Alum-anchored intratumoral retention improves the tolerability and antitumor efficacy of type I interferon therapies

We developed five type I IFNs with different signaling strengths and pharmacokinetics (Fig. 1A and SI Appendix, Fig. S1 A and B and Table S1). We recombinantly expressed murine IFNα subtype A (referred to here as IFNα), and murine IFNβ, which is an order of magnitude more potent than IFNα (35). Since PEGylated IFNα is used in the clinic (7, 8), we also generated an extended half-life format of IFNα by fusion to mouse serum albumin (MSA). Albumin extends cytokine half-life by FcRn-mediated recycling and reducing renal clearance, and albumin-IFNα fusions have been previously validated (9, 36, 37).

Next, we employed a system that we recently developed to enable long-term retention of intratumorally injected cytokines (22). This strategy relies on aluminum hydroxide (alum) particles that are retained at their injection site as a physical depot for weeks. By fusion to an alum-binding peptide (ABP), ABP cytokines that are mixed and coinjected with alum are also retained long term. The design of ABP contains serine motifs that are phosphorylated upon coexpression with the Fam20C kinase (22) because phosphates bind tightly to alum through a ligand exchange reaction with surface hydroxyls (38, 39). Indeed, ABP-IFNs had higher phosphorylation (SI Appendix, Fig. S1C) and tighter alum binding (Fig. 1B) compared to IFNs without ABP. To test translatability, we confirmed that human IFNα2b could be expressed (SI Appendix, Fig. S1D) and phosphorylated (SI Appendix, Fig. S1E) as an ABP fusion. All five murine IFNs activated RAW-Lucia Interferon-stimulated genes (ISG) cells, an IFN-responsive macrophage line, albeit with a mild increase in half maximal effective concentration (EC₅₀) for fusion proteins (Fig. 1C). IFNβ constructs were an order of magnitude more potent than IFNα constructs, and ABP-IFNs maintained their activity in the presence of alum (SI Appendix, Fig. S1F).

To evaluate pharmacokinetics, we injected AF647-labeled IFNs intratumorally (i.t.) into apigmented B16F10 tumors and tracked their retention at the tumor for 5 d (Fig. 1 D and E). Upon coinjection with alum, ABP-IFNα and ABP-IFNβ were retained in the tumor for all 5 d. In contrast, IFNα, IFNβ, and MSA-IFNα leaked out of the tumor within 1 d. Five days after treatment, AF647-IFN levels in homogenized organs were measured by fluorescence spectroscopy. ABP-IFNs were highly specific for the tumor, with no signal detected in the serum, spleen, liver, lung, or kidney (Fig. 1F). Serum fluorescence levels confirmed that MSA-IFNα had extended circulation, while the other IFNs were not detected in the serum at the time points tested (Fig. 1G).

The five IFNs were tested for therapeutic efficacy as a single agent. Mice bearing established MC38 colon carcinoma tumors or B16F10 melanoma tumors were treated with 0.5 nmol IFN i.t. on days 5, 11, and 17. Survival (Fig. 2 A and B) and tumor growth (Fig. 2C and SI Appendix, Fig. S2 A and B) demonstrated that IFNα had poor efficacy in both tumor models and only extended median survival by 4 to 5 d compared to the Tris-buffered saline (TBS) control. IFNβ was not significantly different from IFNα (P = 0.18 for MC38, P = 0.81 for B16F10), indicating that increasing signaling strength alone was insufficient to improve efficacy.

Intratumoral retention improved survival from type I IFN therapies. Alum + ABP-IFNα significantly extended survival over IFNα (P = 0.0005 for MC38, P = 0.01 for B16F10). Similarly, alum + ABP-IFNβ was more efficacious than IFNβ (P = 0.0005 for MC38, P < 0.0001 for B16F10). As a control, even large doses of alum alone did not slow tumor growth, indicating that the efficacy was IFN mediated (SI Appendix, Fig. S2C) (22).

In immunologically cold B16F10 tumors, the more potent alum + ABP-IFNβ was more efficacious than alum + ABP-IFNα (P = 0.004). In contrast, in the MC38 model, the less potent alum + ABP-IFNα was more efficacious than alum + ABP-IFNβ (P = 0.002). MC38 tumors have higher basal type I IFN signaling and immune infiltration compared to B16F10 tumors (40–42). There may be a threshold for productive type I IFN signaling, and overactivation may be detrimental in tumors with preexisting high IFN signatures. Consistent with this hypothesis, CD8⁺ T cells in MC38 tumors treated with alum + ABP-IFNβ had high levels of exhaustion marker TIM3 and suppressive marker PD-L1 (SI Appendix, Fig. S3A) (43, 44). Optimal type I IFN signaling is complex and tumor dependent.

MSA-IFNα had similar efficacy to alum + ABP-IFNα in both models (P = 0.48 for MC38, P = 0.43 for B16F10). However, mice treated with MSA-IFNα experienced mild body weight loss on day 7 that was not observed from the alum-anchored IFNs, raising potential toxicity concerns for prolonging half-life without intratumoral retention (Fig. 2D and SI Appendix, Fig. S2B). Intraperitoneal (i.p.) treatments of the same MSA-IFNα dose had poor efficacy in B16F10-bearing mice (SI Appendix, Fig. S3B), indicating that intratumoral exposure is key to type I IFN efficacy.

Next, we asked how intratumoral retention improved type I IFN therapies. We focused on the poorly inflamed B16F10 model because type I IFNs are particularly promising for turning cold tumors hot (5). We also focused on alum + ABP-IFNβ because this had the strongest efficacy in B16F10 tumors (Fig. 2 B and C). Type I IFNs are pleiotropic and can have a wide range of impacts on tumor cells, endothelial cells, and immune cells (1, 2). We examined each of these compartments to uncover the therapeutic mechanism behind alum + ABP-IFNβ in B16F10 tumors.

To study impacts on tumor cells, B16F10 tumors were treated with TBS, IFNβ, or alum + ABP-IFNβ, followed 4 d later by flow cytometry of live tumor cells (Fig. 3A and SI Appendix, Fig. S4A). In addition to wild-type (WT) mice, we included Ifnar1^−/− mice whose host cells are deficient in type I IFN sensing, so only the implanted B16F10 cells can directly respond to IFN treatment.

Type I IFNs can alter how tumor cells interact with immune cells via transcriptional regulation of thousands of genes (2). Importantly, type I IFNs up-regulate major histocompatibility complex I (MHC-I) on tumor cells, improving immune recognition (45). Alum + ABP-IFNβ led to a greater proportion of B16F10 cells with high MHC-I expression in both WT mice (P = 0.04 compared to TBS) and Ifnar1^−/− mice (P = 0.009 compared to TBS) (Fig. 3B and SI Appendix, Fig. S4B). The altered MHC-I phenotype was substantial even in Ifnar1^−/− mice (Fig. 3 B and C), suggesting that alum-anchored payloads were unexpectedly uniformly accessible throughout the tumor. TBS-treated WT mice had higher MHC-I levels than TBS-treated Ifnar1^−/− mice (P = 0.02, Fig. 3B), highlighting that endogenous type I IFN signaling also impacts B16F10 cells considerably in this therapy. Type I IFNs can also up-regulate PD-L1 on tumor cells, which suppresses T cell function by binding to the PD-1 receptor expressed on activated T cells (12). Alum + ABP-IFNβ increased the percentage of B16F10 cells with high PD-L1 expression (P = 0.03 compared to TBS) (Fig. 3D and SI Appendix, Fig. S4C). Although not significant, PD-L1 levels also trended upwards in treated Ifnar1^−/− mice.

Type I IFNs can be antiproliferative and proapoptotic to tumor cells (2). No antiproliferative activity was observed in tumor cells as measured by Ki67 (SI Appendix, Fig. S4D). However, alum + ABP-IFNβ caused live tumor cells in WT mice to be more apoptotic, as measured by Apotracker Green, which detects phosphatidylserine residues on the cell surface (P = 0.0001 compared to TBS, Fig. 3E). No changes in Apotracker Green levels occurred in Ifnar1^−/− mice (P = 0.77), indicating that the proapoptotic effect observed in WT mice was mediated indirectly.

This study showed that alum + ABP-IFNβ altered tumor cells, both directly (MHC-I and PD-L1 levels) and indirectly (apoptosis). The direct effects on tumor cells were impactful and led to an increase in total CD45⁺ immune cells in the tumor in Ifnar1^−/− mice (SI Appendix, Fig. S4 E and F). However, the direct effects on tumor cells alone were insufficient to slow tumor growth. The strong efficacy of alum + ABP-IFNβ in WT mice (Fig. 2B) was completely lost when the treatment regimen was repeated in Ifnar1^−/− mice (Fig. 3F), indicating that interactions with IFN-responsive host cells are essential for efficacy.

Endothelial cells are an important consideration because type I IFNs are antiangiogenic (1, 24). In order to examine the role of nonhematopoietic cells in treatment efficacy, bone marrow chimeras were generated. WT and Ifnar1^−/− hosts were lethally irradiated, then reconstituted with bone marrow (BM) from congenically marked WT donors. Thus, in the Ifnar1^−/− hosts, the nonimmune compartment is deficient in type I IFN sensing. After 6 wk, BM engraftment was confirmed (SI Appendix, Fig. S5A). Mice were inoculated with B16F10 tumors, followed by TBS or alum + ABP-IFNβ treatments. As expected, survival (Fig. 3G) and tumor growth (SI Appendix, Fig. S5B) showed strong therapeutic efficacy in WT hosts (P = 0.001). The treatment still promoted tumor growth delay and extended survival in Ifnar1^−/− hosts (P = 0.001). However, comparing the two alum + ABP-IFNβ groups revealed significant reduction in efficacy by removing the nonhematopoietic compartment’s ability to respond to type I IFNs (P = 0.0004).

To begin assessing the immune compartment, B16F10 tumors were treated with TBS, IFNβ, or alum + ABP-IFNβ, and tumor lysates were analyzed 4 d later for chemokines and cytokines. All concentrations are reported (SI Appendix, Fig. S6), and a heatmap visualizes fold change compared to TBS (Fig. 4A). Compared to TBS, alum + ABP-IFNβ significantly increased CXCL10 (14-fold, P < 0.0001), IFNγ (7.4-fold, P = 0.0009), RANTES (5.8-fold, P < 0.0001), MCP-1 (3.2-fold, P = 0.0002), IL-12 (2.6-fold, P = 0.0003), IL-10 (1.7-fold, P = 0.0009), TNFα (1.7-fold, P = 0.047), IL-1β (1.5-fold, P = 0.01), and GM-CSF (1.4-fold, P = 0.003). Compared to unanchored IFNβ, alum + ABP-IFNβ had significantly higher CXCL10 (P < 0.0001), RANTES (P = 0.001), and MCP-1 (P = 0.001), highlighting that intratumoral retention induced a stronger chemokine signature.

To understand which immune cells play a role, the alum + ABP-IFNβ treatment regimen was repeated using mice lacking specific immune compartments via genetic knockout or confirmed antibody-mediated depletions (SI Appendix, Fig. S7A). Depleting natural killer (NK) cells and neutrophils did not harm efficacy (SI Appendix, Fig. S7B). The Nlrp3 inflammasome, which can be activated by alum, was also dispensable (SI Appendix, Fig. S7C) (46). However, treatment efficacy was significantly reduced in mice lacking Batf3 dendritic cells (DCs) (P = 0.0002) or CD8⁺ T cells (P = 0.0003) (Fig. 4B). The type I IFN pathway plays a pivotal role in DC antigen presentation and CD8⁺ T cell priming and expansion (3–5). Indeed, 4 d after treatment with alum + ABP-IFNβ, there was a greater proportion (Fig. 4C) and total count (SI Appendix, Fig. S7D) of Ly6C⁻ MHCII⁺ CD24⁺ DCs in the tumor draining lymph node (tdLN) that expressed activation marker CD86. In addition, 6 d after treatment with alum + ABP-IFNβ, there were more CD8⁺ T cells per milligram of tumor (SI Appendix, Fig. S7E) and a greater proportion of CD8⁺ T cells in the tumor were activated compared to TBS, as measured by Ki67 (P = 0.003), TIM3 (P = 0.002), and TCF1 (P = 0.01) (Fig. 4D and SI Appendix, Fig. S7E).

These experiments revealed a pleiotropic mechanism of action in B16F10 tumors. Alum + ABP-IFNβ induced important changes in tumor cells, and therapeutic efficacy relied on multiple contributions from nonhematopoietic cells, DCs, and CD8⁺ T cells. After investigating how alum + ABP-IFNβ interacted with the tumor microenvironment as a single agent, we moved on to study whether combining IFNs with other immunotherapies could enable tumor cures.

First, we combined the panel of IFNs with an extended half-life form of IL-2 (MSA-IL2) (47). IFNα and IL-2 are currently the only FDA-approved cytokines for cancer. Although IFNα and IL-2 play complementary roles in the cancer immunity cycle (48–50), their combination has been generally unsuccessful in the clinic to date (51–53). B16F10 tumors were treated on days 5, 11, and 17 with IFNs (i.t.), previously shown not to be curative as a single agent (Fig. 2B). Thirty micrograms of MSA-IL2 (i.p.) was dosed concurrently. Survival (Fig. 5A) and tumor growth (SI Appendix, Fig. S8A) showed that MSA-IL2 alone delayed median survival by only 2 d. Almost complete cure rates were achieved when MSA-IL2 was combined with alum + ABP-IFNα (6/6 cures) or alum + ABP-IFNβ (6/7 cures). In stark contrast, no mice were cured when MSA-IL2 was combined with IFNα, MSA-IFNα, or IFNβ. However, we and others have observed that cytokine combinations can exacerbate treatment-related toxicities (21, 54, 55). Extended half-life MSA-IFNα combined with MSA-IL2 led to substantial weight loss, and four out of seven mice were killed due to toxicity (Fig. 5B). In contrast, alum-localized IFNs combined with MSA-IL2 avoided this overt toxicity.

Second, we combined IFNs with αPD1, in part motivated by the PD-L1 up-regulation on tumor cells following localized IFN treatment (Fig. 3D and SI Appendix, Fig. S4C). There are also reports of improved antitumor efficacy in mice from combining type I IFN pathway agonists with αPD1 or αPD-L1 (11, 12, 56). Mice bearing B16F10 tumors were treated on days 5, 11, and 17 with IFNs (i.t.) and 200 μg αPD1 (i.p.). Survival (Fig. 5C) and tumor growth (SI Appendix, Fig. S8B) demonstrated that αPD1 alone could not cure B16F10 tumors. Mice were cured when αPD1 was combined with IFNα (5/6 cures), alum + ABP-IFNα (6/6 cures), IFNβ (4/6 cures), or alum + ABP-IFNβ (4/5 cures). Mice did not lose weight (SI Appendix, Fig. S8B).

After 100 to 110 d, cured mice were rechallenged with 0.1 M B16F10 cells in the opposite flank and monitored for an additional 100 d. We observed poor resistance to rechallenge in mice cured from MSA-IL2 combined with alum + ABP-IFNα (1/6 survivors) or alum + ABP-IFNβ (1/6 survivors) (SI Appendix, Fig. S8C). Mice cured from αPD1 + alum + ABP-IFNβ also had poor resistance to rechallenge (1/4 survivors) (SI Appendix, Fig. S8D). In contrast, we observed strong resistance to rechallenge in mice cured from αPD1 combined with IFNα (4/5 survivors), IFNβ (4/4 survivors), or alum + ABP-IFNα (5/6 survivors) (SI Appendix, Fig. S8D). We were specifically intrigued by the difference between alum + ABP-IFNα in combination with MSA-IL2 compared to αPD1. Both therapies achieved 100% cures at the primary tumor (Fig. 5 A and C) but led to starkly different resistance to rechallenge (P = 0.02) (Fig. 5D).

To investigate this mechanism, B16F10 tumors were treated on days 5 and 11 with alum + ABP-IFNα, either alone, with MSA-IL2, or with αPD1 (Fig. 5E). On day 13, spleens were excised, and T cells were analyzed by flow cytometry (SI Appendix, Fig. S9A). Markers KLRG1 and CD127 (a subunit of IL-7R) predict if CD8⁺ T cells are short-lived effector cells (SLECs) or memory precursor effector cells (MPECs) (57–60). Alum + ABP-IFNα alone did not appreciably raise the percentage of CD3⁺ CD8⁺ CD44⁺ effector T cells that were SLECs (%SLECs), which may explain why alum + ABP-IFNα failed to achieve cures as a single agent (Fig. 5F and SI Appendix, Fig. S9 B and C). In contrast, the MSA-IL2 combination substantially increased %SLECs (P < 0.0001 compared to untreated), helping explain the 6/6 cure rate at the primary tumor. However, the increase in %SLECs came at the cost of a drastic decrease in %MPECs (P < 0.0001 compared to untreated), consistent with the poor rechallenge results from the MSA-IL2 combination. On the other hand, the αPD1 combination increased %SLECs compared to untreated (P = 0.03), but to a lesser extent than the MSA-IL2 combination (P = 0.02). Conversely, the αPD1 combination had mild reduction in %MPECs compared to untreated (P = 0.03), but still maintained high %MPECs compared to the MSA-IL2 combination (P < 0.0001).

The MSA-IL2 combination had several additional key differences compared to the αPD1 combination. The MSA-IL2 combination-treated mice had enlarged spleens (P < 0.0001) (Fig. 5G), although there was no significant increase in total CD8⁺ T cells in the spleen (SI Appendix, Fig. S10A) (61). The MSA-IL2 combination also increased the proportion of CD8⁺ T cells that expressed CD25 (P < 0.0001), proliferation marker Ki67 (P = 0.01), and exhaustion marker TIM3 (P = 0.0001) (Fig. 5H and SI Appendix, Fig. S10B). Of note, excessive CD25 expression and prolonged IL-2 signaling have been reported to make T cells prone to terminal differentiation (62). All treatments increased TCF1 levels compared to untreated (SI Appendix, Fig. S10C). Overall, the MSA-IL2 combination with poor rechallenge data had spleens with a dramatic increase in SLECs, reduction in MPECs, and overactivation of T cells.

C57BL/6 mice (Taconic C57BL/6NTac and JAX C57BL/6J) were purchased. Ifnar1^−/− (JAX 032045), Batf3^−/− (JAX 013755), and Nlrp3^−/− mice (JAX 021302) were purchased and bred in-house. Six- to 10-wk old females were used in experiments. All animal work was conducted under the approval of the Massachusetts Institute of Technology Committee on Animal Care in accordance with federal, state, and local guidelines.

B16F10 (American Type Culture Collection [ATCC]), MC38 (a gift from J. Schlom, National Cancer Institute, Bethesda, MD), B16F10 Trp2KO (generated previously) (78), and RAW-Lucia ISG (InvivoGen) cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (ATCC) supplemented with 10% fetal bovine serum. RAW-Lucia ISG cells were maintained with 200 μg/mL Zeocin (InvivoGen) every other passage. Adherent cells were cultured at 37 °C and 5% CO₂. HEK293-F cells (Gibco) were cultured in FreeStyle293 Expression Medium (Gibco) shaking at 37 °C and 8% CO₂. All cell lines tested negative for mycoplasma.

Murine IFNα, IFNβ, MSA-IFNα, ABP-IFNα, ABP-IFNβ, and human IFNα, ABP-IFNα, were cloned with 6-His tags into the gWiz vector (Genlantis) using In-Fusion cloning (Takara Bio) (SI Appendix, Table S1). Fam20C kinase with a KDEL tag (22) and MSA-IL2 with a 6-His tag (47) were previously cloned into the gWiz vector. αPD1 was cloned into the gWiz vector starting with the sequence of 29F.1A12, which was a generous gift from the G. Freeman Laboratory (Dana-Farber Cancer Institute, Boston, MA) (79,80). This clone is typically used in the rat IgG2a format, but we cloned it with a murine IgG2c isotype with a kappa light chain and the LALA-PG mutations to ablate Fc effector functions and prevent target cell clearance (81). Plasmids were amplified in Stellar cells and purified using NucleoBond Xtra endotoxin-free kits (Macherey-Nagel). For transfections, 1 mg plasmid DNA with 2 mL polyethylenimine (Polysciences 23966) in 40 mL OptiPRO Serum Free Medium (Thermo Fisher) was added dropwise to 1 L HEK293-F cells. For ABP-IFNs, the 1 mg plasmid DNA was a mixture of 950 μg ABP-IFN plasmid and 50 μg Fam20C plasmid. After 6 d, proteins were purified from cell supernatants using TALON Metal Affinity Resin (Takara Bio) for IFNs and MSA-IL2, and rProtein A Sepharose Fast Flow Resin (Cytiva) for αPD1. ABP-IFNs and MSA-IL2 were further purified by size exclusion chromatography on an ÄKTA fast protein liquid chromatography system (GE Healthcare). Typical yields per 1 L HEK293-F cells for murine IFNs were ∼10 mg IFNα, ∼5 mg IFNβ, ∼20 mg MSA-IFNα, ∼1 mg ABP-IFNα, and ∼1 mg ABP-IFNβ. Proteins were buffer exchanged into TBS (IFNs) or phosphate-buffered saline (PBS) (MSA-IL2 and αPD1) using Amicon filters (EMD Millipore). Proteins were sterile filtered and confirmed for minimal endotoxin (<0.1 EU per dose) using the LAL Chromogenic Endotoxin Quantitation Kit (Pierce). Molecular weight was confirmed by running proteins alongside a Novex Sharp Pre-Stained Protein Standard on a NuPAGE 4 to 12% Bis-Tris gel (Invitrogen) with 2-(N-morpholino) ethanesulfonic acid running buffer and SimplyBlue Safe Stain (Life Technologies). Phosphorylation of ABP-IFNs was confirmed by malachite green assay (Pierce Phosphoprotein Phosphate Estimation Assay Kit). Proteins were flash frozen in liquid nitrogen and stored at −80 °C.

All alum used in this paper was Alhydrogel (InvivoGen vac-alu-250). To fluorescently label proteins, IFNs were incubated with AF647 N-hydroxysuccinimide ester (Invitrogen A20006) in PBS with 0.1 M K₂HPO₄, pH 9 for 2 h at room temperature. Free dye was removed using PD SpinTrap G-25 columns (Cytiva). Fluorescently labeled IFNs (1 μg) and alum (10 μg) were incubated for 20 min in TBS, pelleted, then resuspended in PBS 10% mouse serum for 1 h. Alum was pelleted by centrifugation at 15,000 × g to separate the alum-bound fraction from the released fraction, and fluorescence was measured by a Tecan Infinite M200 Pro.

RAW-Lucia ISG cells were seeded onto 96-well tissue culture plates at 100,000 cells per well in 200 μL media and varying concentrations of IFNs. For some wells, alum and ABP-IFNs were first mixed at a 5:1 (wt/wt) ratio for 20 min. After 18 h, luciferase levels in cell supernatants were assayed using QUANTI-Luc (InvivoGen) and a Tecan Infinite M200 Pro with an injector, following vendor instructions.

IFNs were fluorescently labeled as described above. AF647-labeled and unlabeled IFNs were mixed such that each dose contained 0.5 nmol IFN and 0.1 nmol dye in 20 μL TBS. Low dye levels can minimize quenching artifacts. ABP-IFN groups included 60 μg alum in the dose. On day 0, C57BL/6 mice given alfalfa-free feed were inoculated with 1 M B16F10 Trp2KO cells subcutaneously (s.c.) in the right flank. On day 5, tumors were treated (i.t.). Fluorescence at the tumor was imaged by in vivo imaging system (IVIS) (Perkin-Elmer) using a 640-nm excitation filter and 680-nm emission filter. Image analysis was done by Living Image software. Data were normalized to the maximum signal throughout the experiment for each IFN. On days 7 and 10, serum was collected. On day 10, mice were killed and organs were homogenized in tubes containing zirconium beads (Benchmark Scientific) using a minibeadbeater (BioSpec Products) in PBS with 1 mg/mL collagenase/dispase (Roche) and 25 μg/mL DNase I (Roche). Fluorescence in serum and homogenized organs was measured on a Tecan Infinite M200 Pro. Standard curves were generated using serum and organs from untreated mice with AF647-IFNs added ex vivo.

On day 0, mice were inoculated with 1 M B16F10 or MC38 cells in 50 μL PBS s.c. in the right flank. Treatments began on day 5 when tumors were established (average 24 mm² for B16F10 and 30 mm² for MC38). Mice were sorted so groups had equal average tumor area at the start of treatment. IFNs were always treated at 0.5 nmol (10 μg IFNα; 10 μg IFNβ; 43 μg MSA-IFNα; 12 μg ABP-IFNα; 12 μg ABP-IFNβ), based on prior IFNα treatments in the laboratory at this dose (54). I.t. doses were in 20 μL TBS. ABP-IFN doses also contained 60 μg alum and were incubated at room temperature for at least 20 min before treatment. Systemic MSA-IFNα was treated i.p. in 100 μL PBS. IFNs were dosed on days 5, 11, and 17, except for the BM chimera study where IFNs were treated on days 6, 12, and 18. For the alum only study, 120 μg alum in 20 μL was treated i.t. on days 5, 7, and 9. MSA-IL2 (30 μg in 50 μL PBS) and αPD1 (200 μg in 100 μL PBS) were treated i.p. on days 5, 11, and 17. Depleting antibodies aCD8α (Bio X Cell 2.43), aNK1.1 (Bio X Cell PK136), and aLy6G (Bio X Cell 1A8) were dosed at 400 μg i.p. every 4 d from day 3 until day 30. Tumor area (length × width) and body weight were monitored. Most survival data are compiled from multiple experiments. Mice were killed when tumors exceeded 100 mm², weight loss exceeded 20%, or mice exhibited poor body condition. Mice who died from tumor burden (as defined by the most recent tumor measurement exceeding 70 mm²) were included in the data. Any other unexpected deaths were omitted from the study. For rechallenge studies, mice were inoculated with 0.1 M B16F10 cells (s.c.) in the left flank at days 100 to 110. Mice were monitored for an additional 100 d and killed when their rechallenge tumor exceeded 100 mm².

Host mice (CD45.2⁺) were irradiated with 500 rad, allowed to recover for 3 h, and irradiated again with 550 rad. The next day, femurs and tibias were excised from congenically marked donor mice (CD45.1⁺), and the ends of the bones were cut. BM was harvested by centrifugation at 2,500 × g, 70-μm filtered, then washed and resuspended in PBS. The 7.5 M BM cells were injected retroorbitally into irradiated host mice within 24 h postirradiation. Engraftments were confirmed after 6 wk.

Tumors were excised, weighed, and ground with a pestle in 1:20 (wt/vol) of tumor in Tissue Protein Extraction Reagent (Thermo Fisher) supplemented with Halt Protease Inhibitor Mixture (Thermo Fisher) and 5 mM ethylenediaminetetraacetic acid (EDTA). Samples were incubated for 30 min at 4 °C, debris was pelleted by centrifugation at 10,000 × g, and supernatants were filtered through Corning Costar Spin-X tubes. The LEGENDplex mouse anti-virus response panel (13-plex) (BioLegend) was used following vendor instructions and analyzed with BD FACS LSR Fortessa.

Blood was collected by cheek bleed into K2 EDTA tubes (MiniCollect). Tumors, tdLNs, and spleens were excised, weighed, mechanically dissociated, and 70-μm filtered into single-cell suspensions. Spleens and blood were resuspended in ACK lysing buffer (Gibco). Cells were blocked with CD16/CD32 antibody (eBioscience Clone 93). For intracellular markers, cells were fixed and stained in permeabilization buffer (Invitrogen). Some tumors were stained with Zombie NIR (BioLegend 423105), Apotracker Green (BioLegend 427402), BUV395-CD45 (BD Biosciences 30-F11), AF647-TA99 (labeled in-house), PE/Cy7-MHC-I (BioLegend 28-8-6), PE-PD-L1 (BioLegend 10F.9G2), and BV605-Ki67 (BioLegend 16A8). Other tumors were stained with Zombie UV (BioLegend 423107), BUV395-CD45, BV785-CD3 (BioLegend 17A2), BUV737-CD8α (BD Biosciences 53-6.7), BV605-Ki67, BV711-TIM3 (BioLegend RMT3-23), and AF488-TCF1/TCF7 (Cell Signaling Technology C63D9). Other tumors were stained with Zombie Aqua (BioLegend 423101), BUV395-CD45, PE/Cy7-CD3, BUV737-CD8α, APC-PD-L1, PE-TIM3, and FITC-PD1 (BioLegend 29F.1A12). tdLNs were stained with Zombie NIR, BUV395-CD45, PE/Cy7-Ly6C (BioLegend HK1.4), AF647-MHCII (BioLegend M5/114.15.2), BV605-CD24 (BioLegend M1/69), and PE-CD86 (BioLegend GL-1). Spleens were stained with Zombie Aqua, BUV395-CD45, PE/Cy7-CD3, and BUV737-CD8α. Spleens were also stained with BV605-CD44 (BioLegend IM7), PE-CD127 (BioLegend A7R34), and APC-KLRG1 (BioLegend 2F1) in one panel, and AF647-CD25 (BioLegend PC61), BV605-Ki67, PE-TIM3, and AF488-TCF1/TCF7 in a separate panel. To confirm BM engraftment, blood was stained with Zombie Aqua, APC-CD45.1 (BioLegend A20), and PE-CD45.2 (BioLegend 104). To confirm antibody-mediated depletions, blood was stained with Zombie Aqua, BUV395-CD45, BV785-CD3, BUV737-CD8α, BV711-Ly6G (BioLegend 1A8), APC/Cy7-Ly6C (BioLegend HK1.4), BV421-CD11b (BioLegend M1/70), and PE/Cy7-NKp46 (BioLegend 29A1.4). BD FACS LSR Fortessa or BD FACS Symphony A3 analyzers were used. Data were analyzed with FlowJo V10.

Statistics were performed with GraphPad Prism software V7. Survival was compared by log-rank Mantel–Cox text. As described in figure captions, other metrics were compared by one-way or two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test, or unpaired t test. The n and P values are indicated in the legends.