Clonal expansion and epigenetic inheritance of long-lasting NK cell memory

Rückert, Timo; Lareau, Caleb A.; Mashreghi, Mir-Farzin; Ludwig, Leif S.; Romagnani, Chiara

doi:10.1038/s41590-022-01327-7

Download PDF

Article
Open access
Published: 26 October 2022

Clonal expansion and epigenetic inheritance of long-lasting NK cell memory

Nature Immunology volume 23, pages 1551–1563 (2022)Cite this article

23k Accesses
37 Citations
97 Altmetric
Metrics details

Subjects

Abstract

Clonal expansion of cells with somatically diversified receptors and their long-term maintenance as memory cells is a hallmark of adaptive immunity. Here, we studied pathogen-specific adaptation within the innate immune system, tracking natural killer (NK) cell memory to human cytomegalovirus (HCMV) infection. Leveraging single-cell multiomic maps of ex vivo NK cells and somatic mitochondrial DNA mutations as endogenous barcodes, we reveal substantial clonal expansion of adaptive NK cells in HCMV⁺ individuals. NK cell clonotypes were characterized by a convergent inflammatory memory signature enriched for AP1 motifs superimposed on a private set of clone-specific accessible chromatin regions. NK cell clones were stably maintained in specific epigenetic states over time, revealing that clonal inheritance of chromatin accessibility shapes the epigenetic memory repertoire. Together, we identify clonal expansion and persistence within the human innate immune system, suggesting that these mechanisms have evolved independent of antigen-receptor diversification.

Replicative history marks transcriptional and functional disparity in the CD8+ T cell memory pool

Article 07 April 2022

The transcription factor Fli1 restricts the formation of memory precursor NK cells during viral infection

Article 14 March 2022

Cardinal features of immune memory in innate lymphocytes

Article 12 October 2023

Main

Vertebrates have evolved several mechanisms of immune adaptation, enabling optimized secondary responses to the same challenge. In a traditional paradigm, persistent immune memory has exclusively been associated with T and B cell clones selected from a pool expressing enormously diversified antigen receptors. Competitive clonal expansion after antigen encounter is accompanied by pronounced epigenetic remodeling, forming a repertoire of memory cells with enhanced and accelerated recall responses¹. The dogma that adaptation mechanisms are exclusively confined to B and T cells has been challenged by observations of adjusted secondary responses within the innate immune system, with the most prominent example being ‘trained immunity’ of myeloid cells². As for adaptive immune cells, these functional differences are epigenetically imprinted but are generally considered to be more transient and shorter lived than classical adaptive immunity. However, long-term maintenance of trained immunity has been observed in different cell types^3,4,5,6. Further, in sharp contrast with the massive expansion of individual, epitope-specific T and B cell clones, trained immunity is mostly described at the population level, as innate immune cell activation occurs via widely expressed pattern recognition receptors.

More specific innate recognition mechanisms for several pathogens are used by natural killer (NK) cells, with the most prominent example being cytomegalovirus (CMV)⁷. CMV closely interacts with the host immune system to establish lifelong latent infection⁸. Consequently, CMV infection strongly imprints both the innate and adaptive immune repertoires, resulting in expansion of T cell clones and the emergence of ‘adaptive’ NK cells. While in C57BL/6 mice, these cells are expanded in response to the engagement of the activating receptor Ly49H by the murine CMV ligand m157 (ref. ⁹), human adaptive NK cells specifically recognize peptides derived from the human CMV (HCMV) protein gpUL40 through the receptor CD94/NKG2C, facilitating affinity-dependent expansion of NKG2C⁺ NK cells¹⁰. Recognition via conserved innate receptors expressed by cell populations is an important difference to unique antigen receptors, likely affecting the size of the pool recruited into the response. In this context, the narrow, self-major histocompatibility complex (MHC)-specific killer cell immunoglobulin-like receptor (KIR) profile of human adaptive NK cells has sparked discussions on an oligoclonal origin¹¹. Along this line, animal transfer studies of barcoded NK cells or progenitors have demonstrated the potential of NK cells to clonally expand in immunodeficient or depleted hosts^12,13,14, raising conceptual parallels to adaptive immune cells¹⁵. However, how these findings apply to natural infection in healthy individuals, where much larger populations of cells compete with each other and clonal diversity is not biased by transfer efficiency, is still unclear. Moreover, whether adaptive NK cell expansions induced by HCMV emerge from oligoclonal ‘founder’ populations and persist to form long-lasting memory remains elusive.

Here, we generated a single-cell multiomic map of human NK cells from HCMV⁺ and HCMV⁻ individuals, integrating measurements of gene expression and chromatin accessibility with lineage tracing using somatic mitochondrial DNA (mtDNA) mutations as endogenous barcodes^16,17. We detected substantial clonal expansion of adaptive NK cells with a convergent inflammatory memory signature accompanied by private, clonotype-associated accessible chromatin sites, demonstrating a human innate immune population clonally persisting in a memory state.

Results

Mapping NK cell subsets onto transcriptional and epigenetic landscapes

To resolve the relationship of human NK cell subsets and the impact of HCMV on these populations, we performed single-cell assay of transposase-accessible chromatin (ATAC) with select antigen profiling by sequencing (ASAP-seq)¹⁸ and cellular indexing of transcriptomes and epitopes (CITE-seq)¹⁹ of NK cells from peripheral blood of HCMV⁺ donors with NKG2C⁺ NK cell expansions and HCMV⁻ individuals (Fig. 1a and Supplementary Table 1). NKG2C⁺/NKG2C^– NK cells (sorted as CD3⁻CD14⁻CD19⁻CD7⁺; Methods) from all donors were equally enriched to a 1:1 ratio to correct for different frequencies of NKG2C⁺ cells between HCMV⁺ and HCMV⁻ individuals (Extended Data Fig. 1a). After filtering for high-quality cells (Methods), 39,106 cells and 49,530 cells were analyzed for their epigenetic and transcriptional profiles, respectively. We detected expression of 19,975 genes and accessibility of 147,299 chromatin regions with consistent enrichment of transcription start sites (TSSs) across donors (Extended Data Fig. 1b). Using cell surface protein information, we annotated clusters defined by their transcriptomes and chromatin accessibility with long-standing NK cell subset definitions, while at the same time testing the validity of these population definitions in an unbiased manner (Fig. 1b–e). Namely, we observed four main clusters in both modalities that could be designated as CD56^bright, early CD56^dim, mature CD56^dim and adaptive NK cells (Fig. 1b,d) with clearly distinct genomic signatures (Extended Data Fig. 1d–f), in line with previous data^{20,21,22,23,24}. An additional cluster of proliferating cells was only resolved via transcriptional profiling (Fig. 1d and Extended Data Fig. 1c). CD56^bright (CD16^–/loCD57⁻CD62L⁺NKG2A⁺) cells were characterized by high accessibility around loci of characteristic transcription factors (TFs) TCF7, RUNX2, BACH2, ZEB1 and MAML3, and expression of signature transcripts, including GZMK, XCL1 and IL7R. By contrast, CD56^dim (CD16⁺CD57⁺/CD57^–CD62L⁺/CD62L^–NKG2A⁺/NKG2A^–) NK cells were characterized by chromatin accessibility in the proximity of ZBTB16, BCL11B, GLI3 and ZEB2 and expression of GZMB, PRF1 and CX3CR1. Between CD56^bright and CD56^dim NK cells, a population of seemingly intermediate, early CD56^dim (CD16⁺CD57^–CD62L⁺NKG2A⁺) NK cells shared signatures of both subsets, such as high gene scores for TCF7, ZEB1 and ZBTB16 (Fig. 1b–g and Extended Data Fig. 1d–f). Label transfer between modalities further allowed for directly linking gene expression and chromatin accessibility, generating an integrated resource that enables correlation-based prediction of subset-specific gene regulatory elements for core regulators of NK cell identity, such as TCF7, RUNX2 and ZEB2 (Extended Data Fig. 2a–d). Overall, we provide a multimodal map of human NK cells, combining single-cell measurements of chromatin accessibility, gene expression and surface phenotypes.

Distinct signatures for naive and adaptive NKG2C⁺ NK cells

Our multiomic analysis of human NK cells clearly revealed the presence of a separate and genomically distinct adaptive NK cell cluster (Fig. 1b–g), characterized by high transcript levels of KLRC2, CD3E and PATL2 together with marked downregulation of CD7, KLRB1 and FCER1G (Fig. 1g) and distinct surface marker expression of CD57 and CD2 coinciding with low NKp30, CD161 and NKG2A (Fig. 1c,e). While adaptive NK cells shared large parts of their genomic effector signature with conventional CD56^dim NK cells, they displayed a marked reduction of ZBTB16 and ZNF516 and further increased gene scores for ZEB2 and the metabolic regulator ARID5B²⁵ compared to CD56^dim NK cells (Fig. 1f). Moreover, adaptive NK cells were characterized by expression and accessibility of ZBTB38 and JAKMIP1 (Fig. 1f,g and Extended Data Fig. 1e,f). Importantly, the integrated analysis demonstrated remodeling at chromatin regions around the KLRCX locus, with several of those being positively correlated with increased KLRC2 expression and one region having a negative correlation with KLRC1 expression (Extended Data Fig. 2e). Their functional role was supported by the presence of several TF binding motifs, with the most strongly correlating putative KLRC1 enhancer containing a TCF7 motif, consistent with expression and accessibility in CD56^bright NK cells, while AP1 motifs were notable in the putative KLRC2 enhancers specifically accessible in adaptive NK cells, which together may regulate the core adaptive NK cell phenotype characterized by high surface expression of NKG2C combined with lack of NKG2A²⁶.

While expansion of adaptive NKG2C⁺ NK cells is associated with HCMV infection, NKG2C⁺ NK cells are present also in HCMV⁻ individuals, albeit at lower frequencies. To directly compare the transcriptional and epigenetic landscape of NKG2C⁺/NKG2C^– NK cells and their distribution within NK cell clusters in HCMV⁺ and HCMV⁻ individuals, we divided the datasets by HCMV serostatus (Fig. 2a–h). As expected, the adaptive cluster was selectively present in HCMV⁺ individuals and was mostly comprised of NKG2C⁺ NK cells (Fig. 2a–d). Comparative analysis of chromatin accessibility and gene expression after downsampling to equal cell numbers for both donor groups highlighted the specific impact of HCMV on NKG2C⁺ cells. A total of 232 genes were differentially accessible between the NKG2C⁺ and NKG2C^– populations in HCMV⁺ individuals, including key adaptive genes, such as JAKMIP1, ZBTB38, ZBTB16 or ZNF516 (Fig. 2i), along with 269 differentially expressed genes, including adaptive NK cell markers CD3E, IL32, FCER1G and KLRB1 (Fig. 2k). Conversely, NKG2C⁺ NK cells from HCMV⁻ donors were randomly dispersed among NKG2C⁻ NK cells (Fig. 2e–h), and differences in chromatin accessibility and gene expression, apart from KLRC2 (encoding NKG2C), were minor (Fig. 2m,o), highlighting their largely similar epigenomes and transcriptomes. Genes differentially accessible or expressed between NKG2C⁺ and NKG2C^– cells in HCMV⁺ individuals, including ZBTB16 and FCER1G, followed a coordinated pattern separating conventional from adaptive NK cells, whereas this was not observed in HCMV⁻ individuals (Fig. 2j,n,l,p), underlining that these differences reflect the adaptive remodeling induced in NKG2C⁺ cells by HCMV.

**Fig. 2: Distinct signatures of naive and adaptive NKG2C⁺ NK cells.**

Interestingly, even HCMV⁺ individuals displayed a small fraction of NKG2C⁺ NK cells that were present in all three subsets of the conventional NK cell compartment (Fig. 2a–d). Indeed, separate analysis of NKG2C⁺ NK cells alone from HCMV⁺ individuals revealed the coexistence of adaptive NKG2C⁺ NK cells with a minority of ‘naive’ NKG2C⁺ NK cells (Extended Data Fig. 3a,d) that lacked adaptive remodeling (Extended Data Fig. 3b,e) and spanned the whole spectrum of conventional NK cells, recapitulating the signatures identified from the full integrated analysis (Extended Data Fig. 3c,f). We confirmed by flow cytometry that the frequencies of naive NKG2C⁺ NK cells were comparable between HCMV⁺ and HCMV^– individuals and that they were similarly distributed between the CD56^bright and CD56^dim compartment (Extended Data Fig. 3g–i).

Together, this comparative analysis of HCMV⁺ and HCMV⁻ donors highlighted the pronounced remodeling that HCMV imposes on the NKG2C⁺ NK cell pool that coexists with a minority of naive NKG2C⁺ NK cells during the latent phase of HCMV infection.

HCMV leaves an inflammatory memory footprint enriched in AP1 motifs

While gene expression profiles and accessibility of regulatory elements in the proximity of trans-acting factors provide a view on the current state of a cell, analysis of TF motif enrichment in cis-regulatory elements also captures molecular events that have left footprints in the generation of this state. Global analysis using chromVAR²⁷ identified a strong signature in CD56^bright NK cells with enhanced activity of motifs for TFs such as TCF7, RUNX2 or ZEB1 (Fig. 3a and Extended Data Fig. 4a,b,e), consistent with an important role of these TFs in shaping the epigenetic and transcriptional identity^21,24 of this subset (Fig. 1f,g) and increased motif activity for NF-κB-related TFs (REL) and STAT4 (Fig. 3a and Extended Data Fig. 4a,b,e), reflecting the enhanced responsiveness of CD56^bright NK cells to activation by proinflammatory cytokines²⁸. Accordingly, an NF-κB- and STAT4-motif-containing peak in the proximity of the IFNG gene was specifically accessible in CD56^bright cells, while this was closed in the more mature subsets (Extended Data Fig. 4f). While CD56^dim NK cells were mostly characterized by a lack of the CD56^bright-specific TF motif activity, both CD56^dim and adaptive NK cells were further distinguished from CD56^bright NK cells by enhanced CTCF motif activity (Extended Data Fig. 4a,e), as previously described²⁴. Importantly, adaptive NK cells showed a strikingly separated signature compared to both conventional NK cell subsets, characterized by enhanced activity of AP1 motifs, with the strongest difference observed for FOS–JUNB (Fig. 3a,b and Extended Data Fig. 4d), and AP1 motifs were the most significantly enriched in adaptive NK-cell-specific accessible regions (Fig. 3c,d). De novo motif prediction on these peaks also yielded a canonical AP1 motif as the most significant hit, further corroborating these results (Fig. 3e). Recently, AP1 motifs have been described to be at the core of a unifying inflammatory memory signature in different immune cell types and murine epithelial stem cells, the latter characterized by specific accessibility in the proximity of the Aim2 and Cadm1 loci⁴. Notably, their human orthologs also showed highly specific opening of AP1-motif-containing regions in adaptive NK cells (Fig. 3c and Extended Data Fig. 4g) and were among the genes with the most pronounced difference in accessibility between NKG2C⁺ and NKG2C^– cells in HCMV⁺ individuals (Fig. 2i). Finally, among the more accessible peaks with AP1 motifs were several enhancers of the IFNG gene (Extended Data Fig. 4f), providing a possible mechanism for the increased interferon-γ (IFNγ) production by adaptive NK cells in response to activating receptor stimulation^29,30. Overall, these findings demonstrate how differential TF activity shapes NK cell epigenetic remodeling during differentiation and identify AP1 TFs as potential drivers shaping adaptive NK cell chromatin accessibility, an intriguing parallel to the inflammatory imprinting used to establish immune memory in different cell types^4,31,32.

**Fig. 3: HCMV infection leaves an inflammatory memory footprint enriched in AP1 motifs.**

Synergistic imprinting by HCMV peptides and proinflammatory cytokines

Activation by interleukin-12 (IL-12) and IL-18 in combination with engagement of NKG2C by HCMV-derived peptides induces the specific expansion of NKG2C⁺ NK cells with some of the transcriptional characteristics also observed ex vivo¹⁰. To assess whether these signals drive epigenetic reprogramming, including AP1 TF activity, and mimic HCMV-induced activation of naive NKG2C⁺ NK cells, we cocultured NK cells from HCMV^– donors for 12 h with IL-15 and RMA-S/HLA-E target cells pulsed with the NKG2C-activating VMAPRTLFL (LFL) or non-activating VMAPQSLLL peptide in the presence or absence of IL-12 and IL-18. We marked the different conditions with individual nucleotide-barcoded hashtags, sorted for NKG2C⁺ cells and performed ASAP-seq (Fig. 4a).

Analysis of high-quality cells clustering by condition and irrespective of donors (Methods) enabled us to identify the epigenetic signatures acquired by activated NKG2C⁺ NK cells in a stimulus-dependent fashion, defining four main clusters clearly enriched for cells cultured under the respective conditions, that is, non-activated (‘control’), activated by LFL peptide (‘LFL’), proinflammatory cytokines (‘IL-12 + IL-18’) or a combination thereof (‘LFL + IL-12 + IL-18’; Fig. 4b, Extended Data Fig. 5a,b and Methods). The resulting chromatin accessibility profiles had a similar TSS enrichment (Extended Data Fig. 5c) as observed ex vivo (Extended Data Fig. 1b), irrespective of activation. Activation status of cells recruited in stimuli-associated clusters was exemplified by surface upregulation of CD137 (4-1BB), with maximal expression observed on cells receiving both peptide and cytokine stimulation (Fig. 4c). Differences in expression levels of CD137 across conditions were confirmed by flow cytometry (Extended Data Fig. 5d,e), validating the proteogenomic-based cluster annotations.

While IL-12 + IL-18 stimulation alone promoted the activation of NKG2A⁺NKG2C⁺ naive NK cells, hardly any NKG2A⁺NKG2C⁺ cells were recruited into response patterns induced by peptide stimulation (Fig. 4d), in line with the dominant inhibitory function of NKG2A engagement³³. As, in contrast to adaptive NKG2C⁺ NK cells²⁶, a large fraction of naive NKG2C⁺ NK cells coexpress NKG2A (Extended Data Fig. 5g,h), these data suggest that peptide recognition might be decisive in biasing the pool of naive NKG2C⁺ cells recruited during HCMV infection toward NKG2A⁻NKG2C⁺ cells.

Activation via proinflammatory cytokines or by peptide recognition via NKG2C induced pronounced chromatin remodeling, manifesting in 1,232 and 936 differentially accessible regions (DARs), respectively, compared to the control condition (Fig. 4e). Most regions underwent remodeling only after synergistic activation by the two combined stimuli (2,693 DARs; Fig. 4e), with different TF families contributing to these effects (Fig. 4f). Besides IRFs, EGR, MYC and NF-κB, LFL peptide activation induced remarkable activity of AP1-family TFs. Consistent with reports in mice³⁴, proinflammatory cytokines IL-12 and IL-18 were highly specific in their induction of POU TF-family activity, along with NF-κB. However, clear synergism between cytokines and peptide became apparent also on the TF motif activity level, in particular for AP1 (Fig. 4f and Extended Data Fig. 5f). Importantly, the two stimuli induced several features associated with adaptive NK cells ex vivo. This included opening of chromatin in the ZEB2 and ARID5B loci, which was mainly dependent on proinflammatory cytokines or similarly enabled by each individual stimulus, respectively, as well as closing of the ZBTB16 locus (Fig. 4g–i and Extended Data Fig. 5i). However, most of these changes were more pronounced after synergistic activation, as was also the case for chromatin accessibility in the AIM2 region (Fig. 4g–i and Extended Data Fig. 5i). To globally assess which of the activation-induced signatures best reconciles the adaptive signature, we performed anchor-based integration of the in vitro stimulated cells with our ex vivo dataset (Methods)³⁵, thereby enabling a classification of the in vitro induced signatures based on chromatin profiles observed ex vivo. The highest adaptive prediction score was observed for cells activated by both stimuli, in concert with a reduction of the prediction score for conventional CD56^dim NK cells (Fig. 4j,k). The close relatedness of the adaptive ex vivo signature to the chromatin remodeling induced by synergistic in vitro activation was further supported by a strong correlation of overlapping DARs when comparing CD56^dim to adaptive NK cells and when comparing control to LFL + IL-12 + IL-18 activated cells (Extended Data Fig. 5j). Among these conserved DARs, we identified core features of adaptive NK cells, such as reduced accessibility at the ZBTB16 locus and increased accessibility of AIM2 or IFNG loci (Extended Data Fig. 5j). Importantly, these modulated chromatin regions showed a strong enrichment of AP1 motifs, whereas NF-κB, NFAT and STAT4 motifs were much less enriched (Extended Data Fig. 5k). Together, these findings directly connect proinflammatory cytokines and NKG2C-activating peptide, both provided during HCMV infection, to the characteristic phenotype and epigenetic remodeling of adaptive NK cells as observed ex vivo. Their synergism manifested on several layers from surface upregulation of CD137, over TF motif activity, to global chromatin remodeling, thereby recapitulating features of adaptive NK cell signatures including open chromatin enriched for AP1 motifs.

Convergent and divergent epigenetic features of adaptive NK cells

The integrated analysis of different HCMV⁺ donors revealed consistent epigenetic, transcriptional and phenotypic features of the adaptive NK cell signature ex vivo. However, integration of donors masked donor-specific heterogeneity within the adaptive NK cell compartment. Indeed, separate analysis of the epigenetic landscape of individual donors revealed clearly defined adaptive subclusters (Fig. 5a and Extended Data Fig. 6a). These populations were characterized by an adaptive phenotype, largely lacking surface expression of NKp30 and NGK2A, and were mostly positive for NKG2C and self-MHC-specific KIRs (Extended Data Fig. 6b and Supplementary Table 2). Interestingly, subcluster 4 in donor P1 likely represents an NKG2C^– adaptive NK cell expansion (Extended Data Fig. 6b), as previously described by others^11,36,37. To assess the degree of convergence of the adaptive NK cell signature, we analyzed the overlap of DARs between the total adaptive and the CD56^dim compartment across HCMV⁺ donors (P1, P2, P3 and P4), revealing that around half of all DARs were shared between at least three donors, similar to the signatures distinguishing CD56^bright from CD56^dim NK cells (Fig. 5b,c). Consistent with integration into a donor-overarching adaptive cluster (Fig. 1b), this adaptive signature was shared by all adaptive subclusters, separating them from the conventional CD56^dim populations (Extended Data Fig. 6c). Many of the features identified in the integrated analysis were highly penetrant across donors and subclusters, such as the peak in the proximity of AIM2 (Extended Data Fig. 6d) and the strongly enhanced AP1 motif activity (Extended Data Fig. 6e), further supporting the idea that this convergent epigenetic signature reflects a coordinated program induced in response to signals received during HCMV infection. In contrast to these shared aspects, DARs between adaptive subclusters were much less conserved (Fig. 5d). Each adaptive subcluster showcased highly unique open chromatin regions (Fig. 5e,f), underlining their distinctive epigenetic makeup. Nevertheless, hierarchical clustering divided the adaptive subclusters into two groups, mainly driven by overlapping signatures of three subclusters (Fig. 5e). Importantly, when directly comparing the gene scores of the two groups, two clear signatures emerged, which had an intriguing overlap with those defining CD56^bright/early CD56^dim NK cells, such as RUNX2, BACH2, TCF7, IL7R and SELL (encoding CD62L), on the one end and CD56^dim-associated genes, such as GLI3 or ZBTB16, on the other end (Fig. 5g). Indeed, differential accessibility of the respective chromatin regions between the adaptive subcluster groups correlated significantly with the differences between CD56^bright/early CD56^dim and CD56^dim NK cells (Fig. 5h and Extended Data Fig. 7a) and followed consistent patterns as exemplified for chromatin regions in the proximity of TCF7 or PCNT (Extended Data Fig. 7b), suggesting that conventional NK cell maturation signatures are preserved after adaptive differentiation. This was also supported by differential activity of CD56^bright/early CD56^dim-associated motifs, such as TCF7L2, on the one hand and the CD56^dim-associated CTCF motif on the other hand (Extended Data Fig. 7c). Strikingly, the group resembling CD56^bright/early CD56^dim NK cells was also marked by surface expression of CD62L (Extended Data Fig. 7e), consistent with its expression by CD56^bright and early CD56^dim NK cells within the conventional compartment³⁸ (Fig. 1c,e). Additionally, the clusters that shared part of the CD56^bright/early CD56^dim signature had relatively higher AP1 motif activity (Extended Data Fig. 7d), suggesting that the two groups might also reflect a different extent of adaptive chromatin remodeling. Overall, this donor-centric analysis revealed an unappreciated heterogeneity within the adaptive NK cell compartment defined by a highly diverse set of unique, subcluster-specific peaks and signatures resembling gradients in conventional NK cell maturation, possibly reflecting the cell of origin.

Clonal expansion underlies divergent signatures of adaptive NK cells

In contrast to the convergent remodeling at key adaptive genes and a pronounced enrichment of AP1 motifs, the large number of unique peaks detected only in individual adaptive subclusters represents a diversification that is difficult to reconcile with a coordinated differentiation program. We hypothesized that this apparent diversification might result from the expansion of individual naive NKG2C⁺ cells with unique accessible chromatin regions. Expansion of rare founder cells would amplify these regions beyond the detection threshold while at the same time reduce the epigenetic heterogeneity of their progeny. Hence, as a measure for cluster heterogeneity, we analyzed the average distance between the k nearest neighbors within the latent semantic indexing (LSI) space representing their chromatin accessibility landscape (Methods). Notably, all adaptive subclusters were less heterogeneous than their corresponding conventional CD56^dim compartment, as demonstrated by a uniformly lower average distance (Fig. 6a and Extended Data Fig. 8a). This heterogeneity measure was robust over a wide range of values for k (Extended Data Fig. 8b), indicating a focused epigenetic profile of individual adaptive NK cell clusters, potentially resulting from a selection bottleneck during recruitment, followed by clonal expansion. To test the hypothesis that divergent epigenetic profiles of adaptive subclusters are associated with clonal expansions, we applied a recently published method that exploits mtDNA mutations as endogenous barcodes to reconstruct clonal relationships and concomitantly links them to epigenetic and cell surface phenotypes^16,17,18. Applying this method to four HCMV⁺ and three HCMV^– donors, we identified informative mtDNA mutations by high strand concordance and variance-to-mean ratio (Extended Data Fig. 8c). In the HCMV⁺ donors, many of these mutations were specifically enriched in the adaptive NK cell compartment (Fig. 6b and Extended Data Fig. 8d). Importantly, clonotypes defined by individual mutations (Extended Data Fig. 9a and Methods) were significantly associated not only with adaptive NK cells as a whole but also with specific subclusters, as demonstrated by the χ² statistic of the observed compared to randomly permuted clonotype–cluster relationship (Fig. 6c,d). This striking concordance between mitochondrial mutations and epigenetic identities reveals the inheritance of epigenetic states as a defining clonal mark. Conversely, such an association was absent in the conventional NK cell compartment (Fig. 6e and Extended Data Fig. 9b), suggesting that mitochondrial mutations of conventional NK cells do not flag clonal expansions but might instead be present already on the progenitor level. Similarly, there was no significant association between clonotypes and epigenetic identities in HCMV^– donors (Extended Data Fig. 9c), underlining the specific clonal expansion of NKG2C⁺ NK cells in response to HCMV. Donor-specific patterns further highlighted the degree of clonal expansion of the adaptive subclusters. For example, a large fraction of cells in subcluster 3 of donor P2 carried the 5590G>A mutation at near homoplasmic frequencies (Fig. 6b), coinciding with expression of a second self-MHC-specific KIR3DL1, which was absent from the other adaptive subclusters in this individual (Extended Data Fig. 6b).

**Fig. 6: Clonal expansion underlies divergent epigenetic signatures of adaptive NK cells.**

Finally, association of clonotypes to individual subclusters defined by chromatin accessibility highlighted the epigenetic similarity of cells belonging to one clonotype. This was further supported by the significant association between individual open chromatin regions and clonotypes (Fig. 6f). Importantly, we found regions that were uniquely accessible in individual donors and clonotypes (Extended Data Fig. 9d), while a part of this diversity was again driven by the opposing accessibility in the proximity of genes associated with conventional NK cell maturation, such as IL7R and GLI3 (Extended Data Fig. 9e), suggesting that these signatures are indeed clonally inherited. Together, we demonstrate that subclusters of adaptive NK cells contain individual clonotypes with private chromatin accessibility profiles, strongly supporting clonal expansion as a driving force of adaptive NK cell generation and epigenetic diversification.

Adaptive NK cell clonotypes are stably maintained over time

The specific association of non-overlapping clonotypes to adaptive subclusters suggests a degree of clonal and epigenetic stability. To test this hypothesis, we performed a longitudinal follow-up analysis of three HCMV⁺ donors after 11, 7 and 19 months, respectively. Clusters defined by chromatin accessibility were unchanged between the two time points, supporting stable maintenance of the frequency and composition of the adaptive NK cell pool down to the subcluster level (Fig. 7a and Extended Data Fig. 10a,b). Further, the subcluster-specific open chromatin regions remained stable in this time frame (Fig. 7b,c and Extended Data Fig. 10c–f). Importantly, the same clonotype-defining mutations were detected at both time points and remained associated with the original subclusters (Fig. 7d and Extended Data Fig. 10g,h), highlighting their persistence and demonstrating that adaptive subclusters represent independent and stably maintained clonal expansions. Along these lines, even frequencies of clonotypes significantly associated with the adaptive compartment were stably maintained over time (Fig. 7e and Extended Data Fig. 10i,k), which was further supported by an unskewed total clonotype distribution across time points (Fig. 7f and Extended Data Fig. 10j,l). Overall, the longitudinal follow-up demonstrates the long-term clonal maintenance of adaptive NK cells with stable chromatin accessibility profiles in progeny arising from individual clonal founders, which are key features of ‘epigenetic memory’, as previously ascribed to the adaptive immune system.

Discussion

Previous studies have revealed conserved signatures of adaptive NK cells driven by coordinated changes in their gene regulatory networks, such as downregulation of the TF PLZF³⁶ or increased activity of the T cell lineage TF BCL11B²⁴. Conversely, insights into their heterogeneity have remained limited to the diversification of signaling adaptor expression between and within individuals³⁶. Here, we provide an integrated, multiomic single-cell analysis of human NK cells, enabling us to study both convergent and divergent aspects of adaptive NK cell biology.

Our finding of a strong enrichment of AP1 motifs in adaptive NK cell chromatin extends observations made in different cell types^{4,31,39,40,41,42}, further supporting the concept of a conserved inflammatory memory signature imprinted by AP1 TFs^4,32. Here, we demonstrate that HCMV infection can induce this characteristic inflammatory memory in human NK cells, revealing a surprising conservation of this signature across stimuli and species. Consistent with reports in mice^43,44 and our own studies in humans¹⁰ we demonstrate that synergistic activation of naive NGK2C⁺ NK cells via CD94/NKG2C and proinflammatory cytokines most closely recapitulated the adaptive NK cell signature observed ex vivo and induced the most pronounced activity of AP1 TFs.

Conversely, our donor-specific analysis revealed two different layers of adaptive NK cell heterogeneity. First, we identified two underlying signatures conserved across adaptive NK cells from HCMV⁺ donors, which resembled maturation differences within the conventional NK cell compartment. Second, we unveiled unique, donor-specific, that is, private, open chromatin regions characterizing individual subclusters. Based on the high degree of clonality of adaptive NK cell expansions, we propose that both of these levels of heterogeneity can be explained by clonal inheritance of epigenetic traits, enabling an epigenetic founder effect. Acting as a clonal bottleneck, the recruitment of individual NK cells into the adaptive response and their massive expansion might skew the distribution of open chromatin regions. Hence, the two layers of adaptive NK cell heterogeneity likely reflect the degree of diversity and maturation stage within the original naive NKG2C⁺ pool before acquisition of the superimposed adaptive differentiation program. Together with our observation that naive NKG2C⁺ NK cells span the whole spectrum of conventional NK cell maturation, these data suggest that naive NKG2C⁺ NK cells might be recruited into the adaptive response at these different stages and preserve a part of their initial epigenetic identity. This is consistent with findings in mice, where pre-existing NK cell subtypes were preserved after murine CMV-induced expansion⁴⁵. Alternatively, these might represent different stages of adaptive NK cells, although the association of clonotypes with individual subclusters and their stability in time argue against this interpretation. Regarding the unique open chromatin regions associated with adaptive NK cell subclusters, we envisage that they become evident due to the drastic expansion of individual NK cell clones, which at the same time underlies the reduced heterogeneity of adaptive subpopulations. Consequently, we propose that the macroscopic picture of epigenetic diversification, manifesting, for example, in variable expression of signaling adaptors³⁶, is created by the coexistence of several clonal expansions that are, in fact, epigenetically focused.

Importantly, our findings enable us to generalize some of the mechanisms driving clonal expansion and their consequences within the immune system as a whole. One such unifying prerequisite for clonal expansion is the competitive recruitment of a limited number of cells to emerge as clones from a diverse population. Despite their reliance on a conserved receptor for HCMV recognition, other cell-intrinsic features of NK cells are strongly diversified, especially their combined cell surface receptor expression profiles⁴⁶, leading to selection of optimal receptor combinations. The low frequency of cells expressing NKG2C before HCMV infection is the first limitation of the original pool activated by HCMV peptides. Further, in response to peptide and cytokine stimulation, only NKG2A^– NK cells underwent remodeling that resembled adaptive NK cells the closest, suggesting that NKG2A expression further limits the pool of naive NKG2C⁺ cells optimally responding to HCMV. Expression levels of NKG2C might also contribute to selection, as adaptive NKG2C⁺ NK cells express higher levels of NKG2C than their naive counterparts⁴⁷, and a similar mechanism occurs for Ly49H⁺ NK cells in mice^12,48. Clonal selection can further be influenced by complex biological traits that are heterogeneously distributed among cells before or after activation, such as proliferation and differentiation kinetics^49,50 or metabolic capacity⁵¹. More studies will be needed to distinguish the adaptive NK cell features modulated by external signals from those conferring a selection advantage within the naive repertoire and the ones that are stochastic passengers of the clonal expansion process. The potential of non-genetic inheritance mechanisms to impact cell fitness has recently been demonstrated in the context of tumor cell resistance against anticancer therapeutics⁵². As T cells can also clonally inherit transcriptional and epigenetic diversity⁵³, these analyses might provide general insights into the drivers of clonal success in different cell types independent of receptor specificity.

Finally, prolonged survival of adaptive NK cells has been postulated based on the observation of stable expression of KIR and signaling adaptor patterns³⁶, and evidence from individuals with mutations in GATA2 or PIGA supports this concept^54,55. Here, we demonstrated long-term persistence on the clonal level for at least 19 months, although the exact cellular mechanisms enabling clonal maintenance remain to be defined.

Here, we demonstrate the clonal expansion and long-term persistence of an innate immune population in a memory state in response to a naturally occurring infection and introduce clonal selection and epigenetic inheritance as mechanisms that shape human innate immune composition for optimized secondary responses. Understanding the mechanisms determining clonal fitness in adaptive and innate immune cells independent of diversified antigen receptors might open new avenues for fundamental research and therapeutic interventions.

Methods

Human studies

All analyses were performed in compliance with the relevant ethical regulations, and all donors gave informed consent. Primary NK cells were isolated from freshly drawn peripheral blood of healthy donors (Supplementary Table 1) or from buffy coats obtained from Deutsches Rotes Kreuz (DRK) Blutspendedienst Nord-Ost. The Charité Ethics Committee approved the study (EA4/196/18 and EA4/059/17).

Cell lines

RMA-S/HLA-E cells⁵⁶ were maintained in complete medium (RPMI-1640 containing glutamine and supplemented with 10 % (vol/vol) fetal bovine serum (FBS), 20 μM β-mercaptoethanol and 100 U ml^–1 penicillin–streptomycin; all Thermo Fisher) in the presence of 400 μg ml^–1 hygromycin B (InvivoGen).

Cell isolation

Peripheral blood mononuclear cells were isolated by density gradient centrifugation (Ficoll Paque Plus, GE Healthcare) and either processed immediately or cryopreserved in FBS containing 10% DMSO.

HCMV serology

For buffy coats, CMV serology was performed at DRK Dresden. Serological status of fresh blood donors was analyzed by HCMV IgG enzyme-linked immunosorbent assay (IBL International) following the manufacturer’s instructions. All samples were measured in duplicate and were either clearly above or below the cutoff control.

Flow cytometry

Cell suspensions were stained with combinations of the following fluorochrome-conjugated antibodies following established guidelines⁵⁷: CD56 PE/Dazzle 594 (1:200), CD137 PE/cyanine7 (1:50) and CD57 BV605 (1:25; all BioLegend); CD7 BV786 (1:25), CD56 BUV737 (1:50), CD3 BUV805 (1:50), streptavidin BUV395 (1:100), CD16 BUV496 (1:50) and CD337 (NKp30) BV421 (1:25; all BD Biosciences); CD159a (NKG2A) biotin/PE-Vio770 (both 1:50) and CD159c (NKG2C) PE (1:100; all REAfinity, Miltenyi Biotec); CD337 (NKp30) eFluor 450 (1:25), CD3 APC-eFluor 780 (1:50), CD14 APC-eFluor 780 (1:50) and CD19 APC-eFluor 780 (1:50; all Thermo Fisher) and anti-FcεRI, γ subunit-FITC (1:50; Merck). Dead cells were excluded using Fixable Viability Dye eFluor 780 (Thermo Fisher) or a Zombie Aqua Fixable Viability kit (BioLegend). Data were acquired on an LSR Fortessa (BD Biosciences). Cell sorting was performed on a FACSAria II (BD Biosciences). FlowJo v10 was used for analysis of flow cytometry data, and GraphPad Prism v8.4.3 was used for statistical analysis.

KIR ligand genotyping

DNA was isolated from 100 µl of peripheral blood using spin-column-based purification implemented with the DNeasy Blood and Tissue kit (Qiagen), and concentrations were measured using a Nanodrop 2000c spectrophotometer (Thermo Fisher). KIR ligand genotyping was performed using the Olerup SSP KIR HLA ligand typing kit (CareDx) following the manufacturer’s instructions. Reactions were analyzed on a 2 % (wt/vol) agarose gel prestained with GelRed (Biotium) in 0.5× TBE buffer. Gels were documented on a UV transilluminator and interpreted using the manufacturer-supplied reference tables. All lanes displayed the required control bands.

Cell preparation for scATAC-seq

NK cells were enriched from peripheral blood mononuclear cells isolated from freshly drawn peripheral blood by magnetic depletion using microbeads against CD3, CD14 and CD19. The enriched fraction was stained with the following fluorochrome-conjugated antibodies: CD7 BV786 (1:25; BD Biosciences), CD3 APC-eFluor 780 (1:50), CD14 APC-eFluor 780 (1:50) and CD19 APC-eFluor 780 (1:50; all Thermo Fisher), CD159a (NKG2A) biotin (1:50) and CD159c (NKG2C) PE (1:100; both REAfinity, Miltenyi Biotec). Fc receptors were blocked using Human TruStain FcX (1:50; BioLegend) for 15 min at 4 °C. Dead cells were excluded using Fixable Viability Dye eFluor 780 (Thermo Fisher). In a second step, cells were stained with combinations of the following nucleotide barcode-labeled antibodies (BioLegend) for 30 min at 4 °C: TotalSeq-A0084 CD56 (1:200), A0083 CD16 (1:500), A0436 anti-biotin (1:100), A0147 CD62L (1:100), A0168 CD57 (1:100), A0367 CD2 (1:1,000), A0801 CD337 (NKp30; 1:100), A0149 CD161 (1:100), A0420 CD158 (1:100), A0592 CD158b (1:100), A0599 CD158e1 (1:100), A0390 CD127 (1:400), A0902 CD328 (Siglec-7; 1:1,000), A0867 CD94 (1:100), A0896 CD85j (ILT2; 1:100), A0911 anti-phycoerythrin (1:100), A0250 KLRG1 (1:100), A0061 CD117 (1:100), A1018 HLA-DR, HLA-DP and HLA-DQ (1:400), A0355 CD137 (1:100), A0152 CD223 (1:100), A0366 CD184 (1:100), A0251 hashtag 1, A0252 hashtag 2, A0253 hashtag 3, A0254 hashtag 4, A0255 hashtag 5, A0256 hashtag 6, A0257 hashtag 7, A0258 hashtag 8, A0259 hashtag 9, A0260 hashtag 10, A0262 hashtag 12, A0263 hashtag 13, A0264 hashtag 14 and A0265 hashtag 15 (all 1:200–1:400). As described in the ASAP-seq and mitochondrial scATAC-seq (mtscATAC-seq) protocols, cells were fixed with 1% paraformaldehyde for 10 min. Fixation was quenched by adding glycin to a final concentration of 0.125 M, and cells were washed twice with PBS/bovine serum albumin (BSA). NK cells were sorted on a FACSAria II (BD Biosciences) as viable, single cells being CD3⁻CD14⁻CD19⁻CD7⁺, separating NKG2C⁺ and NKG2C⁻ cells. We used CD7 to sort for NK cells, as it is widely expressed on all NK cell subsets⁵⁸ and enabled us to stain markers informative for NK cell differentiation, such as CD56 and CD16, with nucleotide barcode-labeled antibodies. After sorting, NKG2C⁺/NKG2C^– NK cells were pooled and lysed for 3 min on ice using modified lysis buffer (10 mM Tris-HCl (pH 7.5), 10 mM NaCl, 3 mM MgCl₂, 0.1% NP40 and 1% BSA). Afterward, lysed cells were washed with washing buffer (10 mM Tris-HCl (pH 7.5), 10 mM NaCl, 3 mM MgCl and 1% BSA) and resuspended in diluted nuclei buffer (10x Genomics). Lysis was confirmed by trypan blue staining, and cells were counted and further processed as described below.

scATAC-seq

scATAC-seq libraries were prepared using the Chromium Next GEM single-cell ATAC reagent kit v1.1 following the manufacturer’s instructions with the modifications described in the ASAP-seq and mtscATAC-seq protocols to retrieve TotalSeq antibody-derived tags (ADTs), hashtag oligonucleotides (HTOs) and mtDNA^16,17,18. The saved eluate from the silane bead elution and the supernatant of the first SPRIselect purification were combined for amplification of ADT/HTO libraries using the KAPA HiFi ready mix (Roche) with sample-specific index primers (Illumina small RNA RPIx/Truseq D7xx) and purified using SPRIselect reagent. Library size and quality were analyzed using a Fragment Analyzer (Advanced Analytical) before fragmentation and after final purification. Final library concentrations were measured on a Qubit 2.0 fluorometer (Thermo Fisher). Libraries were sequenced on a NextSeq500 or NovaSeq6000 sequencer (Illumina) using longer read1/2 configurations than suggested by 10x Genomics to improve mitochondrial genotyping (NextSeq500: R1 72 cycles, R2 72 cycles, I1 8 cycles, I2 16 cycles; NovaSeq6000: R1 88 cycles, R2 88 cycles, I1 8 cycles, I2 16 cycles).

Cell preparation for scRNA-seq

NK cells were enriched, stained and sorted as described above for scATAC-seq (without fixation). NKG2C⁺ and NKG2C^– cells were separately labeled with different nucleotide-labeled hashtag antibodies (BioLegend). NKG2C⁺/NKG2C^– populations were pooled at a 1:1 ratio, and the cell concentration was adjusted to 1,000 cells per µl for further processing.

scRNA-seq

scRNA-seq libraries were prepared using the Chromium Single Cell 3′ Reagent kit v2/v3.1 chemistry (10x Genomics) following the manufacturer’s instructions with the modifications required for recovery of ADTs and HTOs¹⁹. ADT and HTO libraries were amplified using the KAPA HiFi ready mix (Roche) with sample-specific index primers (Illumina small RNA RPIx/Truseq D7xx)¹⁹ and purified using SPRIselect reagent. Library size and quality were analyzed using a Fragment Analyzer (Advanced Analytical) before fragmentation and after final purification. Final library concentrations were measured on a Qubit 2.0 fluorometer (Thermo Fisher). Libraries were sequenced on a NextSeq500 or NovaSeq6000 sequencer (Illumina) using the read configurations recommended by 10x Genomics (v2: R1 26 cycles, R2 98 cycles, I1 8 cycles; v3: R1 28 cycles, R2 91 cycles, I1 8 cycles).

Stimulation of NK cells from HCMV^– donors

RMA-S/HLA-E target cells were prepared by seeding cells in 24- to 96-well plates at a cell concentration of 2 × 10⁶ cells per ml in OptiMEM (Thermo Fisher) and irradiated with 3,000 cGy. Afterward, cells were pulsed either with the negative control peptide VMAPQSLLL or the activating peptide LFL (Peptides&Elephants) at a concentration of 300 µM for 6–12 h.

For scATAC-seq, NK cells were isolated from freshly drawn blood by magnetic depletion of CD3⁺ cells, followed by enrichment of CD56⁺ cells. For read-out by flow cytometry, cells were isolated from buffy coats, and CD56⁺ cells were enriched using CD56 microbeads and were cryopreserved in FBS containing 10% DMSO. Viable CD56⁺CD3⁻ cells were sorted from the cryopreserved CD56-enriched fractions. Purified NK cells were cocultured for 12 h with peptide-pulsed RMA-S/HLA-E cells in complete medium supplemented with 10 ng ml^–1 IL-15 (Miltenyi Biotec) and the respective peptides at 300 µM in the presence or absence of 10 ng ml^–1 IL-12 (Miltenyi Biotec) and 100 ng ml^–1 IL-18 (MBL International). Cells were stained with fluorochrome-labeled antibodies and either analyzed by flow cytometry or additionally stained with nucleotide barcode-labeled antibodies, including hashtag antibodies, marking experimental conditions and donors as described above before sorting for viable CD7⁺NKG2C⁺ cells. Sorted cells were pooled at equal numbers per condition and processed for scATAC-seq as described above.

Removal of sensitive genetic variant information

Raw BAM files for all experiments were processed with BAMboozle⁵⁹ to remove potentially identifying donor-related single nucleotide polymorphisms and indels by replacing them with the sequences present in the corresponding reference genomes. To enable reproducibility of the clonotype analysis based on mitochondrial mutations, the mitochondrial genotyping results were deposited with the raw and processed data for each experiment.

Preprocessing of scATAC-seq data

Base call files were demultiplexed using cellranger-atac v1.2.0 mkfastq and bcl2fastq v2.20.0.422 into scATAC-seq and ADT/HTO libraries. scATAC-seq reads were mapped with cellranger-atac count to the GRCh38 reference genome, hardmasked for regions that would otherwise interfere with mapping to the mitochondrial genome¹⁷, generating count tables and fragment files as output for further analysis. ADT and HTO reads were preprocessed using ASAP to kite¹⁸, followed by pseudoalignment with kallisto⁶⁰ to a mismatch map generated with kite and counting with bustools⁶¹.

Integrated analysis of ex vivo scATAC-seq data

Peak sets called from different experiments were reduced to a joint peak set as a basis for combined analysis of individual experiments using Signac⁶². Cells were filtered for outliers based on nucleosome signal (<1–1.2), TSS enrichment (>2.5–4) and the frequency of reads in peaks (>55–70 %) and blacklisted regions (<0.0001). ADT and HTO counts were imported, centered log ratio normalized and joined with the scATAC-seq data. Normalized hashtag reads were used for exclusion of doublets and to demultiplex donors. Chromatin accessibility counts were normalized by term frequency inverse document frequency, data from different experiments were merged, and dimensionality was reduced by singular value decomposition. The resultant LSI was integrated across experiments using harmony⁶³ and used as input into further dimensionality reduction by uniform manifold approximation (UMAP) and neighborhood graph-based clustering. A small population of non-NK cells, as judged by their lack of CD56 and CD16 expression concomitant with high levels of CD127 and CD117, was excluded. The merged and integrated dataset was used for peak calling using MACS2 (ref. ⁶⁴) to enable more sensitive calling of cluster-specific open chromatin regions. This new assay was again processed by LSI, followed by anchor-based integration of donors³⁵. The UMAP embeddings and clustering for final analysis were based on this integrated LSI. Unbiased clustering initially yielded three major clusters that were annotated based on surface protein expression, with early CD56^dim NK cells separated from CD56^bright NK cells by further subclustering. Gene activities per cell were calculated by summing up counts within the gene body and up to 2,000 bases upstream of the TSS.

For analysis of the global signatures, differential accessibility analysis of individual chromatin regions was performed using logistic regression with the number of peaks as latent variable⁶⁵, while gene activities were assessed by Wilcoxon rank-sum test. For both, the log₂ (fold change) threshold was set at 0.1, and significant hits from the direct comparison of CD56^bright to CD56^dim were added for visualization to include the CD56^dim signature that was mostly shared with adaptive NK cells. Averaged values were visualized in heat maps using tidyheatmap, an implementation of pheatmap.

To compare HCMV⁺ and HCMV⁻ donors, these were separated from the integrated dataset followed by UMAP embedding and clustering as described above. To directly compare the number of DARs between NKG2C⁺ and NKG2C⁻ cells, these populations were defined based on anti-PE ADT counts marking NKG2C surface expression and then downsampled to equal numbers (4,300 NKG2C⁺/NKG2C^– cells) for both donor groups. Further, NKG2C⁺ NK cells from HCMV⁺ donors were extracted, followed by UMAP embedding, and their cluster identities were assessed based on the signatures from the integrated analysis of the full dataset.

Motif analysis

Position frequency matrices for TF motif analysis were downloaded from the JASPAR2020 human core database and amended by Stat-TF motifs from the murine database⁶⁶. Motif activities per cell were calculated using chromVAR²⁷. Enrichment of motifs in differentially accessible peaks between clusters was assessed using the hypergeometric test implemented in Signac. De novo motif analysis on the adaptive NK-cell-specific peaks was performed with homer v4.11 findMotifsGenome.pl using the standard parameters⁶⁷.

Mitochondrial genotyping and clonotypes

Mitochondrial genotyping was performed with mgatk¹⁷ in tenx mode using the barcodes identified as cells by cellranger. Only cells with at least 5× coverage of the mitochondrial genome were included in the analysis, achieving a median coverage of 11–20× across experiments. As sensitivity and positive predictive value are relatively independent of coverage for mutations with high heteroplasmy¹⁷, those could be confidently detected at this coverage. Using a combination of highly abundant, homoplasmic mitochondrial mutations characteristic of each donor, we noted that donor demultiplexing was slightly incomplete in some cases, so we used these mutations as additional donor barcodes to further improve demultiplexing before the donor-specific analysis. Mitochondrial mutations called by mgatk were filtered for high-confident variants detected in at least three cells, with strand concordance of >65% and a variance-to-mean ratio of >0.01. Clonotypes were identified by clustering on a neighborhood graph constructed on mitochondrial mutation frequency using the Euclidean distance metric, because we noted that the majority of adaptive clonotypes were defined by mutations with high allele frequencies and thereby achieved better clonotype resolution than with the cosine distance metric originally suggested for mtscATAC-seq¹⁷. Clustering parameters were empirically optimized per donor so that clonotypes were defined by approximately one mutation, and only those clonotypes were included that had at least one significantly enriched mutation. Clonotypes only defined by differential frequency of highly abundant mutations were excluded from the analysis. Association of clonotypes to clusters defined by their chromatin accessibility profiles was analyzed by χ² test and comparing the resultant false discovery rates (FDRs) to a randomly permuted matrix¹⁷. Analysis of clonotype association to open chromatin regions was performed using a χ² test on the binarized accessibility matrix and compared to a randomly permuted matrix.

Donor-specific and donor-unique signatures

Individual HCMV⁺ donors were analyzed based on the peak set called for the integrated dataset. Both time points were included in this analysis and integrated using harmony⁶². Dimensionality reduction and clustering was performed for each individual donor as described above. To analyze overlaps and differences between the signatures defining clusters across donors, differential accessibility analysis was performed per donor as described for the integrated dataset, and overlaps were visualized using eulerr. To assess the actual signatures, adaptive subclusters defined in each donor were merged, and averaged accessibility per cluster was visualized with tidyheatmap. Differential gene scores and motif activities between the two emerging adaptive subcluster groups were analyzed by directly comparing them to each other within the merged object. To assess similarity of their signatures to those of CD56^bright/early CD56^dim and CD56^dim NK cells, fold changes of the adaptive subcluster group-defining DARs were plotted for both comparisons and analyzed for linear correlation.

Cluster heterogeneity

Cluster heterogeneity was assessed based on the median Euclidian distance of 200 randomly sampled cells per cluster to their 10 nearest neighbors in LSI space (components 2–30). Sampling was repeated 100 times, and results were plotted as box plots of the distribution of median distances of all simulations. The k parameter was varied from 5 to 30 to test for robustness.

Longitudinal analysis

The two time points for donors P2, P3 and P4 were separated from the merged, harmony-integrated objects, and the signatures defining adaptive subclusters were analyzed separately for each time point using hierarchical clustering and visualization with tidyheatmap. Allele frequency of clonotype-defining mutations was visualized for each time point. To analyze the stability of clonotypes over time, only clonotypes that had an association with chromatin accessibility-defined clusters with an FDR of <0.05 were included; these were further curated by assessment of their association with adaptive subclusters in UMAP embeddings. Stability of clonotypes significantly associated with adaptive subclusters was analyzed by testing association of clonotypes with time points by Fisher’s exact test. The stability of the total clonotype distribution was further assessed by calculating their log₂ (fold change) values between time points and comparing the distribution to a randomly permuted clonotype–time point relationship by Kolmogorov–Smirnov test.

Analysis of scATAC-seq data after activation

Quality control, joining of ADT, HTO and scATAC-seq counts, demultiplexing, peak calling on the full dataset with MACS2, normalization and dimensionality reduction by LSI were performed as described for the ex vivo analysis. The two analyzed donors were integrated with harmony, and the corrected LSI was used for UMAP embedding and clustering. After inspection of the distribution of cells from the different conditions and donors, the majority of cells were clearly separated based on whether they had received peptide or cytokine stimulation, but we also noted two smaller clusters, each enriched for cells from one donor, respectively, in which cells from different conditions were mixed, likely driven by underlying donor-specific signatures that were not corrected by harmony integration. To analyze stimulus-induced changes in chromatin accessibility, we focused the analysis on the majority of cells that clustered by conditions and irrespective of donor origin. The cell barcodes used for the final analysis were deposited together with the raw and processed data. After clustering these cells at higher resolution, we annotated clusters according to a clear enrichment of cells derived from different conditions (Extended Data Fig. 4b). To assess the similarity of the in vitro induced signatures to those defining NK cell subsets ex vivo, the MACS2-called peak set of the ex vivo analysis was quantified for the cells stimulated in vitro, and the two datasets were integrated using anchor-based integration, setting the ex vivo dataset as reference and the in vitro dataset as query³⁵. Integration anchors were then used to transfer cluster labels from the ex vivo dataset to the in vitro dataset, and the per-cell prediction scores were plotted for each cluster. To identify which individual chromatin regions overlap between adaptive NK cells ex vivo and the cells stimulated with cytokines and peptide, fold changes for DARs overlapping between the comparisons of CD56^dim cells to adaptive NK cells ex vivo and control cells to LFL + IL-12+IL-18 cells in vitro were analyzed for linear correlation.

Preprocessing of scRNA-seq data

Base call files were demultiplexed using cellranger v3.0.2 mkfastq and bcl2fastq v2.20.0.422 into scRNA-seq and ADT/HTO libraries. scRNA-seq reads were mapped to the hg19 reference genome with cellranger count, generating gene-per-cell count tables as output for further analysis. ADT and HTO reads were either quantified with CITE-Seq-Count⁶⁸ or pseudoaligned with kallisto⁶⁰ to a mismatch map generated with kite and counted with bustools⁶¹.

Downstream analysis of scRNA-seq data

scRNA-seq counts were imported into Seurat⁶⁹ and filtered for outliers regarding frequency of mitochondrial transcripts (v2: <5–6%; v3: <10%), total number of transcripts (v2: <6,000; v3: <10,000) and number of genes (v2: >500–800, <2,000–2,500; v3: >1,000, <5,000) per cell. Contaminating erythrocytes and B cells were excluded based on the expression of HBA1/HBA2 and IGJ, respectively. ADT/HTO counts were imported, centered log ratio normalized and joined with the scRNA-seq data. Normalized hashtag reads were used for exclusion of doublets and to demultiplex NKG2C⁺/NKG2C^– populations and donors. Initially, donors were analyzed individually. Counts were normalized with scran⁷⁰ and used for principal-component analysis (PCA) on the 2,000 most variable genes. The number of principal components used for UMAP embedding and clustering was chosen based on elbow plots for each individual. After initial high-resolution clustering, a small population of non-NK cells, as judged by their lack of CD56 expression concomitant with high levels of IL7R, GATA3, IL2RA and CD40LG, were excluded. The PCA was recalculated on the remaining cells and used as input for UMAP embedding and clustering as before. HCMV⁺ and HCMV⁻ donors were then integrated separately based on the variable features overlapping between HCMV⁺ donors by using anchor-based integration. The integrated data were scaled, regressing out the number of transcripts per cell. The HCMV⁺/HCMV^– objects were merged, and the integrated data were rescaled, also regressing out the number of transcripts per cell. A PCA was performed on the shared variable features and was used as input for UMAP embedding and clustering. The normalized, non-integrated RNA assay was used for all differential expression analyses. For differential expression analysis between NKG2C⁺ and NKG2C^– populations in HCMV⁺ and HCMV⁻ donors, cells were downsampled to equal numbers per population and donor group (6,000 cells each), and only genes that were detected with a minimum count of 100 in both objects were included. Further, NKG2C⁺ NK cells from HCMV⁺ donors were extracted using hashtag counts, followed by UMAP embedding. Signatures from the integrated analysis were assessed to confirm the cluster identities.

Integration of scRNA-seq and scATAC-seq data

Fully processed and donor-integrated datasets of both modalities were integrated by anchor-based integration as implemented in Seurat^35,69 using the integrated RNA assay as reference and the gene scores calculated from scATAC-seq data as query. Integration anchors were used to impute RNA expression of clusters defined by chromatin accessibility and inspected for consistency of key marker genes. Imputed RNA expression was assessed for correlation with chromatin accessibility to identify links between individual chromatin regions and gene expression.

Statistical analysis and reproducibility

Reported P values were corrected for multiple testing. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications^20,22. Data distribution was assumed to be normal, but this was not formally tested. No randomization was performed. Data collection and analysis were not performed blind to the conditions of the experiments. No data points were excluded.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

scRNA-seq and scATAC-seq BAM files and processed data as fragment files, gene count table and antibody count tables have been deposited in the Gene Expression Omnibus under accession code GSE197037. Single-nucleotide variants and indels were removed from BAM files as described in the Methods. To enable reproducibility of the clonotype analysis based on mitochondrial mutations, the mgatk results were deposited for each experiment. scATAC-seq data were mapped to the GRCh38 (GCF_000001405.39) reference genome, and scRNA-seq data were mapped to the hg19 (GCF_000001405.25) reference genome, as supplied by 10x Genomics. Source data are provided with this paper.

Code availability

All code has been deposited in GitHub (https://github.com/timorueckert/Clonal_NK).

References

Pradeu, T. & Du Pasquier, L. Immunological memory: what’s in a name? Immunol. Rev. 283, 7–20 (2018).
Article CAS PubMed Google Scholar
Netea, M. G. et al. Defining trained immunity and its role in health and disease. Nat. Rev. Immunol. 20, 375–388 (2020).
Article CAS PubMed PubMed Central Google Scholar
Kaufmann, E. et al. BCG educates hematopoietic stem cells to generate protective innate immunity against tuberculosis. Cell 172, 176–190 (2018).
Article CAS PubMed Google Scholar
Larsen, S. B. et al. Establishment, maintenance, and recall of inflammatory memory. Cell Stem Cell 28, 1758–1774.e8 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kleinnijenhuis, J. et al. Bacille Calmette–Guerin induces NOD2-dependent nonspecific protection from reinfection via epigenetic reprogramming of monocytes. Proc. Natl Acad. Sci. USA 109, 17537–17542 (2012).
Article CAS PubMed PubMed Central Google Scholar
Mitroulis, I. et al. Modulation of myelopoiesis progenitors is an integral component of trained immunity. Cell 172, 147–161 (2018).
Article CAS PubMed PubMed Central Google Scholar
Hammer, Q., Rückert, T. & Romagnani, C. Natural killer cell specificity for viral infections. Nat. Immunol. 19, 800–808 (2018).
Article CAS PubMed Google Scholar
Berry, R., Watson, G. M., Jonjic, S., Degli-Esposti, M. A. & Rossjohn, J. Modulation of innate and adaptive immunity by cytomegaloviruses. Nat. Rev. Immunol. 20, 113–127 (2020).
Article CAS PubMed Google Scholar
Sun, J. C., Beilke, J. N. & Lanier, L. L. Adaptive immune features of natural killer cells. Nature 457, 557–561 (2009).
Article CAS PubMed PubMed Central Google Scholar
Hammer, Q. et al. Peptide-specific recognition of human cytomegalovirus strains controls adaptive natural killer cells. Nat. Immunol. 19, 453–463 (2018).
Article CAS PubMed Google Scholar
Béziat, V. et al. NK cell responses to cytomegalovirus infection lead to stable imprints in the human KIR repertoire and involve activating KIRs. Blood 121, 2678–2688 (2013).
Article PubMed PubMed Central Google Scholar
Grassmann, S. et al. Distinct surface expression of activating receptor Ly49H drives differential expansion of NK cell clones upon murine cytomegalovirus infection. Immunity 50, 1391–1400 (2019).
Article CAS PubMed Google Scholar
Truitt, L. L. et al. Impact of CMV infection on natural killer cell clonal repertoire in CMV-naive rhesus macaques. Front. Immunol. 10, 2381 (2019).
Article CAS PubMed PubMed Central Google Scholar
Wu, C. et al. Clonal expansion and compartmentalized maintenance of rhesus macaque NK cell subsets. Sci. Immunol. 3, eaat9781 (2018).
Article PubMed PubMed Central Google Scholar
Adams, N. M., Grassmann, S. & Sun, J. C. Clonal expansion of innate and adaptive lymphocytes. Nat. Rev. Immunol. 20, 694–707 (2020).
Article CAS PubMed Google Scholar
Ludwig, L. S. et al. Lineage tracing in humans enabled by mitochondrial mutations and single-cell genomics. Cell 176, 1325–1339 (2019).
Article CAS PubMed PubMed Central Google Scholar
Lareau, C. A. et al. Massively parallel single-cell mitochondrial DNA genotyping and chromatin profiling. Nat. Biotechnol. 39, 451–461 (2021).
Article CAS PubMed Google Scholar
Mimitou, E. P. et al. Scalable, multimodal profiling of chromatin accessibility, gene expression and protein levels in single cells. Nat. Biotechnol. 39, 1246–1258 (2021).
Article CAS PubMed PubMed Central Google Scholar
Stoeckius, M. et al. Simultaneous epitope and transcriptome measurement in single cells. Nat. Methods 14, 865–868 (2017).
Article CAS PubMed PubMed Central Google Scholar
Crinier, A. et al. High-dimensional single-cell analysis identifies organ-specific signatures and conserved NK cell subsets in humans and mice. Immunity 49, 971–986 (2018).
Article CAS PubMed PubMed Central Google Scholar
Collins, P. L. et al. Gene regulatory programs conferring phenotypic identities to human NK cells. Cell 176, 348–360 (2019).
Article CAS PubMed Google Scholar
Yang, C. et al. Heterogeneity of human bone marrow and blood natural killer cells defined by single-cell transcriptome. Nat. Commun. 10, 3931 (2019).
Article PubMed PubMed Central Google Scholar
Smith, S. L. et al. Diversity of peripheral blood human NK cells identified by single-cell RNA sequencing. Blood Adv. 4, 1388–1406 (2020).
Article CAS PubMed PubMed Central Google Scholar
Holmes, T. D. et al. The transcription factor Bcl11b promotes both canonical and adaptive NK cell differentiation. Sci. Immunol. 6, eabc9801 (2021).
Article CAS PubMed PubMed Central Google Scholar
Cichocki, F. et al. ARID5B regulates metabolic programming in human adaptive NK cells. J. Exp. Med. 215, 2379–2395 (2018).
Article CAS PubMed PubMed Central Google Scholar
Gumá, M. et al. Imprint of human cytomegalovirus infection on the NK cell receptor repertoire. Blood 104, 3664–3671 (2004).
Article PubMed Google Scholar
Schep, A. N., Wu, B., Buenrostro, J. D. & Greenleaf, W. J. chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. Nat. Methods 14, 975–978 (2017).
Article CAS PubMed PubMed Central Google Scholar
Cooper, M. A. et al. Human natural killer cells: a unique innate immunoregulatory role for the CD56^bright subset. Blood 97, 3146–3151 (2001).
Article CAS PubMed Google Scholar
Foley, B. et al. Cytomegalovirus reactivation after allogeneic transplantation promotes a lasting increase in educated NKG2C⁺ natural killer cells with potent function. Blood 119, 2665–2674 (2012).
Article CAS PubMed PubMed Central Google Scholar
Luetke-Eversloh, M. et al. Human cytomegalovirus drives epigenetic imprinting of the IFNG locus in NKG2C^hi natural killer cells. PLoS Pathog. 10, e1004441 (2014).
Article PubMed PubMed Central Google Scholar
Ostuni, R. et al. Latent enhancers activated by stimulation in differentiated cells. Cell 152, 157–171 (2013).
Article CAS PubMed Google Scholar
Yukawa, M. et al. AP-1 activity induced by co-stimulation is required for chromatin opening during T cell activation. J. Exp. Med. 217, e20182009 (2020).
Article PubMed Google Scholar
Valés-Gómez, M., Reyburn, H. T., Erskine, R. A., López-Botet, M. & Strominger, J. L. Kinetics and peptide dependency of the binding of the inhibitory NK receptor CD94/NKG2-A and the activating receptor CD94/NKG2-C to HLA-E. EMBO J. 18, 4250–4260 (1999).
Article PubMed PubMed Central Google Scholar
Wiedemann, G. M. et al. Deconvoluting global cytokine signaling networks in natural killer cells. Nat. Immunol. 22, 627–638 (2021).
Article CAS PubMed PubMed Central Google Scholar
Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902 (2019).
Article CAS PubMed PubMed Central Google Scholar
Schlums, H. et al. Cytomegalovirus infection drives adaptive epigenetic diversification of NK cells with altered signaling and effector function. Immunity 42, 443–456 (2015).
Article CAS PubMed PubMed Central Google Scholar
Liu, L. L. et al. Critical role of CD2 co-stimulation in adaptive natural killer cell responses revealed in NKG2C-deficient humans. Cell Rep. 15, 1088–1099 (2016).
Article CAS PubMed PubMed Central Google Scholar
Juelke, K. et al. CD62L expression identifies a unique subset of polyfunctional CD56^dim NK cells. Blood 116, 1299–1307 (2010).
Article CAS PubMed Google Scholar
Moskowitz, D. M. et al. Epigenomics of human CD8 T cell differentiation and aging. Sci. Immunol. 2, eaag0192 (2017).
Article PubMed PubMed Central Google Scholar
Scharer, C. D., Barwick, B. G., Guo, M., Bally, A. P. R. & Boss, J. M. Plasma cell differentiation is controlled by multiple cell division-coupled epigenetic programs. Nat. Commun. 9, 1698 (2018).
Article PubMed PubMed Central Google Scholar
Lau, C. M. et al. Epigenetic control of innate and adaptive immune memory. Nat. Immunol. 19, 963–972 (2018).
Article CAS PubMed PubMed Central Google Scholar
de Laval, B. et al. C/EBPβ-dependent epigenetic memory induces trained immunity in hematopoietic stem cells. Cell Stem Cell 26, 657–674 (2020).
Article PubMed Google Scholar
Nabekura, T. & Lanier, L. L. Tracking the fate of antigen-specific versus cytokine-activated natural killer cells after cytomegalovirus infection. J. Exp. Med. 213, 2745–2758 (2016).
Article CAS PubMed PubMed Central Google Scholar
Sun, J. C. et al. Proinflammatory cytokine signaling required for the generation of natural killer cell memory. J. Exp. Med. 209, 947–954 (2012).
Article CAS PubMed PubMed Central Google Scholar
Flommersfeld, S. et al. Fate mapping of single NK cells identifies a type 1 innate lymphoid-like lineage that bridges innate and adaptive recognition of viral infection. Immunity 54, 2288–2304 (2021).
Article CAS PubMed PubMed Central Google Scholar
Horowitz, A. et al. Genetic and environmental determinants of human NK cell diversity revealed by mass cytometry. Sci. Transl. Med. 5, 208ra145 (2013).
Article PubMed PubMed Central Google Scholar
Lopez-Vergès, S. et al. Expansion of a unique CD57⁺NKG2C^hi natural killer cell subset during acute human cytomegalovirus infection. Proc. Natl Acad. Sci. USA 108, 14725–14732 (2011).
Article PubMed PubMed Central Google Scholar
Adams, N. M. et al. Cytomegalovirus infection drives avidity selection of natural killer cells. Immunity 50, 1381–1390 (2019).
Article CAS PubMed PubMed Central Google Scholar
Buchholz, V. R. et al. Disparate individual fates compose robust CD8⁺ T cell immunity. Science 340, 630–635 (2013).
Article CAS PubMed Google Scholar
Grassmann, S. et al. Early emergence of T central memory precursors programs clonal dominance during chronic viral infection. Nat. Immunol. 21, 1563–1573 (2020).
Article CAS PubMed Google Scholar
Hartmann, F. J. et al. Single-cell metabolic profiling of human cytotoxic T cells. Nat. Biotechnol. 39, 186–197 (2021).
Article CAS PubMed Google Scholar
Shaffer, S. M. et al. Memory sequencing reveals heritable single-cell gene expression programs associated with distinct cellular behaviors. Cell 182, 947–959 (2020).
Article CAS PubMed PubMed Central Google Scholar
Mold, J. E. et al. Clonally heritable gene expression imparts a layer of diversity within cell types. Preprint at bioRxiv https://doi.org/10.1101/2022.02.14.480352 (2022).
Corat, M. A. F. et al. Acquired somatic mutations in PNH reveal long-term maintenance of adaptive NK cells independent of HSPCs. Blood 129, 1940–1946 (2017).
Article CAS PubMed PubMed Central Google Scholar
Schlums, H. et al. Adaptive NK cells can persist in patients with GATA2 mutation depleted of stem and progenitor cells. Blood 129, 1927–1939 (2017).
Article CAS PubMed PubMed Central Google Scholar
Borrego, F., Ulbrecht, M., Weiss, E. H., Coligan, J. E. & Brooks, A. G. Recognition of human histocompatibility leukocyte antigen (HLA)-E complexed with HLA class I signal sequence-derived peptides by CD94/NKG2 confers protection from natural killer cell-mediated lysis. J. Exp. Med. 187, 813–818 (1998).
Article CAS PubMed PubMed Central Google Scholar
Cossarizza, A. et al. Guidelines for the use of flow cytometry and cell sorting in immunological studies (third edition). Eur. J. Immunol. 51, 2708–3145 (2021).
Article CAS PubMed Google Scholar
Milush, J. M. et al. Functionally distinct subsets of human NK cells and monocyte/DC-like cells identified by coexpression of CD56, CD7, and CD4. Blood 114, 4823–4831 (2009).
Article CAS PubMed PubMed Central Google Scholar
Ziegenhain, C. & Sandberg, R. BAMboozle removes genetic variation from human sequence data for open data sharing. Nat. Commun. 12, 6216 (2021).
Article CAS PubMed PubMed Central Google Scholar
Bray, N. L., Pimentel, H., Melsted, P. & Pachter, L. Near-optimal probabilistic RNA-seq quantification. Nat. Biotechnol. 34, 525–527 (2016).
Article CAS PubMed Google Scholar
Melsted, P. et al. Modular, efficient and constant-memory single-cell RNA-seq preprocessing. Nat. Biotechnol. 39, 813–818 (2021).
Article CAS PubMed Google Scholar
Stuart, T., Srivastava, A., Madad, S., Lareau, C. A. & Satija, R. Single-cell chromatin state analysis with Signac. Nat. Methods 18, 1333–1341 (2021).
Article CAS PubMed PubMed Central Google Scholar
Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019).
Article CAS PubMed PubMed Central Google Scholar
Zhang, Y. et al. Model-based analysis of ChIP–seq (MACS). Genome Biol. 9, R137 (2008).
Article PubMed PubMed Central Google Scholar
Ntranos, V., Yi, L., Melsted, P. & Pachter, L. A discriminative learning approach to differential expression analysis for single-cell RNA-seq. Nat. Methods 16, 163–166 (2019).
Article CAS PubMed Google Scholar
Fornes, O. et al. JASPAR 2020: update of the open-access database of transcription factor binding profiles. Nucleic Acids Res. 48, D87–D92 (2020).
CAS PubMed Google Scholar
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
Article CAS PubMed PubMed Central Google Scholar
Roelli, P., bbimber, Flynn, B., santiagorevale & Gui, G. Hoohm/CITE-seq-Count: 1.4.2. Zenodo https://doi.org/10.5281/zenodo.2590196 (2019).
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 (2021).
Article CAS PubMed PubMed Central Google Scholar
Lun, A. T. L., McCarthy, D. J. & Marioni, J. C. A step-by-step workflow for low-level analysis of single-cell RNA-seq data with Bioconductor. F1000Res. 5, 2122 (2016).
PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank members of the Romagnani group, G. Gasteiger, D.R. Withers, T. Schüler and T.N. Haschler for critically reading the manuscript. We are grateful for support by the DRFZ flow cytometry core facility for assistance with cell sorting. We thank O. Knight and P. Durek for help with setting up bioinformatic pipelines and the DRFZ lab managers, M. Klemm and G. Guerra, for technical support. This work was funded by the ERC Advanced Grant ‘MEM-CLONK’ (101055157) to C.R., the Deutsche Forschungsgemeinschaft (DFG) grants SPP 1937 (RO3565/4-2) and SFB TRR241 B02 to C.R., Leibniz-Science Campus Chronic Inflammation, Leibniz-Kooperative Exzellenz K259/2019 to C.R. and Berlin Health Innovations (BHI) Validation Fund to C.R. and T.R. Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.

Funding

Open access funding provided by Deutsches Rheuma-Forschungszentrum Berlin (DRFZ)

Author information

Authors and Affiliations

Innate Immunity, Deutsches Rheuma-Forschungszentrum Berlin (DRFZ), ein Leibniz Institut, Berlin, Germany
Timo Rückert & Chiara Romagnani
Department of Pathology, Stanford University, Stanford, CA, USA
Caleb A. Lareau
Therapeutic Gene Regulation, Deutsches Rheuma-Forschungszentrum Berlin (DRFZ), ein Leibniz Institut, Berlin, Germany
Mir-Farzin Mashreghi
Berlin Institute of Health at Charité Universitätsmedizin Berlin, Berlin, Germany
Leif S. Ludwig
Max‐Delbrück‐Center for Molecular Medicine in the Helmholtz Association (MDC), Institute for Medical Systems Biology (BIMSB), Berlin, Germany
Leif S. Ludwig
Charité—Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin and Humboldt Universität zu Berlin, Medizinische Klinik für Gastroenterologie, Infektiologie und Rheumatologie, Berlin, Germany
Chiara Romagnani
Leibniz-Science Campus Chronic Inflammation, Berlin, Germany
Chiara Romagnani

Authors

Timo Rückert
View author publications
You can also search for this author in PubMed Google Scholar
Caleb A. Lareau
View author publications
You can also search for this author in PubMed Google Scholar
Mir-Farzin Mashreghi
View author publications
You can also search for this author in PubMed Google Scholar
Leif S. Ludwig
View author publications
You can also search for this author in PubMed Google Scholar
Chiara Romagnani
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

Conceptualization, T.R. and C.R.; methodology, T.R. with input from C.R., L.S.L. and C.A.L.; formal analysis, T.R. and C.A.L.; investigation, T.R.; resources, C.R. and M.-F.M.; writing the original draft, T.R. and C.R. with input from all authors; visualization, T.R. with input from C.R., L.S.L. and C.A.L.; supervision, C.R.; project administration, C.R.; funding acquisition, C.R.

Corresponding authors

Correspondence to Timo Rückert or Chiara Romagnani.

Ethics declarations

Competing interests

T.R. and C.R. are listed as inventors on a patent series covering NKG2C-activating peptides. The other authors declare no competing interests.

Peer review

Peer review information

Nature Immunology thanks the anonymous reviewers for their contribution to the peer review of this work. N. Bernard was the primary editor on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Mapping NK cell subsets onto transcriptional and epigenetic landscapes.

(a) Sorting strategy for one representative HCMV⁻ and HCMV⁺ donor, respectively. (b) Transcription start site (TSS) enrichment plot overlaid for all donors. (c) Expression of proliferation genes MKI67 and STMN1. (d-f) Row-scaled differentially accessible regions (D), genes (E) and differentially expressed genes (F) per cluster. Exemplary cluster-associated genes are listed in the respective colors. (h) Gene accessibility, imputed expression (violins), and their Pearson correlation (indicated as links) for RUNX2 and ZEB2. Created with BioRender.com.

Source data

Extended Data Fig. 2 Integration of chromatin accessibility and gene expression.

(a) Pearson correlation between open chromatin regions and gene expression, p-value from one-sided z-test of background-normalized correlation-coefficient as implemented in Signac⁶². (b-e) Chromatin accessibility, imputed expression (violins) and their Pearson correlation (indicated as links) for indicated genes TCF7 (B), RUNX2 (C), ZEB2 (D), KLRC2 and KLRC1 (E).

Extended Data Fig. 3 Distinct signatures dissect NKG2C⁺ NK cells into naïve and adaptive.

(a + d) UMAP embedding of scATAC-seq (A) and scRNA-seq (D) of NKG2C⁺ NK cells from HCMV⁺ individuals. (b + e) Accessibility (B) and expression (E) of key genes regulated in adaptive NK cells. (c + f) Row-scaled accessibility (C) and expression (F) of NK cell subset-specific signatures from fully integrated datasets (Fig. 1) in NKG2C⁺ NK cell clusters from HCMV⁺ donors. (g−i) Representative expression (G) and frequencies (H) of FCER1G⁺ NKp30⁺ (‘Naive’) and FCER1G^- NKp30⁻ (‘Adaptive’) NKG2C⁺ NK cells and distribution of naïve NKG2C⁺ NK cells into CD56^bright and CD56^dim subsets (I) in HCMV⁻ (n = 9) and HCMV⁺ (n = 9) donors. Kruskal-Wallis test with Dunn’s post-hoc test. Created with BioRender.com.

Extended Data Fig. 4 HCMV infection leaves a persistent chromatin footprint enriched for AP1 motifs.

(a) Differentially active motifs between CD56^dim and CD56^bright NK cells by Wilcoxon rank sum test. (b-c) Motif enrichment in chromatin regions specifically accessible in CD56^bright (B) and CD56^dim NK cells as determined by hypergeometric test. (d) Differentially active motifs between adaptive and CD56^dim NK cells by Wilcoxon rank sum test. (e) Activity of the respective TF motifs projected onto UMAP embedding. (f) Accessibility of IFNG-associated open-chromatin regions containing AP1 or STAT4 and NFκB motifs. (g) Per cluster accessibility of AP1-motif containing chromatin regions in the proximity of CADM1 and AIM2. Created with BioRender.com.

Extended Data Fig. 5 Synergistic imprinting by HCMV peptides and pro-inflammatory cytokines.

(a) Distribution of cells colored by donor origin and quantification per cluster. (b) Distribution of cells colored by culture condition and quantification per cluster, utilized for cluster annotation. (c) Transcription start site (TSS) enrichment plot overlaid for all clusters. (d + e) Representative (D) and mean (E) CD137 expression on NKG2C⁺ NKG2A⁻ NK cells from HCMV⁻ donors (n = 9) cultured under the indicated conditions; One-way ANOVA with Holm-Šídák adjustment. (f) JUNB motif activity per cluster; Wilcoxon rank sum test with Benjamini-Hochberg adjustment. (g-h) Representative (G) and quantification (H) of NKG2A expression in naïve and adaptive NKG2C⁺ NK cells from HCMV⁻ (n = 9) and HCMV⁺ (n = 9) donors; see Extended Data Fig. 2g for gating strategy; Kruskal-Wallis test with Dunn’s post-hoc test. (i) Accessibility of regions near key adaptive NK cell-related genes regulated after stimulation. (j) Linear correlation between differentially accessible regions both induced by in vitro activation and observed ex vivo between adaptive and CD56^dim NK cells; error bands show 95 % confidence interval, p-value from two-sided F-test. (k) Motif enrichment in open chromatin regions displayed in (J) as determined by one-sided hypergeometric test with Benjamini-Hochberg adjustment. Created with BioRender.com.

Source data

Extended Data Fig. 6 Convergent and divergent epigenetic features of adaptive NK cells.

(a) UMAP embedding of NK cells from HCMV⁺ individual P1 analyzed by scATAC-seq. (b) Surface expression of the indicated proteins projected on the UMAP embeddings for each donor. Educating KIRs are underlined, see Supplementary Table 2. (c) Column-scaled accessibility of differentially accessible regions (DARs) shared between at least three donors comparing the total adaptive and CD56^dim compartments. (d) Accessibility of adaptive NK-cell defining AIM2 region within each donor and adaptive subcluster. (e) FOS::JUNB motif activity within each donor. (f) Accessibility of representative subcluster-specific chromatin regions for each donor.

Extended Data Fig. 7 Adaptive subcluster groups are defined by a maturation gradient.

(a) Linear correlation between DARs comparing the two adaptive subcluster groups and CD56^dim to early CD56^dim NK cells; error bands show 95 % confidence interval, p-value from two-sided F-test. (b) Accessibility of adaptive group-defining regions within TCF7 and PCNT for each donor and adaptive subcluster. (c) TCF7L2 and CTCF motif activity for each donor and adaptive subcluster; two-sided Wilcoxon test with Bonferroni adjustment. (d) FOS::JUNB motif activity for each donor and adaptive subcluster; two-sided Wilcoxon test with Bonferroni adjustment. (e) Normalized CD62L surface expression per cluster; two-sided Wilcoxon rank sum test with Bonferroni adjustment.

Extended Data Fig. 8 Clonal expansion underlies divergent epigenetic signatures of adaptive NKG2C⁺ NK cells.

(a) Cluster heterogeneity of donor P1 NK cells as assessed by measuring the median distance of 200 randomly sampled cells to their k nearest neighbors and repeating this process 100 times. (b) The k parameter was varied from 5-30 for donor P4 to assess robustness of this heterogeneity metric. The upper and lower hinges of boxplots correspond to the first and third quartiles, respectively. The upper and lower whiskers extend to the largest/smallest value no further than 1.5 times the interquartile range from the hinges. Outliers beyond whiskers are displayed as individual points. (c) Strand-concordance and variance-to-mean ratio applied for extraction of informative mutations for each donor. (d) Allele frequency of representative somatic mtDNA mutations projected onto UMAP embedding for donor P1.

Extended Data Fig. 9 Adaptive clonotypes are associated to unique epigenetic identities.

(A) Clonotypes defined by clustering on per-cell allele frequency for all high-confidence variants. (b) Allele frequency of representative somatic mtDNA mutation within the conventional compartment of donor P2. (c) Association of clonotypes to clusters defined by chromatin accessibility for all NK cells from HCMV⁻ donors (n = 3). (d-e) Representative open chromatin regions specifically associated to individual clonotypes.

Extended Data Fig. 10 Adaptive NK cell clonotypes are stably maintained over time.

(a) UMAP embedding of NK cells from HCMV⁺ donor P3 and P4 analyzed by scATAC-seq at two different time points. (c-f) Stability of overall (C-D) and representative (E-F) subcluster-defining differentially accessible regions (DARs) over time; column-scaled. (g-h) Representative clonotype-defining mutations projected onto UMAP embeddings at the two time points. (i + k) Clonotype frequency of adaptive NK cell clonotypes within total adaptive NK cell compartment over time; Fisher’s exact test with Monte Carlo simulation. (j + l) Observed and permuted distribution of clonotype log₂ fold-changes between time points; two-sided Kolmogorov-Smirnov test.

Supplementary information

Supplementary Information

Supplementary Tables 1 and 2.

Reporting Summary

Source data

Source Data Extended Data Fig. 1

Statistical source data.

Source Data Extended Data Fig. 5

Statistical source data.

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 license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license 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 license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Rückert, T., Lareau, C.A., Mashreghi, MF. et al. Clonal expansion and epigenetic inheritance of long-lasting NK cell memory. Nat Immunol 23, 1551–1563 (2022). https://doi.org/10.1038/s41590-022-01327-7

Download citation

Received: 28 July 2022
Accepted: 09 September 2022
Published: 26 October 2022
Issue Date: November 2022
DOI: https://doi.org/10.1038/s41590-022-01327-7

This article is cited by

Memory B cell subsets have divergent developmental origins that are coupled to distinct imprinted epigenetic states
- Derrick Callahan
- Shuchi Smita
- Mark Shlomchik
Nature Immunology (2024)
Natural killer cell therapies
- Eric Vivier
- Lucas Rebuffet
- Valeria R. Fantin
Nature (2024)
Expansion of memory Vδ2 T cells following SARS-CoV-2 vaccination revealed by temporal single-cell transcriptomics
- Sara Terzoli
- Paolo Marzano
- Domenico Mavilio
npj Vaccines (2024)
Gene-expression memory-based prediction of cell lineages from scRNA-seq datasets
- A. S. Eisele
- M. Tarbier
- D. M. Suter
Nature Communications (2024)
Control of nutrient uptake by IRF4 orchestrates innate immune memory
- Endi K. Santosa
- Hyunu Kim
- Joseph C. Sun
Nature Immunology (2023)

Subjects

Abstract

Similar content being viewed by others

Main

Results

Mapping NK cell subsets onto transcriptional and epigenetic landscapes

Distinct signatures for naive and adaptive NKG2C+ NK cells

HCMV leaves an inflammatory memory footprint enriched in AP1 motifs

Synergistic imprinting by HCMV peptides and proinflammatory cytokines

Convergent and divergent epigenetic features of adaptive NK cells

Clonal expansion underlies divergent signatures of adaptive NK cells

Adaptive NK cell clonotypes are stably maintained over time

Discussion

Methods

Human studies

Cell lines

Cell isolation

HCMV serology

Flow cytometry

KIR ligand genotyping

Cell preparation for scATAC-seq

scATAC-seq

Cell preparation for scRNA-seq

scRNA-seq

Stimulation of NK cells from HCMV– donors

Removal of sensitive genetic variant information

Preprocessing of scATAC-seq data

Integrated analysis of ex vivo scATAC-seq data

Motif analysis

Mitochondrial genotyping and clonotypes

Donor-specific and donor-unique signatures

Cluster heterogeneity

Longitudinal analysis

Analysis of scATAC-seq data after activation

Preprocessing of scRNA-seq data

Downstream analysis of scRNA-seq data

Integration of scRNA-seq and scATAC-seq data

Statistical analysis and reproducibility

Reporting summary

Data availability

Code availability

References

Acknowledgements

Funding

Author information

Authors and Affiliations

Contributions

Corresponding authors

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Quick links

Distinct signatures for naive and adaptive NKG2C⁺ NK cells

Stimulation of NK cells from HCMV^– donors