Decision letter: Inhibiting the integrated stress response pathway prevents aberrant chondrocyte differentiation thereby alleviating chondrodysplasia

Article Figures and data Abstract Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract The integrated stress response (ISR) is activated by diverse forms of cellular stress, including endoplasmic reticulum (ER) stress, and is associated with diseases. However, the molecular mechanism(s) whereby the ISR impacts on differentiation is incompletely understood. Here, we exploited a mouse model of Metaphyseal Chondrodysplasia type Schmid (MCDS) to provide insight into the impact of the ISR on cell fate. We show the protein kinase RNA-like ER kinase (PERK) pathway that mediates preferential synthesis of ATF4 and CHOP, dominates in causing dysplasia by reverting chondrocyte differentiation via ATF4-directed transactivation of Sox9. Chondrocyte survival is enabled, cell autonomously, by CHOP and dual CHOP-ATF4 transactivation of Fgf21. Treatment of mutant mice with a chemical inhibitor of PERK signaling prevents the differentiation defects and ameliorates chondrodysplasia. By preventing aberrant differentiation, titrated inhibition of the ISR emerges as a rationale therapeutic strategy for stress-induced skeletal disorders. https://doi.org/10.7554/eLife.37673.001 Introduction The Integrated Stress Response (ISR) is a eukaryotic cellular stress response, aiming to restore cellular homeostasis upon different types of extrinsic or intrinsic stresses. The ISR can be stimulated by a range of physiological or pathological changes (Brostrom et al., 1996; Dever et al., 1992; Harding et al., 2003; Ron, 2002; Wek et al., 2006), including hypoxia, amino acid deprivation, glucose/nutrition deprivation, viral infection (Dever et al., 1992; Harding et al., 2003; Wek et al., 2006; García et al., 2007; Rzymski et al., 2010; Ye et al., 2010) and intrinsic endoplasmic reticulum (ER) stress (Harding et al., 1999), which is caused by the accumulation of unfolded or misfolded proteins within ER. Furthermore, in the context of cancer biology, oncogene activation can also trigger the ISR (Denoyelle et al., 2006; Hart et al., 2012). Although the ISR is primarily a pro-survival homeostatic program, aiming to optimize the adaptive cellular response to stress, exposure to severe stress, either in intensity or duration, will overwhelm the capacity of this adaptive response and drive signaling toward cell death. The key early controlling step in the ISR is the phosphorylation of eukaryotic translation initiation factor 2 alpha (eIF2α) by one of four members of eIF2α kinase family: protein kinase R (PKR), PKR-like endoplasmic reticulum kinase (PERK), general control nonderepressible 2 (GCN2), and Heme-regulated eIF2α kinase (HRI) (Dever et al., 1992; García et al., 2007; Harding et al., 1999; Han et al., 2001). These kinases phosphorylate serine 51 in eIF2α, promoting the formation of a p-eIF2α and eIF2B complex, consequently inhibiting the guanine nucleotide exchange activity of eIF2B (Sudhakar et al., 2000). Inactivation of the eIF2 complex leads to a shutdown of global protein synthesis but also the induction of preferential translation of transcripts notably the mRNAs of transcription factors, Activating Transcription Factor 4 (ATF4) and C/EBP homologous protein (CHOP, encoded by Ddit3), and other factors with both pro-survival and pro-death functions which aid cell adaption and recovery (Figure 1—figure supplement 1A) (Harding et al., 2000). eIF2α phosphorylation is transient and can be reversed by PPP1R15A (GADD34) or CReP1, the regulatory subunit of eIF2α phosphatases, acting in a negative feedback loop that allows protein synthesis to restart (Harding et al., 2009; Novoa et al., 2001). When the stress is intense or prolonged, cells fail to adapt, and apoptosis is triggered. It is likely that the duration and level of eIF2α phosphorylation, as well as ATF4 regulation, determine the balance between cell survival and cell death. There are more than 400 human genetic skeletal disorders caused by disrupted cartilage and bone development and growth, commonly resulting in dwarfism and skeletal deformities (Geister and Camper, 2015). Many of these disorders are caused by mutations in genes for extracellular matrix (ECM) proteins (Geister and Camper, 2015; Bonafe et al., 2015; Briggs and Chapman, 2002; Warman et al., 1993), and some mutations cause inappropriate folding, processing or export, leading to retention in the ER, affecting the secretory and stress response pathways (Wilson et al., 2005; Tsang et al., 2007; Tsang et al., 2010; Posey et al., 2012; PirogPiróg-Garcia et al., 2007; Hartley et al., 2013; Cameron et al., 2011; Boot-Handford and Briggs, 2010; Arnold and Fertala, 2013). Accumulation of misfolded proteins in the ER can overwhelm the protein quality control mechanism, causing proteotoxicity, cell cycle arrest and cell death. Cells activate the unfolded protein response (UPR), which mediates cell survival by slowing protein translation, promoting proteostasis via the proteasome and activating transcription factors that upregulate the production of protein chaperones (reviewed in [Hotamisligil and Davis, 2016; Horiuchi et al., 2016]). The UPR employs three arms of sensors in the ER to mediate cell adaptation and survival under ER stress: the key component of the ISR, PERK, inositol-regulated enzyme 1α (IRE1α), and the activating transcription factor 6 (ATF6) family (Hotamisligil and Davis, 2016; Horiuchi et al., 2016; Walter and Ron, 2011). Upon ER stress, the PERK-p-eIF2α signaling modulates the cell adaptation via translational control. ATF6 family factors move from the ER to the Golgi, are processed by S1 and S2 proteases, and translocate to the nucleus to activate ER quality control genes such as Hspa5 (encodes BiP) and Xbp1 (X-box binding protein 1). IRE1α has kinase and endoribonuclease (RNase) activities. It catalyzes the splicing of Xbp1 mRNA, generating the UPR transcription factor XBP1S that upregulates genes encoding chaperones and proteins involved in ER-associated protein degradation (ERAD). Human metaphyseal chondrodysplasia type Schmid (MCDS; MIM156500) is an autosomal dominant disorder caused by heterozygous mutations in the NC1 domain of type X collagen, encoded by COL10A1 in hypertrophic chondrocytes (HC) (Warman et al., 1993; Wilson et al., 2005; Mäkitie et al., 2005). In an MCDS transgenic mouse model (13del), carrying a 13 bp deletion in Col10a1 equivalent to the human mutation, misfolded mutant collagen X induces ER stress suggesting the primary role of ER stress in MCDS pathogenesis (Tsang et al., 2007). Although the chondrocytes survive, their differentiation is reversed by an unknown mechanism to a more juvenile state characterized by the re-expression of prehypertrophic chondrocyte markers (Ppr, Sox9 and Col2a1), disrupting endochondral ossification, and skeletal dysplasia ensues. The causative role of ER stress in MCDS is further supported by studies showing that expression of an exogenous misfolded protein in HCs can induce an MCDS-like phenotype (Rajpar et al., 2009). It is noteworthy that the skeletal defects caused by mutations that induce stress or inactivate critical transducers of the stress response in humans (Julier and Nicolino, 2010) and mouse models (Tsang et al., 2007; Cameron et al., 2011; Horiuchi et al., 2016; Rajpar et al., 2009; Julier and Nicolino, 2010; Nundlall et al., 2010; Cameron et al., 2015a2015; Cameron et al., 2015b) strongly implicate components of ER stress-induced UPR pathways involved in chondrocyte and osteoblast differentiation (Tsang et al., 2010; Horiuchi et al., 2016). Expression of misfolded cartilage oligomeric matrix protein (COMP) in proliferating and hypertrophic chondrocytes results in chondrocyte disorganization and causes Pseudoachondroplasia (PSACH), suggesting triggering the UPR may underlie many chondrodysplasias where mutations cause accumulation of misfolded proteins (Posey et al., 2012; Piróg et al., 2014). Another example that links ER stress signaling to chondrocyte differentiation is illustrated by studies on BBF2H7, an ER stress transducer. Bbf2h7 null mutants show severe chondrocyte abnormality due to proliferation and differentiation defects, indicating its essential role for chondrogenesis (Saito et al., 2009; Saito et al., 2014). Furthermore, BBF2H7 has been shown to suppress chondrocyte hypertrophy by direct regulation of IHH signaling and the IHH-PTHrP pathway (Saito et al., 2014). Pharmacological stimulation of intracellular proteolysis of mutant collagen X in another MCDS mouse model reduces the level of ER stress, partially ameliorating the dwarfism phenotype (Mullan et al., 2017). However, the rescue of the chondrocyte differentiation defect, expansion of the hypertrophic zone and bone growth was incomplete, suggesting stimulating degradation of misfolded protein alone is not sufficient to address the impact of the ER stress on aberrant cell differentiation. The relative contribution of the arms of the UPR and its constituent components to the pathology and a molecular understanding of the consequences of activation of ER stress on cell fate and differentiation in vivo is lacking. Here by studying MCDS in a mouse model, we sought to clarify the mechanism(s) by which the UPR/ISR causes aberrant differentiation defects which may provide the basis for the development of novel therapeutic target(s) for treating this disorder, or other ER stress-related skeletal disorders. We reveal by genetic, biochemical and in vivo functional approaches in mouse models, ISR-mediated preferential translation of ATF4 directly activates inappropriate expression of the key transcription factor SOX9. This ectopic SOX9 expression in HCs reverts chondrocyte differentiation, thereby causing MCDS. By targeting the ISR early, at the level of p-eIF2α induction of ATF4, we ameliorate the pathology, thereby providing a rationale pharmacological strategy for treating MCDS and other skeletal disorders caused by activation of the ISR. Results The ER stress-induced UPR disrupts global transcriptome patterns in the chondrodysplastic growth plate The mammalian growth plate comprises four major sub-populations of chondrocytes: resting, proliferating (PC), prehypertrophic (pHC) and hypertrophic chondrocytes (HC). These chondrocytes have distinct morphologies and gene expression profiles governed by a precisely tuned gene regulatory network (Hojo et al., 2016). To investigate the effect of ER stress on the transcriptome of HCs, the proximal tibial growth plates from 10-day-old WT and 13del mice were fractionated into sub-populations representing proliferating (PZ), prehypertrophic (pHZ) and hypertrophic chondrocytes (HZ) (Figure 1A; Figure 1—figure supplement 1B). The wild-type HZ was fractionated into upper and lower zones (UHZ and LHZ) to capture early onset and late phases of hypertrophy. The 13del HZ was fractionated into three zones: upper (UHZ) corresponding to the early phase of UPR activation, middle (MHZ) where HC adaptation would be initiated, and lower (LHZ) where HC should fully adapt. Figure 1 with 3 supplements see all Download asset Open asset PERK signaling pathway, activated by ER stress, plays an etiological role in MCDS 13del mice. (A) Schematic diagram of the rationale for fractionating the WT and 13del p10 growth plates into different chondrocyte populations. (B) Clustering analysis of differentially expressed genes in chondrocyte subpopulations in p10 WT and 13del proximal tibial growth plates. Expression levels were normalized from −0.5 (blue) to 0.5 (yellow). Four major clusters were identified. (C) The average expression levels (Log2 scale) of the genes in different clusters revealed significant expression pattern changes in 13del mice. (D–E) Genes in different clusters were functionally categorized using DAVID web tools. The enriched biological processes (D) and enriched pathways (E) were sequentially shown for Cluster I, II, III and IV. The values on the X-axis represented the Log10 (p-value−1). Each category with p-value <0.05 was considered as significantly enriched. (F) Enriched motifs on Cluster I genes were identified, using sequences of the promoter region (±2 kb from the TSS) for these genes. Motifs matched to the TFs in the UPR were shown. https://doi.org/10.7554/eLife.37673.002 We used k-means clustering (see Materials and methods) to categorize the gene expression patterns across different zones in wild-type and 13del growth plates into four clusters (Supplementary file 1). Genes (453) in Cluster I increased expression from PHZ to lower HZ specifically in 13del HC (Figure 1B and C). Ontological analyses show these differentially expressed genes are mainly involved in protein processing in the ER and the UPR (Figure 1D and E, Supplementary file 2 and 3). Genes in Clusters II (659) and III (314) showed highest expression in wild-type PZ and pHZ followed by progressive downregulation from pHZ to LHZ but were upregulated in 13del LHZ, reflecting UPR-induced changes in HC differentiation (Figure 1B and C). These genes included Sox9, Ppr, and Ihh, consistent with our previous report of re-expression of pre-hypertrophic markers (Tsang et al., 2007). Cluster IV genes (680) showed increasing expression from pHZ to LHZ in wild-type and can be defined as ‘hypertrophy characteristic’ genes. Consistent with a change in the HC de-differentiation state in 13del, these genes were down-regulated in 13del LHZ. The concomitant down-regulation of Cluster I stress response genes in 13del LHZ is consistent with the alleviation of the stress in the reprogrammed cells and an adapted state. We further compared our microarray dataset with the published data (Cameron et al., 2011) from another 2 MCDS mouse models, expressing a Col10a1 p.N617K mutation or an ER stress-inducing form of thyroglobulin (Tgcog) (Cameron et al., 2011). Overall, the gene expression changes detected in all datasets shared some degree of similarity, and 227 genes were commonly changed, showing the activation of ER stress signaling in the MCDS chondrocytes, up-regulation of genes such as Fgf21 and down-regulation of genes such as Ldb3. Differences were also found, and 473 genes were specifically changed in 13del, such as Apoa4 (up-regulated) and Atp2a1 (down-regulated) (Supplementary file 4). The differences from the published datasets may be due to the different mouse models, the time points analyzed (p14 versus 13del at p10), and also the methods of sampling the various chondrocyte populations. Those transcriptomes were derived from whole proliferative and hypertrophic zones, while ours were generated from precisely fractionated chondrocyte populations from the growth plate. PERK-p-eIF2 signaling is the major contributor to chondrocyte adaptation to ER stress We investigated the contributions of the UPR arms, PERK, IRE1, and ATF6 to the HC response to ER stress. By ontology and pathway analyses of Cluster I, we found enrichment of genes in the PERK pathway and IRE1-Xbp1S regulated ERAD, but not for ATF6 signaling (Figure 1D and E; Figure 1—figure supplement 1C; Supplementary file 2 and 3). Activation of PERK signaling in 13del HC was demonstrated by up-regulation of p-eIF2α and its downstream components (Atf4, Atf3, Ddit3, Ero1l and Ppp1r15a) (Figure 1—figure supplement 2A and C) which were validated by in-situ hybridization and immunostaining (Figure 1—figure supplement 2B and C). Using Motif enrichment analysis, we found that the binding motifs of CHOP (encoded by Ddit3) and ATF4 were highly enriched in Cluster I, but not those for Xbp1S or ATF6 (Figure 1F and Supplementary file 5). By interrogating ATF4 and CHOP ChIP-seq data (Han et al., 2013), we found significant over-representation of ATF4 (odds ratio = 2.87, p<0.0001) and CHOP (odds ratio = 4.33, p<0.0001) binding peaks associated with the genes from Cluster I but not for the other clusters (Supplementary file 6). Cluster I genes are therefore most likely to be directly regulated by UPR-associated transcription factors. Together, these data suggest a more prominent contribution of the PERK-p-eIF2 signaling pathway than that of Xbp1S, which is consistent with another MCDS mouse model study that found inactivation of Xbp1 in HCs did not alter the severity of dwarfism (Cameron et al., 2015b). To test this notion, we ectopically expressed Xbp1S in HCs in transgenic mice (Figure 1—figure supplement 3A). Overexpression of Xbp1S specifically in HCs did not affect the growth plate (Figure 1—figure supplement 3B and C). ATF4 expression in hypertrophic chondrocytes reprogrammes differentiation Apart from its role in the stress response, ATF4 is also required for chondrocyte differentiation through direct activation of Ihh (Wang et al., 2009). In the E14.5-E17.5 fetal growth plate, ATF4 is expressed in differentiating chondrocytes, with the highest expression in pHCs (Figure 2A). ATF4 expression in HCs progressively decreases after birth and by p10 levels are especially lowered in the LHZ (Figure 2A). Therefore, the chondrocyte differentiation defects in the MCDS model might be directly caused by ectopic overexpression of ATF4 in HCs, as a consequence of the preferential translation of Atf4 transcripts modulated by p-eIF2α. Figure 2 with 1 supplement see all Download asset Open asset Ectopic activation of ATF4 in HCs recaptures the phenotypes of 13del mice. (A) Expression profiles of ATF4 (a–f) on the WT growth plates from E14.5 to P10 stages. Higher magnification of the boxed region (a'–f') was shown to demonstrate the differential expression of ATF4 and the positive cells were arrowed. (Scale Bar = 200 μm) (B) Scheme of Atf4 expressing vector. Atf4 cDNA is inserted after the ATG codon in exon 2 of the Col10a1-Bac followed by an IRES-Egfp cassette. (C) Radiographic analysis revealed the dwarfism and skeletal abnormality of C10-Atf4 mice at 4-month-old stage (n = 3). (D) Body lengths of the WT (n = 9) and C10-Atf4 (n = 7) littermates were monitored from birth to 30-week stage, and a consistent reduction of body length in C10-Atf4 mice was observed. (E) Abnormal proximal tibial growth plates with expanded HZ, delimited by dotted lines, were observed in C10-Atf4 mice by Alcian Blue staining. (Scale bar = 200 μm). (F) Ectopic expression of Atf4 in HCs was insufficient for ER stress response induction, indicated by in-situ hybridization of ER stress markers (Hspa5, Atf3 and Ddit3). (Scale bar = 200 μm). https://doi.org/10.7554/eLife.37673.006 To assess the impact of ATF4 overexpression in HCs in the absence of ER stress, we generated a transgenic mouse model carrying a Col10-Bac-Atf4-IRES-Egfp to as (Figure in which ATF4 expression was by the highly promoter of Col10a1 et al., 2011; et al., 2014). We expression of the C10-Atf4 in the growth plates from fetal to (Figure supplement and to 13del C10-Atf4 transgenic mice were than wild-type littermates (Figure and analyses revealed growth plate abnormality in both the and These defects are illustrated by the than expansion of the HZ of the and of C10-Atf4 mice (Figure Figure supplement and the HZ expansion is more severe in C10-Atf4 mice of than that in 13del of which is to the expression level of Although expression of ATF4 in was to survival (Han et al., 2013), cell was not in C10-Atf4 mice (Figure supplement overexpression of ATF4 in HCs in the absence of ER stress did not induce transcription of the UPR-associated genes Hspa5 and (Figure Atf3 was upregulated (Han et al., 2013). Therefore, activation of ATF4 in the absence of the ER stress response, is sufficient to alter HC differentiation, endochondral and cause skeletal to those observed in 13del mice. ATF4 chondrocyte hypertrophy by directly activating Sox9 In C10-Atf4 HCs, ATF4 activation down-regulated the expression of Col10a1 and to expression of prehypertrophic chondrocyte genes Sox9, and Ihh in the HZ (Figure 3A). However, HCs were not in the C10-Atf4 HZ after 2 (Figure supplement suggesting C10-Atf4 HCs did not to have through the The differentiation in growth plate chondrocytes is regulated by transcription factors that control the expression of cell genes and growth factors (Hojo et al., 2016; et al., 2011; et al., 2002; et al., 2005; et al., 2012; et al., 2002; et al., 2017). We the published ER ATF4 data (Han et al., for binding peaks in transcription factor including members of and We found ATF4 binding peaks in the regulatory of Sox9, and the genes were in 13del middle and lower zones, but not or (Figure supplement and suggesting that the family be the of overexpression of Sox9 in HCs in an expansion of HZ and differentiation of HCs, to the phenotypes observed in C10-Atf4 mice et al., the expression pattern of SOX9 that of ATF4 in 13del and C10-Atf4 the of a direct between these factors as of the molecular mechanism the MCDS Figure 3 with 1 supplement see all Download asset Open asset ATF4 de-differentiation of HCs via direct regulation of Sox9. (A) Ectopic expression of Atf4 in HCs leads to accumulation of chondrocytes in C10-Atf4 HZ, indicated by expression patterns of markers Col10a1 Sox9 and Ihh (Scale bar = 200 μm). (B) of by Sox9 promoter with different lengths and or ATF4 binding mutants and to different of ATF4 were in cells and results were as induction compared with cells from three were shown as and was by (C) showed the direct binding of ATF4 to the on the Sox9 promoter in using the from WT and C10-Atf4 were and one was shown. ATF4 ChIP-seq this region has been in SOX9 is highly expressed in chondrocytes, critical matrix genes and chondrocyte differentiation and into hypertrophy et al., 2011; et al., 2002; et al., et al., et al., 2012). It is required for expression of and which with SOX9 to et al., and 2015). We and in the in the Sox9 promoter By in chondrocyte we found ATF4 by the Sox9 promoter (Figure of and or ATF4 activation of the Sox9 (Figure using from wild-type and C10-Atf4 demonstrated that ATF4 directly to the region on the Sox9 promoter in vivo (Figure We the contribution of ATF4 activation of Sox9 in reverting HC differentiation by Sox9 in C10-Atf4 using et al., (Figure supplement In the absence of Sox9, the expansion of HZ in C10-Atf4 mice was and were cells expressing in the HZ (Figure inactivation of Sox9 in 13del mice expression of and in and the HZ expansion was (Figure Figure supplement 1B and C). of Sox9 in wild-type HCs did not affect chondrocyte hypertrophy (Figure supplement these data suggest ER stress-induced overexpression of ATF4 reverts differentiation in 13del HC by direct activation of Sox9 in HCs, thereby chondrocyte hypertrophy. Figure 4 with 1 supplement see all Download asset Open asset rescue of growth plate in C10-Atf4 and 13del mice via inactivation of Sox9. (A) of Sox9 in C10-Atf4 HCs growth plate of C10-Atf4 mice at p10 stage, shown by expression analyses of SOX9 and Col10a1 Higher magnification of the boxed region was shown to demonstrate the differential expression of SOX9 and the positive cells were indicated by (Scale bar = 200 μm). (B) of the HZ lengths of C10-Atf4 and littermates (n = 5). The length of the of the HZ from different of mouse was (C) of positive cells in HZ of C10-Atf4 and littermates (n = 5). the of positive cells was and average on (D) of Sox9 in 13del HCs growth plate of 13del mice at p10 stage (n = shown by expression analyses of SOX9 and Col10a1 Higher magnification of the boxed region was shown to demonstrate the differential expression of SOX9 and the positive cells were arrowed. (Scale Bar = 200 μm). (E) of the HZ lengths of 13del and littermates (n = 5). The length of the of the HZ from different of mouse was (F) of positive cells in HZ of 13del and littermates (n = 5). the of positive cells was and average on were shown as and was by CHOP plays an adaptive and pro-survival role in 13del HC CHOP is another prominent transcription factor that was in 13del HCs, revealed by It is expressed in the PERK signaling pathway, downstream of p-eIF2α and ATF4, which protein synthesis via the PPP1R15A negative feedback loop and protein synthesis and induces stress via Ero1l (Han et al., 2013; et al., Apart from overexpression of CHOP in the bone in transgenic mice has been to leading to et al., while CHOP null mice show bone formation et al., 2006), indicating its role in osteoblast differentiation. Although CHOP is considered as a has and as an adaptive and pro-survival factor in et al., et al., 2002; et al., et al., 2012). expression of ATF4 and CHOP has been to cell et al., 2013). the contribution of CHOP in the adaptation of 13del We found CHOP encoding gene in 13del mice the skeletal defects and growth plate The mice further tibial (Figure and with more significant HZ expansion (Figure and and increased the of chondrocytes expressing markers and in the HZ (Figure and in to 13del, was increased apoptosis in consistent with a pro-survival role for CHOP (Figure and Figure with 1 supplement see all Download asset Open asset CHOP the consequence of ER Stress in 13del mice. The length is further in mice at the p10 The was between 13del and littermates (n = 5). (C) growth plate were observed in 13del mice with global of at p10 stage (n = shown by expression analyses of Col10a1 SOX9 and Higher magnification of the boxed region was shown to demonstrate the