This is a single-centered cohort and consisted of 60 PCR-confirmed SARS-CoV-2 infected individuals (30 LC patients, 15 individuals without LC symptoms who were previously infected with SARS-CoV-2 but recovered without any symptoms per se (recovered group, denoted R), 15 acute hospitalized COVID-19 patients (A)) as well as 15 healthy controls (HCs) without acute symptoms and negative SARS-CoV-2 immunoassays ( Figure 1A , Supplementary Table 1 ). Study participants were age -and sex-matched ( Figure 1B ) and all were infected with the Wuhan strain and recruited at ~12 months post the onset of acute infection ( Supplementary Table 1 ). Considering that the majority of our LC and R participants had a mild acute infection, confounding health conditions were not common ( Supplementary Table 1 ).

HCs were HIV, HCV, and HB seronegative, and mainly recruited pre-COVID-19 pandemic.

A key strength of this study was our unique focus on LC patients who met classification criteria for ME/CFS, which was not rigorously performed in any previous study. Those LC patients who did not fulfill ME/CFS criteria were excluded. Our LC patients were evaluated based on diagnostic criteria developed by CDC and WHO (30, 31) for ME/CFS. Moreover, we used a set of validated clinical questionnaires to capture the severity of symptoms for ME/CFS. This was conducted through in-person meetings using 7 different questionnaires including 1) De Paul Symptom Questionnaires (DSQ); 2) Functional Assessment of Chronic Illness Therapy (FACIT) Fatigue scale questionnaire; 3) FACIT-Dyspnea; 4) Fibromyalgia (FM) diagnostic criteria; 5) Cognitive Failure Questionnaire (CFQ); 6) The Pittsburgh Sleep Quality Index (PSQI): 7) The Hospital Anxiety and Depression Scale (HADS). We evaluated each patient based on different components of each questionnaire. For instance, the DSQ is used to assess symptoms related to ME/CFS (32). In this survey, patients were asked to rate 54 symptoms, and based on their responses we characterized them into 6 groups: fatigue (I), malaise (II), sleep difficulties (III), pain (IV), cognitive/neurological manifestations (V), and other (VI) which included autonomic, neuroendocrine, and immune manifestations. At the same time, the frequency of symptoms was evaluated using a 5-point Likert scale with 0: none of the time, 1: a little of the time, 2: about half the time, 3: most of the time, and 4: all of the time. Similarly, the severity of the symptoms was also recorded with 0: no symptom, 1: mild, 2: moderate, 3: severe, and 4: very severe. Each patient was considered positive if scored frequency and severity of ≥2 on at least one symptom from each category. For neurologic/cognitive manifestations, patients were considered positive if they scored a frequency and severity of ≥2 on at least two symptoms. For other categories (category VI), patients were considered positive if they had a minimum of one symptom from two of the autonomic, neuroendocrine, and immune manifestations. The diagnosis of ME/CFS was made if the patients met the criteria for categories I, II, III, IV, V, and VI (32–34).

A similar evaluation approach was utilized for the other questionnaires mentioned above. Thus, based on the obtained information from these questionnaires and clinical analyses, we characterized our LC patients.

A total of 2584 metabolites were detected from 75 plasma samples including LC n=30, acute n=15, R n=15, and HC n=15. First, we compared the whole metabolome in our four study groups. Partial least squares-discrimination analysis (PLS-DA) and the heatmap showed a clear classification of the metabolomes between four groups ( Figures 2A, B ). The acute COVID-19 group showed clearly a distinct metabolome profile from the other three groups. Once the acute group was excluded from analysis, the LC group displayed a more distinct metabolome profile compared to the HC and R groups ( Figures 2C, D ).

While the LC and HC groups were separated, it appeared that the R group was located between these two groups ( Figure 2C ). Next, we characterized the difference at the single metabolite level between groups. Volcano plots showed 478 metabolites were significantly reduced while 294 metabolites were significantly increased in acute (A) COVID-19 patients compared to HCs ( Figure 2E ). Compared to HCs, 103 metabolites were elevated but 123 were declined in the R group ( Figure 2F ). Interestingly, the number of altered metabolites were much lower between LC and HC groups, showing 61 and 76 decreased and increased, respectively ( Figure 2G ). The number of altered metabolites between acute patients and the R group were very similar to those of acute/HCs with 419 increased and 285 decreased metabolites ( Figure 2H ). Compared to acute patients, we noted 495 metabolites were significantly reduced but 167 metabolites were elevated in the plasma of LC patients ( Figure 2I ). Finally, compared to the R group, LC patients had 49 reduced but 111 increased metabolites ( Figure 2J ). When we compared the top 100 altered metabolites between groups, we found that acute COVID-19 patients had a distinct metabolomic profile from HCs and R groups ( Supplementary Figures 1A, B , Supplementary Table 2 ). Even though in general R individuals were segregated from HCs, three R patients assimilated with HCs ( Supplementary Figure 1C , Supplementary Table 2 ). Next, we compared the pattern of metabolomic profile in LC patients versus the R and HC groups. These analyses indicated that when the top 100 altered metabolites were compared the LC group was segregated from both the R and HC groups ( Supplementary Figures 2A, B , Supplementary Table 2 ). Overall, these observations demonstrate that LC and even R individuals present an altered metabolomic profile even ~12 months post the acute disease onset.

To obtain further insight into the differential metabolic characteristics of LC from R and HCs, we performed a metabolic pathway enrichment analysis (https://www.metaboanalyst.ca/) (35–37). Compared to HCs, these analyses revealed that 18 pathways were significantly altered in LC patients ( Figure 3A ). These alterations in metabolic pathways, resulted in a significant decline in the levels of aspartate, uracil, serine, sarcosine, arginine, dehydroalanine, thymine, and porphobilinogen in LC patients versus HCs ( Figure 3B , Supplementary Figure 3A ). However, the levels of other metabolites such as 5-Aminolevulinate, cysteate, putrescine, 4-Aminobutyraldehyde, kynurenine, serotonin, Formyl-5-hydroxykynurenamine, 5-Hydroxykynurenine, 2-Aminomuconate, xanthine, and 5-Aminolevulinate were significantly increased in LC patients ( Figure 3C , Supplementary Figure 3A ). Given the role of Metabolic pathway analyses between the LC and R group revealed also the alteration of 18 pathways, however, 15 out of these 18 pathways were similar to those observed between LC versus HCs ( Figure 4A ). The three differentially altered pathways between the LC vs HCs were glutamine, nitrogen, and butanoate mechanisms whereas cysteine and methionine, taurine, and glutathione metabolisms were altered between the LC and R groups ( Figure 4A ). Overall these alterations in metabolism pathways resulted in a substantial reduction in the levels of metabolites such as aspartate, asparagine, glutamine, histidine, N-formimino-glutamate, sarcosine, and ethanolamine phosphate in LC patients ( Figure 4B , Supplementary Figure 3B ). In contrast, metabolites such as 4-Aminobutanoate, 4-aminobutyraldehyde, 3-hydroxyanthranilate and porphobilinogen were elevated in the plasma of LC patients compared to the R group ( Figure 4C , Supplementary Figure 3B ).

Finally, we conducted a metabolic enrichment pathway analysis between the R and HCs. Although these individuals did not have any clinical symptoms and appeared to be healthy, they exhibited alterations in 15 metabolomics pathways compared with HCs ( Figure 5A ). Interestingly, 13 of these pathways were similar to those altered pathways between LC and HCs. Of note, the other two altered pathways in the R group were vitamin B and glutathione metabolisms ( Figure 5A ). These 15 altered metabolomics pathways resulted in increased levels of glutamate, 4-Aminobutanoate, carnosine, thymine, porphobilinogen, homogentisate and 4-Guanidinobutanoate in the R group ( Figure 5B , Supplementary Figure 4A ). At the same time, compared to HCs, the levels of a wide range of metabolites such as glutamine, asparagine, N-Formimino-glutamate, putrescine, serotonin, kynurenine, 2-Aminomuconate, Cys-Gly, 5-Aminolevulinate, and pyridoxine were reduced in the plasma of R group ( Figure 5C , Supplementary Figure 4A ). Despite the lack of clinical symptoms, the R group also exhibited a differential metabolomics profile compared to HCs. Taken together, these observations imply substantial metabolic pathways alterations in LC patients.

In agreement with previous reports (38), we observed alteration in plasma metabolite levels of acute patients vs HCs. Pathway analysis between acute and HCs revealed alteration in 29 pathways as indicated in Supplementary Figure 4B . These modifications in metabolism pathways resulted in a substantial increase in the levels of 48 metabolites such as Xanthine, ascorbate, Cys-Gly, kynurenine, phenylalanine, 5-hydroxykynurenine, and others as outlined in Supplementary Figures 5A, B . In contrast, we found that these metabolomic alterations led to a reduction in the plasma level of different metabolites including uridine, thymine, carnosine, theophylline, aspartate, taurine, serine, glutamate, tryptophan, arginine, and 53 other metabolites in the acute patients compared to HCs ( Supplementary Figures 6A, B ).

We also compared a metabolomic profile of those recovered from COVID-19 infection with the acute stage of the disease. Pathway analysis between acute and R indicated alteration in 36 pathways as indicated in Supplementary Figure 7 . These modifications in metabolism pathways resulted in a substantial increase in the levels of 39 metabolites in acute COVID 19 patients compared to the R group such as phenylaniline, pyrimidodiazepine, xanthine, ascorbate, kynurenine, 5-hydroxykynurenine and others as outlined in Supplementary Figures 8A, B . In contrast, we found that these metabolomic alterations led to a reduction in the plasma level of 66 different metabolites in acute patients versus the R group ( Supplementary Figures 9A-C ). Finally, we compared the metabolomic profile of LC vs acute COVID-19 patients. These analyses revealed alteration in 30 metabolism pathways ( Supplementary Figure 10 ). These changes in metabolism pathways were associated with a significant increase in 29 metabolites as shown in Supplementary Figure 11 . In contrast, we observed a significant reduction in 16 metabolites in LC patients compared to those with acute COVID-19 disease ( Supplementary Figure 12 ).

Considering the alteration in a variety of pathways such as tryptophan-kynurenine and Xanthine/hypoxanthine, we quantified the levels of inflammatory cytokines. We found a significant increase in the plasma IL-1α, IL-6, TNF-α, IP-10, and liver-associated active phase proteins (CRP and SAA) in LC patients versus R and HCs ( Figures 6A-F ). However, the plasma IL-13 and IL-1β concentrations remained unchanged between groups ( Figures 6G, H ). Although we did not find any significant difference between soluble Flt-1, also known as vascular endothelial growth factor receptor 1, levels in the plasma of R and LC patients, LC group exhibits a significant plasma elevation compared to the HC cohort ( Figure 6I ). We also quantified sCD14, a nonspecific monocyte activation marker, levels in the plasma of our study participants, which showed a significant elevation in LC patients compared to the HC and R groups ( Figure 6J ). In light of hypocalcemia in COVID-19 patients and its association with hypoparathyroidism (39, 40), we detected significant levels of the anti-CaSR antibody (calcium receptor) in LC cohort ( Figure 6K ). This observation forced us to measure the soluble form of CaSR in our different cohorts. Interestingly, we observed the elevation of soluble/shed CaSR in the plasma of acute COVID-19 patients, which was not the case in LC patients ( Figure 6L ). Finally, considering the impact of COVID-19 infection on the purinergic system (41), we quantified the plasma levels of ATP in our different cohorts. We found a significant reduction in the plasma ATP levels in acute COVID-19 patients compared to HCs and R groups ( Figure 6M ). Notably, LC patients had a substantially reduced level of ATP in their plasmas compared to either R and HCs ( Figure 6M ). Intriguingly, the R groups exhibited a significant elevation in the plasma ATP compared to the HC and R groups ( Figure 6M ). Overall, these observations imply a dysregulated and inflammatory immune response in LC patients.

In our investigations of metabolite levels associated with symptoms commonly seen in LC patients such as cognitive function, anxiety, chronic pain, and depression in LC patients, we conducted additional analyses. Despite finding no association between serotonin, tryptophan, aspartate, and kynurenine plasma levels with these symptoms, interestingly, we observed a subtle inverse correlation between sarcosine concentrations and both cognitive failure function scores ( Figure 7A ) and depression scores in LC patients ( Figure 7B ). Moreover, our observations revealed a similar association between the plasma serine levels and anxiety and depression scores in LC patients ( Figures 7C, D ). These initial findings suggest a potential link between reduced levels of sarcosine and serine metabolites and some of the clinical symptoms seen in LC patients.

Sixty PCR-confirmed SARS-CoV-2 infected individuals comprised of 30 LC patients, 15 individuals without LC symptoms who were previously infected with SARS-CoV-2 but recovered without any symptoms per se (recovered group, denoted R), 15 acute hospitalized COVID-19 patients) as well as 15 healthy controls (HCs) without acute symptoms and negative SARS-CoV-2 immunoassays ( Figure 1A , Supplementary Table 1 ). All individuals were infected with SARS-CoV-2 Wuhan viral strain in 2020, which was confirmed by PCR at University of Alberta Hospital, Edmonton. Both LC and R groups were recruited ~12 months after the onset of SARS-CoV-2 infection ( Supplementary Table 1 ).

This study was approved by the Human Research Ethics Board (HREB) at the University of Alberta (protocol # Pro00099502). A written informed consent form was obtained from all participants but a waiver of consent was obtained for those admitted to the ICU.

Plasma samples were collected following centrifugation of fresh blood samples and kept frozen at -80 until use.

Frozen plasma samples were thawed at RT, then were centrifuged at 600 g for 5 min. The 100 μl of soluble fraction was transferred into a new tube and mixed with 300 μl of LC-MS grade methanol. The samples were stored at -20°C for 30 min, then were centrifuged at 16,260 g to let precipitate proteins. The supernatants were transferred into new tubes and dried up completely. The dried samples were reconstituted by dissolving in 85 μl of water. The metabolomic analysis was performed by following the Chemical Isotope Labeling (CIL) LC-MS protocol described in our previous report (81). Briefly, each individual sample was labeled with ¹²C₂-dansyl chloride. The pooled sample was generated by mixing each of the individual samples, and labeled by ¹³C₂-dansyl chloride. After mixing each ¹²C-labeled sample with an equal volume of the ¹³C-labeled pool, the ¹²C-/¹³C- mixtures were injected onto LC-MS for analysis. The LC-MS system was Agilent 1290 LC linked to Brukler Impact II QTOF Mass spectrometer. The samples were injected into an Agilent Eclipse Plus reversed-phase C18 column (2.1 mm × 150 mm, 1.8 μm particle size, 95 Å pore size) for separation. Solvent A was 0.1% (v/v) formic acid in water, and solvent B was 0.1% (v/v) formic acid in acetonitrile. The chromatographic conditions were: t= 0 min, 25% B; t= 10 min, 99% B; t= 15 min, 99% B; t= 15.1 min, 25% B; t= 18 min, 25% B. The flow rate was 400 μL/min. All MS spectra were obtained in the positive ion mode. The MS conditions used for Q-TOF were as follows: nebulizer, 1.0 bar; dry temperature, 230°C; dry gas, 8 L/min; capillary voltage, 4500 V; end plate offset, 500 V; spectra rate, 1.0 Hz. The raw data were exported by Bruker Data Analysis 4.4. as.csv files, and the.csv files were processed by IsoMS Pro 1.2.15. The ¹²C/¹³C- peak pairs were extracted, and missing values in the data matrix were filled. Metabolite identification was performed by searching against libraries with different confidence levels. Tier 1 library (DnsID library) contains accurate mass, MS/MS, and retention time information; Tier 2 library (Linked ID library) contains accurate mass and predicted retention time information; Tier 3 library (My Compound ID library, www.MycompoundID.org) contains accurate mass of predicted metabolites. IsoMS Pro was also used for performing the univariate analysis (volcano plot) and multivariate analysis (PCA and PLS-DA). The data file containing Tier 1 metabolites was uploaded to Metaboanalyst 5.0 (www.metaboanalyst.ca) for generating Heatmap. Pathway enrichment analysis was performed by using Metaboanalyst 5.0.

Frozen plasma samples at -80 °CC were thawed and centrifuged for 15 min at 1500g followed by dilution for quantifying cytokine and chemokine profiles, respectively. The concentration of cytokines and chemokines was quantified using the V-PLEX from Meso Scale Discovery (MSD). Data were acquired on the V-plex^® Sector Imager 2400 plate reader. Analyte concentrations were extrapolated from a standard curve calculated using a four-parameter logistic fit using MSD Workbench 3.0 software according to reported protocols (82–84).

The plasma concentration of ATP was quantified using the ATPlite luminescence assay system (PerkinElmer, MA) as we have reported elsewhere (85). The plasma was subjected to ELISA kit sCD14 (R&D, 383CD-050) (80) and Anti-CaSR (EAGLE Biosciences).

We initially determined the distribution of data using the Wilks-Shapiro test and then based on the distribution of data the appropriate test was used. When data were not normally distributed the non-parametric tests such as the Mann-Whitney U-test or Kruskal–Wallis analysis of variance were used. Depending on the data set different post hoc tests were used. The Dunn’s multiple comparisons as the post hoc test for the Kruskal–Wallis analysis. The Tukey-Kramer test was used for one-way ANOVA with multiple comparisons. For nonparametric correlational studies tow-tailed Spearman correlation was used. P-values are shown in the graphs and measures are expressed as mean ± SEM and P-value < 0.05 was considered to be statistically significant. In violin plots, the middle line represents median, the bottom line 1^st quartile, and to top line 3^rd quartile. No randomization was performed and no data points were excluded.