A rationale for surgical debulking to improve anti-PD1 therapy outcome in non small cell lung cancer

Anti-PD1 immunotherapy has emerged as a gold-standard treatment for first- or second-line treatment of stage IV NSCLC, with response rates ranging from 10 to 60%. Strategies to improve the disease control rate are needed. Several reports suggested that debulking surgery enhances anti-tumor immunity. We aimed at examining tumor burden as a predictive factor of anti-PD1 tretment efficacy and to evaluate the role of cytoreductive surgery in anti-PD1 treated NSCLC. Immunocompetent DBA/2 mice engrafted with various amount of allogeneic lung squamous cancer KLN-205 cells were treated with anti-PD1 monoclonal antibody. Mice engrafted with two tumors also underwent a debulking surgery or a sham procedure. Tumor volume was monitored to assess treatment efficacy. Tumor infiltrating lymphocytes were assessed by flow cytometry. In a retrospective study of 48 stage IV NSCLC patients treated with Nivolumab who underwent a 18-FDG PETscan before treatment onset, the prognostic role of metabolic tumor volume was analysed. Anti-PD1 treatment effect was greater in mice bearing smaller tumors. Treatment with higher doses of anti-PD1 antibody did not improve the outcome, independently of the size of the tumor. In mice bearing 2 tumors, excision of 1 tumor improved the anti-PD1 treatment effect on the remaining tumor. In 48 NSCLC patients receiving anti-PD1 treatment, high metabolic tumor volume was associated with poor overall survival and the absence of clinical benefit. Treg infiltration, but not effector T cells, was positively correlated to tumor volume. Taken together, our results suggest that tumor volume is a predictive factor of anti-PD1 efficacy in NSCLC. Additionally, an experimental murine model suggests that tumor debulking may improve control of residual tumor.

Anti-PD1 immunotherapy has emerged as a gold-standard treatment for first-or second-line treatment of stage IV NSCLC, either in monotherapy or in combination with chemotherapy. However, the majority of patients experiences primary resistance. Strategies to improve the disease control rate are needed. Among them, combination with other immunotherapeutics agents, chemotherapies and/or anti-angiogenic drugs have been reported. The combination of radiotherapy with anti-PD1 treatment is also under investigation, based on a strong biological rationale and case reports of abscopal effect.
Cytoreductive "debulking" surgery has long been considered a standard of care for eligible patients with metastatic renal cell carcinoma and is still widely recommended for selected patients 1,2 . This recommandation relies on many retrospective data and on two prospective clinical trials showing an improved overall survival in patients undergoing nephrectomy before receiving Interferon therapy 3,4 . The mechanisms by which nephrectomy may improve survival in this setting are unclear 5 . Several reports suggested that debulking surgery enhances anti-tumor immunity: (i) reports of spontaneous metastasis regression in patients who underwent primary tumor removal [6][7][8] , (ii) animal data showing lower immune cytotoxicity and reduced B-cell funciton when the primary tumor is not removed 9,10 and (iii) studies showing enhanced cell-mediated immunity in patients after nephrectomy [11][12][13] .
More recently, Oronsky et al. reported three cases of abscopal effect after surgery in patients that were previously treated with an anti-PD1 agent 14 . In all three cases, tumor was progressing at the time of surgery. For two of them, anti-PD1 was continued after surgery. In this setting, surgery may contribute to the enhancement of For each tumor allograft, cells were collected by trypsinisation and washed twice in PBS. 1.10 4 to 1.10 7 cells were resuspended in 100 μL PBS and subcutaneously injected in the flank of a 6 to 12-week-old female DBA/2 mouse (Charles River, L' Abresles, France).
Tumor engraftment led to visible tumors at 1 week in 95% of mice. Tumors were measured every 2-3 days using a caliper by researchers blinded for treatment allocation. Tumor volume was calculated as [length*(width) 2 ]/2.
At day 35 after tumor allograft, mice were euthanized and tumors were harvested for further analysis. Euthanasia was performed immediately after the last procedure by mean of 0.8 g/kg Phenobarbital intra-peritoneal injection.
Treatment regimens. Treatment allocation group was random. Random allocation sequence was generated using a random numbers table and the animal ID numbers. Six animals were used in each group (i.e. mice receiving the same number of cells and the same treatment). Mice received either anti-PD1 (J43 clone, BioXCell, West Lebanon, New Hampshire, USA) or isotype IgG (BioXCell), 100 µg every 7 days or 200 µg twice a week, starting on day 14 after tumor allograft. Treatment administration was not blinded but tumor measurement was.
In the debulking experiments, mice underwent procedure under anesthesia at day 14. A preventive intraperitoneal injection of 0.1 mg/kg Buprenorphine was given before the procedure. Animal anesthesia was performed with inhaled isoflurane. The right flank area was shaved and cleaned with 0.2% chlorhexidine. A 0.5 to 1 cm skin incision was performed ahead of the right flank tumor. Excision of the tumor was performed in half of the mice. Flow cytometry. KLN-205 tumors from DBA/2 allografted mice (n = 9) were harvested at day 21 after treatment initiation and measured using a caliper to determine tumor volume. Single cell suspension of the harvested tumor was prepared with Gentle MACS dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany) and the corresponding Tumor Dissociation Kit (Miltenyi Biotec), according to manufacturer's instructions. Cell suspension was passed through a 70-μm cell strainer. Cells (5.10 5 ) were incubated with 5 μl/ml purified rat anti-mouse CD16-CD32 mAb (TruStain fcX, Biolegend Inc, San Diego, CA) before incubation with specific anti-mouse antibodies. Cells were marked with CD3-FITC, CD4-APC-Cy7, and CD25-PE (Biolegend Inc), CD8-PerCP-Cy5.5 and FoxP3-APC (eBioscience Affymetryx, Santa Clara, CA, USA), washed with PBS and fixed with 0.5% paraformaldehyde. Flow cytometry analysis was performed on FACS LSR Fortessa (BD biosciences, San Jose, CA, USA), and using FlowJo software (Tree Star, Ashland, OR). Cells were gated, based on forward and side scatter and on living cells. Acquisition of multiparameter data was carried out with an appropriate forward scatter (FSC) threshold to exclude debris. At least 50,000 CD3+ cells per sample were analyzed. Gating strategy is illustrated in Supplementary Fig. S1. T regulatory cells were defined as Foxp3/CD4/CD3 positive cells of all living cells.
Clinical study. All patients who were treated with Nivolumab as second-or third-line treatment for stage IV NSCLC at the Rouen University Hospital between February 2015 and July 2017 were screened. Patients who underwent 18-FDG PET-scan within the 3 months prior to Nivolumab treatment onset were selected for further analysis. Clinical data, imaging, pathology results and molecular analysis were collected from the electronic medical record. 18-FDG PET-scan were reviewed and Metabolic tumor volume was determined using PET VCAR semi-automatic software of AW server 3.2 (General Electric ® , Milwaukee, USA) for segmentation. Non malignant FDG avidity areas were not included in this analysis. MTV cut-off to segregate low versus high MTV patients was determined using ROC curve analysis for overall survival.
The protocol received approval from our Institutional Review Board (CHB review board for non-interventional research, agreement #1809B).

Results
To investigate the influence of tumor burden on the efficacy of anti-PD1 agent, we injected 1.10 4 to 1.10 7 KLN-205 cells in allogeneic DBA/2 mice that we treated with intra-peritoneal injection of anti-PD1 monoclonal antibody. In all groups, the anti-PD1 treatment resulted in slower tumor growth (Fig. 1). In mice receiving fewer tumor cells, the treatment effect was greater than in mice receiving 1.10 7 cells (relative tumor volume at day 21 after treatment initiation: 1.97 vs 4.26, p < 0.05). Treatment with higher doses of anti-PD1 antibody did not improve the outcome, independently of the number of cells engrafted (Fig. 1).
In order to test the synergistic effect of debulking surgery and anti-PD1 treatment, we engrafted mice with two tumor grafts: 1.10 4 or 1.10 7 cells in the right flank and 1.10 4 cells in the left flank. At day 14 after engraftment, mice were randomly allocated to either a debulking surgery (group B and C) of the right tumor or a sham procedure (group A) ( Fig. 2A). Mice were treated with 100 µg per week intra-peritoneal injection of isotype IgG (group B) or anti-PD1 (groups A and C). Here again, the anti-PD1 treatment effect was greater in mice receiving fewer tumor cells (Fig. 2B). When mice underwent debulking surgery and received anti-PD1 treatment (group C), the growth of the remaining left tumors was slower than the left tumors in group A (sham procedure and anti-PD1 treatment, p = 0.01). In mice that underwent the sham procedure, the growth of the left tumors was slightly faster than in mice from the corresponding debulking groups (p = 0.04).
In patients, we examined the role of metabolic tumor volume (MTV) as a surrogate marker of tumor burden and aimed to test wether MTV was associated with patients' outcomes in second-line treatment with Nivolumab for stage IV NSCLC. Among 200 patients treated in this setting at our institution from February 2015 to July www.nature.com/scientificreports www.nature.com/scientificreports/ 2017, 48 had undergone 18FDG PET-scan prior to treatment. In this cohort, high MTV was associated with poor overall survival (p = 0.01) ( Table 1).
We studied the immune infiltrate of KLN-205 mouse tumor allografts. Treg infiltration was positively correlated to tumor volume (Fig. 3). There was no association between tumor volume and total lymphocytes, CD8+ lymphocytes and CD4+ lymphocytes.

Discussion
Our study showed that tumor volume is a predictive factor of anti-PD1 efficacy in lung cancer in a mouse model and based on clinical data. Moreover, partial tumor excision enhanced anti-PD1 treatment outcome in the mouse model.
Tumor volume is considered in the TNM classification of NSCLC through the T descriptor. In fact, the size of the primary tumor site has been shown to be a strong prognostic factor in NSCLC 15 . Here we used metabolic  www.nature.com/scientificreports www.nature.com/scientificreports/ tumor volume as a surrogate marker of tumor volume in stage IV NSCLC patients and showed that it is an independent risk factor for poor survival. In melanoma, a recent report showed similar findings, with baseline tumor size being an independent prognostic factor for overall survival in patients treated with the anti-PD1 Pembrolizumab 16 .
Large tumors may lead to an immunosuppressed or tolerogenic state. We showed that bigger tumors have higher densities of T regulator lymphocytes, but similar densities of effector T cells. This may contribute to establish a tolerogenic state in large tumors. Other mechanisms of immunosuppression related to tumor volume have been reported, including metabolic competition between immune cells and cancer cells 17 and hypoxia-induced recruitment of immunosuppressive cells and dysfunction of effector immune cells 18 .
Debulking surgery in metastatic renal cell carcinoma has long been considered a standard of care for selected patients but is not recommended in NSCLC. Nevertheless, it may be considered in the setting of immunotherapy. In fact, surgical debulking may enhance the immune system efficiency through the elevation of T-lymphocyte to tumor cell ratios, the destruction of physical barriers which prevent immune-cell infiltration of tumor, and the reduction of systemic release of immunosuppressive cytokines such as GM-CSF, IL-10, and PGE2 19 . Moreover, it has been shown that therapeutic vaccination achieves better results in small residual tumors than in larger tumors 20 . In a mouse model of mesothelioma, Broomfield et al. showed that partial resection of tumor enhances the effect of chemo-immunotherapy using gemcitabine and anti-CD40 antibody 21 . Tumor resection was also reported to exert a protective tumor specific CD8+ immunity in the same mesothelioma model 22 . In our study, tumor debulking achieved tumor growth delay while anti-PD1 treatment was ongoing. The effect was greater when the residual tumor was smaller. This is consistent with the abovementioned reports.
One limitation of our mouse model is that we did not take metastasis into account. In fact, tumor burden may have been underestimated since we could not perform assessment of metastasis at the time of debulking. Nevertheless, treatments were allocated randomly and post-mortem examination of control animals showed similar tumor burden between individuals (data not shown).
So far, tumor debulking in stage IV NSCLC was not considered achievable because (i) most patients have widely metastatic disease and altered PS, (ii) thoracic surgery was associated with high morbidity and (iii) onset of chemotherapy has to be delayed after surgery for recovery purpose and to lower the risk of post-operative infections, preventing efficient treatment of the non-debulked disease. Nevertheless, the development of minimally invasive techniques has minimized the morbidity and recovery time of thoracic surgery, while the good toxicity profile of immunotherapy allows its peri-operative administration. Thus, our results support the idea of tumor debulking in the context of anti-PD1 immunotherapy, in selected patients such as those with good PS and oligo-metastatic disease.
Taken together, our results suggest that tumor volume is a predictive factor of anti-PD1 efficacy in NSCLC. That effect may be mediated by T regulator lymphocytes. Tumor debulking appeared to improve treatment outcome in a mouse model and warrants further investigations in the clinical setting.
18FDG PET-scan data from patients treated with anti-PD1 for stage IV NSCLC was reviewed. MTV cut-off to segregate low versus high MTV patients was determined using ROC curve analysis for overall survival (AUC = 0.81, p = 0.02). Kaplan-Meier method was used to estimate progression-free survival. MTV, age, histology and Performance Status were compiled in a proportional hazard Cox model to perform multivariate analysis. Comparisons of the 6-months PFS and 1-year OS between groups were performed using Student t-test.