Effects of anti-PD-1 immunotherapy on tumor regression: insights from a patient-derived xenograft model

Immunotherapies, such as checkpoint blockade of programmed cell death protein-1 (PD-1), have resulted in unprecedented improvements in survival for patients with lung cancer. Nonetheless, not all patients benefit equally and many issues remain unresolved, including the mechanisms of action and the possible effector function of immune cells from non-lymphoid lineages. The purpose of this study was to investigate whether anti-PD-1 immunotherapy acts on malignant tumor cells through mechanisms beyond those related to T lymphocyte involvement. We used a murine patient-derived xenograft (PDX) model of early-stage non–small cell lung carcinoma (NSCLC) devoid of host lymphoid cells, and studied the tumor and immune non-lymphoid responses to immunotherapy with anti-PD-1 alone or in combination with standard chemotherapy (cisplatin). An antitumor effect was observed in animals that received anti-PD-1 treatment, alone or in combination with cisplatin, likely due to a mechanism independent of T lymphocytes. Indeed, anti-PD-1 treatment induced myeloid cell mobilization to the tumor concomitant with the production of exudates compatible with an acute inflammatory reaction mediated by murine polymorphonuclear leukocytes, specifically neutrophils. Thus, while keeping in mind that more research is needed to corroborate our findings, we report preliminary evidence for a previously undescribed immunotherapy mechanism in this model, suggesting a potential cytotoxic action of neutrophils as PD-1 inhibitor effector cells responsible for tumor regression by necrotic extension.

Study design. Details on the establishment of the human squamous NSCLC PDX model used for this study and on the preliminary test of response to anti-PD-1 therapy are shown in text file S1 in Supplementary Material.
Pharmacological intervention. Transplanted p2 mice were monitored and their bilateral tumor volumes were measured (Fig. 1). When the tumors reached the appropriate size (~100 mm 3 ), mice were randomly assigned to the following experimental groups (4 mice [and thus 8 tumors]/group) and the treatment was initiated (twice weekly for 6 consecutive weeks -from day 0 to 42): isotype control, cisplatin (monotherapy), anti-PD-1 (monotherapy), cisplatin + anti-PD-1 (concomitant group), sequential treatment with cisplatin and anti-PD-1 (cisplatin → anti-PD-1) or vice versa (anti-PD-1 → cisplatin) (Fig. 1). Finally, 2 days after the last treatment administration (day 46), blood samples were taken from the tail vein and collected in EDTA tubes and the mice were sacrificed as described above. The tumor tissue was harvested and divided into pieces, which were freshly conserved and embedded in Tissue-Tek OCT compound (Sakura Finetechnical Co., Ltd.; Tokyo, Japan) or in paraffin for subsequent analysis.
Outcomes. Tumor stability analysis during consecutive passagingTo verify that the histopathological features of the xenograft tumors were stable and similar to patients´ tumors during the subsequent passages, we performed several analyses. Tumors were fixed (10% formalin), embedded in paraffin (PANREAC Applichem; Darmstadt, Germany), sectioned at 4 µm, immunostained using the Dako Cytomation autostainer (Dako Diagnostics; Barcelona, Spain) or the Leica Bond-Max System (Leica Microsystems; Wetzlar, Germany), and counterstained with hematoxylin and eosin (H&E). Histopathological analysis was then carried out to assess the type and histologic tumor subtype, the degree of cellular differentiation and the tumor infiltration. The differential diagnosis of NSCLC was made on paraffin sections with an immunophenotype panel (Table S1 in Supplementary Material, IHC markers shaded in light gray). The immunophenotype analysis was compared with the expression pattern obtained from the respective patient. The tumor expression of hCD45, hPD-1 and hPD-L1 was also determined by IHC (Table S1 in Supplementary Material, markers shaded in dark gray). In addition, to confirm that the tumor cells were human, we analyzed the presence of Alu sequences on paraffin-embedded tissue using the Alu Positive Control Probe II (Ventana Medical Systems Inc.; Roche Diagnostics; Mannheim, Germany) for the automated Ventana BenchMark Instrument.
Tumor volume and tumor growth rateTumor volumes were measured twice weekly throughout the treatment period and at the day of sacrifice using a caliper, as previously mentioned. We assessed the response to therapy in terms of tumor volume using the rate of tumor growth, which refers to the percentage of tumor growth with respect to the volume at the beginning of the therapy, calculated by the formula (TVdx/TVd0) × 100, where TVdx refers to the tumor volume measured on a specific day and TVd0 is the tumor volume at the beginning of treatment administration (set at 100%). Tumor growth curves were generated for each mouse.
Necrotic indexTo assess tumor regression in response to therapy we calculated the percentage of necrotic areas in paraffin-embedded sections stained with H&E. Four sections of each tumor were scanned and the necrotic areas were quantified using CaseViewer software (3DHISTECH Ltd., Budapest, Hungary) with the formula % necrosis area = (Σ necrosis area/total tumoral mass area) × 100.
Cell identification in tumor-derived fluids (in vivo)We observed some tumor-derived fluids during tumor harvesting in vivo and tumor fragmentation ex vivo (see Results). The fluids were collected, fixed in 95% ethanol for 15 minutes, and stained using the standard trichrome Papanicolaou method. Fluids were analyzed by liquid cytology using the ThinPrep Pap Test (Cytyc Corporation; Boxborough, MA, USA).
Leukocyte identification in peripheral blood (in vivo) and in tumor stroma (ex vivo)Complete peripheral blood (conserved in EDTA tubes) and cell suspensions from tumor tissue were processed by flow cytometry using a FACSCanto II instrument and FACSDiva software v6.1.2 (BD Biosciences; Franklin Lakes, NJ, USA). Individual fresh tumors (~0.1 g) were homogenized, digested with 1 mg/ml collagenase D (Roche Diagnostics; Mannheim, Germany) for 24 hours and filtered through a 40-μm nylon mesh cell strainer (BD Bioscience). Flow cytometry analyses were performed using fluorochrome-conjugated monoclonal antibodies for human and mouse antigens (Table S2 in Supplementary Material). Dead cells were excluded by 7-aminoactinomycin D staining. Specific antibodies for human and mouse immune cells were used for classification of the different cellular populations (the combination of antibodies are shown in Table S3 in Supplementary Material).
Neutrophil nuclear morphology after exposure to anti-PD-1. Nuclear morphology was studied in neutrophils from tumor exudates treated with anti-PD-1 using liquid citology (ThinPrep Pap Test), as well as in the immunofluorescence images of isolated neutrophils exposed to anti-PD-1, applying a negative photo filter with the invert option in Adobe Photoshop CC (version 2017.0.0; Adobe Systems Inc., San Jose, CA, USA).
Immunofluorescence experiments after exposure to anti-PD-1. A suspension of isolated and purified neutrophils was exposed to either anti-PD-1 (nivolumab) or isotype control, both at 50 µg/ml for 1 hour. Subsequently, neutrophils were fixed with 4% paraformaldehyde and incubated with fluorescein isothiocyanate (FITC)-conjugated anti-hIgG for 45 minutes (Table S4 in Supplementary Material) to detect the binding of anti-PD-1, through its antigen-binding fragment (Fab) variable region, to the PD-1 receptor of the neutrophil membrane cell surface. For a second immunodetection, the samples were incubated with a recombinant human PD-1 protein (active)-phycoerythrin (PE) ligand (Abcam; Cambridge, UK [reference # ab246145]) applied for 1 hour and using a concentration of 1/500 diluted in PBS (Table S4 in Supplementary Material). This second staining was used to detect if anti-PD-1 had bound, through its fragment crystallizable (Fc) region, to Fcγ receptors (FcγR) in the neutrophil cell membrane surface, thereby leaving the ligand binding site free. After washing the cells with PBS (twice), they were mounted with a mixture of PBS:glycerol. These experiments were also performed exposing neutrophils for 1 hour to an isotype control, human immunoglobulin G4 (hIgG4).
Images were collected with a TCS SP5 confocal microscope (Leica Microsystems; Wetzlar, Germany) using a 63× HCX PL APO (1.4 numerical aperture) and processed with the ASF Leica software (Leica Mycrosystems) as detailed elsewhere 20 . Excitation and emission parameters: 488 nm/500-540 nm and 546 nm/557-572 nm, for FITC−conjugated anti-hIgG and for PD-1−PE, respectively. Statistical analysis. Data are presented as mean ± SEM for all figure panels in which error bars are shown.
Tumor growth curves were compared between treatment groups with the non-parametric Kruskal-Wallis test and statistical significance was set at 0.05. The extent of necrosis was expressed as the percentage of necrotic tissue and between-group analysis was performed using the Kruskal-Wallis test. Analysis was performed using IBM SPSS 22.0 software (SPSS, Inc.; Chicago, IL, USA). All graphics were made with GraphPad Prism 6, version 6.01 software (GraphPad Software; San Diego, CA, USA).

Results
Tumor stability during consecutive passaging in mice. Fresh tumor tissues were obtained from 17 patients to generate the PDX models (clinical patient data is shown in Table S5 in Supplementary Material). Histopathological characteristics of the passaged NSCLC PDX tumors remained stable in morphology, histological type and tumor cell differentiation grade (Fig. S1 in Supplementary Material).
Sensitivity response test to anti-PD-1 therapy. As shown in the flow diagram ( Fig. S2 in Supplementary Material), a total of 10 NSCLC PDX lines were established. Of those, PDX lines derived from squamous cell lung cancer were chosen, and tumors with a known driver mutation were excluded because of the effect the mutation may have on the treatment outcomes. We also excluded PDX lines that showed tumor degeneration with keratinized phenotypes, because of the difficulty of adequately assessing the tumor sensitivity to therapy caused by the massive production of keratin in these cases. Some PDX lines included mice with atypical Epstein-Barr virus-associated human B cell lymphomas and, for this reason, they were also not used. Finally, only one PDX line showed a clear response in view of changes in tumor volumes in the grafts, and this was selected as a model to study the effect of the immunotherapy treatment in the absence of lymphocytes. The tumor sample of this PDX line (PDX4) came from a 79-year-old Caucasian male patient who was submitted to a right upper lobectomy with complete resection (Table S5 in Supplementary Material). According to WHO criteria for histological classification and staging, the tumor was a basaloid infiltrating and poorly differentiated squamous cell lung carcinoma, and the patient was staged as IIA (pT2a N1 L1 M0).
We also present results from an additional PDX line -PDX6, which was a non-responder to immunotherapy -derived from a patient whose characteristics are shown in Table S5 in Supplementary Material. Tumor volume and tumor growth rate. We analyzed the tumor growth rates in all experimental groups of PDX4 from the beginning to the end of the study ( Fig. 2A). Briefly, in the anti-PD-1 group, we found a progressive increase in the tumor volume over time as compared with the other treatment groups (p = 0.011 for the group effect), and a similar pattern was observed in the isotype control group, showing the natural evolution of the disease. As anticipated, growth rates in the cisplatin-administered group were significantly lower than in the isotype control group. However, the lowest tumor growth rate was found in the cisplatin + anti-PD-1 (concomitant) group. With regard to the sequential treatments (anti-PD-1 → cisplatin and vice versa), the tumor growth rates followed a specular path, with an increase in tumor volumes during the period of anti-PD-1 administration.
Considering the tumor volume at sacrifice, we found the highest volume in both the anti-PD-1 and isotype control groups and the lowest volume in the concomitant treatment group (Fig. 2B, where all the numerous post hoc between-group differences are shown), which was even lower than that measured in the cisplatin group. Tumor volumes between the sequential treatment groups were not significantly different; however, the final Scientific RepoRtS | (2020) 10:7078 | https://doi.org/10.1038/s41598-020-63796-w www.nature.com/scientificreports www.nature.com/scientificreports/ tumor volume in the anti-PD-1 → cisplatin group was lower than that in the isotype group and higher than in the concomitant group (Fig. 2B).
However, as shown in Figs. S3A,B in Supplementary Material, no changes were found in the tumor volume and growth rate of PDX6 mice with regard to anti-PD-1 monotherapy.
Necrosis. Given these results, in particular the increase in tumor volume of PDX4 after anti-PD-1 administration, we next evaluated the extent of tumor necrosis for the different treatments. After calculating the necrosis index in the different experimental groups (expressed as the percentage of necrotic areas by total surface estimated by H&E staining), we found a significant group effect (p = 0.014) with the highest values observed in the concomitant treatment group (anti-PD-1 + cisplatin) and the lowest values in the isotype control group (corresponding to normal conditions of squamous histology) (Fig. 2C).
Concomitant treatment increased tumor necrosis and significantly decreased the viable component of the tumor (Fig. 2C, where all the numerous post hoc between-group differences are shown, and Fig. S4A in Supplementary Material), achieving the greatest tumor regression, which was superior to that achieved by anti-PD-1 → cisplatin sequential treatment. However, the aforementioned sequential treatment group presented a significantly higher necrotic index than the isotype control group (p = 0.043) but statistical significance was not reached when compared to the cisplatin → anti-PD-1 group. There was a trend for an increase in the necrotic index in the anti-PD-1 group as compared with the isotype control group (p = 0.068). Overall, the percentage of necrotic areas tended to be higher in all experimental groups that included the anti-PD-1 and/or cisplatin treatment as compared with the isotype control group. By contrast, in the PDX6 model ( Fig. S3C in Supplementary Material) we only found a significant difference (p = 0.043) in the percentage of necrotic areas after anti-PD-1 → cisplatin sequential treatment compared with the isotype control group, with no differences between the latter and the remainder of groups.

Identification of cells in tumor-derived fluids.
Tumors treated with anti-PD-1 contained a fluid with a serous appearance at the time of tumor removal (Fig. 3A), and this was also evident through a break in the skin www.nature.com/scientificreports www.nature.com/scientificreports/ during in vivo measurements (Fig. 3B). Because the fluid was only observed in the anti-PD-1 group (in both monotherapy and in combination), it was unlikely that this was due to necrosis associated with the tumor itself. Indeed, tumors from the anti-PD-1 + cisplatin group exuded fluids prominently (Fig. 3B), despite presenting the smallest tumor volume among the experimental groups. Moreover, although the tumors treated with anti-PD-1 in monotherapy were the largest at the end of the experiment ( Fig. 2A), many of them were liquified (Fig. 3A) and had large internal cavities. After two expert pathologists had analyzed all the samples with liquid cytology, we determined that this fluid corresponded to a cellular exudate compatible with an acute inflammatory reaction, as reflected by the presence of debris of dead squamous cells and a large number of myeloid cells or polymorphonuclear (PMN) leukocytes specifically neutrophils, as well as debris of dead squamous cells (Figs. 3C,D).

Leukocyte identification in peripheral blood and in tumor stroma.
We tested for the presence of human leukocytes (hCD45+) in peripheral blood in the study groups by flow cytometry, which was negative in all cases (Fig. S5A in Supplementary Material). However, consistent with our observations in the tumor/cellular exudates in the anti-PD-1 groups, we observed the accumulation of inflammatory PMN cells coinciding with the extensive necrotic areas in tumor sections stained with H&E (Figs. 4A, S4B,C in Supplementary Material). By contrast, no such phenomenon was observed in the tumors of the non-responder (PDX6) mouse line after anti-PD-1 treatment (Figs. 4B, S4D,E in Supplementary Material). In the latter, we found no inflammatory infiltrate in necrotic areas, which solely included debris of necrotized tissue and dead epithelial (carcinoma) cells.
To ascertain whether aforementioned PMN cells found in tumor exudates in were possible candidates for anti-PD-1 therapy effector cells, we first determined whether they were lymphocytes (a priori, the immunotherapy target cells). However, after several different analyses, we determined that these cells were neutrophils, as detailed below.
Using a human-specific Alu-sequence, we determined that the xenograft stroma included both human and mouse stromal components, (Figs. S5D,E in Supplementary Material). Further, IHC analysis using a panel of antibodies ruled out the presence of specific lymphocyte subpopulations in tumors, treated or not with anti-PD-1 (Fig. S5F in Supplementary Material).
We next utilized flow cytometry panels to examine the infiltrated leukocyte component from tumor homogenates as well as from fluids collected from some tumors treated with anti-PD-1. In both cases, analysis showed 0% human (hCD45+) and 100% murine (mCD45.1+) leukocyte components, and the results were very similar Neutrophil nuclear morphology. Liquid cytology and subsequent immunofluorescence analysis performed in tumor exudates from mice treated with anti-PD-1 allowed us to study if the morphology of PMN after treatment exposure corresponded to that of neutrophils. As shown in Fig. S6 in Supplementary Material, a multilobed and hypersegmented nucleus that is characteristic of neutrophils, and specifically of neutrophils with the N1 (i.e., anti-tumoral) phenotype, was recognized in the analyzed images. Confocal microscopy images corroborated this finding.

Neutrophil activation in tumors.
Given the identification of neutrophils as the potential anti-PD-1 therapy effector cells, we evaluated elements of the neutrophil oxidant pathway, specifically MPO and nitric oxide (NO), in tumors from the cisplatin and anti-PD-1 (monotherapy) groups and from the isotype control group, as a possible mechanism for the cytotoxic action of neutrophils during immunotherapy treatments. As shown in Fig. 5A, neutrophil infiltration, measured by MPO staining, coincided with the necrotic areas from tumors treated with anti-PD-1, which was not the case with the isotype control group. Also, necrosis in the cisplatin group was accompanied by tissue degradation and structural disruption (see TO-PRO 3 staining in cisplatin treatment, Figs. 5A,B). We assessed the presence of NO indirectly by evaluating nitrotyrosine modification. Whereas the expression of nitrotyrosine was essentially absent in the isotype control group, some labeling could be detected in the groups treated with cisplatin (Fig. 5B). By contrast, there was a very strong labeling of nitrotyrosine in the anti-PD-1 group (Fig. 5B), which was particularly evident around the necrotic areas where presumably NO is released.

Immunofluorescence experiments in tumors: Identification of tumor-infiltrating cells. Double
immunofluorescence experiments revealed the co-localization of MPO and nitrotyrosine in neutrophils from tumors treated with anti-PD-1 (Figs. 5C and S7A in Supplementary Material), but not in the isotype control or the cisplatin groups (Fig. S7B in Supplementary Material), which would suggest that NO production in neutrophils is a response to anti-PD-1 treatment. In accord with this, nitrotyrosine and MPO co-localization was detected in the sequential anti-PD-1 → cisplatin treatment group, but to a lesser extent than that in monotherapy treatment with anti-PD-1 (Supplementary Fig. S7B).
Our results so far suggest that neutrophils are the effector cells in response to anti-PD-1 treatment in the PDX NOD-SCID gamma model. To address the underlying mechanisms implicated in this response and the target "location" of the therapy, we used IHC to analyze human PD-1 and PD-L1 expression, with the aim of identifying the anti-PD-1 target and its ligand in treated and untreated tumor samples. However, the commercial antibodies showed no reactivity in our assays (data not shown). We had more staining success with anti-PD-1 treatment antibody as a primary antibody, which was revealed with an anti-hIgG secondary antibody. We detected PD-1-like binding sites in tumor sections of the isotype control and the anti-PD-1 treatment groups, but not in the cisplatin group (Fig. 6A). The anti-PD-1 labeling was localized to the cell membrane surrounding the necrotic areas within the tumor and was more delocalized in the cytoplasm (Fig. 6B). The anti-PD-1 monotherapy tumor showed stronger labeling than the tumors from the other experimental groups (Figs. 6A,B). Double immunofluorescence staining using the anti-PD-1 treatment antibody together with the anti-MPO antibody identified neutrophils as the target cells for anti-PD-1 (Fig. 6C). Finally, we assessed anti-PD-1 treatment antibody binding to the original patient tumor tissue by immunofluorescence (Fig. S7C in Supplementary Material), which indicated the presence of immunotherapy binding sites, presumably PD-1 receptors. However, their expression in the tumor tissue of the patient could not be confirmed by conventional IHC using the available commercial anti-PD-1 antibody.

Immunofluorescence experiments in tumors: Identification of anti-PD-1 binding sites in neutrophils.
The degree of the neutrophil purity achieved in whole peripheral blood from NOD-SCID gamma mice assessed by flow cytometry (CD11b+ Ly6G+ cells) was 99%. To identify the specific receptors through which anti-PD-1 binds to the neutrophil membrane surface, a double immunostaining was applied to analyze both FcγR and PD-1 receptors by confocal microscopy (schematic representation of the design shown in Fig. 7A). We observed that (i) anti-PD-1 (shown in green colour) bound to the PD-1 receptors located in the cell membrane surface of isolated murine neutrophils, through its  www.nature.com/scientificreports www.nature.com/scientificreports/ Supplementary Material). In the control experiment there was total absence of immunofluorescence staining. These data indicate that neutrophils might be potential targets of treatment with anti-PD-1 immunotherapy.

Discussion
Immunotherapies such as PD-1 inhibitors in lung cancer have resulted in unprecedented improvements in patient survival 1,2 . However, these therapies do not benefit all patients and many issues remain to be worked out, including the mechanisms of action and the possible effector function of immune cells from non-lymphoid lineages. We based our study on an early-stage NSCLC PDX model, which allowed us to develop sensitivity assays to anti-PD-1 therapy in the absence of a reconstituted immune system in NOD-SCID gamma mice. An antitumor effect was observed in animals that received anti-PD-1 treatment, alone or in combination with cisplatin, possibly due to a mechanism independent of T lymphocytes, which has not previously been described. The anti-PD-1 treatment induced myeloid cell mobilization to the tumor, together with the production of exudates compatible with an acute inflammatory reaction mediated by murine PMNs, specifically neutrophils. Accordingly, we have provided preliminary evidence for a new immunotherapy mechanism, suggesting a potential cytotoxic action of neutrophils as PD-1 inhibitor effector cells that might be responsible for tumor regression by necrotic extension.
Only about 20% of patients with NSCLC receiving immunotherapy respond to treatment 8,9,21 . Because the purpose of this study was to investigate immunotherapy sensitivity in the absence of lymphocytes, an anti-PD-1 responder PDX line was selected. Of note, a non-responder line (PDX6) was also studied. As expected, chemotherapy (cisplatin) alone produced a good response, with tumor regression >50% in terms of necrotic extension, and stabilized the tumor graft growth rate along the treatment. By contrast, the response to anti-PD-1 therapy, alone or sequentially combined with cisplatin, was paradoxical, and led to an increase in tumor growth rate (in the anti-PD-1 phase) with large and friable tumors in some cases, which were associated with exudates containing inflammatory PMNs from areas of reactive necrosis. This phenomenon is reminiscent of unconventional responses of checkpoint inhibitor-based immunotherapy, such as pseudoprogression, which can be observed in patients´ tumors treated with this type of immunotherapy [22][23][24] . These 'paradoxical' or 'unconventional' responses are associated both with immune cell (PMNs and lymphocytes) recruitment and with the intratumoral inflammatory environment triggered by those cells 22,23 . We also found an inflammatory necrotic process accompanied by fluids and exudates in the concomitant chemo-immunotherapy (anti-PD-1 + cisplatin) treatment group, although the tumor growth rate and volume were reduced drastically. The best results in terms of tumor regression were found with the aforementioned concomitant treatment, with a necrotic index ~90% and a very reduced viable tumor component, which suggests a synergic effect of both therapies.
In view of the current clinical dilemmas of immunotherapy 25 , we used the PDX model to test whether there were differences in the application of sequential treatment with chemo-immunotherapy; that is, anti-PD-1 first www.nature.com/scientificreports www.nature.com/scientificreports/ and then cisplatin (anti-PD-1 → cisplatin) or vice versa (cisplatin → anti-PD-1). We also wanted to know if delivering treatment sequentially was more efficient than concurrently 25,26 . However, our results do not provide a basis for concluding that any of the options would have more benefits over the other. While anti-PD-1 → cisplatin combination did reduce tumor volume and increase necrosis as compared with the isotype control group, this effect is probably explained by the chemotherapy administration. Nonetheless, sequential treatments were inferior to the concomitant/concurrent administration of chemo-immunotherapy in terms of tumor regression.
Theoretically, infiltrating cytotoxic T lymphocytes are the anti-tumor effector cells in anti-PD-1 therapy 3,7 . NOD-SCID gamma mice, the model used in this study, are characterized by the absence of T and B lymphocytes and NK cell functionality. However, the innate immune system is functionally active in these animals. Our results show that the original human infiltrate was replaced by a murine tumor infiltrate, which is expected in this model 27,28 . Indeed, we found a rich murine myeloid component in the tumor stroma of mice treated with anti-PD-1. In this context, it is known that some myeloid cells such as dendritic cells 29 , macrophages 29,30 and neutrophils 31,32 , have important roles as mediators in the tumor microenvironment. In addition, these cells could contribute to the anti-tumor action, blocking immune checkpoints 29,33 . The recruitment of myeloid cells during immunotherapy treatment has been related to an anti-tumor response 29,34 with cytotoxic capacity as their main mechanism 29,35 . It is possible that these cells replace cytotoxic lymphocytes in response to anti-PD-1 3,7 under immunosuppression conditions, increasing their tumor recruitment in tumors treated with anti-PD-1, as suggested here.
The mechanisms by which intra-tumor myeloid cells function are not clear, particularly the duality of mechanisms regarding the polarization towards pro/anti-inflammatory and pro/anti-tumor capacity in tumor-associated macrophages 30 . Indeed, we are beginning to realize that there are two different populations of tumor-associated neutrophils (TANs), some of them anti-tumor (N1) and others pro-tumor neutrophils (N2) 31,32,36 . This phenomenon is not well understood in detail, but it seems to involve a balance between pro-and anti-tumor populations 31,32,37 .
We detected inflammatory exudates in tumors from the anti-PD-1 groups (treated with anti-PD-1 alone or in combination with chemotherapy) associated with large necrotic areas containing destructed tissue and PMNs, mostly neutrophils. The clinical significance of this is controversial because some studies associate a tumor infiltrate rich in neutrophils with poor prognosis 31,33,38 , while others establish a good association between this infiltrate and a positive response to immunotherapies 33,39 . Our data suggest that neutrophils located in the necrotic areas of tumors from the anti-PD-1 groups are active, and respond to anti-PD-1 through NO production and the consequent formation of nitrotyrosine, a marker of oxidative damage 31,32,36,39 . The most probable mechanism would be the known cytotoxic action of TANs, especially of those with an N1 phenotype 31,32,36,39 .
The necrotic areas evident in the anti-PD-1 treatment histologies are very different morphologically from those produced by cisplatin, indicating that different mechanisms of cell death exist between treatments. Apoptosis is the main mechanism of action of cisplatin 40,41 , although we found no evidence of apoptosis in the tumors from the anti-PD-1 treatment groups (data not shown). Overall, the finding of strong nitrotyrosine labeling in necrotic areas of tumors from the anti-PD-1 treatment group, and its co-localization with MPO, an accepted marker of neutrophils, suggests that infiltrated neutrophils are responsible for the mechanism of cell death by cytotoxicity and NO production, phenomena usually associated with acute inflammation foci 31,42 . This finding was also observed in the sequential treatment groups, but was less marked.
The results of the present study suggesting a role of neutrophils in response to immunotherapy are novel. The action principle of targeted therapies consisting of monoclonal antibodies is their specific binding to a particular antigen (protein) and its blockade 2,43,44 . In the isotype control group, anti-PD-1 binding sites were detected indirectly in the membrane of some cells as well as in delocalized locations (in the cytoplasm of other cells). Because we used nivolumab for this analysis, we presume that PD-1 was blocked in murine cells. This seems in accordance with aforementioned paradoxical or unconventional responses to immunotherapy and contradicts the IHC results in the PDX tumors with commercial antibodies, which were negative for this staining. By contrast, in the immunofluorescence experiments where anti-PD-1 was used as a primary antibody, we found PD-1 receptors in tumor stroma. According to our results in tumors whose treatments included cisplatin, the anti-PD-1 did not bind to any specific site in the tumor, suggesting that cisplatin may be destroying cells that expressed the PD-1 receptor or prevents the expression of PD-1 receptor. The detection of anti-PD-1 in tumors that received anti-PD-1 treatment was increased, indicating that this therapy stimulates the expression of its own target 45,46 .
On the other hand, neutrophils, both infiltrated or located in necrotic areas of tumors treated with anti-PD-1, appear to express the PD-1 receptor on their membrane. Indeed, our immunofluorescence experiments indicated binding of anti-PD-1 to PD-1 receptors in these cells. To our knowledge, there is no precedent for direct neutrophil activation by an anti-PD-1 monoclonal antibody. Preliminary data has suggested that PD-L1 and, occasionally, PD-1 can be expressed on the surface of neutrophils, but there must be T lymphocyte mediation 47,48 , which is not the case in our model. Thus, although more research is needed, especially in patients, our results are novel and suggest that neutrophils might be potential targets of treatment with anti-PD-1 immunotherapy.
Based on our results and while keeping in mind the limitation that the study was done in a murine PDX model and in fact in only one immunotherapy-responder line of PDX mice with a small number of mice in each experimental group, neutrophil action appears to be the consequence of several processes (shown in Fig. 8), that would occur acutely as a ripple effect with the production of NO, causing necrotic expansion. One process would involve the direct activation of neutrophils by the binding of anti-PD-1 to PD-1 receptors located on the neutrophil surface and another process would be the binding of the anti-PD-1 antibody to FcγR, as previously suggested elsewhere 33,49 and corroborated here. It has been described that neutrophils can perform powerful and fast cytotoxic functions (known as antibody-dependent cellular cytotoxicity) in the presence of monoclonal antibodies against tumor cells 36,50 . The neutrophil antitumor effector function mediated by Fc is not a well understood mechanism, but it is known that it could include opsonization and death mediated by necrosis 33,49,51,52 , as shown here.