Objective. Osteoarthritis (OA) is polygenic, with more than 90 risk loci currently mapped, including at the singlenucleotide polymorphism rs6516886. Previous analysis of OA cartilage DNA identified 6 CpG dinucleotides with methylation levels that correlated with the rs6516886 genotype, forming methylation quantitative trait loci (mQTLs). We undertook this study to investigate these mQTLs and to map expression quantitative trait loci (eQTLs) across joint tissues in order to identify a particular gene as a target of the rs6516886 association effect. Methods. Nucleic acids were extracted from the cartilage, fat pad, synovium, and peripheral blood from OA patients. Methylation of CpGs and allelic expression imbalance of potential target genes were assessed by pyrosequencing. A chondrocyte cell line expressing deactivated Cas9 (dCas9)-TET1 was used to directly alter CpG methylation levels, with effects on gene expression quantified by polymerase chain reaction. Results. Multiple mQTLs were detected, with effects strongest in joint tissues and methylation at CpG cg20220242 correlating most significantly with the rs6516886 genotype. CpG cg20220242 is located upstream of RWDD2B. Significant rs6516886 eQTLs were observed for this gene, with the OA risk-conferring allele of rs6516886 correlating with reduced expression. CpG methylation also correlated with allelic expression of RWDD2B, forming methylation-expression QTLs (meQTLs). Deactivated Cas9-TET1 reduction in the methylation of cg20220242 increased expression of RWDD2B. Conclusion. The rs6516886 association signal is a multi-tissue meQTL involving cg20220242 and acting on RWDD2B. Modulating CpG methylation reverses the impact of the risk allele. RWDD2B codes for a protein about which little is currently known. Further analysis of RWDD2B as a target of OA genetic risk will provide novel insight into this complex disease.

Summary Objective In cartilage, the osteoarthritis (OA) associated single nucleotide polymorphism (SNP) rs11780978 correlates with differential expression of PLEC , and with differential methylation of PLEC CpG dinucleotides, forming eQTLs and mQTLs respectively. This implies that methylation links chondrocyte genotype and phenotype, thus driving the functional effect of this genetic risk signal. PLEC encodes plectin, a cytoskeletal protein that enables tissues to respond to mechanical forces. We sought to assess whether these PLEC functional effects were cartilage specific. Method Cartilage, fat pad, synovium and peripheral blood were collected from patients undergoing arthroplasty. PLEC CpGs were analysed for mQTLs and allelic expression imbalance (AEI) was performed to test for eQTLs. Plectin was knocked down in a mesenchymal stem cell (MSC) line using CRISPR/Cas9 and cells phenotyped by RNA-sequencing. Results mQTLs were discovered in fat pad, synovium and blood. Their effects were however stronger in the joint tissues and of comparable effect between these tissues. We observed AEI in synovium in the same direction as for cartilage and correlations between methylation and PLEC expression. Knocking-down plectin impacted on pathways reported to have a role in OA, including Wnt signalling, glycosaminoglycan biosynthesis and immune regulation. Conclusions Synovium is also a target of the rs11780978 OA association functionally operating on PLEC . In fat pad, mQTLs were identified but these did not correlate with PLEC expression, suggesting the functional effect is not joint-wide. Our study highlights interplay between genetic risk, DNA methylation and gene expression in OA, and reveals clear differences between tissues from the same diseased joint.

Osteoarthritis (OA) is a polygenic disease of older people resulting in the breakdown of cartilage within articular joints. Although a leading cause of disability, there are no disease-modifying therapies. Evidence is emerging to support the origins of OA in skeletogenesis. Whilst methylation QTLs (mQTLs) co-localizing with OA GWAS signals have been identified in aged human cartilage and used to identify effector genes and variants, such analyses have never been conducted during human development. Here, for the first time, we have investigated the developmental origins of OA genetic risk at seven well-characterized OA risk loci, comprising 39 OA-mQTL CpGs, in human foetal limb (FL) and cartilage (FC) tissues using a range of molecular genetic techniques. We compared our results to aged cartilage samples (AC) and identified significant OA-mQTLs at 14 CpGs and 29 CpGs in FL and FC tissues, respectively. Differential methylation was observed at 26 sites between foetal and aged cartilage, with the majority becoming actively hypermethylated in old age. Notably, 6/9 OA effector genes showed allelic expression imbalances during foetal development. Finally, we conducted ATAC-sequencing in cartilage from the developing and aged hip and knee to identify accessible chromatin regions, and found enrichment for transcription factor binding motifs including SOX9 and FOS/JUN. For the first time, we have demonstrated the activity of OA-mQTLs and expression imbalance of OA effector genes during skeletogenesis. We show striking differences in the spatiotemporal function of these loci, contributing to our understanding of OA aetiology, with implications for the timing and strategy of pharmacological interventions.

No abstract

Running headline: Epigenetic analysis of the PLEC OA risk locusObjective 1 Osteoarthritis (OA) associated single nucleotide polymorphism (SNP) rs11780978 correlates 2 with differential expression of PLEC, and methylation quantitative trait loci (mQTLs) at PLEC 3 CpGs in cartilage. This implies that methylation links chondrocyte genotype and phenotype, 4 thus driving the functional effect. PLEC encodes plectin, a cytoskeletal protein that enables 5 tissues to respond to mechanical forces. We sought to assess whether PLEC functional effects 6 were cartilage specific. 7 8 Method 9Cartilage, fat pad, synovium and peripheral blood were collected from patients undergoing 10 arthroplasty. PLEC CpGs were analysed for mQTLs and allelic expression imbalance (AEI) 11 was performed. We focussed on previously reported mQTL clusters neighbouring cg19405177 12 and cg14598846. Plectin was knocked down in a mesenchymal stem cell (MSC) line using 13 CRISPR/Cas9 and cells phenotyped by RNA-sequencing. 14 15Results 16 Novel mQTLs were discovered in fat pad, synovium and peripheral blood at both clusters. The 17 genotype-methylation effect of rs11780978 was stronger in cg14598846 than in cg19405177 18 and stronger in joint tissues than in peripheral blood. We observed AEI in synovium in the 19 same direction as for cartilage. Knocking-down plectin impacted on pathways reported to have 20 a role in OA, including Wnt signalling, glycosaminoglycan biosynthesis and immune 21 regulation. 22 23 Conclusions 24Synovium is also a target of the rs11780978 OA association functionally operating on PLEC. 25In fat pad, mQTLs were identified but these did not correlate with PLEC expression, suggesting 26 the functional effect is not joint-wide. Our study highlights interplay between genetic risk, 27 DNA methylation and gene expression in OA, and reveals clear differences between tissues 28 from the same diseased joint. 29 30

Osteoarthritis (OA) is a polygenic disease of older people resulting in the breakdown of cartilage within articular joints. Although a leading cause of disability, there are no disease-modifying therapies. Evidence is emerging to support the origins of OA in skeletogenesis. Whilst methylation QTLs (mQTLs) co-localizing with OA GWAS signals have been identified in aged human cartilage and used to identify effector genes and variants, such analyses have never been conducted during human development. Here, for the first time, we have investigated the developmental origins of OA genetic risk at seven well-characterized OA risk loci, comprising 39 OA-mQTL CpGs, in human fetal limb (FL) and cartilage (FC) tissues using a range of molecular genetic techniques. We compared our results to aged cartilage samples (AC) and identified significant OA-mQTLs at 14 CpGs and 29 CpGs in FL and FC tissues, respectively. Differential methylation was observed at 26 sites between fetal and aged cartilage, with the majority becoming actively hypermethylated in old age. Notably, 6/9 OA effector genes showed allelic expression imbalances during fetal development. Finally, we conducted ATAC-sequencing in cartilage from the developing and aged hip and knee to identify accessible chromatin regions, and found enrichment for transcription factor-binding motifs including SOX9 and FOS/JUN. For the first time, we have demonstrated the activity of OA-mQTLs and expression imbalance of OA effector genes during skeletogenesis. We show striking differences in the spatiotemporal function of these loci, contributing to our understanding of OA etiology, with implications for the timing and strategy of pharmacological interventions.

The importance of temperature data in minimum postmortem interval (minPMI) estimations in criminal investigations is well known. To maximise the accuracy of minPMI estimations, it is imperative to investigate the different components involved in temperature modelling, such as the duration of temperature data logger placement at the crime scene and choice of nearest weather station to compare the crime scene data to. Currently, there is no standardised practice on how long to leave the temperature data logger at the crime scene and the effects of varying logger duration are little known. The choice of the nearest weather station is usually made based on availability and accessibility of data from weather stations in the crime scene vicinity. However, there are no guidelines on what to look for to maximise the comparability of weather station and crime scene temperatures. Linear regression analysis of scene data with data from weather stations with varying time intervals, distances, altitudes and microclimates showed the greatest goodness of fit (R 2), i.e. the highest compatibility between datasets, after 4-10 days. However, there was no significant improvement in estimation of crime scene temperatures beyond a 5-day regression period. The smaller the distance between scene and weather station and the higher the similarity in environment, such as altitude and geographical area, resulted in greater compatibility between datasets. Overall, the study demonstrated the complexity of choosing the most comparable weather station to the crime scene, especially because of a high variation in seasonal temperature and numerous influencing factors such as geographical location, urban 'heat island effect' and microclimates. Despite subtle differences, for both urban and rural areas an optimal data fit was generally reached after about five consecutive days within a radius of up to 30km of the 'crime scene'. With increasing distance and differing altitudes, a lower overall data fit was observed, and a diminishing increase in R 2 values was reached after 4-10 consecutive days. These results demonstrate the need for caution regarding distances and climate differences when using weather station data for retrospective regression analyses for estimating temperatures at crime scenes. However, the estimates of scene temperatures from regression analysis were better than simply using the temperatures from the nearest weather station. This study provides recommendations for data logging duration of operation, and a baseline for further research into producing standard guidelines for increasing the accuracy of minPMI estimations and, ultimately, greater robustness of forensic entomology evidence in court.

Notably, at cg18131582, residing at the COLGALT2 locus, where we observed stronger mQTLs in the developmental samples, ATAC-seq peaks, corresponding with open chromatin and indicative of active enhancers, were observed in fetal cartilage samples, but not in comparable adult samples. Conversely, at 3 RDWW2B CpGs, ATAC-seq narrow peaks were observed in all samples, irrespective of joint site or sample type. However, at the locus harbouring RUNX2, chromatin was inaccessible to the transposase in both fetal and OA cartilage, indicating that the functional effect is potentially active in a distinct tissue of the articulating joint, not in cartilage. Conclusions: This is the first study to report the activity in developing human limbs of OA-associated mQTLs that were initially discovered in the cartilage of elderly patients. It suggests that at a molecular genetics level, the increased OA susceptibility encoded by risk-conferring alleles functionally operates during development, and that the susceptibility is laid down when limbs and joints are forming. OA may be an age-related disease, yet the genetic risk is functionally active from the beginning of life. In summary, our data reveals that OA mQTLs are active in fetal chondrocytes but, importantly, that there are striking differences. These distinctions point toward regulatory variability in gene expression mediated by the same association signal over time, and a developmental switch in sensitivity to epigenetic regulation of the genes and subsequent pathways underlying OA pathophysiology.