Polymorphonuclear myeloid-derived suppressor cells limit antigen cross-presentation by dendritic cells in cancer

1Immunology, Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA, 19104; 2Department of Environmental and Occupational Health, Departments of Chemistry, Pharmacology and Chemical Biology, Radiation Oncology, University of Pittsburgh, 15219; USA # current affiliation H. Lee Moffitt Cancer Center, Tampa, FL, 33616 *-equal contributions as senior authors Correspondence: Valerian Kagan kagan@pitt.edu, Pitt, 130 DeSoto Street, 4126 Public Health, Pittsburgh, PA 15261, 412-624-9479 Dmitry Gabrilovich-dgabrilovich@wistar.org, The Wistar Institute, 3601 Spruce Str. Philadelphia, PA, 19104, Ph. 215-495-6955 Filippo Veglia – Filippo.Veglia@moffitt.org, Moffitt Cancer Center, 12902 USF Magnolia Drive, Tampa, FL 33612-9416. Ph: 813-745-4098


Introduction
Dendritic cells (DCs) are highly specialized antigen presenting cells (APC) (1,2). They can be broadly divided into three subsets: classical DCs (cDC), plasmacytoid DCs (pDCs), and inflammatory DCs. Classical DCs consist of two large groups: cDC type 1 (cDC1) and type 2 (cDC2). cDC1 are key players in the regulation of cancer immune responses due to their ability to cross present antigens to CD8 + T cells and to generate cytotoxic effector T cell responses (3)(4)(5)(6).
Their presence is critically important for the success of immunotherapies, such as adoptive T cell transfer and checkpoint blockade (7)(8)(9). Defective function of cDC1 was described in several types of cancer, which contributed to ineffective immune responses (3,5). We and others have found that accumulation of lipids dramatically affected the antigen cross-presentation by cDC1 resulting in an impaired induction of specific anti-tumor CD8 + T cell responses (10)(11)(12)(13). Recent studies demonstrated specific role of lipid bodies (LB) containing oxidatively truncated species of lipids in negative regulation of cross-presentation by cDC1 (10). However, the source of oxidized lipids in DCs remained unclear since the machinery for lipid oxidation is largely missing in cDC1. We and others have shown that accumulation of lipids in PMN-MDSC is a critical driver of their suppressive functions (16)(17)(18)(19). PMN-MDSC are characterized by high production of reactive oxygen species, as well as elevated level of myeloperoxidase (MPO), which may contribute to 6 mediated inhibition of DC cross-presentation. Separation of PMN-MDSC and CD103 + cDC1 using semi-permeable Transwell (0.4 µm) membrane did not abrogate the negative effect of PMN-MDSC on cross-presentation by cDC1. Direct presentation of peptide was not changed (Fig. 1G).
The data with Transwells suggested that PMN-MDSC affected DC cross-presentation via soluble factors. PMN-MDSC release different cytokines and factors (arginase, S100A8/A9, IL-6, PGE2, etc) that can inhibit DC function. However, they are known to affect broad functional activity of cells (20)(21)(22)(23), rather than selectively cross-presentation as observed in our experiments. We did not find changes in the ability of DCs to activate T-cell proliferation after direct presentation of short peptide or stimulation of allogeneic T cell proliferation (a hallmark of DC activity), as well as expression of MHC class I and major costimulatory molecules or PD-L1 ( Fig. 1C-E).
Since, the accumulation of lipids was previously shown to be responsible for the defective antigen cross-presentation by DC (10) and for the regulation of PMN-MDSC suppressive functions (18), we hypothesized that PMN-MDSC might inhibit antigen cross-presenting ability of DC by acting as a source of transferable lipids. To test this hypothesis, we assessed the lipids in cDC1 after coculture with PMN or PMN-MDSC and BODIPY 493/503 staining. By using confocal microscopy ( Fig. 2A) and flow cytometry (Fig. 2B) we found that co-culture with PMN or PMN-MDSC caused very substantial, but similar accumulation of lipids in cDC1. To confirm these data, we stained PMN or PMN-MDSC with BODIPY 493/503 prior to the co-culture with DC, washed the excess of BODIPY and cultured for 24h with DCs. Both PMN and PMN-MDSC were able to transfer lipids to cDC1 (Fig. 2C). Transfer of lipids did not require direct cell-cell contact. PMN-MDSC were stained with BODIPY 493/503. Excess of BODIPY was washed away and cells then were co-cultured with cDC1 separated by Transwells (0.4 µm). Transwells did not prevent the transfer of lipids from PMN-MDSC to DC (Fig. 2D). 7 To quantitatively characterize the transfer of lipids from PMN-MDSC to DCs, we loaded PMN-MDSC with deuterium labeled linoleic acid (LA-d4) and assessed the incorporation of LA-d4 into different types of lipids in DCs. PMN-MDSC derived free LA-d4 and its elongation product, free arachidonic acid (AA-d4), were found in DCs as non-oxidized fatty acids as well as their monooxygenated species (Fig. 2E, Fig. S3). Moreover, LA-d4 and AA-d4 esterified into non-oxidized TAGs and oxidized TAGs were detected in both PMN-MDSC and, importantly, in DCs (Fig. 2F,   Fig. S3). Thus, taken together, these data indicate that although both PMN and PMN-MDSC can transfer lipids to CD103 + cDC1, only PMN-MDSC affected cross-presentation by DCs. This suggested that the nature of lipids released from PMN-MDSC could be a factor that affects crosspresentation.
Previous studies implicated accumulation of lipid bodies (LB) in defective DC cross-presentation (10,11). We evaluated whether PMN and PMN-MDSC were able to facilitate accumulation of LB in DCs. Co-culture of cDC1 with either PMN or PMN-MDSC resulted in accumulation of LB in cDC1 ( Fig. 2A) indicating that appearance of large LB was not enough for defective crosspresentation. Previous studies demonstrated that only LB containing oxidatively truncated lipids, particularly TAGs, impaired DC cross-presentation (10). Thus, it appears that the chemical nature rather than the amounts of accumulated lipids may have an impact on DC cross-presentation. Therefore, we employed quantitative LC-MS analysis to assess the amounts of oxidativelytruncated TAGs in PMN and PMN-MDSC. We found that PMN-MDSC from TB mice had markedly higher amounts of oxidatively truncated lipids than PMN from tumor free mice (Fig.   3A). Similar species of lipids accumulated in DCs in the presence of tumor explant supernatants (TES) (Fig. S4) consistent with the previous observation (10). 8 PMN-MDSC have efficient oxidative machinery with MPO and NADPH oxidase as its key components. To determine whether MPO or NADPH oxidase are responsible for the generation of oxidized lipids in PMN-MDSC, we analyzed lipid profile in PMN-MDSC isolated from spleen of EL4 TB WT, MPO, or GP91 (component of NADPH oxidase complex) KO mice. We found that the amount of oxidatively truncated TAG was dramatically reduced to essentially the same low levels in both types of KO mice. In contrast only modest decrease in the amount of total TAG was detected (Fig. 3B).
Next we sought to determine the impact of MPO and GP91 on the ability of PMN-MDSC to block cross-presentation by DCs. We co-cultured cDC1 with PMN-MDSC obtained from spleen of LLC and EL4 TB MPO KO or GP91 KO mice. cDC1 were then loaded with OVA long or short peptides and used to stimulate OT-1 CD8 + T cells. The amount of lipids transferred to cDC1 by WT and KO PMN-MDSC was the same (Fig. S5A). Co-culture with WT PMN-MDSC altered crosspresenting ability of cDC1. However, co-culture with PMN-MDSC from MPO KO or GP91 KO mice did not have an effect cross-presentation of DCs (Fig 3C,D). The direct presentation of short peptide by cDC1 (Fig. 3C,D) and the expression of MHC I, CD40, CD80, CD86 and PDL-1 molecules on DC surface were not affected (Fig. S5B).
Next, we sought to assess the impact of MDSC on DC cross-presentation in vivo by depleting MDSC using anti-DR5 antibody described previously (24,25). Lung cancer cells with expression of OVA (LLC-OVA) were injected s.c. Two weeks later when tumors reached 0.5 cm, mice were treated with control immunoglobulin or DR5 antibody twice with 3 days interval. Reduction of PMN-MDSC and M-MDSC in spleens was verified by flow cytometry (Fig. S6A). Short treatment with DR5 antibody did not significantly affect tumor growth (Fig. S6B). cDC1 (CD11c + MHCII + CD103 + CD11b -CD172a -) and cDC2 (CD11c + MHCII + CD103-CD11b+CD172a+) were isolated from draining lymph nodes (dLNs) and used to stimulate OT-1 CD8 + T cells. cDC1 isolated from mice with MDSC depletion showed a higher ability to stimulate the proliferation of specific CD8 + T cells, compared to cDC1 isolated from untreated mice (Fig.   S6C). antitumor CD8 + T cells. To ascertain that observed effect was dependent on cross-presentation, we used BATF3 KO mice, which are depleted of cDC1 (27). In the absence of cDC1 in BATF3 KO mice the effect of combined treatment was abrogated (Fig. 5D).

Discussion
Our study suggested novel mechanism of negative regulation of cross-presentation by DCs in cancer, involving possible transfer of oxidized lipids from PMN-MDSC to cDC1. PMN-MDSC are dramatically expanded in many tumor models in mice and different types of human cancers and infiltrate tumors and lymphoid organs (15,28). These cells are characterized by immune suppressive activity, distinct biochemical and transcriptomics profile compared to normal neutrophils (15). Classical PMN respond to various pathogens and trauma by respiratory burst and degranulation with the release of anti-bacterial and antiviral enzymes. They rapidly die in the process. As a result, even activated classical PMN usually lack immune suppressive activity.
However, their antitumor activity is also difficult to detect. In contrast, PMN-MDSC, produce number of factors (arginase 1, prostaglandin E2, sustained level of superoxide and peroxynitrite, etc.) that inhibit function of T cells and other cells of immune system (15). As a result, these cells may blunt antitumor response and promote tumor progression and limit the effect of immune therapy of cancer. In mice, tumor PMN-MDSC are more suppressive than in spleens due to upregulation of Nos2 expression and several other suppressive mechanisms regulated by hypoxia as well as other factors present in tumors (28). It is known that spleen PMN-MDSC have relatively weak suppression of T cells function as compared to M-MDSC and tumor associated macrophages (14,28). Therefore, well established association between increased presence of PMN-MDSC in blood and negative outcomes in cancer patients was rather puzzling. Our study showed that PMN-MDSC can blunt the ability of DCs to cross-present antigens and thus demonstrated potential role of PMN-MDSC in blocking the priming phase of immune responses in cancer by affecting crosspresenting cDC1 and thus contributing to the resistance to checkpoint blockade therapy. 12 The conclusion that lipid transfer from PMN-MDSC to DCs can inhibit cross-presentation is based on several previous and current findings.
1). LB containing oxidatively truncated lipids blocked cross-presentation of DCs (10,11). LB containing electrophilic oxidatively truncated lipids covalently bound to chaperone heat shock protein 70. This interaction prevented the translocation of peptide-MHC class I complexes to cell surface by causing their accumulation inside late endosomes/lysosomes. As a result, DCs were no longer able to stimulate adequate CD8 + T cells responses (10). Loading of DCs with oxidized but not with non-oxidized fatty acids inhibited cross-presentation by DCs (10).
2 DCs, which make this mechanism less likely. More studies will be needed to clarify this mechanism. However, regardless of the specific mechanism, our findings suggest that targeting of MPO enhanced the effect of check-point inhibitor, which may suggest novel therapeutic opportunity. After 30h (for direct presentation) or 48h (for cross-presentation) or 96h (for allogenic MLR), 3 [H]-thymidine was added at 1 μCi per well for an additional 18 h followed by cell harvesting and a radioactivity count on liquid scintillation counter. Rf level was set to 60. Analytical data were acquired and analyzed using Xcalibur software.