Phenotypic Variation in Vietnamese Osteogenesis Imperfecta Patients Sharing a Recessive P3H1 Pathogenic Variant

Osteogenesis imperfecta (OI) is a syndromic disorder of bone fragility with high variation in its clinical presentation. Equally variable is molecular aetiology; recessive forms are caused by approximately 20 different genes, many of which are directly implicated in collagen type I biosynthesis. Biallelic variants in prolyl 3-hydroxylase 1 (P3H1) are known to cause severe OI by affecting the competence of the prolyl 3-hydroxylation—cartilage associated protein—peptidyl-prolyl cis-trans isomerase B (P3H1-CRTAP-CyPB) complex, which acts on the Pro986 residue of collagen type I α 1 (COL1A1) and Pro707 collagen type I α 2 (COL1A2) chains. The investigation of an OI cohort of 146 patients in Vietnam identified 14 families with P3H1 variants. The c.1170+5G>C variant was found to be very prevalent (12/14) and accounted for 10.3% of the Vietnamese OI cohort. New P3H1 variants were also identified in this population. Interestingly, the c.1170+5G>C variants were found in families with the severe clinical Sillence types 2 and 3 but also the milder types 1 and 4. This is the first time that OI type 1 is reported in patients with P3H1 variants expanding the clinical spectrum. Patients with a homozygous c.1170+5G>C variant shared severe progressively deforming OI type 3: bowed long bones, deformities of ribcage, long phalanges and hands, bluish sclera, brachycephaly, and early intrauterine fractures. Although it remains unclear if the c.1170+5G>C variant constitutes a founder mutation in the Vietnamese population, its prevalence makes it valuable for the molecular diagnosis of OI in patients of the Kinh ethnicity. Our study provides insight into the clinical and genetic variation of P3H1-related OI in the Vietnamese population.


Introduction
Osteogenesis imperfecta (OI) is a connective tissue disorder with bone fragility as the predominant feature. The disease shows a continuum of clinical severity, the heterogeneity

General Description of the Cohort
A total number of 17 patients from 14 Vietnamese families harboured pathogenic variants in the P3H1 gene, which comprises 11.6% of the whole Vietnamese OI cohort of the UT biobank (n = 146) [18]. The cohort included 13 pediatric and four adult patients.
Ages ranged from four months to 35 years, with a mean age of 13.7 ± 10.7 years, locktime 2016 (Table 1). Eleven subjects (64.7%) were female, and six (35.3%) were male. Patients originated from different provinces of Vietnam, but all families belonged to the Kinh ethnic group (Figure 2). None of the families had history of consanguinity or history of OI in previous generations. In seven families, parents reported previous miscarriages. However, only in family VN36 was it clinically confirmed that the two miscarriages were prenatally fatal ( Figure 3).    Patients in three families (VN11, VN20, and VN36-1, VN36-2, VN36-3) received bisphosphonate treatment prior to the phenotype descriptions in this study.

Genetic Analysis
The index patient of the 55 OI families was analysed by a targeted NGS panel for the common OI-related genes and revealed 17 patients in 14 families with a P3H1 (NM_022356.3) pathogenic variant ( Table 2). The c.1170+5G>C p.? pathogenic variant was identified in 15 patients of 12 families; in 13 patients, it was found in homozygous state and in two patients, VN10 and VN42, it was compound heterozygous with the c.2197A>G p.(Lys733Glu) and c.1934G>A p.(Cys645Tyr) variants, respectively. Furthermore, patient VN101 was found to have the homozygous c.1224-79G>A variant, whereas patient VN75 had this variant in compound heterozygous form with c.257del p.(Glu86Glyfs*22). All identified variants apart from c.1170+5G>C were absent from the OI variant database. All pathogenic variants were confirmed by Sanger sequencing to be present in the patient parents whose gDNA sample was available ( Table 2). The only exception was one of the parents in families VN10 (Table 2). Paternity was confirmed in family VN10, which raises the possibility that the c.2197A>G variant is de novo. However, we cannot exclude the possibility of gonadal mosaicism in the father or monoallelic state of the P3H1 variants. Patients in three families (VN11, VN20, and VN36-1, VN36-2, VN36-3) received bisphosphonate treatment prior to the phenotype descriptions in this study.

Genetic Analysis
The index patient of the 55 OI families was analysed by a targeted NGS panel for the common OI-related genes and revealed 17 patients in 14 families with a P3H1 (NM_022356.3) pathogenic variant ( Table 2). The c.1170+5G>C p.? pathogenic variant was identified in 15 patients of 12 families; in 13 patients, it was found in homozygous state and in two patients, VN10 and VN42, it was compound heterozygous with the c.2197A>G p.(Lys733Glu) and c.1934G>A p.(Cys645Tyr) variants, respectively. Furthermore, patient VN101 was found to have the homozygous c.1224-79G>A variant, whereas patient VN75 had this variant in compound heterozygous form with c.257del p.(Glu86Glyfs*22). All identified variants apart from c.1170+5G>C were absent from the OI variant database. All pathogenic variants were confirmed by Sanger sequencing to be present in the patient parents whose gDNA sample was available ( Table 2). The only exception was one of the parents in families VN10 (Table 2). Paternity was confirmed in family VN10, which raises the possibility that the c.2197A>G variant is de novo. However, we cannot exclude the possibility of gonadal mosaicism in the father or monoallelic state of the P3H1 variants.
Among the unaffected siblings, we found three heterozygous carriers of the c.1170+5G>C (families VN11, VN37, VN107) and one carrier of the c.1124-79G>A (family VN101) variant. Other unaffected siblings tested negative for the pathogenic variant found in the index patient of the family (Figure 3). In order to explore the genetic variation in the P3H1 gene and possible common haplotype between carriers of the same c.1170+5G>C variant, we examined the presence of the c.465+489delG and c.465+499delG SNPs in the 14 index patients. All index patients contained the two SNPs in the homozygous state, with the exception of patient VN101, in whom they were found in the heterozygous state.
The mean total number of fractures per patient was 17.9 ± 11.7 (0-40), whereas the median number of fractures per year was 1.4 (0-8). Patient VN10, who was 13 y.o., did not have any fractures; however, she was suspected of having OI, based on light-blue eye sclera, DI, and brother with clinically diagnosed OI. She and her affected brother (not included in the analysis due to unavailability of gDNA sample) were the only cohort patients able to ambulate without assistance. The mean number of fractures per year was 3.3 times higher in pediatric patients (n = 2.2) compared to adults (n = 0.7). Only ten patients (58.8%) experienced intrauterine fractures.
Newborn patients VN33 and VN100 (up to 1 y.o.) were male and shared bowed legs ("frog-leg posture") and hydrocele (Figure 4). Patients showed signs of rhizomelia, with shorter and more severely affected and deformed lower limbs. Limb bowing was noticed directly after birth. Individuals had long hands in comparison to forearms and disproportionally long phalanges. Thirteen patients (76.5%) developed chest deformities, and four patients (23.5%) had extremely severe pectus carinatum (VN36-3, VN75, VN101, VN114). Severe kyphoscoliosis was also typical among this cohort (Figures 4 and 5).
The mean total number of fractures per patient was 17.9 ± 11.7 (0-40), whereas the median number of fractures per year was 1.4 (0-8). Patient VN10, who was 13 y.o., did not have any fractures; however, she was suspected of having OI, based on light-blue eye sclera, DI, and brother with clinically diagnosed OI. She and her affected brother (not included in the analysis due to unavailability of gDNA sample) were the only cohort patients able to ambulate without assistance. The mean number of fractures per year was 3.3 times higher in pediatric patients (n = 2.2) compared to adults (n = 0.7). Only ten patients (58.8%) experienced intrauterine fractures.
Newborn patients VN33 and VN100 (up to 1 y.o.) were male and shared bowed legs ("frog-leg posture") and hydrocele (Figure 4). Patients showed signs of rhizomelia, with shorter and more severely affected and deformed lower limbs. Limb bowing was noticed directly after birth. Individuals had long hands in comparison to forearms and disproportionally long phalanges. Thirteen patients (76.5%) developed chest deformities, and four patients (23.5%) had extremely severe pectus carinatum (VN36-3, VN75, VN101, VN114). Severe kyphoscoliosis was also typical among this cohort (Figures 4 and 5).  The majority of patients (14/17, 82.4%) developed brachycephaly. Out of them, ten (58.9%) had typical OI triangular faces, while others had long rounded faces. Patients also showed a tall forehead with frontal bossing and low set ears. Patient VN33 had a soft unmineralised skull (Figures 4 and 5).

Radiographic Features
Based on available radiographs of patient VN11, both the cortical and trabecular bone mass of the proband are similarly decreased ( Figure 6A,B). Femural shift is observed, and the knee joint is abnormal and deformed. Lower limbs reveal bowing, widening, and shortening of diaphyses. Popcorn-shaped abnormalities in the metaphyses (i.e., bulbous metaphyses) can be noted. Thin ribs reveal decreased bone mass. The majority of patients (14/17, 82.4%) developed brachycephaly. Out of them, ten (58.9%) had typical OI triangular faces, while others had long rounded faces. Patients also showed a tall forehead with frontal bossing and low set ears. Patient VN33 had a soft unmineralised skull (Figures 4 and 5).

Radiographic Features
Based on available radiographs of patient VN11, both the cortical and trabecular bone mass of the proband are similarly decreased ( Figure 6A,B). Femural shift is observed, and the knee joint is abnormal and deformed. Lower limbs reveal bowing, widening, and shortening of diaphyses. Popcorn-shaped abnormalities in the metaphyses (i.e., bulbous metaphyses) can be noted. Thin ribs reveal decreased bone mass. Similarly, lower limbs' diaphyses bowing, shortening, and widening together with bulbous metaphyses were observed in radiographs of patient VN75 ( Figure 6C). Furthermore, here the cortical and trabecular bone mass were decreased. Similarly, lower limbs' diaphyses bowing, shortening, and widening together with bulbous metaphyses were observed in radiographs of patient VN75 ( Figure 6C). Furthermore, here the cortical and trabecular bone mass were decreased.

Discussion
Despite harbouring variants in the same P3H1 gene, the phenotype of patients varied extensively in terms of severity and clinical OI features and ranged from type 1 to type 3, 4, and even 2, reported as two confirmed OI miscarriages in family VN36. This data underlines the broad range of possible phenotype variability in cases of P3H1-associated OI, which lacked previous attention.
Previous studies have confirmed severe presentation of OI as a result of biallelic P3H1 pathogenic variants. Severe adult type 3 and perinatal lethal type 2 OI have been described in numerous cases as a result of loss of function defects arising from mRNA instability due to nonsense, frameshift, or splice site pathogenic variants [6,19,[24][25][26]. A study also described the loss of the Lys-Asp-Glu-Leu (KDEL) endoplasmic reticulum (ER) retrieval sequence as a consequence of a P3H1 pathogenic variant, which also led to a lack of intracellular P3H1 and type 3 OI [27]. It remains unclear what is the underlying reason that triggers the progression of the disease to either of the two severe types in loss of function P3H1 defects. Considering that the expression of P3H1 and CRTAP is dependent on each other, it could be speculated that the level of CRTAP expression remains sufficient in cases of missense P3H1 pathogenic variants, which influences the phenotypic outcome [28].
Moderate type 4 OI has been reported in cases of missense P3H1 variants; numerous cases were documented in Turkish and other OI cohorts [7,16]. Although it is tempting to assume that maintenance of protein expression can have a protective role in diminishing disease severity, it is important to note that missense P3H1 pathogenic variants have also been reported as a cause of type 3 OI [29,30]. However, the fact that missense pathogenic variants have not been found in type 2 OI may indeed support the presentation of a milder phenotype when protein expression is not abolished. Sustained protein expression in patients VN42 and VN10 with missense pathogenic variants may partially account for the presentation of the milder OI types 4 and 1, respectively.
It might be speculated that the OI type 1 phenotype in patient VN10, who harbours a compound heterozygous combination of a loss of function and missense variant, might be associated with a decreased amount of protein with partial functionality. It remains to be seen if the discovery of more missense pathogenic variants will further uncover more about the milder end of the P3H1 clinical spectrum. Additionally, the severity of P3H1-related OI might also be influenced by modifying factors and epigenetics, as it is in the case of other OI forms [4,[31][32][33][34].
Another factor that may be influencing, to a great extent, the severity of P3H1-related OI development is the remaining functional competence of the P3H1-CRTAP-CyPB complex in different P3H1 defects. This has been measured by the 3-hydroxylation of COL1A1 Pro986 of patient fibroblasts on which this complex exerts its function; this is found to be highly decreased and correlates with altered electrophoretic migration of the collagen chains [6,8,19,25] (Table S3). Although the mechanistic details of the relation between the two processes are not known, this results in collagen over-modification because of the excessive exposure of collagen chains to lysyl hydroxylases and prolyl 4-hydroxylases during the delayed folding of the triple helix. As such, the molecular mechanism resembles the over-modification process observed in autosomal dominant forms of OI in which defects such as glycine substitutions in COL1A1 or COL1A2 lead to the same effect, albeit with normal levels of Pro986 3-hydroxylation [8]. Interestingly, P3H1 OI cases are also found to present collagen type 5 over-modification, although this was also detectable in an unaffected member of a family with severe lethal OI [24,35,36]. However, exceptions to this correlation exist; 3-hydroxylated COL1A1 Pro986 was found to be only slightly decreased in patient fibroblasts in which the KDEL ER retention motif was lost; yet, collagen type I over-modification was detected. In this case, the remaining COL1A1 Pro986 3-hydroxylation may be accounted for by some residual activity of the KDEL-lacking P3H1 protein [27]. In all these studies, severe OI types 3 and 2 were reported, which is also commonly found in cases of COL1A1 and COL1A2 glycine substitutions [6,19,25,36]. It is noteworthy that type 3 was also found in the OI patient with moderately decreased Pro986 3-hydroxylation, which points to the contribution of additional elements in OI progression and the lack of direct causative effects by Pro986 3-hydroxylation. Unfortunately, no biochemical analyses of collagen have been performed in cases of missense pathogenic variants or milder phenotypes which would enable us to compare the state of collagen Pro986 3-hydroxylation and over-modification with the more severe cases; this is also a limitation that applies to the characterisation of our OI cohort. Another decisive parameter in the OI phenotype can be differences in the Pro986 3-hydroxylation between fibroblasts and bone cells. A patient with a P3H1 splice pathogenic variant was reported to have 4% of Pro986 3-hydroxylation in collagen produced by fibroblasts, but this was found to be increased to 55% in bone cells [19]. Clearly, more studies are required to explain the significance of this intriguing tissue-specific finding.
Regarding the group of 13 patients with the prevalent homozygous c.1170+5G>C pathogenic variants, two patients presented the moderate type 4 OI instead of the severe type 3. Interestingly, these two patients (VN11, VN36-3) received bisphosphonate treatment, although details are lacking about the treatment type, regime, and responsiveness of the patients. Thus, it is not clear if the treatment can account for the milder phenotype compared to patients VN20, VN36-1, and VN36-2 who also received bisphosphonates, but presented nonetheless OI type 3. A reported case of moderate pediatric P3H1-associated OI was described to show improvement in bone quality after pamidronate therapy [7]. Pamidronate treatment may also account for the extended survival of a pediatric OI patient with the Irish traveller homozygous P3H1 c.232delC pathogenic variant compared to lethal OI noted in other patients with the same pathogenic variant [19]. Similarly, pamidronate treatment also improved bone density in two pediatric OI type 3 patients with the homozygous splice site pathogenic variant c.2055+18G>A and the compound heterozygous c.484delG and c.2155dupC pathogenic variants [25].
We have identified four novel, previously unreported variants in our cohort. The P3H1 1224-79G>A variant is an intronic variant, which, according to results of in silico analysis, activates a cryptic splice donor site at the position c.1224-83. Use of this cryptic site might cause insertion of 92 nucleotides, with frameshift and presence of in-frame stop codons. Therefore, this variant is likely to cause nonsense mediated decay and haploinsufficiency of P3H1 protein leading to OI. The haploinsufficiency is also associated with the c.257del nonsense variant. The c.1934G>A, p.(Cys645Tyr) is a missense variant with a highly deleterious effect, whereas the 2197A>G, p.(Lys733Glu) variant alters the conserved Cterminal KDEL motif of P3H1, necessary for ER retention of the protein, similarly to OI type 3 patients, as described by Takagi et al. [27].
The c.1170+5G>C P3H1 variant was found to dominate our OI cohort. This variant has been previously reported in a homozygous state in three infants of Vietnamese origin [16]. The in silico prediction for the variant shows loss of the splice donor site. Due to the absence of a cryptic splice donor site in the nearest sequence, it can be predicted that the variant causes either intron retention or exon skipping. According to the functional analysis performed by Baldridge et al., the variant results in a single base pair substitution at the splice donor site at the start of intron 6; this subsequently leads to the in-frame skipping of exon 6, resulting in deletion of 90 nucleotides [19] and 360-390 amino acid residues. It is unclear if exon 6 skipping results in no or decreased expression of a non-functional protein that has been reported to commonly associate with severe lethal phenotypes [6,19,[24][25][26]. The level of mRNA has been found to be normal. Collagen analysis of cultured fibroblasts showed over-modification, but the level of Pro986 3-hydroxylation was not reported. Despite the fact that most of the homozygous c.1170+5G>C patients in our cohort did not die prenatally, they did show phenotypic similarities (Figure 4).
Considering the high frequency of the c.1170+5G>C variant in our cohort (10.3%), it is plausible that it may present a founder mutation in the Vietnamese population. We found that all index patients shared the c.465+489delG and c.465+499delG SNPs in P3H1 in the homozygous state with the exception of one patient. Unfortunately, we did not have further information about the haplotype, which could enable the determination of an estimate of when the mutation originated. In addition, the patients in our cohort are equally distributed in provinces in distant geographic locations in Vietnam; but it is not known if they originate from a common location that would potentially point to common ancestry. It is worth noting that Vietnam is generally a country with high population movement; thus, isolated communities and consanguinity are not common. Furthermore, considering the OI cohort of 146 patients, it is unknown whether this cohort is reflective of the entire OI population in Vietnam. Although a 0.016% minor allele frequency (MAF) is known in the East Asian population, there is no data about the carrier frequency in Vietnam, which could allow us to estimate the frequency of this form of OI.

Conclusions
We identified the c.1170+5G>C variant as highly prevalent in our patient cohort of recessive OI cases in which COL1A1 and COL1A2 variants were excluded. This can be very informative for the molecular diagnosis of the disease, especially in a country where NGS is limited in diagnostic services. Although it is unclear what the exact percentage of the OI population in Vietnam is with the c.1170+5G>C variant, we expect that a considerable number of cases could be solved by specifically testing this variant. This could provide a cost-effective method for the molecular diagnosis of recessive OI with the potential to improve clinical counselling and treatment allocation.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/genes13030407/s1, Table S1: List of OI panel genes; Table S2: Primers used for Sanger sequencing analysis of the P3H1 gene in patients' siblings and parents; Table S3: Collagen type I α chain Pro986 residue 3-hydroxylation described previously in patients with P3H1 pathogenic variants.  Informed Consent Statement: An informed written consent from the patients or legal guardians of participants under the age of 16 was obtained before inclusion in the study.

Data Availability Statement:
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The variants identified in the study were submitted to the LOVD Osteogenesis Imperfecta variant database https://databases.lovd.nl/ shared/genes/ (accessed on 23 February 2022).