Background
Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a life-limiting illness characterized by persistent debilitating fatigue, cognitive dysfunction often referred to as “brain fog”, pain/myalgias, post-exertional malaise, autonomic disturbances and gastrointestinal and immune system dysfunction lasting for at least 6 months, but usually much longer as prognosis for recovery is poor [
1]. This symptomatology has led to the suspicion that ME/CFS is an inflammatory disorder; therefore, researchers have been investigating potential biomarkers including, oxidative stress [
2,
3], energy metabolism [
4,
5], or cytokine profiling (reviewed in [
6‐
8]). Dysregulation of cytokine profiles has been demonstrated and often associated with an enhancement of pro-inflammatory cytokines in patients with ME/CFS [
9‐
13]. However, results are frequently inconsistent between studies and it remains difficult to determine a specific cytokine signature that could be implicated in the etiopathogenesis of ME/CFS.
Cytokines are important modulators of immune function and inflammatory responses and are easily affected by several factors which could be responsible for the high heterogeneity observed among studies. They are generally considered to function as soluble factors that mediate cell-to-cell communications in multicellular organisms. Their secretion can occur in classical secretion manner or via extracellular vesicles (EVs).
EVs are heterogeneous membrane-enclosed structures, spherical in shape with different origins, size and composition and released by cells into the extracellular milieu. A wide variety of EV subtypes have been proposed, defined by their size, biogenesis pathway, cargo, cellular source, and biological function, leading to a historically heterogeneous nomenclature including terms like exosomes and ectosomes. They may be released from damaged or stressed cells and carry proteins, pathogen-associated and damage-associated molecular patterns, lipids and miRNA, which are encapsulated and thus protected from degrading enzymes [
14]. Cytokines may be selectively sorted to EVs and act as signals to regulate and propagate the immune and inflammatory response [
15]. EV-associated cytokines are different from those released in a soluble form by cells of the same type [
16]. Once secreted, EVs can be rapidly captured by dendritic cells, phagocytes or macrophages [
17], bind neighboring cells that express specific cytokine receptors, or circulate passively through the bloodstream to sites distant to the local inflammatory lesion. A recent study reported that EV-associated and encapsulated cytokines were more stable than free cytokines and became biologically active upon interacting with sensitive cells [
16].
There is increased evidence that EVs are one of the main participants in cell-to-cell communication [
18] and drive inflammatory, autoimmune and infectious disease pathology [
19‐
22]. They can be sources of proinflammatory cytokines such as Tumor Necrosis Factor-α (TNF- α), Interleukins -6, -1β, -8 (IL-6, IL-1β, IL-8), Monocyte Chemoattractant Protein-1 (MCP-1), can stimulate their production in a variety of cells, and can also induce activation and proliferation of B and T cells as well as migration of granulocyte into inflamed tissues, promoting inflammatory pathways in recipient cells.
In previous reports, circulating EVs have been shown to be increased in number in ME/CFS [
23‐
25], results in line with other diseases such as cancers, Alzheimer’s and Parkinson’s disease [
21,
26‐
28]. Furthermore, a recent study was able to distinguish ME/CFS from idiopathic chronic fatigue and clinical depression by analyzing the content of EVs by means of proteomics [
25]. Observations in that study were an upregulation of proteins involved in focal adhesion, regulation of the actin cytoskeletal signaling pathway and pathways relevant to Epstein-Barr virus infection. Another report showed that neuronal and endocrine system pathways were among differentially expressed miRNA in EVs in ME/CFS, which suggests that there are defects in endocrine tissue functioning [
23].
Unlike our pilot study described below, the prior studies did not report on the cytokine composition of extracellular vesicles isolated from plasma of ME/CFS patients. We utilized an immune-profiling approach to determine whether an abnormal set of EV-associated cytokines could be identified in ME/CFS.
Discussion
Extracellular vesicles were isolated from the plasma of 35 patients with ME/CFS and 35 healthy controls. Transmission electronic microscopy and Western blot analyses on isolated EVs revealed mainly an exosomal morphology with mean sizes ranging from 30 to 130 nm along with the detection of exosomal markers (CD63, CD81, HSP70 and TSG101) (Fig.
2). Contrary to two previous reports that found that the mean size of ME/CFS EVs was reduced [
23,
24], we did not find any statistical significant difference in our study population. Although the total concentration of particles per ml of plasma did not differ between patients and controls in our study, the concentration of 30-100 nm vesicles was significantly increased in the diseased group (Fig.
1c), confirming findings recently reported for ME/CFS [
23‐
25]. These results are also in line with other conditions such as breast and gastric cancer [
26,
28], Alzheimer’s disease [
21] or cerebrovascular disease [
27] in which EVs were found to be elevated in the disease state. Oxidative stress, changes in intracellular calcium level and synaptic activity are contributing factors for inducing exosome secretion [
40,
41].
We further investigated the cytokine content in EVs and plasma from a subset of 19 ME/CFS and 19 healthy females. Using a linear model regression analysis that adjusted for age and BMI on the 45 cytokines measured, we did not find any statistically significant differences between groups in the plasma or EV cytokine levels. Furthermore, the data distribution in both plasma and EVs assessed by PCA did not show any clustering that could distinguish ME/CFS patients (Fig.
3a, b). Within subjects, the cytokine levels in EVs and plasma appeared as two different clusters (Fig.
3c) confirming that EV and plasma are two different compartments.
In a larger cohort analyzed by Hornig et al. [
10], nine plasma cytokine levels significantly differed between cases and controls. Montoya et al. [
12] found only two cytokines out of 51 measured to be significantly different in plasma of ME/CFS patients when compared with healthy individuals (Resistin was lower and TGF-β was elevated). Considering the limited number of analyzed samples in our pilot study (ME/CFS, n = 19 and controls, n = 19) in comparison to Hornig et al. (ME/CFS, n = 298 and controls, n = 348) or Montoya et al. (ME/CFS, n = 186 and controls, n = 38), it is not surprising that we did not find any differences between groups, as a result of low statistical power.
When comparing short duration (≤ 3 years) to long duration (> 3 years) of illness, several studies reported differences in plasma cytokine levels between ME/CFS subgroups and healthy individuals. Landi and colleagues [
42] compared the plasma cytokine levels of 100 ME/CFS of long duration to 79 controls and found reductions in 3 out of 31 cytokines measured (IL-7, IL-16, and Vascular Endothelial Growth Factor A, VEGF-A) while Hornig et al. observed that early ME/CFS cases (duration ≤ 3 years) showed statistically significant differences in comparison to controls for more than half of the 51 cytokines analyzed [
10]. These findings were not reproduced by Montoya et al., who did not find any cytokine to be significantly different between short duration cases (≤ 3 years) or long duration cases (> 3 years) and healthy individuals [
12]. Unfortunately, our analysis is limited as ME/CFS cases in this study were ill for more than 3 years, which prevented us from analyzing cytokine levels between short and long duration of illness.
Despite a lack of significance, higher and lower levels of several cytokines were found in plasma and EV samples as shown by the ratio of cytokine levels of ME/CFS vs. controls (Table
2). In both EVs and plasma samples, MIP-1alpha and IP-10 were both elevated in the ME/CFS group but not statistically different, a trend previously observed in a report that considered disease severity and compared mild ME/CFS patients to controls [
12] but opposite to another study showing significant lower levels of these cytokines in ME/CFS plasma samples [
10]. Interestingly, in our study, the Monocyte Chemoattractant Protein-1 (MCP-1 or CCL2) was significantly higher in both sample types in the ME/CFS group but lost statistical significance after correction for multiple comparisons (Table
2). MCP-1 is of relevance within the central nervous system (CNS), as it is expressed in different parts of the brain by neurons and microglia [
43‐
45]. MCP-1 acts mainly as an attractor for mononuclear cells, a mediator of inflammation, but is also involved in neuroprotection against excitotoxic injuries [
46,
47]. In ME/CFS, decreased levels of MCP-1 were observed in Cerebrospinal Spinal Fluid (CSF) samples in comparison to controls and these levels were higher in comparison to Multiple Sclerosis’ CSF samples [
48]. Also, higher levels were reported in plasma obtained from ME/CFS patients with short duration of illness versus controls [
10]. Elevated levels of MCP-1 have been also observed in several CNS-related pathologies and neurodegenerative diseases such as in the plasma of patients with Alzheimer’s disease [
49,
50], amyotrophic lateral sclerosis (ALS) [
51], in the cerebrospinal fluid of patients with ischemic stroke [
52], in peripheral blood mononuclear cells obtained from Parkinson’s disease and ALS patients [
51,
53], and in HIV-associated dementia [
54], schizophrenia [
55] and epilepsy [
56]. Furthermore, correlations between concentration of MCP-1 and symptoms such as depression, anxiety and fatigue [
57] or severity of disease have been reported [
57,
58]. Due to its role in the development of the inflammatory and immune responses, MCP-1 could be considered as an indicator that might allow the detection and quantification of the progression of ME/CFS.
Several EV-associated cytokine levels appeared to be elevated in the ME/CFS group (Table
2) but were not statistically different probably due to our limited sample size. Amongst them, IL-7 which was previously found to be elevated in serum samples of severely afflicted ME/CFS patients [
59], is a hematopoietic cytokine with critical functions in both B- and T-lymphocyte development. It is secreted by stromal cells, dendritic cells as well as neurons and neuronal progenitor cells [
60] and previous studies have revealed that IL-7 promotes neuronal differentiation [
61]. It can also stimulate cytotoxic functioning in mature T cells and NK cell proliferation, activities reduced in ME/CFS [
62‐
64]. Hardcastle et al. [
59] found a positive correlation between Interferon-γ (IFN-γ) and IL-7 in severely affected ME/CFS patients. Higher levels of IFNγ are associated with reduced Natural Killer cell cytotoxic activity observed in this illness and therefore, elevated levels of both cytokines may be involved in disease severity. In turn, lower IL-7 has been correlated with cognitive decline during aging [
65] and a reduction in IL-7 in plasma previously observed in ME/CFS [
42] suggests a reduction in immune activation along with a potential neuropathology similar to the process of aging.
Another example in our study, although not significant, were the elevated levels of the anti-inflammatory cytokine IL-10 in EVs isolated from ME/CFS patients. In contrast, extracellular vesicles levels of IL-10 levels were found to be reduced in gastric cancer [
26]. IL-10 receptors are found on different populations of brain cells including astrocytes, oligodendrocytes, and microglia [
66,
67]. The stimulation of these receptors with IL-10 reduces synthesis of proinflammatory cytokines, allowing the survival of brain cells [
68‐
70]. Its expression is elevated during the course of several CNS-related diseases such as multiple sclerosis, Alzheimer’s disease [
71] or meningoencephalitis [
72] and promotes survival of neurons and all glial cells. In ME/CFS, studies showed contradictory results about levels of IL-10, but none measured it in extracellular vesicles. A longitudinal study reported increased blood IL-10 levels in patients [
73], while others reported decreased [
9,
74] and increased [
75] levels in cerebrospinal fluid or plasma [
10]. These findings are inconsistent and this may be due to the heterogeneity of the disease, different analytical methods, and the presence of different patient subgroups. Compromises to brain cells may contribute to low levels of IL-10, and a decrease in CNS IL-10 may be related to symptoms reported in ME/CFS cases. The role of IL-10 in ME/CFS requires further investigation. The elevated levels observed in this pilot are consistent with an immune activation.
While the levels of IL-2 in our plasma samples did not differ between groups, they were elevated in the ME/CFS EVs in comparison to controls (Table
2) but not statistically significant. IL-2 levels have been previously reported to be higher in CSF [
48] and plasma from ME/CFS patients [
76]. We can only speculate that these elevated cytokine levels found in EVs from our cohort of ME/CFS patients are part of a specific immune response in ME/CFS.
The content of different types of EVs reflect that of the parent cells and are enriched in certain molecules, including cytokines, chemokines, functional microRNAs, and cell-specific antigens. EVs maintain characteristics of the antigen presenting cell from which they are derived, exposing antigen‐presenting MHC I and MHC II molecules on their surface [
77].
We further investigated if EV or plasma cytokine levels correlated with age, BMI and scores of the Bell, SF-36 and SSS forms, by performing Spearman’s and partial Spearman’s rank correlation adjusting for age and BMI. Sex was not included as a confounding variable in the analysis as all our participants were females. Even though several cytokines correlated either positively or inversely (Additional files
1 and
2: Figs. 1, 2) there were no commonalities between EV and plasma samples. IL-1RA was significantly and positively correlated with age and BMI in plasma samples from the ME/CFS group. It is well documented that IL-1RA, a natural antagonist to the proinflammatory cytokine IL-1, increases dramatically in obese subjects [
78]. Furthermore, it has been shown that IL-1RA levels were positively correlated with serum leptin levels in ME/CFS [
13] and that leptin was associated with fatigue severity in patients with ME/CFS [
12,
13] but also with chronic hepatitis C and irritable bowel syndrome [
79,
80]. Leptin has been identified as a major proinflammatory cytokine that induces IL-1RA secretion [
81] and is involved in NK cell activation, and innate and acquired immune responses [
82]. Unfortunately, our cytokine panel did not measure leptin and because of the close relationship existing between IL-1RA and leptin, we can only speculate that the higher levels of IL-1RA observed in ME/CFS plasma (Table
2) and the significant positive correlations of IL-1RA observed in the ME/CFS group and not in the control group may be due to increased inflammation in patients.
Interestingly, in EV samples from the control group, 3 cytokines strongly positively correlated with the 30-130 nm extracellular vesicle population (MIP-1β, MIP-3β and IL-2, Additional file 2: Fig. 2a–c) but not in the ME/CFS group. MIPs are crucial for immune responses towards infection and inflammation [
83] and are produced by macrophages and monocytes, and stimulated by proinflammatory cytokines such as IL-1β [
84]. IL-2 is a potent stimulator of T-cell proliferation and inhibits the development of inflammatory Th17 cells [
85]. The fact that these cytokines correlated with a particular population of EVs in the control group suggest that ME/CFS patients may be lacking specific vesicles carrying important proteins able to respond to inflammatory challenges. Cytokines and chemokines can be selectively sorted to EVs into multivesicular bodies (MVBs) and secreted via exosomes, into microvesicles shedding from the plasma membrane, or into apoptotic bodies (AB). This selective sorting towards EV subspecies has been demonstrated in Type 1 diabetes in which MCP-1 was expressed in all sorts of EVs while IL-27 was solely expressed in apoptotic bodies [
86]. Fitzgerald et al. [
16] observed that the number and pattern of cytokines packaged in EVs changed depending on the stimulus and suggested that the packaging of cytokines in EVs is not simply the property of a particular cytokine, but rather a tightly controlled biological process. In pathological conditions such as HIV or diabetes, it has been shown that the profile of cytokines in EVs was significantly increased and that specific cytokines were associated with disease duration [
87,
88].
We may assume that in ME/CFS, inflammation may induce changes in the packing of specific cytokines/chemokines into selective EVs and that their release is intimately linked to the state of the releasing cells, healthy versus diseased, near or at a distance.
We further analyzed cytokine-cytokine interactions by performing Spearman’s correlations for each pair of cytokines in plasma and EV samples for both groups to find whether a disrupted profile of EV-associated or circulating cytokines in plasma could be identified in ME/CFS patients. Overall, there was more significant cytokine-cytokine correlations in the plasma from the ME/CFS group and less in the EV samples in comparison to healthy individuals (Figs.
4b and
5b). In ME/CFS plasma, there were no inverse correlations in the control group while there were 13 within the ME/CFS group. IP-10 was highly interconnected with 12 other cytokines in the plasma from ME/CFS (Fig.
4b) and appeared to be a negative regulator. We found IP-10 to be negatively correlated with CD40 Ligand, but this inverse relationship was not observed in a previous study [
10] in which CD40 Ligand was the main driver of the inverse relationships with other cytokines. Recently, Moneghetti and colleagues found IP-10 to play a central role in the cytokine network of plasma from ME/CFS patients [
11], but the IP-10 interactions with other cytokines that they reported differed from those we observed in our study population.
During neuroinflammation, NK cells and activated CD4 + and CD8 + T-cells are attracted to sites of inflammation, infected or tumoral areas in the presence of IP-10 [
89]. In addition to its chemotactic properties in mediating an influx of inflammatory leukocytes into infected or inflamed tissues, IP-10 has non chemotactic functions such as inhibition of angiogenesis [
90] and α-defensins-like antibacterial properties [
91]. Furthermore, IP-10 has been shown to be involved in neurodegenerative disorders. In patients with Alzheimer’s disease, receptors for IP-10 are expressed on neurons and the induction of IP-10 by astrocytes and in microglia [
92], along with its binding on neurons and the accumulation of beta amyloid, result in neuronal dysfunction and apoptosis [
93]. In both multiple sclerosis and experimental autoimmune encephalomyelitis, IP-10 is highly expressed by astrocytes [
94], and largely associated with the influx of inflammatory leukocytes into neural tissue, and disease severity. Abnormal levels of IP-10 have been observed in body fluids of individuals infected with respiratory syncytial virus (RSV) [
95] and contributed to LCMV (Lymphocytic choriomeningitis virus) or West Nile Virus infections, in which IP-10 expression was restricted mainly to neurons [
96]. During Herpes virus infection it has been shown that infected IP-10-deficient mice showed higher viral loads in the CNS, decreased numbers of natural killer cells and CD8
+ T-cells [
97]. Altogether, these studies highlight the important role of IP-10 in mediating CNS inflammation, a hallmark of ME/CFS, thus further focus on this chemokine is needed.
EV cytokine–cytokine correlations in ME/CFS showed a pattern that differed from controls as well, with fewer significant correlations (Fig.
5c). The network diagrams showed that both groups had a unique negative cytokine-cytokine interaction; levels of IL-1RA were inversely associated with eotaxin in the control group (Fig.
5a), and IL-7 with MCP-1 in ME/CFS EVs. Hormig et al. found an inverse relationship of IL-1RA with M-CSF, GM-CSF and IL-17 but not with eotaxin in CSF samples from ME/CFS patients [
48].
Eotaxin is a member of the C–C chemokine family originally implicated in the selective recruitment of eosinophils into inflammatory sites during allergic reactions. It is induced by Th-2 cytokines, including interleukin IL-13, IL-10, and IL-4, and is produced by B cells, endothelial cells, lymphocytes, macrophages, epithelial cells, and chondrocytes [
98‐
100]. Although it has been been thoroughly investigated in allergic reactions, eotaxin has been shown to be involved in a skewed immune response toward a type-2 (Th2) profile and was recently associated with aging, neurogenesis and neurodegeneration, being able to influence neural progenitor cells, and microglia. Increased circulating levels of eotaxin have been described in fibromyalgia [
101,
102] and in several neuroinflammatory disorders [
103] such as Alzheimer’s disease [
104,
105], amyotrophic lateral sclerosis, Huntington’s disease, and secondary progressive multiple sclerosis [
105]. It has been associated with markers of aging and degeneration and correlated with cognitive measures. Since eotaxin is capable of crossing the blood–brain barrier, it is plausible that eotaxin can exert physiological and pathological actions in the central nervous system of ME/CFS patients. Hornig et al. suggested the possibility that increased levels of eotaxin along with dysregulation of IL-1 signaling observed in ME/CFS are part of an allergic process in central compartments also seen in a CNS infections [
106‐
108]. Altogether, eotaxin seems to be very promising to further study in ME/CFS to determine its prognostic value along with careful cognitive phenotyping monitoring and neuroimaging studies to evaluate its association with neurodegenerative changes.
Although several intercytokine network analyses were conducted by other groups in plasma and CSF [
10,
11,
13,
48], there were no commonalities with our results, which may suggest again the heterogeneity of the disease and the presence of different patient subgroups. Nevertheless, our results and those of previous reports suggest unusual regulatory relationships among cytokine in plasma and EVs, indicating a disruption of intercytokine networks.