Differential expression of predisposing HLA-DQ2.5 alleles in DR5/DR7 celiac disease patients affects the pathological immune response to gluten

The DR5-DQ7/DR7-DQ2 genotype is very frequent among patients affected by celiac disease (CD), in Europe. This genotype, associated to high risk of CD, carries the HLA-DQA1*05 and HLA-DQB1*02 predisposing alleles, in trans configuration. The alleles encode the DQ2.5 heterodimer responsible of gluten peptide presentation on the surface of antigen-presenting cells (APCs), and consequent pathogenic CD4+ T cell activation. We demonstrated that DR5/DR7 APCs induce an anti-gluten CD4+ T cell response, of comparable intensity to that observed with APCs carrying DR1/DR3 genotype, which risk alleles are in cis configuration. In addition, we showed that DR5/DR7 APCs from celiac patients stimulated an effector CD4+ T cell response higher with respect to that induced by DR5/DR7 APCs from healthy subjects. To explain these findings, we assessed the DQ2.5 RNA and protein quantity. We showed that the expression of DQA1*05 and DQB1*02 risk alleles is much higher than the expression of non-CD-associated alleles, in agreement with the previous results obtained with DR1/DR3 genotype. The differential expression of transcripts influences the quantity of DQα1*05 and DQβ1*02 chains and, as consequence, the cell surface density of DQ2.5 heterodimers. Moreover, both RNA and proteins, are more abundant in APCs from celiac patients than controls. Finally, to unravel the mechanism regulating the expression of predisposing DQA1*05 and DQB1*02 alleles, we quantified the new synthetized RNA and found that the differential expression is explained by their transcription rate. Our results confirmed that the strength of antigen-specific CD4+ T cell response is mainly determined by the amount of gluten in the diet and provided a new possible approach for a personalized diagnosis and for risk stratification.


Scientific Reports
| (2020) 10:17227 | https://doi.org/10.1038/s41598-020-73907-2 www.nature.com/scientificreports/ CD-associated DQA1*05 and DQB1*02 alleles are in cis configuration, that these risk alleles are significantly more expressed than non-associated ones on the other chromosome. The preferential expression of risk alleles induces high expression of DQα1*05 and DQβ1*02 protein chains, resulting in a comparable amount of DQ2.5 heterodimers on cell surface of the antigen presenting cells (APC) in celiac patients carrying DQ2.5 genes both in homozygous and heterozygous assets. This result supported our interest for DR5/DR7 genotype 5 highly frequent in Southern Italy, from where our patients come. In the present work, we investigate the responsiveness of antigen-specific CD4 + T cells, stimulated with gliadin-pulsed APC carrying the DR5-DQ7/DR7-DQ2 genotype. Moreover, we assessed the transcripts and protein levels and investigated on the possible mechanism that might explain the risk alleles differential expression.
In addition we compared the IFN γ response of TCL stimulated with B-LCLs from celiac patient with respect to healthy control. The capability of DR5/DR7 B-LCLs from healthy control to stimulate celiac TCL resulted significantly reduced compared to celiac B-LCLs respectively of 33% (compared to DR5/DR7) and 35.9% (compared to DR1/DR3), at the sub-optimal peptide concentration of 1 µM, (Fig. 1B), (p = 0.02). www.nature.com/scientificreports/ These results clearly highlight that DQ2.5 genes, either in cis (DR1/DR3) or in trans (DR5/DR7) configurations, provide to APCs high and comparable ability to stimulate antigen-specific immune response. This phenomenon is much evident with APCs from subjects with a diagnosis of CD with respect to healthy controls.
The expression of DQα1*05 and DQβ1*02 chains is different among B-LCLs from HLA DR5/ DR7 CD patients and controls. To explain the difference in the ability of DR5/DR7 B-LCLs from celiac patients and controls, we quantified the amount of DQα1*05 and DQβ1*02 protein chains forming the DQ2.5 heterodimeric molecules. The evaluation of the expression level of HLA DQ2.5 heterodimers was assessed by specific monoclonal antibodies against DQα1*05 and DQβ1*02 proteins, since antibodies directed against the DQ2.5 heterodimer are not available. The expression of DQα1*05 chain on B-LCL from each patient and control subjects is shown in Fig. 2A, while Fig. 2B showed the average variation in the Mean Fluorescent Intensity (MFI) value observed between the two groups. Similarly, the expression of DQβ1*02 molecule on B-LCL from DR5/DR7 patients and control subjects was shown in Fig. 2C and the statistical comparison of MFI between two group was reported in Fig. 2D. Our results clearly demonstrated a greater surface expression of DQα1*05 and DQβ1*02 chains in CD patients with respect to healthy controls carrying the DR5/DR7 genotype. These results explain the higher gliadin-specific CD4 + T cell activation when stimulated by DR5/DR7 APCs from celiacs with respect to controls (Fig. 1).
CD-associated DQA1*05 and DQB1*02 alleles are more expressed in B-LCL carrying DR5-DR7 genotype. To explain functional data, we measured the mRNA amount of CD-associated DQA1*05 and DQB1*02 alleles when they are in trans configuration with respect to the RNA amount of alleles not associated with the disease. DQA1* and DQB1* transcripts were quantified by qRT-PCR in B-LCLs with DR5/DR7 genotype from 8 CD patients (#21 to #28 B-LCL) and 7 healthy subjects (#29 to #36 B-LCL) ( Table 1). The amount of mRNA encoded by each allele was expressed as percentage of total DQA1* transcripts (Fig. 3A) and as percentage of total DQB1* transcripts (Fig. 3C). We observed that the mean percentage of DQA1*05 mRNA was 75%, significantly higher than the DQA1*02 mRNA mean percentage (Fig. 3A). Similarly, the mean amount of DQB1*02 mRNA was 76%, significantly greater than DQB1*03 mRNA (Fig. 3C). However, the differences between the messengers transcribed by the two alleles were lower when we analyzed both DQA1* and DQB1* genes in B-LCL from the healthy subjects with the same genotype (DR5/DR7). In fact, we observed a significant difference when we compared the mean percentages obtained for DQA1*05 mRNA in B-LCL from patients (75%) and controls (64%) (Fig. 3B), as well as when we compared the mean percentages obtained for DQB1*02 mRNA in patient (76%) and control (63%) cells (Fig. 3D). These differences correspond to a 7-8% of increment  www.nature.com/scientificreports/  The differential expression of risk alleles is determined by de novo transcription. In order to unravel the mechanism responsible of differential alleles expression we used click-iT chemistry 7 able to monitor nascent transcripts. Indeed, we investigated if the differential expression of DQA1*05 and DQB1*02 alleles with respect to DQA1*02 and DQB1*03 might be determined by a different rate of transcription. We labeled growing B-LCL#27 with a uridine analog, the 5-ethynyluridine (EU) that is incorporated into the nascent RNA allowing us to assess the rate of transcription of CD associated DQA1*05 and DQB1*02 alleles with respect to non-CD associated DQA1*02 and DQB1*03. After an overnight incubation, cells are harvested and EU-labeled RNA, deriving from de novo transcription, is prepared and used to synthetize cDNA to quantify the RNA transcribed by each allele by qRT-PCR. As reported in Fig. 4A, the new synthetized HLA-DQA1 mRNA is mainly transcribed by DQA1*05 with respect to DQA1*02 allele (82% versus 18%). Similarly, the new transcribed DQB1*02 mRNA is higher than DQB1*03 mRNA (71% versus 29%, Fig. 4B). Our results clearly demonstrated a greater rate of transcription of CD risk alleles either in cis (data not shown) or in trans (Fig. 4) configurations and explain the different amount of DQA1*05 and DQA1*02 RNA found in DR5/DR7 B-LCLs.

Discussion
The CD risk alleles HLA DQA1*05 and DQB1*02 encode the HLA DQ2.5 molecule either when they are in cis (DR3-DQ2 haplotype), or in trans configuration (DR5-DQ7/DR7-DQ2 genotype). Both are associated to high risk for CD. The risk to develop CD has been associated to the ability of DQ2.5 molecule to present full repertoire of gliadin-derived antigenic peptides to CD4 + T cells 8 . In our previous work, we demonstrated that the APCs carrying the risk genes in cis configuration, both homozygous (DR3/DR3) and heterozygous (DR1/ DR3) for DQ2.5, have the same stimulatory capability on gluten-reactive CD4 + T cells 4 . This result was explained by the high expression of DQA1*05 and DQB1*02 alleles also in heterozygous (DR1/DR3) genotype and, as a consequence, by the great density of DQ2.5 heterodimers on APC membrane surface. In the present work, we expanded our analysis to APCs carrying the predisposing alleles in trans configuration, as the DR5/DR7 genotype is very frequent in Southern Europe and associated to high risk. We showed that APCs with DR5/DR7 genotype induces an anti-gluten CD4 + T cell stimulation as strong as APCs carrying DR1/DR3 from celiac patients, and that the magnitude of CD4 + T cells response depends on the antigen concentration. Moreover, we observed that the strength of CD4 + T cell activation is lower when stimulated by DR5/DR7 APC from healthy subjects. This result was confirmed by the reduced expression of DQα1*05 and DQβ1*02 surface chains in controls that impacts on the density of heterodimeric DQ2.5 molecule. When we measured the expression of both DQA1*05 and DQB1*02 risk alleles, we found a higher amount of RNA with respect to non-disease associated DQA1*02 and DQB1*03 transcripts. The differential expression of alleles was much stronger in APC from celiac patients that non-affected subjects, although the latter showed an RNA quantity higher than 50%. To gain information regarding the molecular mechanism that induces the differential expression of risk alleles and to formulate a functional hypothesis, we investigated on the transcription rate. Through click-iT chemistry, monitoring the incorporation of an analog of uridine during transcription, we quantified the new synthetized DQA1 and DQB1 mRNAs. This experiment demonstrated that the great expression of risk alleles DQA1*05 and DQB1*02 is mainly due to a difference in the transcription rate with respect to DQA1*02 and DQB1*03 alleles non-associated to the pathology. These findings and others, relative to Type 1 diabetes 9 and multiple sclerosis 10 , suggested that the high transcription is typical of HLA-DQ or HLA-DR risk alleles associated to autoimmune diseases, regardless of the www.nature.com/scientificreports/ haplotype to which they belong. Several papers, in the recent literature, demonstrated the relationship between genomic variants in the intergenic regions across the MHC locus, and the expression of risk alleles predisposing various autoimmune diseases [11][12][13] . The correlation between eQTL and MHC haplotype sequence variation has been assessed 14 , as well as the differential expression of DQA1 and DQB1 alleles in heterozygous CD4 + T cells 15 , depending on lymphocytes activation. In addition, we might also speculate differences in the regulative function of genes and/or factors included in the upstream activation pathways of risk alleles that, in pathological conditions, might cause stronger promoter activity. We propose that the high transcription rate should be attributed to only one allele, encoding alfa or beta chain, because we demonstrated a coordinate regulation of two genes encoding the HLA class II heterodimer 9 .
In conclusion the preferential expression of risk alleles impacts on DQ2.5 surface density and on the efficacy of gliadin antigen presentation. The risk to develop CD is associated to the threshold of CD4 + T cells activation, strictly dependent on the amount of gluten immunogenic peptides exposed by intestinal T lymphocytes. This conclusion was proven by the comparable strength of antigen-specific CD4 + T cells activation when stimulated by antigen-pulsed DR3/DR3 homozygous or heterozygous DR3/DR1 and DR5/DR7 APCs. All APCs, regardless their haplotypes, stimulated the gliadin-specific CD4 + T cells reaching the plateau of activation, thanks to the large amount of HLA-DQ2.5-gluten epitopes complex presented on APC membrane surface. In addition, our data demonstrated that the expression level of predisposing alleles, found much higher in CeD patients with respect to healthy controls, might be quantified with a Δ value representing the difference in the DQ expression genes between the two groups of subjects. This opens the possibility to further stratify the CD genetic risk and to support an early disease diagnosis. However the high risk associated to DR3/DR3 patients highlighted by many papers might be explained by the contribute of other loci accounting for an additional 18% of genetic heritability 16 in addition to the role of the B*0801 gene of the HLA class I region in LD with DR3-DQ2.5 genes 17 .
Our molecular data, demonstrating the high risk associated to DR5/DR7 genotype, are fortified and supported by genetic studies demonstrating that DQA1*05:01 and DQB1*02:01 alleles are the primary contributors to CD susceptibility. These studies are based on the availability of GWAS 18 data and on several HLA imputation programs that define genetic models of association between classical HLA alleles and autoimmune diseases. Lenz et al. 19 using the SNP2HLA 20 imputation algorithm demonstrated that, in addition to the additive model, in which the risk associated to a specific haplotype depends on the dosage effect or on the sum of risk for each allele, there is a non-additive model in which the synergistic interactions between different haplotypes confer a great disease risk. This genetic point of view confirmed the strong celiac disease risk associated with DR5/DR7 genotype, herein analyzed.

Conclusions
The molecular data we presented highlight the pivotal role that mRNA has in the risk stratification and opens to a new approach in supporting an early diagnosis of CD and autoimmune disorders on the assessment of DQ2.5 gene expression.

Methods
Antigen presenting cells. The EBV-transformed, B lymphoblastoid cell lines (B-LCL) were established from PBMCs of celiac patients or non-celiac donors by infection with a culture medium of Marmoset cells 21 , previously titrated for virulence, in the presence of cyclosporine A at 0.5 mg/ml. All CD volunteers and healthy controls, recruited as a source of APCs, as well as for the generation of gluten reactive T cell lines and clones, were typed for DQA1, DQB1 and DRB1 haplotypes by PCR using the following HLA typing kits: AllSet Gold SSP HLA-DR Low Res kit, AllSet Gold SSP HLA-DQ Low Res kit and DQA1 SSP UniTray kit, all from Invitrogen (Life Technologies). The haplotypes of the B-LCL are reported in Table 1.
Monoclonal antibodies and flow cytometry analysis. B-LCLs were harvested at sub-confluence and were suspended at 10 6 cells/ml in ice cold PBS, 10% FCS and 1% NaN3. Then, 100 ml of cell suspension was plated in a 96 U-bottom plate and labelled with 1 μg/ml of primary or isotypes control monoclonal antibodies, or with 10 μl of the hybridoma supernatant, previously titrated. The cells were incubated at 4 °C in the dark for 30 min, washed and thereafter labelled with secondary antibodies at a final concentration of 10 μg/ml in 3%  23 . Fluorochrome conjugated anti-rat IgG (-PE) and anti-mouse IgG (-FITC) were used as secondary antibodies. All phenotypes were analysed with FACSCanto II system and elaborated using the DIVA software (BD Biosciences).
Steady-state and nascent and mRNA quantization. Total RNA from B-LCL was prepared with the Aurum Total RNA kit (Bio-Rad), and 1 μg of RNA was used for reverse transcriptase reactions, performed using an iScript cDNA Synthesis kit (Bio-Rad).The amount of specific transcripts was measured by qRT-PCR using the Quanti Tect SYBR Green PCR Kit (Bio-Rad) through the DNA Engine Opticon Real-Time PCR Detection System (Bio-Rad). Each reaction was run in triplicate in the presence of 0.2 mM primers synthesized by Eurofins, and each experiment was performed four times 4 . The primer sequences are reported in Table 2. The relative amount of specific transcripts was calculated by the comparative cycle threshold method 24 and β-actin transcript was used for normalization.
The newly synthesized RNA transcripts were captured by Click-iT Nascent RNA Capture Kit (Thermo Fisher Scientific) according to manufacturer's instructions. Briefly, B-LCLs, seeded at 50% confluency, were labelled with 0.2 mM ethynyl uridine (EU) and incubated at 37 °C for 16 h. Total RNA was prepared with TRIzol reagent (Life Technologies). The EU-labeled RNAs were biotinylated with 0.5 mM biotin azide in Click-iT reaction buffer. The biotinylated RNAs were precipitated and resuspended in distilled water. 0.5 µg of purified RNA was bound to 25 μl of Dynabeads MyOne Streptavidin T1 magnetic beads in Click-iT RNA binding buffer. The RNA captured on the beads was used as template for cDNA synthesis. Reverse transcription was performed using the SuperScript VILO cDNA Synthesis Kit Invitrogen (Thermo Fisher Scientific) following the manufacturer's instructions. The amount of specific transcripts was measured by qRT-PCR using Sso Advanced SYBR Green PCR Kit (Bio-Rad). The apparatus and primers used were the same described above. Statistical analysis. All results shown are the mean of at least three independent experiments. Statistical analysis was performed using the unpaired Student's t-test with two-tailed distribution and assuming two samples equal variance parameters. In the figures, a single asterisk corresponds to p < 0.05 and double asterisks correspond to p < 0.01.