Human peripheral blood DNAM-1^neg NK cells are a terminally differentiated subset with limited effector functions

Natural killer (NK) cells are innate lymphoid cells that play important roles in the elimination of malignant or virally infected cells.¹  In addition to their cytotoxic functions, NK cells secrete cytokines and chemokines that contribute to the development of adaptive immune responses. NK cell activation is controlled by an array of receptors. Human NK cells express killer-cell immunoglobulin-like receptors (KIRs) that transmit negative signals upon binding to class I molecules of the major histocompatibility complex.²  Conversely, natural cytotoxicity receptors (NKp30, NKp44, and NKp46) and NKG2D sense stress-induced molecules and contribute to NK cell activation. NK cell activity results from the integration of signals provided by these receptors as well as coreceptors, adhesion molecules, and cytokines.

Human NK cells are usually defined as CD3^negCD56^pos cells and are divided into 2 main subsets: CD56^bright and CD56^dim NK cells.³  In the peripheral blood, ∼90% of NK cells are CD56^dim, but CD56^bright NK cells predominate in lymph nodes and most tissues.⁴  CD56^dim NK cells express high levels of the Fcγ receptor III (CD16), whereas CD56^bright NK cells are CD16^dim/neg. Furthermore, CD56^bright cells are CD62L^pos and express high levels of the inhibitory receptor CD94/NKG2A but are KIR^neg, whereas CD56^dim NK cells are CD62L^neg, CD94/NKG2A^low, and KIR^high.⁵  There is convincing evidence that CD56^bright NK cells are immature and differentiate into CD56^dim NK cells under cytokine stimulation.^6,7  CD56^bright NK cells are generally seen as an immunoregulatory subset characterized by its high production of cytokines such interferon-γ (IFN-γ), granulocyte-macrophage colony stimulating factor (GM-CSF), and tumor necrosis factor–α, whereas CD56^dim NK cells are endowed with high cytotoxic potential.³  Although NK cell diversity expands far beyond CD56^dim and CD56^bright subsets,^8-10  the functional impact of this vast phenotypic diversity and the progeny relationship between NK cell subsets remain elusive.

DNAX accessory molecule-1 (DNAM-1, CD226) is an adhesion and costimulatory molecule known to promote NK cell cytotoxic activity and IFN-γ production upon binding to its ligands CD112 and CD155.¹¹  DNAM-1 shares its ligands with the inhibitory molecules TIGIT and CD96.¹²  DNAM-1 has been involved in NK cell–mediated tumor immunosurveillance,^13-16  control of metastatic disease,¹⁷  and defense against pathogens.^18,19  Importantly, DNAM-1 expression identifies 2 distinct NK cell functional subsets in mice, with DNAM-1^neg NK cells arising from DNAM-1^pos NK cells.²⁰  Although mouse DNAM-1^pos NK cells produce high levels of inflammatory cytokines, have enhanced interleukin-15 (IL-15) signaling, and proliferate vigorously, their DNAM-1^neg counterpart produces high amounts of macrophage inflammatory protein-1 (MIP-1) chemokines. In humans, DNAM-1 is expressed on the majority of peripheral blood NK cells, but bimodal DNAM-1 expression has been observed on lymphoid tissue–resident NK cells.²¹  Several studies reported a reduction in DNAM-1 expression on NK cells isolated from patients with cancer.^22-26  Nevertheless, whether DNAM-1^neg human NK cells represent a distinct NK cell subset with specific functions has not yet been studied.

In the current study, we established that human peripheral blood DNAM-1^neg NK cells represent a distinct NK cell subset with a specific gene expression program. We showed that DNAM-1^neg NK cells arise from DNAM-1^pos NK cells and have limited effector functions. Moreover, proportions of DNAM-1^neg NK cells are increased in the blood of patients with lymphoma, highlighting the necessity of considering this subset when designing immunotherapies.

This study was approved by the Human Research Ethics committee at QIMR Berghofer and the Princess Alexandra Hospital, Brisbane. Buffy coat blood packs were collected from healthy volunteer blood donors through the Australian Red Cross Blood Service (Queensland, Australia). Samples from patients with Hodgkin lymphoma (HL) and diffuse large B-cell leukemia (DLBCL) and from age- and sex-matched healthy donors were collected pretreatment at the Princess Alexandra Hospital, Brisbane. All samples used in this study have been supplied with written informed consent.

NK cells were enriched from cryopreserved peripheral blood mononuclear cells (PBMCs) by negative selection using an automated separator (Miltneyi Biotec). NK cell subsets were purified by using flow cytometry (FACSAria II; BD Biosciences). Cells were cultured in complete RPMI 1640. The supplemental Methods present details regarding the NK cell functional assays.

Cells were stained according to standard protocols (supplemental Table 1). Zombie yellow (BioLegend) was used to gate out dead cells. Isotype or fluorescence-minus-one controls were used to set positive staining gates. Intracellular staining was performed by using the BD intracellular staining kit (BD Biosciences). Cell apoptosis was measured by using the BD Annexin V Apoptosis Detection Kit. Data were acquired by using the LSR Fortessa IV (BD Biosciences) and analyzed with FlowJo software (Tree Star). Cytokines in the supernatant of in vitro cultures were analyzed by using cytometric bead array technology according to the manufacturer’s instructions (BD Biosciences).

NK cell motility and NK/target cell interactions were visualized by time-lapse microscopy with an Xcellence IX81 microscope (Olympus Corporation). Live cell images were taken at 40-second intervals over a 40- to 90-minute time period. Imaris version 8.2 software (Bitplane) was used to track live cells and generate motility data. Trajectory graphs were generated in Microsoft Excel using displacement data.

Details regarding RNA sequencing (RNAseq) are given in the supplemental Methods.

The length of telomeric repeats was determined by quantitative flow–fluorescence in situ hybridization by using a telomere-specific, fluorescently labeled peptide nucleic acid probe according to the manufacturer’s instructions (DakoCytomation).

Statistical analysis was performed by using Prism 6 software (GraphPad). Data were compared by using either 1-way or 2-way analysis of variance (ANOVA) with Tukey multiple comparison post hoc testing. P < .05 was considered significant. P values are depicted as *P < .05, **P < .01, ***P < .001, and ****P < .0001.

To determine whether, similarly to mice,²⁰  distinct DNAM-1^pos and DNAM-1^neg NK populations exist in humans, we analyzed DNAM-1 expression on peripheral blood NK cells (CD3^negCD56^pos) from 13 healthy donors. In agreement with previous reports,^25,27  most NK cells expressed DNAM-1 (Figure 1A; supplemental Figure 1A). However, although CD56^bright NK cells were exclusively DNAM-1^pos, a small population of DNAM-1^neg NK cells was detected within the CD56^dim subset. These CD56^dimDNAM-1^neg cells represented 5.6% ± 3.2% of the total peripheral blood NK cells (Figure 1B) and were detected in both fresh and cryopreserved samples (supplemental Figure 1B-C). The observations that CD56^dimDNAM-1^neg cells expressed both the NK cell–associated transcription factor Eomes and the NK cell linage marker CD122 and were negative for CD14 and CD19 confirmed their NK cell identity (supplemental Figure 1D-E). Hence, 3 human NK cell subsets were defined: CD56^brightDNAM-1^pos, CD56^dimDNAM-1^pos, and CD56^dimDNAM-1^neg NK cells. In agreement with previous reports,³  CD56^brightDNAM-1^pos NK cells expressed low levels of CD16, whereas CD56^dimDNAM-1^pos NK cells were mostly CD16^pos. Intriguingly, CD16 expression was bimodal on CD56^dimDNAM-1^neg NK cells (Figure 1C).

Overall, CD56^dimDNAM-1^neg NK cells appeared very similar to CD56^dimDNAM-1^pos NK cells: compared with CD56^bright NK cells, both CD56^dim subsets exhibited decreased expression levels of CD27, CD69, CD94, NKG2D, NKG2A, and natural cytotoxicity receptors (NKp30 and NKp46) (Figure 1D). CD96 and TIGIT also harbored similar expression patterns on DNAM-1^pos and DNAM-1^neg CD56^dim NK cells and were, respectively, decreased and increased compared with CD56^bright NK cells. However, expression levels of inhibitory KIRs were equivalent between CD56^brightDNAM-1^pos and CD56^dimDNAM-1^neg NK cells and were reduced compared with their CD56^dimDNAM-1^pos counterpart. Moreover, the maturation marker CD57 was expressed by one-half of CD56^dimDNAM-1^pos NK cells but only 20% to 30% of CD56^dimDNAM-1^neg NK cells. Finally, CD56^dimDNAM-1^neg NK cells expressed very low levels of the DNAM-1–associated adhesion molecule LFA-1.²⁸  These data identify a new CD56^dimDNAM-1^neg NK cell population that differs phenotypically from CD56^dimDNAM-1^pos NK cells by exhibiting lower expression of CD57 and KIRs.

DNAM-1 is an adhesion molecule²⁷  and might influence NK cell motility. We therefore investigated the migratory behavior of cytokine-activated NK cell subsets. We observed that DNAM-1^pos NK cells were elongated compared with the majority of DNAM-1^neg NK cells that appeared spherically symmetrical (Figure 2A). Compared with DNAM-1–expressing subsets, CD56^dimDNAM-1^neg NK cells displayed minimal motility and tended to move by involuntary drift (Figure 2B; supplemental Figure 2A-B). CD56^brightDNAM-1^pos and CD56^dimDNAM-1^pos NK cells were significantly faster (Figure 2C) and traveled longer tracks (Figure 2D) than CD56^dimDNAM-1^neg cells, even though no significant difference was observed regarding the displacement from origin (Figure 2E). Thus, CD56^dimDNAM-1^neg NK cells present an altered migratory behavior in vitro.

We assessed the proliferation of sorted NK cell subsets after 5 days of culture in the presence of IL-2 and IL-15 (Figure 3A). We observed that CD56^bright NK cells proliferated more than CD56^dim NK cells (as shown by Romagnani et al⁷ ); within the CD56^dim fraction, CD56^dimDNAM-1^neg NK cells exhibited reduced proliferation. We next measured IFN-γ production by using purified human NK cell subsets following in vitro stimulation with IL-12, IL-15, and IL-18. In agreement with previously published data,³  higher percentages of IFN-γ–producing cells were observed within the CD56^bright NK cell subset compared with both CD56^dim subsets (Figure 3B). Interestingly, CD56^dimDNAM-1^pos NK cells were better IFN-γ producers than CD56^dimDNAM-1^neg NK cells. Further to this, we analyzed cytokine secretion in the culture supernatant of 6 purified NK cells subsets based on DNAM-1, CD56, and CD16 expression (gated as in supplemental Figure 3). CD56^brightCD16^negDNAM-1^pos NK cells secreted more IFN-γ and GM-CSF than the other subsets (Figure 3C). CD56^dimCD16^negDNAM-1^pos also secreted relatively high levels of IFN-γ and GM-CSF, in agreement with the idea that this population might represent an intermediate phenotype between CD56^brightCD16^negDNAM-1^pos and CD56^dimCD16^posDNAM-1^pos NK cells.²⁹  Importantly, both DNAM-1^neg subsets (CD56^dimCD16^negDNAM-1^neg and CD56^dimCD16^posDNAM-1^neg), similarly to CD56^dimCD16^posDNAM-1^pos NK cells, were poor producers of IFN-γ and GM-CSF in response to cytokine stimulation. Finally, CD16 expression seemed to determine NK cell ability to produce MIP chemokines. Indeed, within the 3 main NK subsets described in this study (CD56^brightDNAM-1^pos, CD56^dimDNAM-1^pos, and CD56^dimDNAM-1^neg), CD16^pos cells always secreted more MIP-1α and MIP-1β than their CD16^neg counterpart. Taken as a whole, these data show very limited proliferation and IFN-γ secretion of CD56^dimDNAM-1^neg NK cells in response to cytokine stimulation.

We investigated the response of DNAM-1^neg NK cells to receptor triggering through target cell stimulation. When incubated with the chronic myelogenous leukemia line K562, CD56^dimDNAM-1^neg NK cells expressed more CD107a than the other subsets (Figure 4A; supplemental Figure 4A). However, CD56^dimDNAM-1^pos NK cells killed K562 cells more effectively compared with both CD56^brightDNAM-1^pos cells and CD56^dimDNAM-1^neg NK cells (Figure 4B), in agreement with their high intracellular expression of granzyme B (supplemental Figure 4B). Analysis via time-lapse microscopy confirmed that CD56^dimDNAM-1^pos NK cells rapidly killed K562 cells, whereas CD56^brightDNAM-1^pos and CD56^dimDNAM-1^neg NK cells required a longer interaction time (Figure 4C). Thus, CD56^dimDNAM-1^neg NK cells respond to target cells, but this action does not translate to target cell killing.

We explored the possibility that this CD56^dimDNAM-1^neg population might have a regulatory function. To test this hypothesis, we performed a killing assay in which CD56^dimDNAM-1^pos NK cells were incubated with K562 target cells in the presence of conditioned medium (CM) from cytokine-stimulated CD56^dimDNAM-1^neg cultures (Figure 4D-E). The addition of CD56^dimDNAM-1^neg NK cell CM, but not heat-inactivated CM or CD56^bright NK cell CM, decreased CD56^dimDNAM-1^pos NK cell–killing capacity. These results suggest that soluble factors secreted by CD56^dimDNAM-1^neg NK cells may regulate the cytotoxic activity of CD56^dimDNAM-1^pos NK cells. Overall, our data indicate that the CD56^dimDNAM-1^neg NK cell subset presents limited effector functions but is endowed with immunomodulatory potential.

To better understand the relationship between CD56^brightDNAM-1^pos, CD56^dimDNAM-1^pos and CD56^dimDNAM-1^neg NK cell populations, we analyzed their gene expression profiles by using RNAseq. We confirmed that Ncam1 (CD56) messenger RNA (mRNA) was highly expressed within the CD56^bright subset compared with the 2 CD56^dim subsets (Figure 5A). Interestingly, DNAM-1 downregulation in CD56^dimDNAM-1^neg NK cells was detected at the mRNA level, indicating that the absence of DNAM-1 expression on human CD56^dimDNAM-1^neg NK cells is not a consequence of internalization following ligand binding but likely results from repression of the cd226 gene.

Unsupervised hierarchical clustering revealed that the 3 NK cell subsets clustered independently (Figure 5B), with both DNAM-1^pos subsets clustering together apart from CD56^dimDNAM-1^neg cells (Figure 5C). Analysis of the genes differentially expressed between the 3 NK cell subsets identified 6 groups of genes. In total, 2712 genes (groups B and F) were differentially expressed between DNAM-1^pos and DNAM-1^neg NK cells. Analysis of NK cell–related genes revealed high levels of Kit and Il7r mRNAs in CD56^brightDNAM-1^pos cells and high expression of gzmb (granzyme B) in CD56^dimDNAM-1^pos cells (Figure 5D), thereby confirming published observations.⁴  Interestingly, although KIRs and Tigit mRNA expression matched our cytometry data, cd96 mRNA was expressed on both DNAM-1^pos subsets, indicating that posttranscriptional mechanisms might regulate CD96 expression on the surface of CD56^dimDNAM-1^pos cells. Tbx21 mRNA (encoding T-bet) was downregulated in the CD56^dimDNAM-1^neg subset, but these cells still expressed high levels of Ifnγ mRNA. CD56^dimDNAM-1^neg NK cells also expressed high levels of Tnf mRNA.

We analyzed selected genes encoding immunosuppressive molecules along with receptors for inflammatory cytokines and confirmed the high expression of cytokine receptors in the CD56^brightDNAM-1^pos subset (Figure 5E). The CD56^dimDNAM-1^pos subset expressed high levels of mRNAs encoding for subunits of the receptor to transforming growth factor β, a cytokine known to suppress NK cell activity.³⁰  CD56^dimDNAM-1^neg NK cells expressed high levels of prostaglandin E synthase mRNA, suggesting that the synthesis of prostaglandin E2³¹  might contribute to the immunoregulatory activity of CD56^dimDNAM-1^neg NK cells.

Pathway network analyses confirmed the immaturity of CD56^bright cells, as pathways related to mitosis and Notch signaling³²  were preferentially expressed in this subset (supplemental Figure 5). Expression of pathways related to NK cell cytotoxicity, cell differentiation, receptor signaling, and intracellular signaling in the CD56^dimDNAM-1^pos subset suggested that these cells have been activated and have acquired killing capacities. CD56^dimDNAM-1^neg NK cells also appeared as a mature/activated subset that expressed pathways related to receptor/intracellular signaling. However, in contrast to CD56^dimDNAM-1^pos cells, pathways specific to this subset emphasized binding, either through cell adhesion or binding to immune compounds (immunoglobulins). Very interestingly, pathways related to inflammation (eg, inflammatory/immune response, cytokine/chemokine signaling, response to oxygen-containing compounds) were specifically expressed in CD56^dimDNAM-1^neg cells, suggesting that CD56^dimDNAM-1^neg NK cells might emerge in inflammatory conditions. Taken as a whole, these data established that CD56^brightDNAM-1^pos, CD56^dimDNAM-1^pos, and CD56^dimDNAM-1^neg NK cells represent 3 distinct NK cell subsets with different gene expression programs.

To assess the developmental relationship between CD56^brightDNAM-1^pos, CD56^dimDNAM-1^pos, and CD56^dimDNAM-1^neg NK cells, these subsets were purified from healthy donor PBMCs and cultured for 4 days in the presence of IL-2 only or a combination of IL-12, IL-15, and IL-18 (Figure 6A-B; supplemental Figure 6A). In both conditions, CD56^brightDNAM-1^pos NK cells gave rise to CD56^dimDNAM-1^pos NK cells and a minority of CD56^dimDNAM-1^neg NK cells. The appearance of CD56^brightDNAM-1^pos cells in CD56^dimDNAM-1^pos NK cell cultures in the presence of IL-2 might reflect CD56 upregulation after activation.³³  Importantly, a CD56^dimDNAM-1^neg NK cell population emerged in CD56^dimDNAM-1^pos NK cell cultures, suggesting that CD56^dimDNAM-1^neg NK cells derive from the CD56^dimDNAM-1^pos subset. Of note, the lower proliferative capacity of CD56^dimDNAM-1^neg NK cells excludes the possibility that a few contaminant CD56^dimDNAM-1^neg cells may have overgrown in these cultures. Finally, the vast majority of purified CD56^dimDNAM-1^neg NK cells conserved a stable phenotype, indicating that this subset represents a terminally differentiated stage that does not convert into any of the 2 other NK cell subsets. Of note, the pattern of expression of KIRs and CD57 in the 3 populations obtained postculture matched the phenotypes shown in Figure 1 (supplemental Figure 6B-C).

Telomere shortening is a process usually associated with cell replication, and terminally differentiated/memory lymphocyte subsets harbor shorter telomeres.⁷  Although no significant difference was observed, CD56^brightDNAM-1^pos cells had the longest telomeres, whereas CD56^dimDNAM-1^neg had reduced telomeres (Figure 6C). When analyzing NK cell apoptosis in 2 different assays, we found that CD56^dimDNAM^neg NK cells were more resistant than the other subsets to cytokine starvation (Figure 6D), whereas a high proportion of CD56^dimDNAM-1^neg NK cells underwent apoptosis when cultured with inflammatory cytokines (Figure 6E). Taken as a whole, these results support a linear maturation pathway in which CD56^brightDNAM-1^pos NK cells differentiate into CD56^dimDNAM-1^pos NK cells that finally lose DNAM-1 expression to become CD56^dimDNAM-1^neg NK cells.

We analyzed NK cell populations in the peripheral blood of a cohort of patients diagnosed with HL or DLBCL. Total NK cell percentages were reduced in the blood of patients with lymphoma compared with healthy donors (Figure 7A-B). This finding might be explained by a dramatic reduction of CD56^dimDNAM-1^pos NK cells in these patients (Figure 7C). In patients with DLBCL, decreased percentages in CD56^dimDNAM-1^pos NK cells among the whole NK cell population were associated with increased percentages in CD56^dimDNAM-1^neg NK cells; a similar trend, albeit not significant, was observed in patients with HL. Overall, the ratio of CD56^dimDNAM-1^pos NK cells over CD56^dimDNAM-1^neg NK cells was significantly reduced in both HL and DLBCL patients (Figure 7D). Marker analysis on the 3 NK cell populations revealed additional phenotypic alterations of NK cells in HL and DLBCL patients (supplemental Figure 7). An in vitro cytotoxicity assay against K562 target cells confirmed the limited killing capacity of CD56^brightDNAM-1^pos and CD56^dimDNAM-1^neg NK cells and indicated that the killing capacity of these 2 subsets was equivalent between healthy donors and patients with lymphoma (Figure 7E). Strikingly, the cytotoxic activity of CD56^dimDNAM-1^pos cells was decreased in patients with HL and DLBCL.

We have identified a minor population of terminally differentiated NK cells that are characterized by a lack of DNAM-1 expression. This study expands to humans our previous findings that DNAM-1 expression patterns distinguish 2 mouse NK cell subsets with specific functions and gene expression profiles.²⁰ 

The question of subset correspondence between mouse and human NK cells has recently been addressed by Crinier et al,¹⁰  who used single-cell RNAseq to identify 2 NK cell subsets (NK1 and NK2) conserved across organs and species. The direct comparison of NK1 and NK2 signatures with our gene expression data set confirmed the validity of our RNAseq analysis, with CD56^brightDNAM-1^pos and CD56^dimDNAM-1^pos NK cells overexpressing the NK2 and NK1 signatures, respectively (supplemental Figure 8A-B). Interestingly, there was no correspondence between CD56^dimDNAM-1^neg NK cells and any of the blood or spleen subsets defined by Crinier et al, indicating that CD56^dimDNAM-1^neg NK cells constitute a new NK cell subset. We suggest that, in both mouse and human, DNAM-1 expression might allow the definition of alternative NK cell subsets that do not fall into the NK1/NK2 categories. Using the published mouse DNAM-1^neg NK cell signature,²⁰  we have defined a 6-gene DNAM-1^neg NK cell signature conserved across species (supplemental Figure 8C).

In addition to a specific gene signature, human DNAM-1^neg NK cells share several functional characteristics with their mouse counterpart: they arise from DNAM-1^pos NK cells, they display limited proliferative capacity, and they are poor producers of IFN-γ and GM-CSF. However, this report also highlights differences between the 2 species. Although DNAM-1^neg NK cells account for approximately one-half of mouse splenic NK cells,²⁰  DNAM-1^neg NK cells represent a minor population in the human peripheral blood. In mice, DNAM-1^neg NK cells are high producers of MIP chemokines.²⁰  By contrast, we found that in humans, MIP-producing NK cells are defined by the expression of CD16, regardless of whether they express DNAM-1. Finally, mouse DNAM-1^pos and DNAM-1^neg NK cell subsets display equivalent cytotoxic activity,²⁰  whereas in humans, CD56^dimDNAM-1^pos NK cells are much more potent killers than CD56^dimDNAM-1^neg NK cells.

The consensus that conventional NK cells originate via a linear differentiation model has been challenged by some research suggesting that CD56^bright and CD56^dim NK cells might belong to different lineages.³⁴  Our in vitro differentiation experiments confirm the well-accepted linear model of NK cell maturation and add an additional step: immature CD56^bright NK cells give rise to CD56^dimDNAM-1^pos NK cells that ultimately become CD56^dimDNAM-1^neg NK cells. In agreement of the concept of a progressive maturation program, we identified a group of genes (Figure 5C, cluster E) showing high expression in CD56^bright NK cells, moderate expression in CD56^dim DNAM-1^pos cells, and low expression in CD56^dimDNAM-1^neg. Nevertheless, we did not completely rule out the possibility that some CD56^dimDNAM-1^neg NK cells might directly arise from CD56^bright NK cells. In this regard, the low percentage of KIR⁺ cells and the common gene expression program (Figure 5C, cluster D) shared by CD56^brightDNAM-1^pos and CD56^dimDNAM-1^neg NK cells might be seen as an indication of a differentiation pathway independent from CD56^dimDNAM-1^pos NK cells.

Our data indicate that CD56^dimDNAM-1^neg NK cells are terminally differentiated cells: their phenotype remains stable upon cytokine stimulation, their telomeres are shorter than those of DNAM-1^pos NK cell subsets, and they are highly prone to apoptosis in inflammatory microenvironments but more resistant to cytokine withdrawal. However, surprisingly, only 20% to 30% of CD56^dimDNAM-1^neg NK cells express CD57, a marker commonly believed to mark one of the final steps in NK cell maturation.^35-37  We propose that loss of DNAM-1 on both CD56^dimCD57^pos and CD56^dimCD57^neg NK cells represents an alternative maturation pathway that is associated with the acquisition of immunomodulatory functions. In this model, lower frequencies of CD57^pos cells within the DNAM-1^neg population might be explained by the reduced proliferative capacity of CD56^dimCD57^pos compared with CD56^dimCD57^neg cells.³⁵ 

Intriguingly, increased CD107a expression observed on the membrane of K562-stimulated CD56^dimDNAM-1^neg NK cells did not translate into effective cytotoxicity. Several hypotheses can explain these contradictory results. First, the granules released by CD56^dimDNAM-1^neg NK cells may not contain the cytotoxic molecules required for target cell killing, a hypothesis supported by the low expression of granzyme B of CD56^dimDNAM-1^neg NK cells. Alternatively, limited expression of LFA-1 on CD56^dimDNAM-1^neg cells coupled with the absence of DNAM-1–triggered inside-out LFA-1 signaling might have led to nonpolarized degranulation and limited killing of target cells.^38,39  Finally, similarly to what has been described for CD16,⁴⁰  NK cells might downregulate DNAM-1 after degranulation. In this case, higher percentage of CD107a⁺ cells within the CD56^dimDNAM-1^neg NK cell population might only be a consequence of their terminal differentiation from degranulating CD56^dimDNAM-1^pos NK cells.

Although CD56^dimDNAM-1^neg NK cells were limited in their ability to perform “classical” NK cell functions (ie, killing and IFN-γ secretion), our data indicate that these cells might represent an immunomodulatory subset. Although NK cells are known to play an immunoregulatory role in autoimmune diseases by the elimination of chronically activated immune cells,^41,42  this study is the first to indicate the self-regulation of NK cell cytotoxicity by the secretion of soluble factors from a terminally differentiated NK cell subset. The soluble factors implicated in this process remain to be identified. One main candidate is prostaglandin E2^31,43,44  as our RNAseq analysis identified increased expression of ptges in CD56^dimDNAM-1^neg NK cells.

We observed a profound decline in the ratio of CD56^dimDNAM-1^pos over CD56^dimDNAM-1^neg NK cell subsets in the blood of patients with HL and DLBCL that may be attributed to decreased numbers of CD56^dimDNAM-1^pos NK cells. In addition, we showed that, in patients with lymphoma, CD56^dimDNAM-1^pos NK cells present reduced cytotoxic activity. The decreased numbers and functionality of cytotoxic CD56^dimDNAM-1^pos NK cells associated with overrepresentation of potentially immunosuppressive CD56^dimDNAM-1^neg NK cells are likely to affect the efficacy of NK cell–targeting strategies in patients with lymphoma.

In conclusion, we identified a population of terminally differentiated CD56^dimDNAM-1^neg circulating human NK cells that are proportionally more represented in the blood of patients with HL and DLBCL. Because CD56^dimDNAM-1^neg NK cells are poor effectors and potentially immunosuppressive, this population is likely to affect the quality of ex vivo expanded NK cells for autologous cell transfer therapies. Future research should investigate whether depleting the CD56^dimDNAM-1^neg NK cells before culturing could improve the killing activity of patient NK cells and show better antilymphoma effect.

The authors thank the Flow Cytometry and Imaging Facility at QIMR Berghofer Medical Research Institute, the ACRF Centre for Comprehensive Biomedical Imaging, and the members of the Immunology in Cancer and Infection Laboratory for helpful suggestions and discussion. The authors acknowledge the specimen donors and thank the Red Cross Blood Service for their supply of buffy coats for this project.

C.G. was supported by a National Health and Medical Research Council (NHMRC) of Australia Early Career Fellowship (1107417), and M.J.S. was supported by an NHMRC Senior Principal Research Fellowship (1078671). This project was funded by a Priority-driven Collaborative Cancer Research Scheme project grant from Cancer Australia, Cure Cancer Australia, and Can Too (1122183) attributed to C.G., a project seed grant from QIMR Berghofer Medical Research Institute attributed to C.G., and an NHMRC Program Grant (1132519) attributed to M.J.S.

Contribution: K.A.S., M.K.G., M.J.S., and C.G. were responsible for study conception and design; K.A.S. and N.J.W. were responsible for acquisition of data; K.A.S., S.L., N.J.W., M.J.S., and C.G. performed analysis and interpretation of data; K.A.S. and C.G. drafted the manuscript; S.L., F.V., L.C., M.K.G., L.M., and M.J.S. performed critical revisions and editing; and M.J.S. and C.G. were responsible for supervision.

Conflict-of-interest disclosure: M.J.S. has research agreements with Bristol-Myers Squibb, Tizona Therapeutics, and Aduro Biotech. M.K.G. has financial relationships with BMS (research grant), Celgene (research grant), MSD (advisory board), Janssen (advisory board), Gilead (honorarium and research grant), Roche (honoria and travel), and Amgen (honorarium). The remaining authors declare no competing financial interests.

The current affiliation for C.G. is Cancer Immunotherapies Laboratory, Mater Research Institute, The University of Queensland, Translational Research Institute, Woolloongabba, QLD, Australia.

Correspondence: Camille Guillerey, Cancer Immunotherapies Laboratory, Mater Research Institute, The University of Queensland, Translational Research Institute, Woolloongabba, QLD 4102, Australia; e-mail: camille.guillerey@mater.uq.edu.au; or Mark J. Smyth, QIMR Berghofer Medical Research Institute, 300 Herston Rd, Herston, QLD 4006, Australia; e-mail: mark.smyth@qimrberghofer.edu.au