Lung cancer remains the leading cause of cancer related deaths, and despite the transformative effect of immunotherapy for non-small cell lung cancer (NSCLC), most patients develop acquired resistance to these treatments [1]. Immune checkpoint blockade that activates CD8 + T cells is the standard of care for patients with metastatic non-small cell lung cancer (NSCLC) lacking actionable genomic alterations (such as EGFR and ALK alterations) [1, 2]. Long-term survival in patients with metastatic NSCLC approximates 30% in subsets of patients responsive to anti PD1/PD-L1 inhibitors, however most patients do not derive long term benefit [1]. Multiple mechanisms underlie this phenomenon, and promising approaches include activating additional immune cell subsets such as Natural killer (NK) cells. NK cells are part of the innate immune system and exert crucial functions in anti-tumor immunity through both direct tumor cell lysis and amplification of immunomodulatory signaling to T cells and antigen presenting cells. Therapies based on augmentation of NK cell functions are a new frontier in immunotherapy [3]. In this review, we consolidate studies on NK cell biology, NK cell (dys)function in anti-cancer immunity, opportunities for therapeutic targeting, and the current state of clinical investigation into NK cell-based therapies specifically relevant to lung cancer.

NK cell biology

Natural killer (NK) cells are part of the innate immune system responsible for surveillance and clearance of non-self cells. NK cell cytotoxic activity is tightly regulated by a balance of activating and inhibitory receptors, in order to limit uncontrolled NK cell killing. Expression of activating receptor ligands are frequently induced by cell stressors, such as malignant transformation and virally-infected cells. A comprehensive summary of NK cell inhibitory and activating ligands was summarized by Vivier, et al. [4].

Classically, NK cells have been studied in the human peripheral blood where they have been phenotypically characterized as CD3-CD56dimCD16 + cells. To date, two functionally and phenotypically distinct subsets of NK cells are classified by the differential expression of CD56 and CD16 on the cell surface; CD56dimCD16 + and CD56brightCD16dim/- [5, 6]. The CD56dimCD16 + are highly cytotoxic and comprise 90–95% of the NK cells found in peripheral blood, lymph nodes, spleen [5]. High expression of CD16, the low-affinity Fc receptor, allows for antibody-dependent cellular cytotoxicity (ADCC), whereas low CD16 expression abolishes ADCC [6]. CD56dimCD16 + cells express inhibitory killer cell immunoglobulin-like receptors (KIR), activating receptors including NKG2D, DNAM1, and the natural cytotoxicity receptors NKp44, NKp30 and NKp46 [4]. Cytolytic NK cells can induce target cell death through directed release of lytic granules or by inducing death receptor mediated apoptosis via the expression of Fas ligand or TNF-related apoptosis inducing ligand (TRAIL). NK cell recognition of a target cell activates the polarized exocytosis of the lytic granules, specialized organelles containing pore-forming molecules perforin and serine protease granzymes, that aid target cell death. While granule mediated cytotoxicity is fast, death receptor mediated apoptosis requires more time [7]. Rebuffet, et al., utilized a combination of single cell RNAseq and cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) to add a third group of peripheral blood NK cells, which encompasses adaptive NKG2C + NK cells [8]. This population shows high expression of NKG2C and intermediate expression of cytotoxic markers, and was also identified in multiple tumor tissues including lung, colon cancer, head and neck squamous cell carcinoma, nasopharyngeal carcinoma, and prostate adenocarcinoma [8]. This population can be expanded in response to acute cytomegalovirus infection [9]. However, their role in anti-tumor immunity remains largely unexplored.

CD56bright CD16dim/- NK cells are abundant cytokine producers, less cytotoxic, and are predominant in secondary lymphoid tissues. This is likely because this subset expresses high levels of CCR7 and L-selectin, both required for cellular trafficking [6, 10]. Tissue NK cells express cell surface receptors (such as CD49a, CD69, and CD103) unique to each tissue, and exert different functions specific to their microenvironments and distinct from peripheral NK cells [5, 11]. Tissue-resident NK cells are weakly cytotoxic, but they produce several immunoregulatory cytokines, such as interferon-γ (IFN-γ), tumor necrosis factor-α (TNF- α), granulocyte–macrophage colony-stimulating factor (G-MCSF) [5]. In NSCLC, activation of tissue-resident NK cells were reported to have an exhausted phenotype (CD69 + CXCR6+), highlighting their noncytotoxic function resulting in poor anti-tumor immunity [12].

The development of NK cells primarily occurs in the bone marrow but further differentiates in tissues, and NK cell function is associated with tissue distribution and localization [5, 6]. Besides established cytolytic or tissue-resident NK cells, evidence suggests other subtypes of NK cells may exist, each defined by different surface molecules and functions [13, 14]. For example, a subset of HLA-DR + NK cells have been described to have weak antigen-presenting ability in vitro [15]. Similar functional cells have been observed in mice, and HLA-DR + NK cells were also identified in NSCLC patient tumors [15]. The “helper” NK (NKh) cell differentiation pathway is a subset of CD56 + NK cells that could be induced to express CD83+, CCR7+, and CD25+, common surface markers on mature dendritic cells (DCs) [16]. IL-2 activated NK cells promote the differentiation of mature type-1 polarized DCs (DC1s), these cells display high anti-melanoma activity as they are capable of inducing Th1 and CTL responses in T-cells [17]. This effect is contact dependent, but the secretion of IFN-γ and TNF-α from NK cells contributes to DC maturation as well [18]. A regulatory function of NK cells has been described in mouse models with relevance in controlling autoimmunity, and suggested the concept of regulatory NK cells (NKregs, akin to immunosuppressive Tregs) [6, 19, 20]. For example, DC-activated NK cells exhibited the ability to kill immature DCs in vitro, and postulated to serve as a negative feedback loop [21]. In NSCLC, NK cells isolated from a syngeneic Lewis lung carcinoma (LLC) model secreted CCL22 that recruited Tregs into the tumor [22]. Picard, et al., demonstrated a subtype of CD56dim/CD16- NK cells circulating in blood from patients with NSCLC that produced regulatory cytokines [23]. While there are many studies describing these subtypes, several of them were performed in murine models and thus additional investigation of patient tumors will be necessary to determine clinical relevance. Furthermore, it is unknown whether these cells represent unique NK subsets with distinct in vivo functions, or whether inducible expression of markers indicate transient differentiation states of tissue-resident NK cells. Taken together, NK cell function is context dependent and consideration should be given to NK cell tissue localization and subset heterogeneity in the analysis of NSCLC tumors.

As aforementioned, NK cells recognize target cells through activation and inhibitory receptors. NK cell activating receptors include NKG2D, whose ligands (NKG2D-L) can frequently be detected on human cancer cells [24]. NKG2D expression is not restricted to NK cells but has also been detected on subsets of T cells, including recently activated CD8 + T cells, γδ T cells and iNKT cells [25]. NKG2D ligands include MHC Class I chain A and B (MICA and MICB), and UL16-binding proteins [4]. Expression of these ligands in a panel of NSCLC cell lines and one patient-derived model was highly heterogeneous [26]. MICA expression was the most consistent, with 4/6 cell lines showing expression above 90%. MICB expression ranged from 1 to 95%, ULBP1 0–6% in 5 cell lines but one line showing expression in 99% of cells. ULBP2 expression ranged between 0% and 98%, and ULBP3 between 11% and 99%. Out of the 6 cell lines studied, H1155 consistently had expression of these NKG2D ligands above 90%, whereas the other cell lines had variable expression of NKG2D ligands [26].

NKG2A is an inhibitory receptor that forms a heterodimer with CD94 and is found on about 50% of peripheral NK cells and on a subset of CD8 + T cells [27]. In several tumor types such as melanoma, breast, and liver cancers, NKG2A was enriched in the CD56bright NK population [27]. HLA-E binding to NKG2A/CD94 signals through immune receptor tyrosine-based inhibitory motifs (ITIM) that result in NK and T cell inactivation, and physiologically this prevents development of autoimmune diseases [27]. CRISPR-knockout of NKG2A in primary human NK cells enhanced killing of a B lymphoblastoid cell line, however also decreased NK cell expansion [28]. They surmised that through inhibiting proliferative activity of NK cells and activation-induced cell death, and thus while NKG2A does serve as a checkpoint, a downstream consequence could be preserving the expansion ability of NK cells [28]. NK cells in many solid tumors including breast, lung, and liver cancers show increased expression of NKG2A, and likewise HLA-E is frequently increased in melanoma and cancers of the lung, kidney, liver, prostate, and colon [27, 29].

The leukocyte Ig-like receptor (LIR, or LILR) family is expressed on multiple lymphocytes including NK, T, and B cells, and antigen-presenting cells including macrophages and dendritic cells [30]. This family contains receptors that activate (LILRA1-6) and inhibit (LILRB1-5) leukocyte function. Similar to NKG2A, inhibitory signals are transduced through ITIMs upon ligation by HLA class I molecules [30]. LILRB1 was identified in circulating CD56dim NK cells from patients with prostate cancer and multiple myeloma, and LILRB1 expression was upregulated compared to healthy donors [31]. HLA-G expression in melanoma cells signaling through ILT2 (LILRB1) prevented NK cell cytolysis in vitro [32]. In cancer, NK cells can also be induced to express inhibitory checkpoints such as PD-1, TIGIT, TIM-3, and LAG3. These are discussed in more detail in the section “Immune checkpoints on NK cells”. Taken together, NK cell cytotoxic function is tightly regulated by the balanced ligation of inhibitory and activating receptors, which can be dysregulated in solid tumors preventing NK cell cytotoxicity.

NK cell (dys)function in anti-cancer immunity and opportunities for therapeutic interventionsAssociation of NK cell tumor infiltration with clinical outcome

NK cells have been shown to potently eliminate tumors in various orthotopic and spontaneous models of cancer in mice (including melanoma, lymphoma, sarcoma, and colon cancer), highlighting the potency of NK cells in anti-tumor immunity [33,34,35]. Several studies in human cancer patients found that NK cell infiltration and NK cell activity predicts favorable clinical outcome in a number of malignancies, including melanoma, gastric cancer, and head and neck cancer [36,37,38]. The positive prognostic effect of NK cells in metastatic melanoma was further enhanced with concomitant expression of IL15 [37].

NK cell signatures also affect prognosis in lung cancer. A seven-gene signature of NK cells was predictive of response to immunotherapy and favorable prognosis in lung adenocarcinoma [39]. Another study identified a FLT3 (FMS-related tyrosine kinase induced in dendritic cells) axis in the NSCLC Cancer Genome Atlas (TCGA) database. High FLT3 expression correlated with a gene signature indicative of high immune infiltration (particularly of NK cells and DCs), increased expression of cGAS-STING effectors, and increased disease-free survival in both squamous and adenocarcinoma histologies [40]. Presence of NK cells in primary lung tumors have been correlated with improved patient survival in NSCLC [41]. Furthermore, depletion of NK cells reduced tumor clearance and enhanced metastasis in several murine models of lung cancer [7, 42].

Meta-analysis of TCGA datasets showed that several NK cell receptors are associated with increased overall survival across different tumor types, suggesting that NK cells are indeed beneficial for tumor control [43]. However, there was a strong correlation between CD8 + T cell and NK cell infiltration, indicating that T cell immunity, NK cell immunity, or a combination of both may be responsible for improved overall survival in these cases. Moreover, NK cells stimulated recruitment of cDC1 into melanoma tumors, promoting cancer immune control, and highlighting the immunomodulatory role of NK cells [41]. Taken together, there is strong evidence that NK cells promote anti-tumor immunity in multiple malignancies, which may be the result of direct elimination of tumor cells by NK cells, improved T cell responses, and/or immunomodulatory function of NK cells within tumor microenvironments (TME).

Certain subsets of NK cells have been associated with favorable responses. The CD57 + tumor infiltrating NK subset (detected by IHC) was correlated with survival after curative intent resection in squamous cell lung carcinoma [44]. Recent studies using multiparameter flow cytometry of human tissues showed that the proportion of NK cells within primary lung tumors were lower than in adjacent normal lung. NK cells composed about 4.5% in primary lung tumors, compared to 47% of T cells [45]. Interestingly, the CD16-negative NK population was similar between tumor and non-tumor sections, whereas CD16 + NK cells were decreased in tumor sections relative to distal lung [45]. Another study of NK cells in lung cancer demonstrated that the majority of tumor-infiltrating NKs lacked CD16 expression, and observed that the CD69 + tissue-resident NK cells (trNK) subset was enriched in the tumor center vs. distal lung in both adenocarcinoma and squamous lung histologies [46]. Interestingly, expression of checkpoint inhibitors TIM3 and TIGIT (but not PD-1) on trNK cells increased closer to the tumor center. trNK’s enriched in tumor centers had greater enrichment of CXCR6, lower expression of CCR2, and expressed Granzyme A and Granzyme B, but not perforin [46]. In another study including 30 NSCLC patients, NK cells ranged between 1.7% and 34.4% of tumor infiltrating lymphocytes, most being CD56dim, and associated with downregulation of activating receptors such as CD16 and NKG2D [47]. Future studies leveraging spatial transcriptomics and/or multiplexed immunofluorescence would be valuable to better understand the topography of NSCLC tumors including stromal and peri-vascular niches within the tumor.

Understanding how NK cells may evolve over the course of NSCLC treatment could lead to discovery of new vulnerabilities. Only a few studies have described how NSCLC therapies affect NK cells. One study evaluated NK cell precursors in peripheral blood at baseline and after one cycle of standard treatment in a cohort of 18 patients with metastatic lung cancer [48]. 10 patients received chemo-immunotherapy, 7 patients received immunotherapy alone, and 1 patient received Osimertinib (indicating presence of an activating EGFR mutation, of note these patients are not standardly treated with immunotherapy). They identified a 5-fold increase in Lin-CD34 + DNAM-1bright NK cell precursors, but not total circulating CD34 + precursor cells after just one cycle, suggesting a selective induction mobilization of inflammatory precursors. They also showed that the precursor cells in the primary tissue preferentially expressed CXCR4 compared to peripheral blood. Another study compared the effect of two standard NSCLC drugs gemcitabine (chemotherapy), and gefitinib (an EGFR TKI), on NKG2D ligand expression and NK cell activity [49]. The NKG2D ligands examined in this study included MHC Class I chain related genes A and B (MICA and MICB) and UL16-binding proteins (ULBPs 2,5, and 6). Gemcitabine treatment upregulated NKG2D ligands and concomitantly promoted NK cell anti-tumor activity. Conversely, gefitinib downregulated NKG2D expression and decreased NK-mediated cell lysis. Not all cytotoxic agents, including docetaxel, pemetrexed, or vinorelbine, were able to modulate the NKG2D axis. This suggests that some but not all chemotherapies can augment NK cell anti-tumor activity. Taken together, these data suggest that understanding the kinetics of NK cell activity after standard chemotherapies could identify optimal sequencing of NK cell augmentation therapies in combination approaches, and that each chemotherapy (or targeted therapy) may have distinct effects on NK cells.

NK cells in immunosuppressive tumor microenvironments

NK cells isolated from lung tumors exhibited limited cytotoxic activity [47, 50], indicating that NK cell function can be dampened within tumor microenvironments. The following sections discuss potential mechanisms of suppression within TMEs, with particular note to bone and brain metastases given the high frequency and devastating clinical complications of these metastatic sites in lung cancer. Figure 1 depicts common mechanisms of NK cell suppression by tumors.

To dissect potential mechanisms of NK cell dysfunction, Platonova, et al., examined NK cell receptor expression in primary human NSCLC samples compared to matched peripheral blood NK cells and non-tumoral distant tissue [47]. Overall, activating receptors NKp30, NKp80, CD16, NKG2D, and DNAM-1 was significantly downregulated compared to matched peripheral blood NK cells or NK cells from healthy donors. In contrast, NK cells in these lung tumors expressed higher levels of the inhibitory receptor NKG2A [47]. In vitro co-cultures of human peripheral blood NK cells with the A549 NSCLC cell line resulted in downregulation of the activating receptors NKp30, NKp80, and DNAM-1 in the NK cells [47]. Similarly, the inhibitory checkpoint NKG2A was upregulated in an in vivo metastasis model of NSCLC [51]. Evidence that tumor cells downregulate cytolytic functions of NK cells by tumors was demonstrated in vivo, in a syngeneic model of colon adenocarcinoma that employed temporal photoconversion of NK cells to track tumor-infiltrating NK cells [52]. Dean, et al., showed that peripheral NK cells infiltrating into tumors became converted into a tissue-resident phenotype lacking cytolytic function [52]. These dysfunctional NK cells expressed CD49a, increased expression of inhibitory checkpoints, and exhibited decreased chemokine and cytokine production. However their function was improved by administration of an IL-15:IL-15Ra complex [52], suggesting that this NK cell constraint has the potential to be reversed.

Similar to the adaptive anti-tumor T cell immunity, tumor-infiltrating NK cells are susceptible to immune modulating factors within the tumor microenvironment (Reviewed in [53, 54]. For example, extracellular adenosine and prostaglandins suppress anti-tumor NK cell responses [55, 56]. In mouse models of melanoma and sarcoma, tumors produced suppressive cytokines such as TGF-beta that modulated NK cells to a ILC1 phenotype with decreased anti-tumor function [57, 58]. Corresponding with this observation, TGF-beta downregulates several activating receptors in NK cells in the context of NSCLC and breast cancer [47, 59, 60], and plasma levels of TGF-beta inversely correlated with NKG2D expression in patients with lung or colon cancer [61]. Inhibition of TGF-beta using small molecule inhibitors or neutralizing antibodies reversed NK suppression in NSCLC and breast cancer in vitro [47, 59, 60], in ovarian cancer in vivo [62], and increased in vivo NK cell frequency in oral squamous cell carcinoma [63]. Engineered NK cells resistant to TGF-beta (via Smad4 knockout or expression of a dominant negative TGF-beta receptor) led to improved anti-tumor activity against breast cancer, colon cancer, and neuroblastoma [64, 65].

Many solid tumors are hypoxic, which causes mitochondrial fragmentation in tumor infiltrating NK cells, resulting in decreased NK cell function [66]. Furthermore, hypoxia signals through HIF1α to inhibit IL18-signaling in NK cells, reducing NK cell anti-tumor activity [67]. In murine breast cancer models, senescent cancer-associated fibroblasts secreting extracellular matrix dampened NK cell activity [68]. In melanoma, tumor infiltrating NK cells expressed higher levels of the KLRC1 gene (encoding the inhibitory receptor NKG2A) than NK cells in peripheral blood [69]. Thus, NK cell function can be suppressed within TMEs through several mechanisms.

Single cell RNA sequencing (scRNAseq) of early-stage lung cancers after tumor resection showed imbalances in immune cell subsets including NK cells, T cells, and myeloid cells even in stage I NSCLC, suggesting early changes in tumor microenvironment [70]. Of note, 4 out of the 6 tumors analyzed in this manuscript harbored an EGFR activating mutation (3 L858R and 1 exon19 deletion) [70]. EGFR-mutated NSCLC has very limited response to current checkpoint inhibitors [