Detection of PD-L1-expressing myeloid cell clusters in the hyaluronan-enriched stroma in tumor tissue and tumor-draining lymph nodes

Expression of the transmembrane protein PD-L1 is frequently up-regulated in cancer. Since PD-L1-expressing cells can induce apoptosis or anergy of T lymphocytes through binding to PD1 receptor, the PD-L1-mediated inhibition of activated PD1+ T cells considered as a major pathway for tumor immune escape. However, the mechanisms that regulate the expression of PD-L1 in the tumor microenvironment not fully understood. Analysis of organotypic tumor tissue slice cultures, obtained from tumor-bearing mice as well as from cancer patients, revealed that tumor-associated hyaluronan (HA) supports the development of the immunosuppressive PD-L1+ macrophages. Using genetically modified tumor cells, we identified both epithelial tumor cells and cancer-associated fibroblasts (CAFs) as the major source of HA in the tumor microenvironment. HA-producing tumor cells and, in particular CAFs of bone marrow origin, directly interact with tumor-recruited Hyal2+ myeloid cells forming the large stromal congregates/clusters that are highly enriched for both HA and PD-L1. Furthermore, similar cell clusters comprising of HA-producing fibroblasts and PD-L1+ macrophages were detected in the tumor-draining lymph nodes. Collectively, our findings indicate that the formation of multiple large HA-enriched stromal clusters that support the development of PD-L1-expressing antigen-presenting cells in the tumor microenvironment and draining lymph nodes could contribute to the immune escape and resistance to immunotherapy in cancer.


Introduction
The immunosuppressive ligand PD-L1 plays an important role in the regulation of T cell-mediated immune response and tumor-associated immune tolerance (1). Recent studies demonstrated that PD-L1 expression by host's myeloid APCs, is essential for PD-L1 mediated immune evasion and immunotherapy (2)(3)(4). Bladder cancer is characterized by a highly immunosuppressive microenvironment including up-regulated expression of inhibitory ligand PD-L1 (5, 6) and the strong presence of immunosuppressive myeloid cells such as myeloid-derived suppressor cells (MDSCs) and tumor-associated macrophages (TAMs) (7)(8)(9). The tumor cells can up-regulate the PD-L1 expression in tumor-associated macrophages (6), however, the mechanism of tumormediated regulation of PD-L1 expression in myeloid cells remains unresolved.
Tumor stroma plays a major role in tumor growth and is comprised of both cellular and extracellular components. Bladder cancer is enriched for both cancer-associated fibroblasts (CAFs) as well as for one of the major components of extracellular matrix hyaluronan (HA) (10-11). CAFs play very diverse roles in tumor development and progression including stimulation of tumor-promoting inflammation, tumor proliferation, and invasion, neovascularization as well as tumor-associated immune suppression (12)(13)(14). Several lines of evidence suggest a strong interplay between CAFs and myeloid cells. Thus, fibroblasts play pivotal roles in the recruitment of CCR2expressing monocytic myeloid cells and polarization of those recruited cells to the M2 macrophages or immunosuppressive MDSCs (15,16). Roles of CAFs in the recruitment of monocytes/macrophages to the tumor tissue supported by fibroblast's production of monocytemacrophage chemotactic factor CCL2. Moreover, HA deficiency in tumor stroma resulted in a marked reduction of both macrophage recruitment and tumor neovascularization (16)(17). 4 Here we demonstrate that tumors stimulate the gathering of stromal cells into large cell clusters. The major cellular components of these clusters include HA-producing fibroblasts, tumor epithelial cells, and macrophages. Similar stromal structures with HA-producing fibroblasts and PD-L1 + macrophages were detected in tumor-draining, but not in distant lymph nodes. Our results support the idea that stroma-produced HA directly supports the development of immunosuppressive PD-L1-expressing macrophages

L1-expressing macrophages
To explore the development of PD-L1 + macrophages in the tumor microenvironment, we utilized the organoid tumor tissue-slice technique. To this end, we injected MBT2, the murine bladder tumor cells into syngeneic C3/He mice surgically excised the developed tumors, and prepared the precision-cut 200-300 micron tumor tissue slices using the Compresstome tissue slicer. Prepared tissue-slices are cultured in 24-well plates in a complete culture medium. The viability of tissue slices was monitored using the Live/Dead viability assay (Invitrogen).
Six to twelve hours after the initiation of tumor tissue slice cultures, we noticed the development of multiple stromal cell conglomerates/clusters firmly attached to the plastic ( Fig.   1a and Supporting Fig. S1). Staining for PD-L1 revealed that these cell clusters are highly positive for this marker. Intriguingly, the stromal cell clusters are highly enriched for the HA. The majority of the PD-L1 + cells were also positive for pan-hematopoietic marker CD45 ( Supporting   Fig. S2a). These data implicate the potential involvement of tumor stroma-associated HA in the 5 development of PD-L1 + cells. A recently published study indicates that lymph nodes) could be subjected to the preparation of tissue slices to study ex vivo immune response (18). Accordingly, we collected from MBT2 tumor-bearing mice the tumor-draining and distant control LNs. Freshly collected LNs were used for the preparation of precision-cut tissue slices. A few days later, the LN tissue slice cultures were stained for the presence of PD-L1 and HA. To our surprise, the draining LNs (Fig.1b), similar to the tumor tissue slices, was also able to develop the HA-enriched adherent stromal cell clusters with incorporated PD-L1 + macrophages. In contrast, the distant LNs were not able to produce any HA-enriched PD-L1 + stromal cell clusters that we observed in the cultures prepared from tumor tissues or tumor-draining LNs. Co-staining the tumor stroma with antibodies against PD-L1, HA, and macrophage marker F4/80 revealed (Supporting Figs. S2b and S3a) that cluster-associated PD-L1 + cells co-express F4/80. In addition to the mouse bladder tumor model, similar stromal HA-enriched PD-L1 expressing clusters were observed in clinical cancer tissue samples obtained from patients with bladder cancer (Fig.2a and Supporting Fig.   S3b). Taken together, these data indicate that tumor stroma and stroma-associated HA support the development of PD-L1 + macrophages.

HA-producing fibroblasts of bone marrow origin contribute to the development of PD-L1 + macrophages
Cancer-associated fibroblasts (CAFs) represent one of the major components of tumor stroma.
CAFs exert diverse functions, including matrix deposition and remodeling, extensive reciprocal signaling interactions with tumor cells and infiltrating immune cells (10). The origin of these cells not fully understood, however some CAFs clearly demonstrate a bone marrow origin (19). In addition, it was reported that bone marrow-derived mesenchymal cells with fibroblast-like 6 appearance constitutively produce HA (20). Here we show that bone marrow-derived cells obtained from normal naïve mice and stimulated with recombinant fibroblast growth factor (FGF2) and IL-1β give a rise to HA-producing fibroblast-like cells and PD-L1 + macrophages in the absence of tumor cells (Fig.2a). It appears that IL-1β and FGF2 have a synergistic effect on development of HA-enriched PD-L1 + clusters, because stimulation of bone marrow cells separately with these cytokines results in much weaker effects. Furthermore, similar clusters comprising of PD-L1 + macrophages and fibroblast-like HA-producing cell also detected in tumordraining lymph nodes (Supporting Fig. S4a). Collectively, these data suggest that HA-producing fibroblasts of bone marrow origin could contribute to the development of PD-L1 + macrophages in tumor-free conditions.
To get a better insight into the formation of stromal cell clusters in tumor microenvironment, we next applied the time-lapse video take during consecutive 96 hours after initiation of tumorslice culture. Data presented in Supporting Figs S4b, S5 and tame-lapse video (not shown) demonstrate that formation of macrophage-fibroblast clusters is highly dynamic and interactive process, when both large irregularly-shaped fibroblast-like cells and smaller round-shaped macrophages moving around each other during formation of stromal cells clusters. Fig.2b illustrates that fibroblast-secreted HA (green) physically interacts with PD-L1 + macrophages (red).
These data clearly indicate that HA is directly involved in the development of PD-L1 + macrophages. 7 To delineate the roles of CAFS and epithelial tumor cells in the development of PD-L1 + cells, we stably transfected murine MBT2 tumor cell line with GFP using lentivirus. GFP-expressing tumor cells were injected in mice, and 2 weeks later harvested tumor was used for the preparation of tumor tissue-slices (Fig.3a). Data presented in

Tumor cells promote the development of PD-L1 + macrophages in an HA-dependent manner
The co-culture of the tumor cells with Gr-1 + MDSCs leads to the up-regulation of PD-L1 expression and development, promoting differentiation of myeloid-derived cells into PD-  Fig. 3c demonstrate that inhibition of HA synthesis results in a dose-dependent reduction of PD-L1 expression.

PD-L1 expression in myeloid cells, we have co-cultured MBT2 tumor cells and murine Gr-1 + cells with added HAS inhibitor 4-methylumbelliferone (4-MU). Data presented in
The MDSCs may affect the HA metabolism in tumor tissue through membrane-bound enzyme hyaluronidase 2 (Hyal2) (21). Hyal2 is a rate-limiting enzyme, which upon activation able to degrade the extracellular HA into small fragments with low molecular weight (LMW-HA), and its expression was increased in both tumor-associated and blood-derived myeloid cells in cancer patients with bladder cancer. To examine whether Hyal2-expressing cells could potentially contribute to the HA-mediated development of PD-L1 + macrophages, we co-cultured Gr-1 + MDSCs and MBT2 tumor cells for 5 days and then co-stained with CD45 (pan-hematopoietic marker), PD-L1, and Hyal2. Data presented in Fig.3d indicate that PD-L1 + cells co-express the Hyal2 enzyme. Next, we checked the tumor-associated CD11b myeloid cells (Fig.4a) for the expression of PD-L1 and Hyal2. To this end, we isolated CD11b cells from murine MBT2 bladder tumor tissue and stained for those markers. Data presented in Fig. 4b and Supporting Fig. S8a indicate that majority of PD-L1 + cells also-co-express the Hyal2 enzyme.
The activity of Hyal2 in myeloid cells can be stimulated with tumor-conditioned medium (TCM) or IL-1β (21). To examine whether this activation is associated with the up-regulation of PD-L1 expression, we used the murine CD11b cells isolated from normal bone marrow. Data presented in Fig.4c demonstrate that similarly to its's human counterpart, the TCM-activated murine myeloid cells able to degrade extracellular HA. A similar extent of HA degradation was detected while analyzing HA produced by MBT2 tumor-tissue slices, but not by the MBT2 tumor cell line (Fig.4d). These data were confirmed by gel electrophoresis (Fig. 4e). Specifically, the 9 electrophoretic analysis confirmed that HA produced by MBT2 tumor cell line consisted of fragments with intermediate size (MW ~20 kDa), whereas tumor tissue slices-derived HA showed lower molecular weight (<10 kDa). Since the Hyal2 enzyme specifically degrades HA to fragments with MW 20 kD (22,23), other types of hyaluronidases could likely be involved.
To address this question, we measured the expression of Hyal1 and Hyal3 using qRT-PCR in MBT2 tumor cell line, whole tumor tissue from tumor-bearing mice, and CD11b cells isolated from the tumor. Levels of Hyal3 were very low, whereas the expression of Hyal1 was up-regulated in myeloid cells as compared to the MBT2 tumor cell line (Fig.4f). Hyal1 is an intracellular lysosomal enzyme that degrades the internalized HA to very small fragments with low molecular weight less than 5kD (LMW-HA). It has been proposed that membrane-bound enzyme Hyal2, which breaks down the extracellular HA, works in concert with Hyl1 to produce the LMW-HA (22)(23). Taken together, our data indicate that is very likely that both the Hyal2 and Hyal1 enzymes are involved in the degradation of tumor-associated HA.  Fig. 6b) and kidney carcinoma Renca (Supporting Fig. S8b). These findings indicate that the formation of multiple stromal clusters enriched for HA-producing fibrobalasts, tumor cells and PD-L1 + antigen-presenting cells may represent a general mechanism of immune escape in cancer. 11 HA has been implicated in regulating a variety of cellular functions in both tumor cells and tumor-associated stromal cells, suggesting that altered HA levels can influence tumor growth and malignancy at multiple levels. Previously published studies demonstrate that HA increases the proliferation rate of tumor cells in vitro and promotes cell survival under anchorage-independent conditions (11,26). At the molecular level, HA activates the PI3K/Akt pathway and influences the expression of cell cycle regulators. Furthermore, HA also can inhibit tumor cell apoptosis, as demonstrated by experimentally modifying HA levels. Importantly, increased HA production in cancer is frequently associated with enhanced HA degradation due to high levels of expression/activity of hyaluronidases (27,28). Increased HA degradation leads to the accumulation of HA fragments with low molecular weight (LMW-HA) (29)(30). LMW-HA seems to have specific pro-tumoral functions by promoting inflammation, tumor angiogenesis, and metastasis through stimulating the production of cytokines, chemokines, growth factors in TLR2/TLR4 dependent manner (31). Conversely, the high molecular weight HA shows antiinflammatory and anti-oncogenic effects (32, 33). Our data support the idea that in addition to cancer-related inflammation and tumor angiogenesis, the tumor-associate HA also involved in the immune suppression in cancer, since HA supports the development of immunosuppressive PD-L1 + macrophages in both tumor tissues and TDLNs.

Discussion
Tumor-draining lymph nodes (TDLNs) are essential for the initiation of an effective antitumor T-cell immune response. However, tumors may affect the immune-initiating function of TDLNs (24). A recently published study demonstrated that TDLNs in tumor-bearing mice are enriched for both PD-L1 + antigen-presenting cells and tumor-specific PD-1 + T cells (25). TDLN-targeted PD-L1-blockade induced enhanced anti-tumor T cell immunity by seeding the tumor site with T cells, resulting in improved tumor control. Moreover, abundant PD-1/PD-L1-interactions in TDLNs were also observed in non-metastatic melanoma patients.
Our data demonstrate that similarly to the tumor tissue, PD-L1 + antigen-presenting cells in the TDLNs are associated with HA-producing fibroblasts (Fig. 1b, Supporting Fig.4a) and characterized by enhanced degradation of TDLN-associated HA (Supporting Fig. S9). Taking  Evaluation of HA size. Analysis of HA molecular weight was done using polyacrylamide gel electrophoresis as described previously (35). Briefly, conditioned medium tissue slices were centrifuged, aliquoted, and stored at -80 0 C. To prepare samples for HA size analysis, thawed samples were digested with proteinase K to remove proteins, benzonase for the depletion of nucleic acids (RNA, DNA), and ethanol to extract lipids was added. Samples along with HA standards were then subjected for polyacrylamide electrophoresis. The tissue-produced HA was visualized on the gel by staining with "Stains All" dye (Sigma-Aldrich). Time-lapse video. Tissue slices were seeded in a 24-well glass-bottom Grenier plate with 500ul of media. The surrounding wells were filled with 1000 ul of sterile PBS to provide adequate humidification. Slices were incubated for 2h before imaging. Plates were transferred to the BioTek Lionhart FX that was preheated to 37C with 5% CO2. Using 20x magnification, 5 to 10 beacons were chosen per well. To compensate for good variation, z-stacks of 15 slices of 4.2uM thickness.

Immunofluorescent microscopy and
Images were taken at 20 min intervals for 96 h. Images processing and video rendering were done using Gen5 Image Prime 3.10 (BioTek Instruments).  Representative images are shown.  Hyaluronan was visualized by staining with "Stains All" dye.