Remodeling the Tumor Microenvironment Sensitizes Breast Tumors to Anti-Programmed Death-Ligand 1 Immunotherapy

Cancer immunotherapy using mAbs to target ligands [programmed death-ligand 1 (PD-L1)], receptors [programmed cell death protein 1 (PD-1), and cytotoxic T lymphocyte–associated protein 4 (CTLA-4)] of inhibitory immune checkpoints, which can be overexpressed by cancer cells, represents one of the biggest advances in cancer treatment in recent years. Durable responses with cancer immunotherapies have been observed in a variety of solid tumors (1). Two anti-PD-1 mAbs [pembrolizumab (KEYTRUDA, Merck & Co., Inc.) and nivolumab (OPDIVO, Bristol-Myers Squibb Company)] and three anti-PD-L1 mAbs [atezolizumab (TECENTRIQ, Genentech, Inc.), durvalumab (IMFINZI, AstraZeneca, Inc.), and avelumab (BAVENCIO, EMD Serono, Inc.)], have been approved by the FDA for clinical use in multiple tumors, and an anti-CTLA-4 mAb [ipilimumab (YERVOY, Bristol-Myers Squibb Company)] has been approved for the treatment of metastatic melanoma (2–4). However, several common cancer types such as breast, prostate, and colon have shown very low frequency of response to mAbs and there is an ongoing effort to increase the effectiveness of cancer immunotherapy (5). Antibodies targeting immune checkpoints regulate immune responses at different biological levels and by different mechanisms. These antitumor responses may be enhanced by means of combinatorial strategies that are intelligently designed and guided by preclinical models (6).

The extracellular matrix (ECM) is an essential component of the tumor microenvironment (TME) and can prevent immune cell infiltration, as well as promote immunosuppression and resistance to a checkpoint blockade (7, 8). Immune cell infiltration of the TME is important, as access of tumor-infiltrating lymphocytes (TIL), such as CD8⁺ and CD4⁺ T cells and natural killer (NK) cells, to tumors is a prerequisite for response to PD-L1 blockade (9). In addition, presence of TILs correlates with better patient outcomes for individuals receiving various antitumor therapies (10, 11). Myeloid-derived suppressor cells (MDSC) are important components of the TME and mediate T-cell suppression, contributing to resistance to checkpoint inhibitors (12). Furthermore, increased expression of genes involved in ECM organization significantly correlates with response of melanoma tumors to anti-PD-1 therapy (13). Hyaluronan (HA) is a naturally occurring glycosaminoglycan that is distributed widely throughout connective, epithelial, and neural tissues (14). In cancer, deposition of HA is seen early in tumorigenesis and persists during tumor progression and metastasis (15). Accumulation of HA in several solid tumors correlates with local invasion, presence of distant metastases, and poor overall survival (16–21). In breast cancer, inhibition of HA synthesis has been shown to suppress tumorigenicity in vitro and metastatic bone lesions in vivo (22). A strong association has been established between the level of HA around malignant cells, spreading of cancer, and patient outcomes in breast cancer (21). In terms of specific breast cancer subtypes, highly aggressive breast cancer cell lines such as MDA-MB-231 and HS-578T have been reported to express high levels of hyaluronidase-2 and synthesize high amounts of HA (23). In addition, HA has been shown to promote tumor growth in triple-negative breast cancers and is associated with HER2-positivity and reduced overall survival (24, 25).

Accumulation of HA in the TME has previously been shown to increase tumor interstitial pressure, inducing vascular collapse, which was associated with a decreased exposure of tumor cells to chemotherapy. Conversely, removal of HA in the TME can reverse these physiologic characteristics and, ultimately, improve accessibility of the tumor cells to chemotherapy in various mouse models of pancreatic ductal adenocarcinoma (15, 26–29). Furthermore, it was shown that depletion of HA can lead to an increased accumulation of CD8⁺ T cells in tumors of a mouse model of ovarian cancer (30).

Taken together, targeting the accumulation of HA by enzymatically inhibiting HA signaling or synthesis, and thereby degrading HA in the TME, is a promising therapeutic strategy (15, 16, 31–34). Pegvorhyaluronidase alfa (PEGPH20; PVHA; Halozyme Therapeutics, Inc.) is a novel, first-in-class biologic that enzymatically degrades tumor HA (33) and is in clinical development for the treatment of HA-accumulating tumors (35). In animal tumor models of pancreatic and ovarian cancer, PVHA has been shown to reduce tumor interstitial pressure, improve vascular perfusion, decrease hypoxia, and notably increase access of anticancer agents into the tumor (15, 30, 33, 36).

We hypothesized that enzymatic degradation of HA in the TME by PVHA in mouse models of breast cancer may improve response to immune checkpoint inhibition. The work described here indicates that removal of HA creates a more immunogenic tumor and points to potentially converting HA-accumulating breast tumors with low immune cell accumulation from resistant/refractory to sensitive to anti-PD-L1 immunotherapy when used in combination with PVHA.

Anti-mouse PD-L1 (clone 10F.9G2), rat immunoglobulin G2b (IgG2b) isotype control (clone LTF-2), and anti-mouse CD8α (clone 2.43) were purchased from Bio X Cell and diluted in PBS before administration. PVHA and its vehicle were provided by Halozyme Therapeutics, Inc. and used as described previously (33).

Murine syngeneic 4T1 (CRL-2539 Lot 60770568) and EMT-6 (CRL-2755 Lot 63226370) mammary carcinoma cells were obtained from the ATCC on September 24, 2015, and March 18, 2016, respectively. The 4T1 cell line was genetically engineered to express hyaluronan synthase 3 [HAS3 (4T1/HAS3)] by transduction of 4T1 with pLV-EF1a-hHAS3-IRS-Hyg (Lot 15C09) obtained from Biosettia, Inc. Cells were transduced for 6 hours with 2.1 × 10⁷ IU/mL lentiviral stock supplemented with 8 μg/mL Polybrene (Sigma-Aldrich, Inc., catalog no. TR-1003). After 48 hours, transduced cells were selected using 150 μg/mL hygromycin B (Gemini Bio Products, catalog no. 400-123). Cell lines were authenticated and confirmed to be of mouse origin with no mammalian interspecies or Mycoplasma contamination. A genetic profile was generated for the samples using a panel of short tandem repeat markers for genotyping. Sample profiles were confirmed as identical to the genetic profile established for these cell lines. The cell lines were maintained in RPMI1640 Medium (Mediatech, Inc.) supplemented with 10% FBS and 500 μg/mL hygromycin B at 37°C in a humidified environment with 5% CO₂.

Female BALB/c mice aged 6–8 weeks (Taconic) were inoculated orthotopically in mammary fat pads 5th or 10th, with 5 × 10⁵ syngeneic EMT-6 breast carcinoma cells or 1 × 10⁵ 4T1 or 4T1/HAS3 breast carcinoma cells. For treatment studies, mice were randomized into treatment groups when tumors reached approximately 100–150 mm³ (or as indicated size for larger tumor experiments). Anti-PD-L1 (5 mg/kg, or 0.5 mg/kg for EMT-6 experiments) or its isotype control, rat IgG2b, was administered by intraperitoneal injection at indicated dates.

PVHA (dosed at either 1, or 0.0375 mg/kg) or its vehicle (control), were administered by intravenous injection 24 hours prior to anti-mouse PD-L1 or its isotype control. Specific toxicity studies were not performed. PVHA treatment in the presence or absence of anti-PD-L1 appeared to be well tolerated. For depletion of CD8⁺ T cells, mice were injected intraperitoneally with 250 μg of CD8⁺-depleting antibodies (clone 2.43, Bio X Cell) 1 day before and 3 days after tumor inoculation. Tumors were measured twice per week by digital caliper, and tumor volumes calculated using the modified ellipsoid formula: 1/2 × (length × width²). Tumor growth, measured as tumor volume (mm³), assessed prior to study start with digital calipers and twice weekly thereafter until study termination. Tumor growth inhibition (TGI) was calculated using the formula [1 – (T_B − T_A)/(C_B − C_A)] × 100, where T_B is the average tumor volume (mm³) of treatment group at study termination, T_A the average tumor volume of treatment group at the day of randomization and first treatment, C_B the average tumor volume of control group at study termination, and C_A the average tumor volume of control group at the day of randomization and first treatment.

All animal-related experiments were performed in full compliance with approved institutional protocols and in accordance with guidelines established by the Animal Use and Care Administrative Advisory Committee at Halozyme Therapeutics, Inc. Power calculation analyses based on pilot in vivo experiments were performed before the main animal experiments to determine mouse numbers and dosing regimens. Representative experiments presented in this article were repeated three times, verifying results.

Spleens and tumors were harvested from tumor-bearing mice, and single-cell suspensions were prepared. Spleens were manually disassociated, and tumors were enzymatically dissociated with Liberase DH (Sigma-Aldrich, Inc.) and DNase I (Sigma-Aldrich, Inc.) using gentalMACS Octo Dissociators (Miltenyi Biotec). Surface staining was performed with antibodies against mouse CD8b⁺, PD-L1, CD45, CD19, NKp46, CD3, CD4, CD8a, CD44, and PD-1 (all eBioscience). For intracellular staining, cells were stained for surface markers and permeabilized with Foxp3 fixation buffer set (eBioscience) according to the manufacturer's directions. Fixed cells were stained with antibodies against Granzyme B (Thermo Fisher Scientific) and Foxp3 (eBioscience). All samples were acquired on Novocyte (ACEA Bioscience Inc.) and analyzed using NovoExpress software. Flow cytometry data were plotted using Prism Graphical Software version 7 (GraphPad Software, Inc.).

A DyLight 755 Labeling Kit, #84538 (Thermo Fisher Scientific), was used to label anti-PD-L1 antibody, clone 10F.9G2 (Bio X Cell). The kit was designed for excitation/emission 754/776 nm labeling of antibodies and proteins. The In Vivo Imaging System (Xenogen Corp.) was used to detect anti-PD-L1^DL755–expressing tumor cell uptake in 4T1/HAS3 and EMT-6 tumor-bearing mice. Animals were dosed at 5 mg/kg labeled anti-PD-L1 antibody. In vivo fluorescence and epi-illumination (from the top) images were acquired while mice were anesthetized with inhaled 2% isoflurane. After acquisition, images were opened as sequence files, set to the same scale (adjusting every image in the sequence simultaneously), and analyzed/processed for contrast and brightness by Living Image 4.3.1 Software (PerkinElmer Inc.). Fluorescence contrast was quantified by measuring the radiance within identical size ROI on the animal image.

To detect accumulation of HA, formalin-fixed, paraffin-embedded tissue sections were stained with a biotinylated proprietary HA probe (HTI-601; ref. 37) provided by Halozyme Therapeutics, Inc. For CD8⁺ detection, anti-CD8⁺ (eBioscience) was used. Sections were counterstained with hematoxylin. To detect CD8⁺, sections were stained on the DISCOVERY ULTRA (Ventana Medical Systems) and visualized with the DISCOVERY ChromoMap DAB Kit (Ventana Medical Systems).

Whole-section digital scans were obtained using an Aperio ScanScope AT Turbo slide scanner, and representative micrographs were captured using an Aperio ImageScope version 12.2.1.5005 (Leica Biosystems Inc.). A digital scoring algorithm was used to quantify the HA content of tissue sections (Aperio Positive Pixel Count V9, Leica Biosystems Inc.). For each tumor, the entire section was analyzed, apart from necrotic areas. Hyaluronan content was calculated as percentage of DAB- or HA-positive pixels over entire pixel count in the tissue.

Representative images were randomly chosen to reflect overall alterations in tissue staining/architecture of the respective experiments performed.

Portions of excised tumor tissues were incubated with proteinase K at 55°C to dissociate tumor tissue and digest proteins. Proteinase K enzyme was then heat inactivated by incubating at 95°C for 30 minutes and resulting sample lysate was subjected to HA quantification using Hyaluronan DuoSet ELISA (R&D Systems, Inc.) with a stringent quantitation range of 0.37–20 ng/mL. The concentration of HA was normalized by tissue wet weight and was reported as HA ng/mg tissue weight.

All data presented as mean ± SE of the mean, as indicated in the figure legend. Statistical analysis was carried out with Prism Software (GraphPad Prism v7) and P < 0.05 considered statistically significant. Multiple conditions and drug treatment conditions over time were compared using a parametric two-way ANOVA, followed by Tukey, unpaired two-tailed Mann–Whitney, or Bonferroni post hoc tests. Survival curves were estimated by the Kaplan–Meier method and compared using the log-rank test or Wilcoxon test, whichever was appropriate. Specific analyses used for each experiment are described in figure legends.

All other remaining data that support the findings of this study are available within the article and Supplementary Data or are available from the corresponding author upon reasonable request.

The effect of PVHA on intratumoral HA levels was investigated in three tumor models: naturally HA-accumulating murine EMT-6 mammary carcinoma; naturally low HA–accumulating murine 4T1 mammary carcinoma; and 4T1/HAS3 cells transduced to over-accumulate HA. To confirm presence of HA in the different models in vivo, cells from each line were inoculated orthotopically into female BALB/c mice 6–8 weeks of age. Levels of HA in resulting tumors were assessed by ELISA: HA-low (4T1) 361 ± 130 ng/mg; HA-accumulating (EMT-6) 730 ± 133 ng/mg; and HA over-accumulating (4T1/HAS3) 1,458 ± 190 ng/mg.

PVHA treatment degraded HA in all three breast cancer models compared with vehicle controls, as shown in Fig. 1 by ELISA and IHC. Using ELISA to detect HA, PVHA dosed at 0.0375 mg/kg enzymatically reduced HA levels of EMT-6 tumor-bearing animals by 64%, 650 ± 51 ng/mg in controls to 233 ± 95 ng/mg in PVHA-treated animals. PVHA also lowered 4T1 HA-accumulating tumors by 28%, 650 ± 51 ng/mg in controls to 323 ± 67 ng/mg in PVHA-treated animals. Animals bearing 4T1/HAS3 tumors were treated with a higher dose of 1 mg/kg, resulting in a reduction of 94% HA content by ELISA (HA content 1,647 ± 276 ng/mL in vehicle tumors and 95 ± 11 ng/mg in PVHA-treated tumors; Fig. 1A). These results were corroborated by IHC analysis (Fig. 1B and C). These data demonstrate that the breast cancer models tested accumulate HA, which can be depleted by PVHA treatment.

Next, we asked whether reduction in HA would lead to an increased anti-PD-L1 accumulation in tumor-bearing mice treated with PVHA. To this end, EMT-6 and 4T1/HAS3 tumor-bearing animals were divided into two groups: one treated with control buffer, and the other with PVHA as described in Materials and Methods. After 24 hours, animals received an intraperitoneal injection of fluorescently labeled anti-PD-L1 antibody at 5 mg/kg (Fig. 2). At days 3 and 4 post PVHA treatment, significant increases of intratumoral fluorescently labeled anti-PD-L1 antibody were observed in 4T1/HAS3 tumors (P = 0.0018, day 3 and P = 0.0122, day 4) and in EMT-6 tumors (P = 0.02, day 3) in mice treated with PVHA + anti-PD-L1 relative to those treated with anti-PD-L1 alone (Fig. 2A and B, respectively).

While the mechanism by which removal of HA led to an increased accumulation of labeled antibody as described above remains unknown, the data suggests anti-PD-L1 accumulation in the tumor may be enhanced when given in combination with PVHA. To test whether this accumulation resulted in improved efficacy of anti-PD-L1 treatment, 4T1/HAS3 tumor-bearing animals were dosed with either anti-PD-L1 alone or in combination with PVHA. Monotherapy treatment of either PVHA (1 mg/kg) or anti-PD-L1 (5 mg/kg) alone did not result in significant TGI compared with controls (Fig. 3A). However, when anti-PD-L1 was given in combination with PVHA (1 mg/kg), TGI increased by 56% compared with anti-PD-L1 + vehicle alone (P < 0.0001; Fig. 3). IHC staining of CD8 cells in these tumors is shown in Fig. 3B. This phenomenon was also seen when 4T1/HAS3 tumor-bearing animals were treated at a lower dose, 0.0375 mg/kg PVHA + anti-PD-L1 (5 mg/kg), which resulted in 40% TGI compared with vehicle + anti-PD-L1 (Supplementary Fig. S1A). These observations were recapitulated in the HA low-accumulating 4T1 model, albeit to a lesser extent; treatment with the combination of PVHA (0.0375 mg/kg) + anti-PD-L1 (5 mg/kg) led to a 28% TGI compared with vehicle + anti-PD-L1–treated animals (Supplementary Fig. S1B and S1C). To further investigate increased efficacy of anti-PD-L1 in combination with PVHA treatment, a survival study was performed. As shown in Fig. 3C, 4T1/HAS3 tumor-bearing animals treated with PVHA (1 mg/kg) + anti-PD-L1 (5 mg/kg) increased survival, 15–22 days, compared with animals treated with vehicle + anti-PD-L1 (P = 0.0262), indicating PVHA could improve the tumor growth inhibitory effects of anti-PD-L1 treatment in tumors that accumulate HA and that may not be responsive to anti-PD-L1 monotherapy.

To determine whether the tumor growth inhibitory effects of PVHA + anti-PD-L1 in 4T1/HAS3 tumor-bearing animals was associated with changes in the presence of various immune cells, flow cytometry analysis was performed. The immune cell composition of untreated parental 4T1 and 4T1/HAS3 tumors did not reveal significant differences in the density of immune cells (Supplementary Fig. S2A–S2J). Tumor-bearing animals were divided into the four different treatment groups outlined in Materials and Methods. One week after the last dose of anti-PD-L1 tumors were subjected to flow cytometry.

In 4T1/HAS3 tumor-bearing animals treated with PVHA (1 mg/kg) + anti-PD-L1 (5 mg/kg), the number of immune stimulatory cells, such as CD8⁺ T cells and NK cells, was significantly increased per gram tumor (Fig. 4A–D; Table 1), whereas the percent of CD11b⁺/Gr⁺ cells was decreased, when compared with control-treated animals. The density of tumoral CD8⁺ T cells present was increased 5-fold comparing vehicle + anti-PD-L1 (2.9 × 10⁴ cells/g tumor to 1.6 × 10⁵ cells/g tumor; P = 0.0093; Fig. 4A). When comparing the same tumors, NK cells increased 4-fold, from 2.1 × 10⁴ to 8.1 × 10⁴ cells/g tumor (P = 0.0093; Fig. 4C) and MDSCs Gr1⁺CD11b⁺ decreased 2-fold in density, from 4.1 × 10¹ to 1.9 × 10¹ cells/g tumor (P = 0.014; Fig. 4D).

In addition to increased density of immune stimulatory cells with the combination therapy, there was also a 3-fold increase in Foxp3⁺CD4⁺ T cells (from 8.2 × 10⁴ to 2.8 × 10⁵ cells/g tumor; P = 0.0093; Fig. 4B). It cannot be ruled out that these cells may have had an immunosuppressive effect. However, TGI was increased in PVHA and PVHA + anti-PD-L1–treated tumors as opposed to those treated with vehicle and vehicle + anti-PD-L1 (Supplementary Figs. S3A–S3G and S4A–S4K),indicating that the immune stimulatory effects outweighed the immunosuppressive effects of the increased regulatory T cells.

The correlation of increased TGI and increased density of CD8⁺ T cells in tumor-bearing animals, treated with the combination regime versus anti-PD-L1 monotherapy, led us to hypothesize that this observation of increased TGI may be caused by a CD8⁺ T-cell–dependent response. To test this hypothesis, a CD8⁺ depletion study was carried out using 4T1/HAS3 tumor-bearing animals. Both, tumor volumes and survival were examined. As described in Materials and Methods, animals were injected with CD8⁺-depleting antibodies 1 day before and 3 days after tumor cell inoculation. Once tumors reached a volume of 100–150 mm³, tumor-bearing animals were randomized and treated with PVHA (1 mg/kg) + anti-PD-L1 (5 mg/kg) once per week. Tumors of animals pretreated with anti-CD8⁺ followed by PVHA (1 mg/kg) + anti-PD-L1 (5 mg/kg) showed a significant increase in tumor volume compared with those animals that received control IgG2b antibody and combination therapy (1,000 mm³ ± 100 vs. 600 mm³ ± 70, respectively; P < 0.0001; Fig. 4E). These observations were recapitulated in the survival study, where animals pretreated with anti-CD8⁺ T cells exhibited a similar survival as vehicle control, 7–8 days, whereas IgG2b pretreated animals displayed an increase to 15 days (P = 0.010; Fig. 4F), suggesting the increased tumor growth inhibitory effect of the combination treatment (PVHA + anti-PD-L1) is CD8⁺ T-cell–dependent.

To test whether results seen in the HA-overexpressing 4T1/HAS3 breast cancer model would repeat in the EMT-6 naturally HA-accumulating breast cancer model, similar experiments as described above were carried out.

Unlike in 4T1/HAS3, EMT-6 tumors treated at a similar starting volume (∼150 mm³) were sensitive to anti-PD-L1 (5 mg/kg) monotherapy, exhibiting a TGI of 54% (P < 0.0001) compared with the vehicle control-treated group (Fig. 5A). However, adding PVHA (0.0375 mg/kg) increased the antitumor growth inhibitory effects of anti-PD-L1 by 46% compared with anti-PD-L1 monotherapy (P = 0.0069) and 79% compared with vehicle controls (P < 0.0001).

As shown above, the EMT-6 breast cancer model displayed a response to anti-PD-L1 monotherapy treatment, but the 4T1/HAS3 model was resistant. To investigate whether PVHA could overcome resistance to anti-PD-L1 treatment in the EMT-6 breast cancer model, we sought to develop an EMT-6 anti-PD-L1–resistant model via two methods; a dose reduction of anti-PD-L1, as well as by treating tumors of larger volumes.

A limited dose finding study in EMT-6 tumor-bearing mice was conducted to establish at which dose anti-PD-L1, as a monotherapy, was no longer efficacious. As shown in Supplementary Fig. S5, 0.5 mg/kg anti-PD-L1 treatment no longer led to significant TGI compared with vehicle controls. However, when anti-PD-L1 at this dose was combined with PVHA (0.0375 mg/kg), TGI increased to 57% compared with controls (Fig. 5B; P < 0.0001).

It was recently shown in non–small cell lung cancer that baseline tumor size (a measure of tumor load, bulk, and burden), is a negative prognostic factor in response to immune checkpoint inhibitors (38). Along those lines, we executed similar experiments in tumor volumes that were roughly 3-fold larger (∼325 mm³). Larger EMT-6 tumors appeared resistant to anti-PD-L1 (5 mg/kg) monotherapy (Fig. 5C). However, when PVHA (0.0375 mg/kg) was given in combination with anti-PD-L1 (5 mg/kg), tumor growth was significantly inhibited compared with vehicle + anti-PD-L1 (TGI 65%; P = 0.0209; Fig. 5C, top). These data were further corroborated by a survival study. EMT-6 tumor-bearing animals treated with anti-PD-L1 alone, at the larger staging tumor volume, displayed a survival of 7 days. Survival increased in animals receiving PVHA (0.0375 mg/kg) + anti-PD-L1 (5 mg/kg) to 15 days (P = 0.0262; Fig. 5C, bottom).

Taken together, these data suggest resistance to anti-PD-L1 monotherapy in the EMT-6 model as a result of either using a nonefficacious dose, or larger tumors, can be overcome by adding PVHA.

Equally as described for the 4T1/HAS3 breast cancer model, naturally HA-accumulating EMT-6 tumors were also analyzed via flow cytometry. The immune cell composition of EMT-6 tumors, shown by study day and tumor size in Supplementary Figs. S6A–S6M and S7A–S7G, respectively, did not reveal any significant differences in various immune cells over time.

EMT-6 tumors were subjected to flow cytometry 7 days after treatment on study day 12 (18 days post inoculation; Fig. 6A–D). Tumor-bearing animals treated with PVHA (0.0375 mg/kg) + anti-PD-L1 (5 mg/kg) had both CD8⁺ T cells and NK cells increased 2.2- and 2-fold, respectively, when compared with anti-PD-L1 alone (P = 0.043 for CD8⁺ T cells; Fig. 6A and P = 0.043 for NK cells; Fig. 6C; Table 1). In addition, a slight decrease in percentage of CD11b⁺Gr1⁺ MDSCs among CD45⁺ immune cells in the PVHA + anti-PD-L1 group was observed versus anti-PD-L1 treatment alone, although not significant (Fig. 6D; Table 1).

Similar to our observations in the 4T1/HAS3 model, we observed an increase of regulatory T cells in the combinatorial treatment regimen. Foxp3⁺CD4⁺ regulatory T cells displayed a mild 1.8-fold increase when compared with anti-PD-L1 (P = 0.008; Fig. 6B). While regulatory T cells have an immunosuppressive effect, the increased TGI suggests the effect of increased cytotoxic T cells, as well as NK cells, may have been a greater contributor to response than the effect of increased numbers of regulatory T cells.

In brief, the data indicates that in 4T1/HAS3, as well as in EMT-6 models of breast cancer, PVHA increased the tumor growth inhibitory effects of anti-PD-L1, which correlated with an increased number of CD8⁺ T cells, as well as NK cells, via removal of HA.

Breast cancers are generally regarded as poorly immunogenic and with the exception of antibodies targeting HER2, demonstrate poor responses to immunotherapy (39). While the mechanism behind these observations is unknown, there is evidence suggesting HA accumulates in the ECM of breast cancer; levels are significantly higher in serum of patients with metastatic breast cancer compared with healthy individuals (21, 30, 39). PVHA has been reported to remove HA in tumors of pancreatic and ovarian preclinical models (15, 29, 30, 33). While to date, there is no description of a clear relationship between a poor response to immunotherapy and levels of HA in patients with breast cancer, we hypothesized that HA removal may improve efficacy of immune checkpoint inhibitors. This idea based on several lines of evidence showing, in preclinical models, enzymatic degradation of HA can increase efficacy of various types of chemotherapy by reversing vascular collapse, as well as reducing interstitial pressure (32, 36).

We tested this hypothesis in three different orthotopic models of breast cancer EMT-6, 4T1, and 4T1/HAS3. In all three models, PVHA treatment resulted in removal of tumoral HA in vivo. This depletion of HA was correlated with an increased anti-PD-L1 accumulation, as shown with a fluorescently labeled anti-PD-L1 antibody, although the kinetics of tumoral HA removal and peak distribution of anti-PD-L1 did not precisely overlap (Fig. 2). This work does not describe the mechanism that underlies this observation and remains to be elucidated. Historical evidence indicates PVHA increases tumor cell exposure to chemotherapy and antibodies via removal of HA, but it has not been confirmed whether this is solely due to reductions in interstitial fluid pressure and changes in vessel patency (15, 29, 30, 33, 34, 36), or whether other biochemical or genetic changes in the TME may also be contributing. Along those lines, further investigation may help clarify whether tumoral HA depletion elicits changes in antigen expression, which may influence anti-PD-L1 accumulation.

Increased accumulation of anti-PD-L1 in tumor-bearing mice treated with combination therapy correlated with increased TGI in 4T1/HAS3, as well as in the EMT-6 tumor model. Flow cytometry analysis revealed parallel increases in cytotoxic T cells as well as NK cells. However, the density of regulatory T cells (Foxp3⁺CD4⁺ T cells) also increased, albeit to a lesser extent. Regulatory T cells are known to be immunosuppressive (40). The activity of tumoral immune cells was not explored, therefore we cannot rule out the possibility of immune suppression from regulatory T cells. However, CD8⁺ T-cell depletion studies showed that the decreased tumor growth in the combination treatment was dependent on the presence of CD8⁺ T cells. This may suggest that PVHA potentially converts immune checkpoint inhibitor–resistant tumors to sensitive tumors, by increasing the density of cytotoxic T cells.

This concept further supported by experiments in the EMT-6 tumor model used to create anti-PD-L1–resistant tumors. We attempted to create such a model by reducing the dose of anti-PD-L1 and treating larger tumors. It is unknown if these mechanisms are responsible for resistance to immune checkpoint inhibitors in breast cancer. Recently it was shown that baseline tumor size correlated to resistance to anti-PD-L1 treatment in non–small cell lung cancer (39). In both cases, PVHA was able to increase the tumor growth inhibitory effects of anti-PD-L1, as measured by tumor volume, as well as animal survival.

Taken together, these data suggest that PVHA has the potential to convert a resistant refractory tumor to sensitive to anti-PD-L1 treatment. In addition, the data expands our understanding of PVHA and can provide additional support for potentially meaningful clinical responses to treatment in HA-accumulating tumors (32, 36, 41, 42).

Clinical trials evaluating PVHA + anti-PD-L1 treatment in patients with metastatic pancreatic ductal adenocarcinoma (Morpheus – Pancreatic Cancer; NCT03193190), gastric cancer (Morpheus – Gastric Cancer; NCT03281369), and cholangiocarcinoma/gallbladder cancer (HALO 110-101; NCT03267940) are ongoing. The Morpheus studies are phase Ib/II trials investigating multiple immunotherapy-based combination treatments, including the combination of PVHA and atezolizumab. The HALO 110-101 study is a phase I trial assessing PVHA with cisplatin and gemcitabine with or without atezolizumab. All three clinical trials are currently enrolling patients.

R. Clift has ownership interest (including stock, patents, etc.) in Halozyme Therapeutics. J. Souratha has ownership interest (including stock, patents, etc.) in Halozyme Therapeutics. B. Blouw is a senior scientist at and has ownership interest (including stock, patents, etc.) in Halozyme Therapeutics. S.A. Garrovillo has ownership (stock) in Halozyme Therapeutics. No potential conflicts of interest were disclosed by the other author.

Conception and design: R. Clift, J. Souratha, S. Zimmerman

Development of methodology: R. Clift, J. Souratha, S.A. Garrovillo, S. Zimmerman, B. Blouw

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Clift, J. Souratha, S.A. Garrovillo, S. Zimmerman, B. Blouw

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Clift, J. Souratha, S. Zimmerman, B. Blouw

Writing, review, and/or revision of the manuscript: R. Clift, J. Souratha, S. Zimmerman, B. Blouw

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Clift, J. Souratha, S. Zimmerman

Study supervision: R. Clift

The authors thank and acknowledge Yujun Huang for his contribution to this study; Kelly Chen and Trevor Kimbler for assistance with in vitro work; Genaro Ronquillo, Max Bersabe, James Skipper, and Clay Conway for their assistance with in vivo studies; and Michael Jorge and Flavio Araiza for their assistance with HA ELISA processing. Medical writing support and editorial assistance was provided by Talya Underwood, MPhil, and Shaun W. Foley, BSc (Hons), of Paragon, Knutsford, United Kingdom, supported by Halozyme Therapeutics, Inc. This study was supported by Halozyme Therapeutics, Inc.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.