Tumor Recognition and Self-Recognition Induce Distinct Transcriptional Profiles in Antigen-Specific CD4 T Cells

Abstract

Tumors express a wide variety of both mutated and nonmutated Ags. Whether these tumor Ags are broadly recognized as self or foreign by the immune system is currently unclear. Using an autochthonous prostate cancer model in which hemagglutinin (HA) is specifically expressed in the tumor (ProHA × TRAMP mice), as well as an analogous model wherein HA is expressed in normal tissues as a model self-Ag (C3HA^high), we examined the transcriptional profile of CD4 T cells undergoing Ag-specific division. Consistent with our previous data, transfer of Ag-specific CD4 T cells into C3HA^high resulted in a functionally inactivated CD4 T cell profile. Conversely, adoptive transfer of an identical CD4 T cell population into ProHA × TRAMP mice resulted in the induction of a regulatory phenotype of the T cell (Treg) both at the transcriptional and functional level. Interestingly, this Treg skewing was a property of even early-stage tumors, suggesting Treg induction as an important tolerance mechanism during tumor development.

The development of tumors in immunocompetent hosts is generally thought to be accompanied by the subversion of an effective antitumor immune response (1–4). This process, known as immune evasion, is multifactorial and includes alterations in Ag processing, Ag presentation, expression of immune-inhibitory molecules such as TGF-β and IDO, as well as the expression of cell surface molecules such as B7-H1 that inhibit immune function (2–8). As a consequence of these and other mechanisms, the T cell response to tumors is correspondingly blunted or absent, as evidenced by a lack of lytic function in CD8 T cells and a lack of effector cytokine production by specific CD4 T cells (9–17).

To better understand the nature of this functional CD4 T cell tolerance, we examined the transcriptional profile of CD4 T cells that recognize nonmutated tumor Ag, using both self-Ag and viral Ag recognition for comparison. For these studies, we used a well-described adoptive transfer system in which TCR-transgenic T cells specific for hemagglutinin (HA)⁴ are introduced into mice expressing their cognate Ag as a self-Ag or as tumor-restricted Ag (14, 18–21). In the self-Ag system, the C3 promoter drives HA expression in the lung, as well as in other normal tissues (19). In the tumor system, prostate-specific expression of HA is driven by the minimal rat probasin promoter (14, 22). These mice are termed ProHA (Probasin HA). For a tumor model, we utilized the transgenic adenocarcinoma of the mouse prostate (TRAMP) model, which develops autochthonous prostate tumors in a progressive manner (22, 23). By intercrossing ProHA with TRAMP mice, we created a murine system in which a well-defined tumor Ag is expressed in a tumor- and tissue-specific manner (14). In both ProHA × TRAMP and C3HA^high mice, CD4 T cell recognition of cognate Ag is accompanied by division as well as by functional tolerance as evidenced by a lack of effector cytokine production (20, 24). Our data, involving transcriptional profiling of CD4 T cells isolated from these distinct models, supports the notion that CD4 T cell recognition in the context of an evolving prostate tumor results in a phenotype different from that induced by self-Ag recognition, one characterized by a relative up-regulation of the transcription factor FoxP3 and the development of a regulatory T cell phenotype.

These data are consistent with previously published studies suggesting that tumors may specifically expand or induce regulatory T cells and that multiple tumor types are infiltrated with regulatory T cells (Treg) in humans with cancer (25–32). In addition, our data suggest that these cells represent induced as opposed to natural Tregs, which arise spontaneously in the thymus (33–38). Because tumors in TRAMP mice evolve slowly over time, we were also able to examine the relative effect of tumor stage on Treg induction/expansion by evaluating the results of adoptive transfer to younger vs older animals (39). Surprisingly, Treg induction/expansion was a characteristic of even very early precancerous lesions (prostatic intraepithelial neoplasia, or PIN), suggesting that Treg induction by evolving tumors might be a relatively early contributor to T cell tolerance (40).

Materials and Methods

Mice

All mice were on the B10.D2 (H-2^d) background. C3-HA-transgenic mice express influenza HA under the control of the C3 promoter and have been previously described (19, 20). The strain of 6.5 mice are CD4 TCR-transgenic animals that recognize an I-E^d-restricted HA epitope (¹¹⁰SFERFEIF-PKE¹²⁰; Ref. 18). These mice were backcrossed to a Thy1.1-congenic B10.D2 background for >12 generations. TRAMP mice on the C57BL/6J background were backcrossed to the B10.D2 background for >14 generations (22). ProHA mice express HA in a prostate-restricted manner under the control of the same minimal rat probasin promoter used to generate TRAMP mice and have been previously described (14). ProHA × TRAMP mice were generated by backcrossing ProHA to TRAMP mice for >16 generations (14). Animal care and experimental procedures were performed under pathogen-free conditions in accordance with established institutional protocols from Institutional Animal Care and Use Committees of Johns Hopkins University.

Adoptive transfer

Donor TCR-transgenic (6.5) mice were sacrificed via CO₂ asphyxiation. Spleens and lymph nodes were collected and homogenized, and RBCs were lysed. CD4 T cells were purified using Miltenyi magnetically labeled beads according to the manufacturer’s protocol. For some experiments, purified cells were labeled with for 8 min with CFSE (Invitrogen) by adding 0.5 μl of 5 mM stock per 1 ml of cells. After labeling, cells were washed twice and resuspended in HBSS for i.v. injections. Purified cells (5 × 10⁶) were injected per mouse in 0.2 ml total volume by tail vein injection. For CD4 T cell activation controls, nontransgenic B10.D2 mice were infected with recombinant Vaccinia virus expressing wild-type HA protein as previously described (14).

Flow cytometry/intracellular staining

Lymph nodes were harvested 3–10 days post-adoptive transfer, and single-cell suspensions were prepared. RBCs were lysed with ammonium chloride-potassium lysis buffer. All staining reagents were purchased from Pharmingen, with the exception of FoxP3, which was analyzed using a prepared kit according to the manufacturer’s instructions (Ebiosciences). After 10 min of incubation, samples were washed once in PBS plus 1% FBS solution and analyzed using a FACSCalibur instrument (BD Biosciences). Intracellular cytokine analysis was performed as previously described (41). Data were analyzed using the FloJo software package (Treestar).

Direct ex vivo Ag detection

A direct ex vivo Ag detection assay was performed as previously described (42, 43). Briefly, HA-specific memory cells were generated by adoptive transfer of cells into nontransgenic B10.D2 recipients that received primary vaccination of Vaccinia-hemagglutinin (VaccHA). Two weeks after initial transfer; B10.D2 recipient received a secondary boost of Listeria monocytogenes-HA (LMHA). Two weeks after LMHA boost, all lymph nodes and spleena were harvested from host mice, and responder CD4 T cells were enriched using anti-Thy1.1-biotin Ab and streptavidin microbeads (Miltenyi Biotech). class II^highCD11c^high dendritic cells (DC) were enriched from the iliac lymph nodes from ProHA × TRAMP, C3HA^high, and VaccHA-vaccinated B10.D2 animals as stimulators. Responder CD4 T cells (2 × 10⁴) were incubated with 1 × 10³ stimulators, resulting in a 20:1 ratio of responder to stimulator. Reactions were set up in triplicates. Forty-eight hours later, cultures were pulsed with 1 μCi of [³H]TdR and incubated for an additional 24 h before harvest with a Packard Micromate cell harvester. Determination of the amount of incorporated radioactive counts was performed with a Packard Matrix 96 direct beta counter (Packard Biosciences).

Donor CD4 T cell isolation

From 3 to 7 days post-adoptive transfer of HA-specific CD4 T cells, recipient animals were euthanized, and lymph nodes were harvested. CD4⁺ T cells were enriched by depleting CD8 and B cells using biotinylated anti-CD8, anti-B220, and MACS LS separation columns (Miltenyi Biotech) as previously described (21). Ag-specific T cells that had undergone specific division in vivo were sorted using a FACSVantage SE cell sorter (BD Biosciences), gating on CFSE-diluted, CD4⁺Thy1.1⁺ cells. As all recipient animals are of the Thy1.2 phenotype, this technique results in >95% pure donor cells and avoids the use of TCR-specific or CD4 core-ceptor-specific Abs that could potentially alter TCR- or CD4-dependent gene expression patterns. Control naive T cells were isolated in a similar manner, using nontransgenic B10.D2 animals as recipients, but instead of gating on CFSE^low cells, the undivided, CFSE^high cells were isolated.

Transcriptional analysis

Sorted cells were pelleted, frozen under 1 ml of Trizol, and stored at −80°C before RNA extraction using the Trizol reagent. Our microarray experiment was controlled using paired biological replicates. For these experiments, an identical population of CD4 T cells was adoptively transferred to the two groups of mice at a time. Then, cells were flow-sorted from each group of >10 mice separately, and further processing was performed in parallel; i.e., sorting, RNA extraction, template preparation, and analysis were done in parallel, using two separate Affymetrix chips. The integrity of extracted RNA from T cells was analyzed using an Agilent 2100 bioanalyzer and the RNA 6000 Pico and Nano Kits (Agilent Technologies), and concentrations were determined using a NanoDrop spectrophotometer (NanoDrop Technologies). Transcriptional analysis was performed at the Johns Hopkins Microarray Core facility. Per the standard protocol, RNA was amplified from 20 ng of starting total RNA with the Nugen Ovation RNA Amplification System V2, following the manufacturer’s protocol (http://www.nugeninc.com/pdfs/ov-v2_userguide.pdf). cDNA was synthesized using the Nugen FL-Ovation cDNA Biotin Module V2 kit, following the manufacturer’s protocol (http://www.nugeninc.com/pdfs/flbv2_userguide.pdf). After standard labeling, each sample was hybridized to an Affymetrix Mouse 430 Plus2 expression array, followed by interrogation with an Affymetrix GeneChip Scanner 3000. RMA analysis of paired replicates revealed a striking concordance, with fewer than 300 of 36,000 transcripts different at the p < 0.05 level.

Statistical analysis

To estimate the gene expression signals, data analysis was conducted on the CEL file probe signal values of the chip at the Affymetrix probe pair (perfect match probe and mismatch probe) level, using the statistical algorithm robust multiarray average expression measure (44) with Affy. This probe level data processing includes a normalization procedure using a quantile normalization method to reduce the obscuring variation between microarrays, which might be introduced during the processes of sample preparation, manufacture, fluorescence labeling, hybridization, and/or scanning (45). Using the signal intensities estimated above, an empirical bias method with the γ- γ modeling, as implemented in the bioconductor package EBarrays, was used to estimate the posterior probabilities of the differential expression of genes between the sample conditions. The criterion of the posterior probability >0.5, which means the posterior odds favoring change, was used to produce a differentially expressed gene list. Heatmaps were created using TreeView package version 1.60 (46).

Quantitative PCR

Quantitative PCR was performed as previously described (21). Briefly, RNA was immediately extracted from sorted 6.5 CD4⁺ T cells using the Trizol reagent (Invitrogen). Reverse transcription was performed with the Superscript First Strand Synthesis System (Invitrogen). cDNA levels were analyzed by real-time quantitative PCR with the TaqMan system (Applied Biosystems). Each sample was assayed in triplicates for the target gene together with 18S rRNA as the internal reference in a 25-μl final reaction volume, using the TaqMan Universal PCR Master Mix and the ABI Prism 7700 Sequence Detection system. The relative mRNA frequency was determined by normalization to the internal control 18S RNA. Relative mRNA frequencies were calculated as 2^ΔΔCt, where ΔΔC_t = (ΔC_{t calibration} − ΔC_{t sample}).

Suppression assay

In vitro suppression assays were performed as previously described (34). Briefly, 1 × 10⁴ purified T cells (responders) were mixed with 1 × 10³ 6.5 CD4⁺ T cells sorted from various recipients (suppressors), giving a 1:10 suppressor-to-responder ratio. Cells were then incubated in flat-bottom 96-well tissue culture plates precoated with 5 μg/ml anti-CD3e in 200 μl of CTL media. After 72 h, cultures were pulsed with 1 μCi of [³H]TdR and incubated for an additional 16 h before harvest with a Packard Micromate cell harvester. Determination of the amount of incorporated radioactive counts was performed with a Packard Matrix 96 direct beta counter (Packard Biosciences).

Immunohistochemistry

For immunohistochemistry, dorsal lobes of the prostate were harvested from a 6-, 10-, and 24-wk-old ProHA × TRAMP animals and frozen in Sakura Tissue-tek OCT compound (Accurate Chemical and Scientific) as described by the manufacturer. Serial unstained sections from the dorsal prostate lobes were then cut. Unstained slides were fixed in 75% acetone, 25% ethanol for 5 min. Dry slides were washed three times in PBS, followed by a 30-min incubation with Image-iT FX signal enhancer (Molecular Probes). Slides were washed in PBS, blocked with serum, washed, and incubated with polyclonal rabbit Ki67 (Abcam) for 45 min. Slides were washed and incubated with anti-rabbit IgG HRP for 30 min. Diaminobenzidine was used as the chromogen. Slides were then imaged on a microscope.

In vivo Ab blocking

For in vivo TGF-β blocking experiments, ProHA × TRAMP and control B10.d2 mice were injected with 0.2 mg of anti-TGF-β (2G7 or 1D11) Ab (R&D Systems) or the control IgG (Jackson ImmunoResearch Laboratories) at the time of adoptive transfer of naive transgenic T cells with a second dose administered 3 days later. As above, mice were harvested 5–10 days post-adoptive T cell transfer, and T cells were analyzed as described above.

Results

Specific CD4 T cell proliferation to self, viral, and tumor Ag

Previously, we demonstrated that adoptive transfer of HA-specific CD4 clonotypic cells into C3HA^high-transgenic mice, in which there is widespread expression of the HA Ag (self Ag model), generates functionally tolerized T cells that are prone to deletion (19, 20). These CD4 T cells are characterized by a lack of IFN-γ or IL-2 secretion upon restimulation ex vivo. Similarly, adoptive transfer of identical HA-specific CD4 T cells to animals in which HA is expressed as a prostate/prostate cancer-restricted Ag also results in functional tolerance (14). To assess whether the proliferative response to a self Ag or tumor Ag was broadly comparable in these two models, naive HA-specific CD4 T cells were CFSE labeled and transferred into the following recipients: 1) no Ag (B10.D2); 2) viral Ag (VaccHA); 3) self Ag (C3HA^high); and 4) prostate tissue/tumor Ag (ProHA × TRAMP; Fig. 1A). Adoptively transferred HA-specific CD4 T cells proliferated robustly in response to recognition of self or viral Ag. However, proliferation in response to the prostate/prostate cancer-restricted Ag was less robust and was primarily restricted to the prostate-draining (iliac) lymph nodes (14). Additionally, proliferation in single-transgenic ProHA mice is barely detectable under these conditions (data not shown), consistent with the notion that tumorigenesis is required for Ag recognition in this model (14).

FIGURE 1.

Ag-specific proliferation of HA-specific CD4 T cells. A, CFSE-labeled HA-specific Thy1.1⁺CD4⁺ T cells were adoptively transferred into recipients presenting HA Ag in varying contexts; nontransgenic, no Ag; VaccHA, viral Ag; C3HA^high, self Ag; and ProHA × TRAMP, tumor Ag. Flow cytometric analysis of Ag-driven CFSE dilution gated on donor (Thy1.1⁺) CD4⁺ T cells. B, Ag presentation assayed using a direct ex vivo ag detection assay. HA-specific memory CD4 T cells generated from VaccHA-vaccinated recipient mice were incubated with CD11c⁺-enriched DCs from the iliac lymph nodes of the indicated groups. *, p < 0.05 by two-sided Student’s t test; **, p < 0.01.

We next sought to determine whether DCs were capable of driving HA-specific CD4 proliferation by performing a direct ex vivo Ag detection (DEAD) assay (42, 43). Memory HA-specific CD4 T cells were generated by adoptively transferring cells into B10.D2 (nontransgenic) hosts with a primary immunization of VaccHA and boosted with LMHA. Four weeks after initial transfer, memory CD4 T cells were enriched from vaccinated hosts using Thy1.1 microbeads and incubated with CD11c^high DCs isolated from the lymph nodes of various experimental animals. As a positive control, DCs were isolated from the lymph nodes of VaccHA-infected nontransgenic mice. DCs isolated from single-transgenic ProHA and TRAMP (without HA) mice were used as negative controls given that the proliferation of HA-specific clonotypic cells is minimal following adoptive transfer. DCs from the lymph nodes of both tumor- and self Ag-bearing animals were capable of driving significant (p < 0.05) proliferation of memory HA-specific CD4⁺ T cells, confirming similar reports that HA is cross-presented in the C3HA^high self Ag model and demonstrating DC as a likely APC in the ProHA × TRAMP model (Refs. 19 and 47 and Fig. 1B).

Transcriptional analysis of CD4 T cells specific for tumor or self Ag

We next sought to determine whether these various Ag-specific proliferative stimuli would result in distinct transcriptional profiles in vivo. Naive, CFSE-labeled, Thy1.1⁺ HA-specific CD4 T cells were adoptively transferred into nontransgenic (no Ag), VaccHA-vaccinated (viral Ag), C3HA^high (self-Ag), and ProHA × TRAMP (tumor/tissue Ag) recipients. As shown in Fig. 1A, Thy1.1⁺ CFSE diluted clonotypic cells were FACS sorted from VaccHA, C3HA^high, and ProHA × TRAMP recipients, whereas CFSE undiluted cells from nontrangenic recipients were sorted as a negative control for the adoptive transfer and isolation procedures. Transcriptional profiles were determined using the Affymetrix 430A GeneChip microarray representing 39,700 transcripts and analyzed using the statistical criteria previously described (48). Compared with naive cells isolated from nontransgenic controls, microarray analysis revealed 3908 transcripts differentially expressed during viral Ag recognition, 5479 transcripts in self Ag recognition, and 3572 differentially expressed transcripts in tumor Ag recognition (Fig. 2 and supplemental Tables I–III).⁵ Only 780 unique transcripts were significantly differentially expressed between CFSE diluted tumor vs self Ag recognizing CD4 T cells. A curated list of the top 40 up-regulated (Fig. 2A) and down-regulated (Fig. 2B) genes involved in tumor vs self Ag recognition is presented in Tables I and II. These genes represent a broad array of molecules, including transcription factors such as FoxP3, Madh1, and Znfn1a2; anti-apoptosis molecules such as Bcl2 and Bcl-11b; signaling molecules; and cell surface markers. A similar comparison revealed 1930 transcripts that were differentially expressed in tumor Ag-recognizing CD4 T cells as compared with viral-associated Ag-recognizing T cells (Table III and IV). Overall, these data appear to be consistent with previously published data from our group and others (21, 37, 49, 50). Among many others, we noted the expected up-regulation of IFN-γ in CD4 T cells recognizing HA in the context of viral infection, and a relative up-regulation of the T cell surface protein LAG-3 in the self-tolerance (C3HA^high) model (21).

FIGURE 2.

Gene expression profiling of Ag-specific CD4 T cells. Gene expression profiling was performed on pooled FACS-sorted, CFSE-diluted, HA-specific Thy1.1⁺CD4⁺ T cells from the lymph nodes of no Ag (nontransgenic; NT), viral Ag (VaccHA), self Ag (C3HA^high), or tumor (ProHA × TRAMP) Ag recognition models using Affymetrix Mouse 430 Plus2. Differentially expressed genes were identified by comparing the expression profile of the indicated group to the control group in a pairwise analysis. The top 40 up-regulated (fold change >4) or down-regulated (fold change <–2) genes were selected from each comparison pair to generate the heatmap. N, Number of genes statistically different between indicated groups.

Table I.

Genes differentially up-regulated by CD4 T cells experiencing Ag derived from tumor (ProHA × TRAMP) or self (C3HA^high)

Probe Set IDa	Gene Symbol	Gene Title	ProHA × TRAMP	C3-HA	Fold Change
1417409_at	Jun	Jun oncogene	809	38	21.5
1453840_at	Pabpc1	Poly(A)b-binding protein, cytoplasmic 1	1075	83	13.0
1455805_x_at	FoxP3	Forkhead box P3	340	32	10.5
1436222_at	Gas5	Growth arrest-specific 5	357	37	9.8
1440169_x_at	Ifnar2	IFN (α and β) receptor 2	497	55	9.0
1456956_at	Zfn1a2	Zinc finger protein, subfamily 1A, 2	859	103	8.4
1439036_a_at	Atp1b1	ATPase, Na +/K +-transporting, β1 polypeptide	125	15	8.1
1448208_at	Madh1	MAD homolog 1	206	26	7.8
1427275_at	Smc4I1	SMC4 structural maintenance of chromosomes 4-like 1	1290	186	7.0
1440770_at	Bcl2	B cell leukemia/lymphoma 2	686	100	6.9
1438941_x_at	Ampd2	AMP deaminase 2	722	105	6.9
1450339_a_at	Bcl11b	B cell leukemia/lymphoma 11B	2529	369	6.9
1441926_x_at	Tmie	Transmembrane inner ear	705	107	6.6
1437163_x_at	Gtf2h4	General transcription factor II H, polypeptide 4	312	48	6.5
1417679_at	Gfi1	Growth factor-independent 1	855	135	6.3
1428463_a_at	Ppp2r5e	Protein phosphatase 2, regulatory subunit B (B56), ε isoform	128	20	6.3
1440104_at	Ranbp2	RAN-binding protein 2	674	107	6.3
1440837_at	H2-Ob	Histocompatibility 2, O region β locus	219	35	6.3
1436804_s_at	Scyl1	SCY1-like 1	250	40	6.3
1419810_x_at	Gli	GLI-Kruppel family member GLI	1392	227	6.1
1434895_s_at	Ppp1r13b	Protein phosphatase 1, regulatory (inhibitor) subunit 13B	720	118	6.1
1448325_at	Myd116	Myeloid differentiation primary response gene 116	2204	363	6.1
1450095_a_at	Acyp1	Acylphosphatase 1, erythrocyte (common) type	133	22	6.0
1456467_s_at	Nik	Nemo-like kinase	314	53	5.9
1448471_a_at	Tpbpb	Trophoblast-specific protein β	737	125	5.9
1437828_s_at	Bing4	BING4 protein	415	71	5.9
1427186_a_at	Mef2a	Myocyte enhancer factor 2A	187	32	5.9
1425514_at	Pik3r1	PI3K, regulatory subunit, polypeptide 1	156	27	5.7
1439768_x_at	Sema4f	Sema domain, Ig domain, TM domain, and short cytoplasmic domain	419	74	5.7
1438259_at	Strn3	Striatin, calmodulin-binding protein 3	328	58	5.6
1417734_at	Nakap95	Neighbor of A kinase-anchoring protein 95	891	160	5.6
1439440_x_at	Ptk 9I	Protein tyrosine kinase 9-like	1344	246	5.5
1437667_a_at	Bach2	BTB and CNC homology 2	1291	237	5.4
1434399_at	Galnt6	UDP-N-acetyl-α-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 6	1516	278	5.4
1440885_at	Evl	Ena-vasodilator-stimulated phosphoprotein	909	168	5.4
1447757_x_at	Inpp5f	Inositol polyphosphate-5-phosphatase F	180	34	5.3
1418643_at	Tm4sf13	Transmembrane 4 superfamily member 13	5332	1004	5.3
1436983_at	Crebbp	CREB-binding protein	673	127	5.3
1430357_at	H3f3b	H3 histone, family 3B	527	100	5.3
1438211_s_at	Dbp	D site albumin promoter-binding protein	382	73	5.2

^aID, identification number; poly(A), polyadenylate; RAN, Ras-related nuclear protein; GLI, gliotactin; sema, semaphorin; TM, transmembrane; BTB, bric-a-brac, tramtrack, and broad complex; CNC, canary complex; Ena, enabled.

Table II.

Genes differentially down-regulated by CD4 T cells experiencing Ag derived from tumor (ProHA × TRAMP) or self (C3HA^high)

Probe Set IDa	Gene Symbol	Gene Title	ProHA × TRAMP	C3-HA	Fold Change
1421375_a_at	S100a6	S100 calcium-binding protein A6	73.7	683.6	−9.3
1424542_at	S100a4	S100 calcium-binding protein A4	31.0	267.8	−8.6
1449911_at	Lag3	Lymphocyte activation gene 3	48.6	391.3	−8.1
1422557_s_at	Mt1	Metallothionein 1	91.8	739.2	−8.1
1436164_at	Slc30a1	Solute carrier family 30 (zinc transporter), member 1	137.8	1055.9	−7.7
1428942_at	Mt2	Metallothionein 2	209.4	1175.5	−5.6
1447849_s_at	Maf	Avian musculoaponeurotic fibrosarcoma (v-maf) AS42 oncogene homolog	19.4	108.4	−5.6
1416529_at	Emp1	Epithelial membrane protein 1	12.6	59.5	−4.7
1420119_s_at	Phf3	PHD finger protein 3	19.0	87.8	−4.6
1418622_at	Rab2	RAB2, member RAS oncogene family	24.0	105.0	−4.4
1455090_at	Angptl2	Angiopoietin-like 2	156.6	672.9	−4.3
1453228_at	Stx11	Syntaxin 11	114.6	483.7	−4.2
1433648_at	Spag9	Sperm-associated Ag 9	127.0	501.0	−3.9
1455439_a_at	Lgals1	Lectin, galactose binding, soluble 1	743.7	2915.4	−3.9
1419761_a_at	Gabpb1	GA repeat binding protein, β1	47.8	183.4	−3.8
1418401_a_at	Dusp16	Dual-specificity phosphatase 16	30.2	115.6	−3.8
1436325_at	Rora	RAR-related orphan receptor α	21.2	79.6	−3.8
1439348_at	S100a10	S100 calcium-binding protein A10	65.2	244.2	−3.7
1419060_at	Gzmb	Granzyme B	79.2	295.2	−3.7
1419202_at	Cst7	Cystatin F	152.3	554.8	−3.6
1419573_a_at	Lgals1	Lectin, galactose binding, soluble 1	520.8	1853.3	−3.6
1416029_at	Tieg1	TGF- β-inducible early growth response 1	92.8	322.3	−3.5
1455166_at	Arl8	ADP ribosylation factor-like 8	140.0	484.2	−3.5
1426208_x_at	Plagl1	Pleiomorphic adenoma gene-like 1	94.0	318.2	−3.4
1416958_at	Nr1d2	Nuclear receptor subfamily 1, group D, member 2	39.5	130.9	−3.3
1449310_at	Ptger2	PGER2	42.4	138.4	−3.3
1444500_at	Ahsa1	AHA1, activator of heat shock 90-kDa protein ATPase homolog 1	50.9	162.9	−3.2
1417935_at	Mkrn2	Makorin, ring finger protein, 2	243.9	775.2	−3.2
1448370_at	Ulk1	Unc-51-like kinase 1	260.0	814.9	−3.1
1450753_at	Nkg7	Natural killer cell group 7 sequence	111.4	346.7	−3.1
1435874_at	Prkab2	Protein kinase, AMP-activated, β2 noncatalytic subunit	104.8	323.6	−3.1
1449235_at	Tnfsf6	TNF (ligand) superfamily, member 6	79.8	230.8	−2.9
1419091_a_at	Anxa2	Annexin A2	345.1	965.4	−2.8
1419838_s_at	Stk18	Serine/threonine kinase 18	33.6	93.6	−2.8
1450714_at	Oazin	Ornithine decarboxylase antizyme inhibitor	96.7	266.8	−2.8
1428393_at	Nrn1	Neuritin 1	393.1	1073.4	−2.7
1457528_at	Slc4a7	Solute carrier family 4, sodium bicarbonate cotransporter, member 7	48.7	131.0	−2.7
1421963_a_at	Cdc25b	Cell division cycle 25 homolog B	97.8	262.8	−2.7
1427629_at	Ptprj	Protein tyrosine phosphatase, receptor type, J	17.5	46.8	−2.7
1456126_at	Malt1	Mucosa-associated lymphoid tissue lymphoma translocation gene 1	75.2	200.9	−2.7

^aID, identification number; PHD, plant homeodomain.

Table III.

Genes differentially up-regulated by CD4 T cells experiencing Ag derived from tumor (ProHA × TRAMP) vs virus (VaccHA)

Probe Set IDa	Gene Symbol	Gene Title	ProHA × TRAMP	VaccHA	Fold Change
1456956_at	Zfpn1a2	Zinc finger protein, subfamily 1A, 2	859	105	8.2
1417481_at	Ramp 1	Receptor (calcitonin) activity-modifying protein 1	470	63	7.5
1455265_a_at	Rgs16	Regulator of G-protein signaling 16	337	47	7.2
1428834_at	Dusp4	Dual-specificity phosphatase 4	1139	173	6.6
1436759_x_at	Cnn3	Calponin 3, acidic	232	44	5.2
1419156_at	Sox4	SRY-box containing gene 4	261	54	4.8
1450826_a_at	Saa3	Serum amyloid A 3	1370	286	4.8
1441760_at	Rps25	Ribosomal protein S25	866	187	4.6
1416514_a_at	Fscn1	Fascin homolog 1, actin-bundling protein (Strongylocentrotus purpuratus)	611	136	4.5
1449984_at	Cxcl2	Chemokine (C-X-C motif) ligand 2	208	47	4.4
1418449_at	Lad1	Ladinin	542	125	4.3
1438148_at	Gm1 960	Gene model 1 960	52	12	4.3
1433575_at	Sox4	SRY box-containing gene 4	64	15	4.2
1422892_s_at	H2-Ea	Histocompatibility 2, class II Ag Eα	1108	268	4.1
1434499_a_at	Ldh2	Lactate dehydrogenase 2, B chain	186	46	4.1
1426738_at	Dgkz	Diacylglycerol kinase ζ	926	232	4.0
1428329_a_at	Wdr5 6	WD repeat domain 56	370	95	3.9
1423547_at	Lyzs	Lysozyme	867	223	3.9
1438370_x_at	Dos	Downstream of Stk11	430	111	3.9
1437807_x_at	Ctnna1	Catenin (cadherin-associated protein), α1	350	92	3.8
1426112_a_at	Cd72	CD72 Ag	267	71	3.8
1452543_a_at	Scgb1a1	Secretoglobin, family 1A, member 1	101	27	3.7
1417925_at	Ccl2 2	Chemokine (C-C motif) ligand 22	1254	338	3.7
1416714_at	Icsbp 1	IFN consensus sequence-binding protein 1	617	166	3.7
1416111_at	Cd83	CD83 Ag	449	121	3.7
1419297_at	H2-Oa	Histocompatibility 2, O region alpha locus	541	146	3.7
1421038_a_at	Kcnn4	Potassium intermediate/small conductance calcium-activated channel, subfamily N, member 4	539	146	3.7
1456700_x_at	Marcks	Myristoylated alanine-rich protein kinase C substrate	220	60	3.7
1421073_a_at	Ptger4	PGER 4	54	15	3.7
1436959_x_at	Nelf	Nasal embryonic LHRH factor	1288	351	3.7
1456941_at	Tert	Telomerase reverse transcriptase	76	21	3.6
1437270_a_at	Bsf3	Cardiotrophin-like cytokine factor 1	278	78	3.6
1420404_at	Cd86	CD86 Ag	474	135	3.5
1428643_at	Mgat5	Mannoside acetylglucosaminyltransferase 5	263	75	3.5
1419549_at	Arg1	Arginase 1, liver	68	20	3.5
1451285_at	Fus	Fusion, derived from t(12;16) malignant liposarcoma (human)	183	53	3.5
1432466_a_at	Apoe	Apolipoprotein E	73	21	3.4
1438274_at	Zfpn1a4	Zinc finger protein, subfamily 1A, 4	341	99	3.4
1419083_at	Tnfsf1 1	TNF (ligand) superfamily, member 11	321	94	3.4
1435290_x_at	H2-Aa	Histocompatibility 2, class II Ag A, α	1307	384	3.4

^aID, identification number; PHD, plant homeodomain; RAB, Ras-related; RAR, retinoic acid receptor; Ras, rat sarcoma.

Table IV.

Genes differentially down-regulated by CD4 T cells experiencing Ag derived from tumor (ProHA × TRAMP) vs virus (VaccHA)

Probe Set IDa	Gene Symbol	Gene Title	ProHA × TRAMP	VaccHA	Fold Change
1458947_at	Fancc	Fanconi anemia, complementation group C	420	787	−1.9
1444676_at	Ctcf	CCCTC-binding factor	60	112	−1.9
1441669_at	Centb2	Centaurin, β2	125	233	−1.9
1447903_x_at	Ap1s2	Adaptor-related protein complex 1, σ2 subunit	263	493	−1.9
1459736_at	Stk 10	Serine/threonine kinase 10	134	250	−1.9
1438683_at	Wasf2	WAS protein family, member 2	125	233	−1.9
1445337_at	Dnajc13	DnaJ (Hsp40) homolog, subfamily C, member 13	137	255	−1.9
1423727_at	Cnih	Cornichon homolog	260	483	−1.9
1445928_at	March6	Membrane-associated ring finger (C3HC 4) 6	140	261	−1.9
1455886_at	Cbl	Casitas B-lineage lymphoma	448	833	−1.9
1440729_at	Eps15	Epidermal growth factor receptor pathway substrate 15	19	36	−1.9
1443480_at	Rassf3	Ras association (RalGDS/AF-6) domain family 3	159	295	−1.9
1426559_at	Sbno1	Sno, strawberry notch homolog 1	635	1177	−1.9
1448885_at	Rap 2b	RAP2B, member of RAS oncogene family	230	426	−1.9
1423423_at	Pdia3	Protein disulfide isomerase-associated 3	63	116	−1.9
1417270_at	Wdr12	WD repeat domain 12	39	72	−1.9
1422533_at	Cyp51	Cytochrome P450, family 51	26	49	−1.9
1415838_at	Tde2	Tumor differentially expressed 2	124	230	−1.8
1419758_at	Abcb 1a	ATP-binding cassette, subfamily B (MDR/TAP), member 1A	22	40	−1.8
1420021_s_at	Suz12	Suppressor of zeste 12 homolog	328	604	−1.8
1420497_a_at	Cebpz	CCAAT/enhancer binding protein zeta	89	164	−1.8
1419497_at	Cdkn1 b	Cyclin-dependent kinase inhibitor 1B	36	65	−1.8
1423819_s_at	Arl6ip1	ADP ribosylation factor-like 6-interacting protein 1	805	1480	−1.8
1450093_s_at	Zbtb7a	Zinc finger and BTB domain containing 7a	30	56	−1.8
1459635_at	Dlgh1	Discs, large homolog 1	55	101	−1.8
1453361_at	Hells	Helicase, lymphoid specific	18	33	−1.8
1417371_at	Peli1	Pellino 1	528	968	−1.8
1446205_at	Nfyc	Nuclear transcription factor-Yγ	89	162	−1.8
1429525_s_at	Myo1f	Myosin IF	33	61	−1.8
1448339_at	Tmem30 a	Transmembrane protein 30A	66	121	−1.8
1441460_at	Fgfr1op 2	FGFR1 oncogene partner 2	71	130	−1.8
1459457_at	Camk2d	Calcium/calmodulin-dependent protein kinase II, delta	170	310	−1.8
1431197_at	Arl6ip2	ADP ribosylation factor-like 6-interacting protein 2	82	149	−1.8
1422769_at	Syncrip	Synaptotagmin binding, cytoplasmic RNA interacting protein	15	27	−1.8
1434039_at	Appbp2	Amyloid β precursor protein (cytoplasmic tail)-binding protein 2	22	40	−1.8
1452115_a_at	Plk4	Polo-like kinase 4	17	31	−1.8
1449480_at	Sap 18	Sin 3-associated polypeptide 18	46	84	−1.8
1424443_at	Tm6sf1	Transmembrane 6 superfamily member 1	12	22	−1.8
1441319_at	Rbm5	RNA-binding motif protein 5	52	95	−1.8
1438064_at	Nsep 1	Nuclease-sensitive element binding protein 1	262	478	−1.8
1437884_at	Arl8	ADP ribosylation factor-like 8	71	129	−1.8

^aID, identification number; MDR, multidrug resistant; FGFR, fibroblast growth factor receptor; WAS, Wiskott-Aldrich syndrome; Ras, rat sarcoma; BTB, bric-a-brac, tramtrack, and broad complex.

CD4 T cells that recognized tumor-associated Ag have a Treg phenotype

In the set of transcripts up-regulated in response to tumor-recognition, we noted the transcription factor FoxP3, a relatively well-established marker of regulatory T cell (Treg) (33, 34, 51–53). To verify these data, we performed quantitative PCR analysis on CFSE-diluted HA-specific CD4⁺ T cells isolated after adoptive transfer. As shown in Fig. 3A, CD4 T cells isolated from ProHA × TRAMP tumor-bearing animals expressed significantly higher FoxP3 mRNA levels than cells harvested from C3HA^high or Vacc-HA-infected nontransgenic mice (p < 0.05). We next tested whether these transcriptional changes were reflected at the protein level by intracellular staining for FoxP3 (Fig. 3B), and found the expected up-regulation. Tregs may be functionally distinguished from other T cell subsets by their capacity to suppress the proliferative response of activated T cells (54). Thus, we next sought to determine whether these tumor-induced Tregs were capable of suppressing proliferation of responder cells using an in vitro suppression assay. For these studies, anti-CD3-activated T cells from nontransgenic mice were used as responders. Naturally occurring CD4⁺ CD25⁺ T cells were FACS sorted as control suppressors from the same mice that were the source for responders. FACS-sorted HA-specific CD4 T cells from ProHA × TRAMP recipients were able to suppress the proliferation of responders in a manner similar to that of natural Tregs (Fig. 3C). In accordance with our previously published data, Ag-experienced CD4 T cells isolated from self Ag-expressing animals were also able to suppress to some degree in this assay, despite a lack of relative overexpression of FoxP3 (21). These data are in agreement with multiple reports suggesting that FoxP3 expression is not an absolute requirement for suppressive activity in vitro (55).

FIGURE 3.

FoxP3 expression in clonotypic CD4 T cells that recognize tumor-restricted Ag. CFSE-labeled HA-specific Thy1.1⁺ CD4⁺ T cells were adoptively transferred into the indicated Ag recognition models. CSFE-diluted Thy1.1⁺ CD4⁺ T cell were FACS sorted on day 3 from recipient mice. A, FoxP3 mRNA expression as determined by real-time PCR. B, Intracellular staining for FoxP3. C, CSFE-diluted CD4 T cells from a tumor Ag recognition model were able to suppress proliferation of activated responders. An in vitro suppression assay was performed by FACS sorting CFSE-diluted Thy1.1⁺ CD4⁺ T cells from C3HA^high and ProHA × TRAMP recipients. Sorted suppressors were coincubated at a 1:10 ratio with 10⁵ T cell responders activated by anti-CD3-coated plate for 72 h. Proliferation was measured by [³H]TdR incorporation during the final 18 h of incubation. A p value of <0.05 is denoted by a single asterisk (*) for statistical significance.

Treg development is an early event in ProHA × TRAMP animals

In the same manner as TRAMP mice, ProHA × TRAMP mice reliably develop autochthonous prostate tumors in an age-dependent manner (39). In young mice, typically <6 wk, the prostate glands appear grossly normal (Fig. 4A). Between 6 and 10 wk, hyperplasia and cribriform structures appear, a lesion that has been termed murine PIN (39). Finally, older animals develop overtly neoplastic lesions, with metastatic spread to the draining lymph nodes and liver. We took advantage of the progressive nature of this model to query whether Treg development was a function of tumor stage. Naive CFSE-labeled, HA-specific CD4⁺ T cells were adoptively transferred to ProHA × TRAMP mice of various ages, harvested from the prostate-draining lymph nodes, and evaluated for FoxP3 expression by intracellular staining. Divided cells appeared to up-regulate FoxP3 even in mice in which tumor development was not obvious on gross pathological examination (Fig. 4B). These data suggest that induction of Treg might be a relatively early event in tumor development. Glucocorticoid-induced TNFR family-related gene staining did not correlate with the FoxP3-stained population in our system.

FIGURE 4.

Treg phenotype in tumor Ag-experienced CD4 T cells. A, Ki67 immunohistochemistry staining on prostate dorsal lob sections of different age group ProHA × TRAMP animals indicating pathological progression prostate cancer. B, FoxP3 intracellular staining on CFSE-labeled HA-specific Thy1.1⁺CD4⁺ T cells that were adoptively transferred into the tumor Ag recognition model (ProHA × TRAMP recipients) at various stages of disease; 6 wk with no apparent disease pathology, 10 wk with low-disease-grade pathology, and 24 wk with high-disease-grade pathology. C, Apparent de novo FoxP3 expression in HA-specific Thy1.1⁺CD4⁺CD25⁻ T cells adoptively transferred into ProHA × TRAMP mice. D, In vivo TGF-β blocking with 2G7 in ProHA × TRAMP and B10.d2 hosts did not affect the frequency FoxP3⁺ T cells in the iliac lymph node (LN).

Although CFSE labeling experiments (Fig. 4B) suggested that the Tregs observed in this model represent induced Tregs, rather than the expansion of natural CD4⁺ CD25⁺ Tregs, we attempted to address this issue more completely by depleting donor T cells of CD25⁺ T cells before adoptive transfer using FACS sorting. As shown (Fig. 4C), this process was reasonably efficient, the CD25^low CD4 fraction was <1% FoxP3⁺, as opposed to the CD25^high CD4 fraction, which was 95% FoxP3⁺. These CD4⁺CD25^low-sorted Thy1.1⁺ HA-specific T cells were adoptively transferred to ProHA × TRAMP mice and recovered from the tumor-draining iliac lymph nodes 10 days posttransfer. By this time point, >15% of these HA-specific cells were FoxP3⁺ by intracellular staining, supporting the hypothesis that tumor recognition may induce specific Tregs in vivo. Finally, we attempted to investigate a potential role for TGF-β in the development of Tregs in this system, by administering TGF-β blocking mAb (2G7) during and after adoptive transfer of HA-specific CD4 T cells into ProHA × TRAMP mice and naive B10.d2 hosts. As shown in Fig. 4D, blocking TGF-β in vivo did not significantly affect the percentage of FoxP3⁺ Thy1.1 cells compared with control IgG group. Similar results were obtained using a second TGF-β blocking mAb 1D11 (data not shown).

Discussion

In these studies, we used well-established tolerance models to study the transcriptional profiles of CD4 T cells responding to either tumor Ag or self Ag recognition in vivo. For a tumor model, we used TRAMP-transgenic mice (22). These mice develop PIN during puberty followed by a progressive invasive carcinoma and metastasis of epithelial origin in the adult male animals (56). For our studies, TRAMP mice were crossed with our ProHA mice that express influenza HA under the control of the identical probasin promoter used in TRAMP mice (14). Thus, in ProHA × TRAMP mice, expression of the nonmutated tumor Ag HA is a function of tumor progression. C3HA^high-transgenic mice were developed to study T cell tolerance to self-Ag; in these animals HA is widely expressed under the control of the rat C3 promoter (19, 20). We found that both self and tumor recognition resulted in relative up-regulation and down-regulation of large numbers of transcripts as compared with naive cells, but that a number of genes appeared to be differentially expressed between the two conditions. These data generally suggest that tumor and self-recognition result in different CD4 T cell profiles, but this interpretation must be tempered by several other differences between the two models, including a more robust division in the self Ag model, and the far more localized nature of Ag in the ProHA × TRAMP model.

Our transcriptional data also suggested a significant up-regulation of FoxP3 in tumor vs self-tolerance. This was somewhat surprising because our transcriptional profiling studies were designed in an open-ended manner, evaluating tens of thousands of transcripts in parallel without a specific hypothesis regarding the gene sets that would be found. Up-regulation of FoxP3 was confirmed at both the transcriptional level and at the protein level, and correlated with a suppressive phenotype in the in vitro suppression assay (Fig. 3C). These data are consistent with multiple reports of tumor-induced Tregs, by our group and others (27, 57, 58). Tumors may not be unique in their ability to induce or expand Tregs from a naive CD4 T cell population, as the Von Boehmer group has reported generation of regulatory T cells to both ubiquitously expressed Ag and low-level Ag encountered chronically (59–61).

Several other groups have reported a transcriptional profile for Tregs using in vivo models (37, 49, 50). Our data are somewhat unique in this respect, given that we did not initially set out with the goal of profiling Tregs. However, our data appear to be broadly consistent with prior work. For example, both Sakaguchi’s group (49)and Rudensky’s group (37) found up-regulation of ZFN1a2 in natural Tregs, and here we report a similar up-regulation. Similarly, all three studies report up-regulation of Nrp, ZFN1a4, CD86, Myb, CTL4, and BCL2 family members. However, our dataset also includes a number of unique transcripts that were previously not associated with Tregs such as CD83, Madh1, Infra2, and Scy1. Our complete transcriptional dataset is deposited in the Geo database at the National Center for Biotechnology Information (accession number GSE14662; www.ncbi.nlm.nih.gov/geo/); in-depth examination of these profiles using pathway or comparative assays could conceivably provide additional insight into the biochemical pathways involved in Treg induction and function.

Previous studies in TRAMP mice and their derivatives demonstrated T cell tolerance to prostate-associated Ag(s), with several of these reports suggesting an age dependence to this mechanism (9, 16, 62– 65). However, the relationship between Tregs and T cell tolerance in the TRAMP model is less clear, as at least one study reported that Treg depletion in these animals did not affect the vaccine response in vivo (47, 62). Our data provide additional insight into this issue, suggesting that the ability to induce and/or expand Tregs might occur when tumors are barely pathologically recognizable. These data are also consistent with recent data presented by the Blankenstein group, who used a spontaneous transgenic tumor model to show that early recognition of evolving tumors was a critical event in later tumor tolerance (40, 66). Our findings in this regard have potential clinical relevance as well, confirming other work highlighting a potential role for Treg depletion in tumor immunotherapy (67–70).

Several aspects of these data suggest that the Tregs we observe here represent induced Tregs rather than the expansion of naturally occurring CD4⁺CD25⁺ Tregs. First, we observe FoxP3 expression at the protein level in Ag-specific CD4 T cells that have divided in response to tumor recognition, but not in undivided (CF-SE^high) cells (Fig. 4B). These data are not consistent with those of Valzasina et al. (58), who showed that up-regulation of FoxP3 is independent of division of tumor Ag-experienced CD4 T cells. These differences may reflect differences in TCR affinity or Ag expression in the two systems. We attempted to address this question more fully by transferring naive CD4 T cells that were sorted to be CD25^low; only a very small fraction (<1%) of such cells are FoxP3⁺ at the time of transfer. Posttransfer, these cells showed FoxP3 expression in tumor-draining lymph nodes, consistent with possible Treg induction. We are cognizant, however, that such studies are not definitive and that it is certainly possible that these FoxP3 cells represent the result of a substantial expansion of a small population of natural Tregs in the adoptively transferred population. To investigate this more fully will require crossing FoxP3-GFP mice into the 6.5 TCR B10.D2-transgenic background animals, such that only cells that are truly FoxP3⁻ can be definitively sorted out before adoptive transfer.

In summary, we report here the comparative expression profile of Ag-specific CD4 T cells that have divided in response to tumor Ag, self Ag, and viral Ag. Although these profiles are interesting in and of themselves, our data demonstrate the induction and/or expansion of Tregs by tumor recognition, as evidenced by FoxP3 up-regulation at both the transcriptional and protein levels, as well as the acquisition of a regulatory phenotype in vitro. Ag-experienced CD4 T cells both from the self or tumor Ag models suppressed proliferating responder cells, suggesting that the Treg skewing of CD4 T cells is a tumor recognition phenomenon, whereas the ability to suppress is a shared phenomenon between the two models. The observation of a relative skewing of Tregs at a very early stage of disease further supports the notion that the expansion and/or induction of the Treg is a tumor-related event, a result with clear clinical implications. The Tregs in this system appear to represent an induced population, although far more extensive studies will be required to address that particular point more directly. Nevertheless, the transcriptional profile of these cells appears to be unique, and evaluation of particular transcripts or pathways could conceivably provide additional therapeutic targets to potentiate an antitumor immune response. Finally, this study provides evidence that tumor Ag recognition results in a distinct transcriptional profile in CD4 T cells that is unique from that induced by self Ag recognition.