Circulating miRNAs as Potential Biomarkers for Celiac Disease Development

Abstract

Background & aims: Celiac disease (CeD), an immune-mediated disease with enteropathy triggered by gluten, affects ~1% of the general European population. Currently, there are no biomarkers to predict CeD development. MicroRNAs (miRNAs) are short RNAs involved in post-transcriptional gene regulation, and certain disease- and stage-specific miRNA profiles have been found previously. We aimed to investigate whether circulating miRNAs can predict the development of CeD.

Methods: Using next-generation miRNA-sequencing, we determined miRNAs in >200 serum samples from 53 participants of the PreventCD study, of whom 33 developed CeD during follow-up. Following study inclusion at 3 months of age, samples were drawn at predefined ages, diagnosis (first anti-transglutaminase antibody (TGA) positivity or diagnostic biopsy) and after the start of a gluten-free diet (GFD). This allowed identification of circulating miRNAs that are deregulated before TGA positivity. For validation of the biomarkers for CeD and GFD response, two additional cohorts were included in subsequent meta-analyses. Additionally, miRNAs were measured in duodenal biopsies in a case-control cohort.

Results: 53 circulating miRNAs were increased (27) or decreased (26) in CeD versus controls. We assessed specific trends in these individual miRNAs in the PreventCD cohort by grouping the pre-diagnostic samples of the CeD patients (all had negative TGA) by how close to seroconversion (first sample positive TGA) the samples were taken. 8/53 miRNAs differed significantly between controls and samples taken <1 year before TGA positivity: miR-21-3p, miR-374a-5p, 144-3p, miR-500a-3p, miR-486-3p let-7d-3p, let-7e-5p and miR-3605-3p. 6/26 downregulated miRNAs reconstituted upon GFD, including miR-150-5p/-3p, whereas no upregulated miRNAs were downregulated upon GFD. 15/53 biomarker candidates also differed between CeD biopsies and controls, with a concordant direction, indicating that these circulating miRNAs might originate from the intestine.

Conclusions: We identified 53 circulating miRNAs that are potential early biomarkers for CeD, of which several can be detected more than a year before TGA positivity and some start to normalize upon GFD.

Keywords: autoimmunity; celiac disease; pre-clinical marker; pre-diagnostic marker; small RNA sequencing.

Figure 1

IgA anti-transglutaminase levels peak at diagnosis in the patient group only. IgA anti-transglutaminase levels (TGA) in serum samples of PreventCD participants displayed by age of sampling (CeD=individuals who developed CeD; Ctr=age-matched samples of individuals who did not develop CD; M=Months). For the individuals that developed CeD, we also show serology at diagnosis and after initiating a gluten-free diet (GFD). Samples of individuals in the CeD group that were taken at timepoint of first positive TGA (seroconversion) or at the time of the diagnostic biopsy, were grouped in the diagnosis group (age median: 24, range: 13 - 64 months). One control individual showed positive TGA (29 U/L), but this individual did not have or develop CeD in the follow up (see Supplemental Methods for more information). This sample with a positive TGA in the control taken at 3 years of age was grouped with the M18-M24 age group for visualization and analysis purposes. Dashed lines indicate the cutoffs used to assign positivity, depending on the two types of tests used (see Methods). Boxplots were generated using the default parameters in the R package ggplot2 (median, second and third quartiles shown by the hinges, individual datapoints are displayed outside the whiskers beyond 1.5 * interquartile range).

Figure 2

Analyses in the separate cohorts that were performed before combining the results of the differential expression in two meta-analyses. The goals here were to: in part 1) find miRNAs that are potential biomarkers for CeD development and part 2) find miRNAs that change upon the gluten-free diet (GFD). Corresponding sample sizes are shown in grey. *In the PreventCD cohort, the samples “at diagnosis” include samples at seroconversion (first positive IgA anti-transglutaminase (TGA) levels) and samples taken close to the diagnostic biopsy. All samples in the “before diagnosis” groups had negative TGA levels.

Figure 3

53 circulating miRNA biomarker candidates for CeD development. Log2fold changes are depicted for three separate differential expression (DE) analyses (A–C) of 53 microRNAs that were significant in the meta-analysis combining these analyses. (A) PreventCD: pre-diagnostic samples of CeD patients (IgA anti-transglutaminase (TGA) negative) versus controls. (B) PreventCD: CeD at diagnosis (at seroconversion (TGA positivity) or at diagnostic biopsy) versus samples at 4 months of age (before gluten consumption). (C) Milano-Bicocca: CeD at diagnosis versus controls. (D) Milano-Bicocca: CeD at diagnosis versus controls in intestinal biopsy samples. Right panel shows a forest plot for the meta-analysis (beta and 95% confidence interval). miRNAs that are detectable < 12 months before diagnosis are indicated in bold.

Figure 4

Several miRNA biomarkers for CeD change months to years before detection of CeD serology. The levels of eight out of the 53 microRNAs listed in differ from controls < 12 months before seroconversion (first IgA anti-transglutaminase positivity). Shown are mean values ± standard error of the regularized log-normalized miRNA counts, corrected for batch and age. Black: controls; Dark-grey: pre-diagnostic samples of CeD patients grouped by months till seroconversion (all samples had negative IgA anti-transglutaminase levels); Red: samples at diagnosis (samples at seroconversion or at time of biopsy); grey: CeD patients after start of the GFD.

Figure 5

miR-21-3p can be detected at high levels in pre-diagnostic samples of patients but not in age-matched controls and is significantly upregulated in the small intestinal biopsies of CeD patients. (A) PreventCD cohort: grouped by age of sampling (M=Months). (B) PreventCD cohort: pre-diagnostic (IgA anti-transglutaminase negative) samples of CeD patients are grouped by time till seroconversion: more than 24 months before seroconversion (>24), between 24-12 months before seroconversion (24-12), less than 12 months before seroconversion (<12), or at diagnosis (taken at seroconversion or at time of biopsy) and 6 months after starting GFD. Controls: all samples of the PreventCD controls. (C) Circulating miR-21-3p in the Milano-Bicocca cohort (circulation). (D) miR-21-3p expression in small-intestinal biopsies in the Milano-Bicocca cohort.

Figure 6

Fifteen circulating miRNAs change after start of the GFD. Left panel shows the 15 circulating miRNAs that were significant in the meta-analysis when combining the following comparisons: (A) PreventCD: GFD vs CeD at diagnosis (taken at seroconversion or at time of biopsy) (B) Milano-Bicocca: GFD vs CeD at diagnosis and (C) GFD volunteers: GFD vs gluten containing diet. Right panel shows forest plot for the meta-analysis (beta and 95% confidence interval). Bold text indicates miRNAs that are also among the 53 CeD biomarker candidates and show a normalizing trend upon GFD.

Figure 7

MiR-150-5p is significantly decreased in CeD and reverses after start of a GFD. (A) PreventCD: high-risk controls and CeD patients at time of diagnosis (taken at seroconversion or at time of biopsy) and CeD patients after start of a GFD. (B) Milano-Bicocca cohort: controls at time of diagnosis (CeD) and at GFD. (C) GFD volunteers: on gluten-containing diet or on GFD.