Graphical abstract graphic file with name fx1.jpg [57]Open in a new tab Highlights * • The study depicts the immune fluctuation of vigorous exercise at single-cell level * • Effective T cells decrease after marathon with delayed function recovery * • Effector/naive T cell ratio is better for exercise intensity than CD4+/CD8+ T cell ratio __________________________________________________________________ Physical activity; Immune system; Transcriptomics Introduction Exercise immunology has become a growing field in the past few decades, with an emphasis on understanding how different forms of exercise affect immune function. Regular physical activity can promote body health by modulating cardiovascular function and enhancing immune competency.[58]^1 Evidence also demonstrates that moderate-intensity physical activity can reduce the risk of cancers and communicable diseases such as viral and bacterial infections.[59]^2 Intense exercise has a strong effect on immune function; an acute bout of exercise is associated with proinflammatory cytokine production and immunodepression during recovery from fatigue.[60]^3^,[61]^4 Vigorous-intensity leisure-time physical activity, such as marathon, has enjoyed increasing popularity nowadays for its fat-burning benefits. Some authors advocate for the “open window” hypothesis is perceived widely which considers that vigorous exercise heightens the risk of opportunistic infections in the hours following exercise as the proportion of lymphocytes in circulation undergoes transient dramatic reduction.[62]^5^,[63]^6^,[64]^7 Evidence supporting the “open window” hypothesis is mainly from self-reported symptoms of upper respiratory tract infections after exercise[65]^8^,[66]^9^,[67]^10 and observations that total peripheral blood lymphocyte frequency and functional capacity of the lymphocyte pool is dramatically decreased in the hours following exercise.[68]^11 A study done after the 1987 Los Angeles marathon reported that 12.9% of the responders who had completed the marathon claimed to have had an infection in the week after the race, while only 2.2% of the individuals who withdrew from the race had a similar claim.[69]^9 Recent studies indicated a dramatic decrease in frequency of lymphocytes in circulation 1–2 h post exercise, when the lymphocyte numbers were lower than pre-exercise levels. The lymphocyte frequency returned to pre-exercise levels within 24 h.[70]^12^,[71]^13 Other authors argue for the acute stress/exercise immune-enhancement hypothesis. The latter states that exercise does not promote immune suppression, but a transient redistribution of immune cells to peripheral tissues from circulation reflecting a state of immune surveillance and immune regulation. Evidence showed that exercise redeployed immune cells to peripheral tissues (e.g., mucosal surfaces) where those immune cells could identify and eradicate cells infected with pathogens, damaged or malignant. It has been well established that the composition of almost all immune cell populations in the bloodstream changes during and following exercise.[72]^11^,[73]^14 However, a detailed analysis at single-cell level that leads to a comprehensive understanding of the changes in immune populations promoted by exercise is missing as all the pioneer molecular evidence is from omics data of bulk specimen.[74]^15 In the current study, we performed a longitudinal single-cell transcriptomic analysis of time-serial collected peripheral blood samples from participants taking a single bout of acute physical activity in the form of symptom-limited cardiopulmonary exercise (CPX) or marathon. The goal was to characterize the immune function after a single bout of aerobic exercise of different intensities and durations at single-cell level. Our deep longitudinal profiling revealed remarkable fluctuations of key immune cells in the bloodstream as well as changes in expression of certain molecules and biological processes such as inflammation, innate immune response, and antiviral response. Results Experimental design Three participants were enrolled for CPX physical exercise testing and serial peripheral blood collection while another three participants were enrolled for serial peripheral blood collection before, during, and after marathon. The detailed demographic information of all participants at baseline is shown in [75]Table S1. Intravenous blood samples were collected before the physical exercise as baseline and during the recovery period at 5 min, 60 min, and 4 h for CPX or 24 h for marathon after the end of the physical exercise ([76]Figure 1A). In-depth profiling of single-cell RNA transcriptomics on isolated peripheral blood mononuclear cells (PBMCs) was performed. After data curation and annotation, we obtained a total of 275,054 single-cell transcriptomes with information on expression of 23,494 genes. A list containing the general information for the two datasets can be found in [77]Table S2. Figure 1. [78]Figure 1 [79]Open in a new tab PBMC single-cell atlas after acute exercise (A) Workflow of samples collection, processing for single cell and data visualization. (B) TSNE plot of entire cells from 12 single-cell samples from participants of CPX, colored by clusters. (C and D) Cell proportions of main cell types from different samples of CPX along time, colored either by cell types (C) or by samples collected at different time points (D). (E) Same as (B), but for 12 single-cell samples from participants of marathon. (F) Same as (C), but for samples from participants of marathon. (G) Same as (D), but for samples from participants of marathon. Key immune cells in circulation underwent remarkable abundance fluctuations, which was more intense in marathon than CPX To characterize the immunological features in acute vigorous physical exercise, we performed droplet-based scRNA-seq (10x Genomics) on PBMCs. After the single-cell analysis pipeline, we obtained 142,885 cells (51.9%) for CPX participant samples (dataset 1) and 132,169 cells (48.1%) for those of marathon (dataset 2). The high-quality single-cell data from different exercise groups were integrated separately into two unbatched datasets and subjected to principal component analysis after correction for read depth and mitochondrial read counts ([80]Figures S1A–S1D). Using graph-based clustering of t-distributed stochastic neighbor embedding (t-SNE), transcriptomes of 31 clusters were captured out of dataset 1 for CPX ([81]Figure 1B). We applied distinct cell-type-specific biomarkers to resolve five main cell types — T cell, natural killer (NK) cell, B cell, monocyte, and platelet ([82]Figure S2A). The biomarker sets used for cell population definition included expression of: CD3D, CD3E, and CD3G for T cell; XCL2, KLRB1, KLRD1, KLRF1, FCGR3A, and FGFBP2 for NK cell; CD14 for monocyte; IGHD and CD79A for B cell; and PPBP and PF4 for platelet ([83]Figures S2C and S2D). The proportions of each cell cluster varied among different time points ([84]Figures 1C and 1D). We calculated the average relative percentage of the four major immune cell types in the PBMCs at each time point based on scRNA-seq data. First, we observed a slight proportion reduction of T cell at 5 min post CPX (0.92 ± 0.12, p = 0.42, [85]Figure S2D), but afterward, we observed a persistent upward increase of T cells up to the 4 h time point (1.22 ± 0.06, ∗p = 0.03, [86]Figure S2D). Second, the abundance of monocyte (5 min vs. pre: 0.57 ± 0.01, ∗∗∗p = 0.0006, [87]Figure S2E) and B cells (5 min vs. pre: 0.65 ± 0.06, ∗∗p = 0.01, [88]Figure S2F) underwent a significantly remarkable reduction at 5 min, and recovered to normal levels at 1 h post exercise. Third, the proportion of NK cell increased following a single bout of CPX and recovered soon after 1 h (5 min vs. pre: 1.75 ± 0.34, ∗p = 0.05, [89]Figure S2G). A similar analysis was applied to the second dataset containing single-cell transcriptomes of the participants who run the marathon. In total, 27 clusters were captured and assigned into the same cell type groups as those of dataset 1 using the above-mentioned molecular markers ([90]Figures 1E and [91]S3A–S3C). The average relative percentage of T cells decreased remarkable immediately following exercise (5 min vs. pre: 0.50 ± 0.12, ∗p = 0.03) and remained low 1 h later (1 h vs. pre: 0.52 ± 0.06, ∗∗p = 0.007), returning to normal levels around 24 h after the end of the marathon ([92]Figures 1F, 1G, and [93]S3D). Monocytes and NK cells presented the opposite trend to that of CPX. Monocytes in circulation accumulated during the marathon (2.91 ± 0.55, ∗p = 0.04), peaked at 1 h (3.15 ± 0.62, ∗p = 0.04), and were back to baseline levels within 24 h ([94]Figures 1F, 1G, and [95]S3E). NK cells in marathon participants showed similar fluctuation trend as T cells, presenting a slight reduction immediately following the marathon (0.60 ± 0.25, p = 0.20), and were at a lower number 1 h (0.44 ± 0.07, ∗∗p = 0.009) post exercise. NK cells were back to baseline levels at 24 h post exercise ([96]Figures 1F, 1G, and [97]S3F). B cells gradually decreased, reaching a minimum level (0.49 ± 0.04, ∗∗p = 0.004) 24 h post exercise ([98]Figures 1F, 1G, and [99]S3G). Composition variations of NKT cell subsets were observed in acute vigorous physical exercise NKT cells are part of the first line of immune effectors and participate in the rapid response to a wide range of stress-related signaling. Thus, we subclustered and characterized the features of all NKT subtypes following a bout of vigorous physical exercise. The NKT subtype fluctuations were more intensive in marathon than CPX. We analyzed the NKT cell dataset of cells within the marathon dataset, which could be annotated into 12 subtypes ([100]Figure 2A). We defined 4 CD4^+ T cell subtypes (CD3D + CD3E + CD3G + CD4+IL7R+): CD4+_naive (CCR7+TCF7+LEF1+SELL+), CD4+_fh (follicular helper, MAF+CXCR5+), CD4+_CTL (cytotoxic), which were characterized by high expression of genes associated with cytotoxicity, including PRF, GNLY, NKG7, GZMB, and GZMK, and CD4_CTLGZMK+, which expressed high levels of GZMK; two CD8^+ T cell subtypes (CD3D + CD3E + CD3G + CD8A + CD8B+): CD8+_naive and CD8+_CTL; and 5 NK cell subtypes (CD4^−CD8A-FCGR3A+): NK_cytotoxic_KLRF+ which expressed high levels of KLRF; NK_cytotoxic_KLRD+ with high levels of KLRD; NK_ANXA1+ featured expressing high levels of ANXA; NK_resting (XCL1+XCL2+) and NK_cycling (MKI67+) subsets ([101]Figures 2B and [102]S4A). CD4^+ naive cells represented the most frequent cell type (21.54%) while NK cycling cells were only 0.24% ([103]Figure S4B). We analyzed the cytotoxicity score using the expression of PRF1, IFNG, GNLY, NKG7, GZMA, GZMB, GZMH, GZMK, KLRB1, KLRD1, KLRK1, CTSW, and CST7 and naive (resting) score using the expression of CCR7, TCF7, LEF1, and SELL across all NKT cells, found that NK cytotoxic, CD4+_CTL, and CD8+_CTL subsets showed higher cytotoxicity scores than those of the other subsets ([104]Figure 2C). In contrast, CD4+_naive and CD8+_naive subsets showed higher naive scores ([105]Figure 2D). Figure 2. [106]Figure 2 [107]Open in a new tab Identification of NKT cell subsets in sample of marathon (A) UMAP plot of all T cell subtypes identified from single-cell samples of marathon either from the overall data or from samples collected at different time points, colored by fine labels. (B) Heatmap of expression level for canonical markers of all T cell subsets in samples from marathon participants. (C) UMAP feature plot of cytotoxic scores for all T cells in marathon samples calculated with curated gene signatures. (D) UMAP feature plot of naive scores for all T cells in marathon samples calculated with curated gene signatures. (E) Cell proportions of main T cell subtypes from different samples of marathon participants. (F) Fluctuation of cell proportions along time for all CD4-positive T cell in marathon samples. (G) Fluctuation of cell proportions along time for all CD8-positive T cell in marathon samples. (H) Proportions of CD4-positive T cell subtypes for all participants of marathon along time. (I) Proportions of CD8-positive T cell subtypes for all participants of marathon along time. (J) Fluctuation of CD4/CD8 ratio along time for marathon. (K) Fluctuation of effective/naive cell ratio along time for marathon. The paired t-test and single-factor ANOVA were used for statistical analysis. The error bars represent SD while single (∗) and double (∗∗) asterisks represent a p ≤ 0.05 and p ≤ 0.001, respectively. For the CPX group, we obtained 11 subsets according to the expression and distribution of canonical NKT cell markers ([108]Figure S5A): 3 CD4^+ T cell subtypes, 3 CD8^+ T cell subtypes, 4 NK cell subtypes, and one undefined cytotoxic cell group. Of the CD4^+ T cells, we defined subtypes: CD4+_naive T cells, CD4+_Tem (effector & memory) cells (ANXA1+CD40LG+), and CD4+_CTL cells. Of the CD8^+ T cells, we defined subtypes: CD8+_T naive cells, CD8+_Tem cells, and CD8+_CTL cells. The four NK cell clusters were defined including two cytotoxic NK cell subsets (NK_cytotoxic_KLRB1+ cells expressing high level of KLRB1 and NK_cytotoxic_KLRD1+ cells characterized by high level of KLRD1), NK_resting cells and NK_cycling cells ([109]Figures S5B and S5C). Of all the NKT cell subsets, CD8+_Tem cells accounted for more than 17%, while the CD8+_CTL cell were less than 1% ([110]Figure S5D). We evaluated the cytotoxicity and naive scores of all NKT cell subsets of CPX, identifying the more cytotoxic cells, the more right-side in the UMAP plot ([111]Figures S5E and S5F). It is interesting that within those highly cytotoxic clusters, NK cells had higher cytotoxic state than T cells ([112]Figures 2C and [113]S5E). Reduction of effective T cells was observed with the enrichment of cell migration and motility pathways To gain insights into features of NKT cell subsets, we evaluated the distribution of all clusters across four time points. From the marathon dataset, the frequency of naive T cells was stable following exercise, but the cytotoxic T cells and T follicular helper cells were selectively reduced ([114]Figures 2A and 2E). The abundance variations were bigger during different time points than among different participants ([115]Figures S4C and S4D). Most significantly enriched gene ontology (GO) terms for upregulated differentially expressed genes (DEGs) of the T cells were focused on cell migration, motility, and localization. Within 24 h, the abundance of effective T cells went back to baseline levels. A proper balance of CD4^+ T cells and CD8^+ T cells in circulation is believed to be essential for full immune competence.[116]^16 CD4+/CD8+ ratio has been used to monitor the immune system during exercise as an improper balance may lead to disease.[117]^17^,[118]^18 We calculated the overall abundance changes of CD4^+ T cell and CD8^+ T cell as well as the specific subclusters under each cell type. We found that naive and effector cells showed an opposite trend ([119]Figures 2F–2K). The overall frequency of CD4^+ T cells was found to be decreased 5 min after the marathon and increased thereafter ([120]Figure 2F), while the overall frequency of CD8^+ T cells was increased at 5 min and 1 h after the marathon and then decreased ([121]Figure 2G). The proportion of other naive T cell subclusters increased following the marathon and peaked at 1 h, while the cytotoxic cells were found to be decreased up to 1 h after the marathon ([122]Figures 2H and 2I). Furthermore, we explored the CD4/CD8 ratio ([123]Figure 2J) as well as the effector/naive ratio ([124]Figure 2K), both of which exhibited significant declines following the marathon, with the latter being more homogeneous among the volunteers. We observed that from the CPX dataset, NK cycling cells were increased at early stage of the recovery ([125]Figure S5G). We calculated the overall changes in composition of CD4^+ and CD8^+ T cell as well as specific subclusters under each classification. Of particular note, both CD4^+ and CD8^+ T cells decreased following CPX ([126]Figures S6A–S6F). Among all CD4^+ T cell subtypes, we observed that naive and Tem cells presented significant reductions at 5 min post exercise (0.54 ± 0.11, ∗p = 0.03 for naive T and 0.57 ± 0.11, ∗p = 0.03 for Tem) while the cytotoxic cells were relatively stable ([127]Figure S6C). Similar fluctuation trend occurred for CD8^+ T cells, with naive cells decreasing significantly at 5 min post exercise (0.55 ± 0.09, ∗p = 0.02) and Tem cells decreased significantly at 4 h post exercise (0.59 ± 0.02, ∗∗∗p = 0.001) while the cytotoxic cells were relatively stable ([128]Figure S6D). Furthermore, we calculated the ratio of CD4/CD8 for all time, unfortunately finding no significant points, and found no changes following exercise although the ratio increased slightly from 1 h post exercise, and the ratio showed a big dispersion among participants ([129]Figure S6E). Interestingly, a more uniform ratio which could better indicate the intensity of exercise was found for effector/naive T cells peaked at 5 min (∗∗p = 0.006) and returned to normal at 1 h post CPX. These data suggest that this ratio is a better indicator of the intensity of exercise ([130]Figure S6F). Overall, NKT subtypes could be grouped according with their timeline fluctuations into three main groups. First, CD8^+ Tem and two NK cytotoxic cell types showed a remarkable increase at 5 min after CPX and declining sharply after that. Second, the CD4^+ and CD8^+ naive T cells and CD4^+ Tem cells decreased at the early time point after CPX and increased thereafter. Third, subtypes did not show any fluctuations such as CD4^+ CTL cells ([131]Figure S6G). This shows that there is a reshuffling of some immune cell components in circulation induced by a single bout of CPX. Moreover, among all NKT cell subsets, cytotoxic NK cell subsets were the most variable ones ([132]Figure S6H). Functional feature variations of NKT cell subsets were observed following acute physical exercise, which required longer recovery time than abundance changes To further investigate differential transcriptomic changes in NKT cells following acute exercise, we compared the expression profiles of effector NKT subclusters (CD4^+ cytotoxic, CD4^+ Tem or fh, CD8^+ cytotoxic, and NK cytotoxic cell subsets) across different time points to identify core DEGs. Regarding the marathon dataset, we observed that some DEGs were upregulated at 5 min after exercise but quickly returned to baseline (such as CXCR4, METRNL, and CREM), while others peaked at 1 h post exercise (such as S100A8 and S100A9). There were also several DEGs that were upregulated at 5 min and 1 h post exercise (such as FTH1, SYTL3, and BTG1) ([133]Figure 3A). Moreover, the expression pattern of DEGs in the marathon samples was strongly subcluster dependent. Analysis of DEGs from NK cytotoxic and CD4^+ effector cells suggests that those subsets are mainly early and late responders, respectively. Analysis of DEGs from CD8^+ effector cells suggests that this subset participates in the immune changes through the whole recovery time. Several core DEGs were detected to be upregulated across all subclusters. The first one was CXCR4, which is a transmembrane receptor expressing on the cell surface of most leukocytes and playing a crucial role in cell migration.[134]^19 The other two upregulated genes were S100A8/A9, which can form heterodimeric protein that modulates the inflammatory response.[135]^20 Plasma S100A8/A9 levels have been used as biomarkers of inflammation in several inflammatory disorders.[136]^21^,[137]^22 Thus, genes involved in inflammation, such as S100A8/S100A9, antimicrobial agents, such as LYZ, DDIT4, CEMIP2, and ISG20, cell proliferation, and migration, such as CXCR4, VCAN, CNOT6L, and AREG, were among the key DEGs in effector NKT cells upon exercise. Furthermore, using GO enrichment analysis we identified that biological processes such as “response to stress”, “cell migration”, “cell motility”, or “chronic inflammatory response” were active following marathon ([138]Figure 3C). Figure 3. [139]Figure 3 [140]Open in a new tab Gene expression and functional changes in T cells following acute exercises (A and B) Heatmap illustrating scaled expression values of top significantly differential expressed genes a long time points in main T cell subsets for marathon (A) and CPX (B). (C and D) GO pathway enrichment analysis of identified DEGs for marathon (C) and CPX (D). (E) Expression levels of important immune-related pathways in GO biological process terms in NKT cells across different time series. The genes that were found to be differentially expressed and their pattern of upregulation following CPX showed differences from those found in the marathon samples ([141]Figure 3B). Some peaked at 5 min and recovered soon, genes like DUSPs and S100B; some had their peak expression at 1 h and returned to baseline levels within 4 h, like BTG1 and TNFRSF1B, while others were upregulated late around 4 h, e.g., PPBP and TMEM109. GO enrichment analysis showed that those DEGs were involved in biological processes such as “response to stimulus” and “response to stress” et al. ([142]Figure 3D). To investigate the potential differences in immune function such as antiviral and pathogenic immune responses within the several hours following acute physical exercise of different intensities (marathon vs. CPX), we evaluated expression levels or scores of important biological processes in NKT using GO biological process terms across different time series and compared the behavior differences between CPX and marathon. The genes used for determining each GO term score were listed in [143]Table S3. First, we defined the cytotoxicity score and naive/resting score at the different time points. We found that the cytotoxicity score was exhibiting in most cells and maximum at 5 min after exercise and decreased after that. In contrast, the naive/resting score having a minimum value at 5 min and increased afterward. Next, we checked immune response to virus/bacterial/fungal infections. We observed that the immune response to molecules of fungal origin was weak and did not change with time or intensity of the exercise. The antibacterial response was weak in both exercise forms, but slightly enhanced following marathon. The immune response to viral infection was enhanced following CPX but inhibited following marathon. Interestingly, baseline antivirus response in marathon participants was higher than those doing the CPX, suggesting that regular marathon running can improve antiviral capability, but more intensive studies involving larger cohorts of participants are needed to confirm it further. Not only the antiviral response but also the “immune cell-mediated cytotoxicity” and “NK cell proliferation and activation” abilities demonstrated an enhanced baseline levels in regular marathon runners, while those response processes were elevated shortly following CPX. However, all those immune-related capacities showed an inhibition mode during the “open window” period after a single bout of acute exercise. For the CPX group, recovery took place after 4 h and for the marathon group, recovery started around 24 h post exercise. Finally, we checked the “innate immune response”, “adaptive immune response”, and “acute inflammatory response” scores, found that those pathways were enhanced at 5 min following exercise, returned to baseline at 1 h, and were suppressed later. Notably, the overall adaptive immune response was much lower than the innate immune response ([144]Figure 3E). The variations of those biological processes for individual NKT subsets were shown for CPX ([145]Figure S7A) and marathon ([146]Figure S7B). Two different monocyte states were identified following acute physical exercise The role of monocytes in propagating and maintaining many inflammatory diseases is well established,[147]^23 but the features and function of monocytes in acute exercises have not been comprehensively explored. Monocytes account for a large proportion of cells in our datasets. We observed an overall remarkable abundance fluctuation following acute exercise especially after the marathon ([148]Figure 1). We found two monocyte subclusters and two DC subclusters ([149]Figure 4A) after the marathon. The two monocyte subsets exhibited little variation regarding CD14 and CD68 expression ([150]Figure S8A). We further defined the two monocyte subgroups, as S100A8/A9^hiMono group which expressed high levels of S100A8/S100A9 and S100A12, and HLA-DPA1/DPB1^hiMono group with upregulated HLA-DPA1, HLA-DPB1, and HLA-DMB ([151]Figures 4B and [152]S8B). Notably, the two monocyte groups distributed in a time-dependent manner, the S100A8/A9^hi monocyte group was mainly present in PBMC samples at 5 min and 1 h post marathon, while the HLA-DPA1/DPB1^hi monocyte group was detected mainly at baseline and after 24 h of recovery ([153]Figures 4A and 4C). Moreover, this distribution pattern was uniform among different participants ([154]Figure 4D). Several genes related with risk of heart disease were detected to be upregulated in this monocyte group. The first one was VCAN. Versican accumulation is associated with impaired cardiac function. Single-cell RNA-sequencing data in human cardiac tissue has revealed the upregulation of versican expression in a subset of myeloid cells.[155]^24 The other two upregulated genes were RETN and ACSL1, which could serve as molecular markers for the risk of heart diseases.[156]^25^,[157]^26 Increased resistin (RETN) levels after running have been observed in previous studies.[158]^27 NAMPT was also upregulated in this monocyte subgroup after marathon. S100A8/A9^hi monocyte subgroup was mainly enriched in the “response to stimulus” and “response to external stimulus” pathways, while the HLA-DPA1/DPB1^hi monocyte group was enriched in “immune response” pathways, meaning a functional switch of monocytes as antigen-presenting cells at resting state to effectors to acute stress following acute exercises. As well, time-dependent DEGs were identified for all monocytes ([159]Figure 4F). GO pathway analysis showed that monocytes of the marathon samples expressed gene signatures that were enriched for biological processes such as “response to oxygen-containing compound” and “defense response to fungus” ([160]Figure S8C). Figure 4. [161]Figure 4 [162]Open in a new tab Features of monocyte subtypes following acute exercises (A) UMAP plot of monocyte subtypes in samples of marathon either from the overall data or from samples collected at different time points, colored by fine labels. (B) Volcano plot indicating the upregulated genes in either HLA-DPA1/DPB1^hiMono group (left) or S100A8/A9^hiMono group (right), red indicated for up-expressed with |log2 fold change| ≥ 0.6 and P value ≤0.05. (C and D) Cell proportions of monocyte subtypes from all marathon participants along time, colored either by subtypes (C) or by samples collected at different time points (D). (E) GO pathway enrichment analysis of identified DEGs either from S100A8/A9^hiMono group (upper) or HLA-DPA1/DPB1^hiMono group (lower). (F) Heatmap illustrating scaled expression values of top significantly differential expressed genes along time points in main monocyte subsets for marathon. (G) Expression levels of important monocyte-related pathways in GO biological process terms in monocytes across different time series. From the CPX dataset, we identified two monocyte subsets (mono_CD14++CD68^+ and mono_CD14+CD68++) and two dendritic cell subclusters—cDC (conventional dendritic cell) and pDC (plasmacytoid dendritic cell) ([163]Figure S8D). The CD14++CD68^+ monocytes were the more abundant group ([164]Figure S8E), which expressed high levels of CD14 and moderate levels of CD68 ([165]Figure S8F). Its frequency was stable following exercise ([166]Figures S8G and S8H). The CD14^+CD68++ monocytes expressed low levels of CD14 but much higher levels of CD68 ([167]Figure S8E). This subset was slightly increased following CPX, but recovered baseline levels 1 h after the exercise ([168]Figures S8E and S8G). In both monocyte subsets, some genes related with inflammation (GZMB, NKG7, and KLRB1) and chemokines (CCL5) were upregulated after acute exercise ([169]Figure S8I). GO pathway analysis showed that monocytes expressed gene signatures that were enriched for biological processes such as “response to cytokine” and “immune system process” ([170]Figure S8J). Monocyte-related GO pathway scores showed that “monocyte chemotaxis” pathway was elevated early after the marathon and late after the CPX; “macrophage activation involved in immune response” was inhibited at 1 h following CPX, but gradually increased following marathon, peaking at 24 h; “Toll-like receptor signaling pathway” showed similar changes for CPX and marathon, both being enhanced at 5 min following exercise and inhibited later; the two DC-related pathways were much weaker as the small proportion of DC in circulation, both of which showed slightly enhancement following exercise and recovery to baseline after the open window period ([171]Figure 4G). More immature/transitional B cell subtype was identified following marathon comparing with CPX We analyzed the B cell lineage further to trace the dynamic changes of B cell subtypes upon exercise. Interestingly, we subclustered B cells into three subsets for CPX dataset ([172]Figure 5A) and four subsets for marathon dataset ([173]Figure 5B), according to the expression and distribution of canonical B cell markers ([174]Figures S9A and S9B). We defined memory B cell (GPR183+CD27^+), naive B cell (CXCR4+IGHD+IGHM+), and plasma cells (CD38+MZB1+IGHA1+) for both CPX and marathon datasets, while defined a small group of immature/transitional B cells (CD86^+CD84^+CD93^+) present only in the marathon dataset, and this immature/transitional B cell subgroup was present in all marathon participants ([175]Figure 5B). All B cell subclusters distributed in different time points ([176]Figures S9C–S9F). We observed a slight decrease in naive B cell and an increase in memory and plasma cells during the recovery period of exercise, while immature/transitional B cell accumulated in circulation following marathon and decreased sharply at 24 h ([177]Figures S9D and S9F). Figure 5. [178]Figure 5 [179]Open in a new tab Identification of B cell subsets following acute exercises (A) UMAP plot of monocyte subtypes in samples of CPX, colored by fine labels. (B) UMAP plot of monocyte subtypes in samples of marathon either from the overall data or from samples collected at different time points, colored by fine labels. (C) Heatmap illustrating scaled expression values of top significantly differential expressed genes along time points in main B cell subsets for CPX. (D) Heatmap illustrating scaled expression values of top significantly differential expressed genes along time points in main B cell subsets for marathon. (E) GO pathway enrichment analysis of identified B cell DEGs for marathon. (F) Expression levels of important B cell involved pathways in GO biological process terms in B cells across different time series. To further investigate differential transcriptomic changes in B cells following an acute vigorous exercise, we compared the expression profiles of B cell subclusters at different time points. We identified fewer DEGs after CPX than after marathon. Memory B cells upregulated PPBP 1 h after the CPX, while naive B cells expressed high levels of IGLC3 and TXNIP post CPX ([180]Figure 5C). Inflammation-related genes (S100A8/S100A9) and immunoglobulin genes (IGHG2 and IGHG1) were upregulated following the marathon ([181]Figure 5D). DEGs that were most significantly enriched following marathon were genes involved in “immune system process”, “immune response”, and “humoral immune response” processes ([182]Figure 5E). Moreover, we evaluated expression levels or scores of important B cell-related pathways using GO biological process terms longitudinally, analysis revealing that “antigen processing and presentation” was inhibited following CPX but relatively enhanced following marathon, while the “inflammatory response to antigenic stimulus” was enhanced following both CPX and marathon. Genes of other programs like “immunoglobulin production”, “B cell lineage commitment”, and “B cell proliferation and activation” were expressed in a minority proportion of detected cells, but all of them underwent an enhancement period after exercises ([183]Figure 5F). Discussion It is believed that vigorous exercise stimulates the immune system, resulting in lymphocytosis mainly due to a transient NK cell expansion lasting for 45–60 min.[184]^28 This response is driven by both a non-specific flushing of the marginal blood pools to bloodstream[185]^29 and an adrenergic stimulation involving the β-2-adrenergic receptors on lymphocytes.[186]^30^,[187]^31^,[188]^32 It is next followed by lymphopenia, again involving changes in NK cell count, which occurs 1–2 h later.[189]^33 Our data further support this scenario. As the CPX duration was less than 15 min, we could capture the during-exercise lymphocytosis phenomena at the 5 min post exercise sample, which exhibited a dramatic increase of total NK cells. After that, lymphopenia occurred due to a decline of NK cells even below the baseline. Moreover, NK cells in bloodstream were mainly cytotoxic NK cell subsets (KLRB1+KLRD1+KLRF1+), meaning during exercise, blood is predominantly occupied by cells capable of responding strongly to stimuli. Besides NK cells, we also observed accumulation of CD8^+ memory cells during CPX, which are believed to mount rapid responses and effector functions.[190]^28^,[191]^34^,[192]^35^,[193]^36^,[194]^37 On the contrary, we captured a sharp decline of CD4^+ T cell during exercise inferred by the post 5 min sample. This observation is consistent with previous findings showing that the differential expression of β-2-adrenergic receptors on lymphocytes upon exercise is higher in NK cells > CD8^+ T cells > B cells > CD4^+ T cells.[195]^12^,[196]^13^,[197]^28^,[198]^38 We did not capture the typical NK cell-driven lymphocytosis during marathon running. One reason could be that the 5 min sample could not represent profiles during exercise as a bout of marathon usually lasts for several hours, which is already beyond the “lymphocytosis period”. However, we observed a sharp decline of cytotoxic T cells during the “lymphopenia period” following marathon, and compensated somehow by the increase of naive cells. In conclusion, we indeed observed a transient time-dependent “lymphopenia period” following a single bout of vigorous activity. In CPX, it was driven by cytotoxic NK cell while in marathon was due to cytotoxic T cell. The lymphopenia that occurs following exercise is intensity dependent. A depressed CD4/CD8 ratio during exercise and a trend toward elevation during recovery has been observed in pioneer studies. An imbalance in the CD4/CD8 ratio has been implicated as the possible culprit for post-exercise infections. CD4/CD8 ratio has been used to monitor tolerance of exercise intensity and avoid a post-exercise infection.[199]^16^,[200]^17^,[201]^18 Our single-cell deconvolution of T cell subtypes in PBMCs revealed that the effector/naive T cell ratio serves as a better indicator for exercise tolerance than CD4/CD8 ratio for its precision and uniformity, which could also avoid the possibility of CD4^+ T cell-targeted virus infection like HIV. The “open window” hypothesis considered the “lymphopenia period” was concomitant with decreased functional capacity of the lymphocyte pool, thus inducing a short-term window of immune suppression. While an acute stress/exercise immune-enhancement hypothesis proposed that exercise redeploys immune cells to peripheral tissues to conduct immune surveillance,[202]^39 which is supported by studies using fluorescent T cell tracking in rodents.[203]^31^,[204]^40^,[205]^41 Evidence showing the redistribution of immune cells after exercise is still lacking in humans. In our data, the immune functional analyses indicated that most immune-related processes like “immune cell-mediated cytotoxicity”, “immune cell activation”, “innate and adaptive immune response”, and “acute inflammatory response” were activated during exercise in response to adrenergic activity but suppressed following vigorous exercise. In CPX, this suppression effect peaked at 4 h post exercise, while peaked at 1 h post marathon. Furthermore, we observed a delayed immune homeostasis following exercise as a single bout of vigorous exercise resulted in immune homeostatic perturbations. Based on these results, we recommend that people should take enough time for the immune system to recover after a single bout of vigorous exercise. For a short acute exercise like CPX, it requires at least 4 h, while for a prolonged time of exercise like running the marathon, the recovery time is more than 24 h. Selective reduction of effective T cells was observed. However, by GO enrichment analysis, we identified that biological processes such as “cell migration”, “cell motility”, and “localization of cell” were active following marathon, implying a redistribution of effective T cells to peripheral tissue after a long-term vigorous activity. In clinical samples, an exercise-dependent increase of intra-tumoral CD8^+ T cells has been observed.[206]^42 It has been reported that leisure-time physical activity is associated with lower risks of many cancer types.[207]^43 Another highlight of our single-cell work is that we also explored the landscape of monocytes in bloodstream following exercise. Monocytes and cells of the dendritic cell lineage circulate in blood and eventually migrate into tissue where monocytes further mature into different types of macrophages. The monocyte lineage is multifunctional with roles in homeostasis, immune defense, and tissue repair.[208]^44 Monocytes could also serve as antigen-presenting cells to bridge innate immune response and adaptive immune response. Evidence showed more active interactions between T cell and monocyte presented in human tuberculosis.[209]^45 According to previous studies, human blood monocytes can be subdivided into 3 subsets: classical CD14++CD16^− monocytes, intermediate CD14++CD16^+ monocytes, and nonclassical CD14^+CD16++ monocytes.[210]^46 A recent study reported that the nonclassical CD14^+ inflammatory monocytes dominated and incited the inflammatory storm in COVID-19.[211]^47 We identified two subclusters of monocytes in datasets from CPX and marathon. In dataset 1, we defined two monocyte subclusters, CD14++CD68^+ and CD14^+CD68++, with the latter one showing more macrophage phenotype and its abundance increased following exercise. Both monocyte subclusters upregulated cytotoxicity-related genes such as GZMB, NKG7, and CCL5 during and following exercise. The two monocyte subsets identified in dataset 2 were defined as S100A8/A9^hi and HLA-DPA1/DPB1^hi monocytes. Notably, circulating monocytes accumulated following marathon and displayed a time-dependent abundance, as HLA-DPA1/DPB1^hi monocytes were only present at baseline and recovery while S100A8/A9^hi monocytes were activated following the marathon. As HLA-DPA1/DPB1 are CD8^+ T cells activators, it indicated that regular marathon running could enhance the CD8^+ T cell activation ability of monocytes at rest, while those monocytes transform to stress states during marathon. S100A8/A9 has been identified as a master regulator causing cardiomyocyte death in response to ischemic/reperfusion injury. Elevated serum S100A8/A9 associated with major adverse cardiovascular events in patients with acute myocardial infarction.[212]^48 The transient upregulation of the expression of S100A8/A9 in CD4^+ CTL, CD8^+ CTL, NK cytotoxic, and CD4^+ fh cells, together with several genes related with heart diseases (VCAN, RETN, and ACSL1) in monocytes after marathon, indicated a more than 24 h recovery time is required. In conclusion, our deep longitudinal single PBMC profiling demonstrated that the quantity and quality of effector immune cells as well as the immune function in the peripheral circulation were suppressed following a single bout of acute vigorous aerobic exercise. Such immune suppression following exercise required a certain period for recovery, during which the functional recovery was usually lagging behind the quantity recovery. Effector/naive T cell ratio could serve as a better indicator for exercise tolerance than CD4/CD8 ratio, but more studies involving larger cohorts of participants in the future are needed to confirm it. Limitations of the study The sample size in the current study is relatively small and only males are included. The inclusion of only 3 individuals per groups makes it challenging to identify statistically significant differences at the individual patient level. In the future study, we plan to include more participants to avoid bias. We missed some additional clinical and demographic data for the participants, like blood counts, creatine kinase (CPK) level, and troponin level. STAR★Methods Key resources table REAGENT or RESOURCE SOURCE IDENTIFIER Biological samples __________________________________________________________________ Healthy donor whole blood volunteer This paper __________________________________________________________________ Critical commercial assays __________________________________________________________________ Chromium Single Cell 3ʹ GEM, Library & Gel Bead Kit v3.1 10x Genomics PN-1000121 Chromium Single Cell G Chip Kit 10x Genomics PN-1000120 __________________________________________________________________ Deposited data __________________________________________________________________ Raw and processed scRNA-seq data This paper OEP003910 __________________________________________________________________ Software and algorithms __________________________________________________________________ Cell Ranger 10x Genomics Version 3.1.0 Seurat [213]https://satijalab.org/seurat/index.html Version 4.0.5 [214]Open in a new tab Resource availability Lead contact Additional information and requests for resources and reagents should be directed to and will be completed by the lead contact, Yan Zhang ([215]zhangyan2021@cupes.edu.cn). Materials availability This study did not generate any unique reagents. Experimental model and subject details Participants and exercise protocol A total of six healthy participants were enrolled in this study. The detailed demographic information of all participants at baseline is shown in [216]Table S1. Three young volunteers who do not usually exercise, performed a 10–15min controlled symptom-limited cardiopulmonary exercise CPX. They were instructed to run on the treadmill according to the laid down Bruce treadmill test protocols until they became exhausted or met the criteria for a respiratory exchange ratio (RER; VCO2/VO2) > 1.1, heart rate (HR) > 85% of predicted maximum. Three regular marathon runners completed a 4 h marathon. None of the participants has any previous diagnosed diseases. The study has been reviewed and approved by the local ethics committee. All participants provided written, informed consent. Method details Peripheral blood mononuclear cells isolation Venous blood samples were collected before exercise and 5 minutes, 1 hour, 4 hours (for CPX) or 24 hours (for marathon) post exercise. Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood using Ficoll-Paque density gradient centrifugation (HISTOPAQUE®1083, Sigma, USA) according to the manufacturer’s instructions. The mononuclear cells resting in the opaque interface were transferred into a 15 ml conical centrifuge tube and washed with phosphate-buffered saline (PBS) three times. Single-cell RNA sequencing (scRNAseq) - Library preparation and sequencing Viable single cell suspension was loaded into a Chromium controller (10x Genomics, USA) to generate single-cell Gel Beads-In-Emulsions (GEMs). Single cell transcriptomic amplification and library preparation were performed by CapitalBio Technology Corporation (Beijing, China) using single-cell 3’v3 (10x Genomics, USA) according to the manufacturer’s instructions. Libraries were sequenced on Illumina NovaSeq6000 system (Illumina, Inc., San Diego, California, US). Quantification and statistical analysis Single cell data processing Raw FASTQ files were aligned against the Human GRCh38 genome and processed with Cell Ranger (version 3.1.0) and Seurat R package (version 4.0.5). Low quality cells were removed if gene counts were fewer than 200 or more than 5,000, or more than 15% of UMIs derived from the mitochondrial genome. Additionally, all genes that were not detected in at least three single cells were discarded. All remaining cells were considered as high-quality cells. High variable genes were calculated using Seurat FindVariableGenes function with default parameters and the top 2,000 most variable genes were selected for Principal Component Analysis (PCA) and dimension reduction. The expression matrixes of twelve samples of acute exercise were integrated together using aggregating multiple GEM groups (AGCR) by cell ranger. The Harmony (version 0.1.0) R package was used to eliminate batch effect. Clustering and cell subtype annotation After normalization and scaling, PCA was used for dimension reduction. The first 10 PCs and resolution 0.6 was used for clustering. Cell type annotation was conducted manually using the score of expression for canonical marker gene set of specific cell type. Differentially expressed genes identification and pathway enrichment Differentially expressed genes (DEGs) within each annotated cluster at a specific time point were determined using the FindAllMarkers function in Seurat by performing a Wilcoxon signed-rank test. The function performs the test pairwise between each time point and the baseline. Gene markers was filtered to include genes with an adjusted p value lower than 0.05 and an average log fold change higher than 1.0, thus omitting down-regulated genes. Statistics The paired t-test and single-factor ANOVA were used for statistical analysis. Data transparency All the sequencing data related to the human transcriptome described in this study have been deposited in the Genome Sequence Archive for Human (ngdc.cncb.ac.cn). All other datasets used and/or analyzed during the current study are available within the paper and its [217]supplemental information files. Acknowledgments