An “NK2” subpopulation exhibits the largest distribution divergence between young and elderly individuals

We initially profiled fresh human peripheral blood samples collected from 4 young (ages 21–28 years) individuals and 4 elderly (ages 65–68 years) individuals (Additional file 2: Fig. S1A and Additional file 1: Table S1). Lymphocytes were then sorted and subjected to scRNA-seq using the 10X platform (Additional file 2: Fig. S1B). After rigorous quality control (QC) processing, we retained a total of 11,279 high-quality single transcriptomes (Additional file 2: Fig. S2A, B). Of these, 4930 cells were from young individuals, and 6349 cells were from elderly individuals. We applied Seurat [28] (version 3.2.2) to integrate the single-cell transcriptomes from young and elderly individuals and identified 9 unique immune cell subpopulations, which were visualized via t-distributed stochastic neighbour embedding (t-SNE) (Fig. 1A). We then applied Souporcell [36], a genotype-based unmixing method that can deconvolve scRNA-seq data for assigning cells to their donor of origin, to help assess the per donor representation of the different cell clusters and to identify any batch effects for individual samples. We obtained 96.6% (10,894/11,279) of cells with individual sample identity and observed that each cell cluster was composed of cells originating from the 4 young and 4 elderly individuals (Additional file 2: Fig. S2C, D), indicating that our resulting clusters were not driven by any single individual.

Fig. 1
figure 1

Ageing leads to a marked increase in the proportion of memory-like NK2 cells among total blood lymphocytes. A t-SNE representations of the integrated single-cell transcriptomes of 11,279 PBMCs, with 4930 cells from the 4 young individuals (ages 21–28 years) and 6349 cells from the 4 elderly individuals (ages 65–68 years). Cells are coloured by cell type identity. Each dot represents a single cell. B Dot plot showing the log2 (odds ratio) of the comparison between the cell proportions of each cell cluster in the young and elderly samples. C Scatterplot showing the logarithmic ratio between the estimated frequencies of each of the 9 cell clusters in young individuals (n = 10) and those in elderly individuals (n = 10) from bulk RNA-seq datasets (GEO103232). D Heatmap showing the top 30 differentially expressed genes for NK1 and NK2 cells from young and elderly individuals. E Violin plot showing the gene expression in NK1 and NK2 cells. F FACS staining strategy for NK1 (LinCD7+ NKG2CCD122high) and NK2 cells (LinCD7+ NKG2C+CD122low) from young (top) and elderly (bottom) individuals. G Bar graphs showing the proportions of NK2 cells in lymphocytes (left) and NK2 cells in NK cells (right) from young (n = 35) and elderly (n = 27) individuals. H Gene Ontology (GO) enrichment analysis of the differentially expressed genes of NK2 cells from young individuals and those from elderly individuals. The colour indicates the -log10 (P-value) enrichment for each GO term. I Flow cytometry analysis of IFN-γ in NK2 cells from young and elderly individuals with or without IFN-α stimulation in vitro. J Bar graphs displaying the frequencies of IFN-γ+ in the NK2 subpopulation from young (n = 13) and elderly (n = 15) individuals with or without IFN-α stimulation in vitro. K Flow cytometry analysis of CD107a in NK2 cells from young and elderly individuals with or without IFN-α stimulation in vitro. L Bar graphs displaying the frequencies of CD107a+ in the NK2 subpopulation from young (n = 8) and elderly (n = 9) individuals with or without IFN-α stimulation in vitro. The error bars represent the standard deviation (SD). *P < 0.05, **P < 0.01, ****P < 0.0001. P-values were obtained with two-sided Student’s t-tests. The results are representative of at least three independent experiment

Based on the expression of known marker genes, we identified lymphocyte lineages, including two NK cell subpopulations (NK1 and NK2), NKT cells, four T cell subpopulations (T1, T2, T3, and T4), and two B cell subpopulations (B1 and B2) (Additional file 2: Fig. S2E and Additional file 1: Table S2). Among the 9 immune cell subpopulations defined in our scRNA-seq data, we found that the proportion of NK2 cells among total lymphocytes was 1.89-fold higher in the elderly than in the young individuals; this was the subpopulation exhibiting the most dramatic enrichment in the distribution in the elderly individuals (Fig. 1B). We also downloaded bulk RNA-seq data for peripheral blood mononuclear cells [32] (PBMCs) from young (n = 10) and elderly (n = 10) individuals and used the signature genes identified in our single-cell analysis to assess the composition of immune cell subpopulations in this dataset (see the “Methods” section and Additional file 1: Table S3). As in our scRNA-seq data, we again found that the NK2 subpopulation displayed the largest distribution change between the young and elderly individuals (P < 10−4, two-sided Student’s t-tests) (Fig. 1C). These results indicate that the NK2 subpopulation may represent an age-related NK subpopulation in humans.

NKG2C+CD122low NK2 cells increase with ageing and have a memory-like phenotype

We next assessed differentially expressed genes (DEGs) between NK1 and NK2 cells in our scRNA-seq data (Fig. 1D and Additional file 1: Table S4) and noticed that NK2 cells showed reduced expression levels of genes including FCER1G (FcεRγ), SH2D1B (EAT-2), and ZBTB16 (PLZF) and elevated expression of KLRC2 (NKG2C), among others, compared with NK1 cells (Fig. 1E). Since it has been reported that reduced expression of FcɛRγ, PLZF, and EAT-2 and increased expression of NKG2C correlate with a memory NK cell phenotype [37, 38], we considered the NK2 subpopulation to be phenotypically memory-like NK cells.

We further found two surface marker genes, IL2RB (also known as CD122) and KLRC2 (NKG2C), which can be used to distinguish the NK1 and NK2 subpopulations (Fig. 1E). Using flow cytometry, we confirmed the existence of these two NK cell subpopulations within the LinCD7+ NK cell population in the human blood and characterized NK1 cells as NKG2CCD122high and NK2 cells as NKG2C+CD122low (Fig. 1F). Both the two NK cell populations showed high expression (> 95%) of NK-defining surface molecules, including CD56 (Additional file 2: Fig. S3A-D), CD16 (Additional file 2: Fig. S3E-H), and NKp80 (Additional file 2: Fig. S3I-L). We then confirmed in peripheral blood samples from 35 young and 27 elderly healthy donors that the percentages of NK2 cells (NKG2C+CD122low) among total lymphocytes and among NK cells were significantly higher in the elderly than in the young individuals (Fig. 1G); the percentages of NK1 cells (NKG2CCD122high) showed the opposite trend (Additional file 2: Fig. S4A, B).

The expansion of adaptive or memory-like NK cell subsets has been observed in association with the CMV serostatus [39,40,41,42,43]. Therefore, we compared the proportion of NKG2C+CD122low memory-like NK2 cells among total blood lymphocytes and NK cells of CMV-seronegative (CMV−) and CMV-seropositive (CMV+) donors (Additional file 2: Fig. S5A-C). Indeed, NKG2C+CD122low memory-like NK2 cells were detected more reliably in the blood of CMV+ adult donors (~ 19% in NK cells) than that in CMV− adult donors (~ 5% in NK cells), despite the non-significant differences in the subset distribution of NK cells in the young individuals (Additional file 2: Fig. S5A). It is worth mentioning that the presence of NKG2C+CD122low memory-like NK2 cells was correlated with CMV seropositivity, but not all CMV-seropositive donors had detectable (at least by flow cytometry analysis) memory-like NK2 cell subpopulations (Additional file 2: Fig. S5B, C). These results were consistent with those of previous studies showing that not all CMV+ individuals have circulating memory NK cells [37, 44].

Intriguingly, we further confirmed in peripheral blood samples from 35 young and 27 elderly healthy donors that the percentages of NKG2C+CD122low memory-like NK2 cells among total lymphocytes and among NK cells were significantly higher in the CMV+ elderly (~ 40% in NK cells) than in the CMV+ young individuals (~ 19% in NK cells) (Additional file 2: Fig. S5B, C). These results suggest a memory-like NK subpopulation exhibiting an age-related increase (considering CMV serostatus), as measured by scRNA-seq and supported by independent flow cytometry results.

Gene Ontology (GO) analysis indicated that compared to NK1 cells, NK2 cells showed enrichment for genes related to lymphocyte activation in our scRNA-seq dataset (Additional file 2: Fig. S6A). We further found that the DEGs of NK2 cells from elderly individuals, compared to young individuals, were enriched for functional annotations related to interferon alpha/beta signalling (Fig. 1H and Additional file 1: Table S5). NK cell effector functions are mediated by CD107a expression and IFN-γ production [45]. We subsequently examined peripheral blood samples from an independent cohort of young and elderly individuals using flow cytometry to detect the secretion of IFN-γ and CD107a in NK1 or NK2 cells during in vitro IFN-α stimulation. Although there were no significant differences in the secretion of IFN-γ and CD107a by NK1 cells when comparing elderly individuals with young individuals upon IFN-α stimulation (Additional file 2: Fig. S6B-E), we found significantly higher IFN-γ and CD107a levels in NK2 cells from elderly individuals than in those from young individuals following in vitro stimulation with IFN-α (Fig. 1I–L and Additional file 1: Table S1).

We also examined the peripheral blood samples from young and elderly individuals using flow cytometry to detect the production of IFN-γ and CD107a in NK1 or NK2 cells upon co-stimulation with interleukin (IL)-12 and IL-15. In young individuals, we found that memory-like NK2 cells displayed decreased responsiveness to IL-12+IL-15 stimulation compared to non-memory-like NK1 cells, as shown by the fact that IL-12+IL-15 induced significantly higher both IFN-γ (Additional file 2: Fig. S7A-C) and CD107a (Additional file 2: Fig. S7D-F) production by NK1 cells than that induced by NK2 cells from young individuals, which is in line with the findings of a previous report [37]. However, in elderly individuals, the responsiveness to IL-12+IL-15 co-stimulation appeared to be similar between memory-like NK2 cells and non-memory-like NK1 cells, because there were no significant differences in the production of IFN-γ and CD107a between NK2 and NK1 cells from elderly individuals (Additional file 2: Fig. S7A-F).

Next, we assessed the functional differences in NK1 and NK2 cells after interaction with classical NK cell targets (K562 cell line). We found that there were no significant differences in the production of IFN-γ and CD107a in NK1 cells upon K562 stimulation between young and elderly individuals (Additional file 2: Fig. S8A-D). Although slightly but significantly higher IFN-γ levels in NK2 cells from elderly individuals than in those from young individuals following in vitro stimulation with K562 (Additional file 2: Fig. S8E, F), no significant difference in the production of CD107a was detected in NK2 cells after interaction with K562 cells between young and elderly individuals (Additional file 2: Fig. S8G, H). These results indicated that age-related changes in human NK cell functionality may not be related to target cell-mediated killing function, but instead to proinflammatory cytokine secretion and type I interferon response status.

Previous studies have reported age-related impairment in IL-2 signalling in NK cells from elderly individuals [46, 47]; we therefore examined the effects of IL-2 on the distinct NK cell population. No significant difference in the production of CD107a and IFN-γ was detected upon IL-2 stimulation of NK1 cells from elderly individuals (Additional file 2: Fig. S9A-D). However, the response to IL-2 in NK2 cells from elderly individuals was found to be impaired when IFN-γ was considered (Additional file 2: Fig. S9E, F), whereas CD107a production was not significantly affected (Additional file 2: Fig. S9G, H). These results provided additional support for the conclusion that NKG2C+CD122low memory-like NK2 cells are age-related NK cells.

Taken together, our findings are in line with previous studies reporting that exposure of NK cells to a combination of IL-12 and IL-15 results in memory-like cell behaviours in the absence of antigen, characterized by enhanced effector functions and responses when they are restimulated with cytokines [9, 48]. These findings further support the idea that NK2 cells are phenotypically memory-like NK cells.

ScRNA-seq of human blood NK cells identifies a unique subset of memory-like NK2.1 cells that is enriched in elderly individuals

To determine whether additional cellular diversity exists and gain deeper insights into the age-related functional divergence of NK cells, we conducted 10X single-cell transcriptome sequencing on purified NK cells from the same cohort of 4 young and 4 elderly healthy individuals examined above (Fig. 2A). We sorted LinCD7+ NK cells among lymphocytes to obtain cell populations covering all known developmental stages for NK cells and for type 1 innate lymphoid cells (ILCs) [15, 38]. After QC processing, we obtained a total of 12,234 high-quality NK cells, of which 5501 cells were from young individuals and 6733 were from elderly individuals (Additional file 2: Fig. S10A, B). We used Seurat [28] (version 3.2.2) to integrate the young and elderly samples and identified 6 NK cell subsets (namely, NK1.1, NK1.2, NK2.1, NK2.2, NK2.3, and NK2.4 cells), which were represented using uniform manifold approximation and projection (UMAP) (Fig. 2B). We found that compared to both NK1.1 and NK1.2 cells, NK2.1, NK2.2, NK2.3, and NK2.4 cells all exhibited low expression of genes, including FCER1G (FcεRγ), SH2D1B (EAT-2), and ZBTB16 (PLZF), and high expression of KLRC2 (Fig. 2C and Additional file 2: Fig. S10C). This expression pattern correlates with a memory NK cell phenotype, suggesting that NK2.1, NK2.2, NK2.3, and NK2.4 cells are phenotypically memory-like NK cells.

Fig. 2
figure 2

ScRNA-seq analysis of aged human blood NK cells reveals age-associated alterations in memory-like NK cells. A Peripheral NK cells (LinCD7+) were sorted from young and elderly individuals and analysed with the 10X Genomics single-cell sequencing platform. B UMAP projections of 12,234 NK cells. The different colours represent the 6 NK cell subpopulations (left). Each dot represents a single cell. C, Violin plot showing the gene expression of IL2RB, FCER1G, and KLRC2 in each cell subpopulation. D Heatmap showing the differentially expressed genes among the NK cell subpopulations from young and elderly individuals. E GO enrichment analysis showing the terms of the differentially expressed genes for the indicated NK cell subpopulations from young and elderly individuals. The colour indicates the -log10 (P-value) enrichment for each GO term. F UMAP projections of 11,338 NK cells, with 5154 from young individuals (middle) and 6184 cells from elderly individuals (right). Each dot represents a single cell. G Dot plot showing the proportions of the NK cell subpopulations in an elder vs. young comparison. H Dot plot showing a comparison of the number of differentially expressed genes in the indicated NK cell subpopulations in an elderly vs. young comparison. I PAGA graph showing the connectivity between the NK subgroups in the young and elderly groups. Each of the coloured nodes is an NK subpopulation, and the node size is proportional to the number of cells in the subpopulation. The thickness of the edge shows the strength of the connectivity between subpopulations. DEGs, differentially expressed genes

We then characterized the potential functional annotations of these memory-like NK cell subsets. GO analysis of subset-defining DEGs (e.g. NK2.1 vs. the other NK cell subsets) among these memory-like NK cell subsets indicated that the upregulated DEGs of NK2.1 cells exhibited enrichment for interferon alpha/beta signalling, with high expression levels of genes including IL32, IFI6, ISG15, and IFI44L. The DEGs of NK2.2 cells were enriched for functional annotations related to ribosome assembly, with high expression of genes including RPS26, RPS18, and RPL3. The DEGs of NK2.3 cells showed enrichment for antigen processing and presentation, and the highly expressed genes included CD74, HLA-DPB1, and HLA-DAP1 (Fig. 2D, E and Additional file 1: Table S6). Although we performed scRNA-seq analysis using sorted LinCD7+ NK cells and excluded CD3 expression at the protein level, NK2.4 cells still showed CD3D and CD3G expression at the mRNA level (Fig. 2D and Additional file 2: Fig. S10D). Therefore, we speculate that this cell subset may correspond to previously described activated NKT cells with low expression of the TCR complex [49].

Among the 6 NK cell subsets defined in this scRNA-seq data, we found that the proportion of NK2.1 cells among total NK cells was 1.95-fold higher in the elderly than in the young individuals; this was the subset exhibiting the most dramatic change in distribution between the two age groups (Fig. 2F, G). We next performed pairwise comparisons of the NK cell subsets from elderly individuals and the corresponding cell subsets from young individuals, which identified a total of 253 DEGs (Fig. 2H). When assessing the number of DEGs for each of the NK cell subsets, there were clearly more DEGs in NK2.1 cells than in the other subsets (NK1.1 cells, NK1.2 cells, NK2.2 cells, and NK2.3 cells) (Fig. 2H), suggesting that NK2.1 cells showed the largest transcriptomic changes among NK cells during ageing.

To more intuitively quantify the connectivity of partitions (the NK1.1, NK1.2, NK2.1, NK2.2, and NK2.3 cell subsets) of the single-cell graph, partition-based graphical abstraction (PAGA) [33] was used to generate a much simpler abstracted graph (PAGA graph) of partitions, in which edge weights represent confidence in the presence of connections. We noticed that the connectivity of neighbourhoods appeared to be reduced among NK2.1, NK2.2, and NK2.3 cells from the elderly individuals compared to those from the young individuals (Fig. 2I). Summarizing the above findings, we determined the NK cell hierarchies in the peripheral blood in the context of ageing and identified a unique subset of memory-like NK2.1 cells that is enriched in elderly individuals.

NK2.1 cells in elderly individuals represent a terminal stage of human NK cell differentiation

The impacts of ageing on the development and maturation of NK cell subsets are not well understood in humans [44]. We applied PAGA [33], a high-resolution pseudotime prediction algorithm, to construct differentiation potential trajectories for NK2.1, NK2.2, and NK2.3 cells from young and elderly individuals. Several experimental evidences have shown that NK cell development proceeds from a CD56bright to CD56dim phenotype [50, 51]. We then tried to define the starting point of the putative developmental trajectory and identified that the NK1.1 cells showed enrichment of gene expression for the CD56bright NK cell signature genes, whereas the NK2.1, NK2.2, and NK2.3 cells were enriched for CD56dim NK cell signature genes (Additional file 1: Table S7 and Additional file 2: Fig. S11A, B). The NK1.1 (CD56bright-like NK) subset was, therefore, defined as the starting point of the putative developmental trajectory. Pseudotime analysis with the PAGA algorithm indicated that NK2.1 and NK2.3 cells were distributed on the two branches of the trajectory (Fig. 3A, B). In addition, we found that NK2.1 cells in elderly individuals were projected at the end of one branch along the developmental trajectory (Fig. 3A–C), suggesting that these cells were in the terminal differentiated state in elderly individuals.

Fig. 3
figure 3

Pseudotime analysis reveals the distinct trajectories of memory-like NK cell differentiation during ageing. A–C Trajectories predicted using the PAGA algorithm for NK1.1, NK2.1, NK2.2, and NK2.3 cells from young and elderly individuals. Cells are coloured by NK cell subsets (A), by the pseudotime trajectory (B), and by age group (C). D Pie charts showing the composition of NK cell subsets in five bins; each bin was divided equally according to the pseudotime of the cells. E Violin plots of the cell pseudotime distribution for each NK cell subset from young and elderly individuals. P-values were estimated by Wilcoxon rank-sum test. F The box plot shows the expression of the top 50 age-associated genes (obtained from Peters et al. [31]) of NK2.1, NK2.2, and NK2.3 cells from young people and elderly people. P-values were obtained using the Wilcoxon rank-sum tests. G Dot plot representing the expression of age-related genes of NK2.1, NK2.2, and NK2.3 cells in the young and elderly groups. The plot shows genes expressed by at least 20% of the cells in the subset. The size of the dot represents the fraction of cells expressing the gene, and the colour of the dot represents the gene level as a z-score

To quantitatively illustrate the time order of the three subsets on the pseudotime trajectory, we divided the cells into five equal bins along pseudotime. Notably, ~ 95% of the NK2.1 cells positioned within the terminal pseudotime bin were from elderly individuals (Fig. 3D, E). At minimum, this finding suggests that NK2.1 cells from elderly individuals represent a terminal stage of human NK cell differentiation. A previous study performed a meta-analysis of a large cohort of 14,983 individuals and reported the top 50 age-related genes in human peripheral blood [31] (Additional file 1: Table S8). The constituent genes of this set are related to biological characteristics of personal health (e.g. blood pressure, cholesterol level, fasting blood sugar, and body mass index). Using this gene set for an analysis of each NK2.1, NK2.2, and NK2.3 cell subset, we found that the expression of age-related genes of the NK2.1 cell subset was apparently substantially higher than that of the other two subsets (Fig. 3F, G, P = 4.25e−18, Wilcoxon rank-sum test). This suggests that NK2.1 cells represent a terminal differentiation state for NK cells.

NK2.1 cells in elderly individuals display a transcriptional signature of elevated type I interferon signalling

Transcriptional differences in cells that occur during ageing drive altered functions, so we further investigated the age-related transcriptional differences of the NK2.1, NK2.2, and NK2.3 cell subsets in the DEG analysis between young and elderly samples. We observed that NK2.1 cells from the elderly samples had elevated expression of interferon signalling pathway genes (e.g. IFI6, ISG15) compared to that of NK2.1 cells from the young samples (Fig. 4A, B and Additional file 1: Tables S9-S10), suggesting that NK2.1 cells may be continuously exposed to interferon signals during ageing.

Fig. 4
figure 4

Age-associated transcriptional differences of distinct memory-like NK cell subsets. A Volcano plots of the differentially expressed genes of NK2.1 (left), NK2.2 (middle), and NK2.3 cells (right) from comparisons between young and elderly individuals. Each dot represents a gene, with significantly upregulated genes (lnFC > 0.25, P < 10−3) in young and elderly individuals coloured blue and red, respectively. B Heatmap of the enriched GO terms among the DEGs detected for NK2.1, NK2.2, and NK2.3 cells (elderly vs. young). The colour indicates the -log10 (P-value) enrichment for each GO term. P-values were obtained with the Wilcoxon rank-sum tests. C Heatmap of the AUC scores predicted by SCENIC for expression regulation by transcription factors (TFs) in NK2.1, NK2.2, and NK2.3 cells from young and elderly individuals. D, E UMAP plots showing the AUC of the estimated regulon activity for IRF7, POLR2A, JUN, and JUNB (D) in NK2.1, NK2.2, and NK2.3 cells and the expression of these TFs (E). FC, fold change

We also explored transcription factors (TFs) in NK cells that may regulate ageing-associated transcriptional programmes in NK cells. We used SCENIC [34] to predict TFs that may regulate the genes we detected as upregulated in NK2.1, NK2.2, and NK2.3 cells from elderly or young individuals (Fig. 4C and Additional file 2: Fig. S12A). There were 9 SCENIC-predicted TFs—IRF7, IRF9, STAT1, HMGB2, KLF6, POLR2A, EZH2, POU3F1, and KLF10—that appear to affect the observed ageing-associated transcriptomic changes in NK2.1 cells in elderly individuals (Fig. 4C, D and Additional file 2: Fig. S12A, B). IRF7 is a major regulator of type I interferon-dependent immune responses [52]. Notably, the mRNA expression level of IRF7 was higher in NK2.1 cells from elderly individuals than in NK2.1 cells from young individuals (Fig. 4E). Furthermore, SCENIC-predicted IRF7 may regulate 70.5% (31/44) of the upregulated DEGs in NK2.1 cells in elderly individuals, including the type I interferon signal pathway-related genes IFI6, MX1, ISG15, IFI44L, and IL32 (Additional file 2: Fig. S13A, B). Our finding that IRF7 expression is elevated, in combination with our detection of the enrichment of its motif, in NK2.1 cells from elderly individuals, suggests that this TF may transcriptionally regulate ageing-associated activation of the type I interferon signal transduction pathway.

NK2.1 cells in elderly individuals are predominantly CD52+ NK2 cells, exhibit proinflammatory characteristics, and display a type I interferon response state

Our single-cell transcriptome analysis revealed that NK2.1 cells accumulated significantly in elderly individuals (Fig. 2G). We performed pairwise comparisons among NK2.1, NK2.2, and NK2.3 cells and found that NK2.1 cells had relatively high expression of the CD52 gene (Additional file 2: Fig. S14A, B). CD52 has been reported as an active target for the management of CMV reactivation [53]. To validate this as a surface marker for NK2.1 cells, we analysed blood samples from 35 young and 27 elderly healthy individuals by flow cytometry with gating for CD52 and for NKG2C and CD122, memory-like NK2 markers identified in our scRNA-seq analysis (Additional file 2: Fig. S15). Beyond supporting the presence of NK2.1 cells (Fig. 5A), this analysis also confirmed that the proportion of NK2.1 (CD52+ NK2) cells among NK cells or among total lymphocytes was significantly elevated in elderly individuals (Fig. 5B, P < 0.0001, Student’s t-tests). NK2.1 cells also showed high expression (> 95%) of NK-defining surface molecules, including CD56 (Additional file 2: Fig. S16A, B), CD16 (Additional file 2: Fig. S16C, D), and NKp80 (Additional file 2: Fig. S16E, F). In addition, when taking CMV serostatus into account, we also found that the percentages of CD52+NKG2C+CD122low memory-like NK2.1 cells among total lymphocytes and among NK cells were significantly higher in the CMV+ elderly (~ 35% in NK cells) than in the CMV+ young individuals (~ 12% in NK cells) (Additional file 2: Fig. S17A, B).

Fig. 5
figure 5

Age-associated memory-like CD52+ NK2 cells (NK2.1) exhibit proinflammatory characteristics and display a type I IFN response state. A Representative flow cytometry analysis of the percentage of NK2.1 cells within NK2 cells (LinCD7+NKG2C+CD122low) from young (left) and elderly (right) individuals. B Bar graphs showing the proportion of NK2.1 cells among NK cells (left) and among lymphocytes (right) from young (n = 35) and elderly (n = 27) individuals. C Representative density plots showing the expression of IFN-γ in NK2.1 cells from young and elderly individuals with or without IFN-α stimulation in vitro. D Bar graphs displaying the frequencies of IFN-γ+ cells in NK2.1 cells from young (n = 13) and elderly (n = 12) individuals with or without IFN-α stimulation in vitro. E Representative density plots showing the expression of CD107a in NK2.1 cells from young and elderly individuals with or without IFN-α stimulation in vitro. F Bar graphs displaying the frequencies of CD107a+ cells in NK2.1 cells from young (n = 7) and elderly (n = 7) individuals with or without IFN-α stimulation in vitro

As NK2.1 cells from elderly individuals exhibited enrichment of genes related to interferon alpha/beta signalling in our scRNA-seq analysis (Fig. 4B), we investigated the response sensitivities of NK2.1 cells from elderly and young individuals to type I interferon stimulation. We analysed the expression levels of IFN-γ and CD107a in NK2.1 cells from an independent cohort of young and elderly individuals following in vitro stimulation with recombinant human IFN-α (Additional file 1: Table S1). NK2.1 cells from the elderly group had significantly higher IFN-γ and CD107a levels than NK2.1 cells from the young group (Fig. 5C–F). Furthermore, NK2.1 cells from the elderly individuals appeared to be more responsive to type I interferon stimulation than young NK2.1 cells, as evidenced by significantly increased levels of IFN-γ and CD107a (Fig. 5C–F). In sum, these results indicate that a marked increase in the proportion of NK2.1 cells—which exhibit proinflammatory characteristics and display a type I interferon response state—might represent an immune cell distribution signature for immune ageing.

NK2.1 cells from elderly COVID-19 patients are enriched in type I signalling, which is positively correlated with disease severity in COVID-19

Studies have shown that the risk for severe COVID-19 illness increases with age [54] and that NK cells undergo enhanced effector functional changes in COVID-19 patients [55]. However, the impacts of ageing on NK cell subsets in COVID-19 disease remain unclear. We downloaded single-cell RNA-seq datasets [23,24,25, 56] of peripheral immune cells from young (n = 15) and elderly (n = 41) COVID-19 patients and from young (n = 12) and elderly (n = 6) healthy control individuals. Specifically, the COVID-19 single-cell datasets included 41 samples taken from elderly patients with active disease (from severe disease, n = 25; from moderate disease, n = 1) and during the convalescent phase (from severe disease, n = 9; from moderate disease, n = 6) and 15 samples taken from young patients with active disease (with moderate disease, n = 5) and during the convalescent phase (from severe disease, n = 2; from moderate disease, n = 8) (Fig. 6A and Additional file 1: Table S11).

Fig. 6
figure 6

Integrative analyses of a large-cohort COVID-19 single-cell transcriptomic dataset with our single-cell datasets reveal that NK2.1 cells from elderly COVID-19 patients are enriched for type I interferon signalling which correlates with increased disease severity in COVID-19. A Flowchart showing the integrative analysis of the scRNA-seq datasets of peripheral blood NK cells obtained from healthy controls (HC, n = 18) and COVID-19 patients (n = 56) [23, 25, 56]. B, C UMAP projections of the integrated single-cell transcriptomes of 34,388 NK cells, with 17,748 cells from healthy control individuals and 16,640 cells from COVID-19 patients. Cells are coloured by subset identity (B), and the NK cell subsets were defined as in Fig. 2B (C). Each dot represents a single cell. D Bar graph showing the proportion of each NK cell subset in all cells. E Bar graph showing the proportion of cells derived from young or elderly samples for each of the NK cell subsets. F Histograms showing the number of DEGs for each NK cell subset in the total individuals (top), the elderly individuals (middle), and the young individuals (bottom) between COVID-19 patients and healthy controls. G Dot plots of enriched GO terms of differentially expressed genes among NK2.1 cells between elderly COVID-19 patients and elderly healthy control individuals. The colour of the dot indicates -log10 (P-value) enrichment for each GO term, and the size of the dot indicates the number of differentially expressed genes contained within each enriched GO term. H Box plots of the average expression of genes involved in the signalling pathway “response to type I interferon” in NK2.1 cells from young healthy control individuals, from elderly healthy control individuals, from young COVID-19 patients, and from elderly COVID-19 patients

We then extracted the single-cell transcriptomes of NK cells from COVID-19 patients and healthy control individuals and applied Seurat [28] (version 3.2.2) to integrate the COVID-19 single-cell transcriptomes of NK cells with those from healthy control individuals, enabling the analysis of a total of 34,388 NK cells (19,833 cells from elderly individuals and 14,555 from young individuals). We identified 9 NK cell subsets, which were represented using UMAP (Fig. 6B); note that by mapping our aforementioned NK cell subsets (Fig. 2B) to this UMAP, we found that 7 of the 9 subsets were coincident with NK cell subsets in the aforementioned single-cell analysis, and 2 of the 9 subsets appeared to be COVID-19-specific NK cells (Fig. 6B, C), albeit with low cell numbers (Fig. 6D). We again found that the proportion of NK2.1 among total NK cells was higher in the elderly than in the young COVID-19 patients and higher in the elderly than in the young healthy control individuals (Fig. 6E). Previous studies showed that the proportion of memory NK cells from COVID-19 patients is elevated in severe disease compared to moderate disease [55, 57]. Our integrative scRNA-seq data demonstrated that the percentage of memory-like NK2 cells among total NK cells was increased in elderly patients with severe disease compared to elderly patients with moderate disease (Additional file 2: Fig. S18).

Differential analysis of each NK cell subset between COVID-19 patients and healthy controls showed that there were few DEGs (n = 75) for NK cell subsets from young samples but many DEGs (n = 357) for NK cell subsets from elderly samples. Furthermore, we identified NK2.1 cells that showed the largest number of DEGs among all NK cell clusters in a comparison of elderly COVID-19 patients and elderly healthy controls (Fig. 6F and Additional file 2: Fig. S19), suggesting disease progression-related functions in NK2.1 cells from elderly COVID-19 patients. We also compared the predicted functions of NK2.1 cells in COVID-19 patients compared with healthy controls, and GO analysis indicated that the DEGs of NK2.1 cells from elderly COVID-19 patients, compared to elderly healthy controls, were enriched for functional annotations related to response to type I interferon (Fig. 6G, H and Additional file 1: Table S12; Additional file 2: Fig. S20A and Additional file 1: Table S13, P = 8.20e−35, Wilcoxon rank-sum test). We further found that age has a strong influence on the expression of genes related to type I interferon responses in COVID-19 patients (Additional file 2: Fig. S20B). Specifically, the NK2.1 cells of elderly patients have high expression of genes with functional annotations related to response to type I interferon (e.g. ISG15, ISG20, etc.) compared to that of NK2.1 cells from young COVID-19 patients (Fig. 6H, P = 1.11e−36, Wilcoxon rank-sum test).

Because type I interferon molecules are known to exhibit a wide range of antiviral activities and given the numerous reports of more severe type I interferon responses in patients with severe COVID-19 [58,59,60], multiple clinical trials have used such molecules as potential therapeutic agents to treat COVID-19. We observed elevated expression of genes involved in response to type I interferon in NK2.1 cells from elderly COVID-19 patients in the severe stage compared with those from elderly COVID-19 patients in the moderate stage (Additional file 2: Fig. S20B, C). In contrast, the expression of genes involved in response to type I interferon in NK2.1 cells from young COVID-19 patients in the severe stage was significantly lower than that in NK2.1 cells from young COVID-19 patients in the moderate stage (Additional file 2: Fig. S20B, C). These contrasting trends suggest that the effect of type I interferon signalling on NK2.1 cells might differ for COVID-19 patients in an age-related manner. Together, these results indicate that NK2.1 cells of elderly COVID-19 patients showed enrichment of type I interferon responses which was positively correlated with the disease severity.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Disclaimer:

This article is autogenerated using RSS feeds and has not been created or edited by OA JF.

Click here for Source link (https://www.biomedcentral.com/)