instruments; Abingdon, United Kingdom). Analysis was performed using the automated Spots feature. Spots in the red channel were filtered by their quality and fit to the autoregressive motion-tracking algorithm. Tracks generated over longer than 5 frames were selected for analysis. All tracks collected were measured for track length and mean track speed.

To identify double-stranded DNA breaks indicative of cell death, TUNEL assays were performed as previously described with the following modifications (61). Larvae were fixed overnight in 4% PFA, moved to 30% sucrose in PBS, and cryosectioned. TUNEL was performed using the Roche In Situ Death Detection Kit (12156792910, Sigma; St. Louis, MO). Larvae were co-stained with WGA-488 or GFP-488 and DAPI.

Synthesized cDNA was diluted 1:1 with water before use in qPCR. Three biological replicates were performed, with each biological replicate consisting of an independent RNA extraction. qPCR was performed using forward and reverse primers ( Table 1 ), cDNA, and 2X SYBR Green PCR Master Mix (4309155, ThermoFisher Scientific; Waltham, MA) as previously described (62). Samples were run in technical triplicates, and curves were normalized to ef-1 transcript levels.

Statistical analyses were performed with GraphPad Prism 9.3.1 (63). Statistical significance between WT and rgp1^-/- larva gross morphology was assessed by Mann-Whitney U-test (two-tailed) with 95% confidence intervals. For head length and cartilage size measurements, symbols on the graph represent individual larvae, with lines representing means ± SEM.

Statistical significance for jaw protrusion phenotypes following mRNA injection was analyzed by Chi-Square analysis. Numbers on the graph represent larvae across pooled clutches from injections across at least three independent mating crosses.

Statistical significance between mCherry-zRab6a puncta dynamics in WT and rgp1^-/- chondrocytes was assessed by Mann-Whitney U-test (two-tailed) with 95% confidence intervals. Symbols on graph represent individual vesicles with red lines representing means.

Comparisons of intracellular collagen II staining between fluorescent and non-fluorescent chondrocytes was determined by one-way ANOVA with Tukey’s multiple comparison test. Symbols on the graph represent measurements of individual chondrocytes with lines representing means ± SEM.

Statistical analysis of cell death phenotypes in TEM of chondrocytes was not performed. Numbers on the graph represent the fraction of total individual chondrocytes with each phenotype.

Statistical analysis between qPCR expression levels was assessed by Mann-Whitney U-test (two-tailed) with 95% confidence intervals. Symbols on the graph represent individual biological replicates from technical triplicates with lines representing means ± SEM.

No data points or measurements were omitted across any experiment. The number of larvae, vesicles, or cells required for each experiment was determined based on power analysis with 95% confidence intervals. Significance is represented by *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; ns, not significant.

Zebrafish begin to develop a protruding jaw around 3 days post-fertilization (dpf), a process essential for feeding that begins at 5 dpf. Previous work from our laboratory noted that mosaically reduced Rgp1 expression contributes to defects in jaw morphology and craniofacial development (18). However, it remains unknown how genetic mutations in Rgp1 affect craniofacial cartilage development. To generate stable mutant lines, we used CRISPR/Cas9 genome editing and two guide RNAs (gRNAs) targeting exons one (g1) and seven (g3) to generate three independent rgp1 mutant alleles ( Figures 1A, B ). All three zebrafish alleles are predicted to induce frameshift mutations, which would likely result in the translation of a shorter peptide or nonsense mediated decay ( Figure 1C ). All three rgp1^-/- alleles showed a lack of a protruding jaw and were shorter when compared to wild-type (WT) controls ( Figure 1D ). The embryonic zebrafish craniofacial skeleton is primarily cartilaginous until 5 dpf (64). To assess the cartilage structure and integrity, we stained larvae with Alcian blue, a dye that binds glycosaminoglycans commonly found in the cartilage extracellular matrix ( Figure 1E ) (54). In rgp1^-/- larvae, we found malformed craniofacial cartilage elements, leading to an overall shortening of the head skeleton by approximately 20% compared to WT controls ( Figure 1F , p<0.0001). The Meckel’s cartilage, corresponding to the lower jaw, is significantly shorter in rgp1^-/- larvae ( Figures 1E, G , p<0.0001). Additionally, the ceratohyal cartilage element, a ventral structure of the head skeleton, was bent at a wider angle in mutants compared to WT controls ( Figures 1E, H , p<0.0001).

To validate our new model and rule out the possibility of CRISPR-induced off-target effects, we performed an mRNA genetic replacement experiment and assessed jaw morphology in live larvae to determine if clutches overexpressing rgp1 had a majority of larvae with normal protruding jaws ( Figures 1I–J ) (65–67). There was an increase in the proportion of larvae showing a protruding jaw in WT mRNA injected larvae, compared to the non-injected control of 75% larvae with protruding jaws, typical for a recessive phenotype ( Figure 1K , p<0.0001). On the contrary, injection of an mRNA coding for the mutant rgp1^{g1 -5} allele did not result in phenotypic jaw rescue ( Figure 1K , p=0.502). The rgp1^{g1 -5} allele was used in panels D-K, although all three alleles presented with a similar phenotype. Similarly, trans-heterozygote larvae for different rgp1 mutant alleles had similar jaw protrusion defects to homozygous mutant alleles, further ruling out the possibility of off-target effects in individual rgp1 mutant lines (data not shown). From the mRNA genetic replacement experiment, we concluded that the observed craniofacial skeletal defects were due to mutations in rgp1 and its subsequent loss of function.

Rgp1 was previously identified as a component of the Rab6a GEF complex in yeast and mammalian cells (44, 45). To test whether mutations in rgp1 disrupt Rab6a activation and vesicle movement in our zebrafish model, we generated and injected a Tol2 DNA construct to overexpress mCherry-fused WT zebrafish Rab6a (WT zRab6a) under a Col2a1α promoter (57, 58, 68). We co-injected the construct along with transposase into the single-cell progeny of heterozygote mutant fish carrying a transgene that labels chondrocyte membrane with caax-eGFP ( Figure 2A ) (69). Because transposase integration is mosaic, not all cells express the construct, and the mCherry fusion protein highlights specific vesicular compartments. This gave us the opportunity to compare activation of Rab6a and vesicular compartment movements as puncta in movies. We chose to analyze hyosymplectic chondrocytes that stack horizontally, develop around 3 dpf, and are close to the zebrafish skin, making them amenable to in situ live imaging ( Figures 2B, C ) (64). We found that the mCherry-zRab6a⁺ vesicular compartments in rgp1 mutants traveled shorter distances ( Figure 2D , p<0.0001) and moved slower ( Figure 2E , p<0.0001) than WT controls ( Supplemental Movies 1, 2 ). These data suggest that Rgp1 is needed for Rab6a⁺ vesicular compartment dynamics within chondrocytes.

Craniofacial chondrocytes predominantly secrete collagen II during larval craniofacial development (68). To assess secretion of collagen II, we performed immunofluorescence with a collagen II antibody in WT and rgp1^-/- larvae crossed to a transgenic zebrafish line expressing membrane-bound caax-eGFP under the Col2a1α promoter ( Figure 3A ) (57, 69, 70). The caax-eGFP transgene labeled the cell membrane boundary to distinguish intracellular from extracellular space. We identified two prominent phenotypes in the rgp1^-/- chondrocytes: collagen II secretion defects and an abnormal cell shape; we have chosen to focus on the collagen II secretion phenotype. We found that rgp1^-/- chondrocytes had significantly higher amounts of intracellular collagen II when compared to WT controls ( Figures 3B, C , p<0.0001). We also observed reduced extracellular collagen II staining in rgp1^-/- cartilage. Thus, we conclude that Rgp1 is required for secretion of collagen II in chondrocytes.

To test whether the intracellular collagen II accumulations stem specifically from loss of Rgp1, we overexpressed WT zebrafish Rgp1 (zRgp1) under a Col2a1α promoter using a Tol2 construct ( Figure 3D ). We took advantage of a viral p2A self-cleavable peptide sequence to identify cells over-expressing WT zRgp1 (56, 71). Single-cell embryos were injected, and the construct mosaically integrated, with cytosolic mCherry expression only visible in cells where the Tol2 construct integrated ( Figure 3E ). WT chondrocytes had no significant difference in the amount of intracellular collagen II regardless of whether they overexpressed zRgp1 or not ( Figure 3F , p=0.960). In rgp1-deficient chondrocytes, expression of zRgp1-p2A-mCherry ( Figure 3E ) reduced accumulations of intracellular collagen II compared to mCherry^- chondrocytes ( Figures 3E, F , p<0.0001). Thus, overexpression of WT zRgp1 in rgp1^-/- chondrocytes was sufficient to rescue collagen II secretion defects in a cell-autonomous manner.

To assess the abundance of extracellular matrix proteins in an antibody-independent method, we used TEM. We found that at 3 dpf, WT chondrocytes have secreted a large amount of protein to the ECM, with increasing crosslinking further away from the cell membrane ( Figure 3G ). However, rgp1-deficient cartilage ECM was significantly less dense compared to WT ( Figure 3H , p<0.0001), although there were ECM proteins present in the extracellular space. This supports our observation that although collagen II is likely not secreted by rgp1^-/- chondrocytes, other ECM proteins are secreted to the extracellular space.

To assess where collagen II accumulates in rgp1^-/- chondrocytes, we compared TEM images of WT and rgp1^-/- hyosymplectic chondrocytes ( Figure 4A ). In rgp1^-/- cells, we found large vacuolar-like structures, ( Figure 4B ) some of which were filled with striated assemblies ( Figure 4B’-B” ). Previous studies have identified this striated TEM ultrastructure as a fully assembled, mature collagen fibril, which is typically found exclusively extracellularly (72, 73). Collagen II is normally trafficked with N- and C-terminal telopeptide domains that prevent aberrant oligomerization intracellularly (4, 74, 75). The presence of striated collagen intracellularly suggests that endopeptidases cleaved the procollagen trimer, thus allowing for spontaneous fibril assembly within the observed vacuolar-like structures.

Lysosomes are small, vacuolar compartments that concentrate proteolytic peptidases for enzymatic digestion (76). To test whether the structures observed in TEM carry hallmarks of lysosomal membranes containing collagen, we co-stained for markers of endolysosomal compartments and collagen. To perform this co-localization assay, we used an antibody against a lipid enriched in endolysosomal membranes, lysobisphosphatidic acid (LBPA), and a collagen hybridizing peptide (CHP; Figure 4C ) (77, 78). We chose to use CHP instead of a collagen II antibody since the collagen II antibody and the LBPA antibody were generated in the same host species. In WT chondrocytes, we found LBPA staining in small puncta, consistent with staining of endolysosomal compartments; we also found very little co-localization between LBPA and CHP, suggesting that collagen is not normally found in LBPA⁺ compartments ( Figure 4C ). However, CHP and LBPA staining often co-localized in rgp1^-/- chondrocytes ( Figures 4C, D , p<0.0001) and was as large as the inclusions in the TEM images. This suggests that collagen II accumulates in an endolysosomal compartment in rgp1^-/- chondrocytes.

Rab8a has been shown to mainly associate with exocytic trafficking from the trans-Golgi network (TGN) to the plasma membrane and co-localizes with Rab6a⁺ vesicular carriers (79–83). To test whether overexpression of hRAB8A is sufficient to affect collagen II secretion in zebrafish chondrocytes, we used WT mCherry-hRAB8A under a Col2a1α promoter ( Figure 5A ). In WT larvae, there was no difference in intracellular collagen II levels between mCherry⁺ and mCherry^- cells ( Figure 5B, C , p=0.9999). In rgp1^-/- chondrocytes, we found that overexpression of the WT mCherry-hRAB8A also had no effect on intracellular collagen II when compared to mCherry^- cells ( Figures 5B , 5C , p=0.9993). Thus, simple overexpression of WT hRAB8A does not alter collagen II secretion.

Point mutations in Rab proteins can alter their affinity for GTP or GDP to lock it in the “constitutively-active” (CA), GTP-bound conformation or the “dominant-negative” (DN), GDP-bound conformation (48, 84–86). To test whether inactivating Rab8a blocks collagen II secretion in zebrafish chondrocytes, we expressed a Tol2 fusion construct of GFP and DN human RAB8A (T22N) under a Col2a1α promoter ( Figure 5D ) (85). We found that GFP⁺ WT chondrocytes accumulated large inclusions of collagen II in higher amounts compared to their adjacent GFP^- counterparts ( Figures 5E, F , p=0.0004). In rgp1^-/- chondrocytes that already accumulate large collagen II inclusions, there was no change in intracellular collagen II staining in GFP⁺ vs GFP^- cells ( Figure 5F , p>0.9999). This result is consistent with Rab8a being required for cargo secretion, including collagen II in chondrocytes.

To test whether constitutive activation of hRAB8A would be sufficient to rescue collagen II accumulations in rgp1^-/- chondrocytes, we expressed a Tol2 fusion construct of GFP and CA hRAB8A (Q67L) under a Col2a1α promoter ( Figure 5G ) (86). In WT chondrocytes, there was no difference in intracellular collagen II levels between GFP⁺ and GFP^- cells ( Figures 5H, I , p=0.9890). In rgp1^-/- chondrocytes, we found that expression of CA GFP-hRAB8A ( Figure 5H ) reduced levels of intracellular collagen II compared to GFP^- cells ( Figures 5H, I , p<0.0001). Therefore, overexpression of CA Rab8a is sufficient to partially rescue secretion of intracellular collagen II in rgp1^-/- chondrocytes, suggesting that Rgp1 functions upstream of the Rab8a activation step.

In TEM analysis at 3 dpf, we observed rgp1^-/- chondrocytes with abnormal nuclei when compared to WT cells ( Figures 4A, B ). The observed mixture of electron-dense and electron-sparse material within the nucleus is suggestive of nuclear condensations, an early sign of cell death. To identify hallmarks of cell death, we further assessed chondrocyte ultrastructure using TEM on hyosymplectic chondrocytes (87, 88). While WT chondrocytes did not have any abnormalities in nuclear structure, rgp1^-/- chondrocytes had begun to undergo nuclear condensation and presented with clear cytoplasmic vacuolar structures ( Figures 6A, B ). Specifically, binary scoring for the presence or absence of these phenotypes revealed that hallmarks of cell death were only sporadically observed in the WT chondrocytes ( Figure 6C ). As such, these findings suggest that rgp1^-/- chondrocytes undergo cell death at higher rates than WT chondrocytes.

To test if rgp1^-/- chondrocytes underwent DNA fragmentation consistent with the nuclear condensation observed in TEM, we performed a TUNEL assay. As a co-stain, we used wheat germ agglutinin (WGA) to label N-glycosylated proteins in the extracellular matrix and to help detect cell boundaries ( Figure 6D ). We found more TUNEL⁺ nuclei in rgp1^-/- chondrocytes compared to WT nuclei, indicative of higher levels of DNA fragmentation ( Figure 6E , p<0.0001). Together, the nuclear changes detected by TEM and TUNEL staining support the increase in chondrocyte cell death in Rgp1-deficient cells.

It remained unclear whether the observed cell death occurred as a result of collagen II accumulations or if another Rgp1-related phenotype induced cell death. To test whether collagen II accumulation induced cell death, we overexpressed DN GFP-hRAB8A ( Figure 5D ) in WT chondrocytes and compared TUNEL staining in GFP⁺ cells with intracellular collagen II accumulations to GFP^- chondrocytes with no collagen II backlogs ( Figure 6F ). We found no difference in the percentage of nuclei with DNA fragmentation between the GFP⁺ and GFP^- chondrocytes at 3 dpf ( Figure 6G , p>0.9999); this observation was consistent even after prolonged accumulation of collagen II at 4 dpf ( Figure 6G , p=0.3294). Thus, we conclude that the chondrocyte cell death in rgp1^-/- chondrocytes is caused independently from collagen II accumulation.

Since collagen II accumulation was not sufficient to induce cell death, we investigated whether defects in Golgi trafficking could cause cellular stress responses along the secretory pathway. To test this possibility, we used qPCR to measure transcript levels of three ER stress transcripts, edem1, hspa5 (BiP), and ddit3 (CHOP), and five Golgi stress transcripts, arf4a, arf4b, tfe3a, tfe3b, and acbd3 in 3 dpf larvae ( Figure 6H ). However, rgp1^-/- larvae significantly upregulated expression of arf4b ( Figure 6H , p=0.0286), an ortholog of human ARF4 (GTPase of the Ras superfamily of small G proteins), that was previously shown to be upregulated in response to Golgi stress conditions impinging on Golgi integrity and homeostasis (89). Upregulation of Arf4b and the fact that TEM imaging of rgp1^-/- chondrocytes failed to identify canonical Golgi stacks (data not shown) similar to those observed in WT siblings, suggests that rgp1 loss-of-function leads to Golgi stress and Golgi fragmentation.