Among 429 individuals screened between May 3, 2021 and February 2, 2022, 22 met inclusion criteria and consented to participate and 16 individuals completed the baseline survey and received a first dose of a COVID-19 vaccine. Two 2-week surveys and two 6-week surveys were excluded because they were submitted before an earlier survey time point or on the same day as another survey; thus, we included 14 2-week surveys, 14 6-week surveys, and 16 12-week surveys. 14 completed the 2-week post-vaccination survey on time, 14 completed the 6-week survey, and 16 completed the 12-week survey. People not enrolled had already received a vaccine, did not plan to be vaccinated, or could not travel to New Haven for biospecimen collection. The median age of the 16 included participants was 54 years (range 21–69), 13 (81%) were female, and 14 (88%) identified as Non-Hispanic White (Table 1). The median number of months from participants’ first onset of symptoms until completing the pre-vaccine survey was 7.2 months (Q1–Q3 4.5–13.9, min-max 2.5–19.6). The median number of months from participants’ first self-reported positive COVID-19 test until completing the pre-vaccine survey was 5.5 months (Q1–Q3 4.3–12.9, min-max 1.7–19.5). Two participants did not report positive tests but one reported hospitalization with COVID-19. Immunophenotyping assays were completed on a subset of 11 out of 16 participants because of loss to follow-up in biospecimen collection and instances of difficulty in blood draws where less than expected volumes were collected. All participants reported that they tested positive for COVID-19 at least once with most reporting a PCR-based test (n = 10, 62%). Three (19%) participants were previously hospitalized due to COVID-19 and 4 (25%) visited the hospital or were hospitalized for COVID-19 more than 2 weeks after onset of acute disease (2 of these 4 participants also reported being hospitalized due to COVID-19).
Pre-vaccination health and symptoms
At baseline, on participants’ worst days, 9 (56%) felt they were 50% or less of their health before COVID-19. On participants’ best days, 7 (44%) reported feeling 51–75% of their health before COVID-19. The median number of symptoms per participant before vaccination was 23 (Q1–Q3, 13.8–27). The median number of symptoms per participant that resolved before vaccination was 9 (Q1–Q3 5–15, min-max 1–28), and the most common symptoms experienced that had resolved before vaccination were cough, diarrhea, and persistent chest pain or pressure (n = 6 for each).
The fifteen most frequently reported symptoms that had not resolved before vaccination were brain fog (n = 13, 81%), fatigue (12, 75%), difficulty concentrating (11, 69%), difficulty sleeping (10, 63%), heart palpitations (9, 56%), shortness of breath or difficulty breathing (9, 56%), anxiety (8, 50%), memory problems (8, 50%), dizziness (7, 44%), feeling irritable (7, 44%), headache (7, 44%), inability to exercise or be active (7, 44%), nerve sensations (7, 44%), post-exertional malaise (7, 44%).
Post-vaccination changes in overall health
Eleven of 16 participants (69%) received the Pfizer-BioNTech vaccine (Comirnaty®), 3 (19%) received the Janssen vaccine as their first dose, and 2 (13%) received the Moderna vaccine (SpikeVax®). Nine of 13 participants (69%) recommended to receive a second dose in the primary series reported doing so (i.e., Janssen’s vaccine in the primary series was single dose). One participant was hospitalized for chest pain and myocarditis three days after receiving their first vaccine dose and again after their second dose. This participant reported being previously hospitalized soon after infection with probable myocarditis.
Two weeks after vaccination, 6 out of 14 participants with completed surveys reported their health was better (43%), 3 (21%) said their health was the same, 1 (7%) reported worse health, and 4 (29%) were not sure of a change (Fig. 1). At 6 weeks after vaccination, 11 out of 14 (79%) said their health was better than before vaccination, 2 (14%) reported the same health, and 1 (7%) reported worse health. The participant with worse health 2 weeks after vaccination reported better health at 6 weeks. At 12 weeks, 10 out of 16 (62%) reported better health, while 3 (19%) reported the same health and 3 (19%) reported worse health. Two participants who reported better health at 6 weeks reported worse health at 12 weeks, which we classified as transient improvement in subsequent analyses.
Participants were asked “Would you say that your overall health, as compared to your health before the vaccine, is worse, better, or the same?” at each post-vaccination survey. Data missing for n = 2 at 2 weeks and n = 2 at 6 weeks.
The median number of symptoms per participant before vaccination was 23 (Q1–Q3 14.5–27, min-max 3–43, n = 16). In terms of pain or discomfort associated with a symptom, the median (Q1–Q3) number of symptoms per participant that resolved at 2, 6, and 12 weeks was 4.5 (3–7.8), 4 (3–7), and 5 (3–12), respectively, and the number of symptoms per participant that improved but did not fully resolve was 6 (6–11), 9.5 (8.3–13.3), and 11 (5–13), respectively (Supplementary Fig. 1). The median (Q1–Q3) number of symptoms per participant that worsened at 2, 6, and 12 weeks was 3 (2.5–5), 3 (2.5–5), 3 (1–4.3), respectively. Similar trends were observed regarding impairment of social and family functioning associated with a symptom, though with a higher number of symptoms per participant resolving. Symptom changes by overall health at 12 weeks is shown in Supplementary Fig. 2. For the 15 most common symptoms experienced before vaccination, the number of participants whose symptoms resolved, improved, remained not an issue, remained an issue, or worsened are presented in Supplementary Data 1 and 2. The median number of symptoms that bothered participants “very much”, “quite a bit”, “somewhat”, “a little bit”, or “not at all” before and after vaccination are provided in Supplementary Table 2.
Symptom burden appeared to decrease after vaccination on both physical and social effect scales (Fig. 2). Before vaccination, the median physical effect score for all symptoms was 63 (Q1–Q3 37–83, min-max 12–137, n = 16) and the median social effect score was 35 (Q1–Q3 13–51, min-max 0–83, n = 16), where higher values represent worse symptom burden. Compared with before vaccination, the median physical effect score decreased to 42 (Q1–Q3 18–60, min-max 4–109, n = 14) at 2 weeks after the first COVID-19 vaccine dose, then 35 (Q1–Q3 16–46, min-max 8–87, n = 14) 6 weeks after vaccination, and 36 (Q1–Q3 14–43, min-max, 0–66, n = 16) 12 weeks after vaccination. At 2 weeks after the first COVID-19 vaccine dose, the median social effect score decreased to 26 (Q1–Q3 6–27, min-max 0–55, n = 14), then 21 (Q1–Q3 9–36, min-max 2–89, n = 14) 6 weeks after vaccination, and 17 (Q1–Q3 3–27, min-max 0–52, n = 16) 12 weeks after vaccination.
To measure physical effect of each symptom from a list of 125 symptoms, participants were asked, “While experiencing these symptoms, how much do/did they bother you in terms of discomfort or pain?” Similarly, to measure social effects, participants were asked, “After quarantine, how much does/did the symptom impair your social or family functioning compared to pre-COVID? Responses for each symptom were scored 0–4 (Supplementary Table 1) and summed for each participant. Boxplots show the distribution of responses, with points indicating the score for each participant; the central lines indicate the group median values, the top and bottom lines indicate the 75th and 25th percentiles, respectively, the whiskers represent 1.5× the interquartile range. Higher values suggest a greater symptom burden, and a value of 0 suggests no symptom burden Data missing for n = 2 at 2 weeks and n = 2 at 6 weeks.
SARS-CoV-2-specifc T-cells and antibody responses
To characterize the T-cell responses to SARS-CoV-2, sequencing of the CDR3 regions of T-cell receptor-β (TCR- β) chains was carried out. There was a significant increase in spike protein (Fig. 3a; P = 0.012, V1(Fig. 3b; P = 0.011) and V3 (Fig. 3c; P = 0.011) classifier scores at 6 weeks post-vaccination, which was indicative of an increase in SARS-CoV-2 specific T-cell clonal depth and breadth. By contrast and as expected, no significant differences were observed in classifier scores for non-spike protein TCRs with vaccination (Fig. 3d). There were some individuals who retained high SARS-CoV-2 specific TCR clonality at 12 weeks post-vaccination, however the differences in model scores were not statistically significant in comparison with pre-vaccination. There was a significant decrease in V3 classifier score at 12 weeks post-vaccination as compared to 6 weeks (P = 0.011), however this observation was not replicated using the V1 or spike specific classifier scores.
a Model scores and binary classifications are plotted against days post-vaccination using spike viral protein specific classifier, b COVID classifier version 1 (v1), c COVID classifier version 3 (v3), d Non-spike-specific protein classifier, e Line plots of matched anti-SARS-CoV-2 S1 IgG concentrations before, 6 and 12 weeks post-vaccination in Long COVID participants, f Line plots of matched anti-SARS-CoV-2 RBD IgG concentrations before, 6 and 12 weeks post-vaccination in Long COVID participants, g Line plots of matched anti-SARS-CoV-2 N IgG concentrations before, 6 and 12 weeks post-vaccination in Long COVID participants. The color codes denote the reported health status at 6 and 12 weeks post-vaccination, better at both timepoints [teal], no change at both timepoints [blue], better at 6 weeks and worse at 12 weeks, [purple] & worse at both timepoints [orange], h Line plots of matched anti-SARS-CoV-2 Epsilon variant reactivity scores against the Spike protein assessed by Rapid Extracellular Antigen Profiling (REAP), i Heatmap of REAP reactivities against 10 viral proteins namely, proteins belonging to common viral pathogens from Coronaviridae (human SARS-CoV-1 viruses), Herpesviridae families, and the Rubella vaccine protein. Each protein and each participant timepoint are represented as a row and a column respectively. The participant IDs are mentioned below each column and the numbers after decimal denote the collection timepoints after vaccination (6 weeks = 2; 12 weeks = 3). Statistical significance determined by Wilcoxon Rank tests and corrected for multiple testing using the Bonferroni method, j EBV p23 REAP scores among outcome groups. Significance was assessed using Kruskal–Wallis tests, k EBV gp42 REAP scores among outcome groups, l Hierarchical clustering of Spearman Rank correlation coefficients of TCR model scores, antibody concentrations and REAP scores at all three timepoints. Only adjusted p-values of <0.05 are mentioned in line plots and denoted by asterisks in heatmaps.
Next, SARS-CoV-2 antibody responses were evaluated. A significant increase in anti-S1 IgG (Fig. 3e; pre vs 6 weeks: P = 0.003, pre vs 12 weeks: P = 0.003) and anti-RBD IgG (Fig. 3f; pre vs 6 weeks: P = 0.003, pre vs 12 weeks: P = 0.003) levels at 6 weeks and 12 weeks post-vaccination was observed without significant rise in anti-N IgG levels (Fig. 3g). The anti-S1 and anti-RBD IgG antibody levels peaked at 6 weeks (median anti-S1 IgG: 8.8 × 104 ng/mL; median anti-RBD IgG: 5.0 × 105 ng/mL) with a marginal decrease at 12 weeks (median anti-S1 IgG: 5.8 × 104 ng/mL; median anti-RBD IgG: 2.2 × 105 ng/mL). To further validate the humoral responses attributed to vaccination, SARS-CoV-2 spike protein reactivities were assessed using REAP. Participant antibody reactivities against Beta, Delta, Epsilon, and Omicron variant RBD epitopes were independently evaluated. A significant increase in reactivity across all non-Omicron RBD epitopes at 6 weeks post-vaccination (Supplementary Fig. 3a–c) and the Epsilon variant across 6 and 12 weeks post-vaccination (Fig. 3h; pre vs 6 weeks: P = 0.011, pre vs 12 weeks: P = 0.023) was observed.
IgG responses to herpesviruses and autoantibodies to the extracellular proteome
Given that latent virus reactivation has been a hypothesis behind Long COVID pathobiology and evidence of recent Epstein-Barr Virus (EBV) reactivation has been reported30,39,40, anti-viral REAP reactivities against two families of common viral pathogens namely, Coronaviridae (human SARS-CoV-1 viruses) and Herpesviridae, were assessed. Rubella vaccine spike antigen served as internal control as no changes were expected in reactivities with COVID-19 vaccination. As expected, there was a significant increase in REAP scores against SARS-CoV-1 RBD upon vaccination at 6 weeks (Fig. 3i; P = 0.047). This increase was maintained at 12 weeks, despite not being statistically significant after multiple testing correction (P = 0.09). Herpesvirus reactivities varied across participants. However, no significant decrease in reactivities was observed post-vaccination among the herpesvirus antigens tested including EBV (Fig. 3i; Supplementary Data 3). Additionally, no differences in median reactivities were observed against EBV proteins p23 and gp42 across outcome groups at 6 and 12 weeks post-vaccination (Fig. 3j, k).
Next, given prior reports of elevated autoantibodies targeting the exoproteome in severe acute COVID-1941, we assessed for changes in extracellularly targeted autoantibodies during vaccination (Supplementary Fig. 4a). No difference in the number of autoantibody reactivities at baseline (Supplementary Fig. 4b) or in the mean REAP score delta, representing the change in autoantibody magnitude over time (Supplementary Fig. 4c), between the groups was observed. Overall, autoantibodies were stable over time during vaccination (Supplementary Fig. 4c–e), with the mean REAP score delta close to 0 for all groups. These results are in alignment with a previous report focusing on autoantibody dynamics during SARS-CoV-2 mRNA vaccination in healthy individuals without Long COVID42.
Correlation between SARS-CoV-2 specific TCR and antibody levels
To further evaluate the relation between SARS-CoV-2 specific TCR scores with antibody levels and to assess the concordance among the orthogonal methods of antibody detection, correlation analyses were carried out. Three distinct clusters emerged when distances were calculated based on correlation values among TCR classifier scores and anti-SARS-CoV-2 antibody concentration as well as between ELISA and REAP assays at different timepoints. Each cluster indicated that there was a general concordance in antibody levels using orthogonal methods and TCR scores based on Spearman’s r (rs) and unadjusted p-values (Fig. 3l, Supplementary Data 4 and 5). It was also observed that higher numbers of pre-vaccination SARS-CoV-2 specific TCR repertoire resulted in higher titers of antibodies both at pre-vaccination, 6- and 12-weeks post-vaccination along with an increase in spike protein specific TCR repertoire. Despite visually strong correlation patterns, due to the small sample size, only anti-SARS-CoV-2 S1 and anti-RBD antibody levels as detected by ELISA at pre-vaccination timepoint and at 12 weeks were statistically significant after multiple testing corrections (pre-vaccination: rs = 0.96, P = 0.021; 12 weeks post-vaccination: rs = 0.98, P = 0.003; Supplementary Data 6).
No significant differences were observed between post-vaccination increase in SARS-CoV-2 specific TCR classifier scores and improvement in overall health status. Similarly, no differences were also observed in self-reported health status and increase in anti-SARS-CoV-2 antibody.
Soluble immune mediators
To understand the impact of vaccination on the cytokine, hormone, and proteomic profiles of individuals with Long COVID, unsupervised hierarchical clustering of 162 analytes measured in their plasma was first conducted (Fig. 4a). Clustering analysis showed a consistent pattern in their plasma expression profiles at 6- and 12-weeks post-vaccination. Samples clustered by individual and not by timepoint post-vaccination, suggesting an entrenchment in the cytokine profile of each individual that was not significantly affected by vaccination.
a Unsupervised hierarchical clustering of plasma-derived analyte expression within the cohort for all three sample timepoints (pre-vaccination, 6 weeks post series completion, and 12 weeks post series completion). Color panel above heatmap shows the symptom outcome subgroup of each individual as indicated by the key. Samples for each individual are labeled by their sample code LC.R.HK.1.00XX.tX, where XX designates the patient ID and tX designates the timepoint (t1 = pre vaccination, t2 = 6 weeks post series completion, and t3 = 12 weeks post series completion). Sample label color indicates further categorization into Same/Worse (orange), Transient improvement (i.e., better then worse; purple), and Improved (teal). Color scale is magma and is normalized for each analyte (data table columns) with darker colors indicating higher relative expression and lighter colors indicating lower expression as shown by the key. b Expression Heatmap of significant differentially expressed factors between symptom outcome groups (Same/Worse, Transient, and Improved), as labeled. Each subgroup is further separated by the vaccine timepoint. Each factor was centered and standardized to generate a z-score and colors are representative of expression as indicated by the legend (positive z-scores in red; negative z-scores in blue). To show significance between groups, samples were organized with outer brackets of the heatmap indicating the symptom outcome group demonstrating significantly lower expression and inner brackets indicating the comparator group from which significance is derived. Significance was determined using linear mixed models (LMM) via restricted maximum likelihood (REML) regression for log-transformed values, accounting for repeated measures across individuals over time as described in the methods and adjusted for multiple comparisons within each parameter using the Tukey method. c–h Example differentially expressed factors between symptom outcome groups as determined by LMM, previously described. i Top 20 bootstrapped predictors of symptom outcome (unimproved vs improved), determined by Partial Least Squares (PLS) optimized at eight components. Predictors are ordered by importance with highest importance on the left. Color and direction of each bar represents the relative regression association to unimproved individuals with positive values showing a positive association (red) and negative values showing a negative association (blue). Color is determined by regression as shown. Details of NIPALS and detailed results can be found in the methods and in extended data, respectively.
To understand the relationship of these plasma-derived analytes with post-vaccine symptom outcomes, the average expression levels of each analyte was compared over all three timepoints amongst three symptom outcome groups: those who did not improve or felt worse at weeks 6 and 12 post vaccination (n = 3; Same/Worse), those who showed transient improvement (n = 2, Transient [i.e. Better week 6; then Worse week 12]) and those who reported improvement (n = 7, Better). To do so a linear mixed model was constructed using restricted maximum likelihood (REML) regression for each cytokine and accounted for both time and the interaction of time with each outcome group. Thirty-five factors were found to be significant amongst these subgroups, with the majority being significantly elevated in the Same/Worse group compared with the Improved group. Type I interferons were higher at baseline and after vaccination amongst the Same/Worse group, including IFN-β and IFN-α, compared to the Improved group (Fig. 4b–d). Ciliary neurotrophic factor (CTNF; a neuropeptide that is released by the hypothalamus), IL-11, and stem cell factor (SCF) were also significantly elevated in the Same/Worse group compared to the Improved group (Fig. 4e). Other neuropeptides were elevated amongst the transient group including oxytocin, neurotensin, substance P and melanocyte stimulating hormone (MSH) (Fig. 4b). Notably, soluble IL-6 receptor (sIL-6R; an anti-inflammatory protein responsible for mitigating IL-6 signaling), was significantly higher amongst those who showed improvement compared to the Same/Worse group (Fig. 4h).
Further Partial Least Squares (PLS) analysis with 5-fold cross validation was employed on all 162 analytes to determine feature importance as predictors of symptom outcome and evaluate concordance with the significant features obtained from our LMM models. Final analysis involved reduction to 8 components, accounting for a sizeable portion of the variance in the data (cumulative pseudo-R-squared = 0.99). The top two significant predictors of the PLS analysis for non-improvement were IFN-β and ciliary neurotrophic factor (CNTF) respectively (Fig. 4i). The top significant predictor of improvement was sIL-6R (Fig. 4i), while sgp130, an important immunological partner to sIL-6R, was also associated with improvement, passing the initial VIP threshold criteria (Supplementary Fig. 5), though not the additional bootstrapping threshold criteria. Taken together these results suggested that high IFN and neuropeptide signaling were predictors of non-improvement while those involved in mitigating cytokine signaling, namely sIL-6R, was a predictor of improvement.
