Maureen R. Hanson, Susan M. Levine, Arnaud Germain
Abstract
Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) is a disabling disease with worldwide prevalence and limited therapies exclusively aimed at treating symptoms. To gain insights into the molecular disruptions in ME/CFS, we utilized an aptamer-based technology that quantified 4790 unique human proteins, allowing us to obtain the largest proteomics dataset yet available for this disease, detecting highly abundant proteins as well as rare proteins over a nine-log dynamic range. We report a pilot study of 20 ME/CFS patients and 20 controls, all females. Significant differences in the levels of 19 proteins between cohorts implicate pathways related to the extracellular matrix, the immune system and cell–cell communication. Outputs of pathway and cluster analyses robustly highlight the ephrin pathway, which is involved in cell–cell signaling and regulation of an expansive variety of biological processes, including axon guidance, angiogenesis, epithelial cell migration, and immune response. Receiver Operating Characteristic (ROC) curve analyses distinguish the plasma proteomes of ME/CFS patients from controls with a high degree of accuracy (Area Under the Curve (AUC) > 0.85), and even higher when using protein ratios (AUC up to 0.95), that include some protein pairs with established biological relevance. Our results illustrate the promise of plasma proteomics for diagnosing and deciphering the molecular basis of ME/CFS.
View Full-Text
Keywords: ME/CFS;
proteomics;
plasma;
ephrin-Eph pathway;
immune metabolism;
adherens junction;
glucose;
SOMAscan®;
diagnosis
Discussion
The plasma proteomics dataset investigated here is remarkable for ME/CFS as it allows an extensive probing of protein abundance differences between 20 ME/CFS patients compared to 20 healthy controls. The quantification of plasma proteins whose abundances vary over nine orders of magnitude, achieved by the SOMAmer technology, segregates highly abundant proteins such as albumin, globulins and fibrinogen, which typically account for up to 99% of blood proteins, from the numerous other proteins of lower abundance, which account for another 4773 unique proteins in our dataset. Such a large dataset coupled with an exploratory sized cohort calls for concessions in the analysis and interpretation of results, a balance we have tried to achieve by combining stringent with more relaxed statistical analyses as well as some exploratory tools such as pathway and correlation analyses.
4.1. Most Significantly Different Proteins Are More Abundant in the Plasma of ME/CFS Patients
The classic statistical approach, using a Wilcoxon test and multiple testing correction, highlighted nine proteins at a low false discovery rate (FDR) of q < 0.05, with one group linked to cellular structure through the cytoskeleton and the extracellular matrix, and a second group linked to the immune system. Of the three proteins related to cellular structure, AIF1L is involved in the cytoskeletal apparatus as an actin-bundling protein [
24]. Al-though lower levels of AIF1L have been linked to poor prognosis during breast cancer, AIF1L overexpression in a cell line, similar to what is observed for our ME/CFS cohort (
Table 2 and
Figure 1), suppressed cell spreading and altered cell shape [
25]. The same study predicted a potential role of AIF1L in tight junctions, cell junctions, and focal adhesion, while showing that its overexpression was correlated with decreased RhoA expression, something we also observe in our ME/CFS patient cohort (
Figure S2). However, FAK1 (focal adhesion kinase 1) was not concomitantly reduced and instead slightly higher in our patient cohort, short of any further comparison to breast cancer. The human protein atlas (HPA,
www.proteinatlas.org) puts most of AIF1L proteins in the kidney and urinary bladder, where it is involved in the stabilization of podocyte morphology and focal adhesions through the actomyosin machinery [
26]. AIF1L is also present in many other organs, including male and female reproductive tissues, brain, lungs, the digestive tract, the skin, as well as adipose and soft tissues.
High expression of the extracellular matrix protein PXDN has been linked to proliferation, invasion, and migration of ovarian cancer cell lines, potentially causing an association with poor prognosis in ovarian cancer [
27]. PXDN contains a heme-peroxidase domain that allows for its crosslinking activity of collagen IV, a structure crucial to basement membrane synthesis [
28]. The basement membrane provides cell and tissue support and acts as a platform for complex signaling. The review article by Peterfi and Geiszt [
28] also points to PXDNL, the peroxidasin-like protein homolog to PXDN, exclusively localized in the cell–cell junctions of cardiomyocytes. PXDNL is an antagonist to PXDN when associated in a complex and while our dataset shows higher level of PXDN in the blood of ME/CFS patients, it also displays less PXDNL (p = 0.02 and q = 0.5,
Supplementary File 1). PXDN has been shown to be increased in the heart after cardiac stress or myocardial infarction after activation by elevated levels of TGF-β1 [
29].
The third protein, MXRA7, is ubiquitous to many organs including the brain, endocrine tissues, lungs, the pancreas, male and female tissues, as well as muscle tissues and the skin. It is hypothesized that MXRA7 is involved in injury recovery, neovascularization and wound healing [
30]. A more recent study shows a high expression of MXRA7 in the basal layer of the human epidermis. The absence of the protein in the mouse model leads to a skin disorder similar to psoriasis, indicating its inhibitory role in skin proliferation [
31]. MXRA7 is activated by a few proinflammatory Th1/Th17-type cytokines [
32,
33]. Other groups have described MXRA7 as necessary to alleviate acute liver injury [
34] and a regulator of stem cell differentiation in bone marrow [
35].
The four proteins related to the immune system included TIMD3, with high level of protein in bone marrow and lungs, according to HPA; IL-18 BPa, also with high levels in bone marrow; Ephrin-A4 and TNF sR-I seem ubiquitous to all organs. If we focus on the blood atlas from HPA, three of them (TIMD3, TNF sR-I and IL-18 BPa) are secreted into the blood and are present at high levels in almost all immune cells, except TIMD3 and TNF sR-I are not detected in B-cells. In contrast, none of the proteins in the cellular structure group are thought to be regularly secreted into the blood.
TIMD3 is highly expressed in T-cells and myeloid dendritic cells, implicating a role in both innate and adaptive immunity, with loss-of-function mutations leading to autoimmune disorders in 20% of patients, resulting from uncontrolled immunological activation [
36]. Additionally noted by Dixon et al. [
36] is the high expression of TIMD3 in ‘exhausted’ T-cells in cancers and chronic viral infections, although the associated upregulation of the expression of PD-1 (programmed cell death protein 1) was not observed in our patient cohort (p = 0.6 and q = 0.9,
Supplementary File 1). Nevertheless, high TIMD3 expression inhibits effector T-cell responses and the infiltration of T lymphocytes in adipose tissues, partially increasing the inflammation around adipose tissues in patients [
36].
TNF sR-I dysfunction has mainly been described as resulting from mutations disrupting the activity of the protein, leading to the autosomal dominant autoinflammatory syndrome known as TNFR1-associated periodic syndromes (TRAPS) [
37]. The underlying mechanism specifically impacts CD4+ conventional T-cells and Tregs, leading to inflammation [
38]. Only membrane bound receptors play a role in cell survival, apoptosis, and inflammation, while its soluble form can capture free tumor necrosis factor alpha (TNF-α), henceforth inhibiting inflammation. The level of TNF ligand is not different between cohorts in our dataset (p = 0.8, q = 1,
Supplementary File 1).
IL-18 BPa prevents the binding of the proinflammatory cytokine IL-18 to its receptor, and in the process reduces both T-helper type 1 and 2 immune responses and overall inflammatory response [
39]. IL-18 levels are comparable in our dataset between the control cohort and ME/CFS patients (p = 0.8, q = 1,
Supplementary File 1), which separates this cohort from Crohn’s disease patients, where both IL-18 and IL-18 BPa are higher in patients compared to healthy controls [
40]. The increase in IL-18 BPa was not the result of increased IFNγ [
41], which is slightly lower in the ME/CFS cohort (p = 0.1, q = 0.7,
Supplementary File 1). IL-18 BPa is strongly upregulated during inflammation, including malignancies such as cancer but also by some viral infections, with some viruses having evolved the capability to express active viral forms of IL-18 BPa [
41].
While none of the other proteins highlighted in
Table 2 have ever been linked to ME/CFS in plasma, cytokines such as TNF-α and IL-18, and by association the proteins involved in their metabolism, have been the focus of several studies [
4,
14,
42]. Unfortunately, studies of different cohorts of ME/CFS cases and controls from different investigators are divergent in findings concerning cytokine levels.
