Introduction
Common variable immunodeficiency disorder (CVID) is an inborn error of immunity with an estimated incidence of 1:20000 in Europe. It is a heterogeneous disorder characterized by hypogammaglobinemia [
1‐
3]. Immunoglobulin replacement therapy reduces the number of bacterial infections [
4]. Many CVID patients develop non-infectious complications that relate to immune dysregulation (ImD) and contribute to increased morbidity and mortality [
4]. CVID-related ImD includes autoimmunity, lymphocytic infiltration of non-lymphoid tissues, lymphoid hyperplasia, and lymphoma [
2]. Patients with increased frequencies of atypical B cells with low expression of CD21 and decreased naïve CD4 T cells are at higher risk of ImD complications [
5‐
7]. CVID-related ImD is associated with increased interferon (IFN)-γ activity [
8‐
10]. Treatment to mitigate ImD in CVID is so far mostly restricted to nonspecific immunosuppressive therapy [
11,
12].
Currently, pathogenic monogenic variants have been identified in 35% of the patients with CVID [
13,
14]. Variants of genes originally identified in non-CVID immunodeficiencies have also been observed in CVID patients, emphasizing the complex underlying genetics of this disorder [
15,
16]. Alterations in epigenetic regulators can cause malfunctions of the immune system [
17]. In addition to genetic variants, epigenetic changes may contribute to the development of CVID [
16,
18]. It was recently reported that epigenetic alterations affecting DNA methylation may contribute to B cell failure in CVID [
19].
MicroRNAs (miRNA) are short non-coding RNAs that are involved in epigenetic regulation and function mainly to repress gene expression. For instance, miRNAs are involved in the regulation of B cell differentiation and have been associated with autoimmune disease and B cell lymphomas [
20‐
23]. Increased T cell expression of mir-201 has been reported in a small cohort of CVID patients [
24]. Changes in circulating miRNAs reflect ongoing biological processes in the tissues and can be used as biomarkers in cancer, autoimmune, and infectious diseases [
25‐
27]. Alteration of miRNA signatures in CVID can be considered the result of epigenetic processes and may contribute to the pathophysiology of CVID and CVID-related ImD. Epigenetic studies on CVID are scarce.
In the present study, we used next-generation sequencing to profile plasma miRNA signatures in CVID patients with and without ImD. Selected miRNAs were validated by polymerase chain reaction (PCR) and correlated to plasma proteins and the levels of naïve CD4 T cells. Predicting disease-associated miRNAs may increase the understanding of the disease mechanisms of CVID and pave the way for new treatment strategies.
Methods
Ethical Approval
This study was approved by the Regional Ethics Committee of Linköping, Sweden (2017/214–32). All study subjects enrolled in this study gave their written informed consent to participate in this study.
Study Subjects
The definition of CVID in this study follows the former Swedish Guidelines on Primary Immunodeficiency Disorders; in brief, serum IgG < 3.0g/L and IgA < 0.07g/L in the absence of any secondary immune deficiency. The cohort included 10 CVID patients with no to moderate ImD,
i.e., infections only (CVID
InO), and 10 CVID patients with severe ImD complications (CVID
C). Clinical data were retrieved from their medical records and covered the time of the study start and 10 years back. An evaluation document was used to score the burden of infections and ImD as previously described [
10]. In brief, patients were assessed for the presence of ImD across three items: autoimmunity, gastroenteropathy and lymphoproliferation. Each item was scored 0–4; hence, the sum formed an ImD score between 0 and 12. Patients were also given an infection score between 0 and 4 based on the number of suspected bacterial infections per year after initiation of IgG-substitution therapy. Information about B cell subsets was retrieved from the clinical records, and patients were classified according to the Freiburg CVID criteria [
28]. Patients were compared to a group of healthy controls recruited among blood donors, who were matched for sex and age (Table
1).
Table 1
Demographics and clinical characteristics of study participants
Age, years, median (range) | 57 (21–76) | 51 (22–74) | 52 (28–68) |
Sex, female/male | 1/9 | 4/6 | 4/6 |
CVID duration, years, median (range) | > 10 (0 to > 10) | 6 (1 to > 10) | N/A |
Freiburg class 1 (1a), n | 5 (2) | 9 (8) | N/A |
CD4+ T cells/µL, median (range) | 491 (338–1260) | 739 (289–1766) | 811 (274–1743) |
CD8+ T cells/µL, median (range) | 482 (225–1126) | 474 (187–2328) | 372 (160–703) |
CD4/CD8, median (range) | 1.4 (0.6–2.2) | 1.7 (0.2–2.6) | 2.2 (1.2–4.3) |
Materials
A detailed list of all key resources including reagents for flow cytometry, critical commercial assays, software, and algorithms is available in supplementary information Table
S1.
Blood Sample Collection and Plasma Preparation
Blood samples were obtained in vacutainer tubes (BD Biosciences). The samples were kept at room temperature for up to 5 h before the plasma was separated and stored at − 80°C until used.
Flow Cytometry for Quantification of Peripheral T Cells
Absolute T cell numbers were determined by flow cytometry with the use of Tru count tubes (BD Biosciences). Briefly, antibodies were added to whole blood and samples were lysed with FACS lysing solution (BD biosciences) before data were acquired using FACS Canto II (BD Biosciences). Kaluza flow cytometry software version 1.5 (Beckman Coulter) was used for data analysis.
RNA extraction, library construction, and high-throughput analyses were carried out by QIAGEN Genomic Services in Hilden, Germany. In brief, RNA extracted from plasma with miRNeasy kit (QIAGEN) was converted into miRNA next-generation sequencing libraries. Adapters containing unique molecular identifiers (UMI) were ligated to the RNA. Then, the RNA was converted to cDNA and amplified using PCR. Libraries were pooled in equimolar ratios and quantified by qPCR before being sequenced on a NextSeq (Illumina Inc., San Diego, CA, USA) according to the manufacturer’s instructions.
The miRNA-seq Data Analysis
Raw data was de-multiplexed and FASTQ files for each sample were generated using the bcl2fastq software (Illumina Inc.). All primary analysis was carried out using CLC Genomics Server 21.0.4 (Qiagen). The workflow “QIAseq miRNA Quantification” of the CLC Genomics Server with standard parameters was used to map the reads to miRBase version 22. In short, the reads were trimmed, before filtering of reads with length < 15 nt or length > 55 nt. The reads were then deduplicated using their UMI. Reads were grouped based on the UMI sequences. Reads that failed mapping to the miRbase were mapped to the human genome, using the “RNA-Seq Analysis” workflow of the CLC Genomics Server.
Identification of Differentially Expressed miRNAs
The empirical analysis of the differential gene expression algorithm of the CLC Genomics Workbench was used for differential expression analysis. For all unsupervised analysis, only miRNAs that had at least 10 counts summed over all samples were considered. A variance stabilizing transformation was performed on the raw count matrix using the function of the R package DESeq2 version 1.28.1. Five hundred genes with the highest variance were used for the principal component analysis. Thirty-five genes with the highest variance across samples were selected for hierarchical clustering.
Targeted Analysis of miRNA with Real-Time qPCR
In brief, extracted RNA was reverse transcribed using the miRCURY LNA RT Kit (QIAGEN). cDNA was assayed in PCR reactions according to the protocol for miRCURY LNA miRNA PCR using miRCURY LNA SYBR Green master mix. The amplification was performed in a LightCycler® 480 Real-Time PCR System (Roche Diagnostics, Basel, Switzerland) in 384 well plates. Target miRNA Cq values were normalized to global mean Cq of hsa-miR-23a-3p and hsa-let-7d-3p, using the formula:
$$\mathrm{Normalized}\;\mathrm{dCq}=\mathrm{global}\;\mathrm{mean}\;\mathrm{Cq}\;\left(\mathrm{sample}\;n\right)-\mathrm{target}\;\mathrm{Cq}\;\left(\mathrm{sample}\;n\right)$$
Target Gene Prediction
The miRNA Enrichment Analysis and Annotation Tool (miEAA) was used for target gene prediction using over-representation analysis (ORA) with at least three hits among input deregulated miRNAs and FDR < 0.01. When CVIDC- or CVIDInO groups were compared to healthy controls, differently expressed miRNAs with FDR < 0.01 were considered deregulated and miRNAs with FDR < 0.1 when the CVIDC group was compared to the CVIDInO group. The Uniprot database was used to find Biological Process GO-terms of predicted target genes.
Targeted Proteomic Analysis
Frozen plasma samples were sent to the Clinical Biomarkers facility, Science for Life Laboratory, Uppsala, Sweden, for analysis. The multiplex protein extension assay (PEA), OLINK Target 96 Inflammation v.3021, was used for the detection of 92 different inflammatory-associated proteins (Table
S2). Seventy-two proteins were detected in > 95% of samples and selected for further analysis. Thirty-two of these proteins were differently expressed (
t-test FDR < 0.05) when comparing CVID patients with healthy controls and were used for the construction of a correlation matrix with normalized miRNA dCq-values.
Protein–Protein Interaction Analysis
Protein–protein interaction networks were created with differently expressed proteins using STRING version 11.5 (
http://string-db.org/) by including interactions with a medium confidence score (> 0.7). STRING database is a curated knowledge database of known and predicted protein–protein interactions [
29].
Statistical Analysis
An Exact Test was used for comparing miRNA profiles generated by high-throughput sequencing, and Benjamini–Hochberg correction was applied to adjust for multiple testing. The Mann–Whitney test and Kruskal–Wallis test with Dunn’s correction were used for univariate statistics comparing two or three groups, respectively. Spearman’s was used for correlation measures.
Discussion
This study is, to our knowledge, the first to delineate the circulating miRNAome in CVID and to correlate the findings to CVID-related ImD. We found various miRNAs that distinguished the CVID patients with severe ImD complications (CVID
C) from the infection-only CVID phenotype (CVID
InO), as well as the CVID groups from healthy controls. The burden of non-infectious complications in CVID was assessed by a clinical severity scoring system assessing lymphoproliferation, autoimmunity, and gastrointestinal complications [
10]. The accuracy of the scoring system was validated using the more extensive CVID disease severity score published by Ameratunga in 2018 [
30]. Incomplete data of mild manifestations should not affect the results significantly since mild manifestations only generated a score of 1 using the Ameratunga scoring system. Twenty-four miRNAs were altered between CVID
C patients and the controls, and 24 between CVID
InO patients and the controls. Several genes involved in adaptive immune mechanisms, including a gene involved in type 2 interferon signaling, were predicted candidate gene targets of deregulated miRNAs in the CVID
C group. We and others have previously shown that IFN-γ correlates with CVID-related ImD [
10,
32]. The top predicted gene targets in the CVID
InO group were merely associated with inflammatory processes and innate immunity. Low naïve CD4 T cells are associated with severe disease and a poor outcome in CVID. Targeted analysis with qPCR showed a negative correlation between naïve CD4 T cell numbers and mir-34a-5p expression, and the plasma levels of four putative mir-34a-5p gene targets,
i.e., CD40, CXCL9, IL10, and TNF, were elevated in CVID
C compared to CVID
InO patients. Determining correlations between miRNAs and clinical and laboratory findings may provide new mechanistic insights into CVID-related ImD since circulating miRNA transcripts reflect intracellular gene network activity in tissues [
26].
The majority of CVID cases are sporadic and the onset can occur at any age. A few pathogenic gene variants have been associated with CVID [
14,
15]. However, only a minority of CVID cases appear to be the result of a monogenic variant. Instead, transcriptional and post-transcriptional modifications may contribute to the clinical penetrance of genetic variants in CVID, alone or in combination with other triggers such as infections [
16]. Increasing evidence indicates that miRNAs play a key role in biological processes by their modulation of gene expression [
33]. While the plasma miRNA profiles clearly discriminated between CVID patients and healthy controls, the differences between the CVID
C- and the CVID
InO patients were less pronounced. However, the distinct circulating miRNA profiles found in CVID
C- and in CVID
InO patients compared to the controls suggest that deregulated expression of circulating miRNAs may contribute to the heterogeneity of CVID phenotypes. In addition, ORA of differently expressed miRNAs identified several potential gene targets involved in T- and B cell homeostasis in CVID patients with severe ImD. These findings imply that miRNAs may contribute to the changes in T- and B-cell populations found in this group of CVID. For instance, DAPK-1 was one of the most highly predicted gene targets in the CVID
C group compared to the controls. DAPK-1 is a tumor suppressor induced by IFN-γ [
34,
35] and can also influence the expression of IFN-γ [
36]. Altered DAPK1 activity may impair CVID T cell functionality, seeing that DAPK1 has been linked to the regulation of T cell function [
37,
38]. WNT5A was another predicted target when comparing the miRNA profiles of CVID
InO patients with healthy controls. WNT5A, like DAPK1, is a tumor suppressor which can also modulate T cell activation [
39‐
41]. Taken together, dysregulation of DAPK1 and WNT5A may contribute to the heterogeneity of the immune defects observed in CVID.
When comparing the plasma miRNA profiles of the two CVID groups, the highest upregulated miRNA in CVID
C was mir-224-5p. Mir-224 expression is upregulated via NFΚB and has been reported to attenuate the inflammatory responses by interaction with the acute phase protein pentraxin 3 [
42,
43]. The CVID
C group had low levels of mir-1-3p, and this miRNA has been found to modulate immune cell functions and contribute to inflammatory damage in an experimental model of sepsis [
44,
45]. Increased expression of miR-224-5p and reduced expression of miR-1-3p may regulate the inflammatory response in CVID related ImD. Bioinformatic analysis of deregulated miRNA in the CVID
C group compared to the CVID
InO group identified several putative gene targets such as NOTCH2, which is important for T cell responses [
46], the PIK3CA, part of the PI3K signaling pathway [
47], and the inflammatory caspase 3, a key protein in apoptosis and pyroptosis pathways [
48]. In summary, deregulated miRNA in CVID may potentially modulate fundamental cellular processes and thereby influence CVID-related ImD.
MiRNA raw count means, significant fold change between the CVID
C- and CVID
InO groups, and putative involvement in IFN-γ signaling were factors considered in the selection of miRNAs for the validation assay with real-time PCR. Among the miRNAs in the validation assay, we found correlating levels of mir-103a-3p with mir-34a-5p and mir-301a-3p. In addition, both mir-103a-3p and mir-34a-5p correlated with inflammatory plasma proteins such as CXCL9, IL10, and IL12B. Protein–protein-interaction (PPI) network analysis indicated proteins regulating immune responses to be overrepresented in the mir-103a-3p cluster and the mir-34a-5p clusters. Experimental models have suggested that mir-103a-3p both promotes and ameliorates inflammatory responses [
49,
50] and that this miRNA is upregulated in rheumatoid arthritis and connected to disease activity [
51]. We found a negative correlation of mir-103a-3p with inflammatory proteins, which may be indicative of an anti-inflammatory activity in CVID. It is unlikely that mir-103a-3p specifically counteracts IFN-γ driven inflammation since plasma levels were similar in the two CVID groups. The net effect of mir-103a-3p in CVID remains to be determined.
Increased levels of mir-34a-5p have previously been reported in the whole blood of lung cancer patients [
52] and in the plasma of people living with HIV, even after prolonged control of viral replication [
53]. Functional in vitro studies have shown that mir-34a-5p modulates T cell signaling and the expression of IFN-γ induced CXCR3 ligands such as CXCL9, CXCL10, and CXCL11, which are chemoattractants for activated T cells; hence, mir-34a-5p may act as a major hub of T cell regulatory networks [
54‐
56]. These findings support a role for mir-34a-5p in the pathogenesis of CVID-related ImD, which is associated with upregulation of the Tbet-CXCR3-axis with increased numbers of T-helper 1 skewed T cells and increased IFN-γ-production [
10,
32]. Moreover, PPI analysis of plasma proteins correlating with mir-34a-5p predicted the enrichment of factors involved in a chronic inflammatory response, and the highest levels of mir-34a-5p were found in CVID
C patients, which may also imply a role of mir-34a-5p in CVID related ImD. Among the predicted immune system process gene targets of mir-34a-5p, we analyzed the plasma levels of four gene products (CD40, CXCL9, IL10, and TNF). They were all increased in CVID
C patients compared to CVID
InO patients and this may reflect upregulated mir-34a-5p activity. CXCL9 and TNF are Th1 cytokines. Mir-34a-5p is considered a tumor suppressor acting by augmenting Th1 T cell responses and preventing cancer cell immune checkpoint upregulation and subsequent immune evasion [
57,
58]. Hence, it can be hypothesized that mir-34a-5p contributes to the enhanced CD4 T cell differentiation in CVID-related ImD, which is supported by negative correlations of naïve CD4 T cells and mir-34a-5p levels in this study. If the significance of mir-34a-5p in CVID-related ImD can be verified in functional studies, blocking of mir-34a-5p may potentially be a new therapeutic approach to consider.
Our study has some limitations. First is the restricted number of patients included and their heterogeneity when it comes to the clinical manifestations of CVID. Despite this, we identified differences in miRNA profiles when comparing CVID patients with- and without severe ImD that predicted enrichment of target genes involved in IFN-γ signaling. Furthermore, the distribution of male vs female in the infection only group was highly skewed towards male, which is a limitation, but when exploring the whole cohort, we did not find any significant differences in dCq-levels of miRNAs between male and female participants, so it should have a limited impact on the study. Moreover, it is an explorative study, and the findings need to be validated by longitudinally analyzing a larger study cohort, and an alternative study design is required to dissect the potential mechanistic role of miRNAs in the development of CVID-related ImD.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.