Inhibition of Adaptive Immunity by IL9 Can Be Disrupted to Achieve Rapid T-cell Sensitization and Rejection of Progressive Tumor Challenges

IL9 is a paradoxical cytokine, as it mediates both proinflammatory events and induction of tolerance. It is secreted by a host of proinflammatory immune cells including Th9 cells (1), Th17 cells (2), CD8⁺ Tc9 cells (3), eosinophils, mast cells, and innate lymphoid cells (1, 4–7). It is also associated with tolerogenic cells such as regulatory T cells (Treg). In this population, IL9 enhances Treg-suppressive potency in an autocrine fashion (8), while promoting T-cell tolerance via a paracrine impact upon mast cells (9–11). This wide range of action is followed by an equally wide range of pathologies involving IL9 secretion.

Most commonly, IL9 is linked to Th2 responses such as parasite expulsion and allergic airway inflammation, but it is also involved in autoimmunity and graft-versus-host disease (reviewed in ref. 6). Interestingly, IL9 can be secreted by cells that promote opposite ends of the immune spectrum. For example, proinflammatory Th17 cells can produce IL9 and exacerbate experimental autoimmune encephalitis (EAE; ref. 12), whereas IL9 secreted by Tregs renders them more suppressive and protects against EAE (8). These discrepancies may be explained by the timing of IL9 secretion in a given pathologic circumstance, and by the range of cells that express the IL9 receptor (IL9R). These include Tregs, CD4⁺ T cells, B cells, and dendritic cells (expression data from the Immunological Genome Project), as well as CD3⁺ T cells and CD11b⁺ Gr1⁺ cells from tumor-bearing mice.

IL9 also has seemingly contradictory roles in tumor biology. In many tumors, the presence of IL9 contributes to the establishment of a tolerogenic/immunosuppressive environment, or acts directly to drive tumor growth. For example, IL9 promotes the proliferation or survival of human lymphoid tumors such as Hodgkin lymphoma, acute lymphoblastic leukemia, myeloid leukemia, diffuse large B-cell lymphoma, and NK T-cell lymphoma (13–18). It also promotes the proliferation, migration, and adhesion of human lung cancer cells (19). However, IL9 has the opposite effect on melanoma biology: it slows subcutaneous growth of B16F10 as well as reduces B16 seeding in the lungs (20, 21); both groups showed that anti-IL9 opposes this effect. Adoptively transferred IL9-secreting CD4⁺ T cells (25% IL9 positivity) reduce melanoma growth, in a manner that is very similar to the transfer of Th2-polarized T cells (20). In addition, in vitro polarized OT-1 CD8⁺ T cells (Tc9), adoptively transferred to B16-OVA tumor-bearing mice, led to tumor clearance (22). However, the authors point out that 2 weeks after transfer, Tc9 cells loose IL9 expression, and instead, secrete IFNγ, suggesting a repolarization to a Tc1 phenotype, which could explain the efficient tumor clearance. In the B16 tumor model, IL9 acts on mast cells, and is not T- or B-cell dependent (20), and also has a direct effect on the lung epithelium, which then recruits dendritic cells (21).

Study of the role of IL9 in mammary carcinomas is limited to a longitudinal study of soluble factors present in sera of patients with breast cancer. Investigators found an increase in serum levels of IL9 over time in patients who later developed metastatic lesions, suggesting a relationship between IL9 and tumor progression, or tumor load (23).

In summary, the majority of observations about the role of IL9 in tumor biology suggest that it has a tolerogenic role. Here, we show that IL9 is a key factor in establishing a permissive growth environment for CT26, a colon carcinoma cell line and two murine breast cancer cells lines: TUBO cells that express Her2/neu (24) and 4T1 cells (25) that resemble aggressive, triple-negative breast cancers.

BALB/c mice were purchased from the NCI (Fredrick, MD). IL9ko mice were originally generated by Townsend and colleagues (26). The IL9ko mice used in this study were in a BALB/c background (27), and were a gift from Simon P. Hogan (University of Cincinnati College of Medicine, Cincinnati, OH). BALB/neuT mice were generated as previously described (24). BALB/neuT/IL9ko mice were created by breeding heterozygous BALB/neuT mice with IL9ko mice. F1 mice were genotyped for Neu and IL9ko allele as well as the WT IL9 allele, using the following primers: (IL9 genotyping F, gcgattcttcctgaaagcag: IL9 genotyping R, accggacacgtgatgttctt; NeomycinF, tgtcgatcaggatgatctgg). IL9^+/−/neuT mice were crossed to IL9ko, and the resulting F2 IL9ko/neuT mice were crossed to IL9ko mice to establish the colony. All mice were housed under specific pathogen-free conditions. TUBO cells are derivatives of a spontaneous mammary carcinoma in BALB/neuT mice (24) and were obtained from Dr. Guido Forni (Molecular Biotechnology Center, University of Turin, Torino, Italy). 4T1 are a mammary carcinoma line derived from 410.4 mammary tumors. BM 185 cells were derived from bone marrow from an acute lymphoblastic leukemia model, originally provided by D. Kohn (University of Southern California, Los Angeles, CA). CT26 cells were purchased through the ATCC. All were maintained in complete RPMI-1640 medium supplemented with 10% FCS, 2 mmol/L glutamine, 5 × 10⁻⁵ mol/L 2-ME, and 50 μg/mL gentamicin. Anti-IL9 mAb (MM9C1; ref. 28) and its isotype control Ab were obtained from Dr. Jacques Van Snick (Ludwig Institute, Brussels, Belgium). Depleting antibodies were obtained as follows: anti-CD4 (clone GK 1.5; BioLegend), anti-CD8α (clone 2.43; LifeSpan Biosciences) antibodies, or the corresponding rat IgG isotype control. CD4⁺ and CD8⁺ T cells were enriched using negative selection kits (Invitrogen).

Wild-type (WT) and IL9ko mice were implanted subcutaneously (s.c.) with 1 × 10⁶ TUBO, 1 × 10⁶ 4T1, or 1 × 10⁵ CT26 cells. Tumors were measured twice weekly and mice were sacrificed when the tumors reached 1 cm² or showed signs of external necrosis. Tumor volume was calculated from two perpendicular measurements using the following formula: [a² × b/2]. For CD4⁺ and CD8⁺ T-cell depletion experiments, anti-CD4 and -CD8 antibodies and the corresponding isotype control were delivered intraperitoneally (i.p.) at 125 μg each. Mice were pretreated 3 to 4 days before tumor inoculation, and then subsequently once weekly for 4 weeks.

For IL9 neutralization experiments, WT mice were injected s.c. with 2.5 × 10⁴ 4T1 cells and anti-IL9, isotype control Ab (100 μg each), or left untreated. Abs were delivered to WT and IL9ko mice through i.p. injections three times per week for 3 weeks.

Splenocytes or lymph node (LN) cells were cocultured with tumor cells on IFNγ ELISpot Kit plates from Mabtech (cat. no., #3321-2HW-Plus) for 40 hours following the manufacturer's instructions. Cell numbers were as follows: 1 × 10⁵ total splenocytes derived from 4T1-bearing mice with either 5 × 10⁴ 4T1 or TUBO cells, 5 × 10⁴ total LN cells from 4T1-bearing mice with either 2.5 × 10⁴ 4T1 or TUBO cells. CD4⁺ and CD8⁺ T cells were enriched from the spleens of naïve and 4T1-bearing WT and IL9ko mice using a negative isolation kit (Invitrogen). T cells (1 × 10⁵) were cocultured with 2.5 × 10⁴ 4T1 cells or with BM185, a nonspecific control tumor line as above. Each experimental condition was executed in biologic triplicates, which, in turn, comprised triplicate wells. The plates were imaged and evaluated by ZellNet Consulting, Inc., and results expressed as average of triplicate spots per condition. Phorbol 12-myristate 13-acetate (PMA) was used as a positive control of cell activation.

4T1 tumors were harvested from IL9ko and WT mice, formalin-fixed, and paraffin-embedded. Four-micrometer sections were incubated overnight with or without anti-CD8α antibody (Thermo Scientific; MA5-16761) overnight, followed by detection using ImmPRESS Reagent Anti-rat IgG, peroxidase (Vector Labs; MP-7404) followed by DAB substrate kit (Vector Labs; SK-4100). Slides were mounted in Permount and visualized with a Leica DMRB microscope. Images were acquired at a magnification of ×200 with a numerical aperture of 2 at room temperature with an Olympus DP71 camera using the DPController software (Olympus).

Total splenocytes were harvested from 4T1-bearing WT and from IL9ko mice that had rejected 1 × 10⁶ 4T1 cells 2 to 3 months prior, and that were rechallenged with 5 × 10⁵ 4T1 cells 1 week before the start of the assay. Splenocytes were mixed with 2.5 × 10⁴ 4T1 cells and coinjected s.c. into WT mice in the following proportions of splenocytes to 4T1 cells: 100:1, 33:1, and only 4T1. The 100:1 ratio of splenocytes contained the following number of T cells: WT+4T1 = 3.2 × 10⁵ CD4⁺ T cells and 1.2 × 10⁵ CD8⁺ T cells, IL9ko+4T1 = 4.4 × 10⁵ CD4⁺ T cells and 1.95 × 10⁵ CD8⁺ T cells. The 33:1 ratio of splenocytes contained the following number of T cells: WT+4T1 = 1.1 × 10⁵ CD4⁺ T cells and 4 × 10⁴ CD8⁺ T cells, IL9ko+4T1 = 1.5 × 10⁵ CD4⁺ T cells and 6.5 × 10⁴ CD8⁺ T cells. Mice were monitored twice a week for tumor growth and tumor growth was compared with that of 4T1 cells mixed with splenocytes derived from tumor-bearing WT. A repeat experiment was carried out using negatively enriched CD8⁺ T from tumor-bearing and naïve WT and IL9ko mice, cells were coinjected with 4T1 cells at a concentration of 25:1 (6.25 × 10⁵ CD8⁺ T cells: 2.5 × 10⁴ 4T1 cells). Each experimental cohort consisted of T cells isolated from three individual mice, and each individual isolate was injected in duplicate, bringing the total per condition to 6.

4T1 cells (1 × 10⁴) in 100 μL PBS were injected via the tail vein into WT and IL9ko mice, and mice were monitored for 18 days. During this time 50 ng recombinant IL9 was injected i.p. three times weekly. Mice were sacrificed and lungs inflated with 10% India ink in PBS, and fixed in Feket Solution (100 mL 70% ethanol, 10 mL 10% formalin, and 5 mL acetic acid). All tumor foci macroscopically visible through a dissecting microscope were counted.

Statistical significance of data was determined in most cases using the Student t test to evaluate the P value. The log-rank (Mantel–Cox) test was used to evaluate significant differences in survival. For the comparison of differences in growth curves, the tumor size was compared over days 9, 12, and 15 between groups using repeated measures ANOVA.

We previously demonstrated that neutralization of IL9 in conjunction with intratumoral CpG-ODN administration led to tumor rejection in two tolerant tumor models (29). We now used IL9ko mice (26) to further investigate the role of IL9 in inhibiting antitumor immune activity. Injection (s.c.) of TUBO cells in the flank of BALB/c (WT) mice resulted in robust tumor growth within 10 days after tumor delivery (n = 7; Fig. 1A). Similar s.c. tumor inoculations in IL9ko mice had a markedly different effect: tumors were rejected in 78% of the IL9ko mice (21 of 27 mice; Fig. 1B). In the 6 mice that developed tumors, the average onset of macroscopic tumor growth was delayed to 60 days (P = 0.0001; Fig. 1B). In addition to significantly delayed tumor onset, we also observed slower growth in IL9ko mice, resulting in 100% survival 85 days after tumor delivery, compared with 0% survival at day 30 in the WT control group (P < 0.0001; Fig. 1C). This result was not due to a direct effect of IL9 on TUBO growth (Supplementary Fig. S1A), because TUBO cells do not express the IL9R or directly secrete IL9 (Supplementary Fig. S1C and S1D). To confirm that tumor rejection was based on an immune component and not due to a systemic effect linked to IL9 deficiency, we repeated the TUBO challenge in 7 of the IL9ko mice that rejected TUBO 2 months after the initial challenge (Fig. 1C, arrow). All the mice failed to develop tumors, suggesting that these rechallenged mice had developed a memory response to TUBO cells.

We repeated this experiment with 4T1 cells, and detected palpable tumors 5 days after injection, which grew exponentially in WT mice (n = 10; Fig. 1D). In contrast, 68% (15 of 22) of IL9ko mice failed to develop or rejected tumors (Fig. 1E). Of the remaining 32% (7 of 22) of IL9ko mice that developed tumors, tumor growth was significantly slower compared with WT mice, and 75% of the IL9ko mice survived past day 50 after tumor inoculation (P < 0.0001; Fig. 1F). Again, IL9 did not promote 4T1 growth (Supplementary Fig. S1B), as these cells express very low levels of IL9R, and do not directly secrete IL9 (Supplementary Fig. S1C and S1D). To confirm that 4T1 rejection also elicited a memory response, we rechallenged 8 of the IL9ko mice that rejected 4T1 with 4T1 (Fig. 1F, arrow). As before, all the mice failed to develop tumors, suggestive of a memory response to 4T1 cells. None of the IL9ko mice in which original 4T1 tumors progressed showed any evidence of macrometastasis in the lungs, liver, or spleen (evaluated 80–100 days after injection), which is routinely observed in WT mice.

To investigate whether the tumor rejection seen in IL9ko mice was confined to mammary carcinomas, we injected the colon carcinoma cell line CT26 into both WT (n = 15) and IL9ko mice (n = 16). CT26 tumors developed in 87% (13 of 15) of WT mice (Fig. 1G), whereas tumors developed in only 25% (4 of 16) of IL9ko mice injected (Fig. 1H). Once more, the tumors that developed in IL9ko mice grew slower than those in WT mice, and 75% of IL9ko mice remained tumor free 30 days after the initial challenge (P = 0.001). At this point, 10 IL9ko and 2 WT mice were rechallenged with CT26 (Fig. 1I, arrow). No tumor growth was observed in 10 IL9ko mice and 1 WT mouse (CT26 tumor developed in the second WT mouse), again suggestive of a memory response in IL9ko mice.

To examine whether IL9 deficiency also had an effect on growth of autochthonous tumors, we used BALB/neuT mice (24), which develop aggressive, autochthonous mammary tumors in females by 16 weeks of age, and by 24 weeks in males. Double transgenic mice deficient in IL9 and expressing activated Her2/neu (IL9ko/Her2/neu) were created by breeding BALB/neuT with IL9ko mice, and selecting mice that were homozygous IL9ko and heterozygous for Her2/neu. Both groups of mice were monitored from birth to track tumor onset and growth as compared with that of BALB/neuT mice. Mice were sacrificed when one or more mammary tumors reached 10 mm², and their life span was recorded in weeks. A survival plot segregating males from females showed a significant increase in survival in Her2/neu transgenic females deficient in IL9 (P = 0.001; Fig. 2).

Evaluation of the immune composition of spleens and LNs of 4T1-bearing WT and IL9ko mice revealed increased total numbers of CD4⁺ and CD8⁺ T cells in IL9ko mice (Supplementary Fig. S2A and S2B), and a concomitant decrease in numbers of CD11b⁺Gr1⁺ cells (P = 0.005; Supplementary Fig. S2C). However, closer scrutiny revealed that any difference in total numbers was directly related to tumor size and not to IL9 status (data not shown). In addition, we evaluated Treg cell number, phenotypic characterization, and function (Supplementary Fig. S3). We found that the percentage of splenic Tregs was significantly higher in 4T1-bearing mice (P < 0.001; Supplementary Fig. S3A), whereas the percentage of Tregs in tumor-draining LN was higher in TUBO-bearing IL9ko mice (P = 0.05; Supplementary Fig. S3B). This relative increase in Tregs led us to examine Treg-related cellular markers from naïve and 4T1-bearing mice. There were no differences in Treg phenotype between naïve IL9ko and WT mice. However, fewer Tregs derived from 4T1-bearing IL9ko mice expressed ITGαE and CTLA-4 and at a lower level of expression than Tregs from WT mice (Supplementary Fig. S3C). Moreover, Tregs from IL9ko mice were less functionally suppressive than their WT counterparts (Supplementary Fig. S3D).

Because we observed that tumor rejection occurred 10 to 14 days after tumor inoculation (Fig. 1E), timing that is reminiscent of an adaptive immune response, and because this resulted in a memory response, we asked whether T cells were involved in tumor rejection in an IL9-deficient context. To test this, we inoculated IL9ko mice with 4T1 tumors, and depleted CD4⁺ and CD8⁺ T cells with mAb. Cohorts comprised 6 to 8 mice, and were treated as follows: anti-CD4, anti-CD8, both anti-CD4 and anti-CD8, isotype control antibody, or untreated. Growth was monitored over 30 days. IL9ko mice untreated and treated with isotype control antibody gave evidence of tumor rejection between days 10 and 14 after tumor injection, with rejection completed 1 week later. In contrast, CD8⁺ T-cell depletion or joint CD4⁺ and CD8⁺ T-cell depletion resulted in 4T1 tumor growth (Fig. 3A) comparable with that observed in WT mice (Fig. 1A). Finally, depletion of only CD4⁺ T cells also prevented 4T1 rejection, but with slower tumor outgrowth than CD8⁺ depletion. These results demonstrate that in an IL9-deficient milieu, both CD8⁺ and CD4⁺ T cells were involved in tumor eradication, and that neither subset alone was sufficient for cure. Interestingly, we found that T cells from tumor-bearing mice express high levels of IL9R mRNA (Supplementary Fig. S1C). Tumor sizes were tabulated 21 days after tumor injection, and confirmed that CD8⁺ depletion resulted in large tumors (averaging 700 mm³), whereas untreated or isotype treated mice harbored very small tumors (averaging 14 mm³) or no tumors at all (Fig. 3B).

These results suggest that the presence of IL9 negatively regulates T-cell function within 10 to 14 days of tumor challenge. The precise source of IL9 in the tumor microenvironment is yet to be determined. However, both 4T1 and TUBO cells do not transcribe either IL9 or IL9R mRNA, nor do they secrete IL9 in vitro, as measured by ELISA using supernatants of tumors generated in IL9ko mice and cultured for 3 days (Supplementary Fig. S1C and S1D).

To test whether the absence of recurrent tumors in IL9ko mice challenged with 4T1 reflected active T-cell immunosurveillance, we depleted CD8⁺ T cells in IL9ko mice (n = 8) that had rejected 4T1 cells 21 days prior (Fig. 3C, arrows). Tumors grew in 3 of 8 mice, and no tumor growth was evident in 5 of 8 mice. These data confirmed that active T-cell surveillance was operative in at least a portion of non-recurrences, and raised the possibility that mice without tumor growth following CD8⁺ T-cell depletion might harbor sufficient CD4⁺ T-cell immunologic memory to compensate for the CD8⁺ T-cell depletion, or that the tumor was completely eradicated.

To confirm that tumor rejection was due to the sensitization of a tumor-specific immune component, we harvested total splenocytes and lymphocytes (from tumor-draining LNs) from 4T1-bearing WT and IL9ko mice 14 days after tumor injection. These cells were cocultured with either 4T1 (target tumor) or TUBO (negative control tumor) cells to measure the number of cells that were activated in a tumor-specific manner, using the number of IFNγ⁺ spots as a reporter of activation (Fig. 4A). The number of IFNγ⁺ spots present in WT splenocytes or lymphocytes cocultured with 4T1 cells was near the levels of negative control (TUBO) tumor. In contrast, IL9ko-derived splenocytes and lymphocytes were activated in a tumor-specific manner. Moreover, IL9ko-derived splenocytes (6.5-fold increase) and lymphocytes (9.7-fold increase) showed a significantly higher degree of activation in the presence of 4T1 as compared with WT cells (P = 0.001; Fig. 4B).

Because we observed that T-cell depletion resulted in tumor growth in IL9ko mice, we repeated the ELISpot assay with isolated CD4⁺ T cells and CD8⁺ T cells. Total numbers of CD4⁺ T cells and CD8⁺ T cells in spleens of naïve and 4T1 tumor-bearing mice are shown in Supplementary Fig. S2. These cells were cocultured with 4T1 or BM185 cells, a BALB/c background murine leukemia cell line used here as a negative control (Fig. 4C). Again using the number of IFNγ⁺ spots as a reporter of T-cell activation, we found a 3.6-fold increase in the number of activated CD8⁺ T cells in the population derived from tumor-bearing IL9ko mice as compared with their WT counterparts: an average of 239 of IFNγ⁺ spots in cells from IL9ko mice, versus 67 spots in WT cells (P = 0.007; Fig. 4D). Furthermore, activation of CD8⁺ T cells was 4T1 specific, because there were no measurable IFNγ⁺ spots when CD8⁺ T cells were cocultured with BM185 cells. PMA was used as a positive control to confirm that the WT CD8⁺ T cells were capable of activation. A similar experiment using CD8⁺ T cells from TUBO-bearing mice yielded similar results (data not shown). CD4⁺ T cells tested under identical conditions produced IFNγ⁺ spots only if exposed to PMA. The lack of CD4⁺ T-cell activation is consistent with the absence of MHC class II–expressing cells: T cells were cultured with MHC II–negative tumor cell lines (verified by FACS; data not shown), in the absence of syngeneic antigen-presenting cells.

Having observed that CD8⁺ T cells were key effectors in tumor eradication, and that they were activated in a tumor-specific manner, we sought to verify the presence of CD8⁺ T cells in 4T1 tumors that were in the process of being rejected. WT and IL9ko mice were injected with 4T1 cells, and tumor growth monitored. After 7 days, the tumors in WT mice were robustly growing (an average of 5 mm²; Fig. 1D), whereas tumors growing in IL9ko mice were decreasing in size (2–3 mm²; Fig. 1E). Tumors were harvested at this point, formalin-fixed, and paraffin-embedded. Staining with anti-CD8 revealed a population of CD8⁺ cells arranged mostly in groups at the margins of shrinking 4T1 tumors in the IL9ko mice (Fig. 4E). No CD8⁺ cells were observed in tumors growing in WT mice, even though our ELISpot analyses revealed 4T1-specific CD8⁺ T cells in the spleens of WT 4T1-bearing mice (Fig. 4D).

We used Winn assays to test whether activated splenocytes from IL9ko mice that had rejected 4T1 tumors were capable of inducing tumor rejection in WT mice. IL9ko mice that rejected 4T1 tumors were rechallenged with 4T1 cells to boost the levels of memory cells. Total splenocytes were then harvested from the rechallenged mice, and mixed with 4T1 cells before injection in the flanks of WT mice. WT mice were segregated into three cohorts, which received increasing numbers of splenocytes, holding the number of tumor cells constant: (i) no splenocytes added, (ii) 33:1, and (iii) 100:1. Splenocytes from 4T1-bearing WT mice were used in the same proportions as a control. Tumor growth was monitored twice weekly, revealing tumor growth in WT mice that received only 4T1, and complete abrogation of tumor growth in all the mice that received IL9ko-activated splenocytes (Fig. 5A). Coinjection of 4T1 and splenocytes derived from 4T1-bearing WT mice grew similarly to the mice that received only 4T1. These results show that splenocytes derived from mice that had rejected 4T1 included cells that were capable of eradicating 4T1 tumor cells in a WT context.

We repeated the experiment with negatively isolated CD8⁺ T cells to verify our finding that depletion of CD8⁺ T cells enables tumor growth in IL9ko mice (Fig. 3A). CD8⁺ T cells were harvested from spleens of the following groups of mice: (i) WT bearing 4T1 tumors, (ii) naïve WT, (iii) IL9ko bearing 4T1 tumors, and (iv) naïve IL9ko. CD8⁺ T cells were mixed with 4T1 cells in a ratio of 25:1 injected into WT mice, and tumor growth was monitored. As observed before (Fig. 5A), cells derived from tumor-bearing IL9ko mice prevented 4T1 growth in 6 of 6 mice (Fig. 5B). Cells isolated from both IL9ko and WT naïve mice permitted 4T1 growth. Surprisingly, half of the mice treated with CD8⁺ T cells from tumor-bearing WT mice prevented tumor growth, and half enabled tumor growth. These findings echo our observations from the ELISpot assays (Fig. 3C), which show that CD8⁺ T cells derived from IL9ko mice have 3.6-fold more tumor-reactive cells than their WT equivalents.

If IL9 is an important factor in tumor development, then neutralizing IL9 in WT mice with nascent 4T1 should lead to slowed tumor growth or tumor rejection. We inoculated WT mice with 4T1 cells and separated them into three cohorts: (i) untreated, (ii) treated with neutralizing anti-IL9, and (iii) treated with isotype control (Fig. 6A) and tumor growth was monitored. Mice treated with anti-IL9 antibody showed significant delay in tumor growth (days 0–15) as compared with untreated mice (P < 0.0001), and also when compared with isotype control antibody (P = 0.03; Fig. 6B). The difference in growth between anti-IL9 and untreated remains highly significant throughout the 3 weeks (P < 0.0001). However, due to the high degree of variation in the isotype control cohort, there was no measurable significance in 4T1 growth in latter time points.

Because we observed that the few IL9ko mice that developed 4T1 tumors never had evidence of macro metastases in the lungs, we used an experimental metastasis model to verify whether IL9 plays a role in 4T1 seeding in the lung. Tail vein injections of 1 × 10⁵ 4T1 cells led to metastatic lesions in WT mice (an average of 87 foci per lung), whereas the majority of IL9ko mice did not develop visible metastases (an average of 3 foci per lung; P = 0.0001; Fig. 6B and C). Interestingly, addition of recombinant IL9 led to enhanced 4T1 seeding in the lungs of IL9ko mice (an average of 13 foci per lung; P = 0.043; Fig. 6B and C).

Perhaps the role of IL9 in tumor biology can be inferred by what is known about its role in Th2-type diseases such as parasitic infections, allergy, and asthma. In such cases, IL9 is rapidly, and transiently, expressed by CD4⁺ T cells (30), CD8⁺ T cells (3), dendritic cells (31), and innate lymphoid cells (32) in response to activating stimuli. It can therefore be considered a marker of early T-cell activation. We hypothesize that upon tumor inoculation, the host immune system is activated and responds by producing IL9. Although several IL9-secreting cells have been identified in the tumor microenvironment, at present, it is unclear which are the first responders. We have observed that in IL9-deficient mice TUBO, 4T1 and CT26 cells grow and are then resorbed around day 7 to 10. This leads us to suggest that IL9 is involved in an immediate, early establishment of a tolerogenic milieu, which hampers the development of an adaptive immune response against a tumor challenge. This is reminiscent of CTLA-4 and PD1, receptors that are expressed on activated T cells, and that, when engaged, act as checkpoints that lead to a dampening of an antitumor immune response (33). We hypothesize that the early secretion of IL9 in the tumor microenvironment may prevent the activation of adaptive immunity, whereas the role of CTLA-4 and PD1 is to curtail an adaptive immune reaction.

The presence of IL9 does not completely inhibit the formation of tumor-specific T cells in WT mice. In our ELISpot assays, we observed a modest tumor-specific activation of CD8⁺ T cells derived from tumor-bearing WT mice. In addition, we observed that when total splenocytes derived from tumor-bearing WT mice are coinjected with 4T1 cells, all the mice developed tumors, whereas when CD8⁺ T cells were enriched and coinjected with 4T1, there was a heightened chance of tumor rejection. Therefore, the lack of effector function is not due to lack of tumor-specific CD8⁺ T cells in tumor-bearing WT mice. Indeed, administration of a higher number of enriched CD8⁺ T cells leads to tumor rejection even in WT mice. On the other hand, the large proportion of myeloid-derived suppressor cells (MDSC) present in the spleens of WT 4T1-bearing mice (averaging 40% of MDSCs in a spleen from a 5-mm² tumor; ref. 34) could be the causative factor. MDSCs impede the activation of splenic T cells unless the latter are isolated experimentally. Because IL9ko mice reject tumors approximately 10 days after tumor injection, they do not develop the same immunosuppressive milieu, and therefore tumors are rejected even when total splenocytes are transferred.

As mentioned above, IL9 and the IL9R are expressed by both innate and adaptive immune cells. We are in the process of identifying the timing and cellular source of IL9 during a nascent antitumor immune reaction, using an IL9^Cre reporter system previously used in the study of airway inflammation (32, 35). However, our work shows that in an IL9-deficient host, tumor specific CD8⁺ T cells are generated, and that these are sufficiently activated to be capable of eliminating a tumor challenge in WT mice. Others have shown that in WT hosts, CD8⁺ T cells can be polarized to secrete IL9 (3) and that these cells are less cytotoxic in vitro than conventional cytotoxic T cells (CTL; refs. 3, 21), expressing lower levels of IFNγ, granzyme B, and perforin. These findings support our observations that more IL9ko CD8⁺ T cells express IFNγ and are more cytotoxic than their WT counterparts. However, three discrepancies between our findings and those of others who report that IL9 promotes antitumor activity remain to be explained. First, addition of recombinant IL9 increases experimental 4T1 metastasis, whereas Lu and colleagues show that anti-IL9 increases B16 experimental metastases. Second, in our BALB/c-based tumor models, tumors are rejected in IL9-deficient mice; however, Lu and colleagues found that coadministration of tumor-antigen primed DCs with in vitro polarized Tc9 cells led to tumor shrinkage and cure, which was partially reversed by anti-IL9 antibody. This antitumor activity was exclusively found in tumor models with strong antigen specificity (B16-OVA and MC38-gp100). One explanation was suggested by the authors, who observed that in vivo Tc9 cells repolarized to a more classical IFNγ-expressing CTL profile; these cells were also long-lived explaining their enhanced antitumor activity (22). Therefore, their antitumor activity in vivo may not be IL9 dependent. Most recently, Vegran and colleagues (36) found that TH9-polarized CD4⁺ T cells, coadministered with IL1β, led to slowed tumor growth, reduced tumor foci in the lungs, and increased survival in C57BL/6-based tumor models. However, in this case also, IL9 was not the driving factor in tumor rejection, because only IL21 neutralization reversed the combined antitumor effect of TH9 cells + IL1β. Taken together, the discrepancy between our findings: who see IL9 as a tolerizing agent, and those who conclude that IL9 leads to tumor rejection, may be partly mouse strain dependent. C57BL/6 mice are intrinsically Th1 polarized, whereas BALB/c mice are Th2 polarized (37). Because IL9 belongs to the overarching Th2 phenotype, its removal may lead to a shift to Th1 profile in BALB/c mice, which could lead to an enhancement of tumor rejection. We are in the process of confirming whether IL9ko mice in a C57BL/6 also reject tumor challenges.

The first step in exploring whether anti-IL9 treatment can be used as an adjuvant to immunotherapy regimens, is to demonstrate an effect in a WT host. We show here that the administration of anti-IL9 indeed led to a significant delay in tumor growth of a very aggressive mammary carcinoma. However, we observed that the effect of IL9 diminished as the tumors increased in size. It is possible that neutralization of IL9 becomes progressively more difficult as the tumor size increases because there is an increase in IL9-secreting suppressor cells, as well as poor penetration of the Ab into the tumor. We are actively exploring the pharmacokinetics of anti-IL9 neutralization as well as other strategies for blocking the IL9–IL9R signaling axis more efficiently. Interestingly, IL9 deficiency also plays a role in the development and growth of autochthonous Her2/neu tumors. Although the increase in survival is modest, it is significant, especially because this tumor model is extremely aggressive, where female Her2/neu–positive mice develop multiple mammary carcinomas by 4 months. Nevertheless, we do not know whether this effect is due to the effect of IL9 on the immune component (for example by reducing the efficiency of immunosurveillance) or the neoplasm itself.

To summarize, we have discovered that IL9 is involved in the promotion of a tolerogenic environment during the time period that an adaptive immune response against a tumor is primed. IL9 deficiency leads to a robust tumor rejection, involving a strong sensitization of tumor-specific T cells, which remain operative as effectors rather than tolerized or otherwise subverted. Activated antitumor cells can be transferred to WT mice and lead to tumor rejection. Finally, treatment with IL9-neutralizing Abs significantly slowed the growth of breast cancer cell lines in WT mice. Conversely, addition of recombinant IL9 increased the number of 4T1 foci in the lungs of IL9ko mice.

These data demonstrate that among its many other actions, IL9 inhibits the activation of adaptive antitumor immunity. IL9 ablation enables CD8⁺ and CD4⁺ T cells to become promptly sensitized to tumor Ag within the formidable microenvironment of a growing tumor, where only IL9 is missing, Remarkably, it was unnecessary to perform vaccine or adjunct cytokine maneuvers to achieve this end. Blockade of the IL9–IL9R axis is clearly a promising target for potentiation of immunotherapy. Even regional blockade may prove effective for T-cell sensitization. Serendipitously, a phase II clinical trial evaluating humanized IL9 neutralizing Ab (MEDI-528) for asthma relief has recently ended, revealing no improvement over conventional therapy, but also (and most importantly) no significant side effects (38). We are currently examining the usage of anti-IL9 as an adjuvant in combinatorial regimens such as anti-IL9 with chemo- or immunotherapeutic agents.

No potential conflicts of interest were disclosed.

Conception and design: D.B. Hoelzinger, P.A. Cohen, S.J. Gendler

Development of methodology: D.B. Hoelzinger, P.A. Cohen, S.J. Gendler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.B. Hoelzinger, A.L. Dominguez, S.J. Gendler

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.B. Hoelzinger, P.A. Cohen, S.J. Gendler

Writing, review, and/or revision of the manuscript: D.B. Hoelzinger, P.A. Cohen, S.J. Gendler

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.L. Dominguez

Study supervision: P.A. Cohen

The authors thank Cheryl Myers and Noweeda Mirza for constructive discussions and insight. This work benefited from the technical support provided by Kevin Pollock, the Flow Cytometry Core, the Immunohistochemistry Core, and by Tammy Brehm-Gibson from the Immunology Core. The authors also thank Amylou Dueck for statistical analyses. In addition, they thank the Mayo Clinic Natalie Schafer Animal Care attendants for excellent animal care.

This work was supported by the NCI: CA155295-01-03A (S.J. Gendler), RO1 A108984 (P.A. Cohen), and 5P50 CA102701 (P.A. Cohen and S.J. Gendler), and the Mayo Foundation.

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.