Background
Rheumatoid arthritis (RA) is a chronic autoimmune disease that results from a complex interplay between genetics, environmental factors, and the immune system. Retrospective studies of RA onset based on archival serum samples have indicated that rheumatoid factor (RF) and anticitrullinated protein antibodies (ACPA) are detectable months to years prior to clinical disease onset, and they exhibit a progressive increase in titer as disease onset approaches. In the case of ACPA, this phenomenon is believed to relate to expansion of an autoantigen repertoire targeted by the ACPA, a process that has been termed
epitope spreading [
1‐
3].
We previously demonstrated a high prevalence of RA in an indigenous North American (INA) population in Manitoba, Canada, an observation that is consistent with those in other INA populations [
4]. In this population, RA is characterized by familial disease aggregation and early age of disease onset [
5,
6]. A high proportion of these INA patients with RA are genetically predisposed by having shared epitope encoding HLA-DRB1 alleles, particularly *1402 and *0404 [
7]. The disease is primarily seropositive, and it is severe and disabling, with frequent large joint involvement. In studying the first-degree relatives (FDRs) of INA patients with RA, we have demonstrated frequent RF and ACPA seropositivity, and we have shown that the serum cytokine profile of the FDRs resembles that of their affected relatives more so than that of control subjects with no family history of autoimmune disease [
6‐
8]. Thus, this population is ideally suited for studying the onset of RA in high-risk individuals and the potential role that genetic, environmental, and epigenetic factors play in the process.
MicroRNAs (miRNAs, miRs) are conserved, small, noncoding, single-stranded RNAs (~18–25 nucleotides) that play a role in posttranscriptional gene regulation. miRNAs bind to the 3′-untranslated region (3′-UTR) of target messenger RNA (mRNA) and induce gene silencing by either promoting mRNA degradation or transcript destabilization, resulting in suppression of target protein synthesis [
9,
10]. In RA, proinflammatory cytokines (e.g., tumor necrosis factor-α, interleukin [IL]-1β, and IL-17) alter the expression of multiple miRNAs (e.g., miR-155, miR-146a, miR-26b, miR-16, and miR-21) in peripheral blood mononuclear cells (PBMCs), synovial fibroblasts, T lymphocytes, and synovial tissues derived from patients with RA [
11‐
13]. In turn, miRNAs regulate inflammatory and signaling pathways influencing cellular differentiation and bone homeostasis within the synovial microenvironment [
11]. Consequently, miRNAs play a central role in the regulation of inflammatory processes, synovial proliferation, and osteoclastogenesis, thus affecting the disease activity in RA [
12‐
14]. Therefore, miRNAs may serve as a critical epigenetic component in the breakdown of immune tolerance and progression toward RA disease onset.
There is limited knowledge on the role of miRNAs in RA pathogenesis, particularly during the preclinical phase of the disease. To define mechanisms underpinning the progression of autoimmunity toward disease onset in at-risk individuals, we sought to evaluate miRNA expression profiles in blood samples derived from INA patients with RA, their seropositive FDRs, and healthy control subjects (HCs). This is the first study to demonstrate unique and reproducible differences in miRNA expression patterns in whole blood between these groups. Furthermore, we demonstrated that miR-103a-3p is uniquely upregulated in both patients with RA and FDRs. The observed miRNA patterns and the molecular networks they represent are of value in defining new mechanisms involved in RA onset while being potentially useful as biomarkers for predicting onset of preclinical RA.
Methods
Study design
INA study participants were recruited from Cree, Ojibway, and Oji-Cree communities in central Canada [
5,
6]. The biomedical research ethics board of the University of Manitoba approved the overall design of the study and the consent forms (ethics, 2005:093; protocol, HS14453). Specific research agreements with the study communities were developed and approved by the community leadership. The conduct of the study was guided by the principles of community-based participatory research, a cornerstone of the Canadian Institutes of Health Research guidelines for Aboriginal health research (
http://www.cihr-irsc.gc.ca/e/29134.html). As such, community leadership provided input into the initial development of the project, as well as ongoing input through advisory board meetings. Local healthcare providers were trained in study methodology and standard operating procedures. Regular knowledge translation activities such as newsletters and local radio appearances by study investigators provided the communities with updates regarding progress and significance. The study participants provided informed consent after the study was explained to them in detail, with the help of an INA translator from their community where necessary. The following three groups were included in this study: (1) ACPA-positive patients with RA, (2) their unaffected ACPA-positive FDRs, and (3) HCs negative for ACPA and RF. The demographics of the study groups are summarized in Table
1. RA diagnosis was made on the basis of fulfilling the 2010 American College of Rheumatology/European League Against Rheumatism classification criteria. None of the FDRs or HCs demonstrated clinical evidence of synovitis, as determined by a rheumatologist (HEG).
Table 1
Clinical characteristics of the study population
Age, years, median (range) | 40 (23–66) | 46.6 (29–70) | 33.65 (28–60) |
Sex, female/male | 10/2 | 14/4 | 12/1 |
Disease duration, years, median (range) | NA | 12.02 (0–35.6) | NA |
CRP titer, median (range) | 3.35 (1.07 –9.25) | 6.91 (2–42.6) | 2.595 (1.01–15.9) |
RF titer, IU/ml, median (range)ml | <20 | 321 (20–1540) | 34.9 (20–570) |
Anti-CCP titer, median (range) | 1 (0.4–2.0) | 201 (19–289) | 114 (7–365) |
BMI, kg/m2, median (range) | 29.54 (19.9–34.4) | 27.37 (20.4–39.6) | 26.09 (19.6–40.7) |
Sample collection
Venous blood was collected into PAXgene® Blood RNA tubes (PreAnalytiX, Hombrechtikon, Switzerland), processed as per the manufacturer’s instructions, and used to isolate total RNA. PBMCs were isolated using SepMate®-50 tubes (STEMCELL Technologies, Vancouver, BC, Canada) as per the manufacturer’s protocol. Briefly, venous blood was drawn into ethylenediaminetetraacetic acid-coated tubes and diluted 1:1 with incomplete Gibco RPMI medium (Life Technologies, Carlsbad, CA, USA), layered onto SepMate®-50 tubes with Histopaque Plus (Sigma-Aldrich, St. Louis, MO, USA), and centrifuged at 1000 × g for 10 minutes at room temperature. Buffy coat was separated, and cells were washed in RPMI 1640 medium prior to RNA isolation.
Immunoassays
Serum C-reactive protein (CRP) levels were monitored in serum by using a human high-sensitivity C-reactive protein (hs-CRP) enzyme-linked immunosorbent assay kit (Biomatik, Cambridge, ON, Canada) as per the manufacturer’s instructions. The concentration of ACPA was monitored in serum using the BioPlex® 2200 anticyclic citrullinated protein antibodies reagent kit (Bio-Rad Laboratories, Hercules, CA, USA).
Total RNA extraction and qRT-PCR
Total RNA was isolated from whole blood and PBMCs using the Ambion
mirVANA miRNA isolation kit (catalogue number AM1561; Life Technologies, Carlsbad, CA, USA) as per the manufacturer’s instructions. RNA quality was determined using Bioanalyzer with the RNA 6000 Nano Kit (Agilent Technologies, Santa Clara, CA, USA). Total RNA with absorbance at 260 and 280 nm ≥ 2.0 and RNA integrity number ≥ 7.0 was used for monitoring miRNA expression using a two-step qRT-PCR protocol as previously described [
15]. Briefly, we used the Applied Biosystems TaqMan® MicroRNA Reverse Transcription Kit (Life Technologies) with miRNA-specific stem-loop primers for reverse transcription (Additional file
1: Table S1). Specific amplification of miRNA targets was performed using TaqMan® Universal Master Mix II and target-specific TaqMan® MicroRNA Assay Mix in the ABI PRISM 7300 Real-Time PCR System (all from Life Technologies). For mRNA amplification, first-strand complementary DNA was synthesized from total RNA (1 μg) using SuperScript® VILO
TM MasterMix (Life Technologies) as per the manufacturer’s instructions. Target mRNA was amplified using Applied Biosystems®
Power SYBR® Green Master Mix (Life Technologies) as per the manufacturer’s instructions. Primers used for mRNA amplification of miR-103a-3p are listed in Additional file
1: Table S2.
Data analysis and statistics
Candidate endogenous control miRNAs for data normalization were selected on the basis of prior literature (RNU48, RNU44, U6 snRNA, RNU6B, and miR-16). Expression of these selected miRNAs was assessed for stable expression across samples in whole blood and PBMCs obtained from HCs, patients with RA, and FDRs. RefFinder, a web-based comprehensive gene analysis platform that integrates geNorm, NormFinder, BestKeeper, and comparative cycle threshold (Δ
C
t) methods, was used to identify the miRNA candidates suitable as endogenous controls for data normalization. On the basis of this approach, RNU48 and RNU6B were identified as optimum reference miRNAs for normalization across all samples in this study [
16]. Reference
C
t values for data normalization were determined by calculating the average
C
t value of RNU48 and RNU6B [reference
C
t = mean (
C
t {RNU48} −
C
t (RNU6B)] and used for each sample. Raw
C
t values for each target miRNA were then normalized with reference
C
t values to obtain Δ
C
t values for each sample [Δ
C
t (target miRNA) =
C
t (target) − reference
C
t]. Δ
C
t values of each miRNA were further corrected using a global mean normalization strategy to obtain normalized Δ
C
t values [normalized Δ
C
t = Δ
C
t (target miRNA) − mean Δ
C
t] for all assessed miRNAs [
17‐
19]. Relative fold changes were calculated using the ΔΔ
C
t method [
20]. Of the 35 miRNAs analyzed, 33 showed detectable expression (
C
t ≤ 35) (Table
2) and were considered for further analyses. Target mRNA expression was determined in samples after normalization using 18S ribosomal RNA as an endogenous control [
20], and relative fold changes were calculated using the ΔΔ
C
t method.
Table 2
Fold change expression of microRNAs
hsa-miR-103a-3p
|
0.0064
|
3.96
|
1
|
0.0238
|
7.68
| 0.1223 | 1.97 |
hsa-miR-155
|
0.0002
|
2.47
|
2
|
0.0115
|
1.98
| 0.3627 | −1.25 |
hsa-miR-29b | 0.0648 | 1.91 | 3 | 0.8636 | −1.55 | 0.0754 | −2.96 |
hsa-miR-132
|
0.0016
|
1.90
|
4
| 0.0687 | 1.37 | 0.2530 | −1.39 |
hsa-miR-26b-3p
|
0.0010
|
1.88
|
5
|
0.0024
|
2.28
| 0.2530 | 1.21 |
hsa-miR-152
|
0.0038
|
1.83
|
6
| 0.1309 | 1.97 | 0.4981 | 1.08 |
hsa-miR-19a
| 0.0732 | 1.73 |
7
|
0.0205
|
1.55
| 0.9662 | −1.11 |
hsa-Let-7a | 0.0569 | 1.73 |
7
| 0.2086 | 1.17 | 0.2358 | −1.47 |
hsa-miR-19b
|
0.0260
|
1.67
|
9
| 0.1169 | 1.34 | 0.4091 | −1.24 |
hsa-miR-146a-5p
|
0.0083
|
1.54
|
10
|
0.0031
|
1.99
| 0.3408 | 1.29 |
hsa-miR-451
|
0.0076
|
1.53
|
11
|
0.0127
|
1.56
| 0.7031 | 1.02 |
RNU44
|
0.0120
|
1.52
|
12
|
0.0162
|
1.30
| 0.9157 | −1.17 |
hsa-miR-125a-5p
|
0.0272
|
1.30
|
13
| 0.0553 | 1.50 | 0.7508 | 1.15 |
hsa-miR-222 | 0.0796 | 1.29 | 14 | 0.2154 | 1.20 | 0.4587 | −1.07 |
hsa-miR-107 | 0.0576 | 1.28 | 15 | 0.2481 | −1.02 | 0.4587 | −1.31 |
hsa-miR-29c | 0.2428 | 1.28 | 16 | 0.1457 | 1.11 | 0.8159 | −1.15 |
hsa-Let-7e | 0.0502 | 1.27 | 17 | 0.5837 | −1.20 | 0.0987 | −1.53 |
hsa-miR-21
|
0.0261
|
1.24
|
18
| 0.0924 | 1.22 | 0.7832 | −1.16 |
hsa-miR-223
|
0.0115
|
1.22
|
19
| 0.5487 | 1.03 | 0.5115 | −1.82 |
hsa-miR-26b-5p
| 0.1191 | 1.21 | 20 | 0.6744 | −1.45 |
0.0277
|
−1.76
|
hsa-miR-323-3p | 0.1123 | 1.17 | 21 | 0.2033 | 1.12 | 0.9831 | −1.04 |
hsa-miR-26a | 0.1849 | 1.14 | 22 | 0.1159 | 1.33 | 0.8822 | 1.16 |
hsa-miR-29a | 0.1031 | 1.14 | 23 | 0.6174 | −1.29 | 0.1124 | 1.46 |
hsa-miR-15a | 0.5594 | 1.11 | 24 | 0.8292 | −1.56 | 0.1223 | −1.72 |
hsa-miR-150 | 0.6104 | 1.10 | 25 | 0.41 | 1.20 | 0.7669 | 1.09 |
hsa-miR-34a*
| 0.1065 | 1.06 | 26 |
0.0495
|
−1.84
|
0.0262
|
−1.94
|
hsa-miR-221 | 0.1842 | 1.06 | 27 | 0.9914 | −1.37 | 0.2356 | −1.45 |
hsa-miR-24 | 0.2956 | −1.05 | 28 | 0.8651 | 1.03 | 0.9831 | 1.04 |
hsa-miR-18a | 0.7615 | −1.07 | 29 | 0.7639 | −1.85 | 0.1124 | −1.72 |
U6
| 0.8069 | −1.11 | 30 |
0.0001
|
−1.59
| 0.0625 | −1.43 |
hsa-miR-125a-3p | 0.1654 | −1.20 | 31 | 0.6178 | −1.10 | 0.4847 | 1.10 |
hsa-miR-16 | 0.9950 | −3.94 | 32 | 0.8276 | −1.24 | 0.2802 | −1.23 |
hsa-miR-346
|
0.0001
|
−8.70
|
33
|
0.0001
|
−20.00
|
0.0338
|
−2.32
|
GraphPad Prism version 5.0 was used for miRNA analysis and generating volcano plots, scatterplots, and bar graphs. Empirical cumulative distribution plots (based on the Kolmogorov-Smirnov [KS] test) and ROC curves were generated using MS Excel (Microsoft, Redmond, WA, USA) and Prism (GraphPad Prism, La Jolla, CA, USA) software, respectively. The KS test is a nonparametric statistical method that does not assume normal distribution [
21]. Differences between the datasets were represented as KS scores (in the range of −1 and 1) corresponding to maximum degree of separation between the cumulative distributions of the datasets being compared and directly proportional to relative expression levels. KS scores > 0.5 were considered significant. Heat maps were generated with unsupervised hierarchical clustering using the TIGR multiple experiment viewer. Ingenuity Pathway Analysis ([IPA]
www.ingenuity.com; QIAGEN Bioinformatics, Redwood City, CA, USA) was used for biomolecular network analyses and to predict mRNAs targeted by the differentially expressed miRNAs identified in this study. The Mann-Whitney
U test, the Kruskal-Wallis test with Dunn’s post hoc method, or Spearman’s rank correlation coefficient analysis was used for statistical analysis as required, and
P values < 0.05 were considered significant. Differentially expressed miRNAs were determined after adjusting
P values with Benjamini-Hochberg correction for multiple comparisons [
22].
Discussion
In the present study, we examined the expression pattern of a wide spectrum of miRNAs in whole blood samples from a cohort of INA patients with RA, their ACPA-positive unaffected FDRs, and unaffected INA control subjects with no clinical or serological evidence of autoimmunity. We demonstrated distinct differences between all three groups, and to our knowledge, we are the first to demonstrate that miR-103a-3p is overexpressed in patients with RA and FDRs compared with HCs. Although aberrant miRNA expression patterns in the peripheral blood of patients with RA has been widely reported [
13,
30], aberrant expression in ACPA-positive unaffected individuals has not been reported previously. This study provides an impetus for evaluating the whole blood miRNA profile, particularly miR-103a-3p expression, as a potential biomarker for predicting imminent disease in individuals at risk for developing RA. It also points to specific biological pathways that may be involved in the transition to clinically detectable disease.
We elected to examine miRNA profiles using whole blood samples collected in PAXgene® RNA tubes for several reasons. First and foremost is the ease with which these samples are collected and stored, along with the remarkable resistance of the miRNA to endogenous ribonuclease activity, as well as stability to extreme pH, temperature, and storage conditions [
31,
32]. An alternative approach that is being widely investigated in a spectrum of chronic diseases is testing miRNA levels in serum or plasma [
33,
34]. Although this latter cell-free approach has the advantage of potentially harnessing large archival serum/plasma sample repositories, it suffers from limitations in providing a complete and unbiased miRNA profile of the circulating peripheral blood compartment of an individual. This relates to factors such as preprocessing of samples, cellular contamination, and inconsistency in miRNA levels in serum vs plasma [
35,
36].
One major advantage of using whole blood to determine miRNA levels is that this approach retains the rich compositional architecture of the circulating blood, thus providing the most unbiased representation of this space. Although this approach may be ideally suited for biomarker discovery, its primary disadvantage is the inability to define the cellular subsets that are contributing to the observed miRNA profiles. Combined with the marked cellular compositional heterogeneity of whole blood, the generation of mechanistic hypotheses is challenging. To address this challenge, most of the previous studies of circulating miRNA expression in RA have been focused on PBMCs and their subsets [
13]. However, attempts to correlate PBMC expression patterns with those evident in whole blood have produced conflicting results [
15]. For instance, Atarod et al. demonstrated discordant expression of miR-146a-5p and miR-155 expression between PBMCs and whole blood [
37]. These findings contradict the findings of our previous study [
15], which demonstrated more concordance between whole blood and PBMC expression patterns. These differences may be attributable to total RNA isolation methodology used in each of these sample types. Alternatively, this discordance can also be attributed to blood cell counts and red blood cell hemolysis [
38]. We acknowledge the absence of such information pertaining to our study participants.
Previous studies on miRNA expression in RA, including our own, have been focused on differences in miR-146a and miR-155 expression between patients with RA and unaffected control subjects, both tending to be increased in RA PBMCs and synovial tissues [
13,
15,
39]. In the present study, we compared the expression levels of these two miRNAs in whole blood and PBMCs and found that the levels were concordantly elevated not only in patients with RA as previously documented but also, surprisingly, in ACPA-positive FDRs with no clinical evidence of arthritis. Moreover, as shown in Fig.
1, the overall miRNA expression patterns in patients with RA and ACPA-positive FDRs were relatively similar to those of unaffected control subjects. These observations suggest that the similarity between patients with RA and unaffected ACPA-positive FDRs in the peripheral blood miRNA profile is more likely to relate to autoimmune than to inflammatory mechanisms. Moreover, we demonstrated that these patterns are relatively stable over a short time frame. It will be of interest to determine how the miRNA patterns evolve as individuals at risk for developing RA transition to clinically detectable synovitis. It will also be of interest to determine whether these RA-associated patterns are retained in patients with RA who have achieved clinical remission.
The large difference in miR-103a-3p expression that discriminated both patients with RA and FDRs from unaffected, population-based control subjects is noteworthy and, to our knowledge, not previously reported. Located within the intronic regions of pantothenate kinase enzymes, miR-103a-3p is a member of the miR-15/107 cluster and regulates lipid, cholesterol, and fatty acid metabolism; adipocyte differentiation; and insulin signaling [
40‐
42]. However, the potential role that these biological functions play in RA pathogenesis remains largely unknown. Some studies have suggested that miR-103 upregulation is associated with obesity and insulin resistance in liver and adipose tissue, as well as with atherosclerosis [
24,
43,
44]. Interestingly, the indigenous First Nations population as a whole, including the cohort we have studied, demonstrates a strikingly high prevalence of obesity, type 2 diabetes, and cardiovascular disease [
45,
46].
To identify potential gene targets of miR-103a-3p and delineate the biological functions that they regulate, we performed computational predictive analysis using IPA. On the basis of the curated IPA target network analysis, we identified TP53 and AGO2 as central nodes in miRNA patterns detected in patients with RA and ACPA-positive unaffected FDRs. AGO2 is an integral component of RNA-induced silencing complex (RISC) that cleaves double-stranded immature miRNAs to single-stranded mature forms, a reaction catalyzed by an RNase III-type enzyme called Dicer [
9]. Altered TP53 expression has been observed in lymphocytes and synovial tissues from patients with RA and is associated with synovial proliferation and increased proinflammatory IL-6 secretion in the synovium [
47,
48]. Interestingly, miR-103a-3p associates with AGO2 within RISC and is known to suppress Dicer [
24,
49]. TP53 also regulates miR-103 expression via targeting components of miRNA biogenesis, including DICER1 and AGO2 [
50]. Together, our observations point to miR-103a-3p-associated miRNA target reorganization in patients with RA and ACPA-positive FDRs at risk for developing RA. It is notable that the regulatory networks of miRNAs, including miR-103a-3p and its target mRNAs, are extremely complex and known to control physiological processes at multiple levels [
51]. Considering that miRNAs are involved in an intricate network of feedback and feedforward regulatory loops, it is likely that the target mRNAs monitored in our study may modulate the expression of other miRNAs [
52,
53]. In this regard, further research is warranted to investigate the interaction network between miR-103a-3p and its target mRNAs in different cohorts, especially FDRs, to examine biological processes prior to onset of RA.