Participants. This current study included n = 18 healthy controls (HC), n = 14 participants with ME/CFS, and n = 15 participants with long COVID. There were no significant differences between participant cohorts for age, sex, or highest level of education achieved (Table 1). Body mass index (BMI) differed significantly between cohorts (adjusted P value [Padj] = 0.021), whereby HC reported a significantly lower BMI compared with participants with long COVID (P = 0.016). Full blood count was determined for all participants. The number of monocytes (P = 0.049) and basophils (P = 0.030) differed significantly between groups. The number of monocytes was significantly higher in HC and participants with long COVID compared with those with ME/CFS; however, significance was lost following Bonferroni corrections for multiple comparisons. Basophils were significantly higher in HC compared with those with ME/CFS; however, significance was lost following Bonferroni corrections for multiple comparisons. All QoL variables differed significantly between cohorts. For all 36-item short-form health survey (SF-36) domains, participants with ME/CFS or long COVID reported significantly lower QoL compared with HC. For all World Health Organization (WHO) Disability Assessment Schedule (DAS) domains, participants with ME/CFS or long COVID reported significantly higher levels of disability compared with HC. No significant differences were reported between participants with ME/CFS and participants with long COVID, excluding the WHODAS Self-Care domain, in which patients with ME/CFS reported more difficulty (Padj = 0.048). No participant with long COVID reported multiple SARS-CoV-2 infections at the time of sample collection. All participants reported any previous or current diagnoses. Reported comorbidities included fibromyalgia (n = 2) and postural orthostatic tachycardia syndrome (POTS, n = 3) in people with ME/CFS. A history of post–Epstein-Barr virus chronic fatigue was reported by n = 2 people with long COVID; however, reported chronic fatigue had resolved prior to SARS-CoV-2 infection. Irritable bowel syndrome (IBS) was reported by n = 2 people with long COVID and n = 2 people with ME/CFS. All clinical characteristics and demographics of participants are summarized in Table 1.

Participant demographics, full blood analysis, and quality of life

Participants with ME/CFS or long COVID were required to report on the symptom they regularly experience, with focus on the past 30 days’ frequency (how often they experienced the symptom), and severity (very mild to very severe) of symptoms. The occurrence of symptoms enabled the determination of case criteria fulfilled. All participants with ME/CFS met the Canadian Consensus Criteria (CCC), excluding 1 participant who reported an improvement in cognitive disturbances since a prior appointment with the Neuroimmunology and Emerging Diseases (NCNED) and fulfilled Fukuda criteria, thus demonstrating the fluctuating nature of the illness. All participants with long COVID fulfilled the WHO working case definition for “Post COVID-19 Condition.” There were no significant differences in the prevalence of symptoms between ME/CFS and participants with long COVID. The prevalence of symptoms is reported in Table 2.

Symptom prevalence of ME/CFS and long COVID

Differential gene expression. Differential expression of genes was filtered according to log fold change (FC) parameters –1.5 to 1.5 and a P value of 0.05, resulting in the selection of 29 genes in long COVID compared with HC (Table 3 and Figure 1). Of the 29 selected genes, 15 were upregulated and 14 were downregulated. Downregulated genes, including HLADQA1 and HLADQB1, had the highest degree of differential expression (log2FC = –4.81925, P = 0.009, and log2FC = –4.34154, P = 0.0102, respectively). Of the upregulated genes, KIR2DL5A/B had the highest degree of change with log2FC of 2.54102 (P = 0.0005).

Differentially expressed genes in long COVID. (A) Volcano plot displaying statistical significance (log10[P value] on the y axis and log2 fold change on the x axis). Selected genes meeting filter criteria are presented as those downregulated (≤-1.5) and those upregulated (≥1.5). (B) Heatmap of selected genes representing log2 normalized expression values from –8 to 6. Red indicates high levels of expression, while blue indicates low levels of expression. Clusters are organized according to upregulated or downregulated genes by participant cohort. HC, healthy control.

Differential gene expression in participants with long COVID

A total of 14 genes was selected as differentially expressed between patients with ME/CFS compared with HC (Table 4 and Figure 2, A and B). Of the 14 genes, 5 genes were upregulated and 9 genes were downregulated. Downregulated genes, including IFNA4/7/10/17/21, IGHG1, and IFNA6, had the highest degree of differential expression (log2FC = –2.42502, P = 0.0005; log2FC = –2.24777, P = 0.000008; and log2FC = –2.20172, P = 0.0009, respectively). Of the upregulated genes, CEACAM3 had the highest degree of change with a log2FC of 1.83403 (P = 0.0014). Hierarchical clustering grouped mRNA expression and samples according to similarity in expression patterns. Interpretation of the heatmap demonstrates similarities within ME/CFS and long COVID cohorts and similarities in expression patterns, except for a few samples (Figure 2A). Overlapping genes can be found in Figure 3, A and D. The full data set outputs can be found in Supplemental Data 2 (supplemental material available online with this article; https://doi.org/10.1172/jci.insight.183810DS1.)

Differentially expressed genes between ME/CFS. (A) Volcano plot displaying statistical significance (log10[P value] on the y axis and log2 fold change on the x axis). Selected genes meeting filter criteria are presented as those downregulated (≤ –1.5) and those upregulated (≥1.5). (B) Heatmap of selected genes representing log2 normalized expression values from –3 to 6. Red indicates high levels of expression, while blue indicates low levels of expression. Clusters are organized according to upregulated or downregulated genes by participant cohort. HC, healthy control.

Overlapping gene expression in ME/CFS and long COVID. (A) Heatmap representing log2 normalized expression values (–5 to 5). Red represents higher expression, blue represents low expression, and gray represents no differential expression. Data exported from Rosalind Bio. (B and C) Heatmap of the top 10 canonical pathways (B) and diseases and functions (C). The darker gradient indicates greater significance. P < 0.05. Data exported from IPA. (D) Unique and overlapping genes. HC, healthy control.

Differential gene expression in participants with ME/CFS

Gene set analysis. The change in regulation within each gene set relative to the baseline was described using gene set analysis (GSA); both undirected enrichment score (UES) and directed enrichment score (DES) for the top 15 gene sets are presented in Table 5. Differentially expressed genes in long COVID were associated with PD-1 signaling (UES = 1.7782, DES = –1.742), IL-6 signaling (UES = 1.6638, DES = 0.9619), TGF-β signaling (UES = 1.6145, DES = 1.2819), antigen presentation (UES = 1.589, DES = –0.6682), mitogen activation protein kinase (MAPK) signaling (UES = 1.5361, DES = –0.4939), and mTOR signaling (UES = 1.4928, DES = 0.3127). In ME/CFS, GSA identified enrichment of gene sets including peroxisome proliferator–activated receptors (PPAR) signaling (UES = 1.9043, DES = –1.4736), fatty acid metabolism (UES = 1.7918, DES = –1.5557), NK receptors (UES = 1.7526, DES = –1.3492), glycolysis and glucose import (UES = 1.7244, DES = –1.5434), anergy (UES = 1.6924, DES = 1.4338), B cell exhaustion (UES = 1.6487, DES = –1.5374), and epigenetic modification (UES = 1.5496, DES = 0.9327).

Gene set analysis for genes differentially expressed in long COVID and ME/CFS compared with HC

Overlapping gene sets included chemokine signaling (long COVID: UES = 1.6212, DES = –1.4999; ME/CFS: UES = 1.9815, DES = –1.8727), type I IFN signaling (long COVID: UES = 1.5237, DES = –0.7452; ME/CFS: UES = 1.8554, DES = –1.5052), type II IFN (long COVID: UES = 1.5971, DES = 0.092; ME/CFS: UES = 1.7313, DES = –0.8088), TNF signaling (long COVID: UES = 1.6121, DES = 1.1078; ME/CFS: UES = 1.5776, DES = 0.7148), CTLA4 signaling (long COVID: UES = 1.5185, DES = –0.7578; ME/CFS: UES = 1.7201, DES = –1.1482), DAP12 signaling (long COVID: UES = 1.4937, DES = 1.115; ME/CFS: UES = 1.5637, DES = 1.1961), Janus kinase/signal transducers and activators of transcription (JAK/STAT) signaling (long COVID: UES = 1.594, DES = 0.978; ME/CFS: UES = 1.6042, DES = 0.3461), and other IL signaling (long COVID: UES = 1.5995, DES = –0.7734; ME/CFS: UES = 1.7592, DES = –1.078).

Cell type abundance. The abundance of cell populations was determined according to the expression of cell marker genes using Rosalind Bio. Hierarchical cluster analysis observations demonstrate heterogeneity within cohorts (Figure 4A). The abundance of exhausted CD8 cells was significantly lower in ME/CFS compared with HC (Padj = 0.0147). There were no significant differences in normal CD8 T cells reported. Furthermore, the abundance of Tregs was significantly lower in ME/CFS compared with long COVID (Padj = 0.0375) (Figure 4B). No other significance was reported; remaining abundance scores for cell types are shown in Supplemental Data 3.

Cell profiles and gene expression. (A) Heatmap extracted from Rosalind Bio. Cell type Z scores for cell populations are populated for samples collected from ME/CFS, long COVID, and HC. (B) Comparison of cell type abundance scores extracted from Rosalind Bio; statistical analysis and the figure were completed using GraphPad Prism. Exhausted CD8 cells were compared using Kruskal-Wallis test with Dunn’s multiple-comparison corrections. Tregs were compared using 1-way ANOVA with Bonferroni’s multiple-comparison test. Graphs show mean with minimum and maximum ranges. Data are presented as mean with maximum and minimum range. *P < 0.05. HC, healthy control; LC, long COVID.

Pathways and disease functions. Ingenuity Pathway Analysis (IPA) was used to determine the association of differentially expressed genes with biological functions and canonical pathways for both ME/CFS and long COVID cohorts when compared with HC (Table 6). The top 5 biological functions in long COVID include abnormal morphology of lymphocytes (P < 0.0001), activation of leukocytes (P < 0.0001), immediate hypersensitivity (P < 0.0001), activation of antigen-presenting cells (P < 0.0001), and lack of lymphocytes (P < 0.0001). Only 2 of the abovementioned biological functions overlap with the ME/CFS cohort: the activation of leukocytes (P < 0.0001) and the activation of antigen-presenting cells (P < 0.0001). The remaining top biological functions in ME/CFS include the formation of rosettes (P < 0.0001) (cell type not specified), immune response of cells (P < 0.0001), and activation of myeloid cells (P < 0.0001).

Top biological functions and pathways in long COVID and ME/CFS

Top 5 canonical pathways differed between long COVID and ME/CFS, excluding the macrophage alternative activation signaling pathway (P < 0.0001 and P < 0.0001, respectively). Canonical pathways identified in long COVID included antigen presentation (P < 0.0001), autoimmune thyroid disease signaling (P < 0.0001), allograft rejection signaling (P < 0.0001), and B cell development (P < 0.0001). Meanwhile, IL-12 signaling and production in macrophages (P < 0.0001), primary immunodeficiency signaling (P < 0.0001), role of macrophages, fibroblasts and endothelial cells in rheumatoid arthritis (P < 0.0001), and neutrophil extracellular trap signaling pathways (P < 0.0001) were reported in ME/CFS. HLA-DQA1, HLA-DQB1, and IGHG1 were found to be pivotal in the top biological pathways and diseases in long COVID, which overlapped with ME/CFS with the addition of IGHG3, CCL2, CEACAM3, and IFNA6. Overlapping pathways and disease function were observed in Figure 3, B and C. The complete pathways and disease function output can be found in Supplemental Data 4.

Network analysis. Interaction network analysis was performed using IPA. This analysis demonstrates the interactions between molecules and the data set imported. One network was identified for long COVID (Figure 5), while 2 networks were identified for ME/CFS (Figure 6, A and B), of which network 1 (Figure 6A) obtained the highest score of 18. Network analysis in long COVID was associated with categories and disease or functions including gastrointestinal disease (chronic colitis, P < 0.0001), humoral immune response and protein synthesis (quantity of IgG1, P < 0.0001), immunological disease, injury or abnormalities (immediate hypersensitivity, P < 0.0001), and cell morphology and abnormalities (morphology of lymphocytes, P < 0.0001). Analysis for the highest scoring network in ME/CFS was associated with cellular assembly and organization (formation of rosettes, P < 0.0001), cellular development, hematological system development and function (maturation of Th cells, P < 0.0001), and cell-to-cell signaling and interactions (activation of antigen presenting cells, P < 0.0001). Network analysis outputs can be found in Supplemental Data 4.

Network analysis in long COVID. Gene interaction network map consisting of top filtered differentially expressed genes. Genes are organized according to subcellular space. Network analysis score = 13.

Network analysis in ME/CFS. Gene interaction network map consisting of top filtered differentially expressed genes. Genes are organized according to subcellular space. (A) Network 1 score = 18. (B) Network 2 score = 3.