Tolerogenic nanoparticles restore the antitumor activity of recombinant immunotoxins by mitigating immunogenicity

Significance

Protein-based drugs are very active in treating cancer, but their efficacy is limited by the formation of neutralizing antidrug antibodies (ADAs). Recombinant immunotoxins are proteins that are very effective in patients with leukemia, in whom immunity is suppressed, but induce ADAs, which compromise their activity, in patients with intact immunity. Here we used an immunomodulator that is encapsulated in a nanoparticle delivery system (SVP-R) to induce specific immune tolerance to immunotoxins in mice. SVP-R induces immune tolerance, prevents ADA formation, and prevents the drug neutralization and clearance that results in restoration of its antitumor activity. Importantly, the combination is also efficacious in mice with preexisting antibodies, indicating that this approach can benefit patients who often have such antibodies.

Abstract

Protein-based drugs are very active in treating cancer, but their efficacy can be limited by the formation of neutralizing antidrug antibodies (ADAs). Recombinant immunotoxins are proteins that are very effective in patients with leukemia, where immunity is suppressed, but induce ADAs, which compromise their activity, in patients with intact immunity. Here we induced a specific, durable, and transferable immune tolerance to recombinant immunotoxins by combining them with nanoparticles containing rapamycin (SVP-R). SVP-R mitigated the formation of inhibitory ADAs in naïve and sensitized mice, resulting in restoration of antitumor activity. The immune tolerance is mediated by colocalization of the SVP-R and immunotoxin to dendritic cells and macrophages in the spleen and is abrogated by depletion of regulatory T cells. Tolerance induced by SVPs was not blocked by checkpoint inhibitors or costimulatory agonist monoclonal antibodies that by themselves enhance ADA formation.

Protein- and cell-based therapies have shown great potential in treating various cancers. However, the efficacy of these biological therapies is often mitigated by elicited immune responses to the actual therapy. Repeated administration of immunogenic anticancer drugs, such as chimeric antigen receptor T cells (1), enzyme therapy (2), monoclonal antibodies (mAbs) (3), antibody drug conjugates (ADCs), recombinant immunotoxins (4) and viral-based gene therapy vectors (5), lead to the formation of antidrug antibodies (ADAs), resulting in drug neutralization, accelerated clearance, or infusion-related reactions and other serious adverse events (6).

Recombinant immunotoxins (RITs) are chimeric proteins active in the treatment of several types of cancer (7, 8). RITs consist of a portion of an antibody linked to a protein toxin, such as Pseudomonas exotoxin A (PE38). PE38 is a very potent toxin, but because of its foreignness to the human immune system, it induces the formation of ADAs that inactivate the RIT (9). The use of RITs in patients whose immune systems are suppressed by cancer or by chemotherapy has produced complete regressions and prolonged survival of patients with chemoresistant hairy cell leukemia (10, 11). In contrast, when PE38 is targeted to solid tumors, immune competent patients developed ADAs against the immunotoxin (12, 13). The ADAs neutralize the RIT, dramatically accelerate its clearance, and prevent further treatment. With the coadministration of systemic immunosuppressive drugs, a mesothelin-targeted RIT (SS1P) enabled 2 of 10 patients being treated for mesothelioma to receive more cycles of therapy, resulting in profound antitumor responses and prolonged survival (14). Therefore, RIT has the potential to be a transformative therapy for chemotherapy-refractory mesothelioma and other solid tumors if ADAs can be mitigated more broadly.

LMB-100 is a second-generation RIT that has a humanized Fab targeting mesothelin fused to a modified PE38 toxin (Fig. 1A) (15). LMB-100 was engineered to reduce immunogenicity in humans compared with its parental counterpart, SS1P. The changes included humanization of the antibody moiety, deletion of domain II of PE38 that contains dominant T cell epitopes, and introduction of seven point mutations in domain III to silence B cell epitopes (15). LMB-100 has shown excellent antitumor activity in animal models (16) and is now being evaluated in clinical trials for the treatment of mesothelioma and pancreatic cancer (https://clinicaltrials.gov/ct2/show/NCT02810418). Despite the deimmunized toxin fragment, almost all patients develop ADAs after two treatment cycles.

Fig. 1.

The combination of LMB-100 + SVP-R prevents the ADA response against LMB-100. (A) A ribbon diagram of LMB-100 and an illustration of SVP-R. (B) Mice were injected seven times every other week with LMB-100 or a combination of LMB-100 and SVP-R one, three, or seven times (indicated by arrows). Anti–LMB-100 antibodies were evaluated by ELISA (n = 8). (C) Mice were injected with LMB-100 and SVP-R as indicated by arrows (n = 7). (D) Mice were injected with LMB-100 and SVP-R as indicated by arrows. Final mean titer on week 10 is shown (n = 7). (E) Neutralization assay using plasma from the mice treated as shown in C (n = 7). KLM-1 cells were seeded and treated with plasma-LMB-100 mixture. Cell viability was assessed after 72 h. Curves represent mean of seven viability curves (n = 7, six replicas per samples). (F) Mice were injected with LMB-100 and SVP-R as indicated by arrows (n = 8). ELISA plates were coated with LMB-100, Fab, or anti–TAC-PE24. Plasma samples from week 6 were evaluated. The dilution factor for 50% of binding is shown. Lines indicate mean; error bars, SEM. For statistical analysis in B and C, the AUC for each curve was calculated, and AUCs were compared using one-way ANOVA.

Rapamycin is an mTOR inhibitor that has both immunomodulatory properties and antitumor activity (17). Previous attempts to combine rapamycin with RIT were unsuccessful, because the doses required to suppress the immune response against RIT were toxic to the mice (18). It was recently reported that encapsulated rapamycin in synthetic vaccine nanoparticles (SVP-R) prevented immune responses in mice against a variety of biological drugs, including adalimumab, PEGylated uricase, and coagulation factor VIII (19, 20). While the exact mechanism for the immunogenicity reduction is not clear, it has been shown that injection of SVP results in its accumulation in the liver, spleen, and draining lymph nodes (21). Nanoparticles are preferentially phagocytosed by antigen-presenting cells due to their size, charge, and shape (22). It has been speculated that the selective delivery of rapamycin to these immune organs generates a tolerogenic milieu that can induce specific immune tolerance to coadministered antigens (18).

Here we report preclinical studies supporting the use of SVP-R in patients to be treated with RITs. We demonstrate that SVP-R induces a long-lasting, specific, transferable, and Treg-dependent immune suppression and tolerance that prevents ADA formation against LMB-100 in naïve mice and in mice with preexisting antibodies. This immune tolerance promotes the antitumor activity in immunocompetent mice that otherwise would be neutralized by the ADAs. Furthermore, the combination of LMB-100 and SVP-R induces favorable cytotoxic activity in human mesothelioma and pancreatic cancer cell lines, more potent than that induced by either agent alone. Finally, the immune tolerance persists when combined with anti–CTLA-4 checkpoint inhibitor or anti–OX-40 costimulatory agonist mAbs.

Results

Combination of LMB-100 with SVP-R Prevents ADA Response.

To evaluate the effect of SVP-R on the ADA response to LMB-100 (Fig. 1A), BALB/c mice were injected every other week with LMB-100 or a combination of LMB-100 + SVP-R. LMB-100 has mutations that diminish human, but not mouse, responses. Mice injected with LMB-100 had a strong and rapid response to LMB-100 (Fig. 1B), with a mean titer of 10,975 ± 2,372 at week 14, indicating that LMB-100 is immunogenic in BALB/c mice. All mice injected with LMB-100 + SVP-R had an undetectable titer during the entire course of the experiment, indicating effective prevention of ADA formation. Furthermore, mice injected seven times with LMB-100 and given SVP-R with only the first three injections had a mean titer of only 371 ± 301 at week 14, indicating induction of immune tolerance. This titer was significantly lower than that of control mice treated with LMB-100 alone at both week 8 (P = 0.03) after only four doses and at week 14 (P = 0.0006) after seven doses. The area under the curve (AUC) for each mouse throughout the experiment, calculated to compare the ADA responses (Fig. S1A), demonstrated a significant decrease in the mice given three doses (P = 0.001) or seven doses of SVP-R (P = 0.002). The mice tolerated treatment well, with no significant weight loss (Fig. S1B).

Timing of SVP-R Immunization Is Critical for Immune Tolerance.

To determine the efficacy of SVP-R with an LMB-100 regimen similar to that used in human patients, mice were treated with successive cycles of LMB-100. Each cycle consisted of three doses per week every other day, and mice were injected with SVP-R once, twice, or three times during the first and second cycles (Fig. 1C). We found that a single dose of SVP-R per cycle was as effective as three doses in preventing ADA formation (P = 0.003). The median titer in mice receiving LMB-100 alone was 47,926, compared with only 881, 1,958, and 993 in mice immunized with LMB-100 + SVP-R given two, four, or six times, respectively, over the two treatment cycles. The ADA suppression was also maintained when mice were challenged with three additional cycles of LMB-100 in the absence of further SVP-R treatment. Six doses of LMB-100 + SVP-R were well tolerated by the mice, with no significant weight loss (Fig. S1C).

We evaluated the effect of timing of SVP-R treatment by staggering the day of SVP-R injection. LMB-100 was injected on days 1, 3, and 5 of each of five cycles, with SPV-R coadministered on day 1, day 3, days 1 and 3, days 3 and 5, or days 1, 3, and 5 of each cycle (Fig. 1D). Control mice treated with LMB-100 showed a mean titer of 44,132 at the end of five treatment cycles. In contrast, mice that received SVP-R on day 1 showed significant decreases in ADA formation regardless of whether they received one, two, or three SVP-R doses during each cycle, with mean titers of 1,413 ± 495 (P = 0.0007), 2,952 ± 1,320 (P = 0.001), and 1,979 ± 807 (P = 0.0007), respectively. Mice that received SVP-R on day 3 or days 3 and 5 had final titers of 29,341 ± 11,705 and 41,934 ± 9,725, respectively, indicating that cotreatment with SVP-R on the first day of each cycle is critical to prevent ADA formation.

SVP-R was also evaluated with the more immunogenic precursor of LMB-100, SS1P. Mice were injected with three doses of SS1P at weeks 1, 3, and 7 (Fig. S2), and SVP-R was given at week 1. Three cycles of SS1P induced a mean ADA titer of 37,734 ± 21,748, and a single cycle of SVP-R completely blocked these ADAs (P = 0.0001).

ADA Response Is Neutralizing and Targets both the Fab and Toxin.

To determine whether ADAs can neutralize the immunotoxin, we performed a functional in vitro neutralization assay using plasma samples from mice injected with LMB-100 (15 doses), LMB-100 (15 doses) + SVP-R (six doses), or vehicle. Plasma samples were mixed with various concentrations of LMB-100 and added to KLM-1 human pancreatic cells. The cells were very sensitive to LMB-100 with an IC₅₀ of 1.1 ng/mL (Fig. 1E). Plasma from mice immunized with LMB-100 alone inhibited the activity of LMB-100 and shifted the IC₅₀ to 93.2 ng/mL (P < 0.0001), indicating that the ADAs are neutralizing. In contrast, incubation of LMB-100 with plasma LMB-100 + SVP-R showed an IC₅₀ 50-fold lower (P < 0.0001) and not significantly different from the IC₅₀ of LMB-100 incubated with plasma from vehicle-treated mice (Fig. S3A). We observed a strong correlation between the anti–LMB-100 titers and IC₅₀ (R² = 0.96) (Fig. S3B).

To determine whether the ADAs against LMB-100 target the Fab, the toxin fragment, or both, we assayed the plasma from mice injected with five doses per week of LMB-100 alone or in combination with SVP-R (n = 8) on plates coated with LMB-100, a human Fab, or an immunotoxin containing the same domain III of the Exotoxin A (PE24) as found in LMB-100, fused to a mouse Fv (anti–TacFv-PE24). Anti–LMB-100 plasma reacted strongly with the human Fab and less strongly with the deimmunized toxin, which had some of its murine B cell epitopes removed (23) (Fig. 1F). As expected, SVP-R reduced the response to both components.

Combination of LMB-100 with SVP-R Induces a Specific and Transferable Immune Tolerance.

To determine whether the suppression of ADA formation is a result of specific immune tolerance rather than a chronic immune suppression, mice were immunized with eight weekly injections of LMB-100 and three doses of SVP-R (i.v.) at weeks 1, 2, and 3. At week 4, mice were challenged with four weekly injections of ovalbumin and LMB-100 (S.C.) (Fig. 2A). The combination of LMB-100 + SVP-R selectively inhibited ADA formation against LMB-100, but did not affect the antibody response to ovalbumin, resulting in similar anti-ovalbumin titers of 4,362 and 4,024. These results indicate that the combination of LMB-100 + SVP-R induces a specific immune tolerance that does not suppress the ability of the mice to mount an immune response against another antigen administered later.

Fig. 2.

Combination of LMB-100 with SVP-R induces a specific, transferable, and regulatory T-cell–mediated immune tolerance. (A) Mice were injected three times weekly with LMB-100 (2.5 mg/kg i.v.) or a combination of LMB-100 with SVP-R (2.5 mg/kg i.v.). During weeks 4–8, mice were challenged with weekly doses of LMB-100 (i.v.) and ovalbumin (s.c.). Plasma was collected and analyzed for anti–LMB-100 and anti-ovalbumin antibodies by ELISA. For statistical analysis, AUC for each curve was calculated and analyzed using the Mann–Whitney U test. Error bars SEM, n = 13. (B) Mice were injected six times with vehicle, LMB-100, SVP-R, or LMB-100 + SVP-R. During week 4, splenocytes from donor mice were isolated and adoptively transferred to recipient naïve mice. Recipient mice were injected with LMB-100 six times. Plasma was collected and analyzed for anti–LMB-100 antibodies by ELISA. Results from two separate experiments with identical schedules were combined. n = 5–10. Error bars represent SEM. (C) Mice were injected with LMB-100 on days 1, 3, 5, 29, 31, 33, 43, 45, and 47. SVP-R was given on days 1, 3, and 5. Anti-mouse CD-25–depleting antibody (PC61) or isotype control were injected i.p. on days 15 and 16. Titers on day 55 are shown. (D) Plasma from mice that were injected seven times every other week with LMB-100 or a combination of LMB-100 + SVP-R. Anti–LMB-100 isotypes were analyzed using sandwich ELISA with subclasses IgG1, IgG2a, IgG2b, IgG3, and IgM specific to LMB-100 (n = 8).

To test whether the immune tolerance could be transferred from tolerant mice to naïve mice, donor mice were treated with LMB-100, SVP-R, or LMB-100 + SVP-R for two cycles. Mice immunized with LMB-100 alone showed a mean titer of 4,521 ± 1,994, compared with 51 ± 25 in mice treated with LMB-100 + SVP-R (Fig. S4). Splenocytes were isolated, pooled, and transferred to naïve recipient mice (Fig. 2B). At 1 wk after cell injection, all recipient mice were challenged with two cycles of LMB-100. Adoptive transfer of cells from mice immunized with LMB-100 followed by an LMB-100 challenge of recipient mice induced a mean titer of 4,884 ± 1,548, not significantly different from the titer in mice receiving cells from vehicle-treated mice or no cells (mean titers of 4,571 ± 1,494 and 6,541 ± 3,079, respectively). Here, adoptive transfer did not induce substantial immune memory. Because these three mice groups had similar mean titers, these mice are referred to as controls.

In contrast, adoptive transfer of 10 × 10⁶ splenocytes from mice immunized with a combination of LMB-100 + SVP-R decreased the titers by 78–85% compared with the controls (P = 0.007, 0.003, and 0.02, respectively). Adoptive transfer of 2.5 × 10⁶ splenocytes reduced titers by 44–61%, but the difference was not statistically significant (P = 0.5). The mean titer of mice that received splenocytes from SVP-R–treated mice was not different from that of control mice, indicating that tolerance induction requires both LMB-100 and SVP-R in the donor mice and is not due to a general immune suppression.

Depletion of Treg Cells.

To study the role of Treg cells in SVP-R–induced immune tolerance, we depleted Treg cells in vivo after SVP-R tolerance induction. Mice were injected with LMB-100 or LMB-100 + SVP-R three times. On days 15 and 16, Treg cells were depleted using an anti-CD25 (PC61)-depleting antibody (24), and then challenged with two more cycles of LMB-100 (Fig. 2C). The depletion of Tregs abrogated the tolerogenic effect of SVP-R, increasing the mean titer from 416 ± 157 to 1,094 ± 304 (P = 0.04). This mean titer of 1,094 was similar to that in mice that did not receive SVP-R (1,348 ± 399).

Ig Subclasses.

To study the effect of SVP-R on class switching, plasma samples were characterized for LMB-100–specific IgG and IgM antibodies (Fig. 2D). Immunization with LMB-100 induced ADAs distributed across all IgG subclasses, with IgG1 the most dominant. This subclass distribution is similar to the IgG subclass distribution previously described after immunization with the parent immunotoxin SS1P (25). Immunization with LMB-100 + SVP-R induced an undetectable signal of LMB-100–specific IgG1, IgG2a, IgG2b, or IgG3 antibodies. Interestingly, the levels of anti–LMB-100 IgM antibodies were similar to the levels in mice immunized with LMB-100 alone. These results indicate that SVP-R prevents isotype switching, but does not prevent IgM production.

LMB-100 + SVP-R Colocalize Preferentially on Dendritic Cells and Macrophages.

We hypothesize that specific immune tolerance requires the presence of the antigen in the same cells as the target cells of SVP-R. To determine the fate of SVP-R and LMB-100 in the spleen, after injection in vivo, we injected Alexa Fluor 488-labeled LMB-100– and Cy5-labeled SVP-R consecutively and isolated the splenocytes at 2 h postinjection (Fig. 3A). Cell phenotype was analyzed using cellular markers according to the gating strategy shown in Fig. S5. We compared the uptake of LMB-100 and SVP-R in macrophages, dendritic cells (DCs), CD4⁺ T cells, CD8⁺ T cells, B cells, neutrophils, and monocytes (Fig. 3 B–D). We found that macrophages and DCs had the highest uptake of both LMB-100 and SVP-R; 38% of the macrophages and 13% of the DCs were positive for LMB-100, and 29% of the macrophages and 11% of the DCs were positive for SVP-R. Interestingly, 22% of the macrophages and 9% of the DCs stained positive for both. This colocalization occurred even though the two agents were injected separately. Relative cell numbers were not changed (Table S1).

Fig. 3.

LMB-100 and SVP-R colocalize preferentially on DCs and macrophages. (A) Dye-conjugated SVP-Cy5 and LMB-100-Alexa Fluor 488 were injected i.v. alone or in combination (n = 3–4 mice per group). Spleen cells were analyzed by FACS at 2 h after injection for dye-conjugate uptake. (B and C) Representative FACS plots show gating for macrophages (F4/80⁺CD11b⁺) and DCs (CD11c⁺MHC-II⁺), and in vivo uptake by the gated populations. Red quadrants indicate the percent of positive cells analyzed for each experimental condition. (D) Summary of SVP-R and LMB-100 in vivo uptake by macrophages, DCs, monocytes, CD4⁺ T cells, B cells, neutrophils and CD8⁺ T cells. The gating strategy for all cells is shown in Fig. S5.

Monocytic cells expressing CD11b^high, Ly6C⁺, and Ly6G⁻ have been involved in immunosuppressive activity (26, 27). We found that 3% of these cells demonstrated uptake of both LMB-100 and SVP-R. Finally, lymphocytes and neutrophils displayed the lowest percentages of colocalization (Fig. 3D). Taken together, these results suggest that preferential uptake of SVP-R and LMB-100 by professional antigen-presenting cells might mediate the immune tolerance.

The Combination of LMB-100 + SVP-R Prevents ADA Response in Mice with Preexisting Antibodies.

To determine whether SVP-R could therapeutically reduce immunogenicity and induce immune tolerance in mice with preexisting ADAs, mice were immunized six times with LMB-100 during weeks 1 and 3 to induce preexisting ADAs. At week 9, mice had a mean titer of 741 ± 66 and were divided into three groups with similar mean titers. At week 10, the groups were immunized with vehicle (PBS), LMB-100, or LMB-100 + SVP-R (Fig. 4A). Titers were then evaluated at week 12. Challenge with LMB-100 alone induced a strong memory immune response, resulting in a mean ADA titer of 9,808 ± 3,608. In contrast, challenge with LMB-100 + SVP-R not only prevented the antibody increase, but also decreased the titer (257 ± 121) compared with the preboost titer (738 ± 320; P = 0.003) and compared with mice injected with PBS at week 12 (502 ± 143; P = 0.002). This response was observed in three additional experiments with groups of eight, eight, and four mice.

Fig. 4.

Combination of LMB-100 with SVP-R induces immune tolerance in mice with preexisting antibodies. (A) Female BALB/c mice were injected six times with LMB-100 (2.5 mg/kg i.v.) on weeks 1 and 3 to induce a titer of ADA against LMB-100. On week 10, mice were challenged with three doses of LMB-100, vehicle (PBS), or LMB-100 + SVP-R. The mice treated with LMB-100 + SVP-R were challenged with three additional doses of LMB-100 on week 12. Plasma was collected and analyzed for LMB-100 ADAs by ELISA. n = 7 or 12. Error bars represent SEM. (B) Female BALB/c mice were injected 12 times with LMB-100 over the course of 14 wk to induce a high titer of ADA against LMB-100. In week 15, mice were immunized with LMB-100 or LMB-100 + SVP-R. ADA titers prechallenge and postchallenge are shown. (C and D) BM and spleen were isolated from mice with preexisting ADA and were challenged with PBS, LMB-100, SVP-R, or LMB-100 + SVP-R. BM cells and splenocytes (100,000 cells/well) were seeded in ELISpot plates precoated with LMB-100 (n = 8).

To evaluate whether SVP-R can induce a lasting immune tolerance that can prevent a response to later challenges in mice with preexisting antibodies, the mice were challenged with three additional doses of LMB-100 (no SVP-R) during week 13 (Fig. 4A). Titer evaluation at week 14 showed that administration of LMB-100 + SVP-R during week 10 maintained a low titer of 634 ± 269, which was significantly lower than the titer of mice treated with LMB-100 alone (11,505 ± 4,172; P = 0.0001). This indicates that the LMB-100 + SVP-R combination on week 10 induced an immune tolerance that prevented the response to a later LMB-100 challenge.

We next evaluated whether SVP-R could also be used to therapeutically reduce high titers of preexisting antibodies. Control mice from Fig. 4A that had an anti–LMB-100 antibody titer >10,000 induced by 12 doses of LMB-100 over the course of 14 wk were injected with LMB-100 or LMB-100 + SVP-R (Fig. 4B). Mice treated with the combination had a significant decrease in titer, from 31,114 ± 13,730 to 7,797 ± 4,558 (P = 0.02).

To determine whether treatment of mice with preexisting antibodies with the combination affected the number of antibody secreting plasma cells in the bone marrow (BM), we treated mice with preexisting antibodies with PBS, LMB-100, SVP-R, or LMB-100 + SVP-R. Cells were collected from BM and spleen at 24 h after injection and then assayed for the number of cells making anti–LMB-100 antibodies by ELISpot (Fig. 4 C and D). The mice had similar numbers of antibody-secreting cells (mean, 9.6 ± 6.7 SFC/1E6 cells) in BM and no detectable spots in spleens. These results indicate that SVP-R does not affect antibody-secreting plasma cells residing in the BM.

The Combination of LMB-100 + SVP-R Restores Antitumor Activity of LMB-100 in Mice with Preexisting Anti–LMB-100 Antibodies.

To study the activity of LMB-100 and SVP-R in immunocompetent tumor-bearing mice, the AB-1 mouse mesothelioma cell line (28) was stably transfected with human mesothelin (AB1-L9; Fig. S6). AB1-L9 cells inoculated into BALB/c mice grew rapidly, reaching 600 mm³ in 15 d (Fig. 5A). To evaluate antitumor activity, tumor-bearing mice were therapeutically treated six times with LMB-100, SVP-R, or LMB-100 + SVP-R, when the tumors reached a mean size of 199 mm³. Mice treated with LMB-100 showed significant tumor growth inhibition (P = 0.003 for AUC of tumor growth curves compared with PBS-treated mice), with 1 of 7 mice achieving complete remission. Mice treated with SVP-R showed only a minor tumor growth delay (P = 0.05). LMB-100 + SVP-R induced the most significant tumor growth inhibition (P = 0.0003), resulting in a 13-fold decrease in tumor size on day 20. Due to the relatively short immunization schedule, all mice had either very low or undetectable titers when evaluated on day 18 of the experiment (Fig. S7), so no significant in vivo neutralization of LMB-100 was observed.

Fig. 5.

Combinations of SVP-R and LMB-100 restores neutralized antitumor activity. (A) AB1-L9 cells were inoculated into mice and treated with PBS, LMB-100, or SVP-R as indicated by arrows (n = 7). (B) Mice were immunized with LMB-100 four times to induce a baseline titer and then inoculated with AB1-L9. Mice were treated with vehicle, LMB-100, or SVP-R as indicated by arrows. Tumor size was measured using a caliper (n = 7). (C) Plasma from days 5 and 19 was analyzed for anti–LMB-100 antibodies by ELISA. Titer was interpolated at 10% of the signal. (D) Mice were treated as described in C. The experiment was terminated on day 31. The Kaplan–Meier plot shows the time to the experimental endpoint (once tumor volume was >400 mm³ or if a mouse lost >30% of its body weight). n = 7. (E) Mice were inoculated with CT26 cells on day 1 and treated with SVP-R or vehicle on days 10 and 16. Values indicate average tumor size (n = 7). Error bars represent SEM. (F) Mice were inoculated with 66C14 cells on day 1 and treated with SVP-R or vehicle on days 10, 12, and 14. Values indicate average tumor size (n = 5). For statistical analysis, AUC for each curve was calculated, and AUCs were compared using one-way ANOVA. Error bars represent SEM.

To study the activity of LMB-100 and SVP-R in mice with preexisting antibodies, mice were first immunized with LMB-100 four times to induce an average baseline titer of 2,597 ± 2,080 before inoculation with AB1-L9. At 5 d after tumor inoculation, when the tumors reached a mean of 135 mm³, mice were treated with two cycles of three injections with LMB-100 or vehicle (Fig. 5B) with or without SVP-R administered on the first day of each cycle (every other week). We found that the tumors treated with LMB-100 alone did not respond to treatment and had a similar growth rate as PBS-treated tumors. We attribute the lack of response to LMB-100 to the high ADA titer (Fig. 5C), which neutralized the activity of LMB-100.

In contrast, mice treated with LMB-100 + SVP-R had an excellent response to LMB-100 and did not develop high ADA titers. Mice treated with LMB-100 + SVP-R had a higher survival rate (i.e., time to reach 600 mm³; P = 0.0001) (Fig. 5D). These experiments were repeated two more times using seven mice per group with similar results. However, mice treated with LMB-100 + SVP-R had decreased weight, perhaps due to increased exposure to LMB-100 as a result of preventing neutralizing ADAs (Fig. S8).

SVP-R Does Not Accelerate Tumor Growth Rate.

To test whether treating mice with SVP-R interferes with tumor immunity and/or enhances tumor growth, we inoculated the CT26 (murine colon carcinoma) and 66C14 (murine breast cancer) cell lines in the flanks of immune competent BALB/c mice and compared the growth rates in SVP-R–treated mice and PBS-treated mice (Fig. 5 E and F). SVP-R delayed the growth of CT26 tumors and showed no change in tumor growth in 66C14 tumors.

SVP-R Enhances the Cytotoxic Activity of LMB-100 in Human Cell Lines.

Because rapamycin has also been reported to have antitumor activity, we measured the cytotoxic activity of SVP-R on human mesothelioma cells (HAY) and human pancreatic cells (KLM-1) in vitro. We found that SVP-R had modest cytotoxic activity by itself (Fig. 6A) in both cell lines. However, when combined with LMB-100, 5 μg/mL of SVP-R improved the cytotoxic activity of LMB-100, shifting the IC₅₀ on KLM-1 cells from 1.1 ng/mL to 0.1 ng/mL (Fig. 6B), and on HAY cells. 1 μg/mL of SVP-R improved the IC₅₀ from 2.9 ng/mL to 0.9 ng/mL (Fig. 6C). HAY cell viability was also evaluated by staining with crystal violet after a 72-h incubation with SVP-R (2 μg/mL) and LMB-100 (0.4 ng/mL), followed by incubation for 72 h with no drug (Fig. 6D). The combination was more effective than either drug alone in killing cells.

Fig. 6.

SVP-R enhances the cytotoxic activity of LMB-100 in human cell lines. KLM-1 and HAY cells were seeded in 96-well plates and treated with various concentrations of SVP-R, LMB-100, or LMB-100 + SVP-R. After 72 h, cell viability was assessed using WST-8 or crystal violet. Viability curves were fitted to each sample and IC₅₀ was calculated. (A) Cytotoxic activity of SVP-R in both cell lines. (B) Activity of LMB-100 in KLM-1 cells with or without 5 μg/mL of rapamycin encapsulated in SVP. (C) Activity of LMB-100 in HAY cells with or without 1 μg/mL of rapamycin encapsulated in SVP. Curves show a mean of six replicates. Error bars represent SEM. (D) Representative well images taken after HAY cells were fixed and stained with crystal violet.

SVP-R Activity Is Not Diminished by Checkpoint Inhibitors or Costimulatory Agonists.

We next investigated whether anti–CTLA-4 antagonist antibody and anti–OX-40 agonist antibody can enhance the formation of ADAs against LMB-100, and whether such ADAs could be blocked by SVP-R. Mice (n = 8) were injected with five weekly doses of LMB-100, along with an anti-mouse CTLA-4 antibody or an anti–OX-40 antibody given on the fifth day of every week (Fig. 7). We found that both antibodies substantially enhanced the formation of anti–LMB-100 ADA titers compared with treatment with LMB-100 alone (P = 0.001 for anti-CTLA-4 and P = 0.02 for anti-OX-40). Injection of SVP-R on the same days as LMB-100 resulted in either elimination (i.e., mean titer below the limit of detection) in the mice treated with anti–CTLA-4 or a dramatic 12-fold decrease in titer in the mice treated with anti–OX-40. SVP-R activity was not compromised by the activity of the immune checkpoint inhibitors or costimulatory agonists. These experiments were repeated two more times with n = 5 and n = 3, with similar results.

Fig. 7.

SVP-R activity is not diminished by checkpoint inhibitor antibodies. BALB/c mice were immunized weekly with LMB-100 or LMB-100 + SVP-R five times (2.5 mg/kg i.v.), and at 5 d after each injection were immunized with anti-mouse CTLA-4 antagonist (A) or anti–OX-40 antagonist (B) or vehicle (i.p.). Plasma samples were collected on day 6 of each week, and LMB-100 ADA titer was evaluated using direct ELISA. n = 8. Error bars represent SEM. The experiments were repeated with n = 3 and 5, with similar results.

Discussion

The immune response to RITs is a major factor limiting their efficacy against solid tumors in cancer patients with intact immune systems. In the present study, we have established the preclinical basis of specific immune tolerance for RITs using rapamycin encapsulated in SVP-R. These nanoparticles are composed of a biodegradable poly (lactic acid) core with a corona of surface PEGylation. We demonstrate that SVP-R produces a long-lasting, specific, and transferable immune tolerance that prevents ADA formation against LMB-100 in naïve mice and reduces ADAs in mice with preexisting antibodies. Induction of immune tolerance to LMB-100 resulted in restoration of its antitumor activity in a syngeneic mesothelioma tumor model in immunocompetent mice that otherwise would be neutralized by ADAs.

Immune Suppression vs. Tolerance.

Previous studies have evaluated several immune suppression approaches to reducing the immunogenicity of RITs in patients. These approaches include B cell depletion using rituximab, which was ineffective in preventing anti-immunotoxin immune response in patients (29), and B and T cell suppression using a combination of cyclophosphamide and pentostatin (14). The success of this approach was limited by the toxicity of the immunosuppressive agents, and while some patients exhibited a dramatic delay in ADA formation, most patients developed strong ADA responses that halted treatment.

Immune Tolerance Mechanism.

Unlike general immune suppressive therapies, SVP-R has a strong affinity to accumulate in the lymph nodes and spleen (21), where it is selectively endocytosed by antigen-presenting cells (Fig. 3). This localization translates to a dramatic reduction in toxicity and enhancement of a tolerogenic effect. Rapamycin-conditioned DCs can become tolerogenic and promote the differentiation of Tregs (30–32). In this study, we have shown that SVP-R specifically targets professional phagocytes, such as macrophages, DCs, and, to a lesser extent, monocytes. Importantly, LMB-100 specifically targets professional phagocytes and colocalizes with SVP-R. These results support the finding that SVP-R can induce tolerogenic DCs in vivo (19), but does not preclude additional involvement of macrophages. The tolerance was abrogated after depletion of Tregs by 14 d after SVP-R treatment (Fig. 2C), supporting the mechanism of myeloid cell tolerance mediated by Treg cells. Of note, the antibody used for Treg depletion (24) was recently shown to efficiently deplete CD25⁺CD4⁺ Treg cells, but not CD25⁻Foxp3⁺ Treg cells (33). The finding that a CD25-targeting antibody negated SVP-R activity but a CTLA4-targeting antibody did not may be explained by the dosing and schedule regimen or by differences in mechanism of the two antibodies.

While SVP-R effectively inhibited IgG antibody responses, we observed that specific IgM antibodies were not affected by SVP-R (Fig. 2D). Similarly, Kishimoto et al. (19) injected C57BL6 mice with KLH and SVP-R and found that while SVP-R induced a >99% decrease in anti-KLH IgG, it only reduced 50% of the IgM. This indicates that SVP-R prevents CD4⁺-mediated class switching, but not the initial (T cell-independent) activation of B cell and IgM production. These results support either a Treg-mediated mechanism or a lack of T cell help.

A major differentiator between immune suppression and tolerance is the ability to mount an immune response against other antigens. We found that mice that were tolerized by injections of LMB-100 and SVP-R mounted an immune response to a second antigen injected s.c. (Fig. 2A). That the mice had an immune response to the second immunogen but not to LMB-100, even though both were administered at the same time, dose, and frequency during the challenge phase, indicates the induction of specific tolerance to LMB-100 rather than global suppression of the immune system. Immune suppression is commonly mediated by drugs that have no lasting effect on the immune system after the cessation of therapy. In contrast, immune tolerance involves the induction of regulatory cells that actively maintain tolerance in the absence of drugs. We observed that transfer of splenocytes isolated from mice treated with the combination of LMB-100 + SVP-R (Fig. 2B) prevented ADA formation in naïve recipient mice. Taken together, our data suggest that the combination of LMB-100 + SVP-R induces immune tolerance.

Therapeutic Activity in a Preexisting Antibody Model.

Preexisting antibodies that target and neutralize therapeutic proteins are a hindrance in many therapeutic regimens. Our findings that SVP-R not only were effective in controlling the boost in anti–LMB-100 titers, but also actually demonstrated a striking prolonged tolerance (Fig. 4), indicate that a combination of RITs with SVP-R may be useful in patients with preexisting antibodies or even in patients who participated in previous clinical trials with SS1P, LMB-100, or moxetumomab pasudotox. Many patients in these trials initially responded to immunotoxin therapy, but the response was halted due to ADA formation (7, 34).

The decrease in titers was not due to the depletion of long-lived BM plasma cells that secrete ADAs (Fig. 4C). It is possible that the reduction in titers is associated with inhibition of short-lived plasma cells from memory B cell activation, or perhaps with clearance of the circulating ADAs through immune complex formation. These results agree with work by Zhang et al. (20) showing that mixing SVP-R with recombinant factor VIII and coinjecting the mixture in mice with factor VIII inhibitors resulted in a very low immune response. A similar decrease in ADA was observed in mice with preexisting antibodies against adeno-associated virus after treatment with nonencapsulated rapamycin and prednisolone (35).

In our mouse studies, we observed a wide range of ADA titers (741–47,926) associated with multiple treatment schedules. For example, in Fig. 2C a mean titer of 741 was detected at 9 wk after six doses of LMB-100. The maximal titer of 47,926 was measured at 1 wk after the final injection in an experiment that used 15 doses. The control titer in the Treg experiment (1,348 ± 399) was measured after nine doses of LMB-100. The titer cannot be compared with any other experiment in this study, because we did not use this schedule in any other experiment. Nevertheless, the titer after nine doses is lower than expected and may be explained by variations in age, food, and microbiota.

Rapamycin and Cancer.

The mTOR signaling network contains various tumor-suppressor genes and proto-oncogenes, including PTEN, PIK3, and AKT (reviewed in ref. 36). Rapamycin and other mTOR inhibitors are being evaluated in clinical trials for anticancer effect as single agents and in combinations for multiple types of tumors (reviewed in ref. 37). Our finding that SVP-R improved the cytotoxic and antitumor activity of the immunotoxin (Figs. 5A and 6) is supported by a recent synergy screen that identified the mTOR inhibitor everolimus as the best enhancer of immunotoxin activity among 459 small molecules (38). Encapsulation of rapamycin in 150 nm SVP adds an additional benefit, as nanoparticles enhance permeability and retention in tumors (39, 40). The slow release of rapamycin at the tumor site could synergize with the targeted immunotoxin. Together, the improved antitumor activity and reduced immunogenicity provide compelling support for this combination in human clinical trials.

SVP-R Did Not Affect Tumor Immunogenicity.

SVP-R must be coadministered with antigen to confer immune tolerance (18). Injection of SVP-R even 2 d after LMB-100 was ineffective in inducing immune tolerance (Fig. 1D). Importantly, SVP-R alone did not cause the tumors in immune competent mice to grow faster (Fig. 5 A, E, and F). These observations alleviate a potential safety concern of the SVP-R inducing tolerance against the tumor or making the tumor grow faster.

Triple Combination with Checkpoint Inhibitors.

The groundbreaking advancements in cancer immunotherapy with checkpoint inhibitors and costimulatory agonists, coupled with clinical observations that tumor killing by immunotoxins may induce immune recruitment and activation in the tumor microenvironment (14), have led us to hypothesize that the combination of LMB-100 with checkpoint inhibitors may provide synergistic antitumor activity. To that end, Leshem et al. (41) evaluated the antitumor effect of combining immunotoxins and checkpoint inhibitor antibodies in mice. They found that intratumoral LMB-100 injection combined with anti-mouse CTLA-4 antibody induced dramatic tumor eradication and resistance to secondary tumor challenges.

We evaluated the effect of anti–CTLA-4 and anti–OX-40 antibodies on the onset and intensity of ADA formation against LMB-100, and the ability of SVP-R to prevent these responses. We found that both anti–CTLA-4 checkpoint inhibition and anti–OX-40 costimulatory agonist expedited and intensified the formation of LMB-100 ADAs (Fig. 7). Importantly, SVP-R given on the day of injection of LMB-100 completely eradicated these exacerbated immunogenicity responses. Future work needs to address the effect of SVP-R on the synergistic antitumor activity of immunomodulatory mAbs with LMB-100.

The anti–CTLA-4 mAb used in our study has been studied for anticancer activity in CT26 and MC38 (mouse colon carcinoma) tumor models (42), and it is thought to act by reducing Treg cells and concomitantly activating effector T cells at the tumor site (42). The anti–OX-40 mAb used in the present study has been studied for anticancer activity in CT26 and MOSEC (mouse ovarian) tumor models (43, 44). It is speculated that the mechanism of action involves activation of the NF_KB pathway and, consequently, CD4⁺ T cell activation. The fact that these immune stimulatory mAbs did not compromise the tolerogenic activity of SVP-R suggests that the tolerogenic signal is not overridden by these immunotherapeutic antibodies.

Concluding Remarks

LMB-100 is currently being evaluated in clinical trials for treatment of mesothelioma and pancreatic cancer (https://clinicaltrials.gov/ct2/show/NCT02810418). Immunogenicity likely will be a major obstacle to the maintenance of therapeutic blood levels after multiple treatment cycles, as has been previously observed with several immunotoxins (12). Moxetumomab has completed phase 3 of clinical evaluation, demonstrating 50% immunogenicity in patients with hairy cell leukemia during Phases 1 and 2 (https://clinicaltrials.gov/ct2/show/NCT01829711). Thus, there is an unmet need to mitigate immunogenicity to unlock the full potential of RITs. SVP-R is currently being evaluated in combination with a PEGylated uricase in a Phase 2 clinical trial in patients with symptomatic gout (https://clinicaltrials.gov/ct2/show/NCT02648269). The strategy that we describe here represents a novel and safe approach to unlocking the full potential of RIT therapy for solid tumors. These results may have implications for other highly foreign, life-saving therapeutic proteins as well.

Materials and Methods

LMB-100 and SVP-R.

LMB-100 was manufactured by Roche as described previously (45) and provided for these studies through a Collaborative Research and Development Agreement (2791). SVP-R was manufactured by Selecta Bioscience as described previously, with a rapamycin content of 500 μg/mL (19).

Animal Experiments.

Female BALB/cAnNCr mice (age 8–14 wk) were used for all experiments. All mouse experiments followed National Institutes of Health guidelines approved by the Animal Care and Use Committee of the National Cancer Institute (Animal Protocol LMB-071). Mice were injected i.v. with antigens and SVP-R unless described otherwise. Mice were injected according to the schedules indicated in each experiment (with RIT injected 5 min after SVP-R), and plasma samples were collected by mandibular bleeding. Mouse weight was measured weekly. All mouse studies were performed with age-matched control groups.

For tumor experiments, female BALB/c mice were inoculated with 1 × 10⁶ AB1-L9 cells (Fig. S6) or 1 × 10⁶ CT26 cells (American Type Culture Collection) in RPMI medium in the flank, or 0.5 × 10⁶ 66C14 cells in IMDM medium in the mammary pad. Tumor sizes were measured using a caliper every 2 or 3 d. Mice with a tumor burden >10% of body weight were euthanized. No animals were excluded from the statistical analysis (41).

Depletion of Treg cells was performed by i.p. injection of 200 μg of anti-mouse CD25 depleting antibody (clone PC61) or isotype control (clone TNP6A7) (both purchased from BioXcell), as described previously (24).

Anti–CTLA-4 (Roche IgG2A; clone 9D9) was generously provided by Roche Pharmaceuticals, and anti-OX40 (clone OX-86; InVivoPlus) was purchased from BioXcell. Antibodies were diluted in PBS, and 5 mg/kg was injected i.p. as in the indicated schedules.

Cytotoxicity and Neutralization Assay.

The KLM1 pancreatic cell line was provided by Dr. U. Rudloff (National Cancer Institute). HAY mesothelioma cells were provided by the Stehlin Foundation for Cancer Research. Cells were cultured in RPMI medium supplemented with 10% FCS, 1% l-glutamine, and 1% penicillin-streptomycin. Cells were seeded in 96-well flat-bottom plates (5,000 cells/well) for 24 h, then treated with various concentrations of LMB-100, SVP-R, and both in four replicates. Cell viability was assessed 72 h later using a WST cell viability assay (Dojindo Molecular Technologies) in accordance with the manufacturer’s instructions. Color change was evaluated at optical density (OD) 450 nm. OD reads were normalized at 0–100% viability, with 100% viability representing no treatment and 0% representing treatment with staurosporine (Sigma-Aldrich)-positive control.

Neutralization assays were performed using KLM1 cells as described previously (46). Serum samples from 21 mice were diluted 1:50.

ELISA.

Total Ig anti LMB-100 and anti-ovalbumin antibodies.

Plasma samples were collected into heparinized tubes, spun, and frozen until titer evaluation. Total Ig anti–LMB-100 and anti-ovalbumin antibodies were measured by direct ELISA as described previously (46).

Isotype determination of anti–LMB-100 and total Ig.

ELISA plates (Thermo Fisher Scientific) were coated with 2 μg/mL LMB-100 or polyclonal donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories). Plates were blocked, and serial dilutions of plasma were incubated for 1 h. Captured antibodies in the plasma were bound by goat anti-mouse IgG1, IgG2a, IgG2b, IgG3, and IgM isotyping kits at dilutions of 1:3,000, 1:4,000, 1:4,000, 1:3,000, and 1:16,000, respectively (Sigma-Aldrich). Anti-goat IgG (H + L) HRP (1:15,000; Jackson ImmunoResearch Laboratories) was used for detection.

ADA against the Fab or the toxin fragments.

ELISA plates were coated with 2 μg/mL of the Fab portion of LMB-100 (generously provided by Roche), or a 2 μg/mL RIT containing a murine scFv that targets an irrelevant epitope (anti-Tac) linked to the deimmunized toxin fragment of LMB-100. ADA determination was performed as described above.

The OD of the wells was read immediately after the addition of H₂SO₄ stop solution at a wavelength of 450 nm with subtraction at 650 nm. Titers were calculated based on a four-parameter logistic curve-fit graph and interpolated on the half maximal value of the anti–LMB-100 (IP12) (15) or anti-ovalbumin (clone TOSG1C6; BioLegend) standard curves.

Transfection of Cell Line with Human Mesothelin and Tumor Inoculation.

The AB-1 mouse mesothelioma cell line (Sigma-Aldrich) was stably transfected with human mesothelin cDNA (41) using Lipofectamine LTX/PLUS reagents (Invitrogen) in accordance with the manufacturer’s protocol. The transfected cells were sorted three times for the top 5% expressing cells by FACS. LMB-100/SS1P–sensitive single clones were then isolated from the population of sorted cells. Clone AB1-L9 (5 × 10⁶) was inoculated in BALB/c mice in 100 μL of PBS. When tumor volume reached 200 mm³, tumors were excised. Digested tumors were prepared as described previously (47).

To make single clones of AB1-L9 cells, digested tumors were diluted (0.5 cells/100 μL) and aliquoted 100 μL in a 96-well culture dish with selection reagent. Fifteen single clones were obtained, and the clones with the highest GeoMean values were selected. The final clone was injected s.c. in BALB/c mice; >95% of the tumors were grown in BALB/c mice.

B-Cell ELISpot.

BM extracted from the femurs of eight immunized mice was washed, filtered through a 70-mm mesh, and lazed to eliminate RBCs. Cells were resuspended in warm RPMI supplemented with heat-inactivated FCS, 1% l-glutamine, and 1% penicillin-streptomycin. PVDF plates (0.45 μm; Mabtech) were coated with 2 μg/mL LMB-100 for 18 h, washed, and blocked with assay media at 37 °C for 2 h. Six replicas of each BM sample were seeded at a concentration of 100,000 cells/well and then incubated for 4 h. Spots indicating anti–LMB-100 antibody-secreting B cells were detected using a capture anti-mouse Ig biotinylated antibody (Mabtech) followed by ALP and BCIP/NTP substrate (KPL).

Spots were counted by computer-assisted image analysis (ImmunoSpot 5.0; Cellular Technology). Results are shown in SFC/1E6 cells.

Flow Cytometry.

Spleens were dissected from mice immunized with Alexa Fluor 488-labeled LMB-100, Cy5-labeled SVP-R, or both and also from untreated mice. Splenocytes were extracted by injecting 3 mL of medium supplemented with Liberase, DNAs, and collagenase (all from Roche) to the spleen, followed by a 10-min incubation at 37 °C. Spleens were minced, passed through a 70-mm mesh, and washed, and then RBCs were lysed. All cells were >90% viable as assessed by trypan blue staining. Cells were fixed, washed, and stained as described previously (48) using the following antibodies, obtained from BioLegend: CD3 (clone 17A2), CD4 (clone GK1.5), CD8 (clone 53–5.8), CD19 (clone 6D5), B220 (clone RA3, 6B2), CD11c (clone N418), IAIE (clone M5/114.15.2), CD11b (clone M1/70,), Ly6G (clone 1A85), and Ly6C (clone HK1.4). Data were collected on a FACSCanto II flow cytometer (BD Bioscience) and analyzed with FlowJo version X (Tree Star).

Statistical Analysis.

Statistical analysis and graphing were done using GraphPad Prism software. Multiple comparisons of parametric variables were performed by one-way ANOVA. The Mann–Whitney U test was used for comparisons of two nonparametric variables, and Friedman’s test with Dunn’s multiple comparisons was used for comparisons of multiple nonparametric variables.