The emergence of the novel coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has resulted in an unprecedented global health crisis. While the acute phase of COVID-19 primarily manifests as respiratory symptoms, a certain portion of the population experiences persistent or newly onset symptoms weeks or months after the resolution of acute infection; this condition is now termed post-acute sequelae of COVID-19 (PASC) [1]. PASC encompasses a broad range of symptoms affecting multiple organ systems, including but not limited to fatigue, dyspnea, cognitive impairment, and myalgia, significantly impacting the quality of life of recovered individuals [2,3].

Although several hypotheses, including dysregulated immune response, persistent viral reservoir, dysregulated microbiota, and endothelial abnormalities, have been proposed, the pathophysiology underlying PASC remains unclear due to the varied definitions of PASC among previous studies [4-8]. In this study, our objective was to examine the patterns of cytokine expression related to particular PASC symptoms to elucidate the underlying mechanism of PASC.

A total of 156 blood samples and symptom reports from 79 participants were eligible for analysis. The baseline characteristics of participants are as follows (Table 1). The mean age was 57.9 years, and approximately 60% of participants were male. Most participants (86.1%) completed a 2-dose vaccination with the adenovirus-vectored vaccine or mRNA vaccines; 72.2% of these participants had received booster shots. Additionally, 15.6% of all participants had been previously healthy, and approximately half of the participants experienced mild COVID-19. The treatments that hospitalized individuals received are summarized in Table 1. Of this cohort, 41 individuals (51.9%) met criteria for PASC and 38 (48.1%) were categorized as non-PASC (Supplementary Table 1). PASC group showed a higher proportion of individuals with multiple infections (more than twice) (p = 0.097), chronic liver diseases (p = 0.051), and use of tocilizumab (p = 0.019).

Because a few participants did not respond to all survey questions regarding certain symptoms, only the responses for each symptom were included in the analysis. The most frequent symptom was fatigue (65.7%), followed by PEM (53.7%), chronic cough (41.9%), thirst (41.5%), brain fog (40.3%), and chest pain (40.0%) (Fig. 1). Among 58 participants with a complete set of symptom records, 5 (8.6%) did not exhibit any symptoms. Participants who had one or more symptoms had a median of four symptoms (interquartile range 2–7). When individuals were divided into groups based on the number of symptoms and observed over time, approximately 20% of individuals still possessed more than eight symptoms even after six months post-diagnosis (Supplementary Table 2).

Based on the data collected, we analyzed whether the presence of PASC symptoms is associated with the time elapsed after the infection. The results of the χ² test indicate that cardiopulmonary symptoms such as PEM, chronic cough, and chest pain tended to decrease as time progressed (Supplementary Table 3, p = 0.003, 0.098, and 0.061, respectively). Except for fatigue (p = 0.083), we could not observe significant changes of prevalence over time for symptoms other than cardiopulmonary symptoms.

When comparing antibody and cytokine concentrations in relation to the presence of symptoms, it was observed that the cytokine expression patterns varied for each symptom, whereas the antibody levels were not different (Fig. 2 and Supplementary Fig. 1). Among the symptoms investigated, the widest variety of cytokine levels showed significant differences in the presence of gastrointestinal symptoms. The IL-10, VEGF, IL-1β, IL-6, IL-8, MIP-1α, MCP-1, granzyme A, and IL-9 levels were significantly higher in participants with gastrointestinal symptoms than in those without gastrointestinal symptoms (p < 0.05). In contrast, there were no significant differences in cytokine levels between participants with and without chronic cough, dyspnea, dizziness, or hair loss.

We categorized the investigated symptoms into three clusters based on cytokine expression patterns. Symptom cluster 1, including gastrointestinal symptoms, chest pain, PEM, change in sense of smell/taste, fatigue, brain fog, abnormal movement, and palpitation, primarily exhibited significantly increased levels of IL-10, VEGF, and various inflammatory cytokines like MIP-1α, IL-1β, IL-6, IL-8, MIG, granzyme A, and CX3CL1. Additionally, the expression of other cytokines was either significantly increased or decreased. A consistent increase in IL-10 levels was observed in participants who experienced gastrointestinal symptom, chest pain, change in sense of smell/taste, brain fog, fatigue, or palpitation (Supplementary Fig. 2). An increased VEGF level was commonly observed within the cytokine profile among participants who experienced gastrointestinal symptom, chest pain, PEM, fatigue, or brain fog. Besides these cytokines, IL-1β, IL-6, IL-8, MIG, sCD40L, and CX3CL1 showed common expression patterns across different symptom profiles. In addition, the FasL, I-TAC, and IFN-α levels in participants with PEM or fatigue were significantly higher than in participants without these symptoms (p = 0.003, 0.0003, and < 0.0001, respectively).

Symptom cluster 2 (thirst, cognitive dysfunction, and changes in sexual desire and capacity) did not show significant differences in the levels of inflammatory cytokines that were elevated during symptom presentation in symptom cluster 1. Nevertheless, symptoms in cluster 2 were accompanied by differences in the expression levels of cytokines rather than IL-10, VEGF, and inflammatory cytokines. In contrast, symptom cluster 3 (chronic cough, dizziness, dyspnea, and hair loss) did not demonstrate significant differences in cytokine levels compared to the levels in the absence of these symptoms.

Cytokine secretion is one of the initial innate immune responses crucial for acute inflammation. Elevated levels of proinflammatory cytokines in circulation can result in systemic inflammatory symptoms and subsequent damage to secondary organs. Thus, revealing cytokine profiles is important to gain insight into the pathophysiological aspects of COVID-19 [13]. Furthermore, it has been reported that certain cytokines remain significant even after recovery from SARS-CoV-2 infection [4,14-18]. Given the impact of PASC on quality of life [19,20], it is important not only to consider the number of symptoms, but also understand which symptoms persist over time and the mechanisms underlying their persistence. In our current study, we analyzed the frequencies and evaluated the cytokine profiles of individuals with certain symptoms of PASC to better understand of the mechanism of PASC. We observed that a few symptoms emerged prominently at earlier time points and subsequently diminished, whereas others remained persistent even six months after diagnosis. This pattern of symptom occurrence seems similar to that reported in the previous studies [19-21]. In addition, our data showed that even when symptoms were classified within the same category based on the dysfunctional organs, e.g., neurological or cardiopulmonary symptoms, the patterns of cytokine expression varied markedly. Moreover, our data suggests that symptoms belonging to different categories may share a common pathophysiology in terms of the elevation of certain cytokines. To our best knowledge, few studies have clustered PASC symptoms based on cytokine profiles. While Wang et al. [22] identified clinical phenotypes using multi-omics analysis, their approach differs from ours. However, cytokine-based clustering has been applied in other conditions [23-26]. The previous studies clustered patients with similar characteristics by cytokine levels, revealing differences in prognosis despite similar clinical characteristics. This suggests that diseases with heterogeneity, it is thought to be meaningful to cluster symptoms and patients based on cytokine profiles, even among patients with similar symptoms. In the case of brain fog (symptom cluster 1), it is often considered synonymous with cognitive dysfunction (symptom cluster 2), but they are not entirely identical symptoms. Both symptoms have been thought to be associated with neuroinflammation [27-29]. Moreover, in our recent work in another independent neurological PASC cohort (not yet published) suggests that brain fog appears to be more closely associated with PEM or fatigue, whereas cognitive dysfunction may represent an independent entity linked to neuronal damage and impaired recovery, accompanied by morphologic changes in magnetic resonance imaging. Alternatively, brain fog frequently accompanied with fatigue and high levels of blood cytokine levels in symptom cluster 1 (fatigue, PEM, and brain fog), as shown in this study, may suggest that these symptoms might be more related with systemic inflammation. Taken together, although the mechanism underlying the pathogenesis of brain fog and cognitive dysfunction have yet to be fully established, these symptoms, while similar in appearance, may arise from distinct mechanisms. Therefore, our clustering of various PASC symptoms based on cytokine characteristics instead of traditional organ-specific symptom classification may provide further insight into the heterogeneity of PASC pathophysiology.

Among the analytes evaluated in this study, S1-IgG level was not related to the presence of any symptom, whereas cytokines showed diverse profiles for each symptom. Participants with gastrointestinal symptoms exhibited the widest variety of alterations compared to those without symptoms, and there were no significant differences in certain symptoms.

Symptoms of the same category based on traditional organ-specific symptom classification had different cytokine profiles, whereas symptoms of different categories based on the involved organs showed certain common cytokine expression patterns. As shown in Fig. 2, IL-10 and VEGF levels were commonly increased in more than one-third of all symptoms. IL-1β, IL-6, IL-8, MIG, sCD40L, and CX3CL1 showed certain common patterns. The shared expression patterns of certain cytokines observed in symptom cluster 1 suggest that although these symptoms have been classified differently, they might have certain shared pathophysiological mechanisms in patients with PASC, as mentioned above about brain fog and cognitive dysfunction.

Previous studies have reported that the IL-10 level was increased in participants with PASC [14,17], whereas Queiroz et al. [18] reported diminished IL-10 levels in participants with PASC. The varying definitions of PASC across studies might contribute to this difference. In our previous study, patients with severe COVID-19 showed higher IL-10 levels than patients with mild COVID-19 [12]. Considering the results of the previous and current studies, IL-10 plays a proinflammatory role in both the acute and recovery phases of COVID-19 and seems to be related to the manifestation of PASC symptoms [30-32].

Consistent with previous studies, we observed that VEGF and sCD40L levels were increased in participants with PASC symptoms [33,34]. VEGF is a growth factor for vascular endothelial cells known as vascular permeability factor, and sCD40L is a costimulatory molecule involved in the activation of T cells, B cells, and platelets. Excessive VEGF and sCD40L expression can cause vascular damage. Endothelial damage is considered to be an important mechanism in acute COVID-19 and PASC [35-37]. Recently, it was reported that VEGF and sCD40L secreted from activated platelets interact with CX3CR1+CCR5+ monocytes and further contribute to the progression of vessel damage [34]. The elevated levels of CX3CL1 and RANTES, which are ligands of CX3CR1 and CCR5, respectively, were observed in this study. Thus, vessel damage caused by VEGF, sCD40L, CX3CL1, and RANTES might be a potential target for the treatment of PASC.

When comparing baseline characteristics, there was no statistically significant differences between the two groups in terms of age, sex, vaccination status, or viral variant. We found that PASC group had a trend toward being a higher proportion of individuals with multiple infections, which is consistent with the previous study reporting that long COVID was more prevalent when the subjects had reinfections [38]. However, there was no significant difference in the severity of the initial infection between the two groups. Previous prospective studies that followed patients from the acute phase or 6 months post-infection have reported a correlation between the severity of the initial disease and the development of long-term complications [39,40]. In contrast, the recently published long COVID definition states that PASC can occur after any severity of infection [41]. While the risk of PASC is somewhat higher in patients with severe infection, PASC can develop in patients with mild infections and due to the greater number of mild cases, they represent the majority of PASC patients [42]. In addition, our study enrolled patients with COVID, more focused on PASC: patients with PASC who had even mild COVID were more followed in this study than those without PASC who had mild COVID. Therefore, some cautious interpretation is needed for the interpretation of our data because of this kind of bias. Taken together, further longitudinal studies are needed to better understand the relationship between infection severity and the development of PASC.

Our study has certain limitations. First, since the participants enrolled in this study were mostly infected with the omicron variant of SARS-CoV-2, the risk and severity of PASC might be different from those infected with wild-type SARS-CoV-2 or variants before omicron. Previous studies have reported that people infected with the wild-type or delta variant had a higher risk and relative rate of post-acute symptoms [43,44]. Second, we did not examine the cells that serve as the source of cytokines. Recent studies have reported that monocytes and SARS-CoV-2-specific functional T cells are important in PASC; inflammation caused by the persistence of viral proteins in monocytes and interactions between cytokines and monocyte surface receptors, including Fcγ receptor, CX3CR1, and CCR5, are considered to play a major role [45-47]. In addition, the persistent presence of viral antigens could lead to the continuous activation of SARS-CoV-2-specific T cells [48,49]. Altogether, examining the association of cytokine profiles in our study with previously reported cellular candidates could enhance our understanding of PASC pathophysiology. Finally, we observed the different cytokine profiles by symptoms, however, it is challenging to determine whether this expression is the cause or a consequence of PASC. Further investigation is needed to provide more definitive insights.

In conclusion, we showed that even after recovery from the SARS-CoV-2 infection, a few degrees of inflammation were maintained. Our data suggest that symptoms classified into different organ-specific categories may share a common pathophysiology in terms of the elevation of certain cytokines. Although PASC symptoms are heterogeneous, our findings suggest that T-cell recruitment, thrombosis, and increased vascular permeability might contribute to the various symptom clusters sharing common pathophysiologic mechanisms.