Targeted TLR9 Agonist Elicits Effective Antitumor Immunity against Spontaneously Arising Breast Tumors

This article is featured in Top Reads, p. 167

Although immune checkpoint inhibitors (ICIs) are currently revolutionizing cancer care, only a minority of cancer patients respond to such therapies (1). For ICIs to be efficacious, pre-existing cytotoxic T cells must regain their ability to kill tumor cells when inhibitory checkpoint signaling is blocked, and consequently, ICIs are most effective for patients with T cell–inflamed (“hot”) tumors (1). This motivates the development of new immunotherapies or combination therapies that can effectively treat tumor-immune landscapes that are less responsive to ICIs, such as non–T cell-inflamed “cold” tumors.

There has been great interest in using immunostimulants, such as TLR agonists, to promote immune activation within the tumor microenvironment (TME) and stimulate antitumor immunity, converting tumors from cold to hot immune phenotypes (1–3). Upon binding to their cognate ligands, TLRs initiate signaling pathways that ultimately lead to the production of proinflammatory cytokines, type I IFNs, chemokines, costimulatory molecules, and other mediators that promote innate and adaptive immunity (2, 4). Although this inflammatory cascade is important for enhancing immune infiltration and activation in the TME, it can also lead to negative anti-inflammatory feedback loops that dampen immune responses (e.g., upregulating inhibitory checkpoints, anti-inflammatory cytokines). Thus, comprehensive assessment of these “positive” proinflammatory and “negative” anti-inflammatory effects may broaden understanding of how immune stimulation modulates the TME and enable identification of possible resistance mechanisms.

Relative to implanted tumor models, tumors that develop spontaneously in genetically engineered mouse models (GEMMs) better recapitulate the complexity and heterogeneity of human tumors (5–7), and thus may provide new or more relevant insights into mechanisms of immunotherapy efficacy and resistance. Furthermore, oncogene-driven spontaneous cancer models are often resistant to immunotherapy, including ICIs, and serve as challenging and highly relevant models to evaluate novel immunotherapy strategies.

In this study, we show that a systemically administered tumor-targeted TLR9 agonist conjugate, referred to as PIP-CpG, can effectively alter the TME and elicit an antitumor immune response against spontaneously arising tumors. PIP-CpG consists of an engineered tumor-targeting peptide, termed PIP, conjugated to CpG, which is a synthetic DNA oligonucleotide containing unmethylated CpG motifs that specifically activate TLR9 (3). This fully synthetic PIP-CpG conjugate is cross-reactive between mouse and human receptors and enables targeting to many types of solid tumors (3). We demonstrate that systemic PIP-CpG therapy stimulates robust T cell–mediated immunity that enables effective treatment of mouse mammary tumor virus (MMTV)–PyMT transgenic mice, which spontaneously develop multiple immunologically cold breast tumors (5, 7). In this study, we further investigate the therapeutic mechanism of action, identify potential resistance mechanisms, and provide an unprecedented assessment of how this targeted immunotherapy modulates the spontaneously arising tumor immune microenvironment.

PIP-CpG was synthesized as previously described (3). Briefly, to produce the PIP-CpG conjugate (∼13.8 kDa) with a 1:1 ratio of PIP to CpG, we synthesized PIP via solid-phase peptide synthesis (33 aa; ∼3.4 kDa) with an unnatural amino acid bearing an azide group (5-azido-l-norvaline), which was then reacted with 5′-dibenzocyclooctyne (DBCO)-modified CpG (∼10.4 kDa) via strain-promoted azide-alkyne cycloaddition (click chemistry). The CpG portion of PIP-CpG consists of the SD-101 CpG sequence, which is a species cross-reactive class C CpG DNA oligonucleotide that is being evaluated in clinical trials (8, 9). The sequence of SD-101 is 5′-TCGAACGTTCGAACGTTCGAACGTTCGAAT-3′ (with phosphorothioate bonds throughout the entire sequence).

Female FVB/N-Tg(MMTV-PyVT)634Mul/J mice (referred to as MMTV-PyMT mice; 5–6 wk old) were purchased from The Jackson Laboratory. All mouse experiments were approved by the Stanford Administrative Panel on Laboratory Animal Care and conducted in accordance with Stanford University animal facility guidelines.

For in vivo fluorescence imaging studies, mice were injected i.v. via the tail vein with 1.5 nmol of Alexa Fluor 680 (AF680)-labeled peptides (PIP-AF680 or nonbinding peptide [NBP]-AF680) to evaluate tumor uptake and organ biodistribution. PIP-AF680 and NBP-AF680 were synthesized as previously described (3). Mice were imaged using a Spectral Instruments Imaging Ami imager (excitation/emission, 640/730 nm; exposure time, 10 s), and results were analyzed using Aura imaging software.

For in vitro studies, MMTV-PyMT tumor cells were isolated from a transgenic MMTV-PyMT mouse and grown in cell culture in DMEM (Life Technologies, 11995073) + 20% heat-inactivated FBS (HI FBS; Life Technologies, 10438026) + 1% penicillin/streptomycin (P/S; 100× solution, Life Technologies, 15140122). A20 murine B cell lymphoma and EL4 murine T cell lymphoma cell lines were obtained from the American Type Culture Collection and were grown in cell culture in RPMI 1640 (Life Technologies, 11875119) + 10% HI FBS + 1% P/S + 50 μM 2-ME (Sigma-Aldrich). All cells were grown at 37°C in 5% CO₂.

For PIP staining experiments, the staining buffer (PBSA+CM) was PBS + 0.5% BSA with 0.9 mM CaCl₂ and 0.5 mM MgCl₂ (divalent cations are required for PIP to bind integrins). Peptide–Fc fusions, which consist of a peptide (either PIP or NBP) genetically fused to the N-terminus of a human IgG1 Fc domain, were used for in vitro staining. The peptide–Fc fusions were labeled with Alexa Fluor 647 (AF647) via N-hydroxysuccinimide (NHS) ester labeling as previously described to produce PIP-Fc-AF647 and NBP-Fc-AF647 (average fluorophore/protein ratio of 4.6) (3). MMTV-PyMT tumor cells (1 × 10⁵ cells/sample) were incubated for 1 h at 4°C in the dark in 100 μl of PBSA+CM only (used for baseline subtraction) or PBSA+CM containing 100 nM peptide–Fc fusion (PIP-Fc-AF647 or NBP-Fc-AF647). Cells were washed three times with cold PBSA (PBS + 0.5% BSA), resuspended in PBSA containing 50 nM SYTOX Green (Invitrogen, S7020), and then analyzed on a BD Accuri C6 Plus flow cytometer.

For intracellular TLR9 staining, all cell types (MMTV-PyMT, A20, and EL4) were tested in parallel and using the same conditions with the exception that the immune cell lines (EL4 and A20) were both pretreated with mouse Fc Block (BioLegend, 156603) prior to permeabilization as specified by the manufacturer. Cells were fixed/permeabilized using the BD Cytofix/Cytoperm fixation/permeabilization kit (BD Biosciences, 554714) in accordance with the manufacturer’s instructions. Briefly, cells were incubated for 20 min at 4°C in BD fixation/permeabilization solution and then washed twice in BD Perm/Wash buffer. Cells (1 × 10⁵ cells/sample) were incubated for 30 min at room temperature in the dark in 100 μl of BD Perm/Wash buffer only (used for baseline subtraction) or BD Perm/Wash Buffer containing anti-mouse TLR9-PE (BD Biosciences, 565640) or the appropriate isotype-PE control (BD Biosciences, 554680) (2.5 μl of Ab in each 100 μl sample). Cells were washed three times with BD Perm/Wash buffer, resuspended in PBS, and then analyzed on a BD Accuri C6 Plus flow cytometer.

The dose of PIP-CpG was 18.2 nmol (250 μg) and was administered i.v. via tail vein in 100 µl of PBS. For T cell depletion, anti-CD4 (100 μg; Bio X Cell, BE0003-1; clone GK1.5, rat IgG2b) and anti-CD8a (100 μg; Bio X Cell, BE0061; clone 2.43, rat IgG2b) Abs were injected i.p. 2 d before therapy, on the day of the first treatment, and at 6, 13, and 20 d posttreatment. T cell depletion in blood was validated by flow cytometry. Treatments (vehicle or PIP-CpG) and tumor growth measurements were initiated when tumors became palpable (typically ∼50 d old; usually two to three tumors per mouse). For all experiments, mice were methodically assigned to treatment groups to ensure that mice in different groups were well matched on the first day of treatment (i.e., similar number of tumors, total tumor burden, age). The size of each tumor, expressed as volume (length × width × height), was monitored continuously with digital calipers (Mitutoyo) until the euthanasia end point was reached (criterion: when any tumor reached 1.5 cm in the largest diameter). Nair hair remover lotion was used to remove fur from mice to facilitate tumor measurements. To minimize potential confounders, mice from different treatment groups were housed together.

This assay was performed as previously described (3) with some modifications. Single-cell suspensions were made from spleens of treated mice, and RBCs were lysed with ammonium chloride and potassium buffer (Quality Biological). Splenocytes were cultured for 24 h at 37°C and 5% CO₂ in the presence of 0.5 μg/ml of purified NA/LE anti-mouse CD28 mAb (BD Pharmingen) with or without 0.5 × 10⁶ MMTV-PyMT tumor cells. For this study, the MMTV-PyMT tumor cells were isolated from a separate untreated mouse and grown in short-term explant cell culture to propagate tumor cells (and remove primary healthy cells). For the positive stimulation control, splenocytes were treated with 0.5 μg/ml of purified NA/LE anti-mouse CD3 (BD Pharmingen) in addition to 0.5 μg/ml of anti-mouse CD28 mAb. For all samples, the incubation medium was RPMI 1640 + 5% HI FBS + 1% P/S (final volume, 500 μl). Monensin (0.3 μl/sample; GolgiStop; BD Biosciences) was added for the last 5–6 h of incubation to inhibit protein transport. Extracellular staining was performed as described for the immune profiling studies using FITC anti-mouse CD8 (BD Biosciences, 553031), PerCP anti-mouse CD4 (BD Biosciences, 553052), and allophycocyanin anti-mouse CD44 (BD Biosciences, 559250) (1 μl of each Ab per sample). The BD Cytofix/Cytoperm Plus kit was used according to the manufacturer’s instructions for assessing intracellular IFN-γ using PE anti-mouse IFN-γ (BD Biosciences, 554412) (2 μl per sample). Samples were analyzed on a BD FACSCalibur flow cytometer. The same MMTV-PyMT mice were used for assessing the CD8⁺ T cell response to shared tumor Ags (spleen), tumor cytokines/chemokines via Luminex (tumor 1), and the immune profile via flow cytometry (tumor 2).

MMTV-PyMT cancer cells were seeded into a 96-well flat-bottom plate at 5000 cells/well in 200 μl of medium (DMEM + 20% HI FBS + 1% P/S). The next day, medium was removed and replaced with 200 μl of fresh medium containing the following treatments: PIP-CpG, PIP, CpG, GpC, PIP + CpG (mixed), or PIP + GpC (mixed), at 100, 500, and 1000 nM. The GpC negative control sequence is 5′-TGCAAGCTTGCAAGCTTGCAAGCTTGCAAT-3′ (with phosphorothioate bonds throughout the entire sequence) and was purchased from the Protein and Nucleic Acid (PAN) facility at Stanford University. After a 24-h incubation, 50 μl of the supernatant was immediately used to assess cytotoxicity using the Pierce LDH (lactate dehydrogenase) cytotoxicity assay kit according to the manufacturer’s protocol. The remaining supernatant was frozen and stored overnight at −20°C. The next day, the supernatant was thawed and cytokine levels were analyzed using the BD CBA (cytometric bead array) mouse inflammation kit on a BD Accuri C6 Plus flow cytometer.

Tumors were excised, weighed, and then homogenized using a Dounce homogenizer containing lysis buffer (∼20 µl/mg tumor): tissue protein extraction reagent (T-PER; Thermo Fisher Scientific, 78510) with 1% Halt protease and phosphatase inhibitors (Thermo Fisher Scientific, 78442). Lysis buffer containing tumor homogenate was then transferred to Eppendorf tubes and incubated for 30 min at 4°C with rotation. To remove debris, samples were centrifuged (10,000 × g for 10 min at 4°C), and the supernatant was subsequently filtered using Corning Costar Spin-X columns. Lysates were aliquoted and stored at −80°C. A bicinchoninic acid assay (Pierce) was used to measure total protein concentration of each sample, and then the appropriate amount of lysis buffer was added to samples to ensure that all samples had the same concentration of total protein. All samples were diluted 2× in PBS prior to running the Luminex assay. The Luminex assay was performed by the Human Immune Monitoring Center at Stanford University using the immune monitoring mouse 48-plex ProcartaPlex panel (Thermo Fisher Scientific, EPX480-20834-901). A detailed protocol is available on the Stanford Human Immune Monitoring Center Web site (https://iti.stanford.edu/himc/protocols.html). Plates were read using a Luminex FLEXMAP 3D instrument with a lower bound of 50 beads per sample per analyte.

Tumors were excised, weighed, and then mechanically dissociated into single-cell suspensions. Staining for flow cytometry was performed as previously described (3) with some modifications. Cell samples were stained with LIVE/DEAD fixable blue dead cell stain (Invitrogen, L23105) in PBS for 30 min in the dark at 4°C and then washed twice. Cells were resuspended in 100 μl of cold FACS buffer (PBS + 1% BSA + 0.02% sodium azide) containing mouse Fc Block (anti-mCD16/32) and incubated for 15 min at room temperature. Without washing, fluorescently labeled Abs against surface markers of interest were added directly to samples (1 μl of each extracellular Ab) in addition to 50 μl of Brilliant Stain Buffer (BD Biosciences, 563794). Samples were incubated for 15 min at room temperature in the dark and then washed twice with FACS buffer. To enable intracellular Ab staining, samples were resuspended in 1 ml of cold fixation/permeabilization buffer (eBioscience, 00-5521-00), incubated for 30 min at room temperature in the dark, and then washed twice with permeabilization buffer (eBioscience, 00-8333-56). Note that for the initial steps, centrifuge speed was 290 × g; however, after cells are permeabilized, centrifuge speed was increased to 800 × g. Samples were resuspended in 100 μl of permeabilization buffer, and fluorescently labeled Abs against intracellular markers of interest were added to the appropriate samples (3 μl of each intracellular Ab). Samples were incubated for 30 min at room temperature in the dark and then washed twice with permeabilization buffer. Samples were then resuspended in 250 μl of fixation buffer (PBS + 2% paraformaldehyde), incubated for 15 min at room temperature in the dark, and washed once with PBS. Samples were resuspended in 250 μl of PBS and stored at 4°C in the dark until they were analyzed on a Cytek Aurora (five laser configuration) spectral flow cytometer.

Prior to sample running, the volume of each sample was adjusted to ensure that all samples had the same known volume. The weight of the tumor, initial sample volume, and collected sample volume (as measured by the Cytek Aurora) were used to calculate the number of cells per milligram of tumor for each sample. UltraComp eBeads plus compensation beads (Invitrogen, 01-3333-42) and the ArC Amine Reactive Compensation bead kit (Invitrogen, A10628) were used for spectral unmixing. Extracellular Abs used included the following: BUV395 anti-CD45 (BD Biosciences, 565967), BUV496 anti-B220 (BD Biosciences, 612950), BUV563 anti-Ly6G (BD Biosciences, 612921), BUV615 anti–PD-1 (CD279) (BD Biosciences, 752299), BUV661 anti-CD11b (BD Biosciences, 612977), BUV737 anti-CD4 (BD Biosciences, 612843), BUV805 anti-F4/80 (BD Biosciences, 749282), BV421 anti-TIGIT (CD226) (BD Biosciences, 565270), BV480 anti-CD49b (BD Biosciences, 746355), BV510 anti–MHC class II (MHC II; I-A/I-E) (BD Biosciences, 742893), BV570 anti–Gr-1 (BioLegend, 108431), BV605 anti-CD69 (BD Biosciences, 563290), BV650 anti-CD80 (BD Biosciences, 563687), BV786 anti-CD134 (OX40) (BD Biosciences, 740945), FITC anti–Siglec F (BioLegend, 155503), PerCP-Cy5.5 anti-CD3e (BD Biosciences, 551163), PE anti–PD-L1 (CD274) (BD Biosciences, 558091), PE-Cy5 anti-CD86 (BioLegend, 105015), PE-Cy7 anti-CD11c (BD Biosciences,558079), and allophycocyanin-H7 anti-CD8 (BD Biosciences, 560182). Intracellular Abs used included the following: BV711 anti–Ki-67 (BD Biosciences, 563755), allophycocyanin anti–CTLA-4 (CD152) (eBioscience, 17-1522-82), and AF700 anti-Foxp3 (eBioscience, 56-5773-82).

GraphPad Prism 9 software was used for all statistical analyses. Significance is represented in figures as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and ns (not significant) for p ≥ 0.05. The specific statistical methods for individual experiments are indicated in the figure legends. Nonparametric statistical tests were used for non-normally distributed data.

Flow cytometry data were analyzed using Cytobank (www.cytobank.org). All heatmaps were generated using Morpheus software (https://software.broadinstitute.org/morpheus). Some figures include schematics that were created using BioRender.com.

The MMTV-PyMT model is one of the most commonly used GEMMs for cancer research as it closely resembles the progression and morphology of human breast cancers (5, 6). In this autochthonous model, the MMTV promoter/enhancer is used to drive mammary-gland specific expression of the oncogenic PyMT, causing female carriers to spontaneously develop palpable breast tumors by 6–8 wk of age (10). Over time, female MMTV-PyMT mice develop independently arising tumors in all 10 of their mammary glands.

To enable immunostimulant delivery to every breast tumor, we used a systemically administered tumor-targeting TLR9 agonist, referred to as PIP-CpG. The tumor-targeting component, PIP (formerly known as 2.5F), is an engineered cystine knot (knottin) peptide that binds multiple tumor-associated integrin receptors, including α₅β₁, α_vβ₁, α_vβ₃, α_vβ₅, and α_vβ₆, thereby enabling targeting to many solid tumor types (11–13). PIP’s tumor-targeting capabilities and desirable biodistribution profile have been extensively characterized in prior positron emission tomography and near-infrared imaging studies (3, 14–17). To verify that PIP could also localize to spontaneous MMTV-PyMT breast tumors, we performed an in vivo fluorescence imaging study using AF680-labeled PIP (PIP-AF680). Following i.v. injection, PIP-AF680 was detected in each of the MMTV-PyMT breast tumors and had significantly higher tumor uptake compared with a non-binding control (NBP-AF680), demonstrating that tumor uptake was dependent on PIP’s integrin-binding capability rather than nonspecific accumulation in the tumor (Fig. 1A, 1B). Consistent with previous studies (3, 14–17), PIP exhibited nonspecific uptake in the kidneys due to renal clearance but did not accumulate in the liver, spleen, or heart (Supplemental Fig. 1A, 1B). PIP binding directly to MMTV-PyMT tumor cells was also confirmed in vitro (Supplemental Fig. 1C).

We next evaluated whether tumor-localized TLR9 stimulation via PIP-CpG therapy would elicit an effective therapeutic response (Fig. 1C, 1D). Tumor-bearing MMTV-PyMT mice treated i.v. with three doses of PIP-CpG had significantly lower tumor burden and exhibited prolonged survival relative to vehicle-treated mice (Fig. 1E–H). Remarkably, 80% of the PIP-CpG–treated mice exhibited complete regression of one or more of their tumors, whereas none of the vehicle-treated mice exhibited tumor regression. Furthermore, 1 mo after treatment, PIP-CpG–treated mice had approximately half the number of tumors as vehicle-treated mice (Fig. 1G). These results suggest that systemic PIP-CpG therapy could be an effective strategy for treating patients with multiple tumor sites.

Importantly, PIP-CpG therapy not only inhibited growth of tumors that were present at the time of treatment (referred to as “initial tumors”), but it also delayed growth of newly arising tumors that developed after the treatment period (Fig. 1I). These newly arising tumors are neither recurrences nor metastases of the initial tumors but are new, independent MMTV-PyMT oncogene-driven cancers that developed in different mammary gland locations relative to the initial tumors. Despite their delayed growth, the newly arising tumors eventually progressed, whereas tumors that were established at the time of PIP-CpG treatment either regressed or exhibited minimal growth throughout the study.

Given that PIP-CpG delayed the growth of newly arising tumors, we hypothesized that an adaptive T cell–mediated immune response was elicited against tumor Ags shared by independently arising tumors (e.g., PyMT-derived peptide Ags) (18). To test this hypothesis, we harvested splenocytes from PIP-CpG–treated mice and coincubated them with MMTV-PyMT tumor cells isolated from a separate untreated MMTV-PyMT mouse (Fig. 2A). Coincubation with tumor cells stimulated IFN-γ production in effector/memory CD44⁺CD8⁺ T cells derived from PIP-CpG–treated mice, but not vehicle-treated mice, indicating that PIP-CpG therapy elicits a T cell response against Ags that are shared among MMTV-PyMT tumors.

We next asked whether T cells were responsible for mediating therapeutic efficacy. Ab-mediated CD4⁺ and CD8⁺ T cell depletion significantly diminished the efficacy of PIP-CpG therapy, confirming that T cells are critical for therapeutic activity (Fig. 2B–E). Although it was not statistically significant, there was a slight reduction in tumor growth in the PIP-CpG + T cell depletion group compared with the vehicle-treated group. This trend may suggest potential therapeutic effects mediated by cells/mechanisms other than T cells, although the effects in this model were minimal.

With regard to the newly arising tumors, this depletion study also showed that T cells were required for PIP-CpG to delay onset of newly arising tumors (Fig. 2F), supporting the hypothesis that this effect is mediated by T cells that recognize shared tumor Ags. Overall, these results demonstrate that systemic administration of PIP-CpG can drive an effective antitumor T cell–mediated response against spontaneously arising mammary tumors.

We next investigated whether PIP-CpG had any direct effects on MMTV-PyMT cells in cell culture. Although some cancer cells have been found to express TLR9, we did not detect TLR9, which is active within endosomal compartments, in MMTV-PyMT cells as measured by intracellular flow cytometry (Supplemental Fig. 1D, 1E). We also verified that stimulating MMTV-PyMT cells with CpG or PIP-CpG in cell culture did not result in TLR9-mediated cytokine production (six analytes: TNF-α, IL-6, IFN-γ, IL-12, IL-10, and MCP-1; Supplemental Fig. 1F, 1G). Interestingly, we found that CpG did lead to slightly elevated production of MCP-1, but this result was not mediated through TLR9 activation, given that similar effects were observed with a GpC oligonucleotide control (CpG motifs are replaced by nonstimulatory GpC), which has similar structural and chemical properties as CpG but does not activate TLR9.

In cell culture, PIP by itself can affect the phenotype of certain adherent cancer cells, causing them to ball up and become less adherent at high concentrations, presumably because high prolonged exposure is sufficient to block integrin adhesion to relevant extracellular matrix proteins in two-dimensional cell culture (19). However, PIP alone has not shown therapeutic efficacy in vivo across multiple tumor models (Ref. 3 and unpublished observations). We also confirmed that none of the treatments tested (PIP-CpG, PIP, CpG, GpC, PIP + CpG mixed, PIP + GpC mixed) was directly cytotoxic to MMTV-PyMT cells in cell culture (Supplemental Fig. 1H).

To gain a comprehensive understanding of how PIP-CpG modulates the TME in vivo in the MMTV-PyMT model, we first evaluated cytokine and chemokine secretion in the tumors using a 48-plex Luminex panel (data for all analytes are shown in Supplemental Fig. 2A). Of the 48 analytes tested, 27 cytokines/chemokines were observed at significantly higher levels with PIP-CpG therapy (Fig. 3A), whereas only one analyte (G-CSF/CSF-3) was lower following the treatment (Fig. 3B). Remarkably, we found that 7 of the 10 chemokines that were analyzed were significantly increased with PIP-CpG treatment, including MCP-1 (CCL2), MIP-1α (CCL3), MIP-1β (CCL4), RANTES (CCL5), MCP-3 (CCL7), EOTAXIN (CCL11), and IP-10 (CXCL10), whereas ENA-78 (CXCL5), GRO-α (CXCL1), and MIP-2 levels were unaffected by therapy (Fig. 3C–I). The chemokines with increased expression are known to recruit various types of immune cells, including macrophages, dendritic cells (DCs), NK cells, myeloid-derived suppressor cells (MDSCs), and T cells (20). Moreover, the chemokines CCL2, CCL3, CCL4, CCL5, and CXCL10 can recruit CD8⁺ T cells and are associated with increased T cell infiltration in human tumors (21).

In addition to induction of chemokine secretion, PIP-CpG treatment had profound effects on cytokine production as well. Notably, PIP-CpG amplified the expression of several Th1 cytokines, including IFN-γ, IL-12, TNF-α, IL-18, IL-1β, and IL-27, which support cytotoxic T cell responses (22). Interestingly, we also observed that PIP-CpG treatment upregulated several Th2 cytokines (i.e., IL-4, IL-5, and IL-13); however, the effect sizes for Th1 cytokines were much higher than for Th2 cytokines, supporting the notion that TLR9 stimulation favors the activation of Th1-oriented immunity (22).

Although T cell depletion abrogated the therapeutic response of PIP-CpG, TLR9 stimulation can modulate the TME through activation and infiltration of many immune cell types. Thus, to explore the cellular landscape of spontaneous breast tumors, we used spectral flow cytometry (>20 channels) to characterize various immune populations in the TME and assess their phenotypes (gating strategy is shown in Supplemental Fig. 2C).

Although spontaneous MMTV-PyMT tumors exhibit substantial histological and transcriptional heterogeneity (5), the immune profiles of vehicle-treated tumors were homogeneous and showed very minimal immune infiltration (<10% of live cells) (Fig. 4, Supplemental Fig. 2B). The two most prominent immune cell types in vehicle-treated tumors were macrophages and putative granulocytic MDSCs (gMDSCs; also known as PMN-MDSCs), which are both frequently observed in human breast cancers (23). Surprisingly, the third most abundant cell type was B220⁺ CD8⁺ T cells, which made up 50–70% of all CD8⁺ T cells (Supplemental Fig. 3A–C). This atypical B220⁺ CD8⁺ T cell subset has been previously described to possess regulatory suppressor functions in mice and humans and has been reported to accumulate in the TME of an orthotopic glioblastoma mouse model (24, 25). Compared to B220⁻ CD8⁺ T cells, we found that these B220⁺ CD8⁺ T cells expressed higher levels of the inhibitory checkpoint molecule TIGIT (Supplemental Fig. 3D), which could contribute to their previously described immunosuppressive activity. In addition, within vehicle-treated tumors, we also observed low levels of other T cell populations, NK cells, B cells, and putative monocytic MDSCs (mMDSCs), whereas DCs and eosinophils were too sparse in these tumors to accurately characterize.

Consistent with enhanced chemokine secretion, PIP-CpG treatment greatly enhanced immune infiltration relative to vehicle, both in terms of percentage of total live cells and in the number of immune cells per milligram of tumor (Fig. 4A, 4B). Whereas PIP-CpG treatment increased infiltration of both innate and adaptive immune cells, innate infiltration was relatively homogeneous among the mice, whereas adaptive immune infiltration was much more heterogeneous (Fig. 4C). Notably, immunosuppressive B220⁺ CD8⁺ T cells exhibited minimal infiltration across all PIP-CpG–treated mice, suggesting that newly infiltrating CD8⁺ T cells were conventional B220⁻ CD8⁺ T cells. Correlation analysis on PIP-CpG–treated tumors revealed that most adaptive immune populations showed a significant positive correlation with one another (Fig. 4D), meaning that high infiltration of B220⁻ CD8⁺ T cells was accompanied by high levels of B cells and CD4⁺ T cell subsets (and vice versa).

Although the chemokines secreted following PIP-CpG therapy facilitate effector immune cell infiltration, they can also indiscriminately lead to increased infiltration of immunosuppressive cell types, such as regulatory T cells (Tregs) and MDSCs. Indeed, Treg infiltration was strongly correlated with infiltration of other adaptive immune cells (Fig. 4D, Supplemental Fig. 3E), and these Tregs expressed several inhibitory checkpoint molecules. In fact, both PD-1 and CTLA-4 were found exclusively on subsets of Tregs, whereas TIGIT was found on Tregs, CD8⁺ T cells, and NK cells (Supplemental Fig. 4A). Given that increased Treg infiltration could potentially lead to a higher density of checkpoint molecules in the TME, we next performed a correlation analysis to assess the number of checkpoint-expressing Tregs relative to total infiltrating Tregs. The infiltration of CTLA-4⁺ Tregs and TIGIT⁺ Tregs was higher following treatment with PIP-CpG therapy and correlated with increasing numbers of Tregs (Supplemental Fig. 4B–D). In contrast, the number of PD-1⁺ Tregs was unaffected by PIP-CpG therapy and thus did not correlate with increasing Treg infiltration, suggesting that most of the Tregs that infiltrated following PIP-CpG therapy were PD-1⁻ (Supplemental Fig. 4E). Overall, despite the elevated levels of Tregs, the immune-stimulating capacity of PIP-CpG was sufficient to alter the immune landscape and promote tumor regression in this aggressive GEMM.

In addition to enhancing immune infiltration, we found that PIP-CpG treatment altered the phenotype of multiple immune populations, and these changes were consistent across all PIP-CpG–treated mice despite their heterogeneous adaptive immune infiltration. Specifically, tumors from PIP-CpG–treated mice were highly enriched with activated (CD86⁺) and proliferating (Ki-67⁺) T cells (Fig. 5A, 5B), consistent with the critical role they are playing in therapeutic efficacy. Although the CD86 costimulatory molecule is typically associated with professional APCs, functional CD86 is upregulated on T cells in response to certain activating stimuli and has the potential to costimulate naive T cell responses (26, 27). T cells can also acquire CD86 via membrane transfer from surrounding CD86-expressing APCs through trogocytosis or exosomes (28, 29); however, these membrane transfer methods seem unlikely in this case given that CD86⁺ T cells were generally negative for other APC surface markers. Similar to T cells, many of the B cells in tumors of mice treated with PIP-CpG were proliferating (Fig. 5B, 5C), and thus the increase in adaptive immune cells is likely driven by both chemokine-mediated recruitment as well as expansion within the TME. In addition, CD86 and MHC II levels were upregulated on tumor-infiltrating B cells following PIP-CpG therapy, indicative of higher Ag presentation capacity (Fig. 5A, 5D).

In addition to stimulating adaptive immune cells, PIP-CpG treatment induced intriguing phenotypes on innate immune cells as well. Notably, we observed an unexpected macrophage phenotype, in which therapy with PIP-CpG led to lower MHC II and CD86 levels, but higher CD69 expression (Fig. 5E–G); these results suggest that whereas PIP-CpG enhances APC function in B cells, it has the opposite effect on macrophages in the TME. In addition to macrophages, we found that PIP-CpG treatment increased the number of tumor-infiltrating NK cells and also altered their phenotype, leading to increased levels of CD11b (Mac-1), CD49b (α₂ integrin), B220, and Ki-67 (Fig. 5H–J), consistent with activated/mature NK cells that have enhanced lytic activity and greater capacity to produce IFN-γ (30, 31).

Collectively, these results indicate that PIP-CpG generates and propagates an inflammatory cascade within the MMTV-PyMT spontaneous breast tumor model, leading to amplified expression of numerous chemokines/cytokines, including IFN-γ, as well as robust immune infiltration and activation of many innate and adaptive immune cells.

In this study, we found that PIP-CpG promoted innate and adaptive immune activation and infiltration in the TME and ultimately led to a T cell–mediated immune response that was essential for tumor regression. Although innate immune cells and B cells were not sufficient to induce tumor regression in the absence of T cells, these cell types can contribute to the antitumor immune response by promoting Ag presentation and by converting the TME into a proinflammatory state that enables effective T cell infiltration and killing; B cells also have the potential to produce tumor Ag-targeting Abs, which could contribute to therapeutic activity. Previous studies have shown that CpG can stimulate DCs and B cells to secrete various proinflammatory cytokines and upregulate MHC II and costimulatory molecules to promote Ag presentation (32–34). Consistent with this prior literature, we also observed elevated MHC II and CD86 expression on tumor-infiltrating B cells following PIP-CpG therapy.

Although macrophages often play immunosuppressive roles in the TME, macrophages stimulated by CpG generally exhibit antitumor activity, as they produce numerous inflammatory cytokines/chemokines and have enhanced phagocytotic and NO-dependent tumoricidal activity (35, 36). In contrast to DCs and B cells, the ability of CpG to modulate expression of MHC II and costimulatory molecules on macrophages is less clear, as many studies have reported conflicting findings for this heterogeneous cell population (34, 35, 37, 38). Furthermore, previous studies have been performed almost exclusively in vitro using macrophages from various sources, which may not represent the effects in vivo within the TME. Interestingly, within spontaneous breast tumors, we found that PIP-CpG therapy dramatically altered the macrophage phenotype, leading to lower levels of MHC II and CD86 but higher CD69 expression. These in vivo findings aligned most closely to the results of an in vitro study using bone marrow–derived macrophages (differentiated with L929 conditioned media), wherein treatment with CpG enhanced cytokine production and phagocytic activity, but Ag presentation capacity was reduced due to decreased expression of MHC II and CD86 (34). Overall, these data suggest that macrophages do not play a major role in Ag presentation following PIP-CpG therapy in this model, but they could support the antitumor immune response through proinflammatory cytokine production, which enables indirect activation of non–TLR9-expressing cells. Indeed, although NK cells do not respond directly to TLR9 agonists, CpG can stimulate surrounding myeloid cells to secrete IL-12, TNF-α, and IFNs, which in turn, would activate NK cells and enhance their lytic activity (39). Consistent with this mechanism, we found that PIP-CpG amplified production of these cytokines in the TME and induced an activated/mature NK cell phenotype that is associated with higher cytotoxic activity and higher IFN-γ production (30, 31). Similar to NK cells, CD8⁺ T cells and Th1 CD4⁺ T cells can also produce IFN-γ upon activation, which then leads to further activation of surrounding cells (22). In alignment with this Th1-oriented inflammatory cascade, we found that PIP-CpG amplified production of numerous Th1-type cytokines and promoted infiltration and expansion of activated CD8⁺ and CD4⁺ T cells in MMTV-PyMT tumors.

Importantly, while tumor immune profiling is useful for investigating how the TME changes in response to PIP-CpG administration, it is difficult to ascertain which of these effects directly link to therapeutic outcome. We also note that we performed these studies at a relatively early time point (day 3 posttreatment) to ensure that there would be sufficient tumor tissue remaining for analysis, and as we were most interested in elucidating changes occurring as a result of the TLR9-mediated inflammatory cascade.

An intriguing observation was that although PIP-CpG elicited an immune response that was sufficient to initially delay growth of newly arising tumors, these tumors eventually progressed; in contrast, initial tumors either regressed or had minimal growth throughout the study, suggesting that a stronger and more durable response was generated against initial tumors. In this GEMM, in addition to mammary gland–specific expression of PyMT, additional genetic alterations are needed for full malignant transformation, resulting in substantial genetic heterogeneity between individual MMTV-PyMT tumors (5, 40). Thus, while all MMTV-PyMT tumors possess some shared tumor Ags (e.g., PyMT-derived peptide Ags), each tumor also possesses its own unique set of tumor Ags due to its genetic differences (18, 40). Consequently, one possible explanation for our observation is that T cells could have been elicited against both shared tumor Ags and unique Ags found in initial tumors upon PIP-CpG therapy, leading to a more robust immune response against initial tumors compared with newly arising tumors. Another non–mutually exclusive possibility is that the stronger and more durable response observed for initial tumors was due to PIP-CpG–induced TME remodeling, whereas newly arising tumors, which were never exposed to PIP-CpG, were able to develop sufficient immune evasion tactics over time. Nevertheless, PIP-CpG’s robust therapeutic activity toward initial existing tumors is highly encouraging for clinical translation.

For nonhematological solid tumors injected intratumorally with CpG, the TLR9-mediated inflammatory cascade is thought to be initiated by CpG uptake in tumor-infiltrating TLR9⁺ immune cells. Some cancer cells also express TLR9, but the extent to which that expression equates to functional activity and how that may influence therapeutic response are not well understood. Given that PIP-CpG directly binds to cancer cells, we questioned whether direct TLR9 engagement in cancer cells could be important for PIP-CpG’s mechanism of action. However, MMTV-PyMT tumor cells did not express TLR9 and did not exhibit a TLR9-mediated functional response in cell culture, indicating that TLR9 expression by cancer cells is not a requirement for the therapeutic activity of PIP-CpG. Furthermore, we recently found that treating TLR9-expressing 4T1 breast cancer cells with CpG or PIP-CpG in vitro did not result in any obvious signs of functional TLR9 activation (e.g., cytokine production, changes in activation marker/MHC expression), suggesting that even if tumor cells do express TLR9, this may not always correspond to functional activity. Thus, we hypothesize that PIP binding to integrins on the tumor cells mainly allows the CpG to accumulate in the tumors and, once there, is taken up by surrounding TLR9⁺ immune cells. Some potential mechanisms for immune uptake include 1) DNA binding receptor-mediated uptake upon direct cell–cell contact, 2) capture of PIP-CpG that detaches from tumor cells (via DNA binding receptors or pinocytosis), or 3) phagocytosis of apoptotic tumor cells/debris or exosomes containing PIP-CpG. Furthermore, PIP can also bind integrins on tumor-infiltrating plasmacytoid DCs and macrophages (3), which could promote direct PIP-CpG uptake in those APCs.

Many studies have aimed to identify combination immunotherapy strategies to improve abscopal responses of intratumorally delivered TLR agonists (8, 41–44). Although systemic targeted delivery bypasses the need to generate abscopal responses, our findings in this study suggest that these combination strategies may also be effective for counteracting potential acquired resistance mechanisms to PIP-CpG therapy. For example, we found that the immunosuppressive IL-10 cytokine was elevated in PIP-CpG–treated tumors, supporting the concept of combining PIP-CpG with Abs that block IL-10 or IL-10R. Furthermore, the therapy-mediated influx of adaptive immune cells led to increased levels of Tregs in the TME, supporting the notion of combining PIP-CpG with immunotherapies that can deplete or neutralize the suppressive effects of Tregs, such as anti–CTLA-4, anti-TIGIT, anti-OX40, anti-GITR, and anti-FR4. Among the combination immunotherapy strategies, intratumoral TLR9 stimulation and anti–PD-1 has thus far demonstrated the most promising results in clinical trials (9, 45). Consistent with the MMTV-PyMT model being resistant to anti–PD-1/anti–PD-L1 therapy (7, 46), CD45⁻ cells in the TME were PD-L1⁻, and PD-1 expression was limited to a small subset of Tregs in both vehicle- and PIP-CpG–treated tumors. However, PD-L1 was broadly expressed by immune cells in the TME, and consistent with increased IFN-γ signaling, PD-L1 was upregulated on several immune populations after PIP-CpG therapy (Supplemental Fig. 4F–H). Although this may suggest that anti–PD-1/anti–PD-L1 therapy could be effective when combined with PIP-CpG, we do note that the relative contribution of PD-L1 expressed by immune cells versus tumor cells in suppressing antitumor T cell responses has yielded conflicting results, particularly for human disease, and may be cell type– or context-dependent (47). Finally, the TLR9-mediated inflammatory cascade has the potential to increase JAK/STAT3 signaling in cancer cells and myeloid cells, which is a known mechanism that counteracts antitumor immune responses (48, 49). Thus, therapeutic strategies that combat this negative feedback regulation could serve as another opportunity to enhance PIP-CpG–based therapy (49, 50).

A key finding in this work was that tumor-localized immune stimulation could be achieved in multiple spontaneously arising tumors using systemic PIP-CpG therapy. Intratumoral delivery can be challenging in human patients, particularly for repeated dosing in tumors that are not easily accessible. Thus, tumor-targeted delivery strategies have the potential to greatly improve the clinical applicability of immunostimulant therapy (3, 51–55), and some first-generation Ab-based conjugates have already entered early clinical trials (52, 54). Our fully synthetic PIP-CpG conjugate is a promising candidate for clinical testing given that it is species cross-reactive and binds to multiple integrins that are overexpressed on tumor cells and their vasculature in murine and human tumors (11–13). In particular, in addition to the GEMM used in this work, we have previously demonstrated that PIP can effectively target numerous types of cancer cells, including carcinomas (colon, lung, liver, ovarian, breast, pancreatic), sarcomas (fibrosarcoma, osteosarcoma, Ewing sarcoma, rhabdomyosarcoma), brain tumors (glioblastoma, medulloblastoma), and melanoma (3, 12, 15, 16, 56, 57). Relative to Ab conjugates, PIP-CpG leverages rapid tumor uptake followed by fast clearance from circulation to minimize systemic exposure. As such, PIP-CpG therapy is able to improve therapeutic efficacy while maintaining a similar safety profile as unmodified CpG (3), which is promising given that both local and systemic CpG therapy has been well tolerated in human clinical trials (2). Overall, this work motivates further development of systemic PIP-CpG therapy, as it could be used to treat a wide variety of solid tumor types, including those that are not easily accessible for intratumoral injection.

C.L.M., I.S.-B., R.L., C.R.B., and J.R.C. are inventors on a patent related to this work that is owned by Stanford University. J.R.C. is a cofounder and equity holder of Trapeze Therapeutics, Inc. and Combangio, Inc., an equity holder of Aravive, Inc., and a member of the Board of Directors of Ligand Pharmaceuticals and Revel Pharmaceuticals. R.L. serves on the scientific advisory boards of the following commercial entities: Quadriga, BeiGene, GigaGen, Teneobio, Nurix, Dragonfly, Apexigen, Viracta, Spotlight, Immunocore, Walking Fish, Kira, Abintus Bio, Khloris, and BiolineRx. C.R.B. is a cofounder and Scientific Advisory Board member of Redwood Bioscience (a subsidiary of Catalent), Enable Biosciences, OliLux Bio, Palleon Pharmaceuticals, InterVenn Bio, and Lycia Therapeutics. The other authors have no financial conflicts of interest.

We thank the Human Immune Monitoring Center at Stanford University for performing the Luminex assay.