Ankylosing Spondylitis (AS) is a form of inflammatory spondyloarthritis predominantly affecting the axial skeleton, which is characterised pathologically by enthesitis (Bridgewood et al., 2020). Extra-skeletal manifestations are also common in AS; these include inflammation of the gut (ranging from low-grade sub-clinical inflammation of the terminal ileum to overt inflammatory bowel disease - IBD), skin (psoriasis), and uveal tract (Stolwijk et al., 2015; Rizzo et al., 2017). AS was one of the first complex diseases in which a specific genetic effect was identified when its strong association with the major histocompatibility complex (MHC) immune response gene HLA-B27 was described nearly 50 years ago (Brewerton et al., 1973) (Schlosstein et al., 1973). However, it is clearly polygenic (Brown et al., 1997); even the MHC association is attributable to several alleles at more than one locus (Cortes et al., 2015) and more than 100 non-MHC genetic associations have now been suggested by genome-wide association studies (Burton et al., 2007; Reveille et al., 2010; Evans et al., 2011; Cortes et al., 2013). Shared genetic susceptibility factors undoubtedly contribute to the excess occurrence of psoriasis, IBD and uveitis not only in individuals with AS but also their relatives (Ellinghaus et al., 2016) (Robinson et al., 2016). One of the strongest non-HLA associations with AS is with the RUNX3 (Runt-related transcription factor (TF) 3) locus. RUNX3 is involved in T-cell function and plays a key role in the development of CD8⁺ T-cells (Egawa et al., 2007). It also influences many other cells, including helper T-cells, innate lymphoid, tissue resident, mucosa and gut cells (Ebihara et al., 2015; Behr et al., 2018). We have recently demonstrated that AS-associated non-coding single nucleotide polymorphisms (SNPs) in an enhancer-like region upstream of RUNX3 affect the binding of different factors: in particular the repressive nucleosome remodelling and deacetylase (NuRD) complex binds preferentially to the risk allele, while conversely interferon regulatory factor (IRF) five to the protective allele (Vecellio et al., 2021). However, the functional effects of these changes on gene transcription are still to be precisely determined. Our earlier observations were made in T-cells, but here we describe some of the functional effects of the lead AS-associated SNP in the vicinity of RUNX3 (rs6600247, p = 2.6 × 10⁻¹⁵ (Cortes et al., 2013) that are more obvious in CD14⁺ monocyte-like cells than CD8⁺ T-cells. First, we evaluate the chromatin landscape surrounding rs6600247 using the ENCODE database (https://genome.ucsc.edu/ENCODE/). Second, we demonstrate differential allelic binding of rs6600247 to the c-MYC TF. Finally, we investigate the chromosomal architecture and physical interactions between AS-associated sequences in the enhancer-like region upstream of RUNX3 and its promoter, showing a probable role of chromosome looping in the regulation of RUNX3.

DNA was extracted using the Qiagen AllPrep DNA/RNA Mini Kit (Qiagen Ltd., Manchester, United Kingdom) and genotyped for rs6600247 using TaqMan SNP assay (custom order by Life Technologies, Paisley, United Kingdom), for the cells (obtained by the buffy coat) used in the functional studies.

We used the UCSC genome browser build hg19 and the Roadmap database [https://genome.ucsc.edu/ENCODE/] to investigate the epigenetic landscape of rs6600247 upstream of the RUNX3 promoter, which is strongly associated with AS (p = 4.2 × 10^–15) (Cortes et al., 2013). Histone modifications and GeneHancer (a database of human regulatory elements and their inferred target genes) tracks were selected to evaluate regulatory elements and chromosome looping between promoters and enhancer regions (Fishilevich et al., 2017).

Blood samples were obtained from AS patients with ethical approval (COREC 06/Q1606/139) and informed patient consent. CD8⁺ T-cells and CD14⁺ monocytes were isolated from AS patients’ peripheral blood mononuclear cells (PBMCs) using a CD8⁺ T-cell or a CD14⁺ monocyte isolation kit (Miltenyi, Bisley, Surrey, United Kingdom), respectively. Jurkat, U937, CD8⁺ and CD14⁺ cells were resuspended at 1×10⁶/ml in pre-warmed Roswell Park Memorial Institute medium supplemented with 10% fetal bovine serum, penicillin/streptomycin and l-glutamine, and rested overnight. Cells were then harvested for experiments.

The impact of rs6600247, which lies in a c-MYC binding-site (Figure 2A), was assessed by EMSA. We designed DNA probes including either the protective T or the AS-risk variant C to evaluate the disruption of a c-MYC consensus motif. The DNA probes used in EMSAs (50-bp single-stranded biotinylated DNA probe incorporating rs6600247) were mixed and annealed at room temperature for 1 h. Probes were then incubated for 20 min with nuclear extracts (NE) obtained either from primary CD8⁺ T-cells or a monocyte cell line from histiocytic lymphoma (U937) stimulated with phorbol-12-myristate-13-acetate (PMA). The sequences of the synthetic single-stranded oligonucleotides are listed below:

C* s (sense): 5′-CTC​CAT​GAC​GCA​ATT​TGG​GCT​CCGTT​ATG​AGT​CAG​CTC​AAG​TAA-3′; T* s: 5′-CTC​CAT​GAC​GCA​ATT​TGG​GCT​CTGTT​ATG​AGT​CAG​CTC​AAG​TAA-3′; C* as (antisense): 5′-TTA​CTT​GAG​CTG​ACT​CAT​AACGGAG​CCC​AAA​TTG​CGT​CAT​GGA​G-3′; T* as: 5′-TTA​CTT​GAG​CTG​ACT​CAT​AACAGAG​CCC​AAA​TTG​CGT​CAT​GGA​G-3′.

(Underlined base highlights the position of rs6600247).

DNA probes as for EMSA (above) were incubated with nuclear extract obtained from U937, CD8⁺ T-cells or purified c-MYC human recombinant protein (Abcam, ab169901 Cambridge, United Kingdom) as previously described (Allen et al., 2017) and separated on DNA retardation gels at 100 V on ice. The samples were transferred on to nitrocellulose membranes for Western blotting (WB), then blocked with 5% milk in Tris Buffer Saline +0.1% Tween (TBST) for 1 h at room temperature (RT) before incubating overnight at 4°C with the primary antibody for c-MYC (Santa Cruz Biotechnology sc-40, Dallas, Texas United States). Secondary goat anti-rabbit antibody (1:10,000 dilution) was added (1 h RT) and the membranes washed before Horse Radish peroxidase substrate (Thermo Fisher Scientific, Waltham, Massachusetts, United States) added for imaging. ImageJ (NIH) was used for quantifying WEMSA bands (Schneider et al., 2012).

Chromosome conformation capture (3C) was performed as previously described (Miele et al., 2006). Briefly, libraries were prepared as follows: 1.5 × 10⁷ of U937 or Jurkat cells were cross-linked with formaldehyde at 1% of the final volume. Glycine [0.125M] was used to quench cross-linking and cells were lysed in cold lysis buffer on ice using a Dounce homogenizer (Sigma Aldrich, Gillingham, United Kingdom). Cells were resuspended in specific restriction enzyme buffer (10 μL were kept as undigested control). The remaining samples were digested overnight at 37°C with 500 units of Sac1 (New England Biolabs, Hitchin, United Kingdom). Digestion was stopped by the addition of 10% sodium dodecyl sulfate incubated at 65°C for 30 min. T4 ligase (Ambion, Thermo Fisher Scientific, Waltham, Massachusetts, United States) was used to perform ligation for 4 h at 16°C. Proteinase K was added prior to reversal of cross-linking at 65°C overnight. Proteinase K was added to the undigested and digested controls saved earlier. DNA was purified using phenol-chloroform extraction, followed by ethanol precipitation. 3C template was resuspended in 500 ul H₂O, while undigested and digested controls in 50 ul. The quality of the chromatin samples was assessed on agarose gels. Bacterial Artificial Chromosome (BAC) preparations were performed similarly as genomic controls. 3C PCR primers were designed along the same strand and same orientation to accomplish specific amplification across 3C ligation junctions. Full list of primers is available in Supplementary Table S1 and their genomic position relative to RUNX3 is shown in Figure 3

We interrogated a genomic region upstream the RUNX3 distal promoter, including few AS-associated SNPs in U937 (monocyte-like) and Jurkat (T-lymphocyte-like) cell lines. The bait was placed at the distal promoter (P2) with amplification primers at the AS-associated SNPs rs6600247 and rs4648889 along with three intergenic regions.

Total RNA from CD8⁺ and CD14⁺ cells was isolated with TRIzol (Invitrogen, Paisley, United Kingdom) and reverse transcribed with Superscript III (Invitrogen, Thermo Fisher Scientific, 168 Third Avenue, Waltham, Massachusetts, United States) to synthesise cDNA as previously described (Vecellio et al., 2018). The specific primers were: RUNX3 sense (s): 5′-ACTCAG CAC CAC AAG CCA CT-3′; RUNX3 antisense (as): 5′-GTC GGA GAA TGG GTT CAG TT-3′. Quantitative PCR was performed in triplicate and the 2−ΔCt method was used to calculate the expression of RUNX3 relative to β-actin (ID Assay qHsaCED0036269, Bio-Rad Laboratories, Kidlington, United Kingdom).

RUNX3 transcription in AS cases and controls was evaluated from previously published data derived from RNA-seq in PBMCs from 72 AS cases and 62 healthy controls and stratified for rs6600247 (Li et al., 2017).

Figure 1A shows the genomic landscape at the RUNX3 locus, with the lead AS-associated SNP rs6600247 lying ∼13kb upstream of the distal promoter while the regulatory SNP rs4648889 is physically closer to the promoter. SNP rs6600247 is situated within a region of open chromatin, defined by a peak for dnase I hypersensitvity (DHS - indicative of regions of open chromatin) (Figure 1B), and a peak of H3K4Me1 histone modification. This sequence also binds the transcription factor c-MYC (ENCODE Factorbook (http://www.factorbook.org/human/chipseq/tf/) (Figure 1B). Taken together, these data suggest an enhancer-type element surrounding rs6600247, so we sought to determine a regulatory role of this SNP. The DHS peak overlapping rs6600247 is seen specifically in CD14⁺ monocytes (Figure 1B). For this reason, we conducted our functional experiments in U937 cells, a pro-monocytic, human myeloid leukaemia cell line, exhibiting monocyte-like features.

We analysed the DNA sequence at rs6600247 and found that the SNP lies within a c-MYC consensus binding motif (Figure 2A). We hypothesised that binding of c-MYC protein to a DNA sequence containing the risk allele C would be reduced. The results of EMSA assessing the relative c-MYC protein binding to the C or T alleles are shown in Figure 2. We first incubated probes with recombinant c-MYC purified protein, and observed a specific DNA/protein complex with both alleles but markedly less to the AS-risk allele C than the protective T allele. (Figure 2B; lane 3-4, n = 3). We then incubated the same probes with NE from U937 (monocyte-like cells) and observed a major protein/DNA complex binding to the protective T allele, but none with the C allele (Figure 2C, lanes 3-4, n = 3). In both cases, successful competition with a 100-fold excess of unlabelled probe confirmed the specificity of the complex (Figure 2B, lane 5-6 and Figure 2C lane 5–6). We next used WEMSA to quantitate the relative binding of c-MYC to each allele of rs6600247. Markedly less c-MYC enrichment was seen with the C risk vs T allele using either c-MYC purified protein or U937 NE (Figures 2D,E, relative band intensities p = 0.01 and p = 0.05, respectively, two-sample t test). We also repeated these experiments using CD8⁺ T-cells and Jurkat NE, showing no differential binding between the two alleles (Supplementary Figure S1).

We used 3C to test plausible chromosome looping interactions between the AS-associated SNP rs6600247 and the RUNX3 distal promoter. Figure 3A shows the RUNX3 genomic region interrogated, the location of the primers and Sac1 restriction sites. Baits were designed to capture Sac1 fragments containing rs6600247, three intergenic fragments with H3K4me1 enrichment, and additionally with a previously-studied AS-associated SNP rs4648889. There was very low interaction frequency between rs6600247 and the distal RUNX3 promoter, either in U937 or Jurkat cells. A stronger interaction frequency was observed between the distal promoter and the region encompassing the AS-associated SNP rs4648889 (Figure 3B) confirming its functional role.

Primary CD14⁺ monocytes and CD8⁺ T-cells from AS patients were used to evaluate RUNX3 mRNA expression stratified on rs6600247 genotype (n = 5 each genotype) (Figures 4A,B). There was a non-significant trend for lower expression in CD14⁺ monocytes with the AS-risk CC genotype compared to protective TT and heterozygous TC genotypes (TT vs CC: 4.6 ± 1.8 vs 2.0 ± 0.4; TT vs CT: 4.6 ± 1.8 vs 1.8 ± 0.2; CC vs CT: 2.0 ± 0.4 vs 1.8 ± 0.2, results are expressed as mean ± standard error mean). We also analysed historical RNA-seq data (Li et al., 2017) obtained from AS case PBMCs measuring RUNX3 mRNA expression, stratified on rs6600247: there was no apparent influence from this SNP on RUNX3 expression (Figure 4C).