Research ArticleCancer

Extended release of perioperative immunotherapy prevents tumor recurrence and eliminates metastases

See allHide authors and affiliations

Science Translational Medicine  21 Mar 2018:
Vol. 10, Issue 433, eaar1916
DOI: 10.1126/scitranslmed.aar1916

Think globally, act locally in cancer

Immunotherapy for cancer is becoming increasingly common, but systemic immunotherapy can have major side effects because it stimulates the activity of the entire immune system and is not necessarily tumor-specific. Surgery, a classic mainstay of cancer treatment, has the drawback of temporarily suppressing the immune response at the site of tumor resection. To address both of these concerns, Park et al. designed hydrogel scaffolds that can gradually release agonists of innate immunity and then implanted these scaffolds into mouse models at the sites of tumor resection. This approach was safe and more effective than systemic or even local injection of immunotherapy, suggesting its potential for clinical testing.

Abstract

Cancer immunotherapy can confer durable benefit, but the percentage of patients who respond to this approach remains modest. The ability to concentrate immunostimulatory compounds at the site of disease can overcome local immune tolerance and reduce systemic toxicity. Surgical resection of tumors may improve the efficacy of immunotherapy by removing the concentrated immunosuppressive microenvironment; however, it also removes tumor-specific leukocytes as well as tumor antigens that may be important to establishing antitumor immunity. Moreover, surgery produces a transient immunosuppressive state associated with wound healing that has been correlated with increased metastasis. Using multiple models of spontaneous metastasis, we show that extended release of agonists of innate immunity—including agonists of Toll-like receptor 7/8 (TLR7/8) or stimulator of interferon genes (STING)—from a biodegradable hydrogel placed in the tumor resection site cured a much higher percentage of animals than systemic or local administration of the same therapy in solution. Depletion and neutralization experiments confirmed that the observed prevention of local tumor recurrence and eradication of existing metastases require both the innate and adaptive arms of the immune system. The localized therapy increased the numbers of activated natural killer (NK) cells, dendritic cells, and T cells and induced production of large amounts of type I interferons, thereby converting an immunosuppressive post-resection microenvironment into an immunostimulatory one. The results suggest that the perioperative setting may prove to be a useful context for immunotherapy, particularly when the release of the therapy is extended locally.

INTRODUCTION

Surgery is the most common intervention in oncology, but the resultant wound healing process can promote metastasis (1), which accounts for 90% of cancer-related deaths (2). Recurrence of primary tumors after surgery is associated with poor outcomes, increased local immunosuppression, and distant metastasis (3). A set of genes that mediate breast cancer metastasis to lung has been identified and includes the immunosuppressive factor cyclooxygenase-2 (COX-2) (4). Consistent with this finding, perioperative administration of nonsteroidal anti-inflammatory drugs, which inhibit COX-2, reduces relapse (5). This treatment is designed to counteract transient systemic inflammation that is caused by surgery. Such inflammation can initiate new metastases and awaken dormant existing micrometastases (6), which can disseminate before manifestation of clinically detectable primary tumors (7, 8). The immunosuppression associated with wound healing not only exerts direct effects on malignant tissue to promote cancer cell motility, invasion, and proliferation but also suppresses the activity of antitumor leukocytes, including cytotoxic T lymphocytes, natural killer (NK) cells, and dendritic cells (6). Factors in the serum can promote a “wound-response” gene signature that predicts local recurrence (9) as well as metastasis and poor survival (10). Because metastasis has such a grave impact on mortality, viable strategies to address metastasis are urgently needed.

Cancer immunotherapy is an attractive means of achieving long-term systemic immunosurveillance, but the response rate to immunotherapy requires improvement. In situ immunomodulation is a promising avenue of investigation. Notably, intratumoral immunotherapy is both safer and more effective than systemic administration (1113). Local delivery can elicit systemic antitumor immunity and lead to eradication of disseminated disease in mice, including in sites considered to be cancer cell sanctuaries (14). The related “abscopal effect” after localized radiation therapy has been observed in a murine model of breast cancer (15) and in patients (16). Concentrating the therapy at the site of disease allows one to break local immune tolerance, thereby enabling generation of systemic antitumor immunity in the absence of systemic exposure that can evoke severe side effects (17).

Biomaterials and nanotechnology can be used to alter the colocalization, biodistribution, and release kinetics of immunomodulatory compounds, thereby improving the efficacy and safety of these molecules relative to systemic administration (18). In one illustrative example underscoring the utility of immunoengineering, it was shown that the co-entrapment of an inflammatory cytokine, an immune danger signal, and a tumor antigen in a biodegradable scaffold controlled the activation and localization of host dendritic cell populations in situ, increasing vaccine efficacy (19, 20). Scaffolds are particularly well suited to immunomodulation (21). In another study, a scaffold was prepared to enhance the effectiveness of adoptive T cell therapy by delivering, expanding, and dispersing tumor-reactive T cells after incomplete tumor resection (22). Here, we sought to determine whether it might be possible to produce curative outcomes among mice harboring established disease through extended release of immunomodulatory small molecules or biologics in the perioperative setting in the absence of a vaccine or adoptive cell transfer. We show that delivering cancer immunotherapy in a controlled spatiotemporal manner using a hydrogel scaffold confers superior efficacy to bolus injection of the therapy administered intraperitoneally, intravenously, or locally. The ability to prevent local recurrence and address metastases, both of which bode poorly for survival, is largely lacking.

We hypothesized that focusing immunotherapy at the site of interest in the intraoperative setting would prove beneficial. Specifically, we reasoned that it would be easier to eliminate a small number of residual cancer cells in the absence of a concentrated immunosuppressive microenvironment than to treat an intact primary tumor, which has many means of immune evasion—both innate and adaptive, cancer cell–intrinsic and cancer cell–extrinsic. Whereas the majority of current efforts in oncology focus on eliminating the primary tumor, modulating the immune compartment of the tumor resection microenvironment may represent a promising approach to treating cancer, including distal metastatic disease.

RESULTS

A hydrogel extends the release of immunomodulatory payloads

A hydrogel scaffold was prepared by cross-linking hyaluronic acid in a mold (Fig. 1A). Hyaluronic acid is a U.S. Food and Drug Administration–approved biopolymer that is naturally found in the body and has many desirable properties for biomedical applications, including its biocompatibility, ability to adopt many physical forms, and ability to control drug release spatiotemporally (23). Most hyaluronic acid–based scaffolds are used for applications that involve encapsulation of cells, such as cell therapy and regenerative medicine (23), but we required a hydrogel with greater stiffness than that commonly used. To ensure sufficient mechanical integrity for manipulation with tweezers, we prepared more than 20 formulations before arriving at an optimal ratio of cross-linker to polymer. To inspect the stability and confirm the degradability of the cross-linked hydrogel, a fluorescent dye was covalently conjugated to the scaffold (fig. S1, A and B). The scaffold was then placed into the tumor resection site of a female BALB/cJ mouse that had been inoculated orthotopically with syngeneic 4T1-Luc2 (luciferase) breast cancer cells. The fluorescently labeled scaffold was monitored by fluorescence in vivo imaging system (IVIS) imaging, and it was determined that noteworthy scaffold degradation began at 5 weeks, with complete resorption achieved by 12 weeks (Fig. 1B and fig. S1C).

Fig. 1 A biodegradable hydrogel scaffold extends local release of payloads in situ.

(A) Picture of a representative scaffold loaded with R848. (B) A hydrogel scaffold to which Alexa Fluor 750 was conjugated was implanted into the resection site of a mouse after tumor removal, and fluorescence IVIS imaging was performed at the indicated time points. The lone mouse to survive surgery in the absence of additional therapy is shown. (C) Quantification of fluorescence IVIS imaging (shown in fig. S1, E and F) of non–tumor-bearing mice to whom dye was administered in solution or conjugated to a hydrogel placed by the fourth mammary fat pad after incision. (D) Fluorescence IVIS imaging depicting the in vivo release profile of a model small-molecule payload (Cy7-CA), administered locally either in solution or after loading into a hydrogel. (E) Quantification of the in vivo release profile of Cy7-CA. The experiment was performed once with n = 5 biological replicates. Fold difference is indicated for each time point. Statistics were calculated using a two-sided unpaired t test. Data are means ± SD. *P ≤ 0.05, ***P ≤ 0.001, ****P ≤ 0.0001.

Mice whose tumors were resected exhibited cachexia, lethargy, and wet fur in the absence of perioperative therapy, whereas untreated mice appeared healthy and rarely died from the primary tumors; they were sacrificed because of their large tumor volumes rather than any apparent morbidity. Untreated mice harboring 4T1-Luc2 tumors had to be euthanized within 7 weeks of tumor inoculation (median survival, 40 days), and surgical removal of the tumors provided little survival benefit (median survival, 44 days), because mice succumbed to tumor recurrence and metastasis in this model (fig. S1D), as is often observed among breast cancer patients (24). Because nearly all mice succumbed to disease after tumor resection alone, we repeated the in vivo degradation study in non–tumor-bearing mice, enabling us to increase the sample size. The shape of the scaffold degradation curve was very similar—including an apparent but modest decrease in signal intensity before the irreversible decline—to that of the lone surviving mouse whose tumor had been resected, although the scaffold’s stability was prolonged by ~50% (Fig. 1C and fig. S1E). After 12 hours (early time point) or 15 weeks (later time point), the site of implantation was subjected to histopathological analysis. No abnormalities were detected by a certified pathologist (R. Bronson), confirming that the hydrogel is highly biocompatible. For comparison, a solution containing the fluorescent dye was administered locally, revealing that the free dye diffuses away very rapidly if it is not conjugated to the scaffold (Fig. 1C and fig. S1F). These data confirm that cross-linked hyaluronic acid can serve as a stable, biodegradable depot.

We next examined the loading and release properties of the hydrogel. First, we assessed in vitro the ability of the hydrogel to release various payloads, including antibodies, cytokines, hydrophobic small molecules, and hydrophilic small molecules (fig. S2). Under sink conditions in phosphate-buffered saline (PBS), the release kinetics ranged from hours to days, depending on the payload. We then confirmed that the hydrogel extends the release of the small molecules and biologics in vivo. Cyanine7 carboxylic acid (Cy7-CA) is a fluorophore that was used as a model small-molecule payload because its physical properties are very similar to those of resiquimod (R848), one of the compounds included in our studies. Cy7-CA was administered to non–tumor-bearing mice, either in solution or loaded in a scaffold placed by the fourth mammary fat pad. The mice were assessed by fluorescence IVIS imaging (Fig. 1D), and the data were quantified (Fig. 1E). Whereas a loss of ~60% of the signal was detected within 2 hours of administration of the fluorophore in solution, this amount of signal decay required 24 hours for the fluorophore loaded in the hydrogel. Over this time course, the signal for fluorophore released from the hydrogel was, on average, roughly threefold higher than fluorophore in solution.

To complement the study of small molecules, we examined the release kinetics of biologics in vivo. Confocal imaging of fluorescently labeled payloads demonstrated that antibodies, cytokines, and small molecules were distributed uniformly across the hydrogel (fig. S3, A to C). Fluorescently labeled versions of anti–PD-1 (programmed death 1), IL-15sa (interleukin-15 superagonist), or 2′3′-cGAMP (cyclic guanosine monophosphate–adenosine monophosphate) were administered to non–tumor-bearing mice, either in solution or loaded in a scaffold placed by the fourth mammary fat pad. The mice were assessed by fluorescence IVIS imaging (fig. S3, D to F), and the data were quantified (fig. S3, G to I). The release rates in vivo exhibited prolonged kinetics relative to those in vitro, as expected for an environment that is more physiologically relevant than sink conditions. These data confirm that the hydrogel scaffold can substantively extend the local release of immunomodulatory compounds relative to local delivery of the same compounds in solution.

To assess whether the function of the biologics was preserved after loading into and release from the cross-linked hydrogel, we compared the activity of unentrapped IL-15sa to IL-15sa that was recovered after entrapment and release. Specifically, we isolated central memory (CD44+CD62L+) CD8+ T cells that express the IL-15 receptor CD122 (25) from the spleens of mice by fluorescence-activated cell sorting and examined their proliferation after exposure to the biologics. The released cytokine phenocopied the effects of the unentrapped molecule, resulting in nearly identical proliferation profiles (fig. S4). Proliferation was blocked by preincubation of CD8+ T cells with anti-CD122, demonstrating the specificity of the biologic’s effect. These data—complemented by the recovery of 100% of the loaded biologics (fig. S2, A and B)—confirm that the cross-linking reaction used to form the hydrogels does not cross-react with or inactivate the biomolecules.

Extended local release of agonists of innate immunity prevents tumor relapse and eliminates metastases

Having confirmed that the release of these immunomodulatory compounds could be extended locally in vivo, we aimed to evaluate the utility of such extended delivery in the therapeutic setting. Female BALB/cJ mice were inoculated orthotopically with 4T1-Luc2 breast cancer cells in their fourth mammary fat pad. Nine days later, the mice were imaged by bioluminescent IVIS imaging, which confirmed that the size of the tumors was consistent across animals and enabled randomization into groups. On day 10 after tumor inoculation, tumors (~100 mm3) were resected, and a hydrogel loaded with an immunomodulatory compound—anti–PD-1, anti–CTLA-4 (cytotoxic T lymphocyte antigen 4), IL-15sa, lenalidomide, celecoxib, 2′3′-c-di-AM(PS)2 (Rp,Rp) (“STING-RR”), or R848—was placed in the tumor resection site. Tumor burden was monitored weekly by IVIS imaging, and it was confirmed that local tumor recurrence was prevented most effectively when an agonist of innate immunity (STING-RR or R848) was administered via the hydrogel (Fig. 2A).

Fig. 2 Extended local release of agonists of innate immunity prevents tumor recurrence and eliminates distal metastases.

Tumors were resected from mice 10 days after orthotopic inoculation of 4T1-Luc2 cells, and hydrogels loaded with the following payloads were evaluated: anti–PD-1, anti-CTLA-4, IL-15sa, lenalidomide, celecoxib, STING-RR, or R848. Mice that did not receive a hydrogel were examined as a negative control. (A) IVIS imaging of 4T1-Luc2 cells is shown for all groups described and illustrates tumor burden. (B) Kaplan-Meier curves are shown comparing hydrogels loaded with antibodies that induce immune checkpoint blockade to no hydrogel. (C) Kaplan-Meier curves are shown comparing a hydrogel loaded with the potent cytokine IL-15sa to no hydrogel. (D) Kaplan-Meier curves are shown comparing hydrogels loaded with various immunomodulatory small molecules to no hydrogel. The number of mice per group (n) and median survival (ms) are listed. The experiment was performed at least three times. Statistics were calculated relative to the group treated with no hydrogel using the log-rank (Mantel-Cox) test. **P ≤ 0.01, ***P ≤ 0.001. d, days; n/a, not achieved.

The STING (stimulator of interferon genes) agonist STING-RR stimulates tumor-resident dendritic cells to produce interferon-β (IFN-β), which is required for spontaneous tumor-initiated T cell priming (26). This molecule has previously been investigated in the context of intratumoral injection into intact primary tumors (26). R848 is a Toll-like receptor 7/8 (TLR7/8) agonist that induces expression of type I IFNs and costimulatory molecules by plasmacytoid dendritic cells as well as phenotypic maturation of conventional dendritic cells (27). It is typically used as a vaccine adjuvant or as a topical treatment for viral or neoplastic skin lesions (28).

Notably, the IVIS imaging revealed that lung metastases had already been established in the mice by the time of surgery, and extended local release of an agonist of innate immunity was again the lone condition that eradicated the existing metastatic lesions (Fig. 2A). The IVIS imaging also provided a meaningful proxy for long-term survival such that only the hydrogel loaded with one of the agonists of innate immunity conferred durable survival benefit to a majority of mice (Fig. 2, B to D).

Agonists of innate immunity are effective only when released locally from the hydrogel

To confirm that extended release from the hydrogel was crucial to the observed efficacy, we compared various additional modes of administration of the agonists of innate immunity. For treatment with R848, in addition to no hydrogel and empty hydrogel controls, the following groups were evaluated: hydrogel (single dose), weekly intravenous injection, weekly intraperitoneal injection, and local delivery in solution (single dose) in conjunction with placement of an empty hydrogel (Fig. 3A). The survival benefit was observed only when R848 was loaded in the hydrogel. To ensure that we were not missing efficacy afforded by delivery of R848 in solution owing to the fact that weekly administration was too long of a window between doses, we performed daily intraperitoneal injections for three consecutive days. Multiple systemic administrations of R848 failed to confer a robust survival benefit (fig. S5A), despite resulting in substantial weight loss (fig. S5B). Similarly, for treatment with STING-RR—which is considerably more potent than its natural analog 2′3′-cGAMP (fig. S6)—delivery via hydrogel was required to induce durable survival benefit among a majority of mice (Fig. 3B and fig. S7).

Fig. 3 Agonists of innate immunity are effective only when released locally from the hydrogel.

(A and B) Tumors were resected from mice 10 days after orthotopic inoculation of 4T1-Luc2 cells. (A) Kaplan-Meier curves are shown for the following groups: no hydrogel, an empty hydrogel, a hydrogel loaded with R848, weekly intravenous (IV) injection of R848, weekly intraperitoneal (IP) injection of R848, or an empty hydrogel plus local administration of R848 in solution. (B) Kaplan-Meier curves are shown for the following groups: no hydrogel, an empty hydrogel, a hydrogel loaded with STING-RR, or an empty hydrogel plus local administration of STING-RR in solution. (C and D) Tumors were not resected but were instead injected intratumorally (IT) with PBS, a single dose of R848, or a single dose of STING-RR 10 days after orthotopic inoculation of 4T1-Luc2 cells. Tumor volume (C) and mouse survival (D) were measured. (E and F) Tumors were resected from mice 10 days after orthotopic inoculation of 4T1-Luc2 cells. (E) Kaplan-Meier curves are shown for the following groups: no hydrogel, a hydrogel loaded with CCL4, a hydrogel loaded with CCL5, or a hydrogel loaded with CXCL10. (F) Kaplan-Meier curves are shown for all groups described: no hydrogel, a hydrogel loaded with paclitaxel, or a hydrogel loaded with doxorubicin. The number of mice per group (n) and median survival (ms) are listed. The experiment was performed at least three times. For (A) and (B), statistics were calculated relative to the group treated with hydrogel containing R848 or STING-RR, respectively, using the log-rank (Mantel-Cox) test. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. For (C) to (F), statistics were calculated relative to the no hydrogel control using the log-rank (Mantel-Cox) test, and no differences between groups were statistically significant.

Perioperative delivery of R848 or STING-RR from a hydrogel (Fig. 3, A and B) was superior to intratumoral injection of either compound, which did not prolong survival (Fig. 3, C and D). This finding is noteworthy because cyclic dinucleotide STING agonists are administered intratumorally in the clinic; perioperative administration may yield superior results and is not limited to superficially accessible lesions. To demonstrate utility for the context of clinical translation, we validated that the efficacy of STING-RR loaded in a hydrogel is retained after 1 week of refrigerated storage at 4°C (fig. S8).

Of note, intraoperative placement of the immunotherapy-loaded hydrogel was required for therapeutic benefit. Placement of hydrogels loaded with either R848 or STING-RR adjacent to nonresected tumors produced no survival benefit (fig. S9), supporting the notion that the intervention is modifying the postsurgical resection microenvironment rather than treating an established tumor. Collectively, these results underscore the importance of having scaffold-mediated extended local release of the immunomodulatory payloads in the tumor resection site relative to administration of the compounds in solution, whether systemically, locally into the resection site as a bolus, or directly into the tumor.

To query the relative importance of immune stimulation versus leukocyte recruitment, we next loaded hydrogels with chemokines central to mediating antitumor immunity: CCL4 (C-C motif chemokine ligand 4), CCL5 (C-C motif chemokine ligand 5), and CXCL10 (C-X-C motif chemokine ligand 10). CCL4 is critical to recruitment of CD103+ dendritic cells (29), whereas CCL5 and CXCL10 recruit T cells and are up-regulated in DNA damage response–deficient breast tumors, which exhibit constitutive activation of the STING pathway (30). None of the chemokines conferred survival benefit (Fig. 3E), nor did combination immune checkpoint blockade of anti–PD-1 and anti–CTLA-4 (fig. S10). Similarly, neither paclitaxel nor doxorubicin loaded in hydrogels was efficacious in producing responses (Fig. 3F), indicating that direct killing of residual cancer cells by cytotoxic agents locally does not protect against lethality caused by distal established metastases. Together, the data suggest that direct agonism of innate immune cells is important to achieving curative outcomes in a majority of treated mice.

Both innate and adaptive arms of the immune system are required for the observed efficacy

Knowing that R848 and STING-RR activate innate immune cells, we sought to discern whether there were particular downstream cells or pathways underlying the enhanced antitumor immune response upon extended local release of these compounds. To this end, we repeated the tumor inoculation and resection procedure described above, but, in addition to placing a hydrogel loaded with R848 or STING-RR in the resection site, we depleted NK cells, CD8+ T cells, or CD4+ T cells (fig. S11) or inhibited type I IFN signaling by neutralizing IFN-α/β receptor 1 (IFNAR1). Survival studies indicated that NK cells, CD8+ T cells, CD4+ T cells, and type I IFN signaling are each required to prevent tumor recurrence and metastasis. Mice in all groups whose innate or adaptive immune system was compromised failed to benefit from extended release of an agonist of innate immunity in the perioperative setting (Fig. 4).

Fig. 4 Both the innate and adaptive arms of the immune system are critical to the observed efficacy.

Tumors were resected from mice 10 days after orthotopic inoculation of 4T1-Luc2 cells, and scaffolds containing (A) R848 or (B) STING-RR were placed in the resection site. Specific immune cell subsets (NK cells, CD8+ T cells, or CD4+ T cells) were depleted, or innate immune signaling (IFNAR1) was inhibited to explore their relative contribution to the observed efficacy. Kaplan-Meier curves are shown for all groups described. The number of mice per group (n) and median survival (ms) are listed. The experiment was performed at least three times. Statistics were calculated relative to the group treated with hydrogel containing the indicated agonist of innate immunity and treated with PBS (no depletion control) using the log-rank (Mantel-Cox) test. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

To dissect the cellular and molecular changes among immune cell subsets after localized release of R848, we assessed the composition, activation status, and function of leukocytes in the spleen. Spleens were recovered from mice 3 and 14 days after surgery for flow cytometric analysis. For the early time point—the so-called priming phase (31)—we focused on the innate arm of the immune system, particularly NK cells and dendritic cells. The number of activated [CD69+, KLRG1+ (killer cell lectin-like receptor G1)] and high effector (CD11b+CD27+) (32) NK cells was increased after treatment with R848 (Fig. 5A). We similarly observed increased numbers of CD8+ and CD103+ dendritic cells, which are crucial to cross-presentation and the production of robust antitumor immunity (29, 33), as well as B220+PDCA1+ (plasmacytoid dendritic cell antigen–1) plasmacytoid dendritic cells, which secrete large amounts of type I IFNs (Fig. 5B) (33). After exposure to R848, more dendritic cells expressed the costimulatory molecules CD40 and CD86 as well as MHC II (major histocompatibility complex class II), indicating that they had been activated. Similar findings were observed for NK cells and dendritic cells isolated from mice treated with hydrogels containing STING-RR (fig. S12).

Fig. 5 Extended local release of R848 increases the number of innate and adaptive antitumor immune cells and cytokines.

Tumors were resected from mice 10 days after orthotopic inoculation of 4T1-Luc2 cells, and scaffolds were placed in the resection site. Spleens were recovered from mice 3 and 14 days after surgery for flow cytometric analysis, and blood was recovered from mice 1.5 hours, 6 hours, 3 days, and 14 days after surgery for cytokine analysis. (A to C) Increased numbers of leukocytes with activated and effector phenotypes are observed. Quantitation of flow cytometry gating of subsets of (A) NK cells (day 3), (B) dendritic cells (day 3), and (C) CD4+ T cells and CD8+ T cells (day 14) is shown. (D) Increased numbers of central memory–like CD8+ T cells are observed. (E) Increased numbers of T cells producing proinflammatory cytokines and cytolytic molecules are observed. Quantitation of flow cytometry gating of CD4+ T cells and CD8+ T cells (day 14) is shown. Splenocytes were cultured for 6 hours in the presence of a specific immunodominant peptide expressed by 4T1 cells (gp70423–431) and brefeldin A before flow cytometry was performed. (F) Increased concentrations of type I IFNs are observed in plasma collected at various time points after surgery. Data were generated by multiplexing laser bead technology. The experiment was performed once with n = 5 biological replicates. Statistics were calculated using a two-sided unpaired t test. Data are means ± SEM. *P ≤0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

The dendritic cells induced a strong adaptive antitumor response, as evidenced by the T cell compartment 14 days after surgery. Increased numbers of CD4+FoxP3 T cells and CD8+ T cells that express markers of activation, including CD69 and glucocorticoid-induced tumor necrosis factor receptor-related protein (GITR), were detected (Fig. 5C). Systemic immunity is required for effective cancer immunotherapy (31), and we confirmed an increase in the number of central memory–like CD8+ T cells that coexpress Ly6C and CD62L (Fig. 5D), which become more prevalent with effective immunotherapy (31).

To assess whether the inhibition of tumor recurrence and lung metastasis upon implantation of R848-loaded hydrogels is associated with systemic expansion of tumor antigen–specific CD8+ T cells, we restimulated splenocytes isolated from the treatment and control groups with SPSYVYHQF, an immunodominant peptide of murine leukemia virus envelope glycoprotein gp70 (amino acids 423 to 431), which is expressed by 4T1 cells (34). The proportion of T cells expressing the proinflammatory cytokines IFN-γ and IL-2 as well as the cytolytic molecule granzyme B was increased among mice treated with hydrogels containing R848 (Fig. 5E). These data confirm that the CD8+ T cells responded to the gp70 peptide restimulation in an antigen-specific manner.

Given the ability of the therapy to eliminate metastases that had already developed in the lungs, we also inspected the composition of leukocytes in lungs recovered 3 and 14 days after surgery. Increased numbers of B cells, T cells, NK cells, and dendritic cells were detected among mice treated with R848-loaded hydrogels (fig. S13). Although these data do not suggest a particular mechanism that is responsible for eradication of metastases, they again support the notion that extended release of R848 promotes a broad systemic antitumor immune response.

To gain insights into the soluble factors that are associated with the activation of innate and adaptive immune cells, we measured the concentrations of cytokines in peripheral blood. At multiple time points after surgery, plasma was collected and subsequently analyzed by multiplexing laser bead technology. Consistent with the loss of efficacy after neutralization of IFNAR1—suggesting a role for type I IFNs—the concentrations of both IFN-α and IFN-β were markedly increased after treatment with a hydrogel containing R848 relative to no hydrogel (Fig. 5F). The concentrations of several other soluble mediators of immunity were also noticeably increased, including CXCL10 (IP-10), CXCL9 (MIG), and IL-15 (fig. S14 and table S1). Together, these results indicate an expansive induction of antitumor immunity. The induction of a memory response was confirmed by rechallenging the surviving mice, 100% of which rejected the 4T1-Luc2 cells that were freshly inoculated into the opposite mammary fat pad (fig. S15).

The hydrogel and intervention are both safe

We confirmed the safety of both the hydrogel and extended local release of immunotherapy therefrom. Three or 14 days after tumor resection and placement of no hydrogel, an empty hydrogel plus local delivery of R848 or STING-RR, or a hydrogel loaded with R848 or STING-RR, whole blood was collected and subjected to a blood panel analysis. The composition of the blood was unaffected by the treatments (fig. S16). The lack of detectable systemic toxicity was confirmed by a lack of changes in the concentrations of liver enzymes (alanine aminotransferase and aspartate aminotransferase) and a surrogate of kidney function (blood urea nitrogen) in the blood across treatment groups (fig. S17). Moreover, body weight was consistent across all groups tested, with the sole modest decrease being observed transiently after surgery, as expected subsequent to such stress (fig. S18).

Efficacy is confirmed in additional models of spontaneous metastasis

Finally, we sought to demonstrate the broad utility of extended localized release of perioperative immunotherapy. In addition to confirming that the parental 4T1 cell line is as responsive as 4T1-Luc2 cells to immunotherapy (fig. S19), as previously reported (22), we inoculated C57BL/6J mice subcutaneously with Lewis lung carcinoma (LLC) cells or B16-BL6 melanoma cells, both of which also spontaneously metastasize to the lung. When tumor volumes reached ~500 or ~650 mm3, respectively, tumors were surgically resected. Again, long-term survival benefit was observed among mice implanted with hydrogels loaded with an agonist of innate immunity (Fig. 6 and fig. S20). Last, we generated hydrogels from alginate to establish that extended release of an agonist of innate immunity can be achieved using additional biomaterials and that such release prevents local recurrence and eliminates distal metastases. As with the hyaluronic acid–derived hydrogel, the amount of cross-linker (in this case, calcium chloride) had to be optimized to provide desirable mechanical integrity. The alginate-based hydrogel exhibited a similar in vitro release profile of R848 to that of the hyaluronic acid–based hydrogel (fig. S21A) and conferred similar survival benefit (fig. S21B).

Fig. 6 Localized release of perioperative immunotherapy is efficacious in an additional model of spontaneous metastasis.

Tumors were resected from mice when tumor volumes reached ~500 mm3 after subcutaneous inoculation of LLC cells. (A and B) Kaplan-Meier curves are shown for no hydrogel or a hydrogel loaded with either (A) R848 or (B) STING-RR. The number of mice per group (n) and median survival (ms) are listed. The experiment was performed at least three times. Statistics were calculated relative to the group treated with hydrogel containing the indicated agonist of innate immunity using the log-rank (Mantel-Cox) test. *P ≤ 0.05.

DISCUSSION

This study demonstrates that extended release of immunotherapy from a hydrogel placed in a tumor resection site can prevent local tumor recurrence and induce systemic antitumor immunity that eradicates existing spontaneous metastases. The study of relapse and metastasis adds value in moving a drug candidate toward clinical development (35). A given drug can exhibit differential efficacy depending on the degree of tumor progression at which the drug is administered (36). Murine models that die owing to metastatic burden enable one to study the effects of a drug in the setting of metastatic cancer, which is the setting that is most frequently observed in the clinic. To mimic patient care accurately, scientists may consider removing the primary tumor, as surgeons do clinically (35).

We primarily evaluated the orthotopic 4T1 tumor model because it robustly recapitulates many features of advanced human breast cancer. It is refractory to most therapeutic agents, including immunotherapy (37). 4T1 breast cancer cells spontaneously metastasize from the primary mammary fat pad to distant sites, particularly the lungs (38). Consistent with what is observed among patients, mice harboring 4T1 tumors generally succumb to metastatic disease, even if the primary tumor has been surgically removed (37, 38). Surgery itself appears to reduce NK cell function, which is important for preventing postoperative metastasis in both mice and patients (39). Our data indicate that orthotopic 4T1 tumors can be successfully treated with monotherapy, provided that the therapy is delivered in an extended manner in the perioperative setting.

Using the related 4T1.2 model, it was recently shown that the efficacy of immunotherapy is greatly enhanced if the therapy is administered in the neoadjuvant (preoperative) setting rather than in the adjuvant (postoperative) setting (40). This was the case at least for immunotherapy directed at T cells, whether regulatory or effector. One of the proposed mechanisms underlying this observation is that the primary tumor is needed to expand tumor-specific T cells after administration of neoadjuvant immunotherapy. Cognate antigen is required to maintain the survival of effector T cells, whose numbers decrease dramatically after surgery (41). In contrast, central memory T cells do not depend on antigen for their maintenance (40), which may partially explain the efficacy that we observe after perioperative immunotherapy. Notably, resection of primary tumors removes not only sources of neoantigens but also effector cells and potentially supportive T helper 1 (TH1) cytokines, which are important for antitumor responses and whose removal may account for the limited efficacy of adjuvant immunotherapy.

Although neoadjuvant immunotherapy is an exciting avenue of investigation, perioperative immunotherapy may be superior. Surgery can reduce primary resistance to immunotherapy by removing cancer cell–intrinsic mechanisms of resistance (42). Moreover, it can remove cancer cell–extrinsic factors that promote primary and adaptive resistance, such as immunosuppressive regulatory cells that are often found in high numbers in the tumor microenvironment. Myeloid-derived suppressor cells promote metastasis and are correlated with reduced survival as well as decreased responsiveness to immunotherapy (42). In addition, the improved survival conferred by neoadjuvant immunotherapy in the 4T1.2 model required combination therapy and may be limited to therapies that activate T cell–mediated antitumor immunity, and neoadjuvant immunotherapy may necessitate inadvisable delays in surgery owing to induction of immune-related adverse events (40). Although it is generally thought that earlier treatment with immunotherapy would be beneficial, side effects after systemic administration have limited such studies to date (43).

Localized therapy is expected to reduce such systemic toxicity. Intraoperative placement of a scaffold also obviates the need to optimize the scheduling of immunotherapy administration, which is required in the neoadjuvant setting and may be particularly challenging for combination therapies. The perioperative setting represents a consequential timeframe for intervention, because surgical stress causes acute immunosuppression that must be overcome to prevent recurrence and dissemination (6). Expression of genes related to wound healing is associated with a lack of response to anti–PD-1 treatment (44). Converting the perioperative period from a pronounced augmenter of metastatic progression to a window of opportunity for halting and/or eradicating residual disease is expected to improve long-term survival rates (45).

The utility of intraoperative placement of a scaffold was previously demonstrated in the context of adoptive cell therapy for treatment of inoperable or incompletely removed tumors, such as 4T1 (22). A porous biopolymer implant was used to enhance the proliferation, migration, and function of tumor-reactive T cells, thereby reducing tumor relapse relative to administration of lymphocytes in solution. More recently, such implantable biopolymer devices were used to deliver chimeric antigen receptor (CAR) T cells directly to the surfaces of solid tumors (46). Although the efficacy of adoptive cell therapy was vastly improved by loading a STING agonist into immunostimulatory microparticles that were incorporated into the scaffolds, the benefit required the STING agonist to be co-delivered in the scaffolds containing tumor-specific CAR T cells. The STING agonist did not confer a profound impact when administered in solution, in the scaffold in the absence of CAR T cells, or in the scaffold in the presence of control CAR T cells. Our data demonstrate that it is possible to confer durable survival benefit among mice harboring established spontaneous metastases using an off-the-shelf scaffold that is not loaded with exogenous T cells, antigen-specific or otherwise, even when gross residual tumor antigen is not intentionally left behind. These findings reveal that relatively high numbers of cancer cells are not required at the time of immunotherapy to enable cross-priming of tumor antigen.

Treating in the perioperative setting, wherein primary tumor antigen and antigen-specific T cells are removed, is distinct from treating intact solid tumors or incompletely resected tumors. Most current tactics in immunotherapy rely on the presence of tumor-infiltrating lymphocytes (TILs), and the dearth of such TILs in many solid tumors represents a major challenge (47). Although extended local release of celecoxib and anti–PD-1 from an alginate hydrogel injected subcutaneously in the region adjacent to 4T1 tumors conferred some survival benefit relative to local administration of free drug, no durable responses were observed, despite the presence of tumor antigen (48). Administration of hydrogels in the perioperative setting may be superior to local injection of hydrogels.

The findings reported herein underscore the importance of having a depot to achieve extended local release of an agonist of innate immunity after surgical resection of a tumor to prevent local tumor recurrence and clear established distal metastases. The resultant impact on the immune system is broad, as evidenced by cellular and molecular analyses of systemic leukocyte populations. Antitumor efficacy required involvement of both innate and adaptive immunity, and type I IFN signaling was indispensable. The data suggest that intraoperative placement of scaffolds containing immunotherapy may be worthy of clinical investigation. Because the hydrogels described in these studies can release both small molecules and biologics, many options—including combinations—can be explored in future work.

This intervention modulates the tumor resection microenvironment and stimulates innate immunity during the period of acute immunosuppression that is associated with wound healing, which is known to promote metastasis. The platform is therefore limited to surgically resectable tumors and is not designed to address intact primary tumor masses in the absence of surgery. The finding that modifying the tumor resection microenvironment is different than addressing intact primary tumors (whole or residual) has bearings on both biological and clinical perspectives. Specifically, agonizing the innate immune system in the perioperative setting—before reestablishment of a concentrated immunosuppressive microenvironment after relapse—is sufficient to confer durable survival benefit and induce antitumor memory.

The results of these studies have particular translational relevance to agonists of STING, a highly promising therapeutic target. Whereas previous studies involved multiple intratumoral injections of the STING agonist (26), our studies involved a single administration to enable us to compare the effects of a single dose of agonist in the presence or absence of concentrated tumor antigen and tumor-infiltrating leukocytes. The data reveal that a single dose of STING agonist is efficacious only in the context of stimulating the post-resection microenvironment and is insufficient to reverse established disease in the 4T1 model. Although multiple intratumoral injections are feasible in the clinical setting, we were not seeking to compare this regimen with perioperative therapy, because the individuals who would receive these therapies come from different patient populations. Rather, our objective was to demonstrate that a single administration of STING agonist (released locally from a hydrogel placed in the tumor resection site) would allow oncologists to intervene even when lesions are not superficially accessible. The requirement for superficial accessibility is a major limitation of intratumoral administration, because the vast majority of cancer patients have tumors that are not readily accessible. In contrast, the hydrogel can be administered during any surgical oncology procedure, which is the main treatment option for most solid tumors.

A second major difference between our data and those reported previously in the literature is the location and context of the tumor. Specifically, previous reports involved intratumoral injection into subcutaneous flank tumors (26), whereas the 4T1 tumors in our study were implanted orthotopically into the mammary fat pad. Orthotopic tumors are much more challenging to cure than subcutaneous tumors, because tissues in different anatomical sites can shape and alter the tumor microenvironment to influence responses to therapy (49).

Agonism of the innate immune compartment of the tumor resection microenvironment is seemingly critical to the observed efficacy. The survival data suggest that activating the adaptive immune system [via immune checkpoint blockade (50) or immunomodulatory imide drug (iMiD) (51)] or inhibiting immunosuppressive myeloid cells [celecoxib (52)] was apparently insufficient to confer efficacy in the perioperative setting as monotherapy. IL-15sa, which is a highly potent complex of IL-15 and IL-15Rα sushi domain that expands NK cells and CD8+ T cells (32), conferred a modest benefit, indicating that driving proliferation of effector cells may be less important than stimulating upstream innate cells that produce type I IFNs. Although doxorubicin has been reported to induce immunogenic cell death that results in cancer cell–autonomous production of type I IFNs (53), surgery removes the vast majority of the cancer cells from the tumor site, which may explain the lack of benefit afforded by doxorubicin in this setting. The observation that this cytotoxic drug is deleterious relative to no treatment may be explained by the fact that doxorubicin induces severe myelosuppression (54). Whereas agonists of innate immunity can boost antitumor responses by stimulating dendritic cells and/or macrophages, doxorubicin has an opposing effect by killing these cells.

Although it is possible that mediators of adaptive immunity may also confer benefit in the context of extended localized release of perioperative immunotherapy, such benefit was not observed at the doses evaluated in the experiments presented herein. To account for the fact that intraoperative administration was performed only once, whereas systemic administration can be performed repeatedly, the experiments were performed with doses that are threefold higher than those routinely administered systemically or even locally for anti–PD-1 and anti–CTLA-4 (22, 37, 40, 48). Even the combination of anti–PD-1 and anti–CTLA-4 administered at 300 μg (rather than 100 μg) each failed to confer benefit. Still, it remains possible that other combinations or further increased doses may prove useful. Local retention of mediators of immune checkpoint blockade after peritumoral injection increases efficacy and decreases systemic toxicity relative to administration of clones of the same antibodies that were not modified to be retained locally but rather rapidly diffuse systemically (55). These data suggest that immune checkpoint blockade in the local environment suffices in the context of intact tumors and may be superior to systemic blockade of PD-1 and CTLA-4. Unfortunately, postoperative combination immunotherapy targeting the adaptive immune system is relatively ineffective (40), again demonstrating that there is an unmet need to address the perioperative setting if preoperative therapy, wherein tumor antigen and antigen-specific T cells are present, is not an option.

The results presented herein suggest that imidazoquinolines such as R848 and cyclic dinucleotides such as STING-RR loaded in biodegradable hydrogels can be applied to the clinic and may have a meaningful impact on patient survival. Agonists of TLR7/8 or STING can serve as adjuvants at the tumor resection site, reprogramming an immunosuppressive environment into an immunostimulatory one and thereby stimulating systemic antitumor immunity. Translation will require selection of a preferred active pharmaceutical ingredient and a preferred polymer, such as hyaluronic acid or alginate. This platform has the potential to deliver immunotherapy in a spatiotemporally defined manner that focuses the therapy at the site of interest during a critical time window.

MATERIALS AND METHODS

Study design

The objective of this study was to determine whether extended local release of particular classes of cancer immunotherapy from a hydrogel placed in the tumor resection site could improve survival relative to local or systemic administration of the same compounds in solution. Mice bearing orthotopic or subcutaneous tumors that spontaneously metastasized were used for these controlled laboratory experiments. It was confirmed in vitro and in vivo that the hydrogel could prolong the release profile of the payloads, whether they were small molecules or biologics. A variety of anticancer modalities were compared, including mediators of immune checkpoint blockade, agonists of innate immunity, chemokines, and chemotherapy. Bioluminescence imaging (BLI) was used to inspect for local tumor recurrence and distal metastatic progression. Tumor volume was measured using electronic calipers when tumors were not surgically removed. Survival was selected prospectively as a primary endpoint. For mechanistic studies, flow cytometry and multiplexing laser bead technology were used to assess cell populations and soluble cytokines, respectively. Safety and tolerability were evaluated by measuring the composition of blood, liver enzyme concentrations, a surrogate of kidney function, and weight. Induction of an antitumor memory immune response was confirmed by rechallenging mice with a fresh inoculate of cancer cells.

The sample sizes were selected on the basis of the results of pilot experiments so that relevant statistical tests could reveal significant differences between experimental groups. No samples or animals that received treatment were excluded from the analysis. Mice were randomized twice before administration of therapy. For the 4T1 model, mice were randomly assigned to new groups of five mice per cage 7 or 8 days after tumor inoculation. After BLI measurements on day 9, mice whose tumors were extremely small or large, as determined by photon number, were excluded from enrollment in the studies. Specifically, mice were excluded if the photon number emitted by their tumors was threefold (or more) greater or less than the average. Mice were randomly assigned to treatment groups, and surgery was performed the following day (10 days after tumor inoculation). For the LLC and B16-BL6 models, tumor volume was measured using electronic calipers, and mice were randomized as tumor volumes approached 500 and 650 mm3, respectively. The investigator was blinded to the group allocation before surgery. The surgeries were performed independently at least three times, and the surgeon was blinded when the experiment allowed.

Cell lines

The metastatic murine 4T1 breast cancer (American Type Culture Collection, CRL2539) and B16-BL6 melanoma [provided by G. Merlino, National Institutes of Health (NIH)] cell lines were cultured in complete RPMI 1640 medium with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and 1% l-glutamine. The 4T1-Luc2 breast cancer cells (Perkin Elmer) used for in vivo BLI were cultured the same way. The murine LLC lung cancer cell line [provided by H. Cantor, Dana-Farber Cancer Institute (DFCI)] was cultured in complete Dulbecco’s modified Eagle’s medium with 10% FBS, 1% penicillin/streptomycin, and 1% sodium pyruvate. Cells were tested for mycoplasma contamination and found to be negative.

Mice

Animal experiments were carried out in accordance with protocols approved by the DFCI Institutional Animal Care and Use Committee. For the metastatic breast cancer model, female BALB/cJ mice (6 to 8 weeks old) were purchased from the Jackson Laboratory (stock #000651). For the lung cancer model, female C57BL/6J mice (6 to 8 weeks old) were purchased from the Jackson Laboratory (stock #000664). Mice were housed in the animal facility of DFCI.

Preparation of hydrogel

HyStem hydrogel kits (ESI Bio, GS1004) were used to prepare the hydrogels. A Teflon mold (9 mm diameter) was first filled with 120 μl of Glycosil (thiol-modified hyaluronic acid), and then 200 μg of R848 (Sigma, SML0196) or 100 μg of c-di-AM(PS)2 (Rp,Rp) (Invivogen, tlrl-nacda2r) was added. For comparative studies, 300 μg of rat anti-mouse PD-1 (anti–PD-1) (BioXCell, clone 29F.1A12), 300 μg of hamster anti-mouse CTLA-4 (anti–CTLA-4) (BioXCell, clone 9H10), 3 μg of mouse IL-15/IL-15R complex recombinant protein carrier-free (IL-15sa) (eBioscience, 34-8152-82), 100 μg of 2′3′-cGAMP (Invivogen, tlrl-nacga23), 200 μg of lenalidomide (Sigma, CDS022536), 1500 μg of celecoxib (Selleckchem, S1261), 10 μg of CCL4 (R&D Systems, 451-MB/CF), 10 μg of CCL5 (R&D Systems, 478-MR/CF), 10 μg of CXCL10 (R&D Systems, 466-CR/CF), 100 μg of paclitaxel (Selleckchem, S1150), or 100 μg of doxorubicin (Selleckchem, S1208) was added. Next, 30 μl of Extralink [thiol-reactive PEGDA (polyethylene glycol diacrylate) cross-linker] was added into the mold, and the hydrogel was allowed to cross-link for at least 1 hour (pH 6.8 to 6.9). For in vitro release studies and confocal imaging, anti–PD-1 and IL-15sa were fluorescently tagged with Alexa Fluor 405 NHS Ester (Thermo Fisher Scientific, A30000) and the VivoTag 680XL Protein Labeling Kit (Perkin Elmer, NEV11118), respectively, according to the manufacturer’s guidelines, and fluorescein-tagged 2′3′-cGAMP (BIOLOG Life Science Institute, C195) was used as a model compound for 2′3′-c-di-PS(2) (Rp,Rp). For in vivo imaging, both anti–PD-1 and IL-15sa were fluorescently labeled with VivoTag 800 (Perkin Elmer, NEV11107), and Cy7-CA (Abcam, ab146499) and sulfo-Cy7–labeled 2′3′-cGAMP (BIOLOG Life Science Institute, custom order) were used. For evaluation of in vivo degradation of the hydrogel, 1.2 μl of Alexa Fluor 750 C5-maleimide (Molecular Probes, A30459) was conjugated directly to the hydrogel. Alginate hydrogels were prepared by filling a Teflon mold with 200 μl of sodium alginate solution (Amsbio, AMS.CSR-ABC-AL) and then adding 200 μg of R848 followed by 15 μl of 1 M calcium chloride (bioWORLD, 40320005). The hydrogel was allowed to set for at least 30 min. Protein conjugation and hydrogel preparation were done under sterile conditions.

Confocal microscopy

To assess the distribution of the payloads within the hydrogel, a hydrogel loaded with fluorescently tagged anti–PD-1, IL-15sa, or 2′3′-cGAMP was imaged under a confocal laser scanning microscope (Leica TCS SP8 STED CW; Leica Microsystems). The obtained images were processed using Leica LAS AF software (Leica Microsystems).

In vitro release study

To determine the release kinetics of each payload from the hydrogel, a hydrogel loaded with fluorescently tagged anti–PD-1, fluorescently tagged IL-15sa, lenalidomide, celecoxib, fluorescently tagged 2′3′-cGAMP, or R848 was immersed in 3 ml of PBS (pH 7.4). At each time point, 1 ml of medium was taken out and the same amount of fresh buffer was added back. The amount of payload that had been released was then measured using a fluorescence plate reader or via high-performance liquid chromatography (HPLC). For anti–PD-1, the fluorescence signal was measured at an excitation of 400 nm and an emission of 430 nm. For IL-15sa, the fluorescence signal was measured at an excitation of 660 nm and an emission of 690 nm. For 2′3′-cGAMP, the fluorescence signal was measured at an excitation of 490 nm and an emission of 520 nm. For the remaining small molecules, the amount of drug was determined via HPLC (Agilent Technologies, 1260 series) using an Inspire column (Dikma Technologies, Inspire 5 μm C18, 250 × 4.6 mm). The mobile phase was prepared by mixing a 20 mM solution of PBS at pH 2.5 with acetonitrile (7:3, v/v). The flow rate and injection volume were 1 ml/min and 20 μl, respectively. The ultraviolet (UV) absorbance was measured at 230 nm.

Functional analysis of released IL-15sa

To assess whether the activity of IL-15sa was preserved after loading into and release from the cross-linked hydrogel, we evaluated the proliferation of CD122+ central memory CD8+ T cells. First, CD8+ T cells were enriched from the spleens of female BALB/cJ mice using the EasySep Mouse CD8+ T Cell Isolation Kit (STEMCELL Technologies, 19853). The enriched T cells were stained with antibodies to enable sorting of central memory CD8+ T cells (CD44+CD62L+CD8+TCR+ cells) using an M Aria II SORP UV (BD Biosciences). The purity of the sorted cells was confirmed to be >90%. To evaluate proliferation, the sorted T cells were labeled with CellTrace Violet (Life Technologies, C34557) and plated in 96-well plates at 105 cells per well in 200 μl of RPMI 1640 medium supplemented with 10% FBS, 1% penicillin/streptomycin, 1% GlutaMAX, 1% nonessential amino acids, 1 mM sodium pyruvate, and 55 μM 2-mercaptoethanol. All supplements were obtained from Life Technologies. Unentrapped IL-15sa or IL-15sa released from a hydrogel (using the conditions described in the preceding section) was added at 100 ng/ml. To block CD122, anti-CD122 (BioLegend, clone TM-β) was also added at 10 μg/ml. The cells were incubated for 4 days, and proliferation was confirmed using a BD LSRFortessa X-20 (BD Biosciences). Dead cells were excluded using 7-aminoactinomycin-D (7-AAD) staining (BioLegend, 420404). Cells were cultured in at least triplicate wells, and the experiments were conducted independently in triplicate.

In vivo evaluation of hydrogel degradation

In vivo degradation of fluorophore-labeled hydrogels was monitored using the IVIS Spectrum In Vivo Imaging System (Perkin Elmer) after the hydrogels were surgically implanted into mice. Fluorescence imaging was obtained weekly and analyzed with Living Imaging software (Perkin Elmer). Both tumor-resected mice and non–tumor-bearing mice were tested for in vivo hydrogel degradation, although only one tumor-resected mouse survived for more than a few weeks in the absence of therapy, because tumors recurred in nearly all animals receiving surgery alone.

In vivo release study

To evaluate the in vivo release profiles of Cy7-CA (a model compound for R848), anti–PD-1, IL-15sa, and 2′3′-cGAMP, hydrogels containing the fluorophore or one of the fluorescently labeled payloads were surgically implanted into non–tumor-bearing mice. For controls, solutions containing the fluorescent molecules were administered into the incision sites. Fluorescence was monitored using the IVIS Spectrum In Vivo Imaging System (Perkin Elmer).

In vivo tumor models and treatment

For the metastatic breast cancer model, 105 4T1-Luc2 or 4T1 cells [in 30 μl of Dulbecco’s PBS (DPBS)] were inoculated orthotopically into the fourth mammary pad of mice to generate a local tumor mass. Cells were injected without any incision to expose the fat pad. Mice were randomly assigned to treatment groups, and surgery was performed 10 days after tumor inoculation. For the metastatic lung cancer and melanoma models, 5 × 105 LLC cells or 106 B16-BL6 cells (in 100 μl of DPBS), respectively, were inoculated subcutaneously into mice to generate a local tumor mass. Mice were randomly assigned to treatment groups, and surgery was performed when tumor volumes reached ~500 or ~650 mm3, respectively. While mice were kept under anesthesia at 2% isoflurane, the tumor was resected, and the hydrogel was placed in the site of the resultant cavity at the time of surgery. The wound was closed with medical clips. In control experiments, therapy was administered in solution at the site of surgery, intraperitoneally or intravenously. For tumor rechallenge experiments, 104 4T1-Luc2 cells were inoculated in the contralateral mammary fat pad. The surgeries were performed independently at least three times, and the surgeon was blinded when the experiment allowed.

In vivo bioluminescence and imaging

After surgery, mice were inspected by BLI weekly for local tumor recurrence and distal metastasis. To this end, 10 min after intraperitoneal injection of d-luciferin (150 mg/kg), a substrate of Luc2, the mice were anesthetized with 2% isoflurane and imaged using the IVIS Spectrum In Vivo Imaging System (Perkin Elmer).

Depletion of NK cells, CD8+ T cells, or CD4+ T cells and neutralization of IFNAR1

To evaluate which immune cells are required to confer the observed antitumor effect, specific cell subsets (NK cells, CD8+ T cells, or CD4+ T cells) were depleted by administering a depleting antibody intraperitoneally every 3 days, beginning 1 day before therapy. The antibodies used for depletion were anti-asialo GM1 (polyclonal, Wako Chemical, 30 μl), anti-mouse CD8a (clone 2.43), and anti-mouse CD4 (clone GK1), respectively. To test the role of type I IFN signaling, a neutralizing antibody against IFNAR1 (clone MAR1-5A3) was administered to the mice. All antibodies were purchased from BioXCell, and 200 μg of antibody was used, unless otherwise specified. Depletion of NK cells, CD8+ T cells, or CD4+ T cells was confirmed by flow cytometry of leukocytes isolated from the blood of mice to which antibodies or PBS had been administered.

In vivo cytokine analysis

Blood was collected from mice 1.5 hours, 6 hours, 3 days, and 14 days after resection of tumor and treatment with an R848-loaded hydrogel or no hydrogel. Plasma was sent to Eve Technologies to measure the concentrations of circulating cytokines that were produced in response to the therapy. The MD-31 panel (days 3 and 14) was complemented by assessment of IFN-α and IFN-β (hours 1.5 and 6).

Flow cytometry

Flow cytometry was performed on a BD LSRFortessa X-20 (BD Biosciences), and all antibodies were purchased from BioLegend, eBioscience, or BD Biosciences (listed in table S2). For restimulation studies, SPSYVYHQF (purity >95%), an immunodominant peptide of murine leukemia virus envelope glycoprotein gp70 (amino acids 423 to 431), was used (New England Peptide, BP10-470). GolgiPlug (BD Biosciences, 555029) was used for inspection of intracellular cytokines and cytolytic molecules.

Statistical analysis

Statistical methods were not used to predetermine sample size. The sample sizes were selected on the basis of the results of pilot experiments so that relevant statistical tests could reveal significant differences between experimental groups. Statistical analysis was performed using GraphPad Prism software version 7.01. Data are presented as means ± SD or means ± SEM, as indicated in the figure legends. To determine statistical significance when comparing two groups, the two-sided unpaired t test was used. For survival analysis, the log-rank (Mantel-Cox) test was used.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/433/eaar1916/DC1

Fig. S1. A hydrogel composed of cross-linked hyaluronic acid is stable but biodegradable in vivo.

Fig. S2. The hydrogel scaffold extends the release of biologics and small molecules in vitro.

Fig. S3. The hydrogel scaffold extends the local release of biologics and small molecules in vivo.

Fig. S4. The activity of a biologic released from the hydrogel is preserved completely.

Fig. S5. Multiple systemic administrations of R848 fail to confer survival benefit and are less well tolerated than R848 released from a hydrogel.

Fig. S6. STING-RR confers superior efficacy to 2′3′-cGAMP upon extended release from a hydrogel in the perioperative setting.

Fig. S7. IVIS imaging confirms that STING-RR must be released from a hydrogel to protect against tumor recurrence and eliminate metastases.

Fig. S8. Efficacy of STING-RR loaded in a hydrogel is retained after storage for 1 week at 4°C.

Fig. S9. Intraoperative placement of the immunotherapy-loaded hydrogel into the tumor resection site is required for therapeutic benefit.

Fig. S10. Extended local release of combination immune checkpoint blockade confers limited survival benefit.

Fig. S11. Flow cytometric analysis confirms that NK cells, CD8+ T cells, and CD4+ T cells are depleted after administration of appropriate antibodies.

Fig. S12. Extended local release of STING-RR increases the number of activated innate immune cells.

Fig. S13. Extended local release of R848 increases the numbers of several leukocyte subsets in the lung.

Fig. S14. Increased concentrations of cytokines are observed in plasma collected at various time points after surgery.

Fig. S15. Induction of an adaptive antitumor memory response is confirmed by rejection of 4T1-Luc2 cells inoculated as rechallenge.

Fig. S16. Extended local release of agonists of innate immunity does not alter the composition of the blood.

Fig. S17. Extended local release of agonists of innate immunity is safe.

Fig. S18. Extended local release of agonists of innate immunity is well tolerated.

Fig. S19. The response of parental 4T1 cells to R848 released locally from a hydrogel is similar to that of 4T1 cells expressing Luc2.

Fig. S20. Localized release of perioperative immunotherapy is efficacious in the B16-BL6 model of spontaneous metastasis.

Fig. S21. Extended release of R848 from a scaffold derived from alginate confers survival benefit.

Table S1. Quantitation and statistics for a panel of cytokines measured in plasma recovered 3 and 14 days after surgery.

Table S2. A table of the antibodies used for flow cytometry experiments.

REFERENCES AND NOTES

Acknowledgments: We thank M. Stephan (Fred Hutchinson Cancer Research Center), S. Dougan (DFCI), and J. Guerriero (DFCI) for providing valuable suggestions regarding the tumor resection surgery. We thank C. Sypher (DFCI), M. Louis (DFCI), and the technical staff from the DFCI Animal Research Facility for their helpful suggestions relating to and assistance with bioluminescence and fluorescence imaging. We thank I.-K. Choi (DFCI) and S. Hwang (DFCI) for advice relating to cell sorting and in vitro assays. We thank H. Cantor (DFCI) and G. Merlino (NIH) for sharing the LLC and B16-BL6 cancer cell lines, respectively. We also thank R. Bronson and the Dana-Farber/Harvard Cancer Center Rodent Histopathology Core for histopathological analyses. Funding: This work was supported by the Carney Family Charitable Foundation (to C.G.P.) and the National Cancer Institute’s SPORE in Breast Cancer at DF/HCC P50CA168504 Career Development Award (to M.S.G.) as well as by the Claudia Adams Barr Program for Innovative Cancer Research and STIMIT Corporation. Author contributions: C.G.P. and M.S.G. conceived the study. C.G.P., H.-J.K., and M.S.G. designed the experiments. C.G.P. performed the experiments. C.A.H., D.S., and E.M.C. helped perform the in vivo experiments. C.G.P., H.-J.K., and M.S.G. analyzed and interpreted the data. M.S.G. supervised the overall research. C.G.P. and M.S.G. wrote the manuscript. Competing interests: C.G.P. and M.S.G. are inventors on patent application PCT/US2017/049424 submitted by DFCI that covers the drug delivery platform presented in this report. All other authors declare that they have no competing interests. Data and materials availability: All data are included in this paper or the Supplementary Materials. Materials can be obtained upon request to M.S.G.
View Abstract

Navigate This Article