A novel selective inhibitor JBI-589 targets PAD4-mediated neutrophil migration to suppress tumor progression

Abstract

Neutrophils are closely involved in the regulation of tumor progression and formation of pre-metastatic niches. However, the mechanisms of their involvement and therapeutic regulation of these processes remain elusive. Here, we report a critical role of neutrophil peptidylarginine deiminase 4 (PAD4) in neutrophil migration in cancer. In several transplantable and genetically engineered mouse models, tumor growth was accompanied by significantly elevated enzymatic activity of neutrophil PAD4. Targeted deletion of PAD4 in neutrophils markedly decreased the intratumoral abundance of neutrophils and led to delayed growth of primary tumors and dramatically reduced lung metastases. PAD4 mediated neutrophil accumulation by regulating the expression of the major chemokine receptor CXCR2. PAD4 expression and activity as well as CXCR2 expression were significantly upregulated in neutrophils from patients with lung and colon cancers compared to healthy donors, and PAD4 and CXCR2 expression were positively correlated in neutrophils from cancer patients. In tumor-bearing mice, pharmacological inhibition of PAD4 with the novel PAD4 isoform-selective small molecule inhibitor JBI-589 resulted in reduced CXCR2 expression and blocked neutrophil chemotaxis. In mouse tumor models, targeted deletion of PAD4 in neutrophils or pharmacological inhibition of PAD4 with JBI-589 reduced both primary tumor growth and lung metastases and substantially enhanced the effect of immune checkpoint inhibitors. Taken together, these results suggest a therapeutic potential of targeting PAD4 in cancer.

Introduction

Polymorphonuclear neutrophils (PMN) are one of the major components of the tumor microenvironment (TME) contributing to primary tumor growth and formation of the pre-metastatic niche (1–5) and limiting success of cancer therapies (6). Neutrophils are derived in the bone marrow (BM) from common myeloid progenitor cells (7) and mobilized out of the BM to the blood in response to stimuli including infection, inflammation, or cancer. The mobilization of neutrophils from the BM to the circulation and their migration to the tissues is a complex process that is highly regulated by multiple factors and requires the upregulation of C-X-C chemokine receptor 2 (CXCR2) together with the downregulation of CXCR4 signaling (8,9). A number of molecules that stimulate neutrophil release from the BM and their migration, including the CXCR2 ligands CXCL1 and CXCL2, are produced by tumor cells and are significantly upregulated in tumor-bearing hosts (10,11).

Cancer progression is accompanied by accumulation of pathologically activated PMN -polymorphonuclear myeloid derived suppressor cells (PMN-MDSC). Despite their common origin and phenotypical similarity, these cells are distinct in their biochemical properties, transcriptome, and functional activity. In contrast to PMN, PMN-MDSC are able to suppress responses of T and B lymphocytes and NK cells and promote tumor progression and metastasis via non-immune mechanisms (12,13). Tumor-associated neutrophils are a functionally diverse population, consisting of PMN with anti-tumor and PMN-MDSC with pro-tumorigenic properties (14–16). The balance between PMN and PMN-MDSC depends on the stage of the cancer. A recent study found that the immune suppressive PMN-MDSC accumulate in tumor tissues at early stages of tumor development (16). As a result, a highly suppressive environment is created in tumors and prevents rejection of tumors via immune-mediated mechanisms. The presence of PMN-MDSC in cancer patients is associated with poor prognosis and therapeutic outcomes (17). Therefore, targeting PMN-MDSC in cancer is an attractive therapeutic goal. The main currently unresolved challenge is how to selectively target PMN-MDSC but not PMN to avoid serious toxicity. The fact that PMN-MDSC and PMN share many functional features and receptors make this task especially difficult.

Citrullination is a posttranslational modification of proteins catalyzed by calcium-dependent peptidyl-arginine deiminase (PAD) enzymes, which hydrolyze a guanidino group of the arginine into a urea group resulting in a 1 Da change in molecular mass and converting a positively charged arginine into a neutral citrulline. This reaction affects hydrophobicity, protein structure, protein-protein interactions, and substrate receptivity to other posttranslational modifications (18). The PAD family comprises 5 members (PAD1–4 and PAD6) encoded by distinct genes (19) that have different tissue expression patterns and substrate specificity. PADs are involved in various physiological processes, including skin cornification, myelin sheath maintenance, and gene expression regulation (18). Excessive citrullination has been implicated in the pathogenesis of several autoimmune diseases mainly rheumatoid arthritis and lupus as well as in other pathologies including multiple sclerosis, psoriasis, Alzheimer’s disease, and cancer (20). PADs may be expressed by some cancer cells and usually function in the cytoplasm or nucleus. PAD4 is the only member of the PAD enzyme family with a nuclear localization sequence, and its enzymatic activity impacts gene expression by directly citrullinating transcription factors (21,22) or by regulating histones (23,24). PAD4 is highly expressed in neutrophils, where it regulates citrullination of histone H3 and H4. Citrullinated histones H3 has been implicated in the formation of neutrophil extracellular traps (NETs), extracellular structures formed by chromatin complexed with intracellular neutrophil proteins expelled from the neutrophils in a process called NETosis (25). An emerging role of NETs in cancer progression has been well documented (26). However, besides NETosis, the role of neutrophil PAD4 in cancer remains largely unknown.

Here, we uncovered a novel mechanism by which PAD4 in neutrophils promotes cancer progression. We found that neutrophil PAD4 regulates neutrophil trafficking, an effect mediated by transcriptional regulation of CXCR2 and specific for the PAD4 member of the PAD family. We also show that targeted deletion of PAD4 in neutrophils, or pharmacological inhibition of PAD4 activity with a novel selective PAD4 inhibitor JBI-589 reduced both primary tumor and metastases and significantly enhanced the effect of checkpoint inhibitors. Thus, our data indicate that PAD4 may be a promising therapeutic target for treatment of cancer patients.

Materials and Methods

Human samples

Collection of PB samples from patients with lung and colon cancer and healthy donors was approved by the Institutional Review Boards (IRB) of the Christiana Care Health System at the Helen F. Graham Cancer Center and the Wistar Institute. Collection of PB samples from patients with multiple myeloma was approved by the IRB of the University of Pennsylvania and the Wistar Institute. Written informed consent was obtained from all patients. The studies were conducted in accordance with the Declaration of Helsinki guidelines. Twenty-one patients with previously untreated stage II-IV non-small cell lung cancer (10 males, 10 females, and one unreported gender, age 50–94, average age 70.6±10.7), 12 patients with colon cancer (4 males and 8 females, age 38–79, average age 60.7±13.3), and 3 patients with multiple myeloma (1 male and 2 females, age 51–78, average age 63.3±11.1) were enrolled in this study.

Cell culture

LLC Lewis lung carcinoma (LL/2, RRID:CVCL_4358), LL/2-Luc2 luciferase expressing Lewis lung carcinoma (referred to as LL2, RRID:CVCL_A4CM), B16F10 melanoma (RRID:CVCL_0159), and EL4 lymphoma (RRID:CVCL_0255) cell lines were purchased from ATCC. Cells were expanded, frozen, and vials were kept in liquid nitrogen. Upon thawing, cells were kept in culture for 4–6 weeks. Cells were maintained in DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37°C with 5% CO2 and were Mycoplasma negative as detected using Universal Mycoplasma Detection Kit (ATCC).

Mouse models and treatment

All animal experiments were carried out in accordance with institutional guidelines and were approved by the Institutional Animal Care and Use Committee of The Wistar Institute. All mice were housed in pathogen-free conditions. Male and female mice aged 6–8 weeks were used in experiments.

C57BL/6 mice were purchased from Charles River Laboratories. Pmel TCR-transgenic mice (B6.Cg-Thy1a/Cy Tg(TcraTcrb)8Rest, strain no. 005023, RRID:MGI:3789507), OT-1 transgenic mice (C57BL/6-Tg(TcraTcrb)1100Mjb/J. strain no. 003831, RRID:IMSR_JAX:003831), PAD4 KO mice (B6.Cg-Padi4tm1.1Kmow/J, strain no. 030315, RRID:IMSR_JAX:030315), PAD4fl/fl mice (B6(Cg)-Padi4tm1.2Kmow/J, strain no. 026708, RRID:IMSR_JAX:026708), and Mrp8creTg mice (B6.Cg-Tg(S100A8-cre,-EGFP)1Ilw/J, strain no. 021614, RRID:IMSR_JAX:021614) were purchased from The Jackson Laboratory. PAD4^fl/fl and Mrp8creTg mice were crossed to generate PAD4^fl/flMrp8Cre⁺ mice and their PAD4^fl/flMrp8Cre⁻ littermates control. PAD2 KO mice (27) were backcrossed ten generations to the C57Bl/6 background prior to use. RET melanoma were obtained from Dr. Victor Umansky (German Cancer Center, Heidelberg, Germany) and KPC pancreatic cancer model from Dr. Robert H. Vonderheide (University of Pennsylvania, Philadelphia, PA).

Tumors were established by s.c. injection of 0.3 × 10⁶ LL2 cells, 0.3× 10⁶ B16F10, or 0.5× 10⁶ EL4 cells into the right flank of the mouse. Disseminated B16F10 model was established by i.v. injection of 0.5× 10⁶ B16F10 cells into tail vein. LL2 lung metastasis were evaluated using Perkin Elmer (Caliper) IVIS Spectrum In Vivo System.

For CD8 depletion, mice were treated with anti-CD8 antibody (clone 2.43, BioXCell, RRID: AB_1125541, 200 μg/mouse, i.p.) administered every 3 days beginning 24h prior to tumor cell injection. JBI-589 was administered at a dose of 50 mg/kg b.i.d., daily, via oral gavage. PD-1 antibody (clone RMP1–14, BioXCell, RRID: AB_10949053, 200 μg/mouse, i.p.) and CTLA-4 antibody (clone 9D9, BioXCell, RID: AB_10949609, 200 μg/mouse, i.p.) were administered twice a week beginning on day 7–9 after tumor cell injection for a total of 4 injections.

For evaluation of cell proliferation, mice were injected with BrdU (100 mg/kg, i.p). Mice were euthanized in 12h, spleens were harvested, and single cell suspension prepared. Proliferation of selected cell populations was evaluated after labeling with surface markers and staining with FITC-BrdU flow kit (BD Biosciences).

Reagents and antibodies

JBI-589, (R)-(3-aminopiperidin-1-yl)(2-(1-(4-fluorobenzyl)-1H-indol-2-yl)-3-methylimidazo[1,2-a]pyridin-7-yl)methanone (Supplemental Figure 9A), was provided by Jubilant Therapeutics. JBI-589 has a molecular weight of 481.57. The purity of the compound was 97.8% as evaluated by HPLC. The detailed JBI-589 synthesis protocol is described in the patent WO 2019/077631. All other antibodies and reagents used are described in Supplementary Tables 1 and 2.

Pharmacokinetic (PK) studies in mice

PK studies in mice were approved by the Institutional Animal Ethical Committee of Jubilant Biosys (IAEC/JDC/2019/188R). Balb/c male mice were obtained from Vivo Biotech, Hyderabad, India. Mice used in experiments were 6–8 weeks old and weighted 22–25g. Mice received JBI-589 at a dose of 10 mg/kg via oral gavage. JBI-589 was given in suspension formulation prepared using Tween-80 and 0.5% methyl cellulose. Sparse sampling was done at 0.25, 0.5, 1, 2, 4, 8, 10 and 24 h; at each time point three mice were used for blood sampling (~100 μL). Blood samples were collected in tubes containing K2. EDTA as anticoagulant and centrifuged for 5 min at 10,000 rpm at +4°C to obtain plasma.

Isolation of mouse cells

Mouse neutrophils were isolated from the BM and spleen using MACS technique. Briefly, single cell suspensions were prepared and incubated with anti-Ly6G MicroBeads (Miltenyi Biotec) followed by selection of Ly6G⁺ cells using LS columns. The purity of isolated Ly6G⁺ cells was more than 95% as detected by flow cytometry. All Ly6G⁺ cells expressed CD11b marker. Mouse CD11b⁺Ly6G^lo/-Ly6C⁺ M-MDSC were isolated from spleens by flow sort. For isolation of tumor-associated myeloid cells, single cell suspension was prepared using Tumor Dissociation Kit (Miltenyi Biotec). CD11b⁺Ly6G⁻F4/80⁺ tumor-associated macrophages (TAM) and CD11b⁺Ly6G⁺Ly6C⁻ tumor-associated neutrophils were isolated by flow sort. Mouse hematopoietic progenitor cells (HPCs) were isolated from the BM using a Lineage depletion kit (Miltenyi Biotec) according to the manufacturer’s instructions.

Isolation of human neutrophils and PMN-MDSC

Neutrophils were isolated from peripheral blood (PB) by centrifugation over a double density gradient Histopaque (Sigma) and collected from the interphase between Histopaque −1.077 and Histopaque −1.119. Cells were washed with DPBS and red blood cells were lysed using RBC lysis buffer (ThermoFisher Scientific). Neutrophil purity was confirmed by flow cytometry staining with anti-CD15 antibody.

Some experiments required comparison between PB neutrophils (PMN) and PMN-MDSC from the same cancer patient. For that, PMN-MDSC were isolated from the mononuclear fraction and neutrophils were isolated from granulocyte fraction obtained after Histopaque density gradient centrifugation using positive selection with CD15 MicroBeads (Miltenyi Biotech).

Preparation of tumor explant supernatant and conditioned medium

Tumor explant supernatants (TES) were prepared from excised LLC tumors of ~1.5 cm in diameter. Tumor tissue was minced into small pieces less than 3 mm in diameter and resuspended in RPMI-1640 medium supplemented with 10% FBS. Supernatants were collected after overnight incubation at 37°C in 5% CO₂, passed through 0.22 μm filters (EMD Millipore) and kept at −80 °C.

HPC differentiation

Mouse HPC were isolated from the BM and cultured in complete RPMI 1640 medium in the presence of GM-CSF (20 ng/mL) and 25% TES prepared as described above. Cells were collected on day 5 and Ly6G⁺ neutrophils were isolated using MACS technique.

Chemotaxis assay

Neutrophil chemotaxis was measured using 3 μm pore transwell system (Neuro Probe Inc. or Sigma-Aldrich). Advanced RPMI 1640 medium with CXCL1 (BioLegend) or fMLP (Sigma-Aldrich) was placed in the bottom of the well whereas neutrophils (0.1 × 10⁶ cells) resuspended in Advanced RPMI 1640 medium were placed on the top of Transwell insert. In some experiments, neutrophils were pre-incubated with 10 μM GSK484 (Cayman) for 30min prior to being placed on the top of the Transwell insert. Plates were kept for at 37°C in 5% CO₂ for 1h followed by counting migrated cells in the bottom well.

PAD4 biochemical assay

Full length recombinant human PAD4 (125 nM) was incubated with different concentrations of JBI-589 for 1h at room temperature followed by 1h incubation with 1.5 mM α-N-benzoyl-L-arginine ethyl ester (BAEE) substrate. Ammonia produced as a result of deimination of BAEE was detected after addition of ortho-phthaladehyde resulting in a fluorescent product. Fluorescence was measured at an excitation 410 nm and an emission 470 nm. IC₅₀ values were determined by sigmoidal dose-response curve (variable slope). Activity of JBI-589 against PAD1 and PAD2 enzymes was measured using a similar assay. Recombinant PAD1 and PAD2 were used at the concentration of 100 nM and 200 nM, respectively.

Flow cytometry

Single cell suspension was prepared for flow cytometry analysis. For tumor tissues, Mouse Tumor Dissociation Kit (Miltenyi Biotech) was used to prepare single cell suspension. For surface staining, cells were incubated with Fc-block for 10 min followed by labeling with indicated specific antibodies or isotype controls on ice for 20 min at 4°C in dark and stained with Live/Dead Fixable Aqua Dead Cell Stain Kit (Invitrogen). Fixation/Permeabilization Solution Kit (BD Bioscience) was used to process cells for intracellular staining. For detection of apoptosis, Annexin V/DAPI staining was performed according to the manufacturer’s protocol (BioLegend). At least 10,000 events were acquired using LSRII flow cytometer (BD Bioscience). Analyses was performed using FlowJo software (BD, Ashland, OR).

ROS production

The level of ROS production in neutrophils was assessed using CM-H₂DCFDA (Invitrogen). Briefly, splenocytes were labeled with surface markers, washed with PBS, resuspended in advanced RPMI 1640 medium, and stimulated with 200 nM phorbol 12-myristate 13-acetate (PMA) or 100 nM ionomycin for 30 min at 37 °C. Cells were then washed and incubated with 5 μM CM-H₂DCFDA for 30 min followed by staining with Live/Dead Aqua and analysis using flow cytometry.

Quantitative real-time PCR

RNA was extracted using E.Z.N.A. total RNA purification kit (Omega Bio-Tek) and cDNA was synthesized using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative PCR was performed using SYBR Green PCR Master Mix and primers listed in the Supplementary Table 3 using ABI 7500 FAST instrument (Applied Biosystems). Expression of specific genes was normalized to the expression of a housekeeping gene β-actin.

Western blotting

Cells were lysed using radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich) supplemented with Halt Protease and Phosphatase Inhibitor cocktail (ThermoFisher Scientific). Lysates were subjected to 10% SDS–PAGE followed by transfer to nitrocellulose membrane (Bio-Rad). Membranes were blocked with 5% nonfat dry milk and incubated with primary antibodies overnight at +4°C followed by incubation with secondary HRP-conjugated antibodies for 1h at room temperature. The following primary antibodies were used: anti-PAD4 (Antibodies Online), anti-Histone H3 (Cell Signaling Technology), anti-Histone H3 (citrulline R2+R8+R17, Abcam), and anti-β-actin (Santa Cruz Biotechnology). Densitometry was performed using ImageJ software (NIH, RRID:SCR_003070).

PAD activity assay

Neutrophils were resuspended in cell lysis buffer (Sigma-Aldrich) at 1×10⁸ cells/mL. Lysates were incubated with 10 mM BAEE substrate for 30 min at 37°C followed by addition of color developing reagent (13.0% (v/v) phosphoric acid, 25.27% (v/v) sulfuric acid, 0.573 mg/mL ammonium iron sulfate, 60 mM DAMO, 1.5 mM thiosemicarbazide) and heating at 95–100°C for 15 min. After cooling, 200 μL of the mixture was loaded per well of 96 well plate and absorbance was measured at 540 nm.

Suppression assay

Single cell suspension was prepared from spleens or tumor tissues. TAMs, M-MDSC, or PMN-MDSC were isolated, plated in U-bottom 96-well plates in triplicate in complete RPMI 1640 medium, and were co-cultured at different ratios with Cell Trace Violet-labeled splenocytes from either Pmel-1 transgenic mice in the presence of 1 ng/mL cognate peptide EGSRNQDWL or OT-1 transgenic mice in the presence of 0.025ng/mL cognate peptide SIINFEKL for 48h. Proliferation of CD8⁺ T cells was measured using flow cytometry.

T cell proliferation and activation

Splenocytes from C57Bl/6 wild type and PAD4 KO mice were labeled with CellTrace Far Red dye followed by plating and stimulation with CD3/CD28 Dynabeads (Thermo Fisher) for 48h. Mononuclear cells (MNC) were isolated from PB of healthy donors or patients with multiple myeloma using Ficoll-Paque gradient centrifugation. PB MNC were labeled with CellTrace Far Red dye, pre-treated for 2h with 10 μM GSK-484 or vehicle control (DMSO), and stimulated with CD3/CD28 Dynabeads for 72h. Cell proliferation was evaluated by flow cytometry in gated population of CD4⁺ and CD8⁺ T cells.

In parallel, human PB MNC pre-treated with 10 μM GSK484 or vehicle control were stimulated for 5h with the Cell Stimulation Cocktail with or without protein transport inhibitors (Thermo Fisher) according to the manufacturer’s protocol. Expression of IL-2 and IFN-γ was evaluated by flow cytometry in gated population of CD4⁺ and CD8⁺ T cells.

NET detection

Mouse neutrophils were kept untreated or cultured in the presence of LL2 supernatant or Calcium Ionophore overnight, then were fixed with 4% paraformaldehyde followed by staining with Sytox Green Nucleic Acid Stain at a final concentration of 250 nM. A Nikon Te300 inverted microscope equipped with a motorized XY stage was used to image NETs. Nikon Nis-Elements Ar software was used for image acquisition and was set-up to acquire 25 random locations per well. The z-stacks per location were then combined into an extended depth focused image. The total NET area was calculated by segmenting each image using a defined threshold pixel intensity setting. The spot detection tool in NIS-Elements Ar was used to count the number of cells per field. The sum of the total NET area in the 25 random fields of view was divided by the total number of cells in the 25 fields of view to obtain NET area (μM/cell).

Statistics

Statistical analysis was performed using GraphPad Prism (RRID:SCR_002798) software version 6.0. Differences between two experimental groups were determined using two-tailed Student’s t-test. Differences in tumor growth experiments were analyzed using two-way ANOVA. P value of less than 0.05 was considered statistically significant.

Data availability

The data generated in this study are available within the article and its supplementary data files.

Results

PAD expression and activity in neutrophils from tumor-bearing mice

Mouse neutrophils were defined as CD11b⁺Ly6C^loLy6G⁺ cells. In tumor-free mice, genes encoding PAD proteins, Padi2 and Padi4, were highly expressed in BM neutrophils, whereas levels of Padi1, Padi3, and Padi6 were undetectable (Supplementary Fig. S1A). In different mouse tumor models, the expression of Padi2 and Padi4 remained unchanged in BM neutrophils. These models included transplantable subcutaneous LLC lung carcinoma, B16F10 melanoma, EL4 lymphoma (Supplementary Fig. S1B), and genetically engineered KPC pancreatic cancer (28) and RET melanoma (29) (Supplementary Fig. S1C). Similar levels of PAD4 protein were detected in PMN from BM and spleen of tumor-free and tumor-bearing (TB) mice (Supplementary Fig. S1D). Growth of LLC tumor was not associated with upregulation of the expression of Padi1, Padi3, and Padi6 in neutrophils. No difference was found in the expression of Padi2 and Padi4 between tumor-associated and BM PMN (Supplementary Fig. S1E). Hematopoietic progenitor cells (HPC) were differentiated with GM-CSF in the presence or absence of LLC tumor explant supernatant (TES). The presence of TES did not affect Padi2 and Padi4 expression in PMN (Supplementary Fig. S1F). Thus, no changes in the gene expression or the protein amount of the members of PAD family were detected in neutrophils in cancer compared to neutrophils in normal mice.

Next, we asked whether tumor growth affected the enzymatic activity of PADs in PMN. Markedly increased PAD activity was detected in BM neutrophils from LL2-, EL4-, B16F10-bearing mice compared to tumor-free mice (Fig. 1A). Pre-incubation of PMN with a selective PAD4 inhibitor GSK484 (30) completely abrogated the tumor-induced up-regulation of PAD activity (Fig. 1B). PMN from PAD4 knockout (KO) tumor-free mice had markedly decreased PAD activation compared to WT mice. No difference in PAD activity was observed between PMN from TB and tumor-free PAD4KO mice (Fig. 1C). In contrast, significant PAD activation was detected in PMN from TB PAD2KO mice compared to tumor-free mice. Western blot analysis of PMN cultured overnight in the presence or absence of LLC TES demonstrated that tumor-derived factors induced citrullination of histone H3 at residues R2, 8, and 17 established to be specific substrates for PAD4 (Fig. 1D,E). Taken together, these data demonstrate that tumor growth is associated with activation of PAD4 in PMN.

Figure 1. Activity of PAD4 in neutrophils from tumor-free and tumor-bearing mice.

Ly6G⁺ cells were isolated from the BM of tumor-free (TF) mice or mice with indicated tumors. A,B, Activity of PAD was measured in (A) freshly isolated cells or (B) Ly6G⁺ cells pretreated for 2h with GSK484 as described in Materials and Methods section. C, PAD activity was measured in neutrophils isolated from the BM of TF and LLC-bearing WT, PAD4KO, and PAD2KO mice. Individual results, mean, and SEM values are shown. Statistics: one-way ANOVA. D,E, BM Ly6G⁺ cells isolated from two individual TF mice were kept overnight in the absence (control) or presence of LLC TES. Expression of indicated proteins was determined by western blotting. A representative image (D) and densitometry of indicated proteins normalized to the β-actin level (E) are shown.

Tumor growth and metastases in mice with PAD4-deficient neutrophils

To investigate the role of neutrophil PAD4 in regulation of tumor progression we generated PAD4^fl/flMRP8Cre⁺ mice that lack expression of PAD4 in PMN (Supplementary Fig. S2A) (31). LL2 and B16F10 models were chosen for our studies as these tumor cells do not express PAD4 (Supplementary Fig. S2B). In PAD4^fl/flMRP8Cre⁺ mice, growth of primary LL2 tumors was modestly but significantly delayed compared to their PAD4^fl/flMRP8Cre⁻ littermates (Fig. 2A). At the same time, the presence of lung metastases in PAD4^fl/flMRP8Cre⁺ mice was dramatically reduced compared to control (Fig. 2B,C). The proportion and absolute number of neutrophils also were substantially decreased in both primary tumors (Fig. 2D) and lungs (Fig. 2E) of mice with PAD4-deficient neutrophils. Similar results were observed in B16F10 mice: growth of primary B16F10 tumors was significantly delayed (Fig. 2F), the number of metastases, examined visually (Fig. 2G) or histologically (Fig. 2H,I) was dramatically reduced, and the proportion and absolute number of neutrophils in primary tumor (Fig. 2J) and lungs (Fig. 2K) were substantially reduced in PAD4^fl/flMRP8Cre⁺ mice compared to their PAD4^fl/flMRP8Cre⁻ littermates. In contrast, no changes in the presence of CD11b⁺Ly6C⁺ monocytic cells were observed in the tumor or lung tissues in either model (Supplementary Fig. S3A–D).

Figure 2. Effect of neutrophil PAD4 deficiency on tumor growth and metastasis.

A-E, LL2 or F-K, B16F10 tumors were established in PAD4^fl/fl MRP8Cre⁺ (Cre⁺) mice and their control PAD4^fl/fl MRP8Cre⁻ (Cre⁻) littermates by s.c. injection of tumor cells (n=5 per each group). A, Kinetics of s.c. LL2 tumor growth. **** - p<0.0001, two-way ANOVA. B, Images and C, quantitation of lung metastases detected by IVIS on day 24 after tumor cell injection. Scale bar, 150 mm. Frequency (out CD45+ cells) and absolute number of neutrophils in tumor (D) and lungs (E) on day 24 after tumor cell injection. F, Kinetics of B16F10 tumor growth. ** - p<0.001, two-way ANOVA. G, Visual assessment of lungs, H, representative microscopy images of H&E stained FFPE lungs tissue sections and I, histology score quantitation, frequency (out of CD45+ cells) and absolute number of neutrophils in tumor (J, per gram tissue) and lungs (K, per lungs) on day 28 after tumor cell injection. Individual values, mean, and SEM are shown. Scale bar, 50 mm (G). Statistics: unpaired two-tailed Student’s t-test.

To alleviate a possible effect of differences in primary tumor size on the presence of lung metastases we used a disseminated tumor model where intravenous injection of B16F10 cells resulted in formation of tumor lesions in the lungs. A dramatic reduction of tumor growth was observed in PAD4^fl/flMRP8Cre⁺ mice compared to control (Supplementary Fig. S4A); this effect was associated with a significant decrease in the presence of neutrophils at the tumor site (Supplementary Fig. S4B).

A similar anti-tumor effect against both primary tumors and lung metastases was observed in the whole body PAD4KO mice compared to WT controls (Supplementary Fig. S4C,D). This effect was not associated with the lack of PAD4 in T cells as PAD4-deficient T cells had a reduced proliferation capacity (Supplementary Fig. S4E). However, observed anti-tumor effect was abrogated by the treatment with anti-CD8 antibody, indicating an important role played by the immune system in regulating the anti-tumor response in PAD4KO mice (Supplementary Fig. S4C,D).

In contrast to PAD4 KO, a lack of PAD2 did not affect tumor progression. No differences in primary tumor growth or lung metastases were observed between PAD2KO and WT B16F10 (Supplementary Fig. S5A,B) or LL2 (Supplementary Fig. S5C,D)-bearing mice.

Thus, deletion of PAD4 in neutrophils had a strong suppressive effect on the formation of metastatic lesions, which was associated with a decreased presence of neutrophils in both the primary tumor and the lung metastases.

Functionality of PAD4-deficient neutrophils

To understand the mechanism by which neutrophil PAD4 regulates tumor progression we first tested for the presence of myeloid cells in PAD4^fl/flMRP8Cre⁺ mice. No differences in the proportion or absolute number of neutrophils in the BM were found between tumor-free PAD4^fl/flMRP8Cre⁺ mice and their control PAD4^fl/flMRP8Cre⁻ littermates (Fig. 3A). Similar results were obtained in LLC-bearing mice (Fig. 3B). In spleens, a small increase in neutrophils was found in PAD4^fl/flMRP8Cre⁺ mice, an effect observed in both tumor-free (Fig. 3C) and LLC-bearing (Fig. 3D) mice. No difference was found in the presence of Ly6C⁺Ly6G⁻ monocytic cells in the BM or spleens of tumor-free or LLC-bearing PAD4^fl/flMRP8Cre⁺ mice and their littermate controls (Supplementary Fig. S6A–D). These data suggested that the reduced presence of neutrophils at the tumor and metastatic sites observed in PAD4^fl/flMRP8Cre⁺ TB mice was not due to diminished myelopoiesis in the BM or spleen.

Figure 3. Presence of Ly6G⁺ cells in the BM and spleens of PAD4^fl/fl MRP8Cre⁺ mice.

A-D, Frequency and absolute number of Ly6G⁺ cells in the BM (A,B, per femur) and spleens (C,D, per spleen) of PAD4^fl/fl MRP8Cre⁺ (Cre⁺) and PAD4^fl/fl MRP8Cre⁻ (Cre⁻) tumor-free and LLC-bearing mice determined in gated live (Aqua-negative) cells. Individual values, mean, and SEM values are shown. E,F, HPC isolated from the BM of PAD4 KO or WT mice were differentiated in vitro with GM-CSF in the presence or absence of LLC TES with or without GSK484 (10 μM). Proportion (E) and absolute number (F) of Ly6G⁺ cells were evaluated after 5 days of culture in gated Aqua-negative cells. Statistics: unpaired two-tailed Student’s t-test.

Lack of PAD4 did not affect in vitro differentiation of HPC toward myeloid lineage. As anticipated, differentiation of HPC in the presence of TES led to the accumulation of CD11b⁺Ly6G⁺ cells after 5 days of culture. PAD4 deficiency did not interfere with this process. Pharmacological blocking of PAD4 activity with GSK484 had no effect on HPC differentiation in the presence or absence of TES (Fig. 3E,F). Furthermore, neither genetic deficiency or pharmacological blocking of PAD4 interfered with the presence of monocytic cells (Supplementary Fig. S6E,F).

As expected, spleen and tumor-associated neutrophils from TB mice had potent immune suppressive activity defining these cells as PMN-MDSC (Supplementary Fig. S7A). Deletion of PAD4 did not change the suppressive activity of PMN-MDSC. No effect of PAD4 deletion was observed on various factors implicated in PMN-MDSC activity: expression of Mpo, Arg1, iNos, Ptgs2, S100a8, S100a9, Mmp9, and Il6 genes (Supplementary Fig. S7B) or ROS production (Supplementary Fig. S7C). Pharmacological blocking of PAD4 activity with GSK484 during HPC differentiation did not abrogate the immune suppressive activity of Ly6G⁺ cells generated in the presence of TES (Supplementary Fig. S7D). However, PAD4 deficiency reduced the formation of NETs by mouse neutrophils (Supplementary Fig. S7E) (31,32).

Given that numbers of neutrophils in primary tumors and lungs were substantially reduced in PAD4 KO mice, we hypothesized that the lack of PAD4 in neutrophils may affect their motility and, therefore, impair their recruitment to the tumor site and pre-metastatic niche. To test this hypothesis, neutrophils were isolated from the BM of WT or PAD4KO tumor-free and LLC- or B16F10-bearing mice and their migration and chemotaxis were assessed using a standard transwell membrane assay. Lack of PAD4 did not affect the ability of neutrophils from tumor-free or TB mice to spontaneously migrate (Fig. 4A) and did not have any effect on migration induced by the chemoattractant fMLP (Fig. 4B). However, substantially reduced migration of PAD4-deficient neutrophils was observed in response to the chemokine CXCL1. This difference was more profound in neutrophils from TB mice (Fig. 4C). Pharmacological blocking of PAD4 activity with GSK484 did not affect the viability of neutrophils (Supplementary Fig. S8A) but it reduced their CXCL1-induced chemotaxis (Fig. 4D).

Figure 4. Effect of PAD4-deficiency on neutrophil chemotaxis.

A-C, Transwell assay evaluating the ability of BM neutrophils isolated from WT and PAD4KO tumor-free and LLC- or B16F10-bearing mice to migrate spontaneously (A) or in response to (B) fMLP and (C) CXCL1. B-C, Number of migrated cells was calculated by subtracting number of cells that spontaneously migrate from number of cells that migrate in response to stimulus. D, Migration toward CXCL1 was evaluated after pre-incubation of BM neutrophils with 10 μM GSK484 for 2h. Individual values, mean, and SEM are shown. Statistics: unpaired two-tailed Student’s t-test. E-H, Gene expression, receptor surface expression, and proportion of neutrophils expressing CXCR2 (E,F) and CXCR4 (G,H) in BM neutrophils from PAD4^fl/fl MRP8Cre⁺ (Cre+) and control PAD4^fl/fl MRP8Cre⁻ (Cre-) TF and LLC-bearing mice evaluated by qPCR and flow cytometry, respectively. MFI – mean fluorescent intensity. Statistics: unpaired two-tailed Student’s test. I,J, CXCR2 (I) and CXCR4 (J) surface expression on neutrophils differentiated in vitro from HPC in the presence or absence of LLC TES with or without 10 μM GSK484. Individual values, mean, and SEM are shown. Statistics: one-way ANOVA.

We next tested for a mechanism underlying the impaired chemotaxis of PAD4-deficient neutrophils. CXCR2 and CXCR4 are two major receptors responsible for retention of neutrophils in the BM and their release into the circulation and tissue migration. In PAD4 deficient neutrophils, expression of CXCR2, a major receptor for CXCL1, was markedly downregulated at the gene and protein levels (Fig. 4E,F). This downregulation was more profound in neutrophils from TB mice. CXCR4 gene expression was also lower in PAD4-deficient neutrophils (Fig. 4G), but the change in surface expression of this receptor was minor (Fig. 4H).

In vitro, the presence of TES during HPC differentiation stimulated the expression of CXCR2 receptor on neutrophils. Inhibition of PAD4 with GSK484 reduced baseline CXCR2 expression and abrogated up-regulation of CXCR2 in the presence TES (Fig. 4I). TES did not affect the expression of CXCR4; GSK484 slightly decreased CXCR4 expression in control and reduced its expression in the presence of TES (Fig. 4J). Thus, it appears that PAD4 regulation of neutrophil migration is mediated by CXCR2.

Expression and activity of PAD4 in neutrophils from cancer patients

To determine the clinical relevance of the mouse data, we evaluated PAD4 expression and activity in neutrophils from cancer patients. A significant increase in PAD4 gene and protein expression (Fig. 5A,B) and marked upregulation of PAD activity (Fig. 5C) were detected in neutrophils from peripheral blood (PB) of patients with lung and colon cancer compared to neutrophils from healthy donors. These changes were accompanied by markedly increased CXCR2 expression (Fig. 5D). At the same time, expression of CXCR4 was significantly upregulated only in patients with colon cancer but not with lung cancer (Fig. 5D). A direct correlation was observed between PADI4 and CXCR2 expression in neutrophils from cancer patients (R²=0.7698, p<0.0001) (Fig. 5E). PMN-MDSC isolated from mononuclear fraction of PB obtained from cancer patients demonstrated significantly increased PADI4 expression compared to PB PMN from the same patients (Fig. 5F), an effect associated with significant upregulation of CXCR2. However, no changes in CXCR4 expression were observed in PMN-MDSC.

Figure 5. PAD4 expression and activity in neutrophils from cancer patients.

A-G, Neutrophils were isolated from PB of healthy donors (HD, n=12) and patients with lung (n=21) or colon (n=8) cancer. Expression of PAD4 gene (A) and protein (B), PAD activity (C), and expression of CXCR2 and CXCR4 genes (D) were evaluated. E, CXCR2 correlation with PADI4 expression was assessed. F, Expression of indicated genes was evaluated in paired samples of PMN and PMN-MDSC isolated PB of cancer patients (n=8). Statistics: paired Student’s t-test. G,H, Neutrophils were treated with 10 μM GSK484 for 4h followed by detection of CXCR2 and CXCR4 expression using flow cytometry (G) and evaluation their migration (H). Statistics: unpaired Student’s t-test. I,J, Neutrophils from HD were treated with 10 μM GSK484 for 1–4h. Cells were collected at indicated time points and expression levels of CXCR2 (I) and CXCR4 (J) were detected by qPCR and flow cytometry. Individual values, mean, and SEM values are shown. Statistics: unpaired Student’s t-test.

Next, we evaluated the effect of pharmacological inhibition of PAD4 on CXCR2 and CXCR4 expression and neutrophil chemotaxis. No difference in neutrophil CXCR4 receptor expression was found between cancer patients and healthy donors and treatment with GSK484 did not affect expression (Fig. 5G). Quite differently, the surface level of CXCR2 receptor was highly upregulated on neutrophils from cancer patients compared to healthy donors. A four-hour treatment with PAD4 inhibitor GSK484 did not affect viability of neutrophils (Supplementary Fig. S8B) but dramatically reduced CXCR2 receptor expression in neutrophils from both healthy donors and cancer patients (Fig. 5G) resulting in decreased chemotaxis of neutrophils toward IL-8 (Fig. 5H). Pharmacological inhibition of PAD4 in T cells from healthy donors or cancer patients slightly decreased their proliferation (Supplementary Fig. S9A) and reduced IFN-γ production by CD4⁺ T cells (Supplementary Fig. S9B,C).

We performed a time-course analysis of CXCR2 and CXCR4 gene and protein expression in neutrophils treated with PAD4 inhibitor. One-hour treatment with GSK484 significantly downregulated CXCR2 gene expression while not affecting the surface level of CXCR2 protein (Fig. 5I). A significant downregulation of CXCR2 receptor was observed only 4h after treatment with PAD4 inhibitor. GSK484 did not affect gene or protein expression of CXCR4 (Fig. 5J). Taken together, these data indicate that PAD4 transcriptionally regulates CXCR2 expression in neutrophils.

Targeting of PAD4 enhances the anti-tumor effect of immunotherapy

We asked whether deletion of PAD4 in PMN and PMN-MDSC would improve the effect of immune checkpoint blockade (ICB). LL2-bearing PAD4^fl/flMRP8Cre⁺ mice and control PAD4^fl/flMRP8Cre⁻ littermates were treated with anti-CTLA-4 and anti-PD-1 antibodies. As anticipated, primary tumor growth was moderately reduced in mice with PAD4-deficient neutrophils. Similar effect was observed in control mice receiving checkpoint inhibitors. However, in PAD4^fl/flMRP8Cre⁺ mice receiving the ICB, primary tumor growth was dramatically reduced (Fig. 6A and Supplementary Fig. S10A). Lung metastases in PAD4^fl/flMRP8Cre⁺ mice were substantially reduced in comparison to control mice (Fig. 6B). In control mice, treatment with checkpoint inhibitors significantly decreased presence of lung metastases, an effect that was markedly improved in PAD4^fl/flMRP8Cre⁺ mice. Similar effects of ICB on primary tumor growth (Fig. 6C and Supplementary Fig. S10B) and lung metastases (Fig. 6D) were observed in B16F10-bearing PAD4^fl/flMRP8Cre⁺ mice and control PAD4^fl/flMRP8Cre⁻ littermates. PAD4 deficiency in PMN and PMN-MDSC, or treatment of control mice with ICB, both led to the increased presence of activated CD8⁺ T cells in tumor tissues; this effect was significantly increased when PMN and PMN-MDSC PAD4 deficiency and ICB were combined (Fig. 6E).

Figure 6. Deletion of neutrophil PAD4 improves anti-tumor effect of immune checkpoint blockade.

LL2 (A,B) or B16F10 (C-E) tumors were established in PAD4^fl/flMRP8Cre⁺ (Cre⁺) mice and their control PAD4^fl/fl MRP8Cre⁻ (Cre⁻) littermates by s.c. injection of tumor cells (n=5 per group). Mice were treated with anti-PD1 and anti-CTLA4 antibodies (200 μg/mouse) or isotype control (iso) given twice a week, i.p. for a total of 4 injections. A,C, Kinetics of primary tumor growth. Statistics: two-way ANOVA. B, LL2 lung metastasis evaluated by IVIS. D, B16F10 lung metastasis evaluated visually (left) and quantitated using Image J software (right). E, Proportions of B16F10 tumor-associated TNF-α and IFN-γ producing CD8⁺ T cells. Individual values, mean, and SEM are shown. Statistics: One-way ANOVA with Tukey’s multiple comparisons test.

We used JBI-589, a novel selective PAD4 inhibitor (Supplementary Fig. S11A and Supplementary Table S4) with an excellent oral bioavailability and potential to be translated into the clinic to investigate the effect of pharmacological blockade of PAD4 on tumor progression. In vitro, JBI-589 inhibited PMA-induced citrullination of histone H3 in mouse neutrophils in a dose-dependent manner (Supplementary Fig. S11B) without noticeable toxicity (Supplementary Fig. S11C). JBI-589 demonstrated excellent oral bioavailability with maximum concentration in plasma achieved 0.5h after oral administration and half-life of 6.3h (Supplementary Table S5 and Supplementary Fig. S11D). No effect of JBI-589 on viability of LL2 and B16F10 tumor cell lines was observed in vitro (Supplementary Fig. S11E). As anticipated, JBI-589 significantly downregulated CXCR2 expression on neutrophil (Supplementary Fig. S12A) and reduced their CXCL1-induced migration (Supplementary Fig. S12B) in vitro while it has no effect on suppressive activity of Ly6G⁺ cells generated in vitro from HPC in the presence of TES (Supplementary Fig. S12C). Treatment of LL2-bearing and B16F10-bearing mice with JBI-589 markedly reduced primary tumor growth (Fig. 7A and Supplementary Fig. S13). Treatment with ICB led to the inhibition of primary tumor growth, an effect that was significantly enhanced by the addition of JBI-589 (Fig. 7A). Analysis of lungs demonstrated the presence of metastasis in control mice while metastatic lesions were almost undetectable in JBI-589-treated mice (Fig. 7B,C). This effect was accompanied by a significant decrease in the presence of tumor-associated neutrophils (Fig. 7D) and increased infiltration of tumor tissues with activated CD8⁺ and CD4⁺ T cells (Fig. 7E,F). A significant down-regulation of CXCR2, but not CXCR4, receptor expression was found in Ly6G⁺ cells from JBI-589 treated mice (Fig. 7G). Treatment with JBI-589 did not affect cellularity of the BM in TB mice but reduced splenocyte number (Supplementary Fig. S14A) due to the decreased presence of CD11b⁺Ly6G⁺ cells, CD11b⁺Ly6C⁺Ly6G⁻ cells, and macrophages (Supplementary Fig. S14B–D). No effect of JBI-589 on the presence of CD8⁺ T cells or NK cells (Supplementary Fig. S14E,F) and a significant decrease in Tregs in spleens (Supplementary Fig. S14G) were observed. Presence of different progenitors was assessed in spleens of LL2-bearing mice receiving JBI-589 to exclude a potential inhibitory effect of this drug on spleen myelopoiesis. No differences in frequency of LSK, GMP, CMP, and MEP (Supplementary Fig. S15A) and their proliferation (Supplementary Fig. S15B) were found in spleens of JBI-589 treated and control mice. Since PAD4 has been previously implicated in NET formation, an ability of JBI-589 to block this process in vivo was evaluated. For that, BM and spleen neutrophils isolated from control and JBI-589 treated LL2-bearing mice were re-stimulated in vitro with LL2 cell supernatant. While neutrophils from control mice were undergoing extensive NETosis, BM (Supplementary Fig. S15C) and spleen (Supplementary Fig. S15D,E) neutrophils from JBI-589-treated mice were unable to produce NETs. As presence of M-MDSC and macrophages were reduced in spleens of JBI-589 treated tumor-bearing mice (Supplementary Fig. S14C,D), an effect of PAD4 inhibition of their functional activity was investigated. In vitro treatment of spleen M-MDSC (Supplementary Fig. S16A) or TAMs (Supplementary Fig. S16B) isolated from LL2-bearing mice with JBI-589 did not affect the expression of the genes encoding different soluble factors involved in their pro-tumorigenic effect. Myeloid cells were isolated from spleens and tumors of JBI-589-treated and control mice and their functional activity was evaluated. Pharmacological inhibition of PAD4 with JBI-589 did not affect expression of the genes encoding soluble factors by spleen and tumor PMN-MDSC (Supplementary Fig. S16C,D), spleen M-MDSC (Supplementary Fig. S16E), and TAM (Supplementary Fig. S16F) and did not affect suppressive activity of M-MDSC (Supplementary Fig. S16G), TAM (Supplementary Fig. S16H), and PMN-MDSC (Supplementary Fig. S16I).

Figure 7. Pharmacological inhibition of PAD4 improves anti-tumor effect of immune checkpoint inhibitors.

A-G, LL2-bearing mice were treated with JBI-589 (50 mg/kg, twice a day, daily, via oral gavage), anti-CTLA-4 and anti-PD-1 antibodies (200 μg/mouse, i.p., twice a week for 2 weeks), or their combination. A Kinetics of tumor growth was evaluated. **** - p<0.0001 in two-way ANOVA. B,C Lung metastases were evaluated by bioluminescent imaging using IVIS. Representative images (B) and individual data, mean, and SEM (C) are shown. Scale bar, 150mm. Statistics: two-tailed Student’s t-test. D-F, Proportion and absolute number of tumor-associated neutrophils (D), proportion of activated CD8 T cells (E) and CD4 T cells (F) in tumor tissues evaluated by flow cytometry in gated population of CD45+ cells. Statistics: one-way ANOVA with Tukey’s multiple comparisons test. G, Surface expression of CXCR2 and CXCR4 was evaluated by flow cytometry in CD45⁺CD11b⁺Ly6G⁺ cells. Statistics: unpaired two-tailed Student’s t-test.

Taken together, these data demonstrated that targeting neutrophil PAD4 could be a new approach for reducing metastatic disease and improving the efficacy of immunotherapy in cancer patients.

Discussion

In this study we identified a novel role of PAD4 in cancer: the regulation of neutrophil migration. Previously, the role of neutrophil PAD4 in cancer has been attributed primarily to its ability to citrullinate histone H3; this led to NETosis, a process of NET formation representing one of the mechanisms of neutrophil death (33). NETs have been detected in tumor-bearing hosts and their presence was associated with cancer progression (26). PAD4 KO mice demonstrated a delay in subcutaneous tumor growth, and this has been linked to an impaired ability of PMN to form NETs (34). However, only a fraction of PMN were shown to undergo NETosis (32), prompting the question of whether PAD4 had other functions in neutrophils.

Using mice with targeted deletion of PAD4 in neutrophils we confirmed a pro-tumorigenic effect played by PAD4 in these cells. Although PAD4-deficient neutrophils were unable to die via NETosis, we did not find an increased presence of these cells in tumor tissues. In contrast, the number of neutrophils detected in tumor tissues and metastatic site (lungs) was significantly decreased in PAD4 KO mice. This decrease was not due to a decrease in myelopoiesis in mice with PAD4-deficient neutrophils or increased apoptosis. Instead, it was due to dramatically reduced BM PMN migration. Consistently, blocking PAD4 activity with the selective inhibitor GSK484 significantly reduced PMN migration.

CXCR4 and CXCR2 are two major receptors that regulate PMN retention in the BM and release into circulation, respectively. In TB mice, CXCR2 signaling is important for attracting MDSC to the TME where they promote metastasis (35); inhibition of CXCR2 suppresses the recruitment of neutrophils into the tumor and premetastatic niche (36). In this study, we found that genetic deletion of PAD4 or pharmacological inhibition of PAD4 activity dramatically down-regulated CXCR2 expression at both RNA and protein levels. In contrast, the effect on CXCR4 expression was minimal. As PAD4 inhibition led first to a decrease in CXCR2 gene expression, followed by downregulation of CXCR2 protein, this suggested that PAD4 regulates CXCR2 on a transcriptional level. The cell surface levels of CXCR2 on neutrophils can be down-regulated rapidly by internalization upon ligand binding (37) and by metalloprotease activity following neutrophil stimulation with various nonligand stimuli (38). However, data on transcriptional regulation of CXCR2 are scarce. A recent computational analysis of the CXCR2 promoter predicted potential binding sites for HIF-1 and NF-κB (39). However, this prediction has yet to be confirmed in PMN experimentally. Besides histone H3, a number of PAD4 substrates have been identified including fibronectin, MMP9 (40), and others (41–43). Of these, NF-κB p65 has been reported in neutrophils (44). Citrullination of NF-κB p65 regulated its nuclear translocation and production of cytokines (44). Thus, it is a possible that PAD4 may regulate CXCR2 expression via citrullination of NF-κB. Further studies are needed to clarify the specific factor(s) responsible for transcriptional regulation of CXCR2 by PAD4.

Protein citrullination is mediated by a family of five PAD enzymes. Of them, PAD2 and PAD4 are highly expressed in neutrophils (33). Although a role of PAD2 in cancer cells has been reported (45), and recent data demonstrated that PAD2 is linked to the ability of macrophages to form extracellular traps (46), the contribution of PAD2 expressed by neutrophils in cancer progression remains largely unknown. In our study, the activity of PAD4 was markedly increased in neutrophils from TB mice, but no changes in PAD2 activity were found. In agreement, lack of PAD2 in mice did not affect tumor growth, whereas PAD4 deficiency significantly delayed tumor progression.

In contrast to neutrophils from mice, PMN from cancer patients showed significantly upregulated expression of PAD4 at the gene and protein levels, compared to healthy donors. Presently, the only transcriptional factor implicated in regulating PAD4 levels is p53. Tanikawa et al. demonstrated transactivation of PADI4 through an intronic p53-binding site in the U373MG cell line overexpressing p53 (47). However, it is unclear whether this mechanism occurs in human PMN and whether p53 activation status is altered in PMN from cancer patients compared to healthy donors. Data supporting this hypothesis demonstrated activation of p53 through phosphorylation at serine 15 in myeloid cells under cellular stress including chronic inflammation (48); additional work is needed to clarify the mechanisms regulating PAD4 expression.

Overcoming PMN-MDSC mediated immune suppression in the TME is one of the main challenges to generating potent antitumor immune responses. A number of strategies have been proposed for targeting MDSC, however, their therapeutic application remains limited (49). One of the biggest challenges is to selectively target PMN-MDSC but not PMN. CXCR2 is critically important for neutrophil migration to various tissues. However, it is not specific for PMN-MDSC and inhibition of CXCR2 blocks migration of all neutrophils. Since PAD4 is up-regulated primarily in PMN-MDSC, its targeting may provide an opportunity for more selective inhibition of PMN-MDSC migration.

In this study, we showed that pharmacological inhibitors of PAD4, including the newly developed compound JBI-589, activated anti-tumor immune response and significantly delayed tumor progression. As this anti-tumor effect was accompanied by a dramatic decrease in lung metastasis, upregulation of CXCR2, reduced PMN-MDSCs in tumors and metastatic sites, activation of T cells, and synergy with immune checkpoint blockade, our data suggest that tumor progression is regulated by PAD4-mediated neutrophil trafficking. Thus, our data describe a novel mechanism of regulation of neutrophil trafficking in cancer and demonstrate that targeting this mechanism has a potent anti-tumor effect. This opens an opportunity for a novel therapeutic approach targeting PAD4 in patients with solid cancers.

Statement of Significance.

PAD4 regulates tumor progression by promoting neutrophil migration and can be targeted with a small molecule inhibitor to suppress tumor growth and metastasis and increase efficacy of immune checkpoint blockade therapy.