The cohort of patients with cervical cancer (n=66) included paraffin-embedded baseline tumor biopsies from patients with locally advanced cervical cancer (LACC) undergoing curative-intent concurrent chemoradiation followed by uterovaginal brachytherapy boost. They were treated in our institution between March 2004 and August 2011.14 Only squamous cell carcinomas, which is the most frequent histology, were included in the analysis. A second cohort of paraffin-embedded baseline NSCLC surgical samples was evaluated from patients (stage I to III-IV according to 7th edition TNM classification) undergoing primary surgery at Hôtel-Dieu Hospital (Paris, France) between 2001 and 2005.15 All tumor specimens were collected before chemotherapy. The patients’ main clinical and pathological features are described in online supplemental tables S1 and S2. A written informed consent was obtained from all patients, in application with the article L.1121–1 of French law and the two studies were approved by the local ethics committee (CPP 2012–0612).

Pearson’s correlation coefficients with their significance value were calculated and depicted in scatter plots. Kaplan-Meier curves were stratified using the median value of the cohort for PAR and PDXK expression and DC-LAMP and CD8 infiltration and the log-rank test was used to compare overall and relapse-free survival of patients between groups. The start of follow-up was the date of biopsy for patients with LACC and NSCLC. Statistical testing was done at the two-tailed α level of 0.05. Statistics were managed using R software V.3.4.2 (R Foundation for Statistical Computing, Vienna, Austria; https://www.R-project.org/, packages survival, survminer, Hmisc and corrplot).

Culture media and cell culture supplements were purchased from Life Technologies (Carlsbad, California, USA). Cells were routinely maintained at 37°C under 5% CO₂, in the following culture medium: Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum, 100 units/mL penicillin G sodium and 100 mg/mL streptomycin sulfate. Mouse Lewis lung carcinoma (LLC) cell line was purchased from American Type Culture Collection. Cisplatin-resistant clones were obtained in vitro by prolonged culture of parental (also known as wild type (WT)) cells with sublethal concentrations of cisplatin (Sigma-Aldrich). In parallel, control clones were obtained from the untreated WT cells. Both parental and cisplatin-resistant cells were never passaged more than 1 month before use in experimental determinations.

Adherent and non-adherent cells were collected and costained for 30 min at 37°C in 300 µL of culture medium containing 40 nM 3,3’ dihexiloxalocarbocyanine iodide (DiOC₆(3)), Molecular Probes-Invitrogen), a mitochondrial transmembrane potential-sensitive dye, and 1 µg/mL propidium iodide, which only accumulates in cells exhibiting plasma membrane permeabilization. Cytofluorometric acquisitions were carried out on a Milteny cytofluorometer (MACSQuant Analyzer 10), and statistical analyses were performed by using the FlowJo software (LLC, Oregon, USA) on gating on events exhibiting normal forward scatter and side scatter parameters.

Cells were trypsinized, collected, washed twice with phosphate-buffered saline (PBS) and lysed in a buffer containing 50 mM Tris HCl pH 6.8, glycerol 10%, 2% Sodium dodecyl sulfate (SDS), 10 mM ithiothréitol (DTT) and 0.005% bromophenol blue. Subsequently, 30 µg of proteins were separated on 4%–12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels (Invitrogen) and electrotransferred to nitrocellulose membranes (Biorad) followed by immunoblotting with a primary antibody specific for PAR (Clone 10H, mAb to poly ADP-ribose, Abcam, 1:1000). An antiactin antibody (mAb to beta actin, ab 49900, Abcam, 1:5000) was used to control equal lane loading. Thereafter, membranes were incubated with appropriate horseradish peroxides-conjugated secondary antibodies (Southern Biotech), followed by chemiluminescence detection with the ECLTM Prime Western Blotting Detection Reagent (GE Healthcare), before being revealed by the ImageQuantTM LAS 4000 Biomolecular Imager (GE Healthcare Life Sciences). Finally, protein expression was quantified by ImageJ software (NIH, USA).

Eight-week-old female C57Bl/6 mice were purchased from Envigo France. Animals were maintained in specific pathogen-free conditions, at 25°C with 12 hours light/12 hours dark cycles. All animals were used under an approved protocol by the local Ethics Committee (C2EA 05 n° B-75-06-12, protocol 7810–20 l6112810578922v2) under conditions in accordance with the EU Directive 63/2010. Sample sizes were calculated to detect a statistically significant effect. For tumor induction, 1.2–1.7×10⁶ WT and cisplatin-resistant (R) LLC cells were resuspended in 100 µL of PBS and subcutaneously injected in the flank of mice under anesthesia (2.5% isoflurane). The estimation of the tumor area (longest dimension × perpendicular dimension) was measured using a common caliper. Tumor ulceration, weight loss superior to 20% as compared with the beginning of the treatment and poor body condition were considered as endpoints.

When the surface of the tumors derived from WT and cisplatin-resistant LLC cells reached 1 cm², corresponding to approximately 30 days after tumor cells injection, tumors were harvested, weighed and transferred on ice in gentleMACS C tubes (Miltenyi Biotec, USA) containing 1 mL Dulbecco's modified eagle medium (DMEM) medium. Tumors were dissociated mechanically with scissors, then enzymatically using the Miltenyi Biotec mouse tumor dissociation kit and a gentleMACS Octo Dissociator following the manufacturer’s instructions. Then, the homogenates were filtered through 70 µm MACS SmartStrainers (Miltenyi Biotec) and washed twice with PBS. Thereafter, tumor cells were resuspended in PBS and 50 mg of the initial tumor sample were stained with LIVE/DEAD Fixable Yellow dye (Thermo Fisher Scientific). Fc receptors were blocked with anti-mouse CD16/CD32 (clone 2.4G2, Mouse BD Fc Block, BD Pharmingen). Surface staining of murine T-cell population infiltrating the tumor was performed with the following fluorochrome-conjugated antibodies: anti-CD3 APC V450 (clone 17A2, Thermo Fisher Scientific), anti-CD8 PE (clone 53–6.7, BD Pharmingen). Myeloid populations were stained with the following antibodies: anti-CD45 APC-Fire750 (clone 30F-11, BioLegend), anti-Ly-6G PE (clone 1A8, BD Pharmingen), anti-Ly-6C FITC (clone AL-21, BD Pharmingen), anti-CD11b V450 (clone M1/70, BD Pharmingen), anti-CD11c PE-Vio770 (REA754, Miltenyi Biotec) and anti-I-A/E (MHC-II) APC (clone M5/114.15.2, BioLegend). Finally, stained samples were run through a BD LSR II flow cytometer. Data were acquired using BD FACSDiva software (BD biosciences) and analyzed using FlowJo software (TreeStar).

We found a positive effect of tumor infiltration by CD8⁺ T lymphocytes (determined by immunohistochemistry, figure 1A) to be associated with improved overall survival (figure 1C) and a trend for improved relapse-free survival (figure 1D) in patients with LACC stratified by tumor stage. We also determined the density of the tumor infiltrating DC-LAMP⁺ cells (which are activated DCs) by immunohistochemistry (figure 1B). None of the women with a DC-LAMP^high tumor died, meaning that high DC-LAMP⁺ density was significantly (p=0.0081, log-rank test) associated with overall survival (figure 1E) as well as a tendency for improved relapse-free survival (figure 1F) (online supplemental table S3). DC-LAMP^high tumors were densely infiltrated by CD8⁺ T cells (p=0.03 by χ² test, p=0.08 when computed as continuous values) (figure 2A,B). The combination of a low infiltration by both CD8 and DCs carried a significantly worse prognosis compared with tumors highly infiltrated (p=0.003, log-rank test) (figure 2C,D). We also determined the expression level of PDXK (figure 3A,B) and PAR (figure 3C,D) by cancer cells. Of note, we observed a positive correlation (Pearson test, p=0.002) between DC-LAMP density and PDXK expression, as well as a significant (p=0.0034) negative correlation between CD8 density and PAR expression (figure 3E,F). PAR^high/PDXK^low tumors showed a tendency for worse overall (figure 4A) and relapse-free (figure 4B) survival. These results suggest that tumor cell-intrinsic metabolic characteristics may affect local immunosurveillance.

Intrigued by the aforementioned results, we decided to evaluate our observations in a different neoplastic disease, NSCLC, for which it was known that low expression of PDXK and high abundance of PAR are negative prognostic markers.7 10 Pearson analyses revealed a negative correlation (p=0.017) between PAR levels and CD8⁺ T lymphocytes (figure 5A) and a positive trend (p=0.057) between PDXK expression and infiltration by DC-LAMP⁺ cells (figure 5B). As expected,17 18 the detection of high levels of CD8⁺ or DC-LAMP⁺ cells in tumors was associated with improved overall survival (figure 5C,D) (online supplemental table S4), and this effect was independent from age, gender, histology, smoking status and tumor stage in a multivariate Cox model (p=0.001 and <0.001 for CD8 and DC-LAMP). The combined low infiltration by CD8 and DCs have a significantly worse prognosis compared with tumors highly infiltrated (p=0.0001, log-rank test) (figure 5E,F). Patients with PAR^high/PDXK^low cancers had a reduced overall survival, particularly in stage I-II tumors (figure 6), and this effect was independent from age, gender, histology, smoking status and tumor stage (p=0.02). Altogether, these results confirm the conjecture that tumor cell-intrinsic metabolic alterations may impact the composition of the local immune infiltrate.

In human cervical and lung cancers, high abundance of PAR correlated with a reduced density of CD8⁺ T lymphocytes in the tumors. To establish a causal relationship between high PAR levels and reduced tumor immunosurveillance, we took advantage of Lewis lung cancer (LLC) cells (a mouse NSCLC cell line) that had been selected in vitro for cisplatin resistance, hence enhancing their capacity to maintain mitochondrial function and viability in vitro (figure 7A,B). Such cells manifested upregulation of PAR, as detectable by immunoblot analyses (figure 7C). Parental WT (online supplemental figure S1A,B) and cisplatin resistant (R) clones were injected subcutaneously into mice and their immune infiltrate was characterized by immunofluorescence and flow cytometry (online supplemental figures S1C and S2) when the tumors had reached a surface of 1 cm². Importantly, the density of CD8⁺ T cells was reduced for resistant (PAR^high) tumors (p=0.028). Moreover, the LLC PAR^high tumors were less infiltrated in antigen presenting cells, (i) activated DCs (CD45⁺CD11c⁺MHCII⁺) (p=0.0859) and (ii) myeloid cells (CD45⁺CD11b⁺Ly6G⁻ Ly6C^{low/intermediate}) (p=0.0012) as compared with their PAR^low counterparts (figure 7D–F). These results indicate that the metabolic phenotype of tumor cells may indeed shape the tumor microenvironment at the immunological level.