Autoimmune gene expression profiling of fingerstick whole blood in Chronic Fatigue Syndrome

This study protocol was approved by the Western Institutional Review Board and divided into three phases: patient recruitment, sample collection and testing & analysis. Figure 1 depicts the study design and its workflow. During the patient recruitment phase, a protocol was written to allow for direct-to-participant (D2P) study implementation. The study protocol and supporting documents were submitted to a central IRB for ethical review and approval. The study originally aimed to collect samples at a single timepoint from 100 CFS patients and 100 healthy control participants, followed by samples from a second timepoint from 10% of each group to assess sampling repeatability. During the sample collection phases, subjects were recruited, qualified, consented, characterized, and sampled per the study protocol. During the testing & analysis phase, whole blood fingerstick samples were tested using the autoimmune profile (AIP) test at DxTerity (Rancho Dominguez, CA). A portion of the whole blood samples was extracted to yield DNA which was tested for viral infection using a real time PCR method at Coppe Laboratories (Waukesha, WI). The urine samples were tested for organic acid profiles at Genova Diagnostics (Asheville, North Carolina). The urine data is not the subject of this report.

Upon IRB approval, participants were recruited directly through online advertising, patient advocates and advocacy groups, registries, clinical sites, and physician referral. CFS participants were able to register for the study in late May 2021 through the study specific website thechromestudy.com. Enrollment kicked off in June with invitations sent to registered participants. The website allowed potential participants to sign up and complete an initial qualification before receiving an email linking to a secure portal for further Qualification and eConsent, including identity verification. No other identifying information was stored by the study until consent was obtained. Informed consent was conducted electronically through a HIPAA compliant mobile app, where participants were able to self-report demographics and medical information. All subject-pertaining materials were reviewed by the IRB. Next, participants (male and female age 18 or older at the time of consent) completed a single study collection event consisting of one Micro Collection Device (MCD) for fingerstick blood and one first morning void urine collected from home. 20% of the ME/CFS and 19% of the control group completed a second study collection event approximately 3 to 4 weeks after the initial study collection to assess sampling repeatability. For the Disease Group, patients must have been either diagnosed with ME/CFS by their physicians, or met CFS definitions (per 2015 Institute of Medicine (IOM) diagnostic criteria for ME/CFS definition) [21] as indicated on the qualification questionnaire. For the Control Group, participants were eligible if they did not have history of fatigue and did not meet the ME/CFS definition [21].

Once enrolled in the study, participants were provided access to a study portal where they could access and complete the electronic patient reported outcome (ePROs) questionnaire. Participants were asked to self-report and complete ePROs about the following:

This study was approved by the Institutional Review Board and all patients gave written, informed consent.

Each qualified participant received a DxCollect® MCD Fingerstick Kit and a Genova Diagnostics® Urine Collection Kit and was also asked to complete a study-related questionnaire. This was an observational study and participation did not affect any aspect of patient treatment.

The MCD blood collection kit was used to facilitate the collection, stabilization, and shipping of a microsample (about 120 μL) of blood for non-clinical use. The device was designed to collect fingerstick blood samples autonomously. The MCD kit was shipped to the subject’s home. The completed kit was returned to the lab for analysis in a pre-labeled, postage-paid box according to applicable federal regulations. The MCD sample was stable at ambient temperatures for up to 14 days.

A urine sample was collected in a 15 mL tube pre-coated with a stabilizer containing a preservative. Upon mixing after urine collection, the urine sample was frozen down at − 20 °C or less before shipment was made within 24 h of collection. Frozen urine from the participant’s home was shipped following manufacturer’s instructions for Genova Diagnostics metabolomic testing. Urine samples that did not conform to shipping and/or temperature requirement were rejected and destroyed by the lab per standard operating procedures at the time the sample was received.

Computational data analyses were performed on the resulting data sets generated from the RNA and the DNA tests to discover potential genes associated with ME/CFS and/or sub-segmentation of ME/CFS patients based on disease severity and autoimmune disease comorbidities.

The following diagram (Fig. 2) depicts the data analysis workflow for AIP gene expression data.

The first participant enrollment was received on the same day the study invitation was sent out and the first sample was received just 8 days after the study invitation. The ME/CFS arm was fully enrolled within 3 months, and the control arm was fully enrolled within 4.5 months. A total of 300 participants were enrolled, with only 3 subjects having withdrawn from the study after the samples were collected. In the study, we included 267 valid AIP results, after excluding QC fails and disqualified samples. Participant questionnaire was completed by 287/297 (96.3%) of the total participants. Ten participants who did not complete the questionnaire were equally split between the CFS group and the control group. Results were compiled using a de-identified participant ID in compliance with HIPPA regulations.

A total of 166 CFS and 83 healthy participants (controls) were recruited for the study. From the participants who completed the questionnaire, there were 125 females (75%) and 38 males (23%) in the ME/CFS group and 44 females (53%) and 34 males (41%) in the control group (Table 1). The average age was 50.8 in the ME/CFS participants and 44.8 in the controls (p = 0.001, Table 1). Average Body Mass Index (BMI) was very similar at 26.8 in the ME/CFS patients and 25.8 in the controls (p = 0.522, Table 1). Based on BMI standard weight status categories, similar percentage of the ME/CFS participants and the controls fell into underweight, healthy weight, overweight and obesity categories (Table 1) with no statistical differences (p > 0.05). Ninety-two percent (92%) of the ME/CFS participants and fifty-two (52%) of the controls were Caucasian. A much smaller percentage of both the groups were Asians, Hispanics, African Americans, and other.

Of the 166 ME/CFS participants, ninety-four percent (94%) reported receiving a diagnosis by a health care professional (Table 2). Three percent (3%) were self-assessed based on descriptions provided in the questionnaire as they considered themselves to have the symptoms listed in the 2015 IOM diagnostic criteria for ME/CFS (Table 2). The other three percent (3%) did not provide information, but they were placed in the ME/CFS group since these individuals identified as ME/CFS specific when enrolling in the study (Table 2). Similarly, out of 83 normal participants, ninety-four percent (94%) of the subjects enrolled as healthy normal controls based on the exclusion criteria (Table 2). Six percent (6%) did not provide information but were placed in the normal control group since these individuals identified as normal when enrolling (Table 2).

Onset of CFS symptoms was reported by the participants. Duration of the CFS disease was calculated and reported (Table 2): 17% of the ME/CFS participants reported to have CFS between 1 and 5 years, 18% between 6 and 10 years, 13% between 11 and 15 years, 12% between 16 and 20 years, 21% between 21 and 30 years, 16% between 31 and 63 years, and 2% did not report.

Approximately half (49%) of the ME/CFS patients reported being house-bound or bedridden due to the symptoms of the disease (Table 2). In this study, we considered house-bound or bedridden patients as having severe symptoms, and the rest as having mild to moderate symptoms [7].

Seventy-six percent (76%) of the ME/CFS patients reported a triggering event before the onset of the disease. Among them, eighty percent (80%) reported an assignable viral or bacterial infection as the trigger, while twenty percent (20%) reported other physical and/or emotional traumatic events as the trigger (Table 2). Among the viral and bacterial infections, organisms listed in the questionnaire included: EBV, Mononucleosis, Flu, H. Pylori, Pneumonia, Strep Throat and COVID (Table 2). In addition, thirty percent (30%) of the ME/CFS patients have or have had autoimmune diseases (e.g. Hashimoto's thyroiditis, Fibromyalgia, IBD, POTS, Grave’s Disease, Crohn’s disease, Psoriasis, Rheumatoid arthritis and Sjogren's syndrome). Thirteen percent (13%) have or have had various cancers. Five percent (5%) have or have had kidney diseases (Table 2). Eighty-five percent (85%) of the CFS patients were taking medications (currently or within the last 4 weeks), including Antibiotics, Antihistamines, Immunomodulatory, and Non-Steroidal Anti-Inflammatory drugs, comparing to 36% of the normal controls listing similar medications (not listed).

MCD blood samples were collected from 163 CFS and 79 control participants at the first timepoint, and samples were also collected from 33 CFS and 15 control participants at the second timepoint (~ 19% of each group). The second timepoint samples were collected voluntarily approximately 4 weeks after the first time point. The purpose for testing the second time point was to verify there was no sampling bias between the collections. Together, the AIP testing was completed for a total of 287 MCD samples (out of 240 participants, including first and second timepoint collections). Out of the 287 tests, 267 tests (93.0%) passed QC acceptance.

The testing was carried out in 2 distinct batches. To assess potential batch effects, both batches were combined and principal component analysis (PCA) was performed to visualize clusters of normalization genes and immune module genes separately. Replicate samples displayed a high Spearman’s rank correlation coefficient (r = 0.86, p < 2.2e−16), but multivariable regression models revealed a downward shift in mean unique to timepoint. COMBAT [24] empirical bayes standardization resolved this deviation, suggesting this relatively small batch effect was technical in nature.

When examining concordance between the samples collected at two timepoints, the scatter plot in Fig. 3a showed a good correlation (r = 0.86) between the timepoint 1 and 2 samples. Each gene signal (raw RFU) is normalized against the geometric mean of the 3 house-keeping genes (ACTB, GAPDH, TFRC). When examining the 3 house-keeping genes between the 2 batch data sets, the PCA plot showed no observable difference in house-keeping genes by the 2 data sets, therefore there is no batch effect in sample testing (Fig. 3b). The AIP results in the subsequent analyses were based on the first time point (excluding repetitive samples from the same participants), in which all samples were collected from unique participants of either the CFS patients or the normal controls.

From the CFS cohort, unsupervised consensus clustering was conducted to reveal a heat map showing the relationship between genes significantly up- or down-regulated in patient subsets with certain demographic or clinical attributes, such as age, gender, disease duration, prior viral infection, bedridden status, autoimmune comorbidity, and immunomodulatory medicine (Fig. 4). Six major clusters were unveiled after merging several candidate clusters (labeled as m01–m06 in the heatmap). These groupings confirm categorical gene expression modulation also reported from prior regression testing, such as enrichment in IFIT1, ISG15, and IFI6 in female patients and higher expression of APOL6 and SLC2A3 in patients 55 years and older. In addition, there were also visible expression differences which may be explained by unreported factors from patients.

Next, we analyzed genes that were differentially expressed between two subsets of the CFS patients: self-reported bedridden patients (n = 72) and self-reported non-bedridden patients (n = 72). The bedridden patients were considered to have severe disease. Regression bootstrapping was performed on the genes which were borderline modulated between CFS bedridden and CFS non-bedridden participants per linear models controlling for demographic and clinical confounders such as age, sex, prior infection, and symptom duration. This analysis revealed six genes with significant (p < 0.05) differential expression by bedridden status: IKZF2, IKZF3, ABCE1, BACH2, CD3D and HSPA8 (Table 3). A forest plot showed the coefficient means and 95% confidence intervals for each gene in Fig. 5. A bootstrapped multivariable model on bedridden status was performed controlling for age and sex. The genes most associated with bedridden status when controlled for age and sex were HSPA8, IKZF2, and IKZF3 (data not shown).

A subset of CFS patients also had an autoimmune etiology [16]. In this study, 30% of CFS participants reported having one or more autoimmune diseases that co-exist with CFS. For these 44 patients, we compared gene expression levels of the 6 genes between the 22 bedridden and the 22 non-bedridden patients. Each of the 6 genes exhibited a greater separation in this subset of CFS patients with other autoimmune diseases (Fig. 6), suggesting an additive effect of fatigue related gene expression in CFS and other autoimmune etiologies.

The MCD blood samples were also used to extract DNA which was tested for several viral infections using a real time PCR method at Coppe Laboratories, including HHV-6A and HHV-6B, HHV-7, EBV and CMV. Among the 240 participants tested with the viral assays, 9 samples showed positive results (Table 4). One CFS sample was positive for EBV, and 8 were positive for HHV6 (including 7 CFS samples and 1 normal). The p-value between the CFS and control groups was p = 0.074. Among the HHV6 positives, further analysis showed 4 of them were with the HHV6B genotype, and 1 of them was with the HHV6A genotype.

In this study, we took a social media-based patient centric recruitment and at-home sample collection approach. Direct-to-participant (D2P) recruitment and remote collection of samples allowed potential enrollment of a broader study population than was not easily achievable through a conventional site-based model. Patient advocates, such as social influencers, were contacted to share resources and aid in the recruitment of participants. Advocates and influencers were also invited to enroll and participate in the study to better understand the study experience and provide context for their shareable content. D2P participants interacted with the study through a study-specific app that was available for both iOS and Android operating systems. The app provided an end-to-end solution for deploying, completing, and managing the D2P study activities.

The D2P approach was advantageous in that it enabled participants to collect samples from geographically diverse locations in the US without the need to set up multiple clinical sites. Participants did not need to visit a physician’s office and were able to self-collect samples at home. This was particularly convenient for CFS patients with severe symptoms who are house- or bed-bound. The MCD fingerstick whole blood collection device allows self-collection of small amounts of blood. The participants were able to simultaneously collect and ship blood and urine samples needed for conducting multiple tests. This approach enabled the study to be conducted at a relatively low cost when compared to the anticipated costs for a site-based approach and allowed faster turn-around time. Some limitations of the approach include: 1) Self-reported patient medical history is not verified. 2) Discarded/missing samples due to improper collection procedure by the participants could lead to loss of samples. In this study, sample attrition rate was 3% for MCD blood samples and 8% for urine samples, which were either discarded or not received. The success rate for sample collection was 92% for urine and 97% for MCD.

CFS is a complex clinical condition of unknown etiology, characterized by persistent or intermittent fatigue that is not the result of recent exertion and does not improve with rest, resulting in a significant reduction in the patient's previous normal activity. CFS is a multi-system disease, with altered immune, musculoskeletal, endocrine, neurological and cardiovascular systems. In this study, we have identified 6 genes using a molecular profiling approach. These genes (IKZF2, IKZF3, ABCE1, BACH2, CD3D and HSPA8) have been implicated in various biological functions, particularly in T cell and B cell biology.

The Ikaros zinc-finger family transcription factors (IKZF TFs) are important regulators of lymphocyte development and differentiation and are also highly expressed in B cell malignancies and are required for cancer cell growth and survival [25‐27]. Moreover, IKZF TFs negatively control the functional properties of many immune cells. Specifically, IKZF3 plays an essential role in regulation of B-cell differentiation, proliferation, and maturation to an effector state. It is involved in regulating BCL2 expression and controlling apoptosis in T-cells in an IL2-dependent manner. Diseases associated with IKZF3 include Immunodeficiency 84 and Chronic Lymphocytic Leukemia. CD3D is a T-cell receptor/CD3 complex and is involved in T-cell development and signal transduction. IKZF2, another member of the Ikaros TFs, controls lymphocyte development, promotes quiescence, maintains the inhibitory function of regulatory T cells, and is frequently deleted in hypodiploid B-acute lymphoblastic leukemias (B-ALLs) [28]. BACH2 is a basic leucine zipper transcription factor expressed in B cells from the pro-B cells to mature B cells and is downregulated during the maturation to plasma cells. BACH2 is involved in primary adaptive immune response involving T cells and B cells and enables sequence-specific double-stranded DNA binding activity [29]. BACH2 is essential for the differentiation of stem-like CD8+ T cells during chronic viral infection. Overexpression of BACH2 upregulates IKZF2 gene [30].

IKZF3 (Aiolos) and IKZF2 (Helios), and BACH2 have been implicated in other autoimmune disease such as systemic lupus erythematosus (SLE) and Rheumatoid Arthritis (RA) [25, 31‐36]. In this study, the expression of these genes was increased in severe cases of CFS and greater upregulation was observed in those who suffer from comorbid autoimmune diseases, suggesting that CFS is, in part, an autoimmune disease. These genes can be utilized as potential biomarkers to distinguish mild CFS from severe CFS with autoimmune comorbidity. Furthermore, accumulating evidence demonstrated that downregulation of IKZF3 (Aiolos) and IKZF1 (Ikaros), two members of the IKZF family, in malignant plasma cells as well as in adaptative and innate lymphocytes is key for the anti-myeloma activity of Immunomodulatory drugs (IMiDs) [37]. In addition, IKZF1, IKZF3 and IKZF2 (Helios) have been implicated in SLE pathogenesis. There is strong evidence that therapeutic targeting of Ikaros and Aiolos can ameliorate key pathogenic processes in human SLE [37].

CFS is characterized by fatigue and severe disability. Besides fatigue, certain aspects of immune dysfunction appears to be present in both CFS and cancer [38]. The underlying cause of CFS is unknown due to its heterogeneity, but in many cases, it is thought to be triggered by an abnormal immune response to an agent, such as a viral infection, that results in chronic immune activation. The immunologic changes in CFS and its possible relationship with infection have prompted investigators to consider whether CFS could also be associated with an elevated risk of cancer. A 2013 study has reported that CFS was present in 0.5% of cancer cases among US elderly and 0.5% of controls [39]. CFS was associated with an increased risk of non-Hodgkin’s lymphoma (NHL) and specifically NHL subtypes diffuse large B cell lymphoma, marginal zone lymphoma, and B-cell NHL not otherwise specified (NOS) [39]. In a recent clinic-based study, with a carefully selected matched cohort of 59 confirmed CFS and 54 control cases, the authors have identified statistically significant increased risks of autoimmune disease and cancer among the first-degree relatives of ME/CFS cases [17].

IKZF2 is highly expressed in leukemic stem cells (LSCs) and its deficiency results in defective LSC function. IKZF2 has been shown to drive leukemia stem cell self-renewal and inhibits myeloid differentiation. Regulation of the AML LSC program by IKZF2 thus provides a rationale to therapeutically target IKZF2 in myeloid leukemia [28].

Abnormalities in ribonuclease (RNase) L and hyperactivation of nuclear factor kappa beta (NF-κB) are present in CFS and in prostate cancer [40]. One of the key antiviral effectors is the IFN-inducible oligoadenylate synthetase/ribonuclease L (OAS/RNase L) pathway, which is activated by double-stranded RNA to synthesize unique oligoadenylates, 2-5A, to activate RNase L. RNase L exerts an antiviral effect by cleaving diverse RNA substrates, therefore limiting viral replication; many viruses have evolved mechanisms to counteract the OAS/RNase L pathway [40]. ATP-binding cassette E1 (ABCE1) transporter, identified as an inhibitor of RNase L, regulates RNase L activity and RNase L-induced autophagy during viral infections [40]. In this study, we identified ABCE1 as one of the six genes upregulated in severe CFS cases. This suggests the RNase L pathway may be impaired in these individuals due to elevated ABCE1 expression.

HSPA8 is implicated in a signal transduction pathway in the abnormal proliferation of chronic myeloid leukemia (CML) cells, suggesting that the chaperone HSPA8 and CCND1 contribute to the abnormal behavior of CML cells and represent an interesting target for new therapies [41].

Clinical activity from B-cell depletion using anti-CD20 antibody rituximab has been utilized in treating CFS patients [42]. However, improvement of fatigue was observed in the responders 3 to 7 months after the treatment and the initial rapid B-cell depletion. This data suggested that CFS may be an autoimmune disease, and the delayed response to rituximab could be due to the elimination of the disease-associated autoantibodies. In a separate study, three patients, with 7–10 years of CFS disease duration, had substantial relief of all symptoms related to CFS after rituximab intervention. Overall, there was a major recovery of all symptoms lasting until 40 weeks after treatment, followed by a gradual returning of symptoms. CFS may be amenable to therapeutic interventions aimed at modifying B-cell number and function [43]. It is also speculated that this response pattern could be related to interaction of B-cell and T-cell in antigen presentation. Our finding of up-regulation of both B-cell and T-cell associated genes support this hypothesis.

In this study, we investigated the expression of genes specific to key immune functions in ME/CFS patients. The six genes we identified are strongly related to T cell and B cell functions. The results could hopefully shed some light in further exploring the biological connections between CFS, autoimmune disease and cancer, with an aim of identifying possible therapeutics of the disease. Research focusing on etiology and pathogenesis of CFS could help unravel biological roots of the disease. Additionally, benefits of discriminating differentiated symptoms of CFS from autoimmune and cancer could prove valuable in searching for immunity and oncogenic related therapeutics.

Before the pandemic, estimated 1.5–3 million Americans were suffering from CFS in US alone. An unprecedented wave of CFS-like illness related to COVID infection will appear over the next few years, with profound societal costs. The striking similarities between CFS and long COVID suggest commonality in the disease mechanisms may exist. Studies on the mechanism and treatment of CFS will facilitate research on the long COVID.