Treatment Regimen, Surgical Outcome, and T-cell Differentiation Influence Prognostic Benefit of Tumor-Infiltrating Lymphocytes in High-Grade Serous Ovarian Cancer

This article is featured in Highlights of This Issue, p. 525

Epithelial ovarian cancer (EOC) is the most deadly gynecologic malignancy with an overall 5-year survival of 38% to 40% (1). EOC is a heterogeneous disease with multiple histologic subtypes, of which high-grade serous carcinoma (HGSC) is the most common (2). The poor prognosis of the disease is largely due to diagnosis at advanced-stage and therapy-resistant disease relapses that occur in the majority of patients following initial treatment consisting of cytoreductive surgery and platinum-containing chemotherapy (2). Nevertheless, a subset of patients with HGSC appears to remain disease-free for prolonged periods of time. To date, several factors have been identified to define this subset of patients.

Arguably the strongest prognostic factor is the amount of residual tumor tissue after surgery (3). Indeed, patients in which a complete resection, leaving no residual macroscopic lesions, is achieved have an approximately 1.6-fold longer survival than patients with remaining macroscopic disease (4). As such, neoadjuvant chemotherapy (NACT) is increasingly considered as the treatment option of choice in patients in whom the chances of upfront complete cytoreduction are minimal (e.g., patients with stage IIIC/IV disease with widespread tumor dissemination) or as an effort to reduce morbidity due to aggressive primary cytoreductive surgery (PS). NACT may reduce tumor load before cytoreductive interval surgery, increasing the likelihood of completely resected tumors (5), as well as reduce perioperative morbidity and mortality in this group of patients.

A second major factor in determining the prognosis of HGSC is the presence of tumor-infiltrating lymphocytes (TIL). Infiltration of CD3⁺ T cells, and particularly CD8⁺ cytotoxic T cells (CTL) is associated with a better prognosis for patients with HGSC (6, 7). Moreover, the ratio between CTL and immune-inhibitory cells (FoxP3⁺ regulatory T cells and CD33⁺ myeloid-derived suppressor cells), the activation status of T cells (CD45RO), and their cytolytic activity (measured by TIA-1/Granzyme B expression) are all predictive for survival (8–10). Despite these well-established prognostic roles of both surgical outcome and TIL, the relationship between both the factors has remained largely unknown. Indeed, most studies on the prognostic value of TIL in ovarian carcinomas used heterogeneous patient populations including various subtypes of EOC, early and advanced stages and different grades and included patients treated with various chemotherapeutic regimens. In addition, many studies use different outcome measures for the result of cytoreductive surgery, further complicating analysis.

Next to this effect of surgical outcome on the prognostic value of TIL, neoadjuvant chemotherapeutic treatment might also alter TIL infiltration and function and therefore the overall effectiveness of a (pre-existing) immune response. Indeed, platinum-based NACT in patients with breast cancer can increase TIL infiltrate (11) and levels of TIL before treatment are reportedly predictive for the subsequent response to chemotherapy (12–14) as well as response to trastuzumab, a monoclonal antibody targeted against HER2 (15). Although no association of TIL infiltrate and NACT has been described in HGSC, a single study found that patients with high TIL were more likely to be completely cytoreduced, potentially due to a better tumor control by the immune system (6).

To address the effects of cytoreductive surgery and NACT on the prognostic role of CD8⁺ TIL, patient cohorts therefore need to be highly standardized in terms of treatment schedule, stage, grade, and outcome of cytoreductive surgery. Here, we generated two such cohorts of patients with advanced-stage HGSC treated with identical chemotherapy, but with primary treatment being either PS or NACT.

We show here that the prognostic benefit of CD8⁺ TIL is restricted to patients in whom a complete cytoreductive surgery was achieved (no residual tumor) and is not demonstrable in patients in which upfront complete cytoreduction was considered to be unattainable (NACT patients). Interestingly, we also found that patients performed better when the tumor was infiltrated by less-differentiated, CD8⁺ T cells, despite the presence of a residual macroscopic tumor after cytoreduction. Taken together, our findings indicate that treatment regimen, surgical result, and the differentiation of TIL should all be taken into account when studying immune factors in HGSC or, by extension, selecting patients with HGSC for immunotherapy trials.

An anonymized database was created containing information on clinicopathologic characteristics and follow-up of patients diagnosed with serous ovarian cancer at the University Medical Center Groningen (Groningen, The Netherlands) between January 2000 and December 2012. Patients were staged according to the International Federation of Gynecology and Obstetrics (FIGO) criteria, and graded by a gynecologic pathologist based on the World Health Organization guidelines. Patients were selected if sufficient formalin-fixed, paraffin-embedded (FFPE) tissue was available for tissue microarray (TMA) construction. Tissue was obtained either from primary cytoreductive surgery (PS cohort) or from interval surgery (NACT cohort). Patients that underwent PS subsequently received six cycles of platinum-based chemotherapeutic regimen often combined with paclitaxel. Patients that were selected for NACT first received three cycles of chemotherapy, followed by interval surgery and three more cycles of chemotherapy. Follow-up was calculated from date of initial treatment (either surgery or NACT) and was last updated in April 2014.

Patient data were retrieved from the institutional database into a new anonymous database, in which patient identity was protected by unique patient codes. According to Dutch law, no approval from our institutional review board was needed. Primary patient TIL were isolated from surgical tumor waste for which no approval from our institutional review board was needed according to Dutch law.

From 265 patients with HGSC, the FFPE tissue was available for the construction of a TMA. A gynecologic pathologist confirmed the presence of tumor tissue on hematoxylin and eosin (H&E) slides and selected representative locations with tumor tissue. Triplicate cores with a diameter of 1 mm were taken from each paraffin-embedded tissue block and placed in a recipient block by using a tissue microarrayer (Beecher Instruments). An asymmetrical grid was chosen with a 14 × 9 layout. Both normal and tumor tissue were included as orientation cores and controls. The seventh column from the fourth row onwards, and the fourth row from the seventh column onwards were left empty as a points of reference for grid layout. From each TMA block, 4 μm sections were cut and applied to APES-coated slides (Starfrost). Core-loss was on average 9.0% (PS cohort) and 10.6% (NACT cohort). The presence of tumor in the arrayed samples was confirmed by H&E staining.

TMA slides were stained with mouse anti-human CD8 antibody [DAKO; clone: C8/144B, 1:25 in blocking buffer (1% BSA/PBS with 1% human AB serum)] or rabbit anti-human CD27 antibody (Abcam; clone: EPR8569, 1:150 in blocking buffer) by use of IHC using standard methods (Supplementary Methods). Furthermore, on the basis of the highest infiltration of CD8⁺ cells, 10 patients were selected from the TMA data set and full tumor tissue slides were retrieved for analysis of CD8 and CD27 costaining by use of multicolor immunofluorescence. Antibody binding was visualized with goat anti-rabbit Alexa Fluor-488 and goat anti-mouse Alexa Fluor-555 (1:150, Life Technologies). Counterstaining was done by 4′,6-diamidino-2-phenylindole (DAPI).

Fresh tumor material was obtained for the isolation of TIL from patients undergoing cytoreductive surgery. With a scalpel, tumor pieces of approximately 0.5 cm³ were cut, and subjected to digestion in digestion medium (RPMI supplemented with 1 mg/mL collagenase type IV (Life Technologies) and 31 U/mL rhDNase (Pulmozyme, Genentech) for 30 minutes at 37°C. Subsequently, the digestion medium containing remaining tumor pieces was filtered over a 70-μm cell strainer (Corning) and cells were pelleted, washed, and cryopreserved until further use.

From the digested tumor samples TIL were phenotyped by multiparameter flow cytometry. The Zombie Aqua Fixable Viability Kit was used for live/dead stain according to the manufacturer's instructions (BioLegend). Antibodies used were CD3-PerCP-Cy5.5 (OKT3), CD8-APC-eFluor780 (RPA-T8), CD45RO-PE-Cy7 (UCHL1), CD137-PE (4B4-1), and PD1-APC (MIH4; all eBioscience); CCR7-BV421 (150503; BD Biosciences); and CD27-FITC (9F4; Sanquin). All flow cytometry was performed on a FACSVerse (BD Biosciences) and samples were analyzed with the Cytobank software (cytobank.org).

Scoring of TMA samples was performed if cores had at least 20% tumor epithelium present, and if at least two cores per patient were analyzable. All CD8⁺ or CD27⁺ stained cells localized in tumor epithelium in each core were counted manually by two individuals that were blinded for patient characteristics. The two individual scores were compared and differences in counts of >10% were re-analyzed until consensus was reached. Cell counts were represented as total number of cells per mm² of tumor epithelium. H&E slides were used for comparison in cases tumor/stroma regions were not clearly definable.

Immunofluorescent slides were scanned using a TissueFaxs imaging system (TissueGnostics). Processed channels were merged using Adobe Photoshop. On each slide an area of 1 mm² of tumor epithelium was selected based on DAPI staining, and cells were counted manually.

All statistical analyses were performed using the IBM SPSS version 22 (SPSS Inc.) or Graphpad Prism. Disease-specific survival (DSS) was defined as the time period from date of surgery or first chemotherapeutic treatment until death due to ovarian cancer or last follow-up and was analyzed by using the Kaplan–Meier method, with log rank test to determine differences between groups. Variables that were significantly associated with DSS in univariate analyses were entered into a multivariate analysis using the Cox proportional hazards model. Differences in cell infiltration between 43 matched ovarian and omental primary tumor tissues were assessed by Wilcoxon signed ranks test, no differences were found and therefore primary ovarian and omental tissues were both used in the analyses. To determine differences in cell populations between clinicopathologic variables or between different TIL subsets, the Man–Whitney U or one-way ANOVA test were used. P values of <0.05 were considered significant, and all tests were performed two-sided.

From a total of 265 patients, the tissue in FFPE blocks obtained at primary or interval surgery was available to construct the TMA. Two cohorts were created on the basis of treatment strategy, a PS (n = 134) and a NACT (n = 121) cohort. From 15 patients tissue from a recurrence and from 5 patients tissue from both primary and interval surgery were included on the TMA. Patients diagnosed with high-grade (grade III and undifferentiated), advanced-stage (FIGO ≥ IIB) serous ovarian carcinoma were selected for analysis in the current study (n = 171), recurrences were excluded (Table 1). The five patients of which both primary and interval tissue was available, were analyzed in the cohort of their primary treatment, which was for all five NACT.

For the PS cohort, 87 patients were included, who received six cycles of platinum-based chemotherapeutic treatment after surgery. Five patients did not receive chemotherapy, either because they received palliative treatment and refused additional chemotherapy, or died due to surgical complications. Surgical cytoreduction resulted in 39 (44.8%) patients with no residual macroscopic lesions. The mean age was 64.1 years (SD = 11.3) and median duration of follow-up was 31.0 months [interquartile range (IQR): 40]. Analysis of the DSS of these patients showed a better prognosis for patients who had no residual macroscopic lesions as compared with optimal (≤1 cm residual tumor tissue; P = 0.036) and suboptimal (>1 cm remaining tumor nodules; P < 0.001) cytoreduced tumors (Supplementary Fig. S1A), confirming the value of complete cytoreduction. Age was also a prognostic parameter in this cohort, older patients (cutoff 59 years of age) had a worse prognosis (P = 0.002, Supplementary Fig. S1B).

In the NACT cohort, all 84 included patients received three cycles of platinum-based chemotherapy before and three cycles following surgical cytoreduction. Complete cytoreduction was achieved at interval surgery in 27 of the 84 patients (32.1%). The mean age of patients in this cohort was 64.4 years (SD: 8.6) and median duration of follow-up was 22.0 months (IQR: 24.5). DSS was significantly shorter in patients in this cohort in comparison with the PS cohort (median DSS 24.0 versus 34.0 months, P = 0.042, Table 1). For DSS analysis concerning residual tumor tissue after surgery, the same trend could be observed as in the PS cohort. Although the difference in DSS between nonmacroscopic disease and optimal cytoreduction was not statistically significantly different (P = 0.073), while the difference between no residual tumor and >1 cm tumors was (P < 0.001, Supplementary Fig. S1C). Age was not a significant prognostic marker in this cohort (Supplementary Fig. S1D).

In these two cohorts of patients with advanced-stage HGSC treated with identical chemotherapeutics, we first validated the previously published observation that CD8⁺ TIL are associated with prognosis in HGSC. TMA slides were analyzed for the infiltration of CD8⁺ CTL in the tumor epithelium. In the PS cohort, 87.9% of the samples had infiltrating CD8⁺ cells with a median infiltration of 21.31 cells/mm² tumor (IQR: 69.78; Fig. 1A and B). Prognostic characteristics were analyzed for differences in CD8⁺ cell count in order to determine whether these factors influence intratumoral CTL infiltration. Age (P = 0.215), FIGO stage (P = 0.172), or surgical cytoreductive outcome (P = 0.194) did not show differences in total CD8⁺ count.

In the NACT cohort, tumors obtained at interval surgery were infiltrated with CD8⁺ cells in 96.2% of patients with a median infiltration of 35.00 cells/mm² (IQR: 81.76; Fig. 1B). There were no differences in number of cells within patient groups regarding age (P = 0.856), FIGO stage (P = 0.564), or surgical result (P = 0.535). Comparing the two cohorts revealed that the infiltration of CD8⁺ TIL did not seem to be affected by the received neoadjuvant chemotherapeutic treatment (Fig. 1B; P = 0.084).

For subsequent survival data analysis, patients were dichotomized based on the group of patients with the highest infiltration of cells (highest tertile) versus the group with low or no infiltration. DSS analysis based on infiltration of CD8⁺ cells revealed a nonsignificant improvement of survival (P = 0.052) in the PS cohort for the group with high infiltration compared with patients with low or no infiltration (Fig. 1C). In the NACT cohort, no differences in survival were detected between the groups with either a high or a low infiltration of CD8⁺ (Fig. 1D, P = 0.992).

Because the result of cytoreductive surgery is a major predictor for prognosis and determines the course of disease, we next analyzed the prognostic benefit of CD8⁺ TIL in relation to primary surgical outcome. Patients were categorized based on patients that had no residual macroscopic lesions after surgery and those with residual macroscopic disease. Within the group without residual tissue, a clear survival benefit was observed for patients who had a high infiltration of CD8⁺ T cells (Fig. 1E; P = 0.028), suggesting a role of CD8⁺ in tumor immunosurveillance. By contrast, no survival benefit of CD8⁺ T-cell infiltration was detected in the group of patients that received an incomplete cytoreduction. No difference was found in total number of infiltrating CD8⁺ TIL in completely resected tumors as compared with tumors of patients with remaining tumor tissue (Fig. 1F), suggesting that the number of TIL present does not affect surgical outcome in these patients. Concluding, a survival benefit of infiltration with CD8⁺ TIL is present only in patients that had a complete tumor resection at time of primary surgery.

Interestingly, in adoptive cell transfer (ACT) studies, undifferentiated T cells provide greater control over established large tumor masses, with terminally differentiated T cells providing only poor tumor control (16). Therefore, we wondered whether less-differentiated T cells were equally associated with better tumor control in patients with HGSC in which complete cytoreduction could not be achieved. Hereto, we analyzed expression of the differentiation marker CD27, known to be a key marker associated with improved outcome in ACT. First, to assess coexpression of CD27 on CD8⁺ TIL, whole slides from 10 patients with a high infiltration of CD8⁺ cells (95.9 to 2,555.0 cells/mm²) were selected from the TMA data set and stained for CD27 and CD8 using immunofluorescence. First, we determined if tumor regions could be differentiated from stromal regions by using DAPI nuclear staining. Tumor slides were stained with anti-EpCAM and anti-fibronectin antibodies, followed by DAPI staining of the nuclei. By means of DAPI, tumor islands were selected and checked for expression of EpCAM and fibronectin, showing that indeed the differentiation between tumor islands (EpCAM^pos) and stromal regions (fibronectin^pos) could be made (data not shown). Figure 2A shows a representative image for the double staining of CD8 and CD27, with the background staining of DAPI. Within the tumor islands, both CD8 and CD27 positive cells could be detected, which were either single or double-positive (Fig. 2B and 2C). Although CD8⁺CD27⁺ cells were the dominant subtype, clear differences were detected in the percentages of CD8⁺CD27⁻ and CD8⁺CD27⁺ cells/mm² tumor epithelium within this group of patients (Fig. 2D). Thus, CD27 is expressed on CD8⁺ TIL in HGSC, with variability in infiltration between patients.

To further analyze the phenotype of CD8⁺ and CD27⁺ infiltrating cells, TIL were isolated from fresh tumor tissue (N = 9, Table 1) and analyzed by flow cytometry. The gating strategy for live CD8⁺CD27⁺ cells is depicted in Fig. 3A. First, we compared TIL populations isolated from fresh tumor tissue with fluorescent staining results of full slides. The percentages of CD27⁺ cells within the CD8⁺ subpopulation found were comparable for samples analyzed by either of the two techniques, confirming the reproducibility of the used methods of analysis (Fig. 3B; P = 0.440). Within the population of CD8⁺ TIL, we then compared the CD27⁺ with the CD27⁻ cells for differentiation markers. Here, no differences were found in expression of CD45RO (Fig. 3C; P = 0.421), but CCR7 was predominantly expressed in the CD27⁺ group (Fig. 3D; P = 0.042). Indeed, most of the CD27⁺CD8⁺ cells were double positive for CD45RO and CCR7 (Fig. 3E), confirming the less-differentiated, yet antigen-experienced phenotype of these TIL. CD8⁺CD27⁻ cells had relatively comparable levels of CD45RO⁺ cells positive or negative for CCR7, suggesting this cell subset represents a further differentiated phenotype compared with the CD8⁺CD27⁺ population (Fig. 3F). PD1 was expressed on most of the CD8⁺ TIL (Fig. 3G) with no difference within the subsets with or without CD27 expression (P = 0.095). On the other hand, the marker for recent T-cell activation CD137, was found to be expressed on a significantly higher percentage of CD27⁺CD8⁺ cells in comparison with the CD8⁺CD27⁻ population (Fig. 3H; P = 0.008). Conversely, within the total CD8⁺CD137⁺ population, most cells were CD27⁺PD1⁺, which was a significantly higher percentage compared with the CD8⁺CD137⁻ subpopulation (Fig 3I; P = 0.016). Comparing the subsets with or without expression of CCR7 within the CD8⁺CD27⁺ population, showed that there was no difference in the percentage of CD137⁺PD1⁺ cells (Fig. 3J; P = 0.440). Taken together, CD27⁺ TIL represent a less-differentiated, potentially activated, antigen-experienced subset of CD8⁺ T cells.

Based on the flow cytometry results, we hypothesized the CD27 subset of TIL to resemble less-differentiated TIL, which are therefore better capable of immune control of the tumor, presumably due to a greater expansion potential (17). In order to further analyze this, expression of this marker was determined in both the PS and the NACT cohort. In the PS cohort, 85.5% out of a total of 76 patients demonstrated CD27⁺ cell infiltration. Median cell count was 13.94 cells/mm² of tumor epithelium (IQR: 30.92), revealing lower total cell counts compared with CD8⁺ (Fig. 4A and B versus Fig. 1B). Infiltration was not influenced by any of the prognostic factors age (P = 0.832), stage (P = 0.161), or surgery status (P = 0.749).

In the NACT cohort, the same trend could be observed. In comparison to CD8, a lower percentage of patients had CD27 infiltration (86.4%) and the total number of cells in these tumors was lower (median: 10.83; IQR: 30.25; Fig. 4B versus Fig. 1B). There were no differences in amount of cells within patient groups regarding age (P = 0.965), FIGO stage (P = 0.773), or surgical result (P = 0.902). In order to determine whether the chemotherapy had influenced the infiltration of CD27⁺ TIL, we compared cell numbers between the two data sets and found no differences between the two cohorts (Fig. 4B; P = 0.891). Thus, in the majority of patients in both cohorts infiltration of CD27⁺ cells was detected, with higher numbers of CD8 as compared with CD27, which was not influenced by any clinicopathologic factors or NACT.

To determine the prognostic value of CD27⁺ TIL infiltration, patients were subdivided based on cutoff for the highest tertile of CD27 cell counts. In the PS cohort, DSS analysis based on CD27 expression showed a clear survival benefit for the highly infiltrated group (P = 0.001; Fig. 4C). In the NACT cohort, no differences in survival could be detected within the groups with a high or low infiltration of CD27⁺ (cutoff highest tertile; Fig. 4D).

If indeed, CD27⁺ TIL show better tumor control, it is to be expected that these cells show prognostic value in completely resected tumors as well as in patients who had remaining tumor tissue following surgery. To determine whether CD27⁺ cells can compensate for incomplete removal of the tumor, we again analyzed groups based on surgical outcome in the PS cohort. Indeed, a clear survival benefit could not only be observed in patients without residual disease (P = 0.011), but also in the group of patients, which were incompletely cytoreduced (P = 0.017; Fig. 4E). The survival benefit in the incomplete group can be attributed completely to the optimally cytoreduced patients (≤1 cm; P = 0.021), although no benefit was observed in the patients that had >1 cm of residual tumor after surgery (P = 0.369; Supplementary Fig. S2).

To confirm the value of CD27 over CD8 as a marker for prognosis in HGSC, we performed a multivariate Cox regression analysis (Table 2), including all variables shown to be associated with survival in univariate analyses (FIGO stage, surgery result and age). In this model, only surgical result [HR: 1.50; 95% confidence interval (CI): 1.24–1.80] and CD27⁺ cell infiltration (HR: 0.23; 95%CI: 0.10–0.56) proved to be of prognostic value.

Taken together, the CD27⁺ TIL subset is more strongly associated with a favorable prognosis compared with the CD8⁺ TIL in patients with HGSC, due to survival benefit in both patients with residual macroscopic disease and patients with no residual macroscopic disease. Whereas CD8⁺ TIL provide survival benefit only in patients where no macroscopic disease is present after cytoreductive surgery.

In the present study, we demonstrate that the prognostic value of TIL in advanced-stage HGSC is variable for patients primarily treated with surgery or NACT. Furthermore, the differentiation status of infiltrating CD8⁺ T cells, as determined by CD27 expression, proved to be associated with a survival benefit superior to that observed for the overall CD8⁺ TIL population. This differential impact of various T-cell subsets is related to surgical result of primary surgery. The less-differentiated CD27⁺ TIL were found to have a prognostic benefit even in patients with residual tumor tissue after surgery.

TIL have long been known to be associated with a favorable prognosis in EOC, with ratios between subtypes of cells, their activation status, and cytolytic activity as determining factors (8,9). We analyzed two patient cohorts of advanced-stage HGSC treated with either surgery or NACT as the primary treatment modality. Patient selection for the latter treatment is based on tumor dissemination, comorbidity, and performance status. Therefore, not surprisingly, the patients in the NACT cohort showed a worse prognosis in general. Also, when studying parameters that were of clear prognostic value in the PS cohort, namely, residual tissue after surgery and age, the effect on prognosis was less clear or not present at all in the NACT group. This may in part explain why the effect of infiltrating CD8⁺ or CD27⁺ cells was also not associated with a survival benefit in this group. Whether this is due to the response of T cells to chemotherapy cannot be excluded, although total cell counts of these cell populations do not indicate a difference in infiltration. This is in line with the observation in mice that high dose carboplatin or paclitaxel treatment does not affect the amount of circulating T cells, nor the number of ovalbumin-specific T cells after vaccination (18). Further studies will have to reveal whether functional differences are present in the TIL of tumors treated with these agents. Also, it is of interest whether response to chemotherapy can be predicted by TIL infiltrate as is the case in breast cancer (12–14).

In both cohorts, the most pronounced indicator of survival is the result of cytoreductive surgery, with a more distinct effect in the primary surgery cohort, consistent with observations by Rosen and colleagues (19). In line with the immune tumor control hypothesis, one would expect that the influence of TIL infiltration on prognosis is affected by the amount of residual tumor tissue after surgery. Indeed, our data indicate that CD8⁺ TIL are of prognostic benefit only when maximum tumor cytoreduction can be achieved. Less-differentiated CD27⁺ TIL were able to compensate for incomplete surgical resection of the tumor, but only to a limited extent (residual tumor mass <1 cm). These data clearly indicate that the level of cytoreduction should be taken into account when studying immunomodulating effects. Of note, surgical outcome may not be a predictor of survival itself, but rather a reflection of the underlying tumor biology. Indeed, it was recently shown that a specific gene expression signature correlates with surgical outcome in ovarian cancer (20). In our cohort, this may have led to a selection bias for the NACT cohort, because patients were selected for NACT if chance of upfront complete cytoreduction in primary surgery was minimal, thus these patients possibly differ in expression of key genes. Therefore, the lack of prognostic effect of infiltrating CD8 and CD27 cells in the NACT cohort could potentially be due to a difference in tumor biology. It will be of interest to determine whether patients with the two profiles show contrasting prognostic effects of immune infiltrates, and if so, whether a combined immune and cytoreduction gene signature can be identified for the stratification of patients or as a selection criterion for patients likely to respond to immunotherapy.

Our first in-human analysis of intraepithelial CD27⁺ TIL suggests these cells to represent a more potent tumor-controlling subset of CD8⁺ TIL in HGSC. These observations are in line with recent ACT studies in melanoma. In particular, CD27⁺ TIL were found to persist longer after ACT and had a higher reactivity upon re-administration (21), presumably due to a greater expansion potential in response to antigen exposure (17). Not surprisingly, the proportion of CD8⁺CD27⁺ cells within the transferred TIL was therefore strongly associated with tumor regression in patients (22). Furthermore, introduction of a CD27 intracellular signaling domain to chimeric antigen receptor T cells led to longer persistence after infusion and improved efficacy in an ovarian cancer xenograft mouse model (23). This matches our results showing longer survival in patients that have infiltrating CD27⁺ cells in their tumors. Interestingly, while CD27 is constitutively expressed on naive T cells, further upregulation is induced upon activation via TCR signaling, although loss of CD27 occurs in more terminal stages of differentiation. Therefore, the CD8⁺CD27⁺ subpopulation found within tumors might represent a more activated, tumor-reactive subset of cells. Indeed, when compared with the CD27⁻ CTL, CD27⁺ CTL demonstrated a significantly higher percentage of cells expressing CD137, the recently described marker for tumor-reactive T cells in HGSC (24). Conversely, CD137⁺ CTL were predominantly (>80%) double-positive for CD27 and PD1. In addition to this increased capacity for tumor recognition, CD27⁺ CTL also expressed CD45RO and CCR7, indicating that these cells are indeed antigen-experienced, but not yet terminally differentiated (25).

Our results suggest CD27 as an interesting target for immunotherapy in HGSC, in which stimulation of T cells via CD27 signaling might evoke antitumor responses in patients. A first humanized anti-CD27 antibody (1F5) was well-tolerated in nonhuman primates and was explored for the treatment of hematologic malignancies overexpressing CD27 (26). Treatment with the 1F5 antibody further inhibited growth of CD27⁺ human lymphoma cells in SCID mice and, in a human CD27-transgenic mouse model, led to an increase in antigen-specific CTL with concomitant antitumor activity (27). The 1F5 antibody is currently in phase I trial for the treatment of hematologic malignancies and solid tumors. Based on the coexpression of CD27 and PD1 on TIL, combining agonistic CD27 antibody with a checkpoint blockade targeting the PD1/PD-L1 axis might be of further interest, because preclinical studies have shown a synergistic effect of combinatorial antibody treatment (28).

Currently, also other treatment strategies for HGSC are being explored including strategies targeting angiogenesis (e.g., bevacizumab) and PARP inhibitors in tumors with alterations in the homologous recombination repair pathway. Whether supplementation of these strategies to current PS or NACT treatment regimens may affect the surgical outcome and overall prognosis of patients remains to be determined. Furthermore, effect it may have on the immune infiltrate and potential synergism with immunotherapy is of interest (29). Exploration of these ideas can help to stratify patients with respect to different treatment modalities to predict the patients who might benefit most from immunotherapy.

In conclusion, the methodology of interpreting TIL infiltrates in tissue of HGSC and the cutoff values for positive samples need to be standardized before it can be considered as a prognostic marker or serve as a selective marker for treatment strategies. Here, we showed that the treatment regimen, surgical outcome, and the differentiation status of TIL should also be taken into account.

H.W. Nijman has a scientific collaboration with the biotech company BioNovion. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.C.A. Wouters, F.L. Komdeur, C.J.M. Melief, T. Daemen, M. de Bruyn, H.W. Nijman

Development of methodology: M.C.A. Wouters, F.L. Komdeur, H.H. Workel, C.J.M. Melief, E.W. Duiker, M. de Bruyn

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.C.A. Wouters, F.L. Komdeur, H.H. Workel, H.G. Klip, A. Plat, N.M. Kooi, G.B.A. Wisman, M.J.E. Mourits, H.J.G. Arts, M.H.M. Oonk, R. Yigit, H. Hollema, M. de Bruyn, H.W. Nijman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.C.A. Wouters, F.L. Komdeur, H.H. Workel, S. de Jong, C.J.M. Melief, H. Hollema, M. de Bruyn, H.W. Nijman

Writing, review, and/or revision of the manuscript: M.C.A. Wouters, F.L. Komdeur, M.J.E. Mourits, M.H.M. Oonk, R. Yigit, C.J.M. Melief, E.W. Duiker, T. Daemen, M. de Bruyn, H.W. Nijman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.G. Klip, A. Plat, N.M. Kooi, G.B.A. Wisman, M.J.E. Mourits, H. Hollema

Study supervision: T. Daemen, M. de Bruyn, H.W. Nijman

The authors would like to thank Klaas Sjollema, Henk Moes, Geert Mesander, Roelof-Jan van der Lei, Tineke van der Sluis, Joan Vos, and Niels Kouprie for their technical assistance.

This work was supported by Dutch Cancer Society/Alpe d'Huzes grant RUG 2014–6719 to M. de Bruyn and Jan Kornelis de Cock Stichting grants to M.C.A. Wouters and F.L. Komdeur. Part of the work has been performed at the UMCG Imaging and Microscopy Center (UMIC), which is sponsored by NWO-grants 40-00506-98-9021 and 175-010-2009-023.

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.