Anti‐PD‐1 therapy redirects macrophages from an M2 to an M1 phenotype inducing regression of OS lung metastases

Abstract Osteosarcoma (OS) pulmonary metastasis translates into poor patient survival. The implication of PD‐1‐PD‐L1 pathway in the context of NK cells and/or macrophages in OS is unknown. We investigated the effect of anti‐PD‐1 in OS lung metastasis and the role of NK cells and/or macrophages in anti‐PD‐1 responses. A human LM7 OS mouse model was used. Immunohistochemistry for tissues (PD‐L1, caspase‐3, Ki‐67, NK cells, macrophages), and Western blotting for OS lung tumors (p‐Stat3, p‐Erk1/2) was performed. NK and macrophages were assessed using flow cytometry. NK cell and macrophage depletion were conducted using anti‐asialo GM1 and clodrosome, respectively. PD‐L1 expression was observed in human OS cells and OS patient lung metastases. Anti‐PD1 antibody led to a significant decrease in the number of OS lung metastases, enhanced tumor apoptosis, decreased tumor cell proliferation, and p‐STAT‐3/p‐Erk1/2 signaling blockade in OS lung tumors. NK cells and macrophages in OS lung tumors expressed PD‐1 and anti‐PD1 increased NK cell and macrophage tumor infiltration. Increased numbers of antitumor M1 macrophages and decreased pro‐inflammatory M2 macrophages were seen. NK depletion did not affect therapeutic effect of anti‐PD‐1, suggesting that NK cells were not directly involved. However, macrophage depletion significantly compromised anti‐PD1 efficacy, confirming their role in efficacy of anti‐PD‐1 against OS lung metastasis. Our findings suggest that OS lung metastases regression by anti‐PD1 can be attributed to activated tumor M1 macrophages and reduced M2 macrophages. Owing to the co‐relation of M1 macrophages with OS patient outcome, we provide a novel mechanism of PD‐1 blockade and a basis for future clinical trials for anti‐PD‐1 antibodies in OS.

Introduction immunity [10]. Current studies are focusing on targeting TAMS in the tumor microenvironment, primarily suppressing the polarization from M1 macrophages to M2 phenotype [11,12]. Here, we investigated the effect of anti-PD-1 in OS lung metastases using a human LM7 OS mouse model, enabling the assessment of PD-1 blocking effects on NK cells and macrophages, in the absence of T cells [13]. We show that anti-PD1 treatment induced NK and macrophage infiltration into OS lung metastases. In addition, increased migration of antitumor M1 macrophages and decreased migration of pro-inflammatory M2 macrophages were observed. However, the therapeutic efficacy of anti-PD1 was mediated by macrophages and not NK cells. Hence, we report that macrophages are crucial in the antimetastatic responses of anti-PD1 in OS lung metastasis.

OS patient samples
Paraffin-embedded tissues of lung metastases from 10 OS patients were obtained. The patient protocol was approved by IRB at M. D. Anderson Cancer Center.

Animal model
Athymic female nu/nu mice purchased from National Cancer Institute (Bethesda, MD) were housed in an animal facility approved by American Association of Laboratory Animal Care. Protocols were in compliance with the Institutional Animal Care and Use Committee (IACUC) at M. D. Anderson Cancer Center. LM7 cells (2 × 10 6 ) were intravenously injected in mice tail vein and anti-mPD-1 (RMPI-14) (Bio X Cell [West Lebanon, NH]); treatment was started after 6-7 weeks, after confirmation of micrometastases formation by H&E staining of the lung tissues. Mice were randomly divided into two groups (five mice per group) and treated intraperitoneally (i.p.) with either PBS or anti-PD-1 antibody (200 μg/mouse) twice a week, for 5 weeks. Mice were sacrificed using 5% CO 2 , and major organs were resected. Tissues were either fixed with 10% formalin acetate (Fischer Scientific, Pittsburgh, PA) and paraffin-embedded or were snap-frozen in 1:1 Tissue-Tek OCT (VWR, PA) and PBS.

Toxicity studies
Analysis of mouse tissues was performed by a pathologist blinded to the study. Complete blood count (CBC) and chemistry profile were determined in serum for evidence of toxicity.

PD-L1 expression in OS patient lung metastases
IHC staining of 10 OS patient lung metastasis paraffinembedded sections was performed using lung adenocarcinoma tissue as a positive control. We found PD-L1 (PD-L1 + ) expression in eight of 10 patients (membrane and cytoplasm) ( Fig. 2A). Negative control showed no staining as well as positive control tissue showed high PD-L1 staining intensity. Any staining (either cytoplasmic or membrane or both) was considered as PD-L1 positive staining. Variable expression patterns were observed within the OS patient samples. Quantification confirmed an average of ~30% positive PD-L1-positive staining intensity (calculated based on the fraction of cells staining positive for PDL-1 in every sample) (Fig. 2B). Our data demonstrated that PD-L1 expression in OS lung metastases varies between patients.
To determine the anti-PD1 response mechanisms, we assessed tumor cell proliferation (Ki-67 IHC) and apoptosis (TUNEL). The number of Ki-67 + tumor cells was In addition, cleaved caspase-3 ( Fig. 3C and E) and TUNEL expression ( Fig. 3C and F) in anti-PD-1 group were significantly higher as compared to PBS controls. Further, Western blotting confirmed apoptosis induction through significant increase in cleaved caspase-3 in lung tumors of anti-PD-1 group compared with the control (Fig. 3G).
In conclusion, anti-PD-1 therapy decreased macro-and micrometastases by inhibiting tumor cell proliferation and increasing tumor cell apoptosis.
Pathological analysis of H-and E-stained organ sections showed no toxicity or inflammation (data not shown). Anti-PD1 did not cause significant changes in the complete blood cell counts and liver function in mice, indicating no significant systemic toxicities (Fig. S3). The normal range of WBC differential in nude mice is 4-7.4 × 10 3 / μL, which is in a comparable range of 4-11 × 10 3 /μL in humans, although the lymphocyte population is significantly less in nude mice and is composed of B cells in the majority with very few T cells [17].

PD-1 upregulation and increased NK cells and macrophages in OS lung tumors after anti-PD-1
PD-1 expression on NK cells and macrophages in OS tumors is unknown. Flow cytometry of OS lung tumor cell suspension demonstrated that 98.8% of macrophages (F4/80 + /CD11b + ) and 100% of the NK (NKp46 + ) cells were PD-1 positive (Fig. 4A). These findings support a potential immune escape mechanism for OS tumor cells.
To analyze anti-PD1 activated immune responses, we quantified NK and macrophage infiltration within the OS lung tumors. Flow cytometry revealed a 2.5-fold increase in the number of NK cells (NKp46 + : 22-56%) and a 2.6-fold increase in the number of macrophages (F4/80 + /CD11b + : 15-33%) in the anti-PD-1 group compared with the control, respectively (Fig. 4B). However, NK cells and macrophages in spleens were unchanged after anti-PD-1 (Fig. S4). This data reflected that in the absence of T cells, the effect of anti-PD-1 on NK and macrophage infiltration was specific to the OS lung tumor site.
NKp46 and F4/80 IHC staining were significantly increased in LM7 OS lung metastases of anti-PD-1 group than the controls, primarily inside the tumor compared Anti-PD1 and Macrophages in Osteosarcoma P. Dhupkar et al.
with the tumor periphery ( Fig. 4C-F). These results support our flow data of increased number of NK cells and macrophages infiltrating into the OS lung tumors after anti-PD1 treatment in mice in the absence of T-cell population.

Anti-PD1 increases M1 and decreases M2 macrophages in OS lung tumors
CD86 IHC staining for M1 macrophages revealed a significant increase in the M1 macrophages within LM7 OS lung metastases after anti-PD1 treatment (P < 0.05) ( Fig. 5A and B). By contrast, CD163 + M2 macrophages were observed in the tumor periphery and were significantly decreased in the anti-PD1-treated mice compared with controls (P < 0.05) (Fig. 5C and D). Thus, PD-1 blockade shifts the balance of pro-inflammatory M2 macrophages toward activated antitumorigenic M1 macrophages, which may play a role in its' therapeutic efficacy.

The antitumor efficacy of anti-PD1 is mediated by macrophages and not NK cells
To investigate if NK cells are crucial in anti-PD-1 response, NK cell depletion was performed using anti-asialo GM1 and confirmed in lung tumors (Fig. 6C) and spleens (Fig.  S5). NK depletion had no effect on the anti-PD-1 antimetastatic activity (Fig. 6A and B). Mice treated with anti-PD1 or anti-asialo GM1 preceding anti-PD1 had a similar response compared with the controls (P < 0.05). Thus, the therapeutic effect of anti-PD1 was not compromised on NK depletion. In addition, NK depletion did not affect the growth of lung metastases. Hence, NK cells were not critical in the anti-PD-1 efficacy against LM7 lung metastases.
To assess the role of macrophages in anti-PD1 response, macrophage depletion was performed. Clodrosome treatment caused a significant decrease in macrophages (CD11b + F480+) in OS lung tumors and spleens (Figs. 6F and S6), but not NK cells. The number of macro-and micrometastases was significantly compromised following macrophage depletion (Fig. 6C and D). The number of metastases in mice pretreated with clodrosome followed by anti-PD1 was similar to the PBS group (P > 0.05). Clodrosome did not affect the number of metastases. These data, along with findings from Fig. 6, suggest that macrophages are responsible for anti-PD1 antitumor responses in LM7 OS model. [18]. Most in vivo studies have discovered T-cell activation as the primary mechanism of anti-PD1 responses [9,19].

PD-1 inhibitors have shown promise in the clinic
Here, we demonstrate that anti-PD-1 antibody mediated OS lung metastasis regression through macrophage activation, through increased M1 and decreased M2 macrophage infiltration.
Tumor PD-L1 upregulation has been known to block antitumor immune responses [20]. Our data of in vitro PD-L1 expression and variability in OS cell lines are similar to previous reports in melanoma [21]. We also show that PD-L1 is expressed in OS patient lung metastases and primary tumors. Recently, PD-L1 expression was demonstrated in 75% of OS metastatic patients, but not in primary OS patients [22]. We observed PD-L1 expression in both primary tumors and lung metastases. The variable PD-L1 expression in patients may be due to the heterogeneity, as observed in non-small-cell lung cancer (35-95%) and melanoma (40-100%) [8,23]. Importantly, anti-PD-1 patient responses may not depend on strong PD-L1 positivity as discovered in non-small-cell lung cancer patients [18]. Thus, our data support that PD-L1 is a potential OS therapeutic target in the clinical setting.
PD-1/PD-L1 inhibitors have shown efficacy in preclinical melanoma and pancreatic cancer models [9,19,24]. We discovered that anti-PD1 caused regression of OS lung metastasis. Due to the significant homology, the binding affinities of mPD-1 to mPD-L1 and mPD-1 to h-PD-L1 are equivalent [15,16]. Hence, in our model, antimurine PD-1 antibody likely exerts therapeutic efficacy due to blockade of mPD-1 on endogenous mouse NK cells/macrophages binding with hPDL-1 on LM7 tumor cells. Our data align with previous study showing reduced OS metastases after PD-L1 blockade [22]. In addition, we uncover efficacy mechanisms including inhibition of tumor cell proliferation, apoptosis induction, inhibition of p-Stat3/

CD163 positivity
Anti-PD-1 * : P < 0.05 * PD-L1 pathway. We also established the safety of anti-PD-1 at the administered dosage. PD-1 upregulation on tumor-infiltrating T cells has been associated with T-cell dysfunction [7]. PD-1 is induced on activated NK cells in infection, multiple myeloma and on macrophages in sepsis [25,26]. Also, that mouse and human TAMs have elevated PD-1 expression and that high expression inhibits macrophage phagocytosis and tumor immunity [10]. Here, we demonstrated for the first time that PD-1 is upregulated on both NK cells and macrophages in OS lung metastases. This indicates suppressed NK and macrophage function, which may contribute to an immunosuppressive OS lung metastatic microenvironment. Although previous research has shown tumor T-cell infiltration after PD-1/PD-L1 blockade, the importance of macrophages and intrinsic immune responses in tumor immunity and therapy efficacy is now being recognized [22]. Recent evidence demonstrated that checkpoint inhibitors enhance NK cell-mediated antimetastatic activity against lung metastasis and are not limited to only T cells [27]. Our nude mouse model allowed us to focus on the specific effects of anti-PD-1 on NK cells and macrophages. Our data showing increased NK and macrophage tumor infiltration after anti-PD-1 treatment are unique and entail targeting PD-1 on the innate immune system, as opposed to PDL-1 on the tumors. The effect of anti-PD-1 in an immune-competent mouse model needs to be further evaluated to understand T-cell contribution.
Our discovery of increased M1 macrophages and decreased M2 macrophages on PD-1 blockade aligns with previous literature of L-MTP-PE causing M1 activation, cytokine release, and OS cell death [5,28]. Tumorinfiltrating macrophages have been associated with OS metastasis suppression. Recent studies have shown that M2 macrophages promoted breast cancer metastasis progression [29]. M2 macrophage depletion using CSF-1R inhibitor has been found to enhance dendritic cell immunotherapy responses [30]. M1/M2 macrophage polarization (increase in M1 versus M2 macrophages) has been corelated with increased survival of non-small-cell lung cancer patients [12]. Thus, suppressing M1 to M2 macrophage polarization is being investigated as a potential therapeutic approach. In conclusion, our data showing an increase in M1 and decrease of M2 macrophages, possibly through macrophage polarization, uncover a unique mechanism of anti-PD1.
Unlike other cancers, TAMs play an antimetastatic role in OS, as the infiltrating TAMs have been shown to correlate with better patient survival [31]. Strikingly, we discovered that macrophages were predominant effector cells involved in anti-PD-1 antitumor responses. Recently, Gordon et al. [10] also revealed decreased colorectal tumor growth on PD-1-PD-L1 blockade, in a macrophage-dependent , causing decreased tumor growth and metastasis [28]. Despite increased NK cells migration in tumors, we demonstrated that NK cells were not crucial for anti-PD-1 efficacy. By contrast, the anti-PD-1 efficacy was lost in macrophage-depleted mice. Recent studies of L-MTP-PE and IL-2/IL-15 showed that macrophages can be directly cause phagocytosis, independent of NK cells [28,32]. Also, PD-1/PD-L1 blockade in TAMs led to direct macrophageled phagocytosis of tumor cells [10]. Our data indicate that NK cells were not critical in direct killing of OS tumor cells. It is possible, however, that NK cells may release cytokines causing M1 activation and the killing of M2 macrophages [33].
Our observations have a translational significance for relapsed OS patients, with lymphocyte immunosuppression postchemotherapy. Previously, low absolute lymphocytes were found in sarcoma patients after prolonged chemotherapy [34]. However, the macrophage cytotoxic function is not suppressed by chemotherapy for newly diagnosed/ relapsed disease OS patients [35]. Hence, our results indicate that anti-PD-1 can be combined with chemotherapy in newly diagnosed patients to eradicate lung micrometastases, thus improving survival. Anti-PD-1 alone or in combination with chemotherapy may eradicate lung metastases in relapsed patients through macrophage activation, and by increasing the M1/M2 ratio.
In summary, targeting PD-1/PD-L1 signaling is a promising approach for the treatment of OS lung metastasis. Notably, macrophages are the key mediators of anti-PD1 efficacy, and redirecting M2 macrophages to M1 resulted in regression of OS lung metastasis. These novel mechanisms provide a strong rationale for clinical translation of anti-PD-1 antibodies alone or in combination with other immunotherapies for the treatment of OS lung metastasis patients for improved clinical outcomes. P. Dhupkar et al.

Supporting Information
Additional supporting information may be found in the online version of this article: Figure S1. IFN-γ induces PDL-1 expression in OS cells. IFN-γ (200 U/mL) 24 h treatment was performed followed by flow cytometry using anti-PDL-1-APC antibody or IgG-APC isotype control. Mean, standard deviations for PDL-1 positivity shown (*P < 0.05). Figure S2. Anti-PD-1 inhibits p-Stat3/PDL-1 pathway in LM7 lung tumors. PDL-1 IHC staining (A) and Quantification (B) was performed on LM7 sections using anti-hPDL-1 antibody, Mean ± SD of PDL-1 positivity was calculated and student's t-test was performed, *P < 0.05; Western blot (C) and Image J quantification (D) was performed for LM7 lung tumors after 5 week treatment, Actin was the loading control. Figure S3. Anti-PD-1 did not change CBC and liver function. Hemoglobin, hematocrit, platelets and WBC counts were determined in mice blood after 5 week treatment (A); AST and ALT liver enzymes were evaluated from mice serum. Student's t-test was used for analysis (P > 0.05) (B). Figure S4. Anti-PD-1 treatment did not affect NK cells and macrophages in spleen. Flow cytometry was performed on lung tumor suspensions using anti-NKp46-PerCP or anti-F4/80-APC and anti-CD11b-FITC antibodies. Figure S5. Anti-asialo-GM1 significantly decreased NK cells in spleen. Flow cytometry was performed for spleen cell suspensions after anti-asialo-GM1 (50 μL, twice weekly) treatment using anti-NKp46-PerCP or anti-F4/80-APC and anti-CD11b-FITC antibodies. Figure S6. Clodrosome significantly decreased spleen macrophages. Flow cytometry was performed with spleen cell suspensions after Clodrosome (200 μL, twice weekly) treatment using anti-F4/80-APC and anti-CD11b-FITC or anti-NKp46-PerCP antibodies.