Oxidative stress is a shared characteristic of ME/CFS and Long COVID

Significance

More than 65 million individuals worldwide are estimated to have Long COVID (LC), wherein individuals after infection report persistent fatigue, postexertional malaise, and other symptoms resembling myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). With no clinically approved treatments or diagnostic markers for these conditions, there is an urgent need to define the molecular underpinnings. By studying bioenergetic characteristics of immune cells in healthy controls, ME/CFS, and LC donors, we find lymphocytes from ME/CFS and LC donors exhibit elevated oxidative stress. Due to excess oxidative stress and consequent mitochondrial damage, ME/CFS and LC donor lymphocytes consume excess host energy, contributing to debilitating fatigue and other sequelae.

Abstract

Over 65 million individuals worldwide are estimated to have Long COVID (LC), a complex multisystemic condition marked by fatigue, post-exertional malaise, and other symptoms resembling myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). With no clinically approved treatments or reliable diagnostic markers, there is an urgent need to define the molecular underpinnings of these conditions. By studying bioenergetic characteristics of peripheral blood lymphocytes in 25 healthy controls, 27 ME/CFS, and 20 LC donors, we find both ME/CFS and LC donors exhibit signs of elevated oxidative stress, especially in the memory subset. Using a combination of flow cytometry, RNA-seq, mass spectrometry, and systems chemistry analysis, we observed aberrations in reactive oxygen species (ROS) clearance pathways including elevated glutathione levels, decreases in mitochondrial superoxide dismutase protein levels, and glutathione peroxidase 4–mediated lipid oxidative damage. Strikingly, these redox pathways changes show sex-specific trends. While ME/CFS females exhibit higher total ROS and mitochondrial calcium levels, males have normal ROS levels, with pronounced mitochondrial lipid oxidative damage. In females, these higher ROS levels correlate with T cell hyperproliferation, consistent with the known role of elevated ROS in initiating proliferation. This hyperproliferation can be attenuated by metformin, suggesting this Food and Drug Administration (FDA)-approved drug as a possible treatment, as also suggested by a recent clinical study of LC patients. Moreover, these results suggest a shared mechanistic basis for the systemic phenotypes of ME/CFS and LC, which can be detected by quantitative blood cell measurements, and that effective, patient-tailored drugs might be discovered using standard lymphocyte stimulation assays.

With more than 660 million documented COVID cases worldwide, it has been estimated that as many as 65 million individuals may have “Long COVID (LC),” a complex multisystemic condition associated with postacute sequelae of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection (1). LC spans all ages, even those who have experienced only moderate COVID infections, with 10 to 12% vaccinated, 10 to 30% of nonhospitalized, 50 to 70% of hospitalized SARS-CoV-2 infection survivors estimated to endure persistent symptoms after infection (1, 2), affecting individual quality of life and functional status (3). Although the clinical presentation of LC is highly heterogeneous, with adverse events spanning multiple organ systems from dysrhythmia and cardiac disorders (1, 4) to neurological and cognitive deficits such as “brain fog” (2, 5), there are several common shared symptoms, including fatigue (estimated pooled prevalence of 47%), shortness of breath (32%), and muscle pain (25%) (6, 7).

Strikingly, the clinical presentation of some LC patients strongly resembles myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), a complex chronic disease estimated to affect between 836,000 to 2.5 million individuals in the United States alone (8). By meta-analysis, it is estimated that 0.68% of the world population has ME/CFS, with estimates around 1.3% of adults in the United States (9, 10). Clinical studies show 2× higher preponderance of ME/CFS among females compared to males (9, 10). Based on four symptoms defined by the National Academy of Medicine criteria (8), ME/CFS is characterized by 1) profound fatigue lasting for at least 6 mo, 2) postexertional malaise, 3) unrefreshing sleep, and 4) cognitive impairment or orthostatic intolerance (8, 11, 12). Prospective studies have noted around half of the patients with LC meet the diagnostic criteria for ME/CFS (13), and roughly 75% of ME/CFS patients report an infection preceding symptom onset (11).

Despite the overlap in clinical symptoms, there is no known reported shared molecular basis between ME/CFS and LC. Due to the lack of standard diagnostics and treatments for LC and ME/CFS, patients can only be diagnosed with a physical and mental examination, based on the presentation of clusters of symptoms such as persistent fatigue and postexertional malaise (8, 14). With the mounting public health burden, there is an urgent need to understand the biochemical underpinnings of these conditions, which can guide the development of new diagnostics and targeted therapies (11).

Due to the symptom overlap, our primary intent was to identify shared molecular signatures between ME/CFS and LC donors versus healthy controls (HCs). We focused on immune cell bioenergetics based on several lines of evidence. Multiple studies (15–18) have found signs of inflammation in LC and ME/CFS patients, including higher serum cytokine levels (15–17) that are associated with patient-reported fatigue severity (15) and changes in both lymphocyte and monocyte cell populations (17) among LC donors. These studies suggest that the immune system may play a role in ME/CFS and LC pathogenesis and clinical symptoms.

Fatigue, especially in the form of postexertional malaise, is a principal hallmark of ME/CFS and also one of the most common LC symptoms (2, 6, 13). Plasma metabolomic and proteomic studies have identified molecular signatures associated with these symptoms, suggesting its link with energy metabolism. Specifically, multiple metabolic pathway aberrations have been detected in LC and ME/CFS samples, including changes in serotonin and tryptophan metabolism (19), cholesterol metabolism (20), dysfunctional gut microbial butyrate biosynthesis pathways (21, 22), oxygen transport to tissues (23), and deficient mitochondrial fatty acid oxidation and ATP metabolism (14, 24).

We focused on immune cells because they are a critical consumer and regulator of host energy and metabolism. Specifically, the immune system accounts for 15 to 20% daily energy expenditure in humans (25, 26). This usage is thought to increase to 25%, during a serious infection, highlighting the significant energy expenditure required for maintenance, activation, and proliferation of immune cells (25, 26). Given these estimates, we hypothesized that fatigue experienced by ME/CFS and LC patients may be connected with the increased energy consumption by the immune system due to lingering pathologies, even after the recovery from a specific infection. Among the intracellular pathways in immune cells, mitochondrial metabolism is especially relevant. Changes in mitochondrial ATP production, morphology, and function accompany lymphocyte differentiation and activation (27), with mitochondria driving cytokine production (28), thereby linking energy metabolic deficits in immune cells with chronic inflammation. Along these lines, other studies have hypothesized and identified signatures of mitochondrial dysfunction among ME/CFS and LC patients, including down-regulation of host mitochondrial genes even after COVID-19 recovery (29). Further, it has been observed that ME/CFS donor T cells exhibit altered mitochondrial morphology (30), with decreases in mitochondrial membrane potential (31, 32) among recovered COVID-19 subjects’ lymphocytes, resulting in hypothesized redox dysregulation in both ME/CFS and COVID-19 (33).

Reactive oxygen species (ROS) are at the nexus of chronic inflammation and metabolic regulation, critically driving mitochondrial oxidative phosphorylation, inflammatory cytokine activation (34), and tissue damage (characteristics of COVID-19 recovery) (34). Motivated by these separate lines of evidence, we directly measured redox parameters within lymphocytes from peripheral blood mononuclear cells (PBMCs) of 25 HCs, 27 ME/CFS, and 20 LC donors. From measurements capturing ROS levels, oxidative damage, and mitochondrial redox pathways, our findings identify elevated oxidative stress among lymphocytes is a shared molecular feature of ME/CFS and LC, associated with specific functional proliferation defects. Strikingly, only female patients had elevated ROS levels in ME/CFS and LC, together with the hyperproliferation of lymphocytes after stimulation in culture, whereas males showed signs of elevated lipid oxidative damage, with both sexes showing elevated levels of glutathione. Thus, while there are major sex-specific differences, they converge on evidence of oxidative stress and mitochondrial damage as the common effect, contributing to the persistent fatigue and other symptoms characteristic of these diseases.

Results

Elevated Reactive Oxygen Species in ME/CFS and LC Donor Lymphocytes, Compared to HCs.

ME/CFS and LC donor PBMCs were obtained from the ME/CFS Collaborative Research Center and the Stanford Post-Acute COVID Syndrome clinic, respectively. HC patient PBMCs were obtained from the Stanford Blood Bank, where donors were screened using a medical history questionnaire (35). ME/CFS patients (12), including patients meeting the National Academy of Medicine (NAM) ME/CFS criteria before and after the start of the COVID-19 pandemic, were diagnosed by a physician using the National Academy of Medicine (8), Fukuda (36), and Canadian Consensus criteria (37). LC donors were diagnosed using the combination of the Center for Disease Control (38) criteria for “LC or post-COVID conditions” and a symptom and functional status questionnaire (39). Specifically, LC participants had their clinical symptoms assessed in the last 7 d before their appointment following a Likert scale for severity and their functional status scale. Together, these measures capture whether patients report new symptoms four or more weeks after COVID infection and incorporate the Post-COVID 19 Functional Status (PCFS) Scale (39). Additional details related to patient characteristics are included in the supporting appendix. Fig. 1A shows that the mean ages are balanced across all three cohorts, and the proportion of females is slightly higher in LC and ME/CFS donors, reflecting the higher incidence of LC and ME/CFS in females compared to males (10, 38, 40–42).

Fig. 1.

Comparison of total ROS levels in lymphocytes of HC, ME/CFS, and LC donors. (A) Peripheral blood mononuclear cells were characterized from 25 HC, 20 LC, and 27 ME/CFS samples. Age and gender distributions are shown. (B) Representative flow cytometry plots (normalized to the mode) for ROS indicator DCFDA are shown for CD4 T cells for a single HC, LC, and ME/CFS donor. For (C–D), * indicates P < 0.05, ** indicates P < 0.01, *** indicates P < 0.001 based on a two-sided t test. (C) Comparing 16 HC, 15 LC, and 15 ME/CFS donors, DCFDA staining shows higher total MFI ROS levels in ME/CFS (CD4T P = 5.60 × 10⁻⁴, CD8T P = 1.91 × 10⁻⁴, CD19B P = 3.04 × 10⁻⁴) and LC (CD4T P = 0.104, CD8T P = 0.0210, CD19B P = 0.0727) lymphocytes compared to HCs. (D) ROS levels are uniquely elevated in ME/CFS and LC females, compared to controls (females: ME/CFS vs. HC: CD19B P = 0.00065, CD4T P = 0.00045, CD8T P = 9.5 × 10⁻⁵; LC vs. HC: CD19B P = 0.144, CD4T P = 0.062, CD8T P = 0.01065). In contrast, males show no significant differences (males ME/CFS vs. HC: CD19B P = 0.357, CD4T P = 0.469, CD8T P = 0.516; males LC v. HC: CD19B P = 0.278, CD4 T P = 0.643, CD8T P = 0.583).

Using flow cytometry, we compared mitochondrial bioenergetic parameters in lymphocytes (sample gating in SI Appendix, Fig. S1A). Although we did not detect differences in mitochondrial mass or membrane potential (SI Appendix, Fig. S2), decreases in mitochondrial ATP levels were observed across ME/CFS (1.6× decrease on average; HC vs. ME/CFS two-sided t test CD4T P = 0.0458, CD8T P = 0.0974, CD19B P = 0.356) and LC donors (2.8× decrease on average; HC vs. LC two-sided t test CD4T P = 0.0076, CD8T P = 0.0182, CD19B P = 0.0842), compared to controls (SI Appendix, Fig. S2). Our findings are consistent with a published study showing oxidative phosphorylation defects among PBMCs in ME/CFS patients, compared to controls (32). As decreases in ATP levels imply functional deficits in mitochondria, which significantly contribute to cellular ROS production through oxidative phosphorylation, we measured total ROS levels. Flow cytometric staining using 2’,7’-dichlorofluorescein diacetate (DCFDA) identified elevated reactive oxygen species (ROS) levels among lymphocytes in ME/CFS and LC donors, compared to HCs (Fig. 1 B and C). Comparisons of median fluorescence intensity (MFI) levels, corresponding to ROS levels, indicate significant increases in both LC and ME/CFS donor lymphocytes, compared to controls (Fig. 1C). We observed a bimodal distribution in the MFI ROS levels among ME/CFS donors (Fig. 1 C and D), with 8/15 donors (Fig. 1C) exhibiting >2-4× higher ROS levels compared to controls. This bimodal distribution could be explained by patient sex, where total ROS levels appear to be uniquely elevated in female ME/CFS and LC donors (Fig. 1D). No differences in ROS levels are found between males with ME/CFS and LC, compared to control males (Fig. 1 D, Bottom row).

We did not detect any association between LC and ME/CFS patient-reported fatigue severity and total ROS levels (SI Appendix, Fig. S3A). This result suggests that while ROS levels beyond a threshold may correlate overall with presence of ME/CFS or LC symptoms, the molecular features that explain variation in patient fatigue severity are more complex. Since the duration of symptoms could be pinpointed for LC patients, we found that female LC subjects consistently showed a weak positive association between symptom duration and total ROS levels across lymphocyte populations (SI Appendix, Fig. S3B, females CD19B R = 0.59 P = 0.12, CD4T R = 0.49 P = 0.22, CD8T R = 0.5 P = 0.21). In contrast, in males, where ROS levels are similar across different groups, we observed a weak negative association. These divergent trends in ROS levels suggest that the LC and ME/CFS disease pathogenesis is also distinct between sexes, specifically the mechanisms underlying resolution of sustained oxidative stress.

We detected no association between ROS levels and age (SI Appendix, Fig. S3C). However, among LC donors, we did identify a positive correlation between BMI and total ROS levels (CD19 B cells R = 0.56, P = 0.028; CD4 T cells R = 0.46, P = 0.081; R = 0.14, P = 0.13) (SI Appendix, Fig. S3D). As BMI has also been positively correlated with the incidence of post-COVID 19 symptoms (42), our findings suggest that oxidative stress levels track with some clinical heterogeneity in ME/CFS and LC, particularly in females and those with higher BMI.

Changes in ROS Levels Are Accompanied by Alterations in Oxidative Stress Pathways.

Fig. 2 summarizes some of the critical ROS generation and clearance pathways (43, 44), for which we conducted several measurements to evaluate differences in ME/CFS and LC groups, compared to controls. Intracellular ROS levels are mediated by the balance between ROS-generation and antioxidant processes. In terms of ROS-generation pathways, calcium in the mitochondrial matrix can drive the production of mitochondrial ROS through electron transport chain activity coupled to NADPH oxidases. To regulate mitochondrial superoxide (O₂^•−) levels, SOD2 converts superoxide to hydrogen peroxide (H₂O₂). Hydrogen peroxide can be subsequently transformed to H₂O and O₂ through catalase and glutathione peroxidase, which is coupled to glutathione oxidation. Without the protection of glutathione peroxidase 4 (GPX4) in the mitochondrial membrane or intermembrane space, hydroxyl ROS molecules can attack lipids to form lipid peroxides. Fig. 2 also summarizes the identified sex-specific differences in ROS generation, clearance, and down-stream pathways, where ME/CFS females show elevated mitochondrial calcium (Fig. 3A) and ROS levels (Fig. 1C), whereas ME/CFS males show larger changes in lipid oxidative damage (Fig. 5B). We demonstrate that oxidative stress and accompanying mitochondrial damage in ME/CFS and LC donors drives aberrant immune signaling, thereby enabling lymphocytes to consume excess host energy. The measurements and analyses to support these differences are shown in Figs. 3–7.

Fig. 2.

Summary of identified sex-specific redox pathway differences in ME/CFS and LC donors, compared to controls. ME/CFS and LC donor lymphocytes exhibit oxidative stress and mitochondrial damage, leading to proliferation defects that drive lymphocytes to consume excess energy. ROS generation and clearance pathways are shown, including mitochondrial superoxide, calcium, and superoxide dismutase mediated-conversion to hydrogen peroxide, and hydrogen peroxide breakdown by catalase and glutathione. Sex-specific alterations in these pathways among ME/CFS, LC donors are shown (females in purple, males in gray). The upward arrow corresponds to elevation of parameter in ME/CFS and LC donors, compared to control counterparts.

Fig. 3.

Comparison of intermediates and proteins involved in ROS generation and clearance within lymphocytes between HC, LC, and ME/CFS donors. (A) Rhod-2 AM flow cytometric staining identifies significantly elevated mitochondrial calcium levels in ME/CFS and LC female donors (females: ME/CFS vs. HC CD19B P = 0.00238, CD4T P = 0.0063, CD8T P = 0.0137; LC v. HC CD19B P = 0.0097, CD4T P = 0.0172, CD8T P = 0.240; males: ME/CFS vs. HC CD19B P = 0.791, CD4T P = 0.514, CD8T P = 0.825; LC vs. HC CD19B P = 0.851, CD4T P = 0.241, CD8T P = 0.358). (B) Flow cytometric staining shows significantly lower mitochondrial superoxide dismutase 2 (SOD2) protein levels in LC donors (HC vs. ME/CFS CD19 B cells P = 0.326, CD4 T cells P = 0.638, CD8 T cells P = 0.384; HC vs. LC CD19 B cells P = 0.073, CD4 T cells P = 0.023, CD8 T cells P = 0.023). (C) ME/CFS lymphocytes show significantly higher glutathione (GSH) levels (females ME/CFS vs. HC CD19B cells P = 0.00048, CD4T cells P = 0.00014, CD8T cells P = 0.00159; LC vs. HC CD19B P = 0.580, CD4T P = 0.827, CD8T P = 0.958; males ME/CFS vs. HC CD19B P = 0.0327, CD4T P = 0.0354, CD8T P = 0.0894; LC vs. HC CD19B P = 0.196, CD4T P = 0.093, CD8T P = 0.278). (D) Based on the reaction between hydrogen peroxide (H₂O₂) and glutathione (GSH) in Fig. 2 (H₂O₂ + 2GSH ←→ 2H₂O + GSSG), the association between GSH and total ROS levels (Fig. 1C) is shown for both females (Top row) and males (Bottom row). With each point corresponding to a donor, the plot shows a statistically significant positive correlation between total ROS and GSH levels (females CD19B R = 0.56, P = 0.0039, CD4T R = 0.63 P = 6.9 × 10⁻⁴; CD8T R = 0.52 P = 7.8 × 10⁻³; males CD19B R = 0.5 P = 0.025, CD4T R = 0.56 P = 0.01, CD8T R = 0.43 P = 0.057).

Fig. 5.

Sex-specific lipid oxidative damages in ME/CFS and LC donor lymphocytes, especially in mitochondria. (A) From Phetsouphanh et al. single cell RNA-seq data (Nat. Commun. 2024; GSE262861 (46)), GPX4 transcript levels are uniquely elevated in female LC donors central memory T cells, compared to recovered controls. Comparisons conducted at 4 (13F, 12M), 8 (11F, 11M), 24 (11F, 11M) months after infection (Female: 4 mo P = 0.0024, 8 mo = 0.00077, 24 mo P = 0.0032; Male: 4 mo P = 0.59, 8 mo P = 0.63, 24 mo P = 0.3, two-sided t test). (B) Flow cytometric comparison of lipid peroxide levels, using ratiometric lipid peroxidation sensor, shows both male ME/CFS and LC donor lymphocytes have significantly higher lipid peroxidation, compared to controls (males ME/CFS vs. HC: CD19B P = 0.0027, CD4T P = 0.0037, CD8T P = 0.0041, LC vs. HC CD19B P = 0.003, CD4T P = 0.0033, CD8T P = 0.0029; females ME/CFS vs. HC: CD19B P = 0.076, CD4T P = 0.448, CD8T P = 0.131, LC vs. HC: CD19B P = 0.0365, CD4T P = 0.236, CD8T P = 0.0977). Lower y-axis values correspond to higher lipid peroxidation. (C) Flow cytometric analysis of 10 females (5 HC, 5 ME/CFS) and 9 males (5 HC, 4 ME/CFS) identified elevated mitochondrial lipid peroxidation, particularly in males (CD4T P = 0.0178, CD8T P = 0.034; female: ME/CFS vs. HC CD4T P = 0.348, CD8T P = 0.085). Lower y-axis values correspond to higher mitochondrial lipid peroxidation. (D) Reanalysis of bulk RNA-seq data (GSE224615) from 13 recovered controls (5F, 8M), 23 LC donors (15F, 8M) from 8-mo after COVID infection identifies statistically significant suppression of PLA2G6 in LC male donors (P = 0.028, two-sided t test) with no significant differences among female donors (P = 0.078). (E) Flow cytometric comparison of lipid droplet levels identifies lower lipid droplet levels in ME/CFS donors compared to HC across both males and females (males ME/CFS vs. HC: CD19B P = 1.52 × 10⁻⁴, CD4T P = 5.92 × 10⁻⁴, CD8T P = 7.14 × 10⁻⁴, LC vs. HC: CD19B P = 7.31 × 10⁻³, CD4T P = 9.39 × 10⁻³, CD8T P = 0.0114; females ME/CFS vs. HC CD19B P = 0.0242, CD4T P = 0.0205, CD8T P = 0.00876; LC vs. HC CD19B P = 0.622, CD4T P = 0.826, CD8T P = 0.974). (F) Negative association between lipid droplet (E) and GSH levels (Fig. 3C) (females: CD19 B cells R = −0.63, P = 0.0072, CD4 T cells R = −0.67 P = 0.003, CD8 T cells R = −0.54 P = 0.025; males: CD19 B cells R = −0.9, P = 2.5*10⁻⁵, CD4 T cells R = −0.75 P = 0.0032, CD8 T cells R = −0.78, P = 0.0015).

Fig. 4.

Mass-spectrometry and immunofluorescence analysis suggests lipid oxidative damage in ME/CFS and LC patients. (A) Using extracted metabolites from 200,000 sorted CD3 T cells from ME/CFS and HC donors, hydrophilic interaction liquid chromatography-mass spectrometry and Turium-based systems chemistry analysis were conducted to map metabolic pathway differences between ME/CFS donor and HC T cells. From 747 identified metabolites, the top five mapped pathway differences are shown. Size of the nodes corresponds to magnitude of the difference in relative abundance of detected metabolites from mass spectrometry between ME/CFS and HC samples. (B) The reactions encoded by the large red circles (A) correspond to reactions significantly up-regulated in ME/CFS donors compared to HCs. The phospholipid synthesis reactions involve the production of lysoPE metabolites and are catalyzed by PLA2. (C) Reanalysis of bulk RNA-seq data (45) between 11 sedentary controls and 14 ME/CFS patients identifies significantly higher PLA2 transcript levels in ME/CFS donors (P = 0.044, two-sided t test). FPKM refers to fragments per kilobase of transcript per million mapped reads. (D) Representative images from immunofluorescence staining are shown for GPX4 in CD3 T cells from 1 HC, 1 LC, and 1 ME/CFS donor (blue corresponds to DAPI, red to CD3, magenta to MitoTracker Far Red, green to GPX4). (E) Quantification of average GPX4 pixel intensity from 1 HC, 1 LC, and 1 ME/CFS donor. Each point corresponds to a CD3 T cell, with comparisons conducted across 426 HC, 603 ME/CFS, and 239 LC T cells from maximum projection images. Significant increases in average pixel intensity indicate higher GPX4 levels in ME/CFS and LC donors, compared to HCs (P < 2.2 × 10⁻¹⁶ for both ME/CFS and LC from two-sided t test).

Fig. 6.

Effects of ROS on adaptive immune response. (A) Representative glutathione (GSH) profile from flow cytometry staining for one HC, LC, and ME/CFS donors’ CD4 T cells shows a long-tailed distribution for both HC and LC donors and a shifted distribution for ME/CFS. (B) To study long-tails, the GSH 95th percentile fluorescence intensity distribution is plotted across HC, LC, and ME/CFS donors. Each group demonstrates distinct tail-decaying behavior, where HCs decay quickly (light-tailed distribution), LC show heavy-tailed decay, and ME/CFS show a shifted and bounded tail distribution. (C) To understand this tail decaying behavior, we compared GSH levels on an additional set of 5 HC, 4 LC, and 5 ME/CFS donors between total, naïve (CCR7⁺CD45RO⁻), and memory CD4 T cell populations (CD45RO⁺). The long tails are found predominantly in memory and not naïve CD4 T cells, indicating that memory T cells are the main contributors to the heterogeneity in oxidative stress responses. For (D–F), PBMCs were labeled with CellTraceViolet and stimulated with anti-CD3/CD28 beads and IL-2. Based on a dye dilution assay, the proportion of proliferating T cells as a function of total ROS levels are shown at 5 d poststimulation. Experiments for (D–F) were conducted separately. (D) Proliferation data from 5 control and 5 ME/CFS females show female ME/CFS T cells hyperproliferate and exhibit higher ROS levels compared to control counterparts. Regression lines also show that control female CD4 and CD8 T cell proliferation has 10× greater slope with respect to ROS, compared to ME/CFS females. Compared to controls, the flat slope and higher line for ME/CFS T cells highlights hyperproliferation, suggesting a potential source of patient fatigue. (E) Data for 4 control and 5 LC females shows hyperproliferation among LC CD4 and CD8 T cells. The regression lines also show that control female CD4 and CD8 T cell proliferation has 10× greater slope with respect to ROS, compared to LC females. (F) Data from 7 control and 5 ME/CFS males shows a distinct pattern from females. CD4 and CD8 T cells from ME/CFS males do not hyperproliferate compared to controls, as ME/CFS males with similar ROS levels to controls do not show higher proliferation. As T cell proliferation among controls show 3.5× and 2.3× greater slope compared to ME/CFS CD4 and CD8 T cell counterparts, these results show greater insensitivity to ROS levels among ME/CFS male donors.

Fig. 7.

Metformin can redress T cell hyperproliferation in ME/CFS donors. (A) Study of fPOP subjects flagged patients 27 and 31, along with the father of 27, based on measured total ROS levels (DCFDA) in lymphocytes. Patient clinical information (table) identifies 27 and 31 as ME/CFS patients, based on the National Academy of Medicine ME/CFS diagnostic criteria (8). (B) Using CellTraceViolet staining, the proportion of proliferating T cells 5-d after stimulation were assessed for fPOP donors. ME/CFS donor CD4 and CD8 T cells with higher ROS are also associated with higher proportion of proliferation, consistent with Fig. 6. (C) To reduce proportion of proliferating T cells, PBMCs were treated at day 0 with ROS-modulating drugs, including N-acetylcysteine, metformin, and liproxstatin-1. Effects of drug on proliferation and ROS levels are shown 5-d after stimulation, where drugs reduce the proportion of high ROS proliferating T cells in Q1. (D) Comparison of drug effects between donor 27 and donor 27’s mother and father shows tested ROS-lowering drugs selectively modulate T cell proliferation in the ME/CFS donor 27 and not in the parents. (E) In-vitro metformin treatment in 5 HC and 6 ME/CFS donors finds a statistically significant reduction in the proliferation of ME/CFS CD4 T cells (P = 0.041). Proportion of proliferating T cells is estimated by proportion of (Q1+Q4)/(total CD4 T cells) as shown in (C). (F) Summary of Precision Medicine methodology for identifying ME/CFS patients and ROS-modulating drugs to address oxidative stress and hyperproliferation of T cells.

From flow cytometry analysis, we compared mitochondrial Ca²⁺ levels using Rhod-2 AM, a high affinity calcium indicator that primarily localizes in mitochondria. Compared to HCs, lymphocytes from ME/CFS and LC donors have 1.67× and 1.32× respectively higher calcium levels, with CD4 T cells showing statistically significant elevations in ME/CFS and LC groups (SI Appendix, Fig. S4A). Mitochondrial calcium levels were uniquely elevated in ME/CFS and LC female donors (Fig. 3A). We also compared SOD2 levels (Fig. 3B), which found that LC donor CD4 and CD8 T cells have significantly lower SOD2 levels, compared to HCs. Differences in SOD2 levels appeared to be similar across both sexes (SI Appendix, Fig. S4B). Combining these differences to capture the balance between mitochondrial ROS drivers and antioxidant pathways, we find the ratio of Ca²⁺ to SOD2 MFI levels is highly elevated in both the ME/CFS and LC donors, compared to HCs (1.75× in ME/CFS, 1.67× in LC). These results underscore that mitochondrial dysfunction appears to be a shared feature between ME/CFS and LC donors.

To avoid cellular oxidative damage, glutathione (GSH) acts as a ROS scavenger by reducing H₂O₂ (33, 47). With the hypothesis that glutathione deficiency is a driver of severe COVID-19 disease (47), we compared GSH MFI levels using flow cytometry. Both ME/CFS males and females exhibit significantly higher GSH levels, compared to controls (Fig. 3C). Since DCFDA inference of total ROS (Fig. 1C) can capture cellular hydrogen peroxide levels (48), we also evaluated the association between total ROS and GSH measurements (Fig. 3D). We noted a significant positive correlation between total ROS and GSH levels across all lymphocyte populations (Fig. 3D). In addition to highlighting a common antioxidant response in both male and female LC and ME/CFS lymphocytes, these results help explain uniquely elevated H₂O₂ levels in females. Given the stoichiometric relationship between H₂O₂ and GSH in the cellular chemical reaction (H₂O₂ + 2GSH ←→ 2H₂O + GSSG), we estimated the maximum tolerable oxidative stress, based on GSH availability. Specifically, we calculated the maximum tolerable H₂O₂ load (e.g., ½ * GSH MFI_{ME/CFS females}/GSH MFI_{HC females}) and compared this change to the measured differences in ROS levels between groups (e.g., DCF MFI_{ME/CFS females}/DCF MFI_{HC females}). Across all lymphocyte populations, our calculations show that measured increases in ROS levels within LC and ME/CFS female donors are far greater than the tolerable load based on GSH availability (SI Appendix, Fig. S5, 2.61× higher for LC, 1.53× higher for ME/CFS). In contrast, the increases in ROS are lower in males compared to the tolerable load (SI Appendix, Fig. S5, 0.56× lower for LC, 0.22× lower for ME/CFS). While ROS levels exceed antioxidant levels in ME/CFS and LC females, highlighting oxidative stress, higher GSH availability in males likely reduces hydrogen peroxide, accounting for the lack of H₂O₂ elevation in ME/CFS and LC males (Fig. 1D). Another potential contributor to this difference includes the thioredoxin-peroxiredoxin antioxidant system, where reanalysis of published RNA-sequencing (RNA-seq) data from Phetsouphanh et al. (46) comparing LC patients to recovered COVID-19 controls in a longitudinal cohort 4, 8, 24 mo after infection identified changes. Unlike females who showed no differences between groups, LC males showed elevated transcript levels of thioredoxin, compared to control counterparts (SI Appendix, Fig. S6), providing another potential reason for why males are able to buffer hydrogen peroxide levels more readily. Subsequent analysis (Figs. 4 and 5) reveals that ME/CFS and LC males also exhibit signs of oxidative stress and damage, although the specific aberrant pathways are distinct from females and are likely driven by different ROS molecules as identified below.

Elevated Oxidative Stress in ME/CFS and LC Donors Is Associated with Lipid Damage, Particularly in Mitochondria.

To evaluate evidence of oxidative damage (49), we separately pooled 200,000 sorted CD3⁺ T cells from both HCs and ME/CFS donors and immediately extracted intracellular metabolites. Using hydrophilic interaction liquid chromatography–mass spectrometry (50), 45,507 unique m/z analytes were detected in samples, of which 747 metabolites were identified using analytical standards. From mass spectrometry data, systems chemistry analysis was conducted using Turium (51), a computational program which enabled the mapping of 747 metabolites to 327 pathways. This approach helped identify the top pathways that are increased or distinctly detected in ME/CFS patients, compared to controls (Fig. 4A). In Fig. 4A, each circle corresponds to a specific reaction, up-regulated in ME/CFS donor T cells, compared to HCs. We found several metabolic pathway differences, with the most significant corresponding to phospholipid synthesis (Fig. 4A). Specifically, we found that phospholipid metabolites were significantly elevated in ME/CFS samples compared to HCs, including lysophosphatidylethanolamine (lysoPE) products [lysoPE(22:5), lysoPE(22:4), lysoPE(22:6), lysoPE(20:4), lysoPE(16:0), lysoPE(20:3)] (Fig. 4B). The reaction to produce these lysoPE products is shown in Fig. 4B, (hosphatidylethanolamine (C₇H₁₂NO₈PR₂) + H₂O → LysoPE(C₆H₁₃NO₇PR) + CHO₂R). These results are consistent with earlier plasma metabolomic studies on ME/CFS that identified significant phospholipid abnormalities (14).

As the identified reaction products (Fig. 4B) are catalyzed by phospholipase A2 (PLA2), we used RNA-seq data from Bouquet et al. (45) to evaluate differences in PLA2 transcript expression from 11 sedentary controls and 14 ME/CFS patients whole blood samples. We found that PLA2G4A, a calcium-dependent cytosolic phospholipase implicated with eicosanoid metabolism, has significantly higher transcript levels in ME/CFS donors compared to controls (Fig. 4C). These findings provide additional external validation for phospholipid synthesis dysregulation in ME/CFS donors.

Lipid peroxidation of fatty acyl groups occurs predominantly in membrane phospholipids, when oxyl, peroxyl, and hydroxyl radicals engage with unsaturated fatty acids to form lipid peroxides. Lipid peroxidation promotes the formation of several lipid byproducts including lysophospholipids (52). The conversion of lipid peroxides to lipid alcohols is catalyzed by GPX4, which protects lipids from oxidative damage. To initially test for differences in this pathway, we conducted GPX4 immunofluorescence staining (Fig. 4D) across CD3 T cells from a single HC, LC, and ME/CFS donor. Our staining showed LC and ME/CFS donor lymphocytes have significantly higher (1.77× and 1.9×, respectively) GPX4 levels (Fig. 4E), compared to HC T cells.

To evaluate whether differences in GPX4 levels are conserved across multiple donors, we reanalyzed single-cell RNA-seq data from Phetsouphanh et al. (46), who compared LC patients to recovered COVID-19 controls in a longitudinal cohort 4, 8, 24 mo after infection. Comparison of GPX4 transcript level counts among central-memory T cells from this dataset (46) shows unique elevation in female LC patients, compared to recovered controls (Fig. 5A). Moreover, we found that GPX4 transcript elevation in females was most pronounced in CD4 central-memory T cells, compared to effector-memory or naïve T cells (SI Appendix, Fig. S7), suggesting that oxidative stress is likely associated with deficient adaptive immune responses, consistent with our analysis.

Based on these differences, we hypothesized that female LC and ME/CFS T cells were better protected against lipid oxidative damage, compared to males. Therefore, we directly measured lipid peroxide levels using flow cytometry. Our measurements found substantially higher lipid peroxide levels in both ME/CFS and LC donors’ lymphocytes, compared to controls (SI Appendix, Fig. S4C). Consistent with our scRNA-seq findings, the differences in lipid peroxide levels between ME/CFS and LC donors compared to controls were significantly more pronounced in males (Fig. 5B). Since investigation of z-stacks from immunofluorescence staining (Fig. 4D) suggested mitochondrial and GPX4 colocalization, we measured mitochondrial lipid peroxidation, using a mitochondrial-targeted lipid peroxidation probe MitoPerOx. Flow cytometric analysis of 10 females (5 HC, 5 ME/CFS) and 9 males (5 HC, 4 ME/CFS) identified significantly elevated mitochondrial lipid peroxidation in ME/CFS T cells, particularly in males (Fig. 5C). As mitochondrial lipid oxidative damage has been implicated with mitochondrial dysfunction (53) (due to changes to mitochondrial morphology, ATP production, etc.), this result provides one underlying explanation for bioenergetic deficiencies in ME/CFS and LC male donors (14, 24, 30). Based on mass-spectrometry data with systems chemistry analysis implicating PLA2 differences in ME/CFS donors and the known role of PLA2G6 in protecting against mitochondrial lipid peroxidation (54), we evaluated whether there was evidence of PLA2G6 loss in males from published bulk RNA-sequencing data. By reanalyzing bulk RNA-seq data from Yin et al. (55), who evaluated transcript differences in PBMCs between 13 recovered controls (5F, 8M) and 23 LC donors (15F, 8M) from 8-mo after COVID-infection, we identified statistically significant suppression of PLA2G6 in LC male donors with slight elevations among female donors (Fig. 5D). These differences provide additional external independent validation that changes in PLA2, which modulate mitochondrial oxidative damage, show sex-specific differences in LC donors.

To protect membranes from lipid oxidative damage, lipid droplets can sequester ROS damaged lipids (56). Additionally, fatty acids derived from lipid droplets can be converted to acylcarnitines, the substrates for mitochondrial fatty acid oxidation. Based on detected changes in acylcarnitine metabolism in ME/CFS donors (Fig. 4A), we compared lipid droplet levels. We found significantly lower lipid droplet levels in both LC and ME/CFS donors across all measured lymphocyte populations (SI Appendix, Fig. S4D). Both ME/CFS and LC males and females exhibited significantly lower lipid droplet levels (Fig. 5E), although levels were higher in control females compared to males. Across both male and female lymphocytes, lipid droplet levels are inversely correlated with GSH levels (Fig. 5F). Our results correlating lipid droplet and glutathione levels link redox homeostasis with lipid composition and fatty acid oxidation, highlighting a sex-specific mechanism governing metabolic consequences of oxidative stress.

Elevated Oxidative Stress in Memory CD4T and Altered T Cell Proliferation Responses to ROS in ME/CFS Donors Suggests a Deficient Adaptive Immune Response, upon Stimulation.

It is well known that initial elevated ROS levels are a critical, albeit transient component of both B and T cell lymphocyte activation (57–59). Upon T cell activation, TCR signaling stimulates calcium influx into the mitochondria, which drives mitochondrial ROS production, activates NFAT signaling, and triggers IL2 cytokine production. Two lines of evidence support the need of mitochondrial ROS for T cell stimulation. One, within in-vitro settings, T cells treated initially with the antioxidant N-acetyl-cysteine at millimolar concentrations do not proliferate (60). Two, in-vivo evidence shows that T cells unable to increase mitochondrial ROS levels (e.g., by calcium influx) do not proliferate upon antigen stimulation (61). Thus, the ROS and mitochondrial calcium elevation (Fig. 3B) in ME/CFS females suggests that there are sex-specific differences in T cell proliferation compared to controls.

Next, reexamination of glutathione data revealed significant variability in GSH distribution among CD4 T cells within each donor, especially among HC and LC patients (Fig. 6A). Comparing the extreme value distributions (95th percentile fluorescence intensity) between HC, LC, and ME/CFS donor CD4 T cells, we observed that HC CD4s decay most quickly (light-tailed distribution), whereas LC CD4s decay with a heavy-tailed distribution (Fig. 6B). In contrast, ME/CFS donor CD4s have a shifted and bounded extreme value distribution. As comparisons of the extreme values alone rather than the medians or means identifies statistically significant elevation of GSH in LC and ME/CFS donors (HC vs. LC CD4 P = 0.00831; HC vs. ME/CFS CD4 P = 8.5 × 10^-6), this result underscores that only comparing the medians overlooks the critical heterogeneity that contributes to LC heavy-tail behavior. These results also suggest that a significantly larger proportion of T cells endure oxidative stress in ME/CFS donors, which is reflected in the shift in both comparisons of medians (Fig. 3C) and extreme value distributions (Fig. 6B) (62). In light of this analysis and GPX4 elevation in memory CD4 T cells (Figs. 4E and 5A), we assessed whether naïve (CCR7+CD45RO−) or memory (CD45RO+) CD4 T cells may account for the differences in these phenomena (SI Appendix, Fig. S1B for gating). Our analysis found that the long tails in CD4 GSH profiles were primarily reflected in memory and not naïve CD4 T cells (Fig. 6C). As antigen stimulation encourages the formation and recall of a T cell memory response, our results suggest that ME/CFS donor T cells would likely respond distinctly to antigen stimulation. Therefore, we investigated the relation between oxidative stress in ME/CFS and LC donor T cells and its proliferation, after stimulation. PBMCs, which were labeled with CellTrace Violet proliferation dyes, were stimulated with anti-CD3/anti-CD28 antibodies and IL-2. The extent of proliferating T cells was measured 5 d poststimulation, along with ROS levels and surface activation marker levels (CD69, CD137) in T cells (see Methods for details). Separate experiments were conducted for ME/CFS females (Fig. 6D), LC females (Fig. 6E), and ME/CFS males (Fig. 6F). Upon stimulation, we observed no differences in CD69 and CD137 activation status for any ME/CFS and LC samples compared to HCs (SI Appendix, Fig. S8).

Consistent with the role of ROS in T cell activation, our analysis found among controls that the proportion of proliferating T cells linearly scales with oxidative stress, measured 5 d after proliferation (Fig. 6 D–F). Specifically, the relation between ROS and T cell proliferation appears distinct in ME/CFS and LC subjects, with additional sex-specific differences. Compared to female controls (Fig. 6D), ME/CFS female CD4 and CD8 T cells show on average 26.5% and 28.6% higher proliferation, respectively. The CD4 and CD8 T cell proliferation regression lines for controls in Fig. 6D have 10× greater slope (blue line) with respect to ROS, compared to ME/CFS females (red). Compared to controls, LC females on average show 39.7% and 42.8% higher CD4 and CD8 T cell proliferation, respectively (Fig. 6E). Consistent with the results in Fig. 6D, CD4 and CD8 T cell proliferation among controls has 10× greater slope (blue) with respect to ROS, compared to LC females. Together, these findings indicate that ME/CFS and LC T cells from female donors hyperproliferate, compared to controls.

In contrast, our analysis found no differences on average in T cell proliferation between controls and ME/CFS male donors (Fig. 6F). Control male CD4 and CD8 T cell proliferation has 3.5 and 2.3× higher slope with respect to ROS, compared to ME/CFS males (Fig. 6F). T cells from ME/CFS male donors exhibit relative insensitivity to ROS levels, compared to controls. Males succumb to mitochondrial lipid oxidative damage, which likely shapes mitochondrial membrane integrity and its insensitivity to ROS during activation and proliferation. These results suggest that the capacity for an individual male ME/CFS T cell to proliferate is insensitive to higher oxidative stress levels, pointing to a potential functional defect in ME/CFS T cell proliferation. As memory T cells account for the differences in GPX4 levels (Fig. 5A), exhibit heterogeneity in the T cell glutathione profile, and stimulation drives the formation of memory cells, our results imply deficient adaptive immune responses in ME/CFS donors. This finding agrees with an influenza vaccination study showing T cell hyperproliferation in the ME/CFS group versus controls, consistent with our result showing higher proportion of proliferating T cells upon in vitro stimulation (63). Additionally, characterization of stimulated CD4 T cells from LC donors (17) found significantly higher levels of intracellular IL-2 levels, which are primarily produced from T cells through ROS-NFAT signaling. While neither of these cited studies probed for sex-specific differences in T cell proliferation, our findings highlight that the sex-specific pathways in redox dysregulation also shape the functional differences in male and female ME/CFS or LC adaptive immune responses, where female T cells hyperproliferate and male T cells have an insensitive response to ROS.

As T cell proliferation is associated with a >10-fold increase in energy usage due to elevated protein synthesis (64, 65), these findings also point to one potential source of patient fatigue especially in ME/CFS female donors, where T cell activation drives rampant proliferation in ME/CFS donors. Based on this hyperproliferation and the specific characterized pathways (Fig. 2), we attempted to assess our findings in a clinical context and evaluated whether ROS-modulating drugs could attenuate this hyperproliferation.

ME/CFS Donor T Cell Hyperproliferation Can Be Attenuated with Metformin.

We evaluated total ROS levels in three patients from the family population-omics profiling (fPOP) cohort (66), who presented with chronic symptoms (Fig. 6A, table). The clinical ME/CFS diagnosis, based on NAM criteria (8), was not available at the time of experimental characterization. Therefore, to contextualize changes in ROS levels, these donors were compared to 6 HCs (4 females, 2 males), along with samples from a ME/CFS-presenting patient’s (patient 27) mother and father.

Comparison of total ROS levels using flow cytometry enabled us to flag patient 27, the father of 27, and patient 31, based on elevation of total ROS levels (Fig. 7A). Overlaying our ROS data with patient symptoms showed that both patients 27 and 31 could be diagnosed as ME/CFS+, based on the NAM criteria (8). Additionally, patient 30, who exhibited chronic postnasal drip but did not report any symptoms of fatigue, was not flagged by our assay, highlighting its sensitivity in distinguishing between patients with chronic symptoms with different pathologies (Fig. 7A). While our assay did not flag the mother of 27, our assay detected higher ROS levels in the father of 27 compared to HCs but still lower than patient 27.

As higher oxidative stress in T cells is associated with proliferation, we used CellTraceViolet staining to track the proliferation of fPOP donor T cells, upon anti-CD3/anti-CD28 and IL-2 stimulation. Consistent with our model, we found higher proportion of proliferating T cells in patients 27 and 31, who both presented with ME/CFS symptoms and exhibited elevated ROS levels, compared to both parents of 27 and patient 30, who showed no fatigue symptoms and no elevation in ROS levels (Fig. 7B). Moreover, we found that the initial ROS load (Fig. 7A) correlates strongly with the T cell proliferation 5 d poststim (Fig. 7B), further supporting the relationship between excess T cell ROS levels at day 0 (initial ROS load) and hyperproliferation (SI Appendix, Fig. S9, CD8 T cells R² = 0.9078, CD4 T cells R² = 0.7472).

Based on this relationship, we assessed whether ROS modulating drugs could lower the proportion of proliferating T cells. From our metabolic characterization (Figs. 2–5), we tested three drugs (N-acetylcysteine, metformin, liproxstatin-1). N-acetylcysteine (NAC), an FDA approved drug, helps replenish glutathione levels. Metformin, also an FDA approved drug, can inhibit mitochondrial complex I proteins and subsequently induce SOD2 expression, thereby reducing ROS formation (67). Our earlier assays identified SOD2 loss as a characteristic of LC and some ME/CFS donor lymphocytes (Fig. 3B). Moreover, both electronic health record analysis and clinical trial studies have shown metformin may reduce LC incidence after SARS-CoV-2 infection, with a double-blinded phase III study on >1,300 patients showing metformin reduces LC incidence by 41%, especially among female subjects (68, 69) or those with higher BMI. Finally, as our lipid metabolic studies showed GPX4 changes and higher lipid peroxidation in ME/CFS and LC donors, we tested liproxstatin-1 (70, 71), a potent GPX4 modulator.

We treated patient 27, 27’s father, and 27’s mother PBMCs at day 0 and compared the T cell proliferation after 5 d of stimulation. Our flow cytometry data suggested that all three drugs could modulate the proportion of high ROS proliferating T cells (Fig. 7C). Compared to 79.3% of proliferating CD4 T cells upon stimulation, NAC, metformin, and liproxstatin-1 treatment were able to lower this proportion to 54.3%, 53.2%, and 64.5% respectively. Moreover, we found that ME/CFS patient 27’s T cell proliferation was uniquely responsive to these ROS-lowering drugs, in contrast to her mother or father (Fig. 7D).

To test the generality of these findings, we tested these drugs on 5 HC and 6 female ME/CFS donors, including those reported in Fig. 6D. While NAC at micromolar levels did not reduce T cell hyperproliferation in ME/CFS or LC female donors (SI Appendix, Figs. S10 and S11), metformin had a statistically significant effect in reducing CD4 T cell hyperproliferation in ME/CFS female donors (Fig. 7E, CD4 9.8% reduction in proliferation P = 0.041, CD8 10.5% reduction in proliferation P = 0.39). Metformin appeared to uniquely reduce the proliferation in ME/CFS donor T cells with no effect among the HCs (Fig. 7E, HC CD4 P = 1, CD8 P = 0.69). Liproxstatin-1 did not show a statistically significant effect in lowering T cell proliferation among female ME/CFS or LC donors (SI Appendix, Figs. S10 and S11). Metformin treatment had a smaller and nonstatistically significant effect in LC female PBMCs (SI Appendix, Fig. S11, CD4 P = 0.31, CD8 P = 0.42), where treatment reduced on average CD4 and CD8 T cell proliferation in female LC donors by 5.2% and 7.6% respectively.

Consistent with the finding that male ME/CFS T cells did not hyperproliferate and displayed relative insensitivity to ROS, we did not observe any reduction in T cell proliferation upon treatment with NAC, metformin, or liproxstatin-1 (SI Appendix, Fig. S12).

Although the comparisons from the fPOP cohort are a small case study, these findings demonstrate a Precision Medicine approach for helping potentially diagnose and treat ME/CFS (Fig. 7F), especially in female donors. Further, tracking T cell proliferation in the context of ROS levels can pinpoint which individuals may potentially benefit from ROS-lowering drugs and even identify novel drug candidates that mediate this link between ROS and lymphocyte proliferation.

Discussion

While ME/CFS and LC share many clinical features along with other postacute infection syndromes (11) and are increasingly diagnosed together, there are no approved molecular diagnostics or treatments for these patients. Previous studies have hypothesized oxidative stress as a common biochemical signature (33) and even shown some evidence in circulating serum redox proteins (72–74) among ME/CFS and LC patients. Unique to our work, the direct measurement of intracellular redox properties and extracted metabolic pathways shows that both diseases share signatures of elevated oxidative stress among lymphocytes that would impact mitochondrial function.

Several epidemiological studies report three-four times higher post-COVID 19 and two times higher ME/CFS symptoms incidence among females (10, 38, 40–42). Clinical studies also report different symptom presentation between ME/CFS males and females (41, 75), hinting that the specific biochemical dysregulation in ME/CFS and LC will likely have sex-specific differences. Our results provide a molecular approach to explain these differences.

Specifically, females show higher ROS levels and insufficient antioxidant levels (GSH), while males show mitochondrial lipid oxidative damage (Fig. 2). These findings suggest that the pathophysiology for ME/CFS and LC are distinct between sexes. While additional studies are needed to understand the mechanistic basis for these differences, we offer one potential explanation. As sex hormones (e.g., estradiol) can regulate serum antioxidant enzyme levels and T cell activation (76, 77), the distinct underlying redox biology for ME/CFS females may be partially explained by these changes (78), which appear to be dysregulated in LC donors too (79). Additionally, both our study and a deep phenotyping study (80) on ME/CFS patients show differences in fatty acid oxidation between ME/CFS males and females, with the results here uniquely highlighting larger changes in lipid peroxidation. As androgens modulate peroxisome proliferator-activated receptor (PPAR)α levels in CD4 T cells, where PPARα regulates fatty acid oxidation, differences in PPARα may explain lipid metabolism differences (81, 82).

Next, our findings suggest that these metabolic differences likely impair the adaptive immune response in ME/CFS and LC individuals. Our findings, which show that CD4 T cells’ oxidative profiles in LC patients exhibit a long-tail phenomenon, also caution against solely comparing medians/means between donors, which will miss out on this heterogeneity, uniquely found in memory T cells. By comparing T cell proliferation upon antigen stimulation, we find oxidative stress linearly scales with the proportion of proliferating T cells, meaning a higher fraction of female ME/CFS donor T cells proliferate upon stimulation. With the higher energy demands needed for T cell proliferation, these differences suggest the availability of energy as a significant and unique contributor to fatigue in ME/CFS donors. One possibility is that this lymphocyte dysfunction, driven by oxidative and mitochondrial damage, acts as an “energy sink,” much as an active infection does, draining the body of its available finite energy reserves and giving rise to debilitating fatigue and other sequelae. Another possibility is that the higher calcium levels in females drive hyperproliferation of T cells through a ROS-dependent mechanism, which subsequently induces the production of cytokines (e.g., IL-2, Type I interferons), creating a feedback loop of sustained immune activation and accounting for persistent fatigue symptoms. In contrast, males may be unable to buffer these ROS as effectively, succumbing to mitochondrial lipid oxidative damage, and consequent deficient mitochondrial function (e.g., lower ATP). More broadly, it suggests that at least in females, ROS levels may serve as a dynamic and system-dependent link for adjusting T cell proliferation.

Additionally, our work demonstrates a potential Precision Medicine methodology to identify specific preexisting FDA-approved and novel candidates that can adjust ROS levels and subsequently curtail T cell hyperproliferation in ME/CFS donors. A clinical study has shown metformin’s efficacy in lowering LC incidence by 41% of treated subjects (68) post SARS-CoV-2 acute infection. While our analysis looks at subjects already with LC, rather than in a prophylactic setting, the parallels between our work and the clinical study are noteworthy. The clinical study (68) shows metformin’s prophylactic benefit in females and those with higher BMI. Similarly, we note elevation of ROS levels in female subjects and those with higher BMI, with the capacity to curtail aberrant T cell proliferation in females, suggesting our results offer one plausible mechanism of action. Broadly, these results suggest that the link between oxidative stress and T cell proliferation can be exploited to identify novel drug candidates.

Although our findings were focused on identifying a conserved signature between ME/CFS and LC patients, based on commonly presented symptoms, our analysis suggests several important differences. For example, the specific aberrant antioxidant pathways appear to differ slightly, where LC patients overall show lower mitochondrial superoxide dismutase, ATP levels, and calcium levels compared to ME/CFS donors. One key difference between LC and ME/CFS subjects in our study is the duration of symptoms, which appear to moderately correlate with ROS signatures in females (SI Appendix, Fig. S3), where LC donors in this study experience symptoms from months to years and ME/CFS donors for multiple years. Based on these differences, we hypothesize that the specific ROS molecules and antioxidant pathway deficiencies contributing to oxidative damage seen in both LC and ME/CFS donors have differences and show sex-specific patterns, an aspect worth pursuing for further studies. Additionally, our comparison of the extreme value distributions found that LC T cells show a heavy-tailed oxidative stress profile and ME/CFS donor T cells show a shifted but bounded extreme value distribution. From these differences, it is plausible that LC donors capture an intermediate but distinct state between HC and ME/CFS donors. One interpretation is that continuous exposure to high oxidative stress in lymphocytes may cultivate tolerance to ROS, as a possible protective mechanism for some patients (83). These extended breaches in endosymbiosis [e.g., mitochondrial membrane integrity (84)] for ME/CFS patients allow oxidants to continually seep out of the mitochondria, leading to T cell mitochondrial reprogramming and sustained aberrant immune signaling (e.g., constitutively elevated calcium).

These results also raise the question as to how these oxidative stress pathways are triggered and maintained. Here, we can only speculate that since LC is clearly triggered by an infection, the initial elevation of ROS levels necessary for lymphocyte activation is overly prolonged in some individuals, perhaps caused by an impaired ability to clear the virus, causing damage to mitochondrial fatty acid oxidation and ATP production. While the kinetics of ROS elevation and clearance may be distinct across sexes, in both sexes with LC the damage is long lasting. In ME/CFS it has long been thought that there was a causative infection, although a specific pathogen has not been identified, despite intensive efforts. Nevertheless, it has been noted that other infections have produced LC-like symptoms in particular individuals (11), so the pathologies we document here might be a feature of other infections as well. It may be that individuals with subclinical immune deficiencies suffer from a more prolonged infection (85) and accompanying ROS elevation that produces lasting mitochondrial damage. Although our findings show consistent trends in ROS dysfunction and can be validated using externally published data, additional studies directly measuring these changes in larger and diverse cohorts, with consideration of confounding factors, will be needed to test the generality of these conclusions further. Such studies can also help pinpoint sex-specific changes in phenotypically relevant lymphocyte populations, thereby helping further elucidate the mechanism for redox dysfunction in LC and ME/CFS. Regardless, we are hopeful that the identified pathways from our findings may provide a blueprint to guide the development of possible diagnostics and therapies to help ME/CFS and LC patients.

Methods

Method details are available in SI Appendix.

Informed consent was obtained from all patients prior to sample collection. Blood collection, demographic, and clinical data from study cohorts were approved by Stanford University’s IRB (IRB-40146, IRB-64344).

Briefly, flow cytometry approaches were used to measure bioenergetic parameters. Intracellular metabolic pathways were mapped using sorted T cells, mass spectrometry, and Turium. Anti-CD3/CD28+IL2-stimulated T cells were used in CellTrace dye dilution assay with and without drugs to study functional deficits.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Previously published data were used for this work (45, 46, 55). All other data are included in the manuscript and/or SI Appendix.