High Numbers of Circulating CD57⁺ NK Cells Associate with Resistance to HER2-Specific Therapeutic Antibodies in HER2⁺ Primary Breast Cancer

Natural killer (NK) cells are innate lymphocytes that can recognize antibody-coated tumor cells via the activating Fcγ receptor CD16A (FcγRIIIA), leading to tumor cell death by antibody-dependent cellular cytotoxicity (ADCC) and the release of immune cell recruiting chemokines and proinflammatory cytokines (1–5). In experimental breast cancer models, NK cells command antitumor adaptive immunity and contribute to the antitumor activity of HER2-specific antibodies (6–9). In patients, indirect evidence suggests that NK cells influence the efficacy of anti-HER2 therapeutic antibodies such as trastuzumab and pertuzumab, which are the standard of care in HER2⁺ breast cancer. An increase in tumor-infiltrating NK cells was reported after trastuzumab-docetaxel and T-DM1 (a trastuzumab–emtansine conjugate) treatment (10, 11). We have described an association between tumor-infiltrating NK-cell numbers in diagnostic biopsies and the achievement of pathologic complete responses (pCR) to HER2-specific antibody-based treatment in patients with HER2⁺ breast cancer (12). Despite this evidence supporting a role for NK-cell–mediated ADCC in the efficacy of anti-HER2 therapeutic antibodies, little is known about whether human NK-cell diversity influences NK-cell homing to and function in solid tumors.

In healthy adults, approximately 90% of NK cells in peripheral blood belong to the cytotoxic CD56^dimCD16⁺ subpopulation capable of developing ADCC responses, whereas approximately 10% constitute the CD56^brightCD16^−/low subset characterized by their cytokine production and preferential homing to secondary lymphoid organs (13). CD56^dimCD16⁺ NK cells include several subsets identified by distinct receptor combinations and differentiation status. Genetic (i.e., KIR and HLA haplotypes) and environmental factors, including infection by human cytomegalovirus (HCMV), respectively, shape the steady-state distribution of the NK-cell receptor repertoire and the development of circulating NK-cell subsets with adaptive features (14–17). HCMV-induced adaptive NK cells, identified by elevated expression of the activating receptor CD94/NKG2C, among other phenotypic and epigenetic features, display enhanced ADCC responses in several in vitro experimental models (18–20). Biological age is another factor associated with changes in the composition of the human NK-cell repertoire (21, 22). The CD57 epitope is a glycan carbohydrate (23) labeling differentiated NK cells, with variable prevalence in the adult CD56^dim CD16⁺ NK-cell repertoire (24, 25). CD57 expression is low in newborn NK cells, increases with age (26–28), and has been associated with chronic viral infections (29). CD57⁺ NK cells display lytic activity when stimulated through CD16 yet reduced proliferative capacity in response to cytokines (24, 30). A polymorphism in the gene encoding CD16A (FCGR3A), which results in two receptor variants with different affinities for IgG1, also modulates NK-cell–mediated ADCC responses (31).

We hypothesized that the configuration of the circulating NK-cell repertoire could influence the composition of tumor-infiltrating NK cells and hence, the efficacy of HER2-specific antibody-based treatment in breast cancer. We approached this question by an integrated analysis of: (i) the relationship between the phenotype of circulating NK cells, the CD16A 158V/F genotype, and efficacy outcomes in a prospective cohort of patients with HER2⁺ breast cancer with primary disease receiving neoadjuvant treatment with HER2-specific antibodies; (ii) the comparison of circulating and tumor-associated NK cells by IHC on FFPE tumor biopsies as well as by multiparametric flow cytometry in fresh ex vivo samples; and (iii) the evaluation of NK-cell function, proliferation, and survival in in vitro assays.

The conformation of the NK-cell receptor repertoire was analyzed in baseline peripheral blood samples from patients prospectively recruited between October 2013 and December 2016 at Hospital del Mar (Barcelona, Spain) and Hospital Clínic de Valencia (Valencia, Spain; prospective cohort n = 66). Cohort diagram and characteristics are summarized in Supplementary Fig. S1 and Table 1. Patients included newly diagnosed, previously untreated, primary breast cancer cases. HER2⁺ subtype classification was defined following 2013 American Society of Clinical Oncology/College of American Pathologists guidelines (32). All patients received a neoadjuvant combination therapy of standard chemotherapy and HER2-specific antibodies. Pathologic response at surgery after completing the neoadjuvant treatment was reported according to Miller–Payne grading system. Grades 1–4 are categorized as a partial pathologic response and grade 5 as pCR. The primary efficacy endpoint in our study was pCR defined as ypT0ypN0 [i.e., no malignant invasive cells identifiable in sections from the site of the tumor (T0) and axillary lymph nodes (N0); only vascular fibroelastotic stroma] based on histopathologic analysis of the resection specimen (33); disease-free survival (DFS) was calculated as the time from surgery until any breast cancer relapse or death by any cause. An extended T-cell immunophenotype was analyzed in a subset of 22 patients selected on the basis of their proportions of circulating CD57⁺ NK cells according to cut-off values providing the best Youden index for pCR: <65% CD57⁺ NK cells corresponding to CD57^low and ≥65% CD57⁺ NK cells to CD57^high.

The study was conducted following Declaration of Helsinki guidelines. All patients and healthy volunteers gave informed consent for the analysis of peripheral blood and tumor biopsies for research purposes and biomarker assessment. This study was approved by the Hospital del Mar Ethics Committee (2013/5307) and is reported according to the REMARK guidelines.

Blood samples obtained by venous puncture in EDTA tubes at diagnosis were fractionated to obtain a basic hemogram and viable peripheral blood mononuclear cells (PBMC), which were cryopreserved for later use. Phenotypic analyses of baseline PBMC samples were performed following standard flow cytometry protocols. Briefly, PBMCs were pretreated with aggregated human IgG, incubated with individual unlabeled specific mAbs, washed and further incubated with the secondary PE-Cy7–conjugated F(ab’)2 polyclonal goat anti-mouse IgG or IgM. Subsequently, samples were stained with combinations of directly labeled antibodies. Prior to intracellular staining with anti-FcRγ-FITC, samples were fixed and permeabilized (fixation/permeabilization kit, BD Biosciences).

Antibodies used include anti-CD45-AlexaFluor700 (clone 2D1), anti-CD56-APC (clone CMSSB), anti-CD16-APC-Cy7 (clone CB16), anti-CX3CR1-PE-Cy7 (clone 2A9-1) and anti-CXCR3-efluor660 (clone CEW33D) from eBiosciences; anti-CXCR1-PE (clone 8F1) from BioLegend; anti-CD3-PerCP (clone SK7), anti-CD45RA-FITC (clone L48), anti-CD4-allophycocyanin (APC; clone SK3), anti-CCR7-PE-Cy7 (clone 3D12), anti-CD8-HV500 (clone RPA-T8), anti-CD28-PE-CF594 (clone CD28.2) from BD Biosciences; anti-NKG2C- Phycoerythrin (PE; clone 134591) from R&D Systems, and anti-FcϵR γ subunit-FITC (polyclonal) from Merck Millipore. Anti-LILRB1 (clone HP-F1), anti-NKG2A (clone Z199) provided by Dr. A. Moretta, University of Genoa, Genoa, Italy; anti-CD57 (clone HNK1) and anti-KIR3DL1 (clone DX9) provided by Dr. L. Lanier, University of California, San Francisco (San Francisco, CA); anti-KIR2DL2/S2/L3 (clone CH-L) provided by Dr. S. Ferrini, National Institute for Cancer Research, Genoa, Italy; and anti-KIR2DL1 (clone DM-1) were produced in our laboratory and used either as cell culture supernatants or upon conjugation to fluorescein (FITC) or Pacific Blue (PB).

Data were acquired on an LSR Fortessa flow cytometer (BD Biosciences) and analyzed with FlowJo software (v10.0.7, Tree Star). All immunophenotypes were performed and analyzed blinded to the study endpoint. Expression of distinct markers was referred to CD56^dim NK cells. Circulating NK-cell absolute numbers were calculated on the basis of hemogram counts and expressed as cells/μL of blood.

For some patients, measurements of NK-cell degranulation (anti-CD107a-FITC, BD Biosciences) and intracellular TNFα (infliximab-Pacific Blue) were analyzed in baseline PBMC samples after 4-hour coculture with trastuzumab-coated SKBR3 cells (a human breast cancer cell line that overexpresses HER2) using standard flow cytometry protocols.

DNA was isolated from total blood using the Pure gene BloodCore kit B (Qiagen). Functional polymorphisms of FCGR3A, CD16A 158 V/F were determined using a PCR with confronting two-pair primers, as described previously (34). Because of the limited cohort size, influence of CD16A 158 V/F genotype has been analyzed by comparing those patients harboring the high-affinity allele (CD16 158V/V and V/F genotypes) to those homozygous for the low-affinity allele (CD16 158 F/F genotype).

Presence of CD57⁺ NK cells was evaluated in tumor-associated and circulating NK-cell paired samples from treatment-naïve breast cancer patients with primary disease undergoing surgical resection at Hospital del Mar (Barcelona, Spain; n = 6). For tumor-associated NK-cell characterization, a section within the core of the malignant area was selected by the pathologist and cleared of fat tissue prior to mechanical disruption. Tumor was digested for 40 minutes under agitation with collagenase type IV (1 mg/mL; Gibco) and DNAse (25U/mL; Sigma Aldrich). After filtration, single cell suspensions were stained with specific antibodies and analysed by multiparametric flow cytometry. vi-Stochastic neighbor embedding (SNE) methodology was implemented for the analysis of CD57⁺ distribution in paired samples of tumor-associated and circulating lymphocytes (35). Raw flow cytometry data were imported into R using flowCore and openCyto packages. A compensation matrix generated in FlowJo (v10.0.7, Tree Star) was exported an applied into R. Lymphocytes were gated by forward and side scatter in the CD45⁺ alive gate. Data from 10,000 gated lymphocytes or 2,000–5,000 NK cells were concatenated. Barnes-Hut t-SNE was conducted using the Rtsne package. Graphics were produced using packages ggplot2 and RColorBrewer.

For proliferation assays, NK cells were purified from CFSE-labeled PBMC [CFDA SE Cell trace (CFSE), Invitrogen, Molecular Probes] by negative selection using a Human NK Cell Enrichment kit (Miltenyi Biotec) and cultured in complete RPMI medium supplemented with 200 U/mL rhIL2 (Proleukin, Chiron) in flat-bottom 96-well plates previously coated with anti-CD16 (clone KD1). After 6 days, cells were surface stained with anti-CD56-APC and anti-CD57-BV605 (clone NK-1, BD Biosciences) and analyzed by flow cytometry. Results were expressed as percent of NK cells undergoing ≥4 divisions.

Data on tumor stromal-infiltrating NK-cell (TI-NK) numbers as determined by the enumeration of CD56⁺CD3⁻ cells by IHC in diagnostic tumor biopsies was available from a previous study including a partially overlapping cohort of patients with HER2⁺ primary breast cancer (12). The concordance between 40 of the patients included in both studies enabled the analysis of the relationship between CD57⁺ NK-cell numbers and tumor-infiltrating NK cells presented here.

Bivariate analyses by Mann–Whitney U test were used to assess the putative association between pCR and NK-cell markers as continuous variables. Multivariate analysis was conducted with binary logistic regression by adding patient age (as continuous variable) and estrogen receptor status (ER⁺ vs. ER⁻) in the model, separately. Thresholds for CD57⁺ NK-cell frequencies and numbers that best discriminated pCR achievement were determined using receiver operating characteristic (ROC) curve analysis and optimal cutoffs defined by the maximum Youden index. Predictive effects on pCR of CD57⁺ NK cells as categorical variables were calculated by Fisher exact test. Cox proportional hazards regression was used to estimate the HR in DFS analysis. Kaplan–Meier curves for DFS were used to compare time to event in patients stratified by their pCR achievement or CD57⁺ circulating NK cells. All P values were two-sided; P values lower than 0.05 were considered significant. Statistical analysis was performed using GraphPad Prism version 6 (GraphPad software) and STATA version 15 (STATA Corp.).

The phenotype of circulating NK cells was studied by multiparametric flow cytometry in baseline PBMC samples from prospectively recruited patients with HER2⁺ breast cancer, undergoing neoadjuvant treatment with HER2-specific antibodies and chemotherapy (n = 66). The analysis included several NK-cell receptors defining distinct NK-cell subsets (NKG2A, KIR2D, KIR3DL1, CD16, and NKG2C) and additional molecules associated with NK-cell differentiation (LILRB1, FcϵRγ, and CD57; gating strategy in Supplementary Fig. S2). Clinicopathologic characteristics of the patient cohort are presented in Table 1. pCR to neoadjuvant treatment was confirmed in 42% (27/64) of patients; ER status was the only conventional factor significantly associated with pCR in the studied cohort [OR 6.89; 95% confidence interval (CI), 2.06–23.05; P = 0.002]. No significant differences were observed in total lymphocyte counts, total NK cells, or in the distribution of most NK-cell receptors (e.g., NKG2A and KIR2D) between patients achieving pCR or not, in response to HER2-specific antibody treatment (Fig. 1A; Supplementary Table S1). Higher LILRB1⁺ and lower CD57⁺ average proportions of NK cells were detected in patients achieving pCR. However, absolute CD57⁺ NK-cell counts were the only parameter significantly different between patients achieving pCR or not. Despite the positive correlation between CD57⁺ NK cells and age (r = 0.44 and r = 0.46, for relative and absolute numbers, respectively, P = 0.002), CD57⁺ NK-cell numbers independently associated with lower likelihood of response to HER2-specific antibodies when adjusting by age [%CD57⁺ NK cells OR 1.04 (95% CI, 1.007–1.09), P = 0.01; CD57⁺ NK cells/μL OR 1.006 (95% CI, 1.0004–1.01), P = 0.03] and ER status [%CD57⁺ NK cells OR 1.05 (95% CI, 1.01–1.09), P = 0.01; CD57⁺ NK cells/μL OR 1.007 (95% CI, 1.0008–1.01), P = 0.02, per unit increase] in multivariate analysis. Higher frequencies of CD57⁺ NK cells were detected in patients with G1–G3 responses according to Miller–Payne criteria (Fig. 1B), hence identifying patients resistant to HER2-specific antibody–based neoadjuvant treatment. CD57⁺ NK-cell numbers were not associated with any clinicopathologic factor [i.e., tumor size, tumor grade, hormone receptor status (Supplementary Fig. S3A–S3D)]. ROC curves and cut-off values providing best discrimination between pCR and non pCR for CD57⁺ circulating NK cells were calculated (Supplementary Fig. S3E). As categorical variables, pCR rates were 60% in patients with <65% as compared with 10% in patients with ≥65% circulating CD57⁺ NK cells (Fig. 1C). Concordantly, pCR rates in patients with <164 CD57⁺ NK cells/μL were 60%, whereas only 18% of patients with ≥164 circulating CD57⁺ NK cells/μL achieved pCR (Fig. 1C).

Frequencies of circulating CD57⁺ NK cells in baseline samples from patients with HER2⁺ breast cancer were comparable with those detected in a sample of age-matched controls (n = 15), previously characterized in our laboratory (Supplementary Fig. S4).

None of the markers analyzed in NK cells and T cells correlated with disease-free survival (DFS) in linear regression analysis (Supplementary Table S1B and S2). Nonetheless, because pCR in the primary tumor is associated with long-term outcome in HER2⁺ breast cancer (36), we analyzed the association between CD57⁺ NK cells as categorical variable and DFS. In Kaplan–Meier analysis, patients with low CD57⁺ NK cells (<65% and <164 cells/μL; Fig. 1D–F) showed a trend for prolonged DFS yet not reaching statistical significance perhaps due to the small number of events guiding these analyses (5 relapses of 64 patients).

Chronic infection by HCMV has been associated with the persistent expansion of NKG2C⁺ adaptive NK-cell subpopulations including high proportions of CD57⁺ and FcRγ-deficient cells (30, 37, 38) and the inflation of CD57⁺ effector memory T cells with a terminal differentiation phenotype (39). We next assessed whether elevated CD57⁺ NK-cell numbers in the studied patient cohort could be related to HCMV-induced immunomodulation. HCMV seroprevalence was 86% in the analyzed cohort. As shown in Supplementary Fig. S5A and S5B, frequencies of NKG2C⁺ and FcRγ⁻ adaptive NK cells were comparable in patients achieving or not pCR to HER2-specific antibody–based treatment. Indeed, patients with high frequencies of CD57⁺ NK cells showed variable proportions of NKG2C⁺ NK cells, despite the positive correlation between both NK-cell markers (Supplementary Fig. S5C, gray area). On the other hand, proportions of CD57⁺ NK cells and HCMV-specific IgG titers did not correlate (Supplementary Fig. S5D) and CD57⁺ T-cell numbers were not associated with pCR achievement in these patients (Supplementary Table S2). An analysis of the T-cell compartment was done in a fraction of patients selected on the basis of their CD57⁺ NK-cell phenotypes (n = 22). The distribution of naïve, central memory (CM), effector memory (EM), and terminally differentiated effector memory (EMRA) CD8⁺ T cells, as identified by CD45RA and CCR7 expression, was comparable in patients with low (<65%) or high (≥65%) proportions of CD57⁺ NK cells (Supplementary Fig. S5E and S5F). Similarly, total CD8⁺ and CD4⁺ T cells as well as frequencies of CD57⁺ and CD28⁻ terminally differentiated CD8⁺ T lymphocytes were comparable between both patient groups (Supplementary Fig. S5E–S5G).

These analyses indicate that the association between CD57⁺ NK cells and the efficacy of HER2-specific antibodies in patients with breast cancer appears uncoupled from the expansion of adaptive NK cells and gross changes in the distribution of CD8⁺ T-cell subsets, reported to be associated with HCMV-dependent immunomodulation.

Improved responses to HER2-specific antibody-based treatments have been described for patients with CD16A 158V/F and 158V/V genotypes in some studies but not in others (40–42). Considering that circulating CD57⁺ NK cells express the CD16 receptor, we analyzed whether the CD16A 158V/F genotype could influence the association between baseline CD57⁺ NK cells and response to treatment with HER2 therapeutic antibodies. In our patient cohort, CD16A 158 V/V, CD16A 158 V/F, and CD16A 158 F/F genotypes, respectively, represented 15% (n = 9), 34% (n = 20), and 51% (n = 30) of the studied population, in conformance with the Hardy–Weinberg equilibrium (MAF: 0.32). Clinicopathologic characteristics of patients with CD16 158 V/V and V/F versus CD16 158 F/F genotypes were comparable (Supplementary Table S3). A trend for higher pCR rate in CD16A 158 V/V and V/F as compared with F/F patients was not statistically significant [2.2 OR (95% CI, 0.777–6.23), P = 0.13; Fig. 2A]. In vitro degranulation assays confronting patient-derived PBMCs with trastuzumab-coated SKBR3 cells showed that NK-cell–mediated ADCC was lower in patients with CD16A 158 F/F as compared to patients with CD16A 158 V/V or V/F genotypes (Fig. 2B). However, average frequencies and numbers of CD57⁺ NK cells were comparable in patients encoding CD16A 158 V/V or V/F and CD16A 158F/F genotype (Fig. 2C and D). The association between circulating CD57⁺ NK-cell numbers and pCR to HER2 antibody-based treatment was independent from the CD16A 158 V/F genotype in multivariate analysis [%CD57⁺ NK OR 1.06 (95% CI 1.0–1.1), P = 0.004; CD57⁺ NK cells/μL OR 1.006 (95% CI 1.0–1.01), P = 0.01]. Hence, although not pointing to the specific mechanism, these data rule out the possibility that the CD16 158 V/F genotype influences the association between circulating CD57⁺ NK cells and pCR neoadjuvant therapy with HER2-specific antibodies.

Given the inverse association between circulating CD57⁺ NK cell numbers and pathologic responses to HER2-specific therapeutic antibodies, we sought to investigate whether these cells could be detected in tumor-associated infiltrates from breast carcinomas. Single-cell suspensions were obtained from pretreatment tumor specimens and paired peripheral blood samples followed by the analysis of CD57 and CD16 expression in NK cells by multiparametric flow cytometry. Unbiased analysis of flow cytometry data was performed using t-distributed SNE (t-SNE; ref. 35), which clusters cells by the expression of analyzed parameters in two-dimensional maps. Maps integrating data from circulating and tumor-associated lymphocyte samples (n = 6 of each) revealed three clusters of CD57⁺ lymphocytes corresponding to CD16⁺ NK cells (1), CD8⁺ T cells (2), and a minor fraction of CD4⁺ T cells (3; Fig. 3A). Density maps separately depicting circulating and tumor-associated lymphocytes revealed the predominance of CD57⁺ NK cells in peripheral blood samples (Fig. 3B). Tumor-associated immune cell infiltrates displayed a consistent reduction in the frequency of NK cells, identified as CD56⁺CD3⁻ lymphocytes (Fig. 3C). The percentage of CD16⁺ tumor-associated NK cells was reduced (Fig. 3D), related to a preferential decrease in CD16⁺ CD57⁺ NK-cell fraction (Fig. 3E and F). Lower proportions of CD57⁺ CD8⁺ T cells were detected in tumor-associated (18% average frequency) as compared with circulating (42% average frequency) lymphocyte samples.

Limited homing and/or decreased proliferation/survival in the tumor niche could underlie the reduced presence of CD57⁺ NK cells in breast carcinoma infiltrates. On that basis, we analyzed the expression of CXCR3, CXCR1, and CX3CR1 in circulating CD57⁺ and CD57⁻ CD56^dim NK cells from patients with HER2⁺ breast cancer (n = 6). CD57⁺ NK cells showed higher expression of CXCR1 and CX3CR1 compared with CD57⁻ NK cells, yet expressed lower CXCR3 levels (Fig. 4A–C), a chemokine receptor involved in intratumoral NK-cell accumulation in experimental models (43). Expression of CCR7 and CD62L, markers associated with lymph node homing, was low in CD57⁺ and CD57⁻ circulating CD56^dim NK cells from patients with HER2⁺ breast cancer (Fig. 4B and C), in agreement with the chemokine receptor profile previously described in CD56^dim NK cells (44). On the other hand, CFSE dilution in vitro assays with purified NK cells from healthy donors showed that CD57⁺ NK cells displayed reduced IL2-dependent proliferation upon activation by CD16 crosslinking with high and low doses of an agonistic mAb, as compared with CD57⁻ NK cells (Fig. 4D and E) as well as upon activation by soluble IL2, in agreement with published data (24, 30). Lower CD57⁺ NK-cell numbers were recovered upon CD16-triggered activation as compared with nontreated or IL2-activated cultures (Fig. 4F).

On the other hand, NK cells from patients with HER2⁺ breast cancer with high and low proportions of CD57⁺ NK cells displayed enhanced degranulation and TNFα production in in vitro ADCC assays against trastuzumab-coated SKBR3 cells (Fig. 5A–C), in accordance with the recognized ADCC proficiency of CD57⁺ NK cells (30). Indeed, circulating CD57⁺ NK cells tended to display enhanced trastuzumab-induced degranulation as compared with their CD57⁻ counterparts in patients with HER2⁺ breast cancer, showing a comparable profile with that detected in healthy individuals paired by CD16 158 V/F genotype and NKG2C⁺ adaptive NK-cell content (Fig. 5D-G).

We have previously described the association between tumor-infiltrating NK-cell (TI-NK) numbers and pCR achievement in patients with early HER2⁺ breast cancer treated with HER2-specific antibody-based neoadjuvant therapy (12). We analyzed the relationship between circulating CD57⁺ NK cells and TI-NK-cell numbers in the group of patients for whom those determinations in baseline tumor biopsies were already available (n = 40; Fig. 6A). In this fraction of the studied cohort, TI-NK-cell numbers in the tumor biopsy and CD57⁺ circulating NK cells maintained an inverse and independent association with pCR to HER2 therapeutic antibodies (Supplementary Fig. S6). The correlation of both parameters evidenced the existence of a fraction of patients with high CD57⁺ circulating NK-cell numbers and lacking tumor-infiltrating NK cells (Fig. 6B). TI-NK-cell–associated and CD57–associated cutoffs complemented each other in the identification of patients with primary resistance (high CD57⁺ NK cells in circulation) or pCR achievement (high TI-NK cells) to HER2-specific therapeutic antibodies. (Fig. 6C and D).

The identification of mechanisms influencing the clinical efficacy of tumor antigen–specific antibodies could broaden their therapeutic benefit. We have described tumor-infiltrating NK cells associated with the antitumor efficacy of HER2-specific antibodies in patients with HER2⁺ breast cancer with primary disease (12). Results herein reported extend our previous findings by describing an inverse association between circulating CD57⁺ NK-cell numbers and the achievement of pCRs to early treatment with HER2 therapeutic antibodies. Prior studies have evaluated the association between the efficacy of HER2 antibody–based therapies and the magnitude of NK-cell cytotoxicity in in vitro assays (10, 45) or genetic factors such as CD16 polymorphisms (41, 42). Our study evidences that the differentiation profile of the circulating NK-cell repertoire correlates with response to HER2 antibody-based therapy, regardless of their cytolytic proficiency in vitro or CD16A 158 V/F genotype. From a clinical perspective, baseline screening for high circulating CD57⁺ NK cells could be implemented for the identification of patients with primary resistance to neoadjuvant treatment with HER2 therapeutic antibodies, complementing the positive predictive value of tumor-infiltrating NK-cell enumeration in diagnostic biopsies. Our results also provide a rationale for testing circulating CD57⁺ NK cells as biomarker of resistance to tumor antigen–specific antibody-based treatment in other cancer contexts.

The CD57 epitope is a sulfated carbohydrate posttranslationally added to glycoproteins and/or glycolipids on terminally differentiated NK and T cells (30, 46). The fact that patients showed comparable frequencies of total CD57⁺ T cells and CD57⁺/CD28⁻ CD8⁺ T lymphocytes regardless of pCR status indicated that resistance to HER2-specific antibodies was not associated with general immune senescence or immune risk profiles associated with frailty (47), but rather identified patients with a highly differentiated NK-cell compartment. Prior studies described associations between increased frequencies of circulating CD57⁺ CD8⁺ T cells and poor prognosis in several solid tumors (48–50), suggesting that CD57⁺ CD8⁺ T cells are unable to control tumor growth despite their cytotoxic activity. Our results indicate that CD57⁺ NK cells associate to a similarly low capacity for controlling tumor growth in the context of HER2 antibody–based breast cancer treatment. The relationship between circulating CD57⁺ NK-cell numbers and HER2-specific antibody efficacy is supported by observations showing the reduced presence of tumor-associated CD57⁺ NK cells and, particularly, the inverse correlation between circulating CD57⁺ and tumor-infiltrating NK-cell numbers.

Unraveling the molecular keys underlying the herein described association could point to strategies for counteracting primary resistance to HER2 therapeutic antibodies in patients with breast cancer with high numbers of CD57⁺ circulating NK cells. In the nervous system, CD57 labels proteins involved in cell adhesion, recognition, and migration, such as NCAM (51). Studies on chemokine receptor expression showed that CD57⁺ circulating NK cells from patients with HER2⁺ breast cancer expressed reduced surface CXCR3, a receptor involved in NK-cell tumor homing (43). Reduced CXCR3-dependent homing into breast carcinomas could be one of the mechanisms contributing to CD57⁺ NK-cell scarcity in tumor immune infiltrates. In addition, circulating CD57⁺ NK cells displayed limited proliferation and persistence upon activation via CD16 in vitro, in agreement with the described replicative senescence of CD57⁺ T cells (24, 52). Thus, reduced presence of CD57⁺ NK cells in tumor immune infiltrates and their association with resistance to treatment with HER2 antibodies could also result from their inability to proliferate and persist upon activation in the tumor microenvironment. CD57⁺ NK cells from patients with HER2⁺ breast cancer displayed proficient degranulation and cytokine production in response to trastuzumab-coated SKBR3 cells in ADCC in vitro assays, indicating that their association with anti-HER2 therapy failure could not be attributed to functional dysfunction but rather depended on features associated with their differentiation stage (i.e., homing and persistence into the tumor). The fact that the efficacy of HER2 therapeutic antibodies was unrelated to the distribution of other receptors related to NK-cell functional maturation (i.e., NKG2A⁺, KIR⁺) or enhanced ADCC in vitro (i.e., adaptive NKG2C⁺ and FcRγ⁻ NK cells) further supported that CD57⁺ NK-cell association with anti-HER2 efficacy was independent of the cells’ functional competence. It is tempting to speculate that differentiated CD57⁺ NK cells might preferentially home to bone marrow, as suggested by their presence in the bone marrow from patients with leukemia in complete remission (53) and their association with reduced relapse rates in hematopoietic stem cell–transplanted patients (54, 55), while showing reduced surveillance of peripheral tissues. Monitoring changes in tumor-infiltrating NK-cell subpopulations after treatment could provide valuable information on the molecular keys underlying the correlation between CD57⁺ circulating NK cells and efficacy of HER2-specific antibody-based therapy. However, the nature of this clinical study based on the analysis of samples obtained from patients receiving standard neoadjuvant treatment limits the possibility of studying sequential tumor samples.

Although factors triggering CD57 expression in NK cells are unknown, CD57 expression is enhanced through chronic stimulation by environmental events: CD57⁺ NK cells increase with age, and show delayed appearance in hematopoietic stem cell transplant recipients (24, 26–28). Chronic viral infections, particularly HCMV, have been associated with CD57⁺ increased expression in both NK and CD8⁺ T cells (27, 39, 56, 57). Nonetheless, the association between CD57⁺ NK cells and response to treatment with HER2 antibodies in the studied patient cohort was independent of chronologic age and uncoupled from changes associated with HCMV infection in both NK- and T-cell compartments (i.e., expansions of NKG2C⁺ or FcRγ-deficient adaptive NK cells, expression of LILRB1, or inflation of CD28⁻ or CD57⁺ effector memory CD8⁺ T cells). Hence, our data support a model in which the CD57 expression profile in an adult NK-cell compartment is determined by a combination of environmental factors, including not only HCMV infection but also other infections (e.g., EBV) and/or different stimuli (e.g., immunocomplexes). The fact that CD57⁺ NK-cell numbers were not associated with any clinicopathologic factor (i.e., tumor size, tumor grade, hormone receptor status) argues against the influence of underlying disease on the NK-cell differentiation profile in the studied cohort. Nonetheless, the analysis of other activating (e.g., NKG2D, NKp30, NKp44, or DNAM-1) and inhibitory receptors (e.g., TIGIT and TIM-3) would provide a broader view of NK-cell receptor repertoire in patients with breast cancer.

Circulating CD57⁺ NK-cell numbers and CD16A genotype could exert a complementary impact on HER2 therapeutic antibody efficacy by influencing NK-cell tumor homing/persistence and the strength of HER2 antibody–dependent activation. Studies on larger cohorts could address whether the combination of variables would provide added value for predicting treatment efficacy.

Limitations of this study include the modest cohort size and the short clinical follow-up that (i) may explain why the trend for CD16A 158 V/F genotype and pCR failed to reach statistical significance and (ii) precluded conclusive association of CD57⁺ NK cells with long-term treatment efficacy. On the other hand, our study's strengths are the hypothesis-driven prospective design, the progression from previous results relating NK cells to anti-HER2 efficacy in primary breast cancer, and the integration of in vitro and ex vivo data.

Overall, our data indicate that highly differentiated NK-cell repertoires, revealed by the frequency of CD57⁺ cells, identify patients with HER2⁺ breast cancer with resistance to anti-HER2 therapy. The study also identifies NK-cell differentiation status, rather than their functional maturation, as a parameter that might influence on their antitumor function in the context of solid tumors. These results encourage future research extending the analysis of CD57⁺ NK cells as putative efficacy biomarkers and introducing NK-cell aging as a target for improving the efficacy of HER2-specific therapeutic antibodies.

A. Muntasell has received speakers bureau honoraria from Roche. B. Bermejo is a consultant/advisory board member for Roche and Pfizer. M. Martinez-Garcia is a consultant/advisory board member for Roche and Pfizer. A. Lluch is a consultant/advisory board member for Novartis, Pfizer, Roche/Genentech, Eisai, and Celgene. J. Albanell reports receiving speakers bureau honoraria and is a consultant/advisory board member for Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Muntasell, S. Servitja, F. Rojo, J. Albanell, M. López-Botet

Development of methodology: A. Muntasell, S. Servitja, B. Bermejo, F. Rojo, O. Arpí, L. Soria

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Servitja, M. Cabo, B. Bermejo, S. Perez-Buira, F. Rojo, O. Arpí, M. Moraru, L. Serrano, I. Tusquets, M.T. Martínez, A. Vera, M. Martínez-García, L. Soria, L. Comerma, S. Santana-Hernández, P. Eroles, C. Vilches, A. Lluch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Muntasell, M. Cabo, F. Rojo, M. Costa-Garcia, G. Heredia, S. Santana-Hernández, J. Albanell, M. López-Botet

Writing, review, and/or revision of the manuscript: A. Muntasell, S. Servitja, M. Cabo, F. Rojo, I. Tusquets, M.T. Martínez, M. Martínez-García, S. Santana-Hernández, P. Eroles, A. Rovira, A. Lluch, J. Albanell, M. López-Botet

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Servitja, F. Rojo, G. Heredia, A. Vera, L. Soria, L. Comerma, S. Santana-Hernández, J. Albanell

Study supervision: A. Muntasell, J. Albanell

Other (interpretation and discussion of results): A. Rovira

We acknowledge A. Moretta (University of Genoa, Genoa, Italy), L. Lanier (University of California, San Francisco, San Francisco, CA), and S. Ferrini (National Institute for Cancer Research, Genoa, Italy) for kindly providing mAbs used in this study. The authors thank the technical help of Oscar Fornas (Flow cytometry Core Service at PRBB). We acknowledge Xavier Duran [Consulting Service on Methodology for Biomedical research (AMIB) at IMIM] for assistance in statistical analysis. The authors also thank all clinical professionals (Mar Biobank, INCLIVA Biobank (PT17/0015/0049; B.000768 ISCIII//) and nurses at Hospital del Mar/Hospital Clinic de Valencia Oncology Departments) what provided support to this project. We are also especially grateful to all patients and healthy volunteers participating in this study. The authors are supported by coordinated research projects from Asociación Española contra el Cáncer (GCB15152947MELE) and Proyecto Integrado de Excelencia ISCIII (PIE 2015/00008); M. López-Botet and A. Muntasell are supported by Worldwide Cancer Research Foundation (15-1146) and by Plan Estatal I+D Retos (SAF2016-80363-C2-1-R), Spanish Ministry of Economy and Competitiveness (MINECO, FEDER), and by Generalitat de Catalunya (2017 SGR 888). J. Albanell, A. Rovira, and F. Rojo are supported by ISCiii/FEDER (PI15/00146, PI18/00006, PI18/00382 and CIBERONC) and by Generalitat de Catalunya (2017 SGR 507) and Fundació Privada Cellex (Barcelona). A. Lluch and P. Eroles are supported by ISCiii/FEDER (PI15/01617, PI18/01219 and CIBERONC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.