Topical arginase inhibition decreases growth of cutaneous squamous cell carcinoma

Cutaneous squamous cell carcinomas (cSCC) are among the most commonly diagnosed malignancies, causing significant morbidity and mortality. Tumor-associated macrophage (TAM) expression of arginase is implicated in tumor progression, and therapeutic use of arginase inhibitors has been studied in various cancers. However, investigating potential cSCC immunotherapies including arginase inhibition in pre-clinical models is hampered by the lack of appropriate tumor models in immunocompetent mice. PDV is a cSCC cell line derived from chemical carcinogenesis of mouse keratinocytes. PDVC57 cells were derived from a PDV tumor in C57BL/6 (B6) mice. Unlike PDV, PDVC57 tumors grow consistently in B6 mice, and have increased TAMs, decreased dendritic and T cell intra-tumor infiltration. Arginase inhibition in cSCC tumors using Nω-hydroxy-nor-arginine (nor-NOHA) reduced tumor growth in B6 mice but not immunodeficient Rag1-deficient mice. nor-NOHA administration increased dendritic and T cell tumor-infiltration and PD-1 expression. The combination of nor-NOHA and anti-PD-1 therapy with nivolumab enhanced anti-PD-1 therapeutic efficacy. This study demonstrates the therapeutic potential of transcutaneous arginase inhibition in cSCC. A competent immune microenvironment is required for tumor growth inhibition using this arginase inhibitor. Synergistic co-inhibition of tumor growth in these results, supports further examination of transcutaneous arginase inhibition as a therapeutic modality for cSCC.

In this study, we used PDVC57, an cSCC cell line derived from the C57BL/6-derived PDV cSCC cell line via in vivo passage, to serve as a murine tumor model for analyzing the immune microenvironment in cSCC 21,22 . In contrast to implantation of PDV cells, which develop tumors in ~ 10% of C57BL/6 mice, PDVC57 cells develop tumors in 100% of C57BL/6 mice. We investigated the role of arginase in PDVC57 tumor progression in vivo using transcutaneous inhibition of arginase. Inhibition of arginase activity was effective in reducing tumor growth in vivo. We determined that PD-1, a T cell activation/exhaustion marker and the prototypical checkpoint inhibitor target, was upregulated in the setting of small molecule inhibition of arginase. The combination of transcutaneous arginase inhibition with anti-PD-1 therapy demonstrated therapeutic synergy in the reduction of PDVC57 cSCC tumor growth.

Results
PDV and PDVC57 are genetically related murine squamous carcinoma cell lines that grow similarly in vitro but are distinct in their growth in vivo. To better characterize the PDVC57 cell line, PDV and PDVC57 cells were cultured and whole exome sequencing was performed. Sequencing demonstrated that PDV and PDVC57 share a majority (2800/3640, 76.9%) of their exonic variants compared to normal blood leukocytes (Fig. 1A, upper) and the two cell lines have similar mutational burdens (PDV: 2896 exonic mutations; PDVC57: 2726 exonic mutations. Supplemental Fig. 1). The cultured cells appeared morphologically different (Fig. 1A, lower) and flow cytometric analysis revealed that PDVC57 cells were significantly larger and more granular than PDV cells (Supplemental Fig. 2). Finally, the cell lines had similar logistic growth curves that demonstrated no significant difference between the two, with PDVC57 having a slightly shorter but not significantly different calculated doubling time (Fig. 1B).
To develop and characterize this syngeneic cSCC tumor model, we injected 1, 5, or 10 million PDVC57 cells intradermally into the flanks of five C57BL/6 (B6) mice per group. All five mice in the 5 and 10 million cell injections developed tumors, and 3/5 (60%) of the mice in 1 million-cell injection developed tumors (Supplemental Fig. 3). The resultant tumors were harvested at day 21 ( Fig. 2A) and either embedded in paraffin for histopathological analysis or frozen for immunofluorescence. The histology of the injected tumors demonstrated cords and nests of malignant epithelial cells with nests of keratinizing cells consistent with cutaneous squamous differentiation 23 (Fig. 2B). Under immunofluorescence microscopic imaging, the tumors broadly expressed p63, a marker of squamous differentiation often used in the identification of SCCs and usually only found in the basal layer of the epidermis and the cutaneous appendages, such as the hair follicles [24][25][26] (Fig. 2C,D).
Given that the PDVC57 cells consistently developed tumors in B6 mice with 5-million-cell injections, we intradermally injected two sets of 10 mice with 5 million PDV versus PDVC57 cells. Comparing PDV and PDVC57 intradermal tumors in B6 mice, injected PDV tumors grew more slowly, reached a growth peak at approximately day 15, and subsequently regressed (Fig. 3A). None of the ten B6 mice bearing PDV tumors Upper: whole exome sequencing revealed that the two cell lines shared 2800/3640 (76.9%) of exonic variants compared to normal mouse peripheral blood leukocytes; lower panels demonstrate light microscopy images of the cell lines growing in vitro, with PDVC57 (right) demonstrating greater size and granularity than PDV (left). (B) Logistic growth curves of the two cell lines that show no significant difference between the growth curves using the F-test, P = 0.6350. www.nature.com/scientificreports/ reached the experimental endpoint and were all censored at day 42, and the tumor weights at the conclusion of the experiment were significantly lower than PDVC57 tumors (Fig. 3C,D). These differences were not demonstrated when the cells are injected into Rag1-deficient mice (B6.129S7-Rag1 tm1Mom /J) mice, which lack mature T and B cells (Fig. 3B-D).

Scientific
Transcutaneous arginase inhibition is sufficient to inhibit PDVC57 tumor growth. The growth of PDV tumors in immunodeficient Rag1 knockout (KO) mice (B6.129S7-Rag1 tm1Mom /J) in contrast to the immunocompetent B6 mice demonstrates that PDV tumors are immunologically rejected in B6 mice; furthermore, PDVC57 tumors may modulate the immune microenvironment to promote tumor growth 27 . Flow cytometry of PDV and PDVC57 tumors injected into B6 mice showed that PDVC57 tumors contained more tumor-associated macrophages (TAM), fewer tumor-infiltrating dendritic cells, and fewer tumor-infiltrating CD4 + and CD8 + T cells ( Fig. 4A-C). The increase in TAMs in SCC has been shown to lead to a worse prognosis, and previous work and existing literature had shown that TAM-derived arginase is important in promoting tumor growth 20,28,29 . In agreement, we observed that PDVC57 tumors demonstrated greater arginase activity compared to PDV tumors (Fig. 4D). Using a novel transcutaneous delivery model of arginase inhibition, Nω-hydroxy-nor-arginine (nor-NOHA) was topically applied to the cSCCs formed after intradermal PDVC57 injection. Compared to vehicle control, daily application of 5 mM nor-NOHA significantly decreased the volume and weight of PDVC57 tumors in WT B6 mice (Fig. 5A,B). This difference was not demonstrated in immunodeficient Rag1 KO mice (Fig. 5C,D), while arginase activity assays confirmed that nor-NOHA was able to successfully decrease arginase activity in both experimental groups (Fig. 5E,F).  Fig. 4)-almost fourfold higher than the concentration used in the in vivo treatments. Further, as shown above, transcutaneously delivered nor-NOHA did not induce a significant reduction of tumor growth in vivo in Rag1 KO mice (Fig. 5). Therefore, we sought to characterize the immunomodulatory effects of arginase inhibition. In B6 mice injected with PDVC57 tumors, nor-NOHA application increased CD4 + and CD8 + T cells, as well as CD11c + MHCII + dendritic cells within the tumor (Fig. 6A,B). Interestingly, proportions of CD3 + T cells expressing PD-1 also increased after treatment (Fig. 6A). Due to the increased PD-1 expression in nor-NOHA treated PDVC57 tumors, combination therapy with the PD-1 inhibitor nivolumab was tested. The combination therapy further decreased PDVC57 tumor growth ( Fig. 7A-D) and final tumor weight ( Fig. 7E-H), compared to either nor-NOHA or nivolumab alone. In addition, the combination treatment increased intratumoral CD4 + and CD8 + T cells, and CD11c + MHCII + dendritic cells in comparison to either nor-NOHA or nivolumab treatment, alone ( Fig. 8A-C).

Discussion
Immunotherapy is a rapidly emerging field in the treatment of cSCC and has largely focused on eliciting an adaptive immune response. The rationale behind treating cSCC by activating adaptive immunity is based on the observation that cSCC exhibits a high load of neoepitope mutations induced by ultraviolet (UV) radiation, which should stimulate cytotoxic T cells 30 . This mutational load was demonstrated in both PDV and PDVC57 cell lines through whole exome sequencing and prediction of more than 1900 neoepitope-contributing exonic variants ( Supplementary Fig. 1). Cemiplimab is an anti-PD-1 checkpoint inhibitor recently approved in 2018 by the United States Food and Drug Administration for the treatment of metastatic and locally advanced cSCC [31][32][33] . Pembrolizumab, another PD-1 inhibitor, is currently being studied as another possible immunotherapeutic in   [34][35][36] . However, only half of patients with unresectable locally advanced and/or metastatic cSCC responded to cemiplimab checkpoint inhibition 31 . As such, other targets in the immune response against cSCC may be exploited for further therapy. TAM-derived arginase is increasingly becoming an important therapeutic target in various cancers. The degradation of arginine, in which arginase plays an integral part, has been demonstrated to lead to T cell suppression 37 . Clinically, the upregulation of arginase activity and the expression of ARG1, has been correlated with various cancers, such as gastric, breast, renal cell, and head and neck squamous cell carcinomas [38][39][40][41] . In this study, we sought to utilize transdermal arginase inhibition as a method of non-invasive drug delivery that is particularly applicable to cSCC, a skin cancer.
Testing this hypothesis required an appropriate murine model with an intact immune system. However, as noted previously, there is a lack of preclinical models available for testing of potential therapeutics, which drastically hinders the further development of new therapies. The current gold standard for analysis of cSCC in immunocompetent mice is chemical carcinogenesis using a two-stage protocol of 7,12-dimethylbenz[a]anthracene (DMBA) and 12-O-tetradecanoylphorbol-13-acetate (TPA) application 42,43 . This method is limited by the multiple-month delay between initiation of carcinogenesis and observation of clinically apparent tumors. Many of the tumors that develop are also more similar, in terms of histology and growth characteristics, to squamous papillomas than true cSCC [42][43][44] . Several injectable murine cSCC tumor models have been derived from DMBA or TPA treatment. HaCa4 were transformed by HaMSV-transforming retrovirus and TPA 45 . CarB and CarC were derived from two anaplastic skin carcinomas induced by DMBA and TPA 46 . The PDV cell line was derived from primary C57BL/6 mouse epidermal cells treated in vitro with DMBA 27 . None of these cell lines form tumors consistently in immunocompetent syngeneic mice 27,45,46 . Herein we characterized PDVC57, a cell line derived by in vivo passaging of the PDV cell line. Morphologically and genetically, we have shown that PDVC57 is a distinct cSCC cell line. Notably, both the PDV and PDVC57 cell lines bear the Hras Q61L mutation, a well-characterized driver mutation in some human cSCC 47,48 . PDVC57 was able to consistently form cSCC We demonstrated that transcutaneous arginase inhibition was sufficient to decrease growth of the PDVC57 tumors, in part due to an increase in intratumoral effector T cells and dendritic cells. The involvement of the adaptive immune system was further demonstrated given the lack of anti-tumorigenic effect when nor-NOHA was applied to Rag1 KO mice, which lack mature T and B cells. Somewhat surprisingly, arginase inhibition in immunocompetent B6 mice also resulted in increased PD-1 expression on the intratumoral T cells. The mechanism of this is unclear; one possible pathway is that arginase inhibition leads to the increased activation of T cells with subsequent upregulation of the hyperactivation/exhaustion marker PD-1. The infiltration of PD-1 + T cells have been documented to be a favorable prognostic biomarker in both SCC of the head and neck and melanoma due to their response to checkpoint inhibition therapy 49,50 . As such, we were able to use systemic administration of the PD-1 inhibitor nivolumab in conjunction with topical nor-NOHA to decrease the rate of tumor growth more than either individual agent alone.
Finally, the development of cSCC due to UV radiation-induced carcinogenesis carries with it the relatively unique concept of field cancerization. Originally described in oropharyngeal SCC, field cancerization in cSCC is defined as the anatomical area adjacent to skin cancers and/or pre-cancers that demonstrate photodamage, and may show histopathologic dysplasia 51,52 . As such, treatment of only the pre-cancerous and/or cancerous lesions will not address the surrounding dysplastic tissue that carry similar mutational burdens and are likely to Arginase activity as assessed by enzymatic fluorometric assay of treated PDVC57 tumors in immunocompetent B6 mice, which was significantly different indicating effective inhibition of arginase. (F) Arginase activity in units of catalytic activity (U) as assessed by enzymatic fluorometric assay of treated PDVC57 tumors in immunodeficient Rag1 KO mice, which was significantly different indicating effective inhibition of arginase. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 using the unpaired t-test adjusted for multiple comparisons. Three separate experiments with n > 5 per group.  53 . Current field-directed therapies include 5-fluorouracil, a topical chemotherapeutic agent, and imiquimod, a toll-like receptor 7 agonist 54 . Topical arginase inhibition may provide an additional avenue of field-directed therapy for cSCC. In summary, cSCC is the second-most commonly diagnosed cancer in the US and has significant morbidity and mortality rates, especially in high-risk populations. The use of immunotherapy such as checkpoint inhibition for cases of advanced cSCC has shown promise, though a majority of patients do not respond. In this study, we have characterized a pre-clinical injectable immunocompetent mouse model of cSCC and demonstrated the efficacy of topical arginase inhibition in reducing tumorigenesis of cSCC, especially in conjunction with checkpoint inhibitors. These results are promising for the development of future topical adjuvant therapies in the treatment of cSCC.  Tumor measurements. Mice were sedated with isoflurane. Calipers were used to measure tumor diameter. Tumor volume was calculated using the following formula: (tumor length) × (tumor width) 2 × 0.52, where length is longer than the width.

Methods
Topical application of nor-NOHA. 20-50 μL (depending on tumor size) of 5 mM nor-NOHA (EMD Millipore) in DMSO or DMSO control was applied twice daily, starting one day after injection of PDVC57 cells. Tumor sizes were measured as above every other day. At twenty-one days, the mice were euthanized using carbon dioxide and cervical dislocation. The tumors were then dissected out and weighed. Histology and immunofluorescence. Freshly collected tumors were either fixed in 4% paraformaldehyde and embedded in paraffin or placed in optimal cutting temperature compound (OCT). The prepared tumors were cut into 5 μm sections, with the paraffin-embedded sections used for hematoxylin and eosin (H&E) staining and the frozen sections for immunofluorescence. The H&E slides were visualized with light microscopy. The frozen section slides were stained with DAPI and anti-p63 antibody conjugated to FITC (Bioss, bs-0723R-FITC) and scanned with Leica SP5 confocal microscope.
Whole exome sequencing. Exome sequencing and mutational burden analysis were performed according to 55 . Neoantigen prediction was performed for nonsynonymous SNV, Frameshift, Indel, and Stop-Loss variants that passed MuTect2 quality control filters to create all possible 8mers, 9mers, 10mers, and 11mers using custom Python scripts. Peptide binding affinity of wildtype and mutant peptides to C57BL6/J MHC I alleles H2-Db and H2-Kb were calculated using netMHCCons 56 .
Tumor processing. In a sterile Petri dish (non-tissue-culture-treated plate), a digestion buffer of 20 mL PBS + Ca 2+ + 200μL Liberase (5 mg/mL, Roche) + 1 mL DNAse (2 mg/mL, Roche) was prepared. After adding Arginase activity assay. Arginase activity was detected using the Arginase Activity Colorimetric Assay Kit (BioVision, Milpitas, CA). Tumors were processed as above and lysed in assay buffer at a concentration of 1 × 10 6 cells/100 µL. The assay was conducted in accordance with manufacturer instructions. The tumor lysates and background samples were measured in triplicate in a 96 well plate. (Corning, Falcon) The plate was warmed to 37 °C and read at 570 nm (BioTek). Optical density was recorded in kinetic mode every 5 min, beginning at time zero until 60 min, or signal saturation. Arginase Activity (Units [U]) was calculated by fitting the corrected sample reading to the generated standard curve. (A) Percentage of intratumoral CD45 + leukocytes that are CD3 + CD4 + , indicating tumor-infiltrating CD4 + T cells; combination nor-NOHA + nivolumab treatments had significantly higher proportions than either treatment alone. (B) Percentage of intratumoral CD45 + leukocytes that are CD3 + CD8 + , indicating tumorinfiltrating CD8 + T cells; combination nor-NOHA + nivolumab treatments had significantly higher proportions than either treatment alone. (C) Percentage of intratumoral CD45 + leukocytes that are CD11c + MHCII + , indicating tumor-infiltrating dendritic cells; combination nor-NOHA + nivolumab treatments had significantly higher proportions than either treatment alone. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 using the unpaired t-test adjusted for multiple comparisons. Two separate experiments with n > 5 per group.