Blockade of immune checkpoints in lymph nodes through locoregional delivery augments cancer immunotherapy

To evaluate whether augmenting the accumulation of administered mAb drug within target tissues can improve the effects of ICB, survival studies were performed in a poorly ICB-responsive tumor model (B16F10 melanoma) using both aCTLA-4 and aPD-1 mAbs and comparing intraperitoneal (i.p.) versus i.t. routes of administration. Modest reductions in tumor growth were induced by i.p. administration of ICB mAb; however i.t. administration instead resulted in profound reductions in tumor growth (Fig. 1A). To explore the potential effects on priming and expansion of T cells in response to endogenous tumor antigen, the systemic response of untreated tumors in the contralateral (c.l.) dorsal skin was concurrently monitored. Untreated tumors were found to be reduced in animals treated i.t. with ICB therapy (Fig. 1B). The net effect was a prolongation of mouse survival with i.t. compared to i.p. administration of ICB mAb (Fig. 1C). These data demonstrate the capacity of tumor-localized ICB therapy to elicit a systemically functional antitumor immune response that exceeds the effects of systemically administered ICB therapy.

The immunological mechanisms underlying the therapeutic responses seen with i.t. and i.p. ICB therapy (aCTLA-4 and aPD-1 in combination) were explored. T cell phenotypes in i.t. saline (control)– or ICB-treated animals bearing single B16F10 tumors were analyzed 12 days after tumor implantation. Administration i.t. led to a reduction in tumor burden (Fig. 2A) and was associated with a reduction in trT_reg frequencies (Fig. 2B), which can be attributed to the particular aCTLA-4 mAb clone used (4–6). ICB therapy was also associated with an increase in CD8 TILs, although this effect was limited to i.t. administration here (Fig. 2C). Of these CD8⁺ TILs, the frequencies of granzyme B–producing (Fig. 2D) and effector [killer cell lectin-like receptor G1–positive (KLRG1⁺)] cell frequencies (Fig. 2E) were similar between i.p. and i.t. administration, suggesting that effective therapy is associated with increased frequencies of CD8 TILs rather than reinvigoration of exhausted TILs. Cycling CD8 T cell frequencies were found to be increased in the TME and TdLN using i.t. administration, with comparable increases observed in the spleen between i.p. and i.t. administration, and with no changes observed in the non-TdLNs (nTdLNs) (Fig. 2, F and G). Similar results for Ki-67⁺ frequencies were also observed in the CD4 T cell compartment (fig. S1). Frequencies of CD8 T cells within each tissue compartment exhibiting stem-like (PD-1⁺Tcf1⁺Tim3⁻) versus effector-like (PD-1⁺Tcf1⁻Tim3⁺) CD8 T cell phenotypes were also assessed. The phenotypes of activated (PD-1⁺) CD8 T cells were predominately effector-like in the TME compared to stem-like in the LNs, with a balance in-between in the spleen regardless of therapy or route of administration (Fig. 2H). Of note, ex vivo staining confirmed that therapeutic aPD-1 mAb did not block the binding of aPD-1 mAb used for flow cytometry staining (fig. S2). These data support the concept that ICB efficacy is, in part, mediated by increasing the frequencies of TILs that may originate from the TME or peripheral tissues including the TdLN or spleen.

Considering that responses were observed in secondary lymphoid tissues in addition to the TME after i.t. administration, we assessed the effect of route of administration on mAb accumulation within the spleen, LNs (tumor-draining or nondraining), and TME as well as other systemic tissues. Alexa Fluor 647–labeled aPD-1 or aCTLA-4 mAbs (fig. S3A) were measured after a single dose using four different administration routes: i.p., in the forelimb skin contralateral to the tumor (c.l.), in the forelimb skin ipsilateral to the tumor (i.l.), and i.t. 5 days after B16F10 tumor implantation (Fig. 3A). Tumor accumulation of mAb was sustained over 24 hours using an i.t. injection but was low to negligible using other administration routes (Fig. 3, B and C). mAb concentrations in the blood and spleen were also equivalent between administration routes (Fig. 3, C and D). When assessing mAb accumulation in LNs, i.p. administration resulted in minimal accumulation in any measured LN, whereas c.l. administration led to accumulation within nTdLNs (Fig. 3, E and F). Using i.l. and i.t. administration led to detectable concentrations of mAb solely within TdLNs (Fig. 3, E and F). These locoregional administration routes allowed for reduced dosing while maintaining mAb accumulation within TdLNs (Fig. 3, G and H). Accumulation of mAb in dLNs was not an effect of dye labeling, because administered nonfluorescent mAb accumulated within dLNs as with Alexa Fluor 647–tagged mAb (fig. S3B). Using these four different routes of administration allowed for subsequent studies to explore the effects on drugging particular tissues of interest and their effects on ICB therapeutic efficacy.

To explore whether lymph-delivered mAb had access to LN T cells, aCD3 (in place of immune checkpoint targeting) mAb was administered in the forelimb to target LN-resident CD3-expressing T cells. A gradual increase in T cell labeling of aCD3 mAb was observed over 24 hours, with nearly 100% of T cells labeled with Alexa Fluor 647–aCD3. Because comparable total LN mAb concentrations measured in LN tissue homogenates were observed within LNs at all time points (Fig. 3I) and this measured LN mAb concentration was sufficient to label ~100% of T cells when LNs were mechanically and enzymatically degraded after resection (fig. S3C), this suggests that labeling of T cells by i.d. administered aCD3 mAb within intact LNs is a diffusion-limited, intra-LN transport process. These results are in line with a recently published study, confirming that mAb has access to LN cells (38). Overall, these results demonstrate that various administration routes can be used to direct the delivery of mAb to T cells in specific tissues, including the TME, LNs, spleen, and blood.

To elucidate the mechanistic effects of modulating immune checkpoints in various tissues on antitumor immunity, ICB mAbs were administered using the injection routes/locations depicted in Fig. 3A. This approach allowed the effects of ICB in specific tissues to be decoupled because i.t. administration results in appreciable mAb accumulation within the TME, TdLN, and spleen; i.d. forelimb injections target only the TdLN or nTdLN and spleen; and i.p. administration results in accumulation only in the spleen but not in the TME or in LNs. On days 5, 7, and 9 after B16F10 melanoma implantation, mice were treated with various ICB therapeutic regimens, including aPD-1, aCTLA-4, or the combination of the two. When used as monotherapies, aPD-1 and aCTLA-4 mAb administered i.p. and in the forelimb c.l. to the tumor had no effect on tumor growth and animal survival, whereas administration of ICB in the forelimb i.l. to the tumor and i.t. reduced tumor growth during treatment to equivalent extents, which in the case of aCTLA-4 monotherapy led to prolonged survival (Fig. 4, A to D). However, ICB therapy was less effective in larger-sized tumors (fig. S4). Similar effects were observed when aPD-1 was combined with aCTLA-4 (Fig. 4, E and F). Overall, these data suggest that targeting of TdLNs (in addition to the spleen) results in the generation of robust antitumor immunity much greater than that seen with systemic administration.

We next explored the effects of ICB therapy in combination with a model tumor vaccine. The rationale was to develop and expand a robust antitumor CD8 T cell pool in tumor-bearing animals before modulation of T cell activation and effector functions resulting from ICB. Mice bearing B16F10 melanomas expressing ovalbumin (OVA) were vaccinated i.d. in each limb with OVA protein as tumor antigen and CpG oligodeoxynucleotide (CpG) as an adjuvant 4 and 10 days after tumor implantation and before ICB mAb administration (aPD-1 and aCTLA-4 in combination) on days 5, 8, 11, and 14 (Fig. 5A). Irrespective of the route of mAb administration, ICB improved vaccine effects during treatment, as measured by tumor outgrowth over the first 16 days (Fig. 5B). After cessation of therapy, ICB administered in the skin (either i.t. or i.d.) conferred improved survival (Fig. 5C) relative to that of systemically administered ICB (i.p.). T cell phenotyping on day 16 after tumor implantation revealed that increased CD8/T_reg ratios and CD8 T cell infiltration correlated with smaller tumors (Fig. 5, D and E), whereas increased T_reg frequencies correlated with increased tumor size (Fig. 5F). ICB administered i.t. reduced proliferating T_regs (CD4⁺FoxP3⁺Ki-67⁺) within the tumor compared to other ICB administration methods (Fig. 5G), an effect attributable to the particular aCTLA-4 mAb isotype used (4–6) and higher CTLA-4 surface expression on T_regs (fig. S5). When mice were treated with ICB therapy, increased infiltration of CD8 TILs was observed (Fig. 5H). However, similar rates of CD8 TIL proliferation were observed regardless of therapy or route of administration (Fig. 5I). Instead, increases in proliferation were observed in lymphoid tissues, specifically the TdLN using an i.d. or i.t. administration and the spleen with the addition of ICB therapy (Fig. 5J). Similar to results exploring the cell state of PD-1⁺ CD8 T cells (Fig. 2H), the PD-1⁺ CD8 T cells within the TME were predominately effector-like cells, whereas lymphoid tissues consisted of both effector- and stem-like CD8 T cells (fig. S6, A to D). Furthermore, CD8 T cells generated in LNs and spleen with ICB therapy were functional and capable of responding to tumor antigen upon ex vivo restimulation (fig. S6, E and F). Similar patterns were observed in the CD4 helper compartment (fig. S7). Together, these results are in line with neoadjuvant studies demonstrating that improved responses are associated with increased CD8 T cell proliferation and tumor infiltration (Fig. 2), which were achieved via concurrent drug modulation of immune checkpoint pathways in the TME and LNs.

Considering the dose-toxicity relationship of ICB therapy, we assessed ICB efficacy in a dose de-escalation study using aPD-1 and aCTLA-4 (fig. S8). Dose-dependent effects were observed in the case of nTdLN (c.l.) and TdLN (i.l.) delivery, whereas i.t. administration did not display a dose-efficacy relationship (fig. S8). This may be explained by our observed reductions in trT_reg after following i.t. administration (Figs. 2B and 5G), which is in line with multiple other reports showing that the efficacy of aCTLA-4 therapy is due, in part, to the depletion of trT_regs when an IgG2a clone is used (4–6). However, T_reg depletion by aCTLA-4 has not been observed in the clinical setting, and instead, mAb clones of aCTLA-4 used in human patients have been shown to act predominately via CTLA-4 receptor blockade and favoring CD28 ligation (39, 40). We therefore investigated the therapeutic effects of CTLA-4 blockade using a mAb clone of an IgG1 isotype (4F10) using an identical dosing schedule to those conducted using the T_reg-depleting (IgG2a isotype) aCTLA-4 mAb clone (Fig. 4). Using this non–T_reg-depleting aCTLA-4 mAb clone (IgG1), i.p. and c.l. administration had small antitumor therapeutic effects (Fig. 6, A to D). Conversely, both i.l. and i.t. administration elicited robust antitumor therapeutic effects even at the lowest tested dose (12.5 μg; Fig. 6, E to H). These data thus demonstrate that the benefits of LN targeting are applicable to multiple aCTLA-4 mAb clones with differing immunomodulatory mechanisms. These results suggest that the efficacy of ICB directed to the TdLNs alone or in addition to the TME is roughly equivalent, at least at the doses tested in this model, pointing to LNs mediating the expansion of CD8 T cell immunity in response to ICB.

To extend these results beyond melanoma models, two different mammary carcinoma models, E0771 and 4T1, were implanted orthotopically in the fourth mammary fat pad. These implantation sites generate TdLNs differing from those generated using the dorsal lateral B16F10 melanoma model, specifically the ipsilateral inguinal (primary draining) and axillary (secondary draining) LNs. To deliver mAb to TdLNs or nTdLNs, mAb administration was performed in the flank skin of mice (Fig. 7A), whereas the systemic i.p. administration was kept the same and i.t. administration consisted of an injection into the mammary fat pad tumor site. After administration of fluorescently labeled mAb (aPD-1), i.t. administration resulted in sustained mAb retention in the TME, whereas other administration routes resulted in low to minimal TME concentrations (Fig. 7B). Accumulation of mAb in the spleen was equivalent regardless of administration route (Fig. 7C). Contrastingly, i.l. and i.t administration of mAb (aPD-1 or isotype) led to higher TdLN accumulation, whereas c.l. administration led to accumulation within nTdLN (Fig. 7D).

Using aPD-1 in the E0771 model, locoregional therapy was as effective as systemic (i.p.) administration (fig. S9, A and B), motivating the exploration of aPD-1 in combination with aCTLA-4. To this end, a single dose of aPD-1 in combination with aCTLA-4 (clone 9H10) administered i.t., i.l., and c.l. resulted in reductions in tumor growth compared to either no treatment or systemic i.p. administration (Fig. 7E). Improvements in survival were found for this combination therapy (40 to 60% overall complete response) and were comparable between all administration routes (Fig. 7F). Dose de-escalation studies demonstrated that survival was dose sensitive, but effects were again roughly equivalent between ICB mAb administration route (fig. S9, C to F). These data demonstrate that the therapeutic benefits of ICB with aPD-1 in this model can be improved with aCTLA-4, which is expected given this aCTLA-4 mAb clone’s pleotropic effects on the antitumor immune response.

To decouple the effects of trT_reg depletion and expansion of CD8 T cell immunity associated with aCTLA-4 mAb treatment, we evaluated the effects of the 4F10 mAb clone of aCTLA-4 that does not result in trT_reg depletion (Fig. 7, G and H). Antitumor therapeutic efficacy of all tested administration routes overall was much less effective, indicative of a major role that T_regs play in the immune physiology of the E0771 model. However, i.l. and i.t. administration did suppress tumor growth and prolong survival compared to i.p. and c.l. administration (Fig. 7, G and H). Effects of aPD-1 mAb in combination with aCTLA-4 mAb (clone 4F10) in the highly metastatic 4T1 model were also tested. Improved responses were observed with c.l., i.l., and i.t. therapy compared to that of systemic i.p. therapy (Fig. 7, I and J), with no signs of metastasis in responding mice (fig. S9, G and H). When the trT_reg-depleting aCTLA-4 clone (clone 9H10) was used, systemic therapy was just as effective as administration in the skin (both i.l. and c.l.; fig. S9, I and J). These data suggest that these breast tumor models are highly infiltrated with T_regs implicated in tumor progression. Nevertheless, as in the B16F10 melanoma model, (Td)LN targeting of mAb improved therapeutic responses to ICB in two breast tumor models compared to conventional systemic administration.

ICB is associated with immune-related adverse events (iRAEs) that can lead to discontinuation of treatment, especially when aPD-1 and aCTLA-4 mAb therapies are used in combination (41). To explore ICB-related toxicities, blood was collected 2 to 3 days after the last mAb administration, and the serum was analyzed. When used with ICB alone, i.p. administration led to increased alanine transaminase (ALT) serum concentrations compared to no treatment or cutaneous injections in both the B16F10 and E0771 tumor models (Fig. 8A). Similar effects of locoregional therapy on ALT serum concentrations were also observed when mice were vaccinated (Fig. 8B). Furthermore, mAb concentrations in the liver, kidneys, and lungs were proportional to the administered dose of ICB mAb (Fig. 8C and fig. S10). Overall, these results suggest that locoregional administration of ICB mAb, which can elicit robust immunity and antitumor efficacy, reduces the toxicity associated with systemic and high-dose ICB therapy.

We explored the effects of sustained LN drugging through the use of a hydrogel formulation to improve local pharmacokinetics and reduce the need for multiple injections. Formed from biocompatible, U.S. Food and Drug Administration (FDA)–approved block co-polymer Pluronic F-127, equivalent hydrogel formulations have been used to prolong injection site protein retention (42) to improve drug bioactivity. After injection of fluorescently labeled mAb, the hydrogel formulation prolonged mAb retention at the injection site to over 72 hours, leading to higher mAb concentrations in the dLN after injection compared to those resulting from administration of free (no polymer) mAb solution (Fig. 8, D and E). After a single injection of both aPD-1 and aCTLA-4 mAb on day 5 after tumor implantation in the B16F10 model, the hydrogel formulation injected in the forelimb i.l. to the tumor afforded improved antitumor efficacy, effects not seen when it was injected in the c.l. limb (Fig. 8, F and G). Furthermore, the hydrogel formulation improved efficacy relative to that of the free, unformulated mAb, suggesting that sustained mAb accumulation within TdLNs improves effects (Fig. 8, F and G). Overall, these results demonstrate the potential for controlled-release strategies directing mAb delivery into the lymphatic-draining tissue basin co-draining the tumor to improve the therapeutic effects of ICB.

This study was designed to explore the effects of targeting ICB mAb to LNs on antitumor efficacy in mouse melanoma and breast cancer models. We evaluated in what tissues immune checkpoint pathways were active and explored how ICB delivery to LNs differed from systemic therapy with ICB alone or in combination with vaccination. Sample sizes were chosen on the basis of previously published studies. For animal studies, mice were randomized into various groups before treatment, with each cage having one mouse per group. Experiments were not performed in a blinded fashion.

Cell lines were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin/amphotericin B and periodically checked for mycoplasma contamination. C57Bl6 and BalbC mice were purchased from The Jackson Laboratory. All protocols were approved by the Institutional Animal Care and Use Committee. Tumors were implanted intradermally in 6- to 12-week-old mice and monitored in anesthetized mice by caliper measurements of tumor width, length, and depth. Mice were euthanized when tumors ulcerated or reached 1.5 cm in any dimension.

The dorsal skin of C57Bl6 mice was shaved, and B16F10 or B16F10-OVA cells (10⁵) were implanted in the right dorsal flank on day 0. After 5 (when all tumors were visible), 7, and 9 days, mice were i.d. injected with 150, 50, or 12.5 μg of anti-mouse CTLA-4 (clone 9H10 or UC10-4F10-11; BioXCell) and/or rat anti-mouse PD-1 (clone RMP1-14; BioXCell) i.t., i.d. in the forelimb, or i.p. in 30 μl of saline. In abscopal tumor immunotherapy experiments, 10⁵ B16F10 cells were injected i.d. on the right dorsal skin of the mouse on day 0 and on the left dorsal skin on day 2. On days 5, 7, and 9, mice were injected with 150 μg of aCTLA-4 (clone 9H10) and aPD-1 (i.d. or i.p.) in saline. For immune cell phenotyping, mice were euthanized on day 12 and tissues were harvested. In vaccination studies, at 4 and 10 days, CpG (3 μg) and OVA (10 μg) were i.d. administered in 30 μl of saline in each limb. On days 5, 8, 11, and 14, mice received 150 μg of aCTLA-4 and aPD-1 mAb in 30 μl of saline either i.t., i.d. in the forelimb, or i.p. In studies evaluating the effects of sustained mAb release, Pluronic F127 (Sigma-Aldrich) was dissolved at 25 weight % in cold phosphate-buffered saline (PBS). Before injection, 25 μg of aCTLA-4 (clone 9H10) and aPD-1 mAb (5 μl) was mixed with 25 μl of the gel solution or PBS and i.d. injected once into one forelimb on day 5.

On day 5 after tumor implantation, mice were administered aCTLA-4 or aPD-1 mAb. Fluorescent imaging was performed with an IVIS Spectrum instrument (PerkinElmer) at the injection site over 24 hours. Twenty-four hours after mAb injection, mice were euthanized and tissues were collected for imaging and homogenization. Concentrations of mAb in homogenized tissues were determined using a standard curve of injected mAb solution in naïve tissue homogenates. Tissue background was subtracted from all measurements. For biodistribution experiments using aCD3 mAb, naïve mice were injected with 6.25 μg of mAb and euthanized after 1, 5, or 24 hours. LNs were either collagenase-treated for 30 min followed by tissue disruption to form single-cell suspensions (a process described in Supplementary Materials and Methods) or immediately cut up and dispersed into a single-cell suspension to prevent ex vivo T cell labeling.

E0771 (5 × 10⁵) or 4T1 (3.5 × 10⁵) cells resuspended in 30 μl of saline were implanted i.d. in the left mammary fat pad (fourth) in C57Bl6 or BalbC mice, respectively. For E0771 experiments, aPD-1 or aPD-1 and aCTLA-4 (clone 9H10) in combination were administered i.d. when tumors were ~100 mm³. Alternatively, 30 μg each of aPD-1 and aCTLA-4 (clone 4F10) mAb were administered on days 10, 14, and 17. For 4T1 experiments, 50 μg each of aPD-1 and aCTLA-4 (clone 4F10) mAb were administered on day 7 or 50 μg each of aPD-1 and aCTLA-4 (clone 9H10) mAb were administered on days 7 and 10. At end point, LNs and the spleen were harvested and imaged and organ sizes were measured using ImageJ.

Statistical significance of differences between experimental groups was calculated with Prism software (GraphPad). All data are expressed as means ± SD except for tumor growth (SEM). ****P < 0.0001, ***P < 0.001, **P < 0.01, and *P < 0.05 by unpaired two-tailed t tests or one- or two-way analysis of variance (ANOVA) followed by Tukey post hoc test for multiple comparisons. For survival curves, log-rank (Mantel-Cox) test was performed. Original data are provided in data file S1.