|Home | About | Journals | Submit | Contact Us | Français|
CD4+25+Foxp3+ regulatory T cells (Treg) play a critical role in the induction of tolerance to tumor associated antigens and suppression of anti-tumor immunity. How Treg are induced in cancer is poorly understood. We reported previously that Treg are significantly elevated in the peripheral blood of patients with pancreas cancer and that in a murine pancreas cancer model induction of Treg appears to be TGF-Beta dependent. Here we provide additional evidence that Treg are increased locally within the tumor microenvironment by a mechanism that appears dependent on TGF-Beta Receptor expression and the presence of tumor derived TGF-Beta. The murine pancreas cancer cell line Pan02 produces high levels of TGF-Beta both in vitro and in vivo. In contrast, the esophageal murine cancer cell line, Eso2, does not. Immunohistochemical staining of Foxp3 in explanted tumors shows an identifiable population of Treg in the Pan02 (TGF-Beta positive) tumors but not Eso2 (TGF-Beta negative). Naïve CD4+25−Foxp3− T cells, when adoptively transferred into Rag−/− mice, are converted into Foxp3+ Treg in the presence of Pan02 but not Eso2 tumors. Induction of Treg in Pan02 mice is blocked by systemic injection of an anti-TGF-Beta antibody. If Rag−/− mice are instead reconstituted with naïve CD4+25− T cells expressing a mutated TGF-Beta receptor, induction of Foxp3+ Treg in Pan02 bearing mice is blocked. Collectively these observations further support the role of TGF-Beta in the induction of Treg in pancreas adenocarcinoma.
Adenocarcinoma of the pancreas, a highly invasive and metastatic disease, is the fourth leading cause of cancer mortality in the United States. Like many tumors, pancreas cancer is weakly immunogenic and is associated with a poor anti-tumor immune response 1;2. This in part is due to the chronically immunosuppressed state that pancreas cancer patients exist. Their serum level of the immunosuppressive cytokine, TGF-Beta 3;4, can as much as 4.5 times that of normal healthy subjects; a finding that in one study was associated with a worse overall survival5. Furthermore, pancreas cancer patients have reduced number and function of circulating dendritic cells6;7. What is interesting is that post-adjuvant treatment dendritic cell numbers and function recover while there also is a reciprocal drop in systemic TGF-Beta levels. Taken together it suggests that logically this state of immunosuppression is detrimental to the host. Adding to this there is now evidence that pancreas cancer patients have elevated numbers of the suppressor T cell, Treg8-10. As well, elevated numbers of Treg in ovarian cancer is associated with less favorable outcomes11. Yet little is known for how modern medical therapeutics can be used to reverse this immunosuppressive state. To do so, it would be necessary to better understand the mechanisms by which tumors suppress anti-tumor effector responses. The elevation of Treg in cancer suggests one mechanism by which tumors induce immunosuppression.
Regulatory T cells (Treg), identified phenotypically as CD4+25+Foxp3+ T cells, are a unique subset of immunosuppressive T cells which prevent autoimmunity in thymectomized mice 12. They can be subdivided into two main categories: naturally occurring Treg (nTreg) and peripherally induced Treg (iTreg). Both types express Foxp3, a gene encoding the forkhead/winged helix transcription factor that in humans is responsible for Immunodysregulation, Polyendocrinopathy, Enteropathy, X-linked (IPEX) syndrome and is believed to be a key regulator of Treg development 13. Where nTreg and iTreg differ is in their origin, antigen specificity, and mechanism of suppression 14. Natural Treg (nTreg) are thymus derived, constitutively express CD25, and their mechanism of suppression is cell contact dependent 15. Inducible Treg (iTreg), also known as adaptive Treg, develop in the periphery from CD4+25−Foxp3− precursor T cells, that variably express CD25, and suppress via a cell contact independent, cytokine dependent mechanism 16;17. Early evidence shows that TGF-Beta plays an important role in the de novo induction of CD4+25+Foxp3+ T cells18-20.
Natural and inducible Treg cells can be protective to the host during states of autoimmunity, infection, or transplant rejection. In malignancy, however, Treg play a critical role in progression of tumor growth and suppression of antitumor immune responses11;21-24. A few preliminary reports suggest that iTreg play a more significant role in suppressing anti-tumor immunity than nTreg25-27. Furthermore human trials that non-selectively neutralized both nTreg and iTreg function using CD25 directed immunotoxins resulted in diseases of autoimmunity and only transient partially reduced populations of Treg28-30. The induction of autoimmunity in these hosts likely results from the unintended neutralization of nTreg function as well as the tumor iTreg. What these findings at least suggest is that selective depletion of certain Treg subtypes may be a more efficient and possibly effective way of overcoming Treg suppression in cancer patients. These techniques, however, cannot be developed until we better understand how tumors induce Treg.
In this study, we demonstrate in a murine pancreas cancer model that a high prevalence of tumor induced Treg is associated with elevated systemic levels of TGF-Beta. Using Rag1−/− mice reconstituted with naïve CD4+25− T cells we demonstrate that tumor-induced Treg conversion is blocked by the systemic use of an anti-TGF-Beta antibody. These tumor-derived Treg have in vitro Treg suppressor function identical to that of nTreg. When tumor bearing mice are reconstituted with naïve T cells expressing a mutated TGF-Beta receptor II (dnTGF-BetaRII) these naïve T cells do not undergo conversion despite the systemically elevated presence of TGF-Beta. Collectively these findings suggest that murine pancreas tumors induce Treg prevalence through a mechanism that appears to be TGF-Beta dependent. Potentially this could represent one mechanism by which iTreg can more selectively be blocked in an effort to promote anti-tumor immune responses while not disrupting nTreg activity.
All animal study protocols were reviewed and approved by the institutional review committee. Six-week old female C57BL/6 mice and C57BL/6 Rag1−/− mice were purchased from Jackson Laboratories. The dnTGF-BetaRII mice were a generous gift from Dr. RA Flavell of Yale University 31. The mice were housed and bred at Washington University School of Medicine Animal Facilities in specific pathogen-free conditions. Animal care and all procedures were in accordance with the guidelines of the animal care committee. The mouse pancreas adenocarcinoma cell line Pan02 is syngeneic to C57BL/6 and was obtained from DCT Tumor Repository (NCI-Frederick Cancer Research and Development Center, Bethesda, MD). The C57BL/6 syngeneic esophageal tumor cell line, Eso-2 was a gift from Dr. R. Battafarano. Both cell lines were grown in RPMI culture medium (Mediatech, Manassas, VA) supplemented with 10% (v/v) fetal bovine serum (FBS, Life Technologies, Carlsbad, CA), and 1% glutamine and antibiotics.
The spleen and LN from C57BL/6 female mice were harvested and homogenized. The splenocytes were briefly treated with ACK buffer to remove red blood cells. Tumor tissue was digested in a buffer containing 1mg/ml collagenase, 2.5 U/ml hyaluronidase and 0.1mg/ml DNase for 30 minutes. The suspension was then separated from tissue debris by filtration through 40 μm cell strainer and centrifugation on Histopaque®-1083. All cells were then washed twice in PBS containing 0.5% FBS and passed again through 40 micron nylon mesh to obtain single cell suspensions. CD4+25− and CD4+25+ Treg populations were separated using a mouse regulatory T cell isolation kit (Miltenyi Biotech, Auburn, CA) per manufacturer’s instructions. Briefly, CD4+ cells were separated from other cells by negative selection with an antibody cocktail provided by the manufacturer. The cells were then passed over a magnetic column and the CD4+ cells were collected. CD25+ cells were then selected by positive selection using anti-CD25 magnetic microbeads and then passed over a magnetic column. The initial flow through was collected as CD4+25− T cells and the bound fraction was eluted from the column and collected as the CD4+25+ T cells. Purity (>93%) of each fraction was checked using FACSCalibur (Becton Dickinson, Franklin Lakes, NJ) and antibodies against mouse CD4 (Cy-chrome labeled clone L3T4), CD25 (FITC labeled clone 7D4), as well as compatible isotype control antibodies (BD Biosciences, San Jose, CA).
In vitro conversion of naive T cells to Treg has been described elsewhere 18. Briefly, freshly isolated C57BL/6 naive CD4+25− T cells were cultured in 6 well plates with anti-CD3 (0.5 μg/ml), and either irradiated APCs or soluble anti-CD28 (2ug/ml) in the presence or absence of 0.02, 0.2, 2, or 20 ng/ml of TGF-Beta1 (R&D Systems, Minneapolis, MN). The cells were kept in these culture conditions for 3 days, after which they were washed and subjected to further analysis. Cytokine assay for TGF-Beta production by our tumor cell lines was done using a commercially available ELISA kit (Biosource International, Carlsbad, CA). Pan02 and Eso2 were both plated at a density of 1×106 cells per well, and kept in culture for 7 days, after which the supernatant was removed and processed according to manufacture’s instructions for TGF-Beta specific ELISA assay. Measurement of TGF-Beta in mice sera was preformed according to the manufacture’s instructions (Biosource International, Carlsbad, CA). In brief, C57BL/6 mice were injected with 0.25×106 Pan02 or Eso2 cells and after 3 weeks serum was collected. Samples were stored at 70°C until ready for processing. The sera were diluted to 1/40 for ELISA.
Freshly isolated tumor infiltrating lymphocytes were labeled using antibodies against mouse CD4 (Cy-chrome labeled clone L3T4), CD25 (FITC labeled clone 7D4) and were analyzed with a FACS Calibur® Flow Cytometer (Becton Dickinson, Franklin Lakes, NJ). At least 100,000 live events were collected per sample. The analyses were conducted using Flojo (Tree Star, Ashland, OR).
RNA was isolated from 5×106 lymphocytes using the RNeasy®Protect Mini Kit (Qiagen, Valencia, CA). Synthesis of cDNA was performed on 100ng of RNA using 8 Oligo (dT)12-18 (SuperScript First-Strand Synthesis System for RT-PCR, Invitrogen, Carlsbad, CA) per manufacturer’s instructions. Real-time PCR for Foxp3 normalized to hypoxanthine-guanine phosphoribosyl-transferase (HPRT) was performed on a 9600 thermal cycler and analysis of the data was performed using Sequence Detection System 5700 software (Applied Biosystems, Foster City, CA). TaqMan® Universal PCR Master Mix and TaqMan® Primer with non-fluorescent quencher probes for Foxp3 (Cat # Mm 00475156) and HPRT (Cat # Mm00446968) were purchased from Applied Biosystems.
Cells were cultured in DMEM medium supplemented with 10% FBS (Aleken Biologicals, Nashville, TN) HEPES, non-essential amino acids, sodium pyruvate, 50 μM β-ME, and penicillin/streptomycin/L-glutamine. αThy1.1-PE (clone OX-7) and αCD4-PE-Cy5.5 (clone RM4-5) were purchased from BD Pharmingen (Franklin Lakes, NJ) and Biolegend (San Diego, CA) respectively. For the in vitro suppression assays, 50,000 CFSE-labeled CD4+ CD25−Thy1.1+ T cells were co-cultured in 96-well plates with 200,000 irradiated thy1.2+ splenocytes, 0.1 μg/ml and αCD3, with or without 50,000 CD4+ CD25+ Thy1.2+ cells for 72 hours. Cells were subsequently harvested, stained for Thy1.1 and CD4, and then analyzed by flow cytometry. The gates were set for Thy1.1 and CD4 and a histogram constructed based on CFSE positivity.
Six-week old Rag1−/− mice received i.p. injections of 200 μl HBSS with or without 5×106 freshly isolated CD4+25− naive T cells from either normal C567BL/6 mice or dnTGFBetaRII transgenic mice. Two days after reconstitution, the mice received subcutaneous injections in the left outer thigh of 0.5×106 Pan02 or Eso-2 tumor cells. Seven weeks later, the mice were sacrificed and the spleen and tumor draining lymph nodes (TDLN) were harvested and analyzed for expression of CD25 and CD4 by FACS, and for Foxp3 by real-time PCR.
Tumors were implanted by subcutaneous injection of 1.0×106 PanO2 or ESO2 cells in the hind limb of C57BL/6 mice. At four weeks post implantation the tumors and TDLN were resected and fixed overnight in buffered formalin (Sigma, St. Louis, MO). Prior to staining antigen retrieval was performed by incubating sections in a decloaking chamber (Biocare Medical, Concord, CA) for 10 minutes at 90°C in Reveal™ buffer. Sections were stained with a monoclonal rat anti-mouse-Foxp3 IgG (FJK-16s, 50 μg/ml, eBioscience, San Diego, CA) followed by Alexa Fluor 488 conjugated goat anti-rat antibody IgG (A11029, Invitrogen, Carlsbad, CA). Sections were cover slipped with Vectashield Hard Mounting Media containing DAPI (Vector Labs, Burlingame, CA). Sections were examined using an Olympus BX61 microscope and captured digitally with an Olympus DP70 camera (Olympus, Center Valley, PA). For the CD4+ and Foxp3+ co-localized tissue sections, Pan02 tumors were section and immediately embedded in base molds (Ted Pella, Redding, CA) containing Tissue-Tek OCT medium (Sakura, McGaw Park, IL ) then frozen over liquid nitrogen. Frozen tumor sections were cut at 10μm and mounted on SuperFrost Plus slides (Fisher, Pittsburgh, PA) then stored at −80°C until use. Tumor sections were fixed in 1% paraformaldehyde for 5 minutes at room temperature then washed several times in 1X phosphate-buffered saline (PBS). The slides were blocked for 30 minutes in Serum Free Protein Block (DAKO) at room temperature and then stained with rabbit anti-mouse Foxp3 (Abcam, Cambridge, MA) and rat anti-mouse CD4 (eBioscience, San Diego, CA) primary antibodies. Sections were then stained with AlexaFluor 488 goat anti-rabbit and AlexaFluor 555 goat anti-mouse IgG secondary antibodies (Invitrogen, Carlsbad, CA). Confocal images were visualized on an Axiovert 100M microscope equipped with a LSM 510 confocal system (Axiovert, San Marcos, CA).
In vivo blockade of TGF-Beta was achieved using a multi-species monoclonal antibody toward TGF-Beta1,2,3 (Clone 1D11, IgG1, R&D Biosystems, Minneapolis, MN). On day 0 Rag1 −/−mice were reconstituted with 5×106 naive T cells given intraperitoneally. On day 2 the mice were injected subcutaneously with 0.25×106 Pan02 cells into the lateral left thigh. After two weeks the mice received intravenous injections via the tail vein of 10μg of anti-TGF-Beta without azide, low endotoxin IgG1 isotype control (BD Biosciences, San Jose, CA). Treatments were given either weekly (Q7) or every 4 days (Q4). The treatments lasted for a total of four weeks at which time the mice were sacrificed and their TDLN analyzed for expression of Foxp3 by real time PCR. Tumor growth curves were based on weekly measurements taken with calipers, in two separate axes and were confirmed by two independent observers. The tumor volume was defined as the multiplication of the diameters taken in two separate axes. Tumor burden was defined as the percentage weight of the tumor relative to the total body weight (tumor weight ex vivo ÷ total body weight at sacrifice).
We first studied whether TGF-Beta induced conversion of naive T cells into Treg in vitro. Freshly isolated CD4+25− naive T cells were purified from the spleens of normal C57BL/6 mice. The purity of these fractions were found to be >93% by FACS (data not shown). The cells were cultured with soluble anti-CD3 and either anti-CD28 or irradiated antigen presenting cells (APCs) in the presence or absence of TGF-Beta. Cells were then purified from culture after 72 hours and then analyzed for expression of Foxp3 by real time PCR. In prior studies, we validated that induction of Foxp3 expression is a marker of Treg conversion and function through the use of T cell suppression assays 8. For this reason functional assays of Treg suppression were not routinely performed. Figure 1 shows that Foxp3 expression, as expected, was significantly higher in freshly isolated regulatory T cells (CD4+25+) when compared with naive non-regulatory T cells (CD4+25− ). Foxp3 expression was induced in cultures of naive T cells treated with anti-CD3/CD28 and TGF-Beta. The upregulation of Foxp3 by TCR crosslinking did not differ whether costimulation was provided by irradiated APCs or soluble anti-CD28. Neither TGF-Beta nor TCR engagement alone was enough to induce Foxp3 expression in CD4+25− T cells. The effect of TGF-Beta was concentration dependent, where the highest expression of Foxp3 was observed at a TGF-Beta concentration of 2ng/ml.
Several clinical studies have demonstrated that pancreas cancer patients have significantly elevated levels of the immunosuppressive cytokines TGF-Beta and IL-10 in their systemic circulation 3;4;6. We investigated the possibility of whether pancreas cancers induce Treg conversion in vivo through secretion of TGF-Beta. Pan02, a murine pancreas adenocarcinoma cell line, is a highly tumorigenic cell line derived from 3-methylcholanthrene (3-MCA)-induced tumors in C57BL/6 mice. ELISA studies of culture medium taken from Pan02 cells demonstrated significantly elevated levels of TGF-Beta when compared with media alone (Fig. 2a). An unrelated murine cell line of esophageal adenocarcinoma, Eso2, exhibited no difference in TGF-Beta levels when compared with media alone. Neither cell line produced significant levels of IL-10 (data not shown). Because with successive cultures cell lines can change phenotype, we also analyzed Pan02 and Eso2 production of TGF-Beta in vivo. Pan02 and Eso2 tumors were established in the left lateral thigh of C57BL/6 mice and allowed to grow for three weeks. At the time of sacrifice all mice exhibited palpable tumors of relatively similar sizes (data not shown). Serum levels of TGF-Beta in Pan02 mice were significantly elevated above levels measured in non-tumor bearing mice (Fig. 2b). There was no statistically significant difference in TGF-Beta serum levels between Eso2 and non tumor bearing mice.
In prior studies, we have shown that regulatory T cells are elevated in the peripheral blood and TDLN of patients and mice with pancreas cancer 8;21. Based on these observations we hypothesized that Treg were actively being recruited to the tumor microenvironment. To test this we took tumors explanted from Pan02 and Eso2 bearing mice and stained them for Foxp3 using immunofluorescence. Hematoxylin and Eosin (H&E) staining of Pan02 and ES02 demonstrated standard morphology of murine pancreas and esophageal adenocarcinoma (Fig. 3A,B). Staining for Foxp3 in Pan02 tumors demonstrated a high number of Foxp3+ cells within the Pan02 tumors (Fig. 3C). These finding are consistent with our real-time PCR and FACS analysis of TDLN taken from Pan02 bearing mice. Tumors established with the non-TGF-Beta secreting Eso2 cell-line showed no tumor infiltrating Foxp3+ cells, suggesting that the increased prevalence of Treg in Pan02 tumors is possibly related to tumor dependent expression of TGF-Beta (Fig. 3D). To further phenotype the Foxp3+ tumor infiltrating lymphocytes we did co-localization of Foxp3 and CD4. Figure B demonstrates the Pan02 tumors have strong positivity for CD4+ tumor infiltrating lymphocytes, some of which co-localize with Foxp3. Taken together, this suggests that populations of tumor infiltrating lymphocytes in Pan02 are CD4+Foxp3+ T cells.
The induced populations of Treg in Pan02 tumors were found to have functional Treg suppressor activity similar to that of natural Treg (Figure 4). To demonstrate this, we labeled freshly naïve CD4+CD25−Thy1.1+ effector cells (Teff) with CFSE and cultured them in the presence of CD4+25+Thy1.2+ regulatory T cells (Treg) derived from the spleen and tumors of Pan02 challenged mice. FACS was used to analyze CFSE staining with gating for CD4+Thy1.1 Teff cells. These experiments demonstrate that tumor induced Treg are capable of suppressing effector T cell proliferation, a function that likely impairs the anti-tumor immune response within the tumor microenvironment.
An increased prevalence of Treg has been demonstrated within the peripheral blood, TDLN, malignant ascites, and effusions of cancer patients 8;9;11;32;33. It is unclear whether this is mediated by recruitment or expansion of existing Treg, or whether the tumor induces de novo generation of Treg from peripherally circulating naive CD4+25− T cells. We established an adoptive transfer model using Rag1-knockout (Rag1−/−) who because of a defect in the recombinase gene have no mature T cells. Here we reconstituted Rag1−/− with an intraperitoneal injection of 5×106 syngeneic naive CD4+25− T cells followed two days later by a subcutaneous injection of either 0.25 × 106 Pan02 or Eso2 in the left medial thigh. The tumors became palpable after approximately two weeks and reached their maximum allowable size by six weeks. Once the mice were sacrificed the spleen and TDLN were harvested and analyzed for expression of Foxp3. There was low expression of Foxp3 in the freshly isolated CD4+25− T cells used to reconstitute Rag1−/− mice and in TDLN of mice that were not reconstituted (Figure 5). By contrast, CD4+CD25− reconstituted mice challenged with Pan02 had marked upregulation of Foxp3 expression within the TDLN compared to those challenged with ES02. Additionally, Foxp3 induction within the Pan02 tumor-bearing mice was restricted to the TDLN as it was not seen in the harvested splenocytes.
To study the importance of TGF-Beta in tumor induced conversion, we utilized the same Rag1−/− adoptive transfer model and studied the effects of systemic TGF-Beta blockade on Treg prevalence. Four weeks following tumor challenge, reconstituted Rag1−/− mice all exhibited palpable tumors with an average size 40mm2 (Fig. 6A). Starting at Week 4 mice received either no treatment, anti-TGF-Beta antibody (10 μ.g., i.v.), or IgG1 isotype-matched control. The decision to treat mice with established tumors at the fourth week was based on prior studies that indicated the maximal increase in prevalence of Treg numbers occurred at four weeks following tumor challenge 33. The mice were treated on a schedule of either every four days (Q4) or weekly (Q7) for a total of four weeks. Tumor volume was recorded as the measured product of diameters taken in two separate axes. The growth curves in Figure 6A show that there was no difference in tumor size between treatment groups at the start of treatment. However over time mice treated with anti-TGF-Beta demonstrated a statistically significant delay in tumor growth when compared to untreated and isotype controls. FACS analysis of total cells purified from the TDLN of these mice showed that mice receiving a Q7 schedule of anti-TGF-Beta had more than a fifty percent reduction in the percentage of CD4+25+ T cells when compared with the no treatment controls (Fig. 6B). (FACS analysis of the no tumor group was not preformed due to insufficient number of cells isolated from normal LN). Analysis of Foxp3 expression by real-time PCR demonstrated that untreated tumor bearing mice had a 2.5 fold induction in Foxp3 expression compared with the non-tumor bearing control (Fig. 6C). Treatment with anti-TGF-Beta given on a Q7 schedule resulted in a significant reduction in Foxp3 expression whereas treatment with isotype or antibody given more frequently than Q7 did not reduce tumor mediated induction of Foxp3. We speculate that the Q4 frequency was less effective at blockade given that the tail veins by often the 4th treatment and mice were unlikely receiving an appropriate dose of the antibody.
TGF-Beta plays an important role in the induction and maintence of immune tolerance. Mice expressing a dominant negative TGF-Beta receptor II (dnTGF-BetaRII) transgene have a mutation that prevents downstream signaling via the receptor and renders T cells insensate to TGF-Beta 31. To study the effect of TGF-Beta receptor signaling on Pan02 mediated conversion of naive T cells, we isolated naive CD4+25− T cells from dnTGFBetaRII transgenic mice, adoptively transferred them into Rag1−/− mice, and subsequently challenged the mice with Pan02. Mice that were reconstituted with wildtype CD4+CD25− naive T cells and then challenged with tumor expectedly showed a marked induction of Foxp3 expression, demonstrating that these naive T cells had undergone conversion (Fig. 7). In contrast, when tumor-challenged Rag1−/− mice were instead reconstituted with dnTGFBetaRII CD4+25− naive T cell there was no observed induction of Foxp3 expression. This suggests that intact TGF-Beta receptor signaling is necessary for tumor induced conversion of naive T cells into Treg. Interestingly, there was no statistically significant difference in the extent of tumor burden in mice reconstituted with wildtype versus those transgenic CD4+25− naive T cells (Fig. 8A). By contrast, if the transgenic line of mice themselves were challenged with tumor versus the wildtype controls, the transgenic mice had statistically less tumor burden than their wildtype controls (Fig. 8B). Previously published studies with dnTGFBetaRII transgenic mouse had shown that these mice were resistant to tumor challenge with thymoma and melanoma cell lines 34.
Cancers have evolved numerous mechanisms by which they evade immune recognition. Evidence suggests that regulatory T cells (Treg) play an important role in tumor induced immunosuppression 35. Published works due in fact show that non-selective removal of host Treg with either an anti-CD25 antibody or CD25-directed immunotoxin results in improved anti-tumor effector responses. This effect however is transient and in some cases comes at a cost to the host. This is because non-selective neutralization of both nTreg and iTreg puts the host at risk of developing autoimmunity22;28-30. In some cancer models this could mean only clinical manifestations of vitiligo whereas in a model for pancreas cancer it could result in treatment induced diabetes or an autoimmune pancreatitis. Therefore our group set out to better understand the specific mechanisms by which Treg are induced in the tumor microenvironment, and hopefully allow for the development of more selective targets to block tumor mediated immunosuppression. The premise for the current study was taken from four observations in the literature: 1) pancreas cancer patients are chronically immunosuppressed and have elevated levels of serum TGF-Beta 4, 2) Treg numbers are elevated in the periphery and tumor microenvironment of patients with pancreas cancer8, 3) TGF-Beta induces in vitro conversion of naïve T cells into Treg18, and 4) preliminary evidence suggests induced Treg within the tumor microenvironment may play a more significant role than thymically derived nTreg in suppressing anti-tumor immunity25-27 .
Other studies have elucidated in both cancer and non-cancer models that increased prevalence of Treg can occur via three different mechanisms: recruitment, expansion, and conversion. Treg recruitment has been evidenced in an ovarian cancer model, where the release of a Treg specific chemokine, CCL22, resulted in active recruitment of circulating Treg to the tumor microenvironment 11. Expansion of resident Treg was exhibited in a diabetic mouse model where pancreatic islet cells engineered to transiently produce TGF-Beta via a tetracycline inducible promoter resulted an increased number of resident “intra-islet” Treg 36. Induction of Treg function was shown early on in vitro through the costimulation of naïve CD4+25− T cells in the presence of TGF-Beta18. The works presented here are an extension of our previous studies where we introduced the role of TGF-Beta in tumor induction of Treg in pancreas cancer. The unique angle of the current study is that it: 1) provides both qualitative and functional data for tumor induced Treg; 2) confirms the presence of a “high TGF-Beta” dependent environment by which pancreas cancer induces Treg; 3) introduces the role of TGF beta receptor signaling as specific mediator in tumor induction of Treg; and 4) places focus on the role of conversion in tumor induced Treg prevalence.
We show here that the pancreas tumor cell line, Pan02, produces elevated TGF-Beta levels in vitro and in vivo. By contrast, Eso2 is an esophageal cancer cell line that does not produce significant levels of TGF-Beta in vitro or in vivo. Presence of TGF-Beta production correlated with increased immunohistochemical staining for the Treg marker, Foxp3, in Pan02 tumors. Not surprisingly, the Eso2 tumors showed no evidence of Foxp3 staining within the tumor sections. To further phenotype the Foxp3+ cells found in the Pan02 tumors we co-stained the cells for CD4 and Foxp3. In separate experiments not shown here, we also confirmed that of the CD4+ T cells present within the tumor 35% of them were CD4+Foxp3+. In the adoptive transfer model naïve T cells were only converted in Treg while in the presence Pan02 and not Eso2. These mice at baseline do not express functional levels of Fop3+ Treg cells. Therefore once these mice were reconstituted with purified wildtype syngeneic CD+25− naive T cells any induction of Foxp3 expression is occurring exclusively within our adoptively transferred T cell population.
An obvious limitation of this study design is that we are comparing the ability of a tumor to induce Treg in two morphologically different cell lines. Thus, it is appropriate to state that the two cell lines possibly exhibit other differences in their tumor biology besides their ability to produce TGF-Beta. Preferably we would utilize an Eso2 cells line transfected with TGF-Beta or alternatively, a Pan02 cells harboring a “knock-out” element for TGF-Beta. Indeed we attempted to engineer such an Eso2 line and were able to maintain a stably TGF-Beta transfected clone in culture, but found by ELISA analysis that our vector expressed an “inactive” form of TGF-Beta that did not allow for release from the latent TGF-Beta binding protein. In light of this limitation, the focus of our study was to show how “tumor-derived” TGF-Beta was accountable for inducing Treg conversion. Thus, utilizing as a “negative-control” a tumor cell line that did not produce elevated levels of TGF-beta both in vitro and in vivo, we felt this was a near comparable model that demonstrated in vivo conversion of Treg required a “high TGF-Beta” environment. The tumors implanted subcutaneously likely do induce local inflammation from their invasive nature and an acute desmoplastic reaction that we know is TGF-Beta driven. This likely was the case with both Pan02 and Eso2. However, it doesn’t appear this minute elevation of TGF-Beta is enough to induce in vivo conversion of naïve T cells into Treg locally within the tumor microenvironment.
To further address this limitation, we chose to then investigate whether neutralization of TGF-Beta made by our Pan02 tumor blocked conversion of naïve T cells into Treg in our adoptive transfer model. This was done in place of giving systemic recombinant TGF-Beta to our Eso2 mice because it was felt to be an artificial rendering of the Pan02 system for the reasons stated above and suggestions in the literature claimed that TGF-Beta given systemically is often neutralized by factors in the serum. Indeed, use of an anti-TGF-Beta antibody showed that reconstituted mice no longer exhibited induction of Tregs while in the presence of Pan02. As a side observation we also noted that mice treated with TGF-Beta blockade also had significantly smaller tumors than the untreated group. Given that these experiments were conducted in immunocompromised models that have no functional T cells we do not assume that this “treatment effect” is secondary to removal of Treg suppressive function and subsequent enhanced anti-tumor T cell immunity. It is more likely that the smaller tumors in our treatment group were because of blocking the tumor growth promoting properties of TGF-Beta, such as tumor cell proliferation, angiogenesis, and evasion of apoptosis37 while also possibly promoting natural killer cell activity38;39.
The observation that antibody neutralization of TGF-Beta blocks tumor induced conversion of naïve T cells into Treg suggests that TGF-Beta receptor signaling plays an important role. To test this hypothesis we again utilized our adoptive transfer model but instead reconstituted them with naïve T cells expressing a mutated TGF-Beta receptor. These mice despite Pan02 implantation did not demonstrate conversion of naïve T cells into Treg. Logically because the adoptively transferred cells were unable to transducer the effects of the “high TGF-beta” environment produced by the tumor. As a side observation we did note that tumors in Rag1−/−mice reconstituted with mutated naïve T cells did not have smaller tumors, which is contrary to what we saw in the anti-TGF-Beta treated model. It is technical limitation of our Rag1−/−adoptive transfer model. Treatment effects would better be demonstrated in an immunocompetent model. Indeed, when dnTGFBetaRII transgenic mice were challenged with Pan02 they developed tumors that were significantly smaller than those in wildtype mice. This observation mirrors what is found in the literature, where these transgenics have been shown to be resistant to tumor challenge and have enhanced anti-tumor immunity. The key focus of our study, however, is not to show treatment effect but instead further understand the process of Treg conversion in the absence of competing mechanisms such as expansion and recruitment. A question that still remains is what percentage of Treg prevalence is the result of increased recruitment by tumor, expansion of resident Treg, or conversion of naïve T cells. Our current study aims only at better understanding the mechanism of tumor induced conversion. For future directions with targeted immunotherapy this would be important to know, since inhibiting only one of the three mechanisms potentially would result in suboptimal blockade of Treg mediated tumor immunosuppression and an infective anti-tumor response.
Since our works began, a few publications have come out suggesting a similar role of TGF-Beta in Treg conversion40;41. Petty et al in a 2008 study of Treg in an AIDS feline model showed the importance of TGF-Beta Receptor signaling as a mechanism for Treg conversion. Another paper recently published by Liu et al constructed a model similar to ours by using prostate cell line in vitro and renal cancer cell line in vivo to look at tumor mediated induction of Treg via TGF-Beta production. They too, utilized antibody directed neutralization of TGF-beta to demonstrate blockade of tumor induced Treg conversion. They show that mice receiving 1D11 had reduced tumor burden and number of CD4+CD25+ T cells within isolated lung metastases. Based on these findings they comment that TGF-Beta neutralization “abolished the conversion” of naïve T cells into Treg. These works support our hypothesis that Treg prevalence in the tumor microenvironment is reduced by blocking TGF-Beta. However, because Treg prevalence in the tumor microenvironment is likely influenced by a triad of expansion, conversion, and recruitment, their studies do not clearly elucidate whether TGF-Beta blockade abolished conversion, recruitment, induction or all three. This is because they used an immunocompetent model that has circulating natural Treg and host tissue Treg that may have expanded or been recruited from the periphery. We show in an immunodeficient Rag−/− model replete with ex vivo isolated naïve T cells that Treg are induced by the presence of TGF-Beta secreting tumor cells and that indeed neutralization of TGF-Beta results in blockade of this induction. Additionally, our study differs from this model in that we show Treg prevalence in both the tumor draining lymph nodes and the tumors themselves which helps demonstrate on what levels the induction of Treg is occurring. These works along with those previously published hopefully will lead to the discovery of more specific immunoregulatory targets that help harness anti-tumor immune effector responses in a cancer greatly in need of more effective therapeutic options.
The authors would like to thank Dr. Richard Flavell, Dr. Richard Battafarrano for making accessible some of the scientific materials used in these studies. Additionally the authors wish to thank the Washington University School of Medicine, Mouse Genetics Core and Gastroenterology Pathology Core for technical services. Funding for this study was supported in part by NCI K08 grant CA87018-01 to D.C.L and NIH T32 grant CA09621-11 to T.J.E. This work was also supported in part by the Washington University Digestive Diseases Research Core Center P30 DK52574
Financial Disclosure: All authors have declared there are no financial conflicts of interest in regards to this work
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.