Why SNP rs227584 is associated with human BMD and fracture risk? A molecular and cellular study in bone cells

Abstract A large number of SNPs significant for osteoporosis (OP) had been identified by genome‐wide association studies. However, the underlying association mechanisms were largely unknown. From the perspective of protein phosphorylation, gene expression regulation, and bone cell activity, this study aims to illustrate association mechanisms for representative SNPs of interest. We utilized public databases and bioinformatics tool to identify OP‐associated SNPs which potentially influence protein phosphorylation (phosSNPs). Associations with hip/spine BMD, as well as fracture risk, in human populations for one significant phosSNP, that is, rs227584 (major/minor allele: C/A, EAS population) located in C17orf53 gene, were suggested in prior meta‐analyses. Specifically, carriers of allele C had significant higher BMD and lower risk of low‐trauma fractures than carriers of A. We pursued to test the molecular and cellular functions of rs227584 in bone through osteoblastic cell culture and multiple assays. We identified five phosSNPs significant for OP (P < 0.01). The osteoblastic cells, which was transfected with wild‐type C17orf53 (allele C at rs227584, P126), demonstrated specific interaction with NEK2 kinase, increased expression levels of osteoblastic genes significantly (OPN, OCN, COL1A1, P < 0.05), and promoted osteoblast growth and ALP activity, in contrast to those transfected with mutant C17orf53 (allele A at rs227584, T126). In the light of the consistent evidences between the present functional study in human bone cells and the prior association studies in human populations, we conclude that the SNP rs227584, via altering protein‐kinase interaction, regulates osteoblastic gene expression, influences osteoblast growth and activity, hence to affect BMD and fracture risk in humans.


| INTRODUCTION
OP is a complex chronic bone disease characterized by reduced bone mass, resulting in microarchitectural deterioration of bone tissue and increased bone fragility. 1 Osteoporotic fracture (OF), the most serious consequence of OP, has high disability and mortality. In China, the prevalence of OP increases from 14.94% before 2008 to 27.96% during 2012-2015, which places an enormous economic burden on the whole society. 2 Bone mineral density (BMD) is a classical diagnosis standard for OP and has high heritability. 3 Over the past decades, plentiful genetic and genomic studies have identified genes that are associated with BMD variation in human populations. However, the underlying molecular and cellular mechanisms for most of the established associations were largely unknown, leaving the molecular pathophysiology of OP/OF unclear yet.
Single-nucleotide polymorphisms (SNPs) make up approximately 90% of human genetic variations. 4 Nonsynonymous SNPs (nsSNPs) account for nearly half of the genetic variations associated with inherited diseases. 5 Notably, almost 70% of nsSNPs in human genome are potential phosphorylation-related SNPs (abbreviated as phosSNPs). 6 Protein phosphorylation is a reversible posttranslational modification, which plays critical roles in various signalling pathways, 7 being involved in transcription factor activation and gene expression, 8,9 and regulates essential cellular processes, for example, metabolism, differentiation, and membrane transportation. [10][11][12] In eukaryotes, different protein kinases (PKs) could specifically recognize and attach phosphate groups to target amino acid residues serine (S), threonine (T), and tyrosine (Y). 13 Previous studies have manifested that abnormal regulation of protein phosphorylation was related to pathogenesis of various diseases, like cancer, diabetes, osteoporosis, and so on. [14][15][16] From the perspective of protein phosphorylation, gene expression, and cellular functions, the present study aims to ascertain the association mechanisms for significant OP-related SNPs and OP.
Through searching public resources for SNPs significant for OP, and predicting the potential impacts of filtered SNPs on protein phosphorylation, we identified significant phosSNPs influencing OP.
Based on the hints from the bioinformatics prediction and previous GWAS studies, we focused our investigation on a representative SNP of interest, that is, phosSNP rs227584, and carried out purposeful downstream assays in vitro to ascertain its molecular and cellular functions.
Evidences, collected from the present molecular and cellular study in bone cells, support that SNP rs227584 changes protein phosphorylation site, regulates osteoblastic gene expression, and influences osteoblast functions. Our findings filled the gap of knowledge and illustrated the mechanism underlying the association between rs227584 and BMD/fracture risk in human populations.
The present study highlights the significance of rs227584 and protein phosphorylation in bone biology, and shed new insights into the molecular pathogenesis of OP in humans.

| Searching database for OP-associated phosSNPs
We searched public resources (dbGap, NHGRI) in Phenotype-Genotype Integrator (available at https://www.ncbi.nlm.nih.gov/gap/phege ni) for SNPs significant for OP, including traits of interest "bone density," "osteoporosis," and "fracture." All the picked SNPs are located in the exon, neargene, and UTR regions. From the list of OP-associated SNPs (P < 0.01), phosSNPs were extracted for follow-up molecular and cellular functional studies. 6 2.2 | Predicting and validating impacts of SNP rs227584 on protein functions GPS2.0 software was utilized to predict the potential impacts of phosSNPs on protein phosphorylation. 17  incubated at 37°C in a 5% CO 2 atmosphere. Penicillin (100U/mL) and streptomycin (100U/mL) were added in the medium.

| Quantitative real-time PCR
Osteopontin (OPN) and alkaline phosphatase (ALP) had been recognized to be involved in the osteoblastic differentiation process and are commonly used as "osteoblastic indexes." In bone tissue, collagen 1 (COL1A1 and COL1A2) constitutes the fundamental structural framework for bone formation and mineral deposition, and the collagen fibre gives tensile strength to bone. Osteocalcin (OCN), also known as bone gamma-carboxyglutamic acid-containing protein, is a noncollagenous matrix protein in bone, which is involved in the bone matrix maturation and ordered deposition of hydroxyapatite.
To evaluate the molecular effect of rs227584 on osteoblastogenesis and bone formation, we compared mRNA expression levels (OPN, COL1A1, and OCN) and enzyme activity (ALP) of MG63 cells that had been stably transfected with wild-type or mutant C17orf53 gene. The experimental procedures are described as followed.
MG63 cells, transfected stably with vector, wild-type, or mutant  Table S1. The quantitative RT-PCR experiments were performed in triplicates for each condition and repeated three times.

| Identification of phosSNPs significant for OP
Based on the findings from published genome-wide association studies and archived in public resources, five phosSNPs significant for OP were identified (Table S2). Most interesting, associations with hip/spine BMD as well as fracture for one of the phosSNPs, that is, SNP rs227584 (major/minor allele: C/A, EAS population) located in C17orf53 gene, were suggested in prior meta-analyses in large human populations of European and East Asian ancestry. Specifically, allele C carriers have significant higher hip and spine BMD and lower low-trauma fracture risk than allele A carriers. 21 3.2 | Predictive impacts of phosSNP rs227584 on protein functions The potential impacts of the phosSNP rs227584 on protein phosphorylation were predicted and summarized in Figure 1. In general, the allele-specific sequences encode different protein isoforms, which could change (create or abolish) the phosphorylation site(s) of the encoded protein product or change the kinase that catalysed the phosphorylation of target site(s).
Taking rs227584 into consideration, the C17orf53 gene harbouring wild-type allele C and mutant allele A encodes two different C17orf53 protein isoforms with amino acid residue P and T at position 126 (abbreviated as C17orf53-P126 and C17orf53-T126), respectively. Bioinformatics prediction using GSP2.0 showed that, compared with mutant C17orf53-T126, the wild-type C17orf53-P126 loses a phosphorylation site at amino acid 126 due to the residue substitution T126P, but makes an upstream residue T120 phosphorylatable. In another words, a new phosphorylation site (T120) was created while a nearby one at residue 126 was abolished. NEK2 kinase was predicted, with the highest score of 6, to specifically catalyse the T120 phosphorylation in the wild-type protein C17orf53-P126. The substrate-kinase relationship between C17orf53 protein and NEK2 kinase and the effect of rs227584 on C17orf53 protein phosphorylation (T120) had not been experimentally validated before.

| Substrate-kinase interaction
Our data indicated that the portion of wild-type protein C17orf53 can be pulled down together with kinase NEK2 by anti-NEK2 antibody during Co-IP process, as visualized on Western blot using anti-C17orf53 antibody (Figure 2A). In contrast, no mutant protein C17orf53 was pulled down by anti-NEK2 antibody or visualized. The data displayed significant difference in interaction with kinase NEK2 between wild-type and mutant C17orf53 protein in MG63 cells, suggesting allele-specific substrate-kinase interaction and impaired interaction between mutant C17orf53 protein and NEK2 kinase.

| Phosphorylation at C17orf53-T120 affected by P126T substitution
Two-sided student's t-test showed that C17orf53 protein phosphorylation level was significantly reduced in MG63 cells transfected ZHOU ET AL.
| 901 with variant C17orf53-A120P126, compared with that transfected with wild-type C17orf53-T120P126 (P = 0.008) or mutant C17orf53-T120T126 (P = 0.013) ( Figure 2B). Nevertheless, no difference in protein phosphorylation level was detected between the wild-type and mutant C17orf53. The data suggested that the SNP rs227584 does not significantly change the total phosphorylation level of protein C17orf53. With the presence of P126, the residue T120 was significantly phosphorylated, though the T120 phosphorylation was quantitatively offset by the loss of phosphorylation at amino acid site 126 due to the amino acid substitution from phosphorylatable T126 to nonphosphorylatable P126. The experimental data were consistent with the predicted effect of rs227584 on C17orf53 protein phosphorylation at sites 120 and 126( Figure 1).
The data imply that C17orf53 total protein phosphorylation is regulated by both amino acid substitutions P126T and T120A. It seems that the sacrifice of phosphorylation site T126 in wild-type

| Effect of rs227584 on cell growth
Dynamic monitoring of cell growth for 24 hours showed that, transfection with wild-type C17orf53-T120P126 obviously promoted MG63 cell growth in contrast to the negative control. In contrast, the cell growth was obviously inhibited by mutant C17orf53-T120T126 ( Figure 3B). MG63 growth in vitro recorded at multiple time-points ( Figure 3C) showed that the growth differences were significant (P < 0.001). The data indicated allele-specific effect of rs227584 on osteoblastic growth.

| Effect of rs227584 on osteoblast activity
ALP staining assay showed that transfection with wild-type C17orf53-T120P126 significantly increased ALP enzyme activity in MG63 cells, as compared with the negative control ( Figure 4B). In contrast, the stimulatory effect on ALP activity was significantly reduced by mutant C17orf53-T120T126, as compared with the wildtype ( Figure 4B). The data indicated allele-specific effect of rs227584 on osteoblastic differentiation and activity.

| eQTL effect of phosSNP rs227584 from phenotype-genotype integrator
There is only one record archived in Phenotype-Genotype Integrator by Dec 6, 2017, supporting the eQTL effect of SNP rs227584 (Table 1). Specifically, SNP rs227584 exerts trans effect on KLF10 gene and cis effect on G6PC3, ASB16-AS1 (C17orf65, 13.5 kb downstream of C17orf53), and SLC4A1 genes. A recent eQTL study on whole blood cells in a large cohort reported its cis effect on the above three genes (Table 1), as well as the ASB16 gene, 20 which is partially overlapping with the ASB16-AS1 gene in DNA sequence.
Besides, it was reported that the mutant allele A at rs227584 correlated with increased C17orf65 expression in monocytes, adipose tissue, whole blood, and lymphoblasts. 21 The above data supported that phosSNP rs227584 plays a significant role in regulating gene expression in multiple other types of human cells, including the osteoclast precursor monocytes.

| DISCUSSION
This study represents our first endeavour to illustrate the molecular and cellular mechanism underlying the association of a key SNP, that is, rs227584, with bone phenotypes (BMD/fracture risk) in humans.
Through in-depth functional studies in vitro, we characterized and validated that the SNP rs227584 changes substrate-kinase interaction between protein C17orf53 and kinase NEK2 and affects which is conserved in mouse, chicken, lizard, and zebrafish. It was expressed in diverse tissues in human body, like brain, colon, fat, kidney, and so on, but most abundantly in bone marrow and testicles (https://www.ncbi.nlm.nih.gov/pubmed). In regard to the C17orf53 protein, there was relatively limited understanding about its molecular and cellular functions and biological effects. Nonetheless, association of rs227584 with BMD variation has been suggested in previous meta-analysis. 21 A large-scale study involving 15,375 individuals of European and Oceanian ancestry revealed that rs227584 was associated with both hip BMD (P < 0.001) and spine BMD (P = 0.058). 22 To be noted, the observed effect on BMD in human populations was consistent with its effects on osteoblastic functions observed in MG63 in the present study. Concretely, carriers of risk allele A has significantly inhibited osteoblast growth and decreased osteoblastic activity and be associated with significantly lower BMD than carriers of the alternative allele C.
The quantitative real-time PCR data showed that C17orf53-T126 (risk allele A at rs227584) transfection significantly decreased the expression of osteoblastic-related marker genes (OPN, OCN, COL1A1) F I G U R E 2 Effects of phosSNP rs227584 on protein phosphorylation (A) substrate-kinase interaction assay for C17orf53 and NEK2 proteins. Two representative Western blots were presented. Samples were Co-IP products of cell lysates from MG63 cells which was transiently transfected with variant, mutant, or wildtype C17orf53 genes, respectively. The Co-IP product was prepared with anti-NEK2 antibody. (B) Allele-specific effects of rs227584 on C17orf53 total protein phosphorylation level. Presented are estimated phosphate levels in the immunoaffinity-purified C17orf53 proteins (mean and SD). The phosphate level under the condition of variant type C17orf53 transfection (A120P126) was scaled as 1.0. *P < 0.05, as compared with the variant type. NS: insignificant and attenuated the growth and differentiation of osteoblast in cell culture, in contrast to C17orf53-P126 (allele C at rs227584). Our findings implied that rs227584 affected osteoblastogenesis via regulating the osteoblastic gene expression. OF is heritable by mechanisms that are partly independent of BMD, 23 but to some extent, BMD and fracture were genetically correlate with each other. [24][25][26] In a study including 8594 cases and 23218 controls, it revealed that rs227584 harbouring allele A was associated with increasing risk of any fracture type. 21 The data highlighted that rs227584 was not only associated with BMD variation but also correlated with fracture risk. To the best of our knowledge, our present work for the first time dissected and characterized the molecular functions of SNP rs227584 which significantly affects the growth, differentiation, and activity of bone formation cells.
Our present work indicated that the risk allele A at SNP rs227584 does not significantly change the overall phosphorylation level of C17orf53 protein. Instead, it could change C17orf53 protein interaction with kinase NEK2 and change the target site of phosphorylation to be T126 instead of T120. Specifically, the interaction with NEK2 was present with wild-type C17orf53-T120P126 but absent with mutant C17orf53-T120T126. In the presence of P126, the overall C17orf53 protein phosphorylation level was significantly changed by the residue substitution T120A, indicating that the amino acid residue T120 is a target site of phosphorylation, which is catalysed by kinase NEK2. The observation of no significant difference in overall phosphorylation level between wild-type and mutant C17orf53 protein isoforms coincides with the bioinformatics prediction, supporting that the creation of a new phosphorylation site at residue 120 accompanies with the removal of a nearby one at residue 126.
In this study, we found that wild-type C17orf53-T120P126 significantly promoted MG63 cell growth compared to the mutant type.
Furthermore, only the wild-type but not the mutant C17orf53 protein could interact with NEK2 kinase. Coincidently, NEK2 kinase is an important protein involved in mitotic process and cell cycle regulation. 27 This protein is located at the centrosome, which accumulates progressively throughout the S phase and reaches maximal levels in late G2 phase. It revealed that NEK2 shows constitutive catalytic activity and phosphorylates proteins involved in centrosome A B C F I G U R E 3 Allele-specific effects of rs227584 on the growth of stably transfected MG63 cells. Cell culture images with fluorescence microscope. Presented are representative images for three types of stably transfected MG63 cells. A, 24-h growth curve. Presented are realtime cell indexes (CI) recorded at every 10-min within 24 h. B, MG63 growth in vitro recorded at multiple time-points. Presented are cell indexes (CI) monitored for 24 h in cell growth assay. The data are described as mean ± SD duplication. 28 Among its related pathways are "Regulation of PLK1 Activity at G2/M Transition" and "APC-Cdc20 mediated degradation of Nek2A" (https://genecards.com/). We infer that the NEK2 kinase may play an important role in mediating the effects of rs227584 on osteoblast growth, and similar molecular pathways might be involved.
Over the past decades, extensive genetic and genomic studies have conducted to discover genes/SNPs associated with BMD variation in human populations. [29][30][31][32][33][34] However, our understanding about the underlying association mechanism is very limited. Besides, a majority of phosSNPs in the GPS 2.0 database were predicted, and their potential effects on protein phosphorylation have not been validated by experiments. 35 From the perspective of protein phosphorylation, gene expression, and cellular functions, the present study represents our pursuant efforts to ascertain the functions of a key phosSNP, rs227584 in bone cell, so as to fill the gap of knowledge about its link with osteoporosis risk in human populations of multiple ancestries.
In summary, this study was focused on phosSNPs significant for human BMD and fracture risk and characterized the functions of a representative phosSNP of interest through in vitro molecular and cellular studies. Based on the collected evidences, we proposed that