Single-cell transcriptomics reveal a hyperacute cytokine and immune checkpoint axis after cardiac arrest in patients with poor neurological outcome

Abstract

Background: Most patients hospitalized after cardiac arrest (CA) die because of neurological injury. The systemic inflammatory response after CA is associated with neurological injury and mortality but remains poorly defined.

Methods: We determine the innate immune network induced by clinical CA at single-cell resolution.

Findings: Immune cell states diverge as early as 6 h post-CA between patients with good or poor neurological outcomes 30 days after CA. Nectin-2⁺ monocyte and Tim-3⁺ natural killer (NK) cell subpopulations are associated with poor outcomes, and interactome analysis highlights their crosstalk via cytokines and immune checkpoints. Ex vivo studies of peripheral blood cells from CA patients demonstrate that immune checkpoints are a compensatory mechanism against inflammation after CA. Interferon γ (IFNγ)/interleukin-10 (IL-10) induced Nectin-2 on monocytes; in a negative feedback loop, Nectin-2 suppresses IFNγ production by NK cells.

Conclusions: The initial hours after CA may represent a window for therapeutic intervention in the resolution of inflammation via immune checkpoints.

Funding: This work was supported by funding from the American Heart Association, Brigham and Women's Hospital Department of Medicine, the Evergreen Innovation Fund, and the National Institutes of Health.

Keywords: Nectin-2; Translation to patients; cardiac arrest; cytokine; immune checkpoint; inflammation; interferon γ; monocyte; natural killer cell; peripheral blood mononuclear cells; single-cell RNA sequencing.

Figure 1.. Distinct single-cell transcriptomic profiles distinguish innate immune cells from post-CA patients with poor or good neurological outcome early after CA.

A) Approach for single-cell RNA-seq analysis of PBMC from patients after OHCA. Gross clustering of scRNA-seq dataset shown in t-SNE plots by B) cell type, and C) by patient subcohort (healthy subjects or CA patients, divided by time post-arrest). D) Fraction of each cell type in PBMC. Good and Poor denotes good and poor neurological outcomes 30 days after CA, respectively. E) Principal component analysis of NK cells and monocytes at 6h and 48h post-CA, with individual patients and their subcohorts shown. CA, cardiac arrest; neuro, neurological outcomes; NK, natural killer; OHCA, out-of-hospital CA; PBMC, peripheral blood mononuclear cells; PC, principal component; RNA-seq, RNA sequencing.

Figure 2.. Hyperacute expansion of NECTIN2^high monocytes enriched for IFNγ-response pathways correlated with poor neurological outcomes after clinical CA.

A) t-SNE plots of fine clustering of monocytes in the scRNA-seq dataset of PBMC. B) Proportions of monocyte subclusters. C) M1 and M2 gene signature scores for monocyte fine clusters 2 and 4. Average expression level for genes in the M1 or M2 signature (listed in Table S2) are shown. D) Cluster-defining genes for monocyte subclusters, with full list in Table S3. E) At 6h post-CA, differential gene expression analysis between the dominant monocyte clusters in patients with poor outcomes (cluster 4) and patients with good outcomes (cluster 2). See also Table S4. D-E. The immune checkpoint gene confirmed in validation cohort is highlighted (red). F) Gene set enrichment analysis of monocyte clusters. Enrichment plot for IFNγ signaling shown. Z score was calculated by average gene expression of genes in the pathway from each cluster. See also Figure S2. CA, cardiac arrest; FC, fold-change; FDR, false discovery rate; neuro, neurological outcomes. C, Mann-Whitney U test, ***P < 0.001.

Figure 3.. Hyperacute expansion of TIM3⁺TIGIT⁺NK cells correlate with poor neurological outcomes after clinical CA.

A) t-SNE plots of fine clustering of NK cells in the scRNA-seq dataset of PBMC. B) Proportions of NK cell subclusters. C) Violin plots for FCGR3A (CD16) and NCAM1 (CD56) expression in NK cell fine clusters. D) Quantification of NK cell subsets of healthy subjects and patients 6h post-CA by CD56 and CD16. Flow cytometry gating and 48h post-CA shown in Figure S3. E) Cluster-defining genes for NK cell subclusters, with full list in Table S3. F) At 6h post-CA, differential expression analysis between the dominant NK cell clusters in patients with poor outcomes (cluster 1) and in patients with good outcomes (cluster 2). See also Table S4. D-E) Immune checkpoint genes confirmed in validation cohort are highlighted (red). G) Gene set enrichment analysis of NK cell clusters. Z score was calculated by average gene expression of genes in the pathway from each cluster. Enrichment plots for glycolysis and hypoxia shown. CA, cardiac arrest; FC, fold-change; FDR, false discovery rate; neuro, neurological outcomes. C, D, Kruskal-Wallis test with Dunn’s multiple comparisons, *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 4.. Validation cohort confirms expansion of Nectin-2⁺ monocyte and Tim-3⁺NK cell states in CA patients with poor neurological outcomes.

A) Approach for confirmation of scRNA-seq results in validation cohort of CA patients, with clinical characteristics in Table S5. B) Flow cytometry analysis of Nectin-2⁺ monocytes. % Nectin-2⁺ monocytes are shown (per patient). C) Flow cytometry analysis of Tim-3⁺ NK cells. % Tim-3⁺ NK cells are shown (per patient). B-C) Flow cytometry gating strategy shown in Figure S4. D) Nectin-2⁺ and Nectin-2⁻ CD14⁺ monocytes were sorted by flow cytometry and assessed by bulk RNA-seq. Differential expression (DE) of genes were calculated for Nectin-2⁺ monocytes from patients with eventual poor outcomes compared to Nectin-2⁻ monocytes from patients with eventual good outcomes (measured at 6h post-CA) or healthy subjects. Fold-change (FC) is shown for patients at 6h-post CA or healthy subjects. See also Table S6. E) Heatmap of bulk RNA-seq dataset is shown for sorted Nectin-2⁺ monocytes from 6h poor outcomes vs Nectin-2⁻ monocytes from 6h good outcomes and healthy subjects. Genes shown are differentially expressed genes identified in the scRNA-seq dataset between Nectin-2⁺ monocyte cluster 4 and Nectin-2⁻ monocyte cluster 2. See also Table S6. F) Gene set enrichment analysis (GSEA) of the bulk RNA-seq of sorted Nectin-2+/− monocytes using the same comparison as in (4D). G) DE of genes were calculated for Nectin-2⁺ monocytes from patients with eventual poor outcomes compared to Nectin-2⁺ monocytes from patients with eventual good outcomes (measured at 6h post-CA) or healthy subjects. Fold-change (FC) is shown for patients at 6h-post CA or healthy subjects. See also Table S5. H) Heatmap of bulk RNA-seq dataset is shown for sorted Nectin-2⁺ monocytes from 6h poor outcomes vs Nectin-2⁺ monocytes from 6h good outcomes and healthy subjects. I) GSEA of bulk RNA-seq of sorted Nectin-2+/+ monocytes using the same comparison as in (4G). J) Tim-3⁺ and Tim-3⁻ NK cells were sorted by flow cytometry and assessed by bulk RNA-seq. Differential expression (DE) of genes were calculated for Tim-3⁺ NK cells from patients with poor outcomes post-CA compared to Tim-3⁻ NK cells from patients with good outcomes post-CA or healthy subjects. FC is shown for patients at 6h post-CA or healthy subjects. See also Table S6. K) Heatmap of bulk RNA-seq dataset is shown for sorted Tim-3⁺ NK cells from 6h poor outcomes vs Tim-3⁻ NK cells from 6h good outcomes and healthy subjects. Genes shown are differentially expressed genes identified in the scRNA-seq dataset between Tim-3⁺ NK cluster 1 and Tim-3⁻ NK cell cluster 2. See also Figure S4. L) GSEA of the bulk RNA-seq of sorted Tim-3+/− NK cells using the same comparison as in 4G. CA, cardiac arrest; FC, fold-change; FDR, false discovery rate; OHCA, out-of-hospital cardiac arrest; RNA-seq, RNA sequencing. Good and Poor denote good and poor neurological outcomes 30 days after CA, respectively. B, C, Kruskal-Wallis test with Dunn’s multiple comparisons or Mann-Whitney U test, *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 5.. Interactome analysis of NECTIN2^high monocyte and TIM3⁺TIGIT⁺ NK cell states identifies immune checkpoint and pro- and anti-inflammatory axes in cardiac arrest patients with poor neurological outcomes.

The NicheNet algorithm was applied to the scRNA-seq dataset and, at 6h post-cardiac arrest, CA patients with poor neurological outcome are compared to patients with good neurological outcome. Significantly increased interactions are shown for: A) Ligands secreted by NECTIN2^high monocytes binding to receptors on HAVCR2⁺TIGIT⁺ NK cells. NK cells ligands to receptors on monocytes. B) Ligands secreted by HAVCR2⁺TIGIT⁺ NK cells binding to receptors on NECTIN2^high monocytes. C) Ligands secreted by NECTIN2⁺ monocytes binding to receptors on NECTIN2^high monocytes (autocrine).

Figure 6.. Hyperacute elevation in plasma levels of pro- and anti-inflammatory cytokines distinguish CA patients with poor neurological outcomes.

A) Approach to assessment of cytokines identified by interactome analysis in a separate validation cohort of CA patients, with clinical characteristics in Table S7. B) Cytokines and chemokines measured by multiplex ELISA of plasma from CA patients with good or poor neurological outcome at time points post-CA or healthy subjects. CCL, C-C Motif Chemokine Ligand; IFNγ, interferon-gamma; IL, interleukin; TNFα, tumor necrosis factor-alpha. Mann-Whitney U test with Benjamini-Hochberg correction for multiple comparisons, *P < 0.05, **P < 0.01, ***P < 0.001. Good and Poor denotes good and poor neurological outcomes, respectively.

Figure 7.. Nectin-2 mediates a negative feedback loop to limit production of IFNγ by Tim-3⁺TIGIT⁺ NK cells.

A) Approach for: in vitro study of monocytes from healthy human subjects (7B); in vitro study of PBMC from healthy subjects (7C-D); and ex vivo study of PBMC from patients at 6h post-CA with poor neurological outcome (7E-G). B) Monocytes from healthy subjects were treated in vitro with cytokines identified by interactome analysis (Figure 5). NECTIN2 mRNA expression levels measured in monocytes by quantitative PCR (qPCR) are shown. Expression is normalized to housekeeping gene and to unstimulated control for each subject. C) Flow cytometric analysis of Nectin-2⁺ monocytes from healthy subjects after in vitro stimulation. Representative plots and % Nectin-2⁺ monocytes (per patient) are shown. D) Intracellular IFNγ in Tim-3⁺TIGIT⁺ NK cells measured by flow cytometry after PBMC from healthy subjects treated in vitro with cytokine (IFNγ+IL-10) stimulation after treatment with anti-Nectin-2 blocking monoclonal antibody (mAb) or isotype control. Representative plot and % IFNγ⁺Tim-3⁺TIGIT⁺ NK cells (per patient) shown. E) % IFNγ⁺Tim3⁺TIGIT⁺ and IFNγ⁺Tim-3⁻TIGIT⁺ NK cells at 6h post-CA in patients with poor neurological outcome, measured by flow cytometry. F) % IFNγ⁺Tim-3⁺TIGIT⁺ NK cells in the presence or absence of Anti-Nectin-2 mAb. G) Intracellular IFNγ levels of early post-CA Tim-3⁺TIGIT⁺ NK cells after treatment with recombinant human (rh)-Nectin-2. OHCA, out-of-hospital cardiac arrest; B, C, Welch ANOVA test with Dunnett’s T3 multiple comparisons; D, F, G, ratio paired t-test; E, unpaired t-test. *P < 0.05, **P < 0.01, ***P < 0.001.