Deficiency for the ER-stress transducer OASIS causes severe recessive osteogenesis imperfecta in humans

Osteogenesis imperfecta (OI) is a clinically and genetically heterogeneous brittle bone disorder. Whereas dominant OI is mostly due to heterozygous mutations in either COL1A1 or COL1A2, encoding type I procollagen, recessive OI is caused by biallelic mutations in genes encoding proteins involved in type I procollagen processing or chaperoning. Hitherto, some OI cases remain molecularly unexplained. We detected a homozygous genomic deletion of CREB3L1 in a family with severe OI. CREB3L1 encodes OASIS, an endoplasmic reticulum-stress transducer that regulates type I procollagen expression during murine bone formation. This is the first report linking CREB3L1 to human recessive OI, thereby expanding the OI gene spectrum.


Findings
We describe a Turkish family ( Figure 1A) with three sibs, two of whom were affected by severe OI (written informed consent of the family was obtained and the study was approved by the Ethics Committee of the Ghent University Hospital (Ghent, Belgium)). Consanguinity was not reported, but the parents originated from neighbouring villages. The first affected child (III:3) developed several fractures in utero and was small for gestational age. His birth length was 40 cm (<P3). At the first day of life he was hospitalized for hyperbilirubinemia and O-bain-like deformities, soft calvarial bones and widely open fontanelles were noticed. He developed several fractures after birth and multiple fractures healed with extremity deformities. He also had a right inguinal hernia. X-rays showed beaded ribs, callus formation and multiple fractured tubular bones with an accordion-like broadened appearance. He was hospitalized several times due to recurrent constipation and pulmonary infections (bronchopneumonia). During this period, he developed abdominal distention and hepatomegaly, the latter due to cardiac insufficiency. No signs of T-cell dysfunction or other immune deficiencies have been noted. He died at 9 months of age. The second affected sib (III:4, Figure 1A) was a male foetus from a pregnancy that was medically terminated at 19 weeks of gestation. Post-mortem examination showed thin ribs and fractures at bowed humerus and femora ( Figure 1B-C).
The parents have a healthy daughter (III:1) and have had one miscarriage (III:2, cause unknown). The adolescent daughter has blue sclerae but had not experienced any fractures. The mother (II:5) at 38 years of age and the father (II:6) at 47 years have blue sclerae, a soft and velvety skin and normal teeth. While the mother has small joint hypermobility, the father has conductive hearing loss.
We selected the CREB3L1 gene [GenBank:NM_052854.2], encoding the ER-stress transducer OASIS (Old Astrocyte Specifically Induced Substance), as an excellent candidate gene based on the observation that OASIS−/− mice were born with severe osteopenia and spontaneous fractures   [29], reminiscent of severe human OI. In those mice, OASIS was shown to be crucial for bone formation through activating col1a1 transcription and facilitating the secretion of matrix proteins. Treatment of murine osteoblasts with BMP-2 (bone morphogenic protein 2) causes mild ER-stress and is associated with accelerated RIP (regulated intramembrane proteolysis) of OASIS. The N-terminal part of OASIS is subsequently translocated to the nucleus, where it binds to the osteoblast-specific UPRE (unfolded protein response element) regulatory region in the murine Col1a1 promoter thereby causing high levels of type I procollagen expression [29]. While the amount of type I procollagen is normal in the murine OASIS−/− skin, reduced amounts of type I procollagen were detected in OASIS−/− calvaria and tibia, which suggested tissue-specific decrease of type I procollagen in the bone matrix but also failure of the OASIS−/− osteoblasts to produce high levels of type I procollagen [29]. OASIS further functions as a tissue-specific ER-stress transducer that alters transcription of target genes involved in developmental processes, differentiation, or maturation upon mild ER-stress. PCR amplification of all exons and flanking introns of CREB3L1 failed in foetus III:4, suggesting a homozygous whole gene deletion. ArrayCGH analysis (1M SurePrint G3 Human CGH Microarray, Agilent Technologies) and copy number profiling (arrayCGHbase) confirmed this genomic deletion, which encompasses CREB3L1 and the first exon of DGKZ (arr11p11.2(46268141-46359490)×0, Figure 2B) [30,31]. Whereas the arr11p11.2(46268141-46359490)×0 homozygous deletion was not reported before, heterozygous deletions or gains of this genomic region are described in the Decipher database [32] and the Database of Genomic Variants [33] but encompassing large genomic regions comprising multiple genes (6 to 86 genes and/ or multiple chromosomal abnormalities) which, in some cases, are associated with intellectual disability. Both parents and the healthy sister were heterozygous for the deletion (data not shown). DGKZ encodes diacylglycerol kinase zeta, an ubiquitously expressed enzyme that is most abundantly present in the brain, thymus and skeletal muscle [34] and which has a regulatory role in T-cell receptor signalling and T-cell activation [35]. Two different isoforms (DGKζ1 in immune cells and DGKζ2 in other cells) are known, in which exon 1 is either present or absent and which have a tissue-and developmental stage-specific expression [28]. Hitherto, no known function in bone formation has been ascribed to DGKζ and thus a possible contributing role to (the severity of ) the bone phenotype of patient III:3 and foetus III:4 cannot completely be excluded. Expression analysis by real time-quantitative PCR (RT-qPCR) on total RNA isolated from dermal fibroblasts of foetus III:4 confirmed complete absence of the CREB3L1 transcript. In order to investigate the expression of the two DGKZ isoforms (DGKζ1 and DGKζ2), two different primer pairs were designed, of which one was specific for exon 1 that is only present in the DGKζ1 isoform. RT-qPCR experiments revealed no amplification for the primer pair specific for exon 1 in cultured dermal fibroblasts, suggesting that the DGKζ1 isoform is not expressed in these cells. For the second primer pair normal DGKZ expression was observed, which implies normal expression of the DGKζ2 isoform in cultured human dermal fibroblasts ( Figure 2C). RT-qPCR analysis of the ER-stress markers BiP, CHOP and the spliced form of XBP1 showed levels comparable to controls, even after stimulation of confluent fibroblasts for 4 hours with the ER-stress inducers Tunicamycin (Tu, 10 μg/ml, Sigma-Aldrich) and Thapsigargin (Th, 1 μM, Sigma-Aldrich) ( Figure 2C). This is in accordance to the observations in OASIS−/− mice. The expression level of CREB3L1 was unchanged in control fibroblasts after treatment with Tu and Th ( Figure 2C), suggesting that OASIS does not play a major role in the ER-stress pathways previously linked to disease pathogenesis [1]. Additionally, our finding that type I (pro)collagen production is normal in human dermal fibroblasts (Figure 2A) confirms that OASIS has a tissue-specific effect on type I (pro)collagen production [29].
In conclusion, the identification of CREB3L1 (encoding the ER-stress transducer OASIS) as a novel gene for autosomal recessive OI expands the spectrum of genes linked to OI and reinforces the role of ER-stress in the pathophysiology of OI.
(See figure on previous page.) Figure 2 Biochemical and molecular results. A. Biochemical collagen analysis was performed on collagens produced by the patients dermal fibroblasts, which were grown for 16 hrs in the presence of 14 C-Proline. Radioactively labelled intracellular and secreted fibrillar collagen proteins were isolated and mature collagens were obtained by pepsin digestion. Foetal secreted (left panel) as well as intracellular (right panel) mature type I collagen revealed a normal electrophoretic pattern when compared to a control (C) sample. Also for the unprocessed, secreted type I procollagen a normal electrophoretic migration pattern was observed (data not shown). B. ArrayCGH analysis on a 1M SurePrint G3 Human CGH Microarray revealed a homozygous deletion of the entire CREB3L1 gene in the affected foetus III:4. C. Expression level analysis by RT-qPCR was performed in duplicate on total RNA extracted from three biological replicates of the fibroblast cell lines from foetus III:4 and three controls (C1, C2 and C3) (LightCycler480 and RealTime ready DNA Probe Master Mix, Roche). The expression level of each investigated gene was quantified using qbase PLUS (Biogazelle) [27]. HPRT1, RLP13a and YWHAZ were applied as reference targets. RT-qPCR for foetus III:4 confirmed the total absence of CREB3L1 expression when compared to control samples (C1, C2 and C3). DGKZ has two alternative (tissue-specific) isoforms [28].