Antigen-Specific Bacterial Vaccine Combined with Anti-PD-L1 Rescues Dysfunctional Endogenous T Cells to Reject Long-Established Cancer Free

Tumors can escape immune control despite expressing tumor-specific antigens arising from mutations (1–3). Although immunogenic cancer cells can induce functional CD8⁺ T-cell responses, this is usually restricted to early stages of tumor growth (4–6). Immunity rapidly decays with tumor growth, concurrent with the establishment of a tumor microenvironment in which cancer cells are embedded in a suppressive tumor stroma (7). Human tumors are regularly infiltrated by dysfunctional endogenous PD-1–expressing CD8⁺ T cells (8–10). Proliferation and effector function of these CD8⁺ T cells are likely impaired because of engagement of PD-1 with PD-L1 expressed by cancer cells and/or antigen-presenting cells (APC; refs. 11, 12). Although clinical trials blocking PD-1/PD-L1 interactions with antagonistic antibodies have reported promising results (13, 14), new therapeutic approaches need to be developed to rescue antitumor T-cell responses for patients who do not respond to PD-1 blockade alone.

Therapeutic vaccination has been extensively tested as a second approach to generate or rescue endogenous tumor-reactive T cells. Because bacterial products can strongly activate APCs through Toll-like receptor (TLR) signaling, many vaccine studies have used bacterial products as adjuvants (15) or used live bacteria that express tumor antigen (16, 17). Nevertheless, therapeutic vaccination approaches have not reported eradication of long-established experimental tumors (18) or human tumors (for review see ref. 19). We define an experimental tumor as long-established if 2 or more weeks old and exceeding 100 mm³. It is only after 2 weeks that artifacts from cancer cell inoculation–significant necrosis, acute inflammation, and an initial functional T-cell response–finally resolve (4–6, 20).

The limited success of previous therapeutic vaccination approaches is likely due to vaccine-generated T cells becoming dysfunctional after infiltrating long-established suppressive tumors (21, 22). There have been extensive studies showing that intravenously-injected attenuated Salmonella Typhimurium strains preferentially accumulate in murine tumors (for review see ref. 23) and reduce immunosuppression in the tumor and tumor-draining lymph node (24, 25). However, previous studies using S. Typhimurium, either unmodified or genetically modified to deliver recombinant antitumor proteins or short hairpin RNA (shRNA) into tumors, have not eradicated long-established tumors in immunocompetent mice (23–26). We hypothesized that intravenous injection of antigen-producing S. Typhimurium could be used effectively to rescue T-cell dysfunction by (i) coupling antigen delivery and TLR stimulation to APCs and (ii) generating a proinflammatory tumor microenvironment. To test this hypothesis, we treated long-established B16 melanoma tumors that expressed the model tumor-specific antigen ovalbumin (OVA). This model provided the following advantages: (i) B16-OVA tumors resembled human tumors that are also infiltrated by dysfunctional endogenous PD-1⁺ CD8⁺ T cells (8–10), (ii) targeting the SIINFEKL (SIINF) epitope of OVA provided valuable immunologic tools to detect the SIINF epitope and track SIINF-specific CD8⁺ T cells, and (iii) the SIINF epitope has high affinity for H-2K^b (27), similar to a natural unmodified tumor-specific rejection epitope also presented by H-2K^b (28). Indeed, treating mice bearing long-established B16-OVA tumors with S. Typhimurium A1-R (A1-R)–producing SIINF rescued the endogenous dysfunctional tumor-specific CD8⁺ T-cell response, resulting in tumor eradication in about one third of the experimental mice. Anti-PD-L1 antibody has been shown to rescue dysfunctional T cells (13, 29), but when used alone, it was ineffective in treating B16-OVA tumors. However, anti-PD-L1 synergized with antigen-expressing A1-R leading to tumor rejection in a large majority of tumor-bearing mice.

Antigen constructs were cloned into the pEGFP (Clontech) plasmid. We codon-optimized the OVA antigen construct (Invitrogen) encoding the first 104 amino acids of the SopE gene, the M45 epitope from the adenovirus E4-6/7 protein (30), and amino acids 248 to 357 of ovalbumin before inserting this antigen construct into the pEGFP backbone. Using standard cloning techniques, the SIINFEKL epitope of OVA was replaced by the irrelevant SNFVFAGI (31) epitope to make a control antigen construct. Expression plasmids were electroporated into A1-R bacteria. Antigen expression by A1-R was verified by Western blot analysis using an antibody against the M45 epitope (30) as described previously (16).

C57BL/6 and C57BL/6 CD8^−/− (B6.129S2-CD8a^tm1Mak/J) mice were purchased from The Jackson Laboratory and maintained in a specific pathogen-free facility at the University of Chicago (Chicago, IL). Female mice were used at 8 to 14 weeks of age. All animal experimentation was carried out in accordance with the University of Chicago Institutional Animal Care and Use Committee (IACUC) protocols.

The B16-OVA M04 cell line, a gift from Mary Jo Turk (Dartmouth Medical School, Hanover, NH) received in 2009, has been described previously (32). OVA expression was verified using the 25-D1.16 antibody that recognizes the SIINFEKL peptide bound to H-2K^b. The M04 line was regularly confirmed to be Mycoplasma-free by using the ATCC Universal Mycoplasma Detection Kit [American Type Culture Collection (ATCC)]. Cancer cells were trypsinized, washed 1× in PBS, and injected at a dose of 5 to 10 × 10⁶ s.c. on the backs of mice. B16-OVA tumor-take was more than 60% and all established tumors invariably progressed rapidly and killed the host (20 of 20 tumors from 8 independent experiments). Tumor size was measured along 3 orthogonal axes (a, b, and c). Tumor volume = abc/2. Mice were sacrificed once tumors exceeded 1.5 cm³.

The leucine–arginine auxotrophic S. Typhimurium A1-R strain (refs. 33, 34; AntiCancer, Inc.) was intravenously injected at a dose of approximately 2 × 10⁷ colony formation units in 200 μL 1× PBS per mouse. To study the effect of PD-L1 and CTLA-4 blockade, mice were treated with 150 μg of anti-PD-L1 (10F.9G2) and/or 100 μg of anti-CTLA-4 (UC10-4F10-11) intraperitoneally every third day. Anti-CTLA-4 was purchased from the Fitch Monoclonal Facility (University of Chicago) and anti-PD-L1 was purchased from BioXCell. Further experimental details are in the Supplementary Experimental Procedures.

A staining solution, referred to as SIINF-dX, containing SIINFEKL peptide-loaded K^b-DimerX [(K^b)₂-IgG], anti-mouse IgG1–phycoerythrin (PE) or -APC (BD Biosciences), and mouse immunoglobulin G1 (IgG1) isotype control was used to detect SIINF-specific CD8⁺ T cells. A staining solution, referred to as SIYR-dX, was loaded with the irrelevant SIYRYYGL peptide and used as control. Details about other antibodies and flow-cytometric analyses are in the Supplementary Experimental Procedures.

Single-cell suspensions from the tumor were used in a peptide restimulation assay. Cells were placed into a 96-well plate with each well containing 1 to 2 × 10⁶ viable cells in 200 μL RPMI with 20 μg/mL Brefeldin A, and 1 μg/mL SIINFEKL peptide (generously provided by S. Meredith, University of Chicago). Cells were restimulated with peptide for 5 hours at 37°C. Experimental details are provided in the Supplementary Experimental Procedures.

A Mann–Whitney test was used to conduct pairwise comparisons of tumor sizes between different groups during outgrowth. Both nonparametric Kruskal–Wallis and parametric analysis of variance (followed by Sidak's multiple comparisons procedure) were used for analysis of 3 or more groups. Fisher exact test was used to compare survival rates of mice at an indicated time or tumor rejection rates between different treatment groups. Wilcoxon-signed rank and paired t tests following parametric transformation of the data were used to compare the percentage of CD8⁺ T cells that were stained by SIINF-dX or SIYR-dX. The P values reported in this article were derived from parametric analysis. All results reported to be significant following parametric analysis were also significant or at least marginally significant (P < 0.06) using the nonparametric tests.

S. Typhimurium oral vaccines depend on type III secretion to deliver heterologous antigen to APCs resulting in the generation of antigen-specific CD8⁺ T cells (16, 35). Similar to these studies, we constructed a fusion protein consisting of the SopE type III secretion/translocation domain (16), an M45 epitope tag (30), and a SIINFEKL (SIINF)-containing OVA domain (amino acids 248–357). As SIINF-negative controls, we used (i) a similar fusion protein in which the irrelevant SNFVFAGI (SNFV) epitope (31) replaced the SIINF epitope or (ii) the enhanced GFP (EGFP; diagrams in Fig. 1A).

A1-R SIINF and A1-R SNFV were generated by introducing expression plasmids encoding the corresponding fusion proteins into S. Typhimurium A1-R. A1-R EGFP has been described previously (33). Western blot analyses were used to verify the expression of the A1-R SIINF and A1-R SNFV fusion protein (Fig. 1B).

We evaluated whether A1-R SIINF could deliver the SIINF epitope into APCs for correct MHC class I processing and presentation to CD8⁺ T cells. J774 K^b-expressing macrophages were infected in vitro with A1-R SIINF or A1-R SNFV. The capacity of the infected macrophages to present the SIINF epitope to CD8⁺ T cells was determined by stimulation of B3Z, a CD8⁺ T-cell hybridoma specific for SIINF. A1-R SIINF- but not A1-R SNFV-infected macrophages stimulated B3Z to secrete IL-2 (Fig. 1C), showing that A1-R SIINF can deliver the SIINF epitope into APCs for MHC class I presentation.

We evaluated the endogenous SIINF-specific CD8⁺ T-cell response to B16-OVA cancer cell inoculation. At 10 days post-B16-OVA inoculation, the peripheral blood from C57BL/6 mice contained a population of SIINF-specific CD8⁺ T cells (Fig. 2A). B16-OVA outgrowth in wild-type mice was delayed compared with both parental B16 in wild-type mice and B16-OVA in CD8^−/− mice (Fig. 2B), suggesting that SIINF-specific CD8⁺ T cells delayed the outgrowth of B16-OVA tumors.

Despite this initial SIINF-specific CD8⁺ T-cell response, B16-OVA tumors grew progressively and killed the host. As tumors reached 100 to 168 mm³, SIINF-specific CD8⁺ T cells reached a high percentage of total CD8⁺ T cells in the tumor (mean of 22%; ranging between 9% and 30%) but were at an undetectable or low percentage in the tumor-draining lymph node, spleen, and lungs (Fig. 2C).

Previous studies showed that intravenously injected A1-R can preferentially colonize established tumors (33, 34). To determine if this occurs in our model, we intravenously injected A1-R SIINF into B16-OVA tumor-bearing mice. Eight days postinjection, the mean number of bacteria per gram of tissue was 10,000-fold greater in the tumor compared with the lungs and spleen, and 800-fold greater in the tumor compared with the liver (Supplementary Fig. S1A). A1-R SIINF colonization in tumors was confirmed by α-Salmonella immunohistochemistry (Supplementary Fig. S1B).

We analyzed the endogenous SIINF-specific CD8⁺ T-cell response to intravenously injected A1-R SIINF in mice bearing established B16 parental (non-OVA) tumors. Nine days postinjection, a high percentage of endogenous SIINF-specific CD8⁺ T cells were present in the tumor-draining lymph node, nondraining lymph nodes, spleen, and tumor. A 1-hour bromodeoxyuridine (BrdUrd) pulse showed that SIINF-specific CD8⁺ T cells were proliferating rapidly in all lymphoid organs and the tumor (Supplementary Fig. S2A). These data show that A1-R SIINF induces a systemic SIINF-specific CD8⁺ T-cell response.

Mice bearing untreated B16-OVA tumors, established for at least 14 days and reaching 100 to 168 mm³, did not have detectable SIINF-specific CD8⁺ T cells in the lymph nodes, spleen, or blood (Figs. 2C and 3A). This group of mice was defined as “Day 0 untreated” as mice were treated throughout this study at this tumor size. Treating B16-OVA tumor-bearing mice with intravenously-injected A1-R SIINF induced systemic SIINF-specific CD8⁺ T-cell proliferation in the lymphoid organs (Supplementary Fig. S2B) and generated a high percentage of SIINF-specific CD8⁺ T cells in the peripheral blood (Fig. 3A). To determine if this effect required A1-R to deliver SIINF, we used 2 independent SIINF-negative controls (A1-R SNFV and A1-R EGFP) collectively referred to as A1-R control. Mice treated with A1-R control did not have a high percentage of SIINF-specific CD8⁺ T cells in the peripheral blood (Fig. 3A), showing that A1-R is required to deliver SIINF to rescue the SIINF-specific CD8⁺ T-cell response in the periphery.

In mice bearing untreated B16-OVA tumors between 100 and 168 mm³, SIINF-specific CD8⁺ T cells reached a high percentage of total CD8⁺ T cells in the tumor (Figs. 2C and 3B). However, these SIINF-specific CD8⁺ T cells were dysfunctional as they produced little IFN-γ and TNF-α upon peptide restimulation (Fig. 3B). Treatment with A1-R SIINF or A1-R control did not significantly change the percentage of SIINF-specific CD8⁺ T cells in the tumor (Fig. 3B). However, treatment with A1-R SIINF, but not A1-R control, rescued the capacity of SIINF-specific CD8⁺ T cells to produce IFN-γ and TNF-α (Fig. 3B), indicating that the A1-R–produced SIINF rescued SIINF-specific CD8⁺ T-cell effector function in the tumor.

Mice bearing long-established B16-OVA tumors were treated with either a single injection of A1-R control, A1-R SIINF, or A1-R SIINF plus αCD8 depletion antibody (Fig. 4; Tables 1 and 2). At 25 days posttreatment with a single A1-R injection, 11 of 13 mice treated with A1-R SIINF were alive compared with 0 of 5 mice treated with A1-R control (P = 0.002) and 0 of 7 mice treated with A1-R SIINF plus αCD8 (P < 0.0005). B16-OVA tumors were rejected in 4 of 13 mice treated with A1-R SIINF. We then tested if treating mice with weekly A1-R SIINF injections could maintain a stronger SIINF-specific CD8⁺ T-cell response leading to more consistent tumor rejection (Supplementary Fig. S3 and Table 1). At 25 days posttreatment, 7 of 9 mice treated with A1-R SIINF were alive compared with 1 of 10 mice treated with A1-R control (P = 0.005). Tumors were rejected in 3 of 9 mice treated with weekly A1-R SIINF. Therefore, A1-R SIINF treatment improves mouse survival by a SIINF-specific CD8⁺ T cell–dependent mechanism, and treatment with repeated doses does not provide additional therapeutic benefit.

We tested whether tumors were fully eradicated in 4 mice that rejected the tumor. Three of the mice were originally treated with a single injection of A1-R SIINF and 1 mouse was originally treated with weekly injections of A1-R SIINF. All 4 mice were held for at least 90 days posttreatment and then depleted of CD8⁺ T cells. The tumors did not relapse (Fig. 4 and Supplementary Fig. S3), strongly suggesting that A1-R SIINF treatment completely eradicated B16-OVA tumors.

We investigated whether tumor relapse following A1-R SIINF treatment was due to outgrowth of antigen-loss variants. Cancer cells reisolated from six relapsed tumors retained the SIINF epitope as shown by staining with the K^b-SIINFEKL antibody (data not shown).

Tumor relapse with retention of the SIINF epitope indicated that the SIINF-specific CD8⁺ T-cell response eventually failed to control tumor progression. Because CD8⁺ T cells can become dysfunctional through PD-1/PD-L1 interactions following chronic antigen exposure (11, 29, 36), we evaluated PD-1 expression before and after treatment with A1-R. As B16-OVA tumors became established at the 100 to 168 mm³ tumor size, more than 90% of intratumoral SIINF-specific CD8⁺ T cells expressed a high level of PD-1 (Supplementary Fig. S4). Treatment with A1-R control or A1-R SIINF did not reverse high PD-1 expression on the majority of SIINF-specific CD8⁺ T cells in the tumor. However, there was a nonsignificant trend suggesting that A1-R SIINF treatment may partially reduce the percentage of high PD-1–expressing SIINF-specific CD8⁺ T cells in the tumor (Supplementary Fig. S4). In the same A1-R SIINF-treated mice, SIINF-specific CD8⁺ T cells from the spleen did not express a high level of PD-1 (Supplementary Fig. S4). This showed that high PD-1 expression was specific to SIINF-specific CD8⁺ T cells from the tumor and suggested a potential mechanism accounting for relapse following A1-R SIINF treatment.

We tested whether the antitumor effects of A1-R SIINF could be enhanced by blocking the immunoinhibitory PD-1 and/or CTLA-4 pathway, which has also been implicated in suppressing T-cell responses to tumors (37). B16-OVA tumors responded minimally to treatment with a combination of αPD-L1 and αCTLA-4 (Fig. 5A and Table 2). Similarly, combining A1-R SIINF with αCTLA-4 was not effective because 0 of 4 tumors were rejected. However, combining A1-R SIINF with αPD-L1 or both αPD-L1 and αCTLA-4 was synergistic because 4 of 5 tumors were rejected in each group (Fig. 5A). Tumor rejection (4 of 5 mice) in each group was significant compared with the lack of rejection (0 of 5 mice) in the αCTLA-4 and αPD-L1 treatment group (P < 0.05; Table 2). Depleting CD8⁺ T cells in mice that rejected tumors did not result in relapse, strongly suggesting that treatment with A1-R SIINF combined with αPD-L1 resulted in complete tumor eradication (Fig. 5A).

To evaluate whether tumor rejection versus relapse was dependent on the magnitude of the SIINF-specific CD8⁺ T-cell response, we measured the percentage of SIINF-specific CD8⁺ T cells in the blood of mice following the different treatments. The peripheral blood was a reliable indicator of the therapeutic response to A1-R SIINF treatment, because mice that rejected B16-OVA tumors had a significantly higher percentage of SIINF-specific CD8⁺ T cells compared with mice that had tumor relapse (Fig. 5B). Consistent with this finding, mice rejecting tumors following treatment with A1-R SIINF combined with αPD-L1 had a similar percentage of SIINF-specific CD8⁺ T cells as mice rejecting tumors following treatment with A1-R SIINF alone (Fig. 5B). These data show that the SIINF-specific CD8⁺ T-cell response generated by A1-R SIINF can be enhanced by blocking PD-L1, most likely accounting for the consistent tumor eradication observed in these mice.

To our knowledge, this is the first study reporting that therapeutic vaccination can rescue the dysfunctional endogenous tumor-specific CD8⁺ T-cell response leading to eradication of long-established tumors. Treating mice with our antigen-producing S. Typhimurium A1-R vaccine rescued the tumor-specific CD8⁺ T-cell response: proliferation was induced in the lymphoid organs and cytokine production was recovered in the tumor. Treatment with our vaccine resulted in improved mouse survival and rejection of established tumors in approximately one third of the mice. We discovered that mice rejecting tumors had a significantly higher percentage of tumor-specific CD8⁺ T cells in their blood compared with mice that relapsed following treatment, suggesting that the magnitude of the generated antigen-specific CD8⁺ T-cell response determined vaccine efficacy in our model. By combining our vaccine with αPD-L1, we enhanced the expansion of vaccine-generated CD8⁺ T cells and achieved consistent tumor rejection.

Our analysis focused on how to rescue tumor-specific CD8⁺ T cells which were specific to the SIINF epitope. By using 2 independent A1-R controls that did not produce SIINF, we determined that A1-R must produce SIINF to rescue the SIINF-specific CD8⁺ T-cell response in the periphery and tumor. However, because A1-R SIINF expressed a large portion of OVA, it is possible that A1-R SIINF activated additional OVA-specific T cells (e.g., CD4⁺ T cells specific to the OVA_323-339 MHC class II epitope) that had direct antitumor effector function or contributed to rescuing SIINF-specific CD8⁺ T cells. The requirement for A1-R to deliver SIINF strongly suggests that APCs mediating T-cell rescue presented the SIINF epitope derived from A1-R rather than from B16 cancer cells. Because A1-R SIINF treatment induced systemic SIINF-specific CD8⁺ T-cell proliferation, this suggests that A1-R delivered the SIINF epitope to APCs localized both within and outside the tumor. This is probable as A1-R SIINF persisted at a low level in normal organs following intravenous injection. Because A1-R SIINF generated a systemic SIINF-specific CD8⁺ T-cell response, it is unknown whether tumor colonization by bacteria is required for tumor eradication. This is an important question to address as clinical experience with intravenously injected S. Typhimurium VNP20009 showed that bacterial colonization of human tumors can be less consistent compared with murine experimental tumors (26, 38). Thus, if bacterial colonization of tumors is required for CD8⁺ T-cell rescue, this would support the continuation of studies investigating the mechanisms responsible for bacterial colonization of tumors and how colonization may be optimized in humans (39).

Rescuing CD8⁺ T-cell dysfunction remains a significant challenge in cancer (13, 14) and chronic viral infection (29). Patients with the strongest responses to αPD-1 have tumors expressing PD-L1 (14), which seems to be indicative of a preexisting T-cell response (12, 40). However, the objective response rate for this PD-L1⁺ patient subset is still only 36% (14), suggesting that blocking PD-1/PD-L1 may be insufficient to rescue T-cell dysfunction in many advanced tumors. In our model, long-established B16-OVA tumors were resistant to treatment with αPD-L1 and αCTLA-4 but were eradicated following treatment with A1-R SIINF combined with αPD-L1. Although the exact mechanism accounting for synergy between A1-R SIINF and αPD-L1 needs to be further investigated, our data suggest that synergy may have occurred in both the lymphoid organs and tumor. A1-R SIINF-mediated expansion of SIINF-specific CD8⁺ T cells was enhanced by blocking PD-L1, suggesting that PD-L1 expression by APCs inhibited T-cell expansion in the lymphoid organs. It is probable that A1-R treatment induced PD-L1 upregulation on APCs by inducing a strong inflammatory response (41) and/or through an LPS-mediated mechanism (40). In A1-R SIINF-treated tumors, it is likely that SIINF-specific CD8⁺ T cells were inhibited by PD-1/PD-L1 interactions as (i) the majority of SIINF-specific CD8⁺ T cells expressed PD-1 at a high level and (ii) SIINF-specific CD8⁺ T cells had rescued capacity to produce IFN-γ, a potent stimulator of PD-L1 expression on B16 (42) and human melanocytes (12) in addition to stromal cells (12).

This study used B16-OVA to determine how to rescue tumor-specific CD8⁺ T cells in long-established tumors. We targeted a model tumor-specific antigen as targeting tumor-associated self-antigens can lead to normal tissue damage and sometimes even patient death (43, 44). One might argue that the tumor eradication observed in this study was biased on the basis of the high affinity of SIINFEKL for MHC class I. However, recent data from Robbins and colleagues strongly suggest that high-affinity peptides may be the only clinically relevant peptide targets in human cancers (45). Future studies will evaluate whether our vaccination approach is similarly effective when targeting natural tumor-specific epitopes on unmodified cell lines. If successful, translating our approach could be achieved using genomic technologies (45, 46) and peptide affinity measurements/algorithms (27, 45) to identify targetable high-affinity tumor-specific peptides for each individual patient. The ease in genetically modifying S. Typhimurium creates the opportunity to construct S. Typhimurium strains that could deliver the individual or a collection of unique tumor-specific antigens present on a patient's cancer cells, thereby rescuing the tumor-specific T-cell response unique to that patient. Targeting a collection of antigens may be required to prevent relapse of tumors as antigen-loss variants (47).

Our approach to cancer vaccine development could also be used to rescue or induce T cells that are isolated from the patient and used for adoptive T-cell therapy protocols. This could be especially beneficial for treating patients with larger tumor loads, as adoptive T-cell therapy has shown great potential for treating large established tumors in experimental and clinical studies (18, 48, 49). As isolating sufficient numbers of functional tumor-specific T cells from patients remains a challenge, treating patients with our antigen-expressing bacterial vaccine combined with αPD-L1 before T-cell isolation may significantly improve the quality of T cells that could be expanded in vitro and reinfused into patients. Therefore, the capacity to rescue a dysfunctional endogenous CD8⁺ T-cell response in tumor-bearers may also create new opportunities to target tumor-specific antigens by adoptive T-cell therapy.

No potential conflicts of interest were disclosed.

Conception and design: D.C. Binder, B. Engels, P. Yu, J.M. Slauch, H. Schreiber

Development of methodology: D.C. Binder, A. Arina, P. Yu, D.A. Rowley

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.C. Binder, C. Idel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.C. Binder, B. Engels, Y.-X. Fu, T. Karrison, C. Idel

Writing, review, and/or revision of the manuscript: D.C. Binder, B. Engels, A. Arina, P. Yu, J.M. Slauch, Y.-X. Fu, T. Karrison, B. Burnette, C. Idel, M. Zhao, R.M. Hoffman, D.H. Munn, D.A. Rowley, H. Schreiber

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.C. Binder

Provided Salmonella typhimurium A1-R for the study: M. Zhao, R.M. Hoffman

The authors thank Karin Schreiber, Michael Spiotto, Ralph Weichselbaum, and Frank Wen for valuable discussions; Dr. Stephen Meredith for synthesis of the peptides; and the University of Chicago flow cytometry core for technical assistance.

This research was supported by the NIH grants P01-CA97296, R01-CA22677, and R01-CA37516 to H. Schreiber and the Graduate Training in Growth and Development grant T32 HD009007 to D.C. Binder.

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.