key: cord-0872395-opjo5kdx
authors: Holmes, Elaine; Wist, Julien; Masuda, Reika; Lodge, Samantha; Nitschke, Philipp; Kimhofer, Torben; Loo, Ruey Leng; Begum, Sofina; Boughton, Berin; Yang, Rongchang; Morillon, Aude-Claire; Chin, Sung-Tong; Hall, Drew; Ryan, Monique; Bong, Sze-How; Gay, Melvin; Edgar, Dale W.; Lindon, John C.; Richards, Toby; Yeap, Bu B.; Pettersson, Sven; Spraul, Manfred; Schaefer, Hartmut; Lawler, Nathan G.; Gray, Nicola; Whiley, Luke; Nicholson, Jeremy K.
title: Incomplete Systemic Recovery and Metabolic Phenoreversion in Post-Acute-Phase Nonhospitalized COVID-19 Patients: Implications for Assessment of Post-Acute COVID-19 Syndrome
date: 2021-05-19
journal: J Proteome Res
DOI: 10.1021/acs.jproteome.1c00224
sha: b20b5009a97ecef7a1c3f92f311dd46bd2b6dca7
doc_id: 872395
cord_uid: opjo5kdx

[Image: see text] We present a multivariate metabotyping approach to assess the functional recovery of nonhospitalized COVID-19 patients and the possible biochemical sequelae of “Post-Acute COVID-19 Syndrome”, colloquially known as long-COVID. Blood samples were taken from patients ca. 3 months after acute COVID-19 infection with further assessment of symptoms at 6 months. Some 57% of the patients had one or more persistent symptoms including respiratory-related symptoms like cough, dyspnea, and rhinorrhea or other nonrespiratory symptoms including chronic fatigue, anosmia, myalgia, or joint pain. Plasma samples were quantitatively analyzed for lipoproteins, glycoproteins, amino acids, biogenic amines, and tryptophan pathway intermediates using Nuclear Magnetic Resonance (NMR) spectroscopy and mass spectrometry. Metabolic data for the follow-up patients (n = 27) were compared with controls (n = 41) and hospitalized severe acute respiratory syndrome SARS-CoV-2 positive patients (n = 18, with multiple time-points). Univariate and multivariate statistics revealed variable patterns of functional recovery with many patients exhibiting residual COVID-19 biomarker signatures. Several parameters were persistently perturbed, e.g., elevated taurine (p = 3.6 × 10(–3) versus controls) and reduced glutamine/glutamate ratio (p = 6.95 × 10(–8) versus controls), indicative of possible liver and muscle damage and a high energy demand linked to more generalized tissue repair or immune function. Some parameters showed near-complete normalization, e.g., the plasma apolipoprotein B100/A1 ratio was similar to that of healthy controls but significantly lower (p = 4.2 × 10(–3)) than post-acute COVID-19 patients, reflecting partial reversion of the metabolic phenotype (phenoreversion) toward the healthy metabolic state. Plasma neopterin was normalized in all follow-up patients, indicative of a reduction in the adaptive immune activity that has been previously detected in active SARS-CoV-2 infection. Other systemic inflammatory biomarkers such as GlycA and the kynurenine/tryptophan ratio remained elevated in some, but not all, patients. Correlation analysis, principal component analysis (PCA), and orthogonal-partial least-squares discriminant analysis (O-PLS-DA) showed that the follow-up patients were, as a group, metabolically distinct from controls and partially comapped with the acute-phase patients. Significant systematic metabolic differences between asymptomatic and symptomatic follow-up patients were also observed for multiple metabolites. The overall metabolic variance of the symptomatic patients was significantly greater than that of nonsymptomatic patients for multiple parameters (χ(2)p = 0.014). Thus, asymptomatic follow-up patients including those with post-acute COVID-19 Syndrome displayed a spectrum of multiple persistent biochemical pathophysiology, suggesting that the metabolic phenotyping approach may be deployed for multisystem functional assessment of individual post-acute COVID-19 patients.

The COVID-19 pandemic continues, and although infection rates are currently reduced in many locations and multiple vaccines are now being deployed, the severe acute respiratory syndrome SARS-CoV-2 virus continues to evolve, yielding more infectious variants that are likely to maintain high infection rates.

There is increasing concern about the alarmingly high number of patients not recovering from SARS-CoV-2 infections, resulting in "Post-Acute COVID-19 Syndrome (PACS)" with multiple persistent and highly variable symptoms (Table 1) . A recent meta-analysis found that up to 80% of patients display one or more post-acute COVID-19 symptoms including fatigue (58%), headache (44%), attention disorder (27%), hair loss (25%), and dyspnea (24%), with joint pain (27%) and chest pain (22%) being among the most common postinfection symptoms. 1, 2 Postinfection lung damage is a problem with several groups reporting 50% or more of patients with lung structure abnormalities and shortness of breath (dyspnea, up to 70% of patients), 3, 4 long-term neurological complications including chronic fatigue (17.5−87% of patients 5 ), cognitive impairment (up to 78% 6 ), and anosmia/hyposmia (8.5−25% 7 ) being reported, as well as cardiomyopathies or other structural cardiac abnormalities (up to 60% 8 ); liver dysfunction; 9 disruption of the microvasculature; skin rashes 10 and discoloration and a range of autoimmune phenomena. 11 Diabetes has been proposed as a severity risk factor, but sadly, the considerable stress imposed on the body following the acute infection has resulted in new-onset diabetes mellitus in a number of patients. 12 Based on multiple study observations including our own, the metabolic sequelae of SARS-CoV-2 infection are complex, engaging multiple-organ dysfunction presumed to be driven by immunopathological events including, for example, cytokine storms. 13, 14 Related coronavirus diseases, Middle East respiratory syndrome (MERS) and SARS, have also been shown to induce longterm metabolic changes in recovered patients pertaining to hypertriglyceridemia, disorders of glucose metabolism, and liver function, further underlying the considerable need to support recovery and repair of the body during the post-COVID period. 15, 16 The similarities between the SARS-CoV-1 and SARS-CoV-2 outbreaks also extend to the neurological consequences, with 54% of patients experiencing fatigue and 60% experiencing sleep disturbances 6 months post infection. 17 In some patients recovering from severe infection, residual postmedication treatment side effects have been reported: (i) hydroxychloroquine and cardiotoxicity; and (ii) long-term use of painkillers and renal and gastrointestinal dysfunction. 18 However, patients who suffered from less-severe manifestations of COVID-19 constitute the largest proportion of the active or post-acute COVID-19 population, now over 160 million worldwide, with few follow-up studies published to date. Within this mild-symptom COVID-19 group, hypometabolism in multiple brain regions 19 and persistent elevation of plasma Ddimer and C-reactive protein have been reported. 20 As is the case with most diseases, mildly affected patients are normally expected to show better functional recovery than seriously affected patients, but there is growing evidence that even mild cases can have persistent effects even a year post acute infection. 17, 21 A recent comprehensive review of post-acute COVID-19 syndrome presents the key persistent manifestations of the disease and illustrates the enormous potential consequences in terms of global healthcare burden and individual suffering. 22 Given the variability of clinical presentation and severity of the disease, recovery from COVID-19 is likely to be heterogeneous, in terms of both time to recovery and completeness of recovery. Severity of COVID-19 has so far been mainly classified in relation to respiratory symptoms in the acute stage, but the complex systemic symptoms require further consideration, with respect to both classifying the acute illness and assessing patient functional recovery. We and others have observed metabolic heterogeneity in the biomarker responses in acute COVID-19 patients showing multiple systemic effects. 23−28 We postulate that full functional recovery from the disease, by most of the patients without pre-existing comorbidities, is likely to be reflected in a normalization of the disrupted sentinel biomarker parameters and the overall metabolic phenotype. The process of multilevel recovery can be considered to be functionally equivalent to the reverse of disease onset and progression "phenoconversion", 23, 24 during which the measured metabolic profile shifts from normality to one typical of the disease or its subphenotypes. From a theoretical point of view, patient "phenoreversion" therefore would be partial or complete, reflecting the degree of systemic recovery from COVID-19 infection. Going forward, this framework which employs sentinel biomarkers and appropriate multivariate models for mapping recovery from COVID-19 can be used to objectively assess other immunologically driven multisystem diseases.

The expected heterogeneity of the recovery responses for a complex system disease such as COVID-19 requires a nonstandard approach to understand and interpret the biochemical profiles of PACS individuals. There are precedents for metabolic modeling of single-organ or complex multiorgan damage and partial recovery in experimental toxicology studies where the onset, progression, and recovery or nonrecovery from toxicity can be followed using biofluid metabolic models. 47 Multiple studies using exemplar toxins that damage different organ systems in experimental animals at various times post dosing have been documented, 48−50 and some studies have shown sequential organ toxicity or recovery trajectories. 51, 52 In such situations, functional biochemical recovery may occur at different rates and vary between individuals. 53 We hypothesize that because SARS-CoV-2 infections frequently result in multiorgan involvement the functional recovery of patients might show similar complex temporal patterns and variations in biochemical normalization through time and that study of these processes can provide an objective functional classification of recovery.

To assess the longer-term effects of COVID-19 infection in nonhospitalized patients, we obtained blood plasma samples at approximately 3 months following the acute phase of the disease and recorded persistent symptoms of patients at 6 months as well as their original symptoms during the acute phase of the disease. We adopted a targeted approach in which validated analytical methods were used to focus on metabolic parameters that had previously been described as biomarkers (Tables S1− S4) in the active phase of COVID-19. 23, 26, 54 This approach enables quantitative phenoreversion to be assessed and patients with incomplete functional recovery to be identified. We examined these variables from both univariate and multivariate modeling perspectives to investigate individual and group recovery behaviors. The data were evaluated with respect to a range of terms of possible phenoreversion outcomes. The principal aim of this study was to develop the basis of an objective metabolic framework for measuring systemic recovery in COVID-19 patients using a range of metabolic technologies and biomarkers.

Three cohorts were examined: (i) a healthy control group (n = 41); (ii) a hospitalized patient group (n = 18 with multiple timepoints) sampled during the acute infection phase; and (iii) a recovery cohort consisting of a nonhospitalized group sampled 3 months post the acute phase of the disease (n = 27) with acutephase symptoms and symptoms 6 months post COVID infection. The participant demographics and the numbers of patients and samples used for each assay are provided in Tables S5 and S6. From the recovery cohort, 57% of the participants recorded one or more persistent symptoms, with the majority having more than one symptom. A graphical description of the follow-up symptomatic patients in relation to the acute-phase symptoms is shown in Figure 1 . In addition to persistent respiratory symptoms including cough, dyspnea, and rhinorrhea, there were also nonrespiratory "systemic" symptoms including fatigue, anosmia, myalgia, and joint pain, which are indicative of wider systematic damage and inflammatory components, highlighting the systemic nature of post-acute COVID-19 syndrome. Many of these longer-term symptoms have been reported in previous studies as itemized in Table 1 , suggesting that this group is similar to other follow-up populations, bearing in mind that in the current case these were previously nonhospitalized patients. Note, however, that with the exception of anosmia, many of these symptoms are common to aging populations, and without an accurate framework of assessment, we cannot be entirely certain of the direct association of these symptoms with PACS. On balance, the fact that these symptoms were measured in the same participants during and post infection should partially standardize the collection of metadata. Thus, we aimed to develop an independent metric of PACS and recovery that does not rely purely on reported symptoms. 1 H NMR Spectroscopy of Blood Plasma 1 H NMR spectroscopy was completed according to a previously published method 55 briefly described as follows: blood was centrifuged at 13 000g for 10 min at 4°C and the plasma supernatant was mixed with buffer (75 mM Na 2 HPO 4 , 2 mM NaN 3 , 4.6 mM sodium trimethylsilyl propionate-[2,2,3,3-2 H 4 ] (TSP) in 80% D 2 O, pH 7.4 ± 0.1) (1:1); 600 μL of each plasma/buffer mixture was transferred into a Bruker SampleJet NMR tube (5 mm) and sealed with POM balls added to the caps. For quality control samples, long-term reference samples were used, prepared using the same methods as for the study samples. A validated sampling preparation and handling protocol for COVID-19 samples was used as previously described. 56 NMR measurements were performed on one of two Bruker 600 MHz Avance III HD spectrometers (in vitro diagnostics research (IVDr)) equipped with a BBI probe and fitted with Bruker SampleJet robots with the cooling system set to 5°C. A quantitative calibration was completed prior to the analysis using the protocol of Dona et al. 55 All experiments were completed using the Bruker in vitro diagnostics research (IVDr) methods. 57 For each sample, three experiments were completed in automation with a total analysis time of 12.5 min: a 1 H onedimensional (1D) experiment with solvent presaturation (32 scans, 98 304 data points, spectral width of 18 028.85 Hz), a Carr−Purcell−Meiboom−Gill (CPMG) spin-echo experiment (32 scans, 73 728 data points, spectral width of 12 019.23 Hz), and a two-dimensional (2D) J-resolved experiment (two scans with 40 t 1 increments). All data were processed in automation using Bruker Topspin 3.6.2 and ICON NMR to achieve phasing, baseline correction, and calibration to TSP (δ = 0). On completion of the run, the samples were transferred directly to the second NMR spectrometer and DIffusional and Relaxation Editing (DIRE) spectra were obtained (64 scans, 98 304 data points, spectral width of 18 028.85 Hz). 54 Lipoprotein reports itemizing 112 lipoprotein parameters for each plasma sample were generated using the Bruker IVDr Lipoprotein Subclass Analysis (B.I.LISA) method. 57 The α-1-acid glycoprotein N-acetyl-glucosamino (N-acetyl) signal integrals were calculated as GlycA from the superimposition of terminal N-acetyl signals (δ 2.03) and GlycB was calculated from branched-chain N-acetyl signals (δ 2.07) determined by the DIRE spectra. The supramolecular phospholipid composite (SPC total ) signals (δ = 3.20−3.30) were determined from the DIRE spectra by integration, a method previously described and applied to COVID-19 studies. 54 The SPC total parameter consists of the N + −(CH 3 ) 3 choline headgroup signals of lysophosphatidylcholines carried out on plasma glycoproteins and from phospholipids in highdensity lipoprotein subfractions together with a phospholipid component of LDL.

Fully quantitative amino acid analysis of 35 molecular species was performed using an ultra-high-performance liquid chromatography (UHPLC)-quadrupole time-of-flight (QToF)-based MS method following derivatization. 58 Amino acid standards and ammonium acetate were purchased from Sigma-Aldrich (MO). Stable-isotope-labeled (SIL) internal standard noncanonical and canonical amino acid mixes were purchased from Cambridge Isotope Laboratories (MA). Water, acetonitrile, methanol, and isopropanol were all Optima-grade (from Thermo Fisher Scientific). Calibrators and quality controls were prepared from a standard solution of physiological amino acids (acids, basics, and neutrals) at 500 μM. Asparagine and glutamine were prepared freshly at 0.5 mM on the day of analysis due to their solution instability. A working aqueous stock solution of all amino acids was prepared at 400 μM and separately diluted to 200, 100, 40, 20, 10, 4, 2, and 1 μM for calibrators and to 300, 75, 15, and 3 μM for analytical quality controls. A stable-isotope-labeled (SIL) internal standard solution (12.5 μM in water) was prepared from stocks of canonical and noncanonical amino acids at 2.5 mM in water and stored at −20°C until use. Following the addition of the SIL working solution to each sample, methanol was added to effect protein precipitation. Following centrifugation, the extract supernatant was taken for a derivatization step with the AccQTag reagent (Waters Corp., Milford, MA). Samples were finally diluted 1:50 with liquid chromatography (LC)-MS-grade water for LC-MS analysis.

Amino acid analysis was performed using a Waters Acquity Iclass UPLC system (Waters Corp., Milford, MA) coupled to a Bruker impact II QToF mass analyzer (Bruker, Daltonics, Billerica, MA). Chromatographic separation was achieved using an Acquity UPLC HSS T3 1.8 μm 2.1 × 150 mm 2 column (Waters, Milford, MA). Eluent A consisted of 2 mM ammonium acetate in water, and eluent B consisted of 2 mM ammonium acetate in acetonitrile/water 95/5 (v/v). The flow rate was 0.6 mL/min, and the column temperature was maintained at 45°C. The autosampler compartment was cooled to 4°C, and a 2 μL injection volume was performed using the full-loop injection mode. Gradient elution was performed starting with 5% B for 0.2 min, increasing to 30% B at 5 min and 100% B at 5.1 min for 1 min before returning to 5% B until 7.5 min. The weak and strong washes were water/acetonitrile 95/5 (v/v) and isopropanol, respectively. The mass analyzer was operated with positive electrospray ionization (ESI) in the broadband collisioninduced dissociation (bbCID) mode, enabling MS and tandem mass spectrometry (MS/MS) spectral data collection within the same run. The ion source settings were as follows: capillary voltage = 4.5 kV; end plate offset = 500 V; drying gas flow = 12.0 L/min; nebulizer gas = 5.0 bar; drying temperature = 250°C. The data acquisition rate was set to 8 Hz over the mass range of m/z 30−1000. The collision energy for the MS scan was set to 6.0 eV, and alternating low and high energies for MS/MS were set at 20 and 50 eV. An internal calibration was performed by injection of 5 mM aqueous sodium formate/isopropanol (50:50 v/v) at the beginning of every run. Mass spectrometric data were collected with Compass HyStar 5.1 and O-TOF Control version 5.2. MS data were processed using TASQ 2.2 (Bruker Daltonics, Bremen, Germany). The molecular formula or exact mass of the derivatized amino acid was used to extract the precursor ions with a mass error of <3 mDa. Calibration curves were linearly fitted with a weighting factor of 1/x.

Quantitative analysis of tryptophan and 14 catabolites in its metabolic pathway was performed as previously reported. 26, 59 Analysis was performed on 50 μL of plasma, with the addition of SIL prior to protein precipitation with methanol containing 2 mM ammonium formate. After mixing, samples were transferred to a Phenomenex PHREETM phospholipid removal solid-phase extraction plate (Phenomenex, NSW, Australia). PHREE plates were washed with methanol containing 2 mM ammonium formate, and the eluent was dried using a SpeedVac vacuum concentrator (Thermo Fisher, MA). Extracts were resuspended in 100 μL of aqueous 0.1% formic acid prior to LC-MS analysis.

LC-MS analysis was performed on a Waters Acquity UPLC (Waters Corp., Milford, MA) coupled to a Waters Xevo TQ-XS MS (Waters Corp., Wilmslow, U.K.). Separation was performed on a Waters HSS T3 1.8 μm, 2.1 × 150 mm 2 column maintained at 45°C. Linear gradient elution was performed at 0.6 mL/min. The mobile phase was composed of 0.1% formic acid in 2 mM ammonium formate (v/v) (A) and 0.1% formic acid in acetonitrile (v/v) (B), starting at 1% B, increasing to 10% B over 3 min, and then increasing to 90% B at 4 min, and finally returning to 1% B at 4.1 min for column re-equilibration, which was completed at 5 min. The weak and strong washes were 95:5 water/acetonitrile (v/v) and 100% isopropanol, respectively. Mass spectrometric data were collected with MassLynx version 4.2 and processed using the TargetLynx package (Waters Corp., Milford, MA).

Following a standard suite of NMR and targeted MS assays, 34 biogenic amines 59 and 15 tryptophan pathway parameters 60 were measured by mass spectrometry and 112 lipoprotein parameters, and glucose, 23 plus glycoproteins, and novel supramolecular phospholipid composite biomarkers previously identified and measured by NMR spectroscopy. 54 In addition, we calculated multiple ratios. Based on modeling these data in relation to SARS-CoV-2 positivity, we selected 45 discriminatory primary parameter measurements and 7 ratios (see Tables S1−S4) to evaluate the extent or recovery of the nonhospitalized patients in relation to the SARS-CoV-2 positive and the healthy control groups. 

Based on a selected subset of parameters that have been shown to be significantly perturbed in the active COVID-19 cases, 23, 26 we estimated the functional recovery of individuals and the 3 month follow-up cohort as a whole. The full parameters sets were evaluated are shown in Tables S1−S4. Journal of Proteome Research pubs.acs.org/jpr Article substantially not normalized, such as taurine in the follow-up patients, which was significantly different to healthy controls (Mann−Whitney test p = 3.6 × 10 −3 ) and the glutamine-toglutamate ratio also significantly different to control (Mann− Whitney test, p = 6.95 × 10 −8 ). For several parameters, the majority of the participants had normalized with serum concentrations falling within the range for healthy participants but leaving a small subset of participants with concentrations or ratios in the same range as the active disease COVID-19 patient group. Parameters that were substantially normalized included GlycA, GlycB, total plasma cholesterol (TPCH), and the kynurenine/tryptophan and SPC/GlycA signal ratios, but in all cases, there were patient outliers that had not normalized underscoring the heterogeneity of response to SARS-CoV-2 infection. The HDL parameters H4A1, H4A2 (apolipoproteins A1 and A2 in HDL subfraction 4), and kynurenine were partially normalized, and the mean of the follow-up cohort lay between the means of the healthy and the SARS-CoV-2-infected groups with approximately equal numbers of participants with values in the range for the healthy and infected ( Figures S10, S30 , and S31). The radar plots for individual patients and participants ( Figure 3 ) clearly show that the SARS-CoV-2 group is defined by elevated plasma levels of tryptophan pathway metabolites. Plasma neopterin levels were markedly elevated in active-phase COVID-19 infections as previously reported, 23,62 but the followup patients were essentially at control neopterin levels in this cohort. Neopterin is generated in macrophages and microglia under the influence of the tumor necrosis factor (TNF)-α and γinterferon 62 and is a marker of early-stage cell-mediated acute inflammation and as such would be expected to recover soon after the acute inflammatory phase. 63 A similar pattern in neopterin levels was observed following infection with SARS-CoV-1, where neopterin levels were directly associated with fever. 64 Further insights come from inspection of the "radar" parameter plots for individuals in each group where each individual is connected across all their measured parameters as shown in Figure 3 . There are notable differences in the individual shapes for the groups with biomarker pointers in each set. Interestingly, plasma concentrations of 3-indole acetic acid were higher in the recovery cohort 3 months after infection than in the infected or healthy groups (Figures 3 and S9) , suggestive of a gut microbiome functional change. Indoles are known to activate the aryl-hydrocarbon-receptor (AhR), a ligandactivated transcription factor that has been shown to reduce inflammation and stimulate metabolic circuits by integrating signals from the host, the microbes, and dietary intake. 65 Additional microbe indole-derived metabolites are, for example, indole-3-acetylaldehyde and indole-3-aldehyde, both known to reduce inflammation via AhR signaling in the intestine and the liver as well as to strengthen barrier function in the intestine. 66 Several other metabolites behave in an analogous fashion with near-complete recovery for all patients (Tables S1−S4). For example, the kynurenine/tryptophan ratio ( Figure 2B ) is also substantially within the normal range for most patients 3 months post infection, with the exception of three individuals, who must be assumed to be still in a heightened inflammatory state. This is consistent with data showing acute systemic inflammatory activity during SARS-CoV-2 infection. 23, 26, 67, 68 The ratio of tryptophan to kynurenine is predominantly mediated by indoleamine 2,3-dioxygenase activity and is generally associated with inflammation as well as immunomodulation in both infectious diseases and chronic inflammatory conditions, including neurological diseases such as Parkinson's disease, 69 diabetes, 70 and several autoimmune conditions such as rheumatoid arthritis, 71 systemic lupus erythematosus, 72 and primary sclerosing cholangitis. 73 Plasma taurine concentrations were also increased in the acute phase of COVID-19 23 but also appear substantially elevated in the 3 month follow-up patients with respect to controls ( Figure  2C ). Elevated taurine levels have previously been associated with liver injury and hepatotoxicity in man and experimental animals, 74 but high levels of taurine are also present in large quantities in cardiac and skeletal muscles 75, 76 and muscle damage has been associated with taurinaemia as a measure of tissue damage. 77 As COVID-19 causes considerable skeletal muscle mass loss, 78 it is possible that high plasma taurine may simply reflect muscle turnover. We noted that Fischer's ratio, a marker of liver injury, was substantially normalized for most follow-up patients ( Figure S5 ). However, persistent liver dysfunction with elevated ALT and AST has also been noted post COVID-19 in other studies ( Table 1 ). The relative sensitivity for Fischer's ratio versus plasma taurine versus serum transaminases as markers of liver injury severity has not been evaluated to our knowledge. Macroscopic cardiac damage is also a known persistent consequence of acute SARS-CoV-2 infection (Table 1) , and it is possible that high plasma taurine is reflective of longer-term multiple-organ damage including the liver, heart, and skeletal muscle components, but this remains speculative. Given that the nonhospitalized patients studied here were relatively mildly affected in the acute phase, these observations are a cause for concern and warrant a larger-scale study of the persistence of elevated plasma taurine in COVID-19 patients and its clinical significance.

Acute-phase COVID-19 patients show elevated plasma glutamate ( Figure S6 ) and reduced glutamine levels 23, 26, 79 and a low glutamine/glutamate ratio, which is also substantially higher in the 3 month follow-up patients ( Figure 2D ). This ratio is driven both by elevated plasma glutamate relative to controls (Mann−Whitney test p = 1.04 × 10 −3 versus controls, Figure  S6 ) and by reduced plasma glutamine relative to controls (Mann−Whitney test p = 7.00 × 10 −5 ). Glutamate is an important anaplerotic carbon source to the citric acid cycle 80 and thus plays major important cellular energy generation roles in multiple organs including the kidney, heart, and brain. Both renal and neurological effects of COVID-19 are now welldocumented in the acute phase, 81 and again these data pose questions about the possible role of these metabolites in COVID-related organ dysfunction. Furthermore, glutamate is the major excitatory amino acid in the central nervous system (CNS) and elevated CNS levels can induce acute injury, 82 with the possible significance of this observation being amplified by knowledge that tryptophan metabolites 3-hydroxykynurenine and quinolinic acid are elevated in SARS-CoV-2 infections 23, 26 and both are also excitatory neurotoxins. 83 Quinolinate and kynurenate (elevated in most acute COVID-19 patients) also modulate glutamate function, with increased plasma quinolinate resulting in cellular release of glutamate. 84 However, quinolinic acid levels are mainly normalized in the follow-up patients so they cannot be directly driving the glutamine/glutamate ratio. The relationship between glutamate and glutamine is important in immune cell homeostasis, especially T-cells and astrocytes within the CNS, 85 where it acts as an immunomodulator upregulating surface glutamate receptors on activation, and the rate of immune cell glutamine consumption increased in catabolic conditions including sepsis in which plasma glutamine Journal of Proteome Research pubs.acs.org/jpr Article levels may also be low. 86 So although the mechanistic significance of a low glutamine/glutamate ratio is not yet understood for COVID-19, the persistence of a significantly reduced ratio in the majority of the follow-up patients implies continuing post-COVID immunometabolic dysregulation. Diabetes is a known risk factor for early progression to severe COVID-19 disease, 87 and new-onset diabetes has also been reported in COVID-19 patients. 41 Plasma glucose, a well-known marker of type 2 diabetes, was elevated in active-phase patients ( Figure 2E) , and high blood glucose on hospital admission has been associated with poor prognosis. 88 The 3 month follow-up patients appear to have a glucose distribution similar to that of the control patients (none of whom were diabetic) although there were several elevated plasma glucose outliers in both groups. Our tentative inference from these data is that diabetes does not appear to be a major persistent effect of SARS-CoV-2 infection in this nonhospitalized patient cohort. However, this statement cannot be generalized to cover other COVID-19 patient severity classes because the data set is relatively small and the patients were not hospitalized. Further studies on COVID-19 follow-up patients are required to evaluate the persistence of diabetes in patients who were more severely affected during the acute stage of the disease.

Lipoprotein dysregulation is a notable feature of acute COVID-19 disease. 24, 89, 90 The NMR spectroscopy-derived apolipoprotein B100/A1 ratios for the three groups are shown in Figure 2F . We have previously reported elevations in this ratio in COVID-19 patients, as have others. 24, 27 The clinical significance of this ratio is that it is generally predictive of long-term atherosclerosis and coronary artery disease 91 and has previously been associated with a range of other cardiometabolic diseases. 92 All of the follow-up patients (except one) appear to be at the control level for this parameter. Other lipoprotein parameters apolipoproteins A1 and A2 were also normalized, respectively, and mostly normalized ( Figures S47 and S48 ) although their exact relationships varied according to the degree of persistent symptoms (see below). The NMR-derived inflammatory markers GlycA and GlycB and the SPC/GlycA ratio were measured from the diffusion and relaxation-edited NMR spectra and are markedly altered in acute COVID-19 patients. 54 The GlycA and GlycB signals from acute-phase reactive proteins (mainly from α-1-acid glycoprotein but with contributions from other glycoproteins) 93 are strongly elevated in the active phase of COVID-19 and several follow-up patients still had elevated levels ( Figure 2G,H) . The SPC/Glyc ratio was reduced in active disease, and likewise, several follow-up patients had reduced ratios ( Figure 2I ). The SPC peak is a supramolecular cluster with several lysophosphatidylcholines and choline species being present bound to HDL and LDL and the α-1-acid glycoprotein. 54 This provides a metric for detection and discrimination of SARS-CoV-2 positivity and would appear to provide a potential measure of recovery from the COVID-19related immunopathology. A reduction in plasma HDL cholesterol has been shown to be associated with COVID-19 infection. 23, 24 The majority of patients showed normal total plasma cholesterol after 3 months, with one or two outliers that were still significantly reduced ( Figure 2J ).

The patient signatures and variations of individual metabolic parameters in the COVID-19 follow-up patients shown in Figure 2 do not demonstrate the full system complexities of the metabolic interactions in the acute patient group. Differences between group behaviors can be investigated by considering the correlation matrix structures of individual sentinel marker data as shown in Figure 4A (left) (healthy versus COVID-19 acute phase) and Figure 4B (right) (follow up versus healthy patients). The asymmetry of the paired matrices observed across the diagonals is an indicator of the systemic differences in metabolic relationships within the groups. Thus, if full functional recovery had occurred in the follow-up patients, the two correlation matrices would be substantially equivalent and, therefore, symmetrical. This is clearly not the case, and so as a group across multiple variables, the follow-up patients were not fully metabolically recovered. The glutamine/glutamate relationships, the liver-related plasma metabolic biomarkers (elevated taurine and low citrulline), and the tryptophan−kynurenine pathway intermediates (high quinolinic acid, kynurenine, 3hydroxykynurenine) were also noticeably different in the three groups as indicated in the earlier univariate data.

To investigate further the multivariate metabolic relationships in the patient groups, we performed principal component analysis (PCA) using three different composite windows on metabolism involving amino acids and tryptophan targeted assays, lipoproteins (targeted), and concentrations of GlycA, GlycB, and SPC and the ratio GlycA/SPC estimated by integrating DIRE spectra, as previously described 54 (Figure 5A −C). Each metabolic window reveals different degrees of variable recovery because they cover different features of the systemic disease. PCA is primarily used here as a simple visualization tool to view group clustering behavior among SARS-CoV-2-positive (red), follow-up (blue), and healthy (green) patients and thus cannot be used in a diagnostic sense, merely as another indicator of general recovery. Figure 5A shows that most follow-up patients were substantially normalized with respect to their overall amino acid profiles with just a few outliers. However, PCA plots do not show all of the variance in the data set and are a crude clustering tool to visualize complex and heterogeneous sets. Lipoprotein data from the same patient samples also show substantial normalization in most patients but with multiple patients at or near the interface between the acute and healthy groups ( Figure  5B ). This is consistent with the intermediate nature of the correlation matrices shown in Figure 3 . The extracted diagnostic features from the DIRE-NMR spectra from the same samples are shown in Figure 5C , with follow-up patient data points being scattered across the healthy and acute metabolic space. We have shown previously that the GlycA, GlycB, SPC, and SPC/Glyc ratios are highly diagnostic for SARS-CoV-2-positive versus controls and SARS-CoV-2 negative respiratory patients. These are from acute-phase reactive glycoproteins, mainly α-1-acid glycoprotein (GlycA and B peaks) and the supramolecular phospholipid composite (SPC total ) peak from phospholipids bound to glycoproteins or HDL and LDL subfractions that reflect underlying immunopathological processes. 54 Thus, Figure 5C also illustrates the incomplete immunometabolic phenoreversion of a subset of follow-up patients as revealed by the SPC/GlycA inflammatory markers.

Having demonstrated the differential clustering behavior by PCA, we applied O-PLS-DA to determine the predictive components that separated the healthy from the follow-up participants with respect to both targeted MS data ( Figure 6A ,B) and lipoproteins ( Figure 6C,D) . The metabolic properties that distinguished these groups were mainly similar to the properties that have been reported previously as separating healthy from positive active-phase patients with higher glutamic acid, lower glutamine/glutamate ratios, higher kynurenine/tryptophan ratios, higher LDL parameters, and lower HDL parameters. Thus, a residual multivariate signature of the disease was still present at three months after the acute phase. In addition, indole-3-acetic acid also marked a difference present in the follow-up patients but was not a strong biomarker of the acutephase disease, 23, 26 as illustrated in the radar plot shown in Figure  3 .

Examination of univariate plots of individual parameters revealed marked differences between the statistical distributions of data for follow-up patients who were symptom-free or symptomatic (for any clinical parameter) at 6 months. In particular, the observed variance in the symptomatic patient data appeared generally greater, suggesting a greater heterogeneity of biochemical states in patients with persistent symptoms. To assess this quantitatively, we performed a series of F-tests on the variances between the parameters of asymptomatic and symptomatic patients, two of which reached absolute statistical significance, viz. H3FC ( Figure S28 ), free cholesterol in HDL fraction 3, p = 0.027; and TPCH ( Figure S49 ), total plasma cholesterol, p = 0.017. However, we observed that, taken as a group, in the majority of cases, the lipoproteins' parameter variance was still greater for symptomatic than asymptomatic patients to a degree that seemed unlikely by chance alone. We confirmed this by applying a χ 2 test to these data, which gave a value of p = 0.014. This indicates that collectively the parameter variances were indeed significantly different between symptomatic and asymptomatic patients. This is not a predictive model, and although there are indications that follow-up metabolic profiles carry data that are potentially predictive of long-term symptomatic outcomes, this needs to be explored in larger patient data sets.

Further studies are required to evaluate the detailed long-term metabolic consequences of patients who suffered more significant acute effects and required interventions such as oxygen therapy and assisted ventilation than in the current nonhospitalized group where long-term effects might be less severe. There are several literature reports of multiple clinical symptoms in COVID-19 patients extending 6 months to a year beyond hospitalization, 17 and serious accompanying metabolic abnormalities would be unsurprising. Long-term metabolic effects of SARS-CoV-1 have been reported up to 12 years post the critical illness phase; these included liver dysfunction and hypertrigylceridemia, 94 and it is a possibility that COVID-19 will create similar long-term outcomes. The original SARS-CoV-1 virus affected 8098 patients worldwide; 95 given that the COVID-19 pandemic has now affected over 141 million people, if disability were then scaled to the same level as seen with SARS, such long-term effects would represent a major increase in global healthcare burdens and associated healthcare costs. Thus, it is imperative to implement large-scale follow-up studies to assess PACS healthcare liabilities in each population and also to understand the long-term immunopathological drivers or disease and to help identify patients with post-acute COVID-19-syndrome healthcare needs to mitigate any persistent disorders and optimize patient recovery.

Metabolic phenotyping appears to be a highly effective tool for assessing phenoreversion and functional systemic recovery of patients following COVID-19 and the evaluation of post-acute COVID-19 syndrome. There are clear metabolic patterns associated with persistent inflammatory and metabolic derangements in a majority of nonhospitalized 3-month-post-COVID-19 patients, with one or more symptoms being persistent in 57% up to 6 months following the acute phase. Most metabolic parameters showed a high level of normalization. Other biomarker parameters, such as plasma taurine and the glutamine/glutamate ratios, showed little normalization across most patients, and the long-term clinical significance of these observations requires further investigation. There was a high degree of interindividual variability in the follow-up patients, which may reflect the heterogeneity of PACS as an emergent clinical syndrome. At present, it is unclear whether PACS is the terminal extension of the COVID-19 disease trajectory or indeed the emergence of collateral onset of additional separate disease entities. The observation of increased levels of 3-indole acetic acid implies a microbiome functionality shift in the postacute COVID-19 patients; this may also be pertinent to understanding the tryptophan pathway perturbations in PACS patients given the general importance of the microbiome in tryptophan metabolism and serotonin production. Larger studies with greater ranges of observed severity should cast further light on the ability to determine predictors of PACS and the individual patterns of longer-term symptoms exhibited by patients. Heterogeneous metabotype patterns were observed for the PACS patient spectrum, and if reflective of the clinical picture, these could enable personalized interpretation of the persisting inflammatory/metabolic abnormalities and give insights into strategies for long-term disease mitigation in affected individuals. Worldwide, over 160 million have been infected by the SARS-CoV-2 virus so far, with 138 million socalled "recovered" patients, of whom many tens of millions are likely to have PACS with multiple-organ involvements. It is therefore important to find metabolic markers that illuminate possible public healthcare "time-bombs" for future medicoeconomic burdens that may be faced by many societies as a result of widespread exposure of populations to the disease. The current data and models provide the basis for a framework of assessment of the extent of multifunctional biochemical patient recovery from COVID-19. With further refinement, including addition of other metabolic system parameters, new biomarker panels can be assembled to assess patients with PACS and assist in clinical decision making in their long-term management and disease mitigation. The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.1c00224.

List of parameters considered from targeted MS assays (Table S1 ); list of parameters considered from IVDr lipoproteins (Table S2) ; list of parameters considered from IVDr small molecules (Table S3) ; list of parameters considered from DIRE-NMR experiments (Table S4) ; demographic information about the population of the recovery cohort (Table S5) ; description of the number of samples (patients) included in each method (Table S6) ; univariate statistics: 3-hydroxykynurenine ( Figure S1 ), 3hydroxykynurenine/kynurenic acid ( Figure S2 Figure S18 ); correlation using the healthy population ordered according to hierarchical clustering obtained from healthy control ( Figure S19 ); correlation using the positive population ordered according to hierarchical clustering obtained from healthy control ( Figure S20 ); correlation using the recovered population ordered according to hierarchical clustering obtained from healthy control ( Figure S21 Figure S50 ); correlation using the healthy population ordered according to hierarchical clustering ( Figure S51 ); correlation using the positive population ordered according to hierarchical clustering obtained from healthy control ( Figure S52 ); correlation using the recovered population ordered according to hierarchical clustering obtained from healthy control ( Figure S53 ); univariate statistics: glucose ( Figure S54 ), GlycA ( Figure  S55 ), GlycB ( Figure S56 ), SPC total ( Figure S57 ), and SPC total /GlycA ( Figure S58 ) (PDF)

More Than 50 Long-Term Effects of COVID-19: A Systematic Review and Meta

Gemelli Against COVID-19 Post-Acute Care Study Group. Persistent Symptoms in Patients After Acute COVID-19

Persistent Symptoms 3 Months after a SARS-CoV-2 Infection: The Post-COVID-19 Syndrome?

Medium-Term Effects of SARS-CoV-2 Infection on Multiple Vital Organs, Exercise Capacity

Long-COVID": A Cross-Sectional Study of Persisting Symptoms, Biomarker and Imaging Abnormalities Following Hospitalisation for COVID-19. Thorax

Frequent Neurocognitive Deficits after Recovery from Mild COVID-19

COVID-19: Recovery from Chemosensory Dysfunction. A Multicentre Study on Smell and Taste

Outcomes of Cardiovascular Magnetic Resonance Imaging in Patients Recently Recovered From Coronavirus Disease 2019 (COVID-19)

Liver Function Recovery of COVID-19 Patients after Discharge, a Follow-up Study

Long COVID in the Skin: A Registry Analysis of COVID-19 Dermatological Duration

New-Onset Diabetes in Covid-19

Cytokine Storm Induced by SARS-CoV-2 Infection: The Spectrum of Its Neurological Manifestations

The Cytokine Storm in COVID-19: An Overview of the Involvement of the Chemokine/chemokine-Receptor System

Long-Term Sequelae of SARS: Physical, Neuropsychiatric, and Quality-of-Life Assessment

Long-Term Sequelae Following Previous Coronavirus Epidemics

6-Month Consequences of COVID-19 in Patients Discharged from Hospital: A Cohort Study

Non-Steroidal Anti-Inflammatory Drugs (NSAIDs) and Organ Damage: A Current Perspective

18F-FDG Brain PET Hypometabolism in Patients with Long COVID

Prolonged Elevation of D-Dimer Levels in Convalescent COVID-19 Patients Is Independent of the Acute Phase Response

The Lancet. Facing up to Long COVID

Integrative Modelling of Quantitative Plasma Lipoprotein, Metabolic and Amino Acid Data Reveals a Multi-Organ Pathological Signature of SARS-CoV-2 Infection

NMR Spectroscopic Windows on the Systemic Effects of SARS-CoV-2 Infection on Plasma Lipoproteins and Metabolites in Relation to Circulating Cytokines

Proteomic and Metabolomic Characterization of COVID-19

Systemic Perturbations in Amine and Kynurenine Metabolism Associated with Acute SARS-CoV-2 Infection and Inflammatory Cytokine Responses

Metabolomic/lipidomic Profiling of COVID-19 and Individual Response to Tocilizumab

Longitudinal Clinical and Radiographic Evaluation Reveals Interleukin-6 as an Indicator of Persistent Pulmonary Injury in COVID-19

Persistent Symptoms 1.5-6 Months after COVID-19 in Non-Hospitalised Subjects: A Population-Based Cohort Study

The COronavirus Pandemic Epidemiology (COPE) Consortium: A Call to Action

Cardiac Involvement in Consecutive Elite Athletes Recovered from Covid-19: A Magnetic Resonance Study

Thyrotoxicosis Following SARS-CoV-2 Infection

A Case of Post-COVID-19 Subacute Thyroiditis

Subacute Thyroiditis Associated with COVID-19

Persistent Hypokalemia Post SARS-coV-2 Infection, Is It a Life-Long Complication? Case Report

The Clinical Course of Smell and Taste Loss in COVID-19 Hospitalized Patients

Prevalence and Correlates of Chronic Fatigue Syndrome and Post-Traumatic Stress Disorder after the Outbreak of the COVID-19

Clinical Characteristics and Short Term Outcomes after Recovery from COVID-19 in Patients with and without Diabetes in Bangladesh

Proportion of Newly Diagnosed Diabetes in COVID-19 Patients: A Systematic Review and Meta-Analysis

Persistent Hallucinations in a 46-Year-Old Woman After COVID-19 Infection: A Case Report

COVID-19 Myocarditis and Postinfection Bell's Palsy

Reactive Arthritis After COVID-19 Infection

COVID-19 as a Trigger of Recurrent Guillain-BarréSyndrome. Pathogens

Late Onset of Guillain-BarréSyndrome Following SARS-CoV-2 Infection: Part of "Long COVID-19 Syndrome

Nuclear Magnetic Resonance Spectroscopic and Principal Components Analysis Investigations into Biochemical Effects of Three Model Hepatotoxins

Geometric Trajectory Analysis of Metabolic Responses to Toxicity Can Define Treatment Specific Profiles

Nuclear Magnetic Resonance Spectroscopy and Pattern Recognition Analysis of the Biochemical Processes Associated with the Progression of and Recovery from Nephrotoxic Lesions in the Rat Induced by mercury(II) Chloride and 2-Bromoethanamine

Elucidating the Time-Dependent Changes in the Urinary Metabolome under Doxorubicin-Induced Nephrotoxicity

Acute Renal Failure Following Accidental Cutaneous Absorption of Phenol: Application of NMR Urinalysis to Monitor the Disease Process

NMR Spectroscopy as a Novel Approach to the Monitoring of Renal Transplant Function

Metabonomics: A Platform for Studying Drug Toxicity and Gene Function

Diffusion and Relaxation Edited Proton NMR Spectroscopy of Plasma Reveals a High-Fidelity Supramolecular Biomarker Signature of SARS-CoV-2 Infection

Precision High-Throughput Proton NMR Spectroscopy of Human Urine, Serum, and Plasma for Large-Scale Metabolic Phenotyping

Quantitative In-Vitro Diagnostic NMR Spectroscopy for Lipoprotein and Metabolite Measurements in Plasma and Serum: Recommendations for Analytical Artefact Minimization with Special Reference to COVID-19/SARS-CoV-2 Samples

Quantitative Lipoprotein Subclass and Low Molecular Weight Metabolite Analysis in Human Serum and Plasma by (1)H NMR Spectroscopy in a Multilaboratory Trial

A Validated UPLC-MS/MS Assay for the Quantification of Amino Acids and Biogenic Amines in Rat Urine

A Simultaneous Exploratory and Quantitative Amino Acid and Biogenic Amine Metabolic Profiling Platform for Rapid Disease Phenotyping via UPLC-QToF-MS

Ultrahigh-Performance Liquid Chromatography Tandem Mass Spectrometry with Electrospray Ionization Quantification of Tryptophan Metabolites and Markers of Gut Health in Serum and Plasma-Application to Clinical and Epidemiology Cohorts

Analysis: Combining the Strengths of PLS-DA and SIMCA Classification

Neopterin Predicts Disease Severity in Hospitalized Patients With COVID-19

Serum Neopterin for Early Assessment of Severity of Severe Acute Respiratory Syndrome

Bidirectional Communication between the Aryl Hydrocarbon Receptor (AhR) and the Microbiome Tunes Host Metabolism

Indoles: Metabolites Produced by Intestinal Bacteria Capable of Controlling Liver Disease Manifestation

Cellular Events of Acute, Resolving or Progressive COVID-Journal of Proteome Research pubs

Increased Kynurenine-to-Tryptophan Ratio in the Serum of Patients Infected with SARS-CoV2: An Observational Cohort Study

Association of Tryptophan Metabolites with Incident Type 2 Diabetes in the PREDIMED Trial: A Case-Cohort Study

Latest Developments in Tryptophan Metabolism: Understanding Its Role in B Cell Immunity

Kynurenine Pathway Is Altered in Patients with SLE and Associated with Severe Fatigue

Associations of Neopterin and Kynurenine-Tryptophan Ratio with Survival in Primary Sclerosing Cholangitis

NMR-Based Metabonomic Analysis of the Hepatotoxicity Induced by Combined Exposure to PCBs and TCDD in Rats

New Implications for an Old Amino Acid

Taurine and Skeletal Muscle Function

Huk, I. Ischemia/reperfusion Injury of Skeletal Muscle: Plasma Taurine as a Measure of Tissue Damage

Cabello-Verrugio, C. SARS-CoV-2/Renin-Angiotensin System: Deciphering the Clues for a Couple with Potentially Harmful Effects on Skeletal Muscle

Understanding the Pathophysiological Changes via Untargeted Metabolomics in COVID-19 Patients

A Simple Mathematical Model and Practical Approach for Evaluating Citric Acid Cycle Fluxes in Perfused Rat Hearts by 13C-NMR and 1H-NMR Spectroscopy

Endogenous Neurotoxins from Tryptophan

Tryptophan Metabolites Are Associated With Symptoms and Nigral Pathology in Parkinson's Disease

Astrocytes Maintain Glutamate Homeostasis in the CNS by Controlling the Balance between Glutamate Uptake and Release

High Admission Blood Glucose Independently Predicts Poor Prognosis in COVID-19 Patients: A Systematic Review and Dose-Response Meta-Analysis

Altered High-Density Lipoprotein Composition and Functions during Severe COVID-19

Association of Serum HDL Cholesterol and Apolipoprotein A1 Levels with Risk of Severe SARS-CoV-2 Infection

The Apolipoprotein B/apolipoprotein A-I Ratio as a Potential Marker of Plasma Atherogenicity

Associations of Apolipoprotein B/apolipoprotein A-I Ratio with Pre-Diabetes and Diabetes Risks: A Cross-Sectional Study in Chinese Adults

Characterization of 1H NMR Plasma Glycoproteins as a New Strategy To Identify Inflammatory Patterns in Rheumatoid Arthritis

Altered Lipid Metabolism in Recovered SARS Patients Twelve Years after

The authors declare no competing financial interest. The original data of the follow up cohort can be found at DOI:10.5281/zenodo.4713391.