Monocyte promotes anti-virus response via distinctly expressing TNFSF10 and TNFSF13B
Two patients (designated PA0130 and PA0131) were admitted to the hospital with 5 or 6-day fever and dry cough. Neither patient had traveled to Hubei, but both had close contact with COVID-19 positive individuals travelling from Wuhan. Their throat swabs were positive for SARS-CoV-2 but negative for Influenza A/B virus, respiratory syncytial virus and adenovirus. Chest computed tomography (CT) further revealed progressive diffuse interstitial opacities and consolidation in the right lower lung fields (rapid progression stage) (Fig. 1A). Accordingly, both patients were diagnosed with moderate COVID-19 infection. PA0130 received lopinavir/ritonavir and low-dose prednisone after admission, and PA0131 additionally received acetylcysteine. The throat swab of PA0130 was negative for SARS-CoV-2 on day 21, while PA0131 showed viral clearance on day 13. Peripheral blood samples were collected from both patients on the day of admission and during the recovery phase, and subjected to single-cell RNA-seq, TCR-seq and BCR-seq (Fig. 1A and 1B).
Based on the single cell gene expression, the peripheral blood mononuclear cells (PBMCs) were clustered and annotated as CD4 T cell, CD8 T cell, NK, NKT, B cell, plasma cell, classical/non-classical monocyte, megakaryocyte, dendritic cell, and plasmacytoid dendritic cell (Fig. 1C, 1D and Fig. S1). The relative proportion of monocytes increased in the disease phase as per the UMAP position in both PA0130 and PA0131 (Fig. 1C and 1D), indicating changes in its gene expression pattern and function during the infection. TNFSF10 and TNFSF13B were distinctly expressed in the classical and non-classical monocytes, and were upregulated during the infection (Fig. 1E and 1F). TNFSF10 induces apoptosis in virus-infected or transformed cells, and mediates the host immune surveillance against these cells11. TNFSF13B is linked to T cell-independent B cell activation12. Thus, the increased number of circulating monocytes in the COVID patients likely plays a role in viral clearance. The megakaryocyte subset showed a 7.9-fold increase in PA0130, but decreased significantly in PA0131. Nevertheless, megakaryocytes were the most abundant source of TGFβ in both patients (Fig. 1E and 1F). Virus-induced damage of lung tissues and pulmonary endothelial cells leads to platelet activation, aggregation and entrapment in the lungs, resulting in the formation of thrombi at the infected site that recruit more platelets and megakaryocytes2. Therefore, it is possible that megakaryocytes are also involved in COVID infection and immunomodulation. Finally, the proportion of CD8 T cells increased in PA0131 but not PA0130, while B cells showed the opposite trend (Fig. 1C and 1D).
Clonal CD4 T cell expansion correlates low CD8 T cell exhaustion and multiple clonotypic expansions, and rapid virus clearance
Since adaptive immune response is central to anti-virus immunity, we integrated the total gene expression data with the TCR/BCR-seq data to gain more insights into the clonal expansion of distinct T and B cell subsets. The clusters 0, 2, 3 and 5 included the SELL+ CCR7+ TCF7+ naïve T cells (CD4 TNV) in PA0130. Cluster 6 and 9 consisted of a transient population between naïve and effector lymphocytes (CD8 TNV-EFF) which showed a decrease in the expression of lymphoid homing markers, and a corresponding increase in cytotoxic factors (PRF1, GZMB, GZMK, GZMH, or GNLY) and T cells activation markers (CCL5 and NKG7). Clusters 1 and 4 were defined as T effector and memory cells (CD8 TEFF and TEM) with low expression of homing markers, high expression of cytotoxic factors and IL7R+ KLRG1+ in the later. The high expression of CX3CR1 in cluster 1 indicated terminally- differentiated CD8 T cell effectors, while cluster 4 expressed high levels of IL7R and KLRG1, and intermediate levels of CD27 that are characteristic of long-lived memory cells (Fig. 2A and 2B).
An exhausted CD8 T cell subset expressing multiple co-inhibitory molecules was also observed in PA0130, and constituted 1.5% of all immune cells in the active disease phase before decreasing to 0.15% in the recovery phase (Fig. 2C). Interestingly, the exhausted T cells showed the lymphoid homing SELL+CD27+IL7Rint signature, and also expressed multiple cytotoxic factors like PRF1, GZMB, GZMH, GZMK and GNLY. However, the high CASP3 expression indicated an exhausted state and short-lived fate. Consistent with this, TCR seq data pointed to non-clonal expansion of the exhausted CD8 T cell population. However, one CD8 TEFF clonotype in cluster 1 expanded at 6.23% compared to the entire population, and was predominant in the recovery stage as well (4.57% on day 43). We hypothesized that this particular clone (TRA: CALRLEYYGNKLVF; TRB: CASSPGQGTLAYEQYF) was SARS-CoV-2 specific, and was activated upon antigen stimulation. In support of our hypothesis, there was no clonal expansion of T cells (> 1%) in two healthy blood samples (Fig. S2), indicating that the increased proportion of T cells in COVID-19 patients is stimulated by the SARS-CoV-2 antigens.
To further demonstrate T cell persistence in PA0130, we screened for clonotypes identical in the disease and recovery stages, and found them localized in the CD8 TEFF and CD8 TEM clusters (Fig. 2D). We did not observe CD4 T cell clonal expansion during the disease stage, although a very small percentage of CD4 TNV-EFF cells was identified (Fig. 2C). Unlike the CD8 subsets, there were no overlapping CD4 T cell clonotypes between the disease and recovery stages, indicating that the CD4 T cells were short-lived and not persistent in circulation during the disease (Fig. 2D), which is supported by the CD4 TNV functional annotation as well (Fig. 2A).
The T cells were similarly annotated in PA0131, and the CD4/CD8 naïve, effector and memory subsets were identified (Fig. 3A and 3B). Interestingly, PA0131 harbored expanded clonotypic CD4 T cells in circulation (Fig. 3C), and there were two major clonotypes with dramatically high expansion rates of 7.29% and 6.29% during the disease stage. Both clones persisted till day 47 and their expansion rates were 4.35% and 3.31% respectively. In addition, multiple CD8 T cell clonotypes were also detected in the disease phase that persisted during recovery (day 47), and no exhausted CD8 T cells were observed in this patient (Fig. 3A). The majority of clonally expanded T cells were CD4 TEM or CD8 TEM, and both populations expressed IL7R indicative of long-lived memory T cells (Fig. 3B).
A pseudotime trajectory was drawn to analyze the divergence between the expanded and non-expanded CD4 TEFF populations, using two clonotypes (1 and 2) with high expansion rates and the non-expanded cells in cluster 3. The pre-branch consisted of the non-expanded CD4 TEFF, while the two branches corresponded to the mixture of clonotypes1 and 2. The non-expanded cells expressed high levels of SELL and ribosome-related genes but had lower cytokine gene expression compared to the expanded clonotypes, indicating that the former is an early-differentiation effector population and the latter is the major anti- viral effector. The difference between two branches on the trajectory is cell fate 1 branch showed enhanced cytotoxic capability with higher granule expressions than the other branch. Both clonotype 1 and 2 had distribution on the two branches, indicating each clonotype comprises heterogeneously differentiated effectors. High usage of TRBV7-9 or TRBV20-1 double confirmed that the populations come from clonal expansion. (Fig. 3E)
Interestingly, the analysis of BCR seq integrated with gene expression of both PA0130 and PA0131 revealed that there are limited clonotypes of B cells with only slightly clonal expansions (Fig. S3A and 3B).
To summarize, the rapid virus clearance in PA0131 correlated to the expansion of multiple CD4 and CD8 T cell clonotypes, and lack of exhausted CD8 T cells. In contrast, PA0130 took longer to recover, and exhibited CD4 clonotypic expansion absence, only one predominant CD8 T cell clone as well as exhausted CD8 T cells expressing multiple co-inhibitory molecules. To validate the correlation between the host immunophenotype and viral clearance, we performed the same analysis on the blood samples of two additional patients with early (PA0133; 14 days) and late (PA0132; 29 days) viral clearance respectively (Table. 1). Similar to PA0130, PA0132 did not show any clonal expansion of CD4 T cells and only one expanded CD8 T cell clonotype (> 2%; Fig. 4A), along with a short-lived CASP3+IL7Rlow exhausted CD8 T cell population expressing 6 co-inhibitory molecules (Fig. 4B). In contrast, PA0133 had multiple expanded CD4 and CD8 TEM clones in circulation, and lack of CD8 T cell exhaustion similar to PA0131 (Fig. 4C and 4D). In addition, similar to those in PA0130 and PA0131, both PA0132 and PA0133 had limited B cell clonal expansions (Fig. S4).
Exhausted CD8 TEFF cells in SARS-CoV-2 infection are featured by GZMA production and programmed by SUB1 and HMGB
To further determine the accurate signature of exhausted CD8 T cells, we profiled the expression patterns of cytokines, chemokines, co-inhibitory/stimulatory molecules and transcription factors across the exhausted, non-expanded and expanded CD8 T cell subsets in PA0130 and PA0132. The genes differentially expressed across the 3 cell populations in more than 25% of the cells are shown in Figure 5A. Multiple co-inhibitory molecules were highly expressed only in the exhausted CD8 T cells, which incidentally also expressed the co-stimulatory molecules CD27int and CD28int. In addition, GZMB and PRF1 rather than IFNγ were identified as the major cytotoxic factors produced by the effector CD8 T cells against SARS-CoV-2 virus. We identified GZMA as the signature effector cytokine of the exhausted CD8 T cells, whereas CCL5 and CCL4 were the signature chemokines of the exhausted and expanded T cell subsets respectively. The highly expanded CD8 effector T cell clones expressed CX3CR1, which is indicative of terminal effector differentiation to the memory cells13,14. Studies on the mouse chronic virus infection such as LCMV and SIV models show that the balance between TBX21 and EOMES determines the T cell effector versus exhaustion fate. In the COVID-19 however, neither transcriptional factor was a determinant of T cell fate, and was expressed by a very small percentage of cells. In contrast, SUB1 and HMGB were the signature transcription factors of exhausted CD8 T cells, while PLEK and KLF2 programmed clonotypic CD8 T cells (Fig. 5A). However, the exact functional role of these transcriptional factors remains to be elucidated.
Co-expression of LAG3, Galectin-9 and SLAMF6 is the predominant inhibitory receptor signature of exhausted CD8 TEFF cells during SARS-CoV-2 infection
Next, we analyzed the co-inhibitory molecule co-expression pattern in the exhausted TEFF cells of PA0130 and PA0132. More than half of the exhausted CD8 TEFF cells in PA0130 and PA0132 co-expressed 4 molecules of LAG3, Gelectin-9 (LGALS9), SLAMF6, PD1 (PDCD1), CTLA4, TIM3 (HAVCR2), and TIGIT, 79.31% and 77.78% of the cells expressed at least 3 receptors, and 91.67% and 93.10% expressed at least 2 molecules in PA0130 and PA0132 respectively (Figure S5). Multiple co-inhibitory molecules co-expression parallels the severity of the exhaustion status. Consistent with this, the disease was more severe in PA0132 compared to PA0130. Notably, LAG3, Galectin-9 and SLAMF6 rather than PD1 and TIM3 are the predominant inhibitory molecules in the exhausted CD8 T cells of COVID-19 patients (Fig. 5B). In addition, 66.67%, 61.11% and 47.22% of the exhausted CD8 T cells co-expressed LAG3 with LGALS9 or SLAMF6 and their combination respectively in PA0130, whereas 58.62%, 46.55% and 34.48% of the cells were in PA0132. In chronic virus infection and cancer immunity, the signature of inhibitory molecule expression of exhausted T cells is featured by co-expression of PD1, TIM3 and LAG315-17. We further included the analysis of the exhausted T cells from non-HBV related or chronic HBV related hepatocellular carcinoma. Consistently, PD1, TIM3 and LAG3 or their combination are predominant co- inhibitory molecules co-expressed in both exhausted populations, while Galectin-9 and SLAMF6 showed low expression and less concurrence with LAG3 expression (Fig. 5C).