Lobular hyperexpression of HLA-A, -B and -C (HLA-ABC) in type 1 diabetes
In accord with earlier reports [9, 14, 21, 22], hyperexpression of HLA-ABC was consistently observed in the islets of patients with type 1 diabetes among all cohorts examined (Fig. 1), but not in controls. The pattern was lobular and mainly restricted to insulin-containing islets (ICIs) (Fig. 1), while insulin-deficient islets (IDIs) displayed normal expression. Islet hyperexpression of HLA-ABC was not confined solely to beta cells, but occurred in all islet endocrine cells (Fig. 1, ESM Fig. 1b).
Classification of donors based on HLA-ABC expression
Since islet hyperexpression of HLA-ABC has been claimed to be artefactual , we monitored the levels of HLA-ABC in a subset of nPOD donors in two independent laboratories using pancreas sections preserved by different methods (frozen vs FFPE). Staining for HLA-ABC was performed using either an immunoperoxidase method coupled with a mouse primary antiserum in FFPE tissue (ESM Fig. 2a) or via immunofluorescence in OCT sections (ESM Fig. 2b) from the same donor, using a different primary antiserum. A blinded analysis was conducted with donors classified into three categories: normal, elevated and hyperexpression (i.e. at least one islet with extremely high expression of HLA-ABC affecting all endocrine cells) (ESM Fig. 2). Unblinding of the analysis revealed a 100% concordance rate between laboratories (ESM Fig. 2c).
Further confirmation of the staining specificity in FFPE tissue was obtained by staining serial islet sections with two different HLA-ABC antibodies. In all cases where hyperexpression of HLA-ABC was detected with one antiserum, this was confirmed in the same islet on the serial section with the second antiserum (ESM Fig. 3, ESM Tables 3, 4).
Examination of patients with increasing disease duration revealed that HLA-ABC hyperexpression was not restricted only to patients with recent-onset type 1 diabetes, but was also observed in individuals with longer-term disease (i.e. up to 11 years) when ICIs were retained (ESM Fig. 4). However, the proportion of ICIs hyperexpressing HLA-ABC decreased as the duration of type 1 diabetes increased (r = –0.883, p < 0.0001; ESM Fig. 4). HLA-I hyperexpression was not found in patients lacking residual ICIs. It was also absent from the ICIs of patients with still longer disease durations (>11 years), even among those who retained insulin immunopositivity (three nPOD patients with a total of 110 ICIs) after this time (ESM Fig. 4).
β2-Microglobulin (β2M) is elevated in islets in type 1 diabetes
Functional HLA-I complexes are heterodimers comprised of an isoform of HLA-I plus β2M. Therefore, the levels of β2M were also assessed and found to be expressed differentially in patients with type 1 diabetes and controls. β2M was present in the islets of individuals without type 1 diabetes (Fig. 2a), but its expression was increased in the ICIs of patients with type 1 diabetes, which also hyperexpressed HLA-ABC. IDIs from the same individuals expressed levels of β2M and HLA-ABC comparable with those seen in non-diabetic controls (Fig. 2b, c).
Expression of RNA transcripts encoding HLA or β2M in laser-captured, microdissected islets
Next, the expression of HLA isoforms and β2M was examined at the RNA level in laser-captured, microdissected islets. RNA was extracted from pooled islets harvested in a manner that did not differentiate between islets with hyperexpression or normal expression of HLA-I or between ICIs and IDIs (Fig. 3). Initially, RNA expression profiles were analysed in islets from the DiViD cohort, since these represent patients with recent-onset disease who retained ICIs [29, 30]. Age-matched control individuals were selected from the nPOD collection. When displayed in a ‘heat map’ format to indicate relative RNA levels using multiple probe sets (Fig. 3a), each of the HLA isoforms (HLA-ABC) and β2M were shown to be markedly elevated. Quantification yielded mean ± SEM increases of 1.9 ± 0.14-fold, 2.15 ± 0.16-fold, 2.02 ± 0.09-fold for HLA-A, -B and -C, respectively, and 2.07-fold for β2M. Data from the nPOD cohort revealed similar trends (ESM Fig. 5), although the effects were less marked. When analysis of the nPOD population was refined by excluding individuals in whom no ICIs could be found in sections adjacent to the pancreatic blocks used for islet RNA isolation, the trend for increased expression of HLA-ABC and β2M was more pronounced (Fig. 3b).
HLA-F expression is also elevated in the ICIs of individuals with recent-onset type 1 diabetes
During analysis of RNA expression in islets from the DiViD patients, it was observed that a non-classical HLA, HLA-F, was also upregulated (by 1.71 ± 0.04-fold) when analysed across all probe sets (Fig. 4a). Therefore, expression at the protein level was assessed in FFPE tissue. This revealed that HLA-F is expressed at low levels in the islets of non-diabetic controls, but is upregulated in the ICIs of patients with recent-onset type 1 diabetes (Fig. 4b, ESM Fig. 6a). The elevated expression was not restricted to beta cells, but could also be observed in alpha cells (Fig. 4c). Similar findings were observed in pancreas tissue from the nPOD, DiViD and UK cohorts (ESM Fig. 6b). Surface localisation of HLA-ABC and HLA-F was observed, but HLA-ABC was also seen in the cytosol of ICIs (ESM Fig. 6c).
NLR family CARD domain containing 5 (NLRC5) expression does not correlate with HLA-ABC hyperexpression
In order to understand the factors that might drive islet HLA-I hyperexpression in type 1 diabetes, we studied NLRC5, a known transcriptional regulator of HLA-ABC and β2M . NLRC5 was readily detected at the protein level in the cytoplasm of beta cells in healthy control pancreas (Fig. 5a). Expression of NLRC5 was similarly detected in the islets of patients with type 1 diabetes but was not elevated, even in islets with demonstrably elevated HLA-ABC expression (Fig. 5a,b). This was confirmed at the RNA level in laser-captured, microdissected islets (Fig. 5c; p = 0.4504).
Signal transducer and activator of transcription 1 (STAT1) expression correlates positively with HLA-ABC hyperexpression in type 1 diabetes
Given that NLRC5 expression was not found to change in parallel with HLA-ABC or β2M in the islets of patients with type 1 diabetes, a second transcriptional regulator, STAT1, was investigated. This protein was present at low levels in the islets of non-diabetic controls (Fig. 6a), and the pattern of staining was similar in tissues from each of the three patient cohorts examined (UK, nPOD and DiViD). STAT1 expression was also low in the IDIs of type 1 diabetes donors (Fig. 6a). However, STAT1 levels were markedly elevated in ICIs that hyperexpressed HLA-ABC (Fig. 6a). STAT1 expression was highest in beta cells and appeared to be localised within both the cytoplasm and the nucleus (Fig. 6a). The fluorescence intensity for immunolabelling of STAT1 and HLA-ABC was measured across a minimum of seven ICIs in seven different individuals. This revealed a striking positive correlation between STAT1 and HLA-ABC expression (Fig. 6b, overall Spearman’s r = 0.5454, p < 0.0001).
Analysis of the expression of STAT1 at the mRNA level in laser-captured islets confirmed the data obtained at the protein level. Moreover, in common with HLA-ABC (ESM Fig. 4), the extent of this increase declined with disease duration (Fig. 6c). Consistent with this, the most pronounced rise was seen in islets harvested by laser-capture microdissection from the six DiViD patients (p = 0.0263; Fig. 6d, ESM Fig. 7), who were studied very close to disease onset.