Combining Local Immunotoxins Targeting Mesothelin with CTLA-4 Blockade Synergistically Eradicates Murine Cancer by Promoting Anticancer Immunity

Most human tumors harbor mutant proteins that are capable of becoming targets for the immune system (1–3). However, regulatory mechanisms disable the immune system and prevent tumor rejection (4). Cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) functions as an immune checkpoint regulator. Antibodies blocking CTLA-4 (anti–CTLA-4) prolong the survival of patients with metastatic melanoma. However, the response rate is low, ranging from 11% to 19% (5, 6). Combination therapies are being pursued to increase the number of patients benefiting from anti–CTLA-4 therapy.

Mesothelin is a cell surface 40kDa glycoprotein that is being pursued as a target for antibody-based therapies, because it is highly expressed on ductal pancreatic carcinoma (80%–85%), mesothelioma (85–90%), ovarian cancers (60%–65%), lung cancers (60%–65%), and other cancers (7). Mesothelin is not expressed on any vital tissues; its expression is limited to mesothelial cells lining the pleura, peritoneum, and pericardium (8).

Recombinant immunotoxins (RIT) are therapeutic agents composed of an antibody fragment attached to a protein toxin. To target the toxin to solid tumors expressing mesothelin, we have constructed RITs composed of an antibody fragment targeting mesothelin and a portion of Pseudomonas exotoxin A (PE). After entering the cell by endocytosis, the RIT reaches the cytosol, where it inactivates elongation factor 2, arrests protein synthesis, and induces apoptotic cell death (9).

SS1P is an antimesothelin immunotoxin that has been evaluated for cancer therapy in several clinical trials. When given by itself, SS1P is well tolerated, but has low activity (10, 11), mainly because it is immunogenic and can be given for only 1 cycle before neutralizing antibodies develop. When SS1P was combined with pentostatin and cyclophosphamide to modulate the immune system and delay antibody formation, SS1P produced delayed and prolonged regressions in several patients with advanced drug-refractory mesothelioma (12). These responses were accompanied by increased metabolic activity in the tumor, documented by PET scan, indicating that the tumors were infiltrated with immune cells. The ability of RITs to induce antitumor immunity was also indicated in a clinical study in patients with brain tumors receiving local immunotoxin therapy, where regressions began and progressed long after therapy had ended (13, 14). Because a direct antitumor effect is expected to occur close to the time of treatment, an additional immune-mediated effect was suggested.

LMB-100 (previously named RG7787) is an improved antimesothelin RIT designed to be more active, less immunogenic, and better tolerated by patients compared to SS1P (15, 16). It contains a humanized antimesothelin Fab fused to a 24kDa truncated PE fragment with mutations that suppress B- and T-cell epitopes. Clinical trials with LMB-100 in patients with mesothelioma and pancreatic cancer are in progress at the NIH. (https://clinicaltrials.gov/ct2/show/NCT02810418).

Based on clinical data, we hypothesized that the antitumor effect of SS1P or LMB-100 can be potentiated by combining it with anti–CTLA-4. To examine this hypothesis, we used a 66C14 BALB/c mouse breast cancer cell line (17) transfected with a cDNA encoding human mesothelin to create cell line 66C14-M. Because tumor cells expressing human mesothelin are rejected by normal BALB/c mice, the cells were grown in BALB/c mice expressing a human mesothelin transgene. We find that the combination of SS1P or LMB-100 injected directly into tumors with anti–CTLA-4 given i.p. causes complete regressions most of injected tumors and half of uninjected tumors and induces antitumor immunity.

We engineered a plasmid encoding a truncated human mesothelin (Supplementary Figs. S1A and S1B). The expression plasmid contains a ferritin promoter and an IG κ leader sequence attached to truncated mesothelin (amino acids 296–585). The megakaryocyte-potentiating factor (MPF) portion of full-length mesothelin was deleted and the glycosyl phosphatidyl inositol membrane attachment site was replaced with the transmembrane domain of a murine transferrin receptor. The mouse mammary cancer cell line 66Cl4 luc was transfected using lipofectamine LTX & Plus (Invitrogen) according to the manufacturer's instructions. A puromycin resistance gene enabled selection of the transfected clones. After the cells were grown in 3 μg/mL puromycin for 1 month, cells expressing high levels of mesothelin were sorted by flow cytometry using 5 μg/mL MN-1 antimesothelin antibody (Rockland Immunochemicals).

To express human mesothelin in mice, a cDNA that consists of full-length mesothelin (Accession Number: NM_005823) under the control of a CAG promoter was produced (Supplementary Fig. S1C). The pronucleus of fertilized oocytes from BALB/c mice were microinjected with the plasmid. Founder animals carrying a human mesothelin transgene were identified by Southern blot analysis followed by PCR screening to establish the founder lines. The founder lines were further characterized for protein expression by ELISA of the blood. After screening four founder lines, we selected line that has serum MPF levels of 50 ng/mL. That mouse line is designed as 01TGZ-N1.

The 66Cl4 luc tumor cell line was provided by Dr. C. L. Jorcyk (Boise State University, Boise, ID). Tumor cells were cultured in IMDM supplemented with l-glutamine, HEPES (Gibco Life Technology), 10% FBS (HyClone, Thermo Scientific), 100 U/mL penicillin, and 100 μg/mL streptomycin. For 66Cl4-M cells, we added 3 μg/mL puromycin (Glibco Life Technology) to maintain mesothelin expression. All cells were grown and maintained at 37°C with 5% CO_2. LMB-100, LMB-100-I, and anti–CTLA-4 (clone 9D9, mouse IgG2a isotype) were manufactured by Roche. LMB-100-I is an inactive form of LMB-100 lacking the glutamic acid in position 553; the furin cleavage site (RHRQPRGWEQ) was replaced with a glycine–serine linker (GGGGGSGGGGSGGS; ref. 18). SS1P was manufactured by ABL, LMB-2 was manufactured by BDP, and HA22 was manufactured by MedImmune. Endotoxin levels for all RITs were below 5 EU/mg. Most immunotoxins were diluted in PBS. HA22 was diluted in a 32 mmol/L citrate and 0.65% v/v tween 80. Paclitaxel was from TEVA. Trypan blue was from BioWhittaker. Western blot analysis was performed with an antimesothelin MORAb-009 antibody (Morphotek) and an anti-GAPDH antibody (Cell Signaling Technology).

66Cl4-M cells were plated at 5000 cells/well in 96-well plates 24 hours before various RITs were added. Cells were incubated for 3 days, and viability was assessed using a WST-8 cell counting kit (Dojindo Molecular Technologies) according to the manufacturer's instructions.

All mouse experiments were approved by the NCI Animal Care and Use Committee. Female mesothelin Tg BALB/c and wild-type (WT) BALB/c mice ages 6 to 12 weeks were used in this study. To produce tumors, we implanted 1 × 10⁶ 66C14-M cells in the right second mouse mammary fat pad. To form a second tumor, we implanted 1 × 10⁵ 66C14-M cells in the left second mammary fat pad. Treatment was initiated when tumors reached 80 to 100 mm³, typically on days 11 to 13 after tumor implantation. Tumor volume was calculated using the formula 0.4 x length x width². RIT was injected into tumors in a volume of 30 μL using a 29-G needle. Before injection, the tumor surface was sterilized using povidone iodine and alcohol pads. Anti–CTLA-4 was administered i.p. Mice were euthanized when the tumor volume reached 700 mm³. The day of euthanasia was used to calculate survival. Mice reaching complete remission were rechallenged 90 days after the first tumor implantation with 1 × 10⁶ 66C14-M or 66C14 cells. Surviving mice were followed for 6 months following the first tumor implantation.

To deplete CD8⁺ T cells we used antimouse CD8 antibody (clone 2.43, Bioproducts for Science [Madison, WI]). The antibody (100–200 μg/dose) was administered i.p. on the same day as the RIT.

Tumors were removed, fixed overnight in 10% formalin, and submitted to the Pathology/Histotechnology Laboratory (Frederick, MD) for staining with hematoxylin and eosin (H&E) and anti-CD8a antibody (clone 4SM15, eBioscience). Automated image processing and analysis software (Toolbox; Leica Biosystems) was used to identify and quantitate the number of CD8⁺ cells.

We used Prism 6 for all statistical calculations and graph fitting. To compare the survival of groups, we used log-rank (Mantel–Cox) test. The Mann–Whitney test was used to analyze the differences in the number of T cells in the tumors. All animal experiments were done twice and often more times. Error bars represent SDs.

Because the targeting moiety of SS1P and LMB-100 is not mouse cross-reactive, we generated a mouse cancer cell line expressing the human mesothelin protein and a transgenic mouse strain expressing human mesothelin and tolerant to the human mesothelin protein to determine whether immune system activation by anti–CTLA-4 can augment the antitumor effect of these RITs. We produced BALB/c transgenic mice expressing the full-length human mesothelin gene as described in Materials and Methods. MPF is encoded by the mesothelin gene and released from the pre-protein into the blood by the action of furin (8). To ensure high mesothelin expression, we engineered a plasmid encoding only mature mesothelin fused to the transmembrane domain of the murine transferrin receptor, and used it to establish cell line 66C14-M (Supplementary Figs. S1A and S1B).

To determine if 66C14-M cells can form tumors, cells were implanted into five WT BALB/c mice and five mesothelin Tg mice. Figure 1A shows the cells formed tumors in all the transgenic mice, but were rejected in 4 of 5 normal mice. To determine sensitivity of the 66C14-M cells to the cytotoxic effect of SS1P and LMB-100, a WST-8 cell viability assay was performed. Figure 1B shows that 66C14-M cells are sensitive to both SS1P and LMB-100. The IC₅₀ of SS1P is 8 ng/mL and that of LMB-100 is 46 ng/mL. We also tested LMB-100-I, mutant immunotoxin with a E553 deletion (see Materials and Methods), and the mutant protein had no effect on cell viability up to 210 μg/mL.

The transgenic mice were found to express human mesothelin in the pancreas where mesothelin is not expressed in normal mice or humans (19, 20). We found that some mice died when treated i.v. with 3 doses of 50 μg of LMB-100, whereas nontransgenic mice were not killed by this dose. To avoid this toxicity, we chose to inject LMB-100 or SS1P directly into 66C14-M tumors.

To evaluate how well a locally administered agent would distribute in a tumor, we injected trypan blue, which is visible to the eye, into five (72–117 mm³) 66C14 tumors. One hour after injection, we harvested the tumors and photographed them. We found that trypan blue reached most parts of the tumors (Supplementary Fig. S2). This suggests that direct injection of SS1P or LMB-100 should enable the injected agent to reach the majority of tumor cells.

To evaluate the therapeutic effect of SS1P alone, we injected SS1P (5 μg) in PBS or PBS alone into 66C14-M tumors five times, beginning when the tumors reached 80–100 mm³ average size. Mice were injected on days 1, 2, 6, 10, and 14, with day 1 being the day of the first injection. As shown in Fig. 2A, tumors injected with PBS grew rapidly and reached the size of the experimental end point (volume > 700 mm³) at a median time of 15 days. Injecting SS1P into tumors delayed the time to reach 700 mm³ by 5 days, but did not cure any mice (Fig. 2B and C).

To increase the antitumor effect of SS1P, we administered anti–CTLA-4 i.p. (days 3, 7, 11, and 15). Fig. 2D shows experiments in which a constant dose of SS1P (5 μg) was combined with increasing doses of anti–CTLA-4, ranging from 10 μg to 100 μg. We found that therapeutic effect of the combination did not improve above a dose of 25 μg, and therefore used this dose in further experiments. Figure 2E–G shows an experiment in which 11 mice were treated with anti–CTLA-4 (25 μg) and PBS or SS1P (5 μg). We observed that there were no tumor regressions in mice treated with anti–CTLA-4 and intratumoral PBS, but in mice treated with anti–CTLA-4 and intratumoral SS1P 8/11 tumors disappeared (Fig. 2F); two became much smaller but eventually grew back. Figure 2G shows that there is a significant survival benefit when treating with SS1P and anti–CTLA-4 (P < 0.0001). Mouse weights were stable during therapy, indicating that the combination therapy is well tolerated (Supplementary Fig. S3A).

LMB-100 is an improved, less immunogenic, form of SS1P that is currently being evaluated in phase I/II clinical trials. To determine if LMB-100 could also induce tumor regressions, we treated 66C14-M tumor–bearing mice with LMB-100 and anti–CTLA-4. Because LMB-100 is 6-fold less potent than SS1P on the 66C14-M cell line (Fig. 1B), we used 30 μg of LMB-100 for antitumor experiments. All tumors treated with LMB-100 alone continue to grow (Fig. 3A). When anti–CTLA-4 was combined with PBS, transient tumor shrinkage was observed in 1 of 6 mice and no complete remissions were obtained (Fig. 3B). However, combining anti–CTLA-4 and LMB-100 resulted in tumor size reduction in 11 of 13 mice, and complete tumor elimination in 8 of 13 mice (61%; Fig. 3C). To determine whether cell killing by the immunotoxin is required for the antitumor effect, we tested LMB-100-I, an inactive mutant form of LMB-100. Of 13 mice treated with 30 μg LMB-100-I and anti–CTLA-4, only one achieved a complete remission (Fig. 3D). Survival was comparable with that of mice treated with anti–CTLA-4 and PBS, and significantly lower than the survival of mice treated with anti–CTLA-4 and LMB-100 (P < 0.001, Fig. 3E). Mice treated with LMB-100 and anti–CTLA-4 did not lose weight during therapy, again implying lack of major toxicity (Supplementary Fig. S3B)

To determine whether CD8⁺ T cells were required for the therapeutic effect, we treated 66C14-M tumor–bearing mice with anti-CD8 antibody (100 μg), intratumoral SS1P (5 μg), and anti–CTLA-4 (25 μg). Anti-CD8 antibodies were given on the same schedule as SS1P (days 1, 2, 6, 10, and 14). Figure 4A and B shows that anti-CD8 reduced the number of complete remissions in mice treated with SS1P and anti–CTLA-4 from 5/9 (55%) to 1/9 (11%) and significantly decreased survival (P < 0.05; Fig. 4C). Two additional CD8 depletion experiments were done using LMB-100 (30 μg), anti–CTLA-4 (100 μg), and anti-CD8 antibody (100-200 μg; Supplementary Fig. S4). In one experiment, a single mouse out of 6 (16%) treated with LMB-100, anti–CTLA-4, and anti-CD8 reached complete remission, whereas in the other experiment none out of 6 mice (0%) were cured. In mice treated with the combination drugs without CD8 depletion, complete remission was achieved in 10 of 14 mice (71%).

Anti–CTLA-4 induces antitumor immunity through the adaptive immune system (21). To evaluate whether the combination of SS1P or LMB-100 with anti–CTLA-4 produced long-term antitumor immunity, we injected 66C14 cells into 38 mice or 66C14-M cells into 6 mice about 45 days after the mice achieved complete remissions. We found that in 37/38 mice, the 66C14 cells were unable to form tumors. Also, no tumors formed in 6 of 6 mice receiving 66C14-M cells (Supplementary Table S1).

We examined tumor sections by H&E staining and used immunohistochemistry to determine the number of CD8⁺ cells infiltrating the tumors of mice responding to treatment. Mice with 66C14-M tumors were treated with SS1P (5 μg) and anti–CTLA-4 (25 μg). When the tumor size was reduced by 20% to 60% from maximum, the tumors were harvested. Also harvested were tumors from nonresponding mice treated with PBS alone, SS1P alone, or anti–CTLA-4 and PBS. Figure 5 shows representative sections stained with H&E and for CD8⁺ cells for all four treatment groups. The areas outlined in red are viable tumor cells; the areas outlined in green and stained pink are necrotic and the areas outlined in yellow are micro-abscesses with necrotic material and admixed/degenerative granulocytes.

In tumors injected with either PBS (n = 3) or SS1P (n = 3) or PBS and anti–CTLA-4 (n = 3), there were small areas of necrosis in the center of the tumors, which may have been caused by the injection procedure or could represent central necrosis seen in many rapidly growing tumors. In 7 of 8 tumors treated with SS1P and anti–CTLA-4, the central necrosis was very pronounced (Fig. 5A) with few viable tumor cells present and these were admixed with PMNs. These tumors usually contain a collar of admixed eosinophils and mononuclear cells, making up to 30% of the total “mass.” Whether this collar is part of the tumor and contains dying tumor cells or is just close to the tumor has not yet been established.

Figures 5B and C show representative sections stained for CD8⁺ T cells. A very large number of CD8⁺ cells were observed in the inflammatory collar of tumors treated with SS1P and anti–CTLA-4. To quantify these differences, the number of CD8⁺ cells in the entire tumor and the number present only in the inflammatory collar were measured. Figure 5D shows that there are more CD8⁺ T cells in the tumors of mice treated with anti–CTLA-4 with SS1P or without SS1P than in mice treated with SS1P alone or PBS alone. Figure 5E shows that there were more CD8⁺ cells in the inflammation collar of mice treated with anti–CTLA-4 and SS1P than in the other mice (P < 0.05).

66C14-M cells are sensitive to the cytotoxic effect of SS1P with an IC₅₀ of 8 ng/mL (Fig. 1B). Nonetheless, we needed to inject 5 μg of SS1P directly into the tumor mass to achieve complete remissions. No cures were obtained in 6 mice treated with 0.5 μg SS1P and anti–CTLA-4 (Supplementary Fig. S5).

To determine if the tumor cells growing in mice lost mesothelin expression, we did a Western blot with an antimesothelin antibody. Mesothelin levels in the tumor sample (Fig. 6A) are about 50% lower than in cultured cells, which is expected because about 50% of the tumor is made up of mouse stromal and inflammatory cells. Thus, loss of mesothelin does not explain the need for high dose antimesothelin RIT. To further examine how much of the therapeutic effect is mesothelin-mediated, we used LMB-2 (anti-TacFv-PE38). LMB-2 has the same toxin moiety as SS1P, but targets human CD25, an antigen not expressed on mouse 66C14-M cells. Using a WST-8 viability assay, we found that LMB-2 has an IC₅₀ of 14 μg/mL (Fig. 6B), which is 1,700-fold higher than the IC₅₀ of SS1P (Fig. 1B).

We then treated 66C14-M tumor–bearing mice with SS1P (5 μg) or LMB-2 (5 μg) and anti–CTLA-4 using our standard protocol. As shown in Fig. 6C, tumors continued to grow in all mice that were treated with intratumoral PBS and anti–CTLA-4, whereas 50% of the mice (7/14) treated with SS1P and anti–CTLA-4 achieved complete remissions (Fig. 6D) and 5 of 15 mice treated with LMB-2 and anti–CTLA-4 achieved complete remissions (Fig. 6E). The survival of the SS1P and LMB-2 groups was greater than that of mice treated with PBS and anti–CTLA-4 (P < 0.0001), but the difference in survival between mice treated with SS1P and LMB-2 was not significant (Fig. 6F).

To determine if other immunotoxins not targeted to mesothelin were also effective in producing an antitumor effect, we evaluated HA22, an immunotoxin that targets human CD22, in combination therapy with anti–CTLA-4. We found that in 66C14-M tumor–bearing mice, no complete remissions were achieved by vehicle, HA22 alone or vehicle and anti–CTLA-4, whereas combining anti–CTLA-4 with HA22 produced complete remissions in 6 of 10 mice (Supplementary Fig. S6A). Survival benefit was found in mice treated with HA22 and anti–CTLA-4 compared with control groups (P < 0.0001; Supplementary Fig. S6A). We also tested HA22 and anti–CTLA-4 in 66C14 tumors not expressing mesothelin and observed that 4 of 7 tumors underwent a complete remission.

To determine the effect of antimesothelin immunotoxin on cells that do not express the target, we examined the effect of intratumoral SS1P (10 μg) combined with anti–CTLA-4 on parental nonmesothelin expressing 66C14 tumors. Supplementary Fig. S6B shows that 10/10 66C14-M tumor–bearing mice underwent complete regressions, while 8 of 10 mice with 66C14 parental tumors also reached complete remissions. We also evaluated the activity of native PE in combination with anti–CTLA-4 but found it was very toxic to the mice so that we could not give more than 100 ng. At this dose, the antitumor effect was minimal.

To determine if another kind of cytotoxic agent could recapitulate the synergy, we injected tumors with paclitaxel and anti–CTLA-4 and found no increase in rate of complete remissions or survival benefit (Supplementary Fig. S7).

The ability of anti–CTLA-4 and intratumoral SS1P to promote antitumor immunity as indicated by rejection of a second tumor challenge led us to hypothesize that this therapy can also control metastatic tumor growth. To determine if uninjected tumors would respond, we implanted 66C14-M cells into both right and left mammary fat pads. We inoculated 10⁶ cells on the right side and 10⁵ cells on the left side. SS1P (25 μg) was injected directly into one tumor and anti–CTLA-4 (25 μg) given i.p. Figure 7 shows a representative experiment in which mice were treated with SS1P (25 μg) on days 1, 5, 9, and 13, and with anti–CTLA-4 (25 μg) on days 2, 6, 10, 14, and 18. Tumor regressions occurred in 10 out of 12 SS1P injected tumors (83%) and in 6 of 12 uninjected contralateral tumors (50%). No tumor regressions occurred with drugs given separately. In two repeat experiments, 91% and 83% of the mice treated with anti–CTLA-4 and SS1P achieved complete remission in the injected tumors (11/12 and 5/6 mice) and in 58% and 50% of uninjected tumors (7/12 mice and 3/6 mice).

We report here that locally delivered RITs combined with systemic anti–CTLA-4 cause eradication of injected as well as uninjected tumors. When given alone, each agent has very little activity. Our data indicate that immunotoxin-mediated cell death is immunogenic and leads to induction of a T cell response against the tumor that contributes to tumor regression. This also results in long-term immune memory and protection against subsequent rechallenge with the tumor.

To carry out these studies, we produced a mouse breast cancer cell line that expresses human mesothelin. Because the cells are rejected by normal mice, we developed a mouse line that expresses human mesothelin so that the mice are tolerant to human mesothelin. Because human mesothelin is expressed in the pancreas of these mice, we could not administer the immunotoxin i.v. and instead administered it directly into the tumors. This route avoided systemic exposure responsible for the on-target toxicity caused by atypical tissue expression in the transgenic animals.

The difference in mesothelin expression patterns between transgenic mice and humans makes it difficult to predict toxicity. Nevertheless, we show that in this model, in which mesothelin is expressed by essential organs, combining anti–CTLA-4 with localized immunotoxin is well tolerated. Another concern arising from combining immunotoxin with anti–CTLA-4 is increasing the immunogenicity of the immunotoxin. Patients given SS1P i.v. establish neutralizing antibodies against the drug after 2 to 3 weeks. The concentration of antidrug antibodies may increase when an immunotoxin is combined with an immune checkpoint inhibitor. However, given the high concentration of SS1P injected into the tumors (5–30 μg in 30 μL for a mouse and potentially 1000 μg/mL for humans), we believe that the levels of antibody against the immunotoxin would not be high enough to neutralize the RIT in the local tumor microenvironment.

The tolerance of the transgenic mice to human mesothelin may account for the lack of requirement of mesothelin as an antigen for rejection of the tumor, and thus for the cross-protection against rechallenge with the parental tumor cell line lacking mesothelin. This indicates that the antitumor immunity develops against tumor neoantigens rather than human mesothelin.

We found that we needed to inject a high dose of SS1P (5 μg) into tumors to achieve a therapeutic response even though the cells are sensitive to low doses of SS1P in cell culture. In addition, combining SS1P, or nonmesothelin-specific LMB-2 or HA22, with anti–CTLA-4 yielded a similar therapeutic effect when tested on 66C14-M tumors or even on 66C14 parental tumors not expressing mesothelin. Altogether, these findings indicate that the immunotoxin does more than target specific cells in this model of localized tumor injection. We confirmed that mesothelin, the target of the immunotoxins, continued to be expressed in 66C14-M tumors. Further work is needed to understand the mechanism underlying the target independency, but after intratumoral injection much of the RIT uptake may be mediated by endocytosis independent of antimesothelin targeting.

The current experiments were stimulated by our clinical observation that several patients with mesothelioma who received both immunotoxin therapy and immunomodulatory drugs pentostatin and cyclophosphamide had tumor responses weeks to months after the immunotoxin therapy (12). Because the mouse tumors we used grew rapidly, we needed to begin treatment about 10 days after tumor implantation and finish treatment within a 2-week period, so the time scale is compressed compared with the human studies.

We observed that the major antitumor response began 12 to 15 days after administration of anti–CTLA-4. The presence of a lag period indicates several days are needed for an adaptive immune response to develop (22). Tumor regressions were accompanied by inflammation and CD8⁺ cell infiltration, most prominently found in the tumor margins. Our observations are consistent with previous studies showing that anti–CTLA-4 therapy increases the percentage of tumor-infiltrating CD8⁺ T cells and depends on this cell subset (23, 24).

The generation of immunity to cancer is a cyclical process that can be self-propagating, leading to an accumulation of immune-stimulatory factors that in principle should amplify and broaden T-cell responses. The cycle is also characterized by inhibitory factors that lead to immune regulatory feedback mechanisms, which can halt the development or limit the immunity. For the cancer-immunity cycle to be efficient, release of antigens from the cancer cell must occur and these antigens must be processed and presented by APCs in order to prime and activate effector T cells that need to infiltrate into the tumor site (25). Immunotoxin killing, as shown, can promote the cancer-immunity cycle by killing tumor cells locally, thereby releasing tumor antigens and allowing for their presentation by professional APCs.

Here, we show that locally delivered RITs combined with anti–CTLA-4 promoted regression of a distant tumor. Unlike other anticancer modalities, immunotherapy does not directly kill cancer cells, but reshapes the conditions in which the immune system targets cancer. Thus, a local immune activation can spread and eliminate distant tumor sites. Our findings with regression of untreated distant or contralateral tumors suggest that locally delivered treatment can be used to treat systemic cancers.

Other therapies that induce tumor cell death, including chemotherapy and radiotherapy, promote antitumor immunity (26). One example is abscopal phenomenon in which local radiotherapy results in regression of a distant tumor foci. Similar to our results, combining check point blockade with local radiotherapy improves the therapeutic effect (27).

Rechallenging mice that had undergone complete tumor regressions with tumor cells 45 days after the regressions occurred demonstrated that antitumor immunity was present. The mice rejected tumors expressing mesothelin, but also rejected 66C14 tumors not expressing mesothelin. Because this long-term immunity does not require mesothelin expression it is probably due to responses to some of the mutant proteins present in the 66C14 cells.

The finding of T cells within tumors treated with PBS alone suggests a mild immune response may be present in untreated mice that is enhanced by the combination of anti–CTLA-4 and immunotoxin. Our finding of occasional partial responses in mice treated with anti–CTLA-4 and PBS or inactive-immunotoxin is probably due to enhancement of low-grade antitumor immunity. Because the toxin is of bacterial origin, we were concerned it might be activating receptors for pathogen-associated molecular pattern. This possibility was eliminated, because an immunotoxin (LMB-100-I) with mutations that abrogate cytotoxic activity in cell culture, showed no synergy with anti–CTLA-4 in the mice.

Clinical studies showed that intratumoral injection of TP-38 targeting the EGF receptor or an immunotoxin scFv(FRP5)-ETA targeting ErbB2 induced regressions of breast, melanoma, and brain tumors (13, 14, 28). We speculate that anti–CTLA-4 can potentiate the therapeutic effect of these RITs as well.

Altogether, this study demonstrates the antitumor effect of local immunotoxin therapy combined with systemic injection of anti–CTLA-4. We believe this combination therapy has potential for clinical use.

M. Terabe is a consultant/advisory board member for Intensity Therapeutics. G. Niederfellner has ownership interest in a patent as co-author. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Leshem, J. O'Brien, M. Terabe, B. Bossenmaier, I. Pastan

Development of methodology: Y. Leshem, J. O'Brien, X. Liu, J.A. Berzofsky, I. Pastan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Leshem, J. O'Brien, T.K. Bera

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Leshem, J. O'Brien, T.K. Bera, J.A. Berzofsky, G. Niederfellner, C.-H. Tai, Y. Reiter, I. Pastan

Writing, review, and/or revision of the manuscript: Y. Leshem, J. O'Brien, X. Liu, T.K. Bera, M. Terabe, J.A. Berzofsky, B. Bossenmaier, G. Niederfellner, C.-H. Tai, Y. Reiter, I. Pastan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. O'Brien

Study supervision: Y. Reiter, I. Pastan

The authors thank Dr. Diana Haines for assistance with pathology and Donna Butcher for assistance with immunostaining.

This research was supported in part by the Intramural Research Program of the NIH, NCI, Center for Cancer Research and with a Cooperative Research Development Agreement (#2791) between the NCI and Roche Pharmaceuticals.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.