Regulatory Rebound in IL-12–Treated Tumors Is Driven by Uncommitted Peripheral Regulatory T Cells

Regulatory T cells (Treg) constitute a significant proportion of the T cell infiltrates in both murine and human tumors (1), and their prevalence can be prognostic of disease progression (2). At the same time, the literature on the origin, the phenotypic characteristics, and the functional dynamics of tumor-associated Treg is limited. Early work suggested that the tumor microenvironment is conducive to both thymic Treg (tTreg) recruitment and generation of peripheral Treg (pTreg) in situ (3–6). Other reports demonstrated that the preponderance of tTreg versus pTreg is variable and dependent on the tumor type and stage (7–9). Many of the above studies, however, relied on adoptive transfer of exogenous populations to track thymic versus peripheral Treg in tumors, which complicated interpretation. With the identification of neuropilin-1 (Nrp-1) as a cellular and the Foxp3 promoter Treg-specific demethylated region (TSDR) methylation status as a molecular marker of tTreg, a recent study confirmed that tumor-associated Treg are a mixture of thymic and peripheral populations (10). The ontogeny of these populations, in contrast, is still a matter of debate (11). In addition, the functional dynamics of Treg subpopulations during tumor progression or following extrinsic perturbation such as immune therapy have not been studied. To this end, a better understanding of Treg biology in posttherapy tumors could have important clinical implications, as in most instances, suppressive homeostasis in the tumor microenvironment is rapidly restored negating the beneficial effects of treatment (12–14).

Previous studies in our laboratory demonstrated that intratumoral IL-12 restored cytotoxic function to tumor-associated effector T cells with concomitant elimination of Treg (15). Activation of pre-existing T effectors was followed by the priming of a secondary antitumor cytotoxic T cell response (CTL) in the tumor-draining lymph nodes (TDLN), which subsequently homed to the primary tumor and the metastatic sites (16). However, the cytotoxic effector window was transient and was followed by reinfiltration of tumors by Treg and the restoration of suppressive homeostasis by day 7 posttreatment (12). More recently, we demonstrated that Treg expansion was driven by the IL-12–IFN-γ–IDO axis (13). Specifically, IFN-γ was found to play a dual role in post–IL-12 immune activity in that it mediated both the activation of the cytotoxic T and NK cell responses and the induction of IDO⁺ tolerogenic dendritic cells (DC), which in turn orchestrated the Treg expansion (17). In the current study, we investigated the phenotypic and functional properties of the rebounding Treg. The results revealed that the post–IL-12 spike was driven by the pTreg subset. These cells displayed uncommitted induced Treg (iTreg)–like qualities including hypermethylated TSDR, CTLA-4, and glucocorticoid-induced TNFR (GITR) loci, unstable Foxp3 expression, reduced suppressive function, and enhanced functional plasticity. The expansion of the CD4⁺Foxp3⁺Nrp-1^low cells was dependent on IDO but not on TGF-β or programmed cell death ligand-1 (PD-L1). We speculate that the iTreg-like properties of tumor pTreg may provide an opportunity to modulate their activity to enhance therapeutic outcome.

BALB/c, BALB/c (PL-Thy1^a), Foxp3gfp (C.Cg-Foxpg3^tm2Tch/J), and C.FVB-Tg(Itgax-DTR/GFP)57Lan/J (CD11c-DTR/GFP) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). C.B.17 SCID mice were purchased from Taconic Farms (Hudson, NY). All experiments were approved by the Institutional Animal Care and Use Committee at the University at Buffalo and the University of Louisville.

Recombinant murine IL-12 was a gift from Wyeth Pharmaceuticals. Recombinant murine GM-CSF, IL-2, and IFN-γ were purchased from PeproTech (Rocky Hill, NJ). Cytokine-encapsulated biodegradable polylactic acid microspheres were prepared using phase-inversion nanoencapsulation as previously described (18). Mice were treated with a single intratumoral injection of 4 mg each preparation (equivalent to 1 μg of cytokine). 1-Methyl tryptophan (1-MT; the D/L racemic mix), D-1MT, or L-1MT (4 mg/ml in 100 μl; Sigma-Aldrich) were injected once intratumorally mixed with microspheres as previously described (13). In addition, mice received the racemic mix or the purified enantiomers in drinking water (4 mg/ml; filter sterilized and supplemented with a small amount of aspartame) for the duration of the study.

Tumors were induced by s.c. injection of 1 × 10⁶ mammary carcinoma 4T1, lung alveolar carcinoma Line-1, or the colon carcinoma colon 26 (CT-26) cells behind the neck just above the scapula. Tumor growth was monitored, and tumors were treated with a single intratumoral injection of IL-12 and GM-CSF particles upon reaching 300–400 mm³ in size. For in vivo blockade of the PD-L1 and TGF-β pathways, 200 μg anti-mouse PD-L1 Ab (10F.9G2; BioXCell) or 2.5 mg/kg body weight anti-mouse TGF-β (1D11; BioXCell) was administered i.p. every 2 d for a total of six treatments beginning 2 d before IL-12 and GM-CSF treatment. For systemic DC depletion, tumor-bearing CD11c-DTR/GFP transgenic mice were injected i.p. with 1.8 ng/g body weight diphtheria toxin (Sigma-Aldrich) in PBS on days 3 and 6 post–IL-12 plus GM-CSF treatment.

For surface phenotyping, the following mAbs were used: anti-CD4 (GK1.5), anti-CD25 (PC61.5), anti-Thy1.1 (OX-7), anti-Thy1.2 (53-2.1), anti-CD11c (HL3), anti–MHC class II (I-A/I-E) (M5/114.15.2), Nrp-1 (761705), and IL-12Rβ2 (305719). Foxp3 and IDO were quantified by intracellular staining (PJK-16s and mIDO-48) performed according to the manufacturer’s protocol. To detect the expression of IFN-γ, cells were stimulated for 4 to 5 h with PMA (Sigma-Aldrich) and ionomycin (Sigma-Aldrich) in the presence of brefeldin A (Sigma-Aldrich) and then stained with anti–IFN-γ (XMG1.2). All mAbs were purchased from eBioscience or BD Biosciences except anti–Nrp-1 and IL-12Rβ2 (R&D Systems). Stained cells were analyzed on a BD LSR Fortessa or BD Accuri C6 (BD Biosciences). Cell sorting was performed on an FACS Aria III (BD Biosciences). Flow cytometry data were analyzed using FlowJo software (Tree Star).

FACS-sorted GFP⁺Nrp-1^low or GFP⁺Nrp-1^high T cells from Foxp3gfp mice (average purity of >98.5%) were incubated in cell-culture plates coated with anti-CD3 (10 μg/ml), anti-CD28 (2 μg/ml) containing cell-culture media with IL-2 (100 U/ml), and IFN-γ (25 ng/ml) for 6 d. Cells were washed and transferred into new plates containing media with IL-12 (10 ng/ml) for an additional 24 h before intracellular staining.

FACS-sorted CD4⁺CD25⁻ T cells from syngeneic BALB/c.PL Thy1^a mice and CD4⁺GFP⁺ Treg cells from Foxp3gfp mice were used as responder T cells and Treg cells, respectively. CFSE-labeled responder cells were transferred with Treg cells at a ratio of 4:1 into SCID recipients. The number of Thy1.1⁺ responder cells and CFSE dilution were determined 4 d posttransfer by analysis of the spleen and the lymph nodes.

CD4⁺GFP⁺ Treg derived from Foxp3gfp mice were purified by flow cytometry and transferred into SCID mice (3 × 10⁵ per recipient) by i.v. injection. The lymph nodes and spleens of recipients were harvested at indicated times, and single-cell suspensions were counted, stained, and analyzed by flow cytometry using Abs against CD4, Thy1.1, Thy1.2, Nrp-1, Foxp3, and IFN-γ.

Animals were injected with BrdU in 0.1 ml sterile PBS (100 mg/kg body weight) i.p. 20 h before euthanasia. Incorporation was detected by intracellular staining using the BD Pharmingen BrdU flow cytometry kit as per the supplier’s instructions.

Genomic DNA was purified from sorted cells with PureLink Genomic DNA Mini Kit (Invitrogen). Methylation analysis was performed by bisulfite conversion of genomic DNA using the EZ DNA methylation-Gold kit (Zymo Research). The following primers were used: Foxp3 TSDR (CNS2), outer primer forward, 5′-TATTTTTTTGGGTTTTGGGATATTA-3′, outer primer reverse, 5′-AACCAACCAACTTCCTACACTATCTAT-3′, inner primer forward, 5′-TTTTGGGTTTTTTTG- GTATTTAAGA-3′, and inner primer reverse, 5′-TTAACCAAATTTTTCTACCATTAAC-3′; CTLA-4 exon 2, forward, 5′-TGGTGTTGGTTAGTAGTTATGGTGT-3′ and reverse, 5′-AAATTCCACCTTACAAAAATACAATC-3′; and GITR exon 5 forward, 5′-GAGGTGTAGTTGTTAGTTGAGGATGT-3′ and reverse, 5′-AACCCCTACTCTCACCAAAAATATAA-3′. The PCR products were purified using the DNA Clean & Concentrator kit (Zymo Research) and subsequently cloned using the TOPO TA cloning kit (Invitrogen). Recombinant plasmid DNA from the individual bacterial colonies were purified using the EZNA plasmid kit (Omega Bio-Tek) and sequenced.

Total RNA was extracted using TRIzol (Invitrogen). The cDNA was synthesized by using TaqMan Reverse Transcription reagents (Applied Biosystems). The prepared cDNAs were amplified using iTaq Universal SYBR Green Supermix kit (Bio-Rad) with the following primers: Nrp1 forward, 5′-GACAAATGTGGCGGGACCATA-3′ and reverse, 5′-TGGATTAGCCATTCACACTTCTC-3′; Dapl1 forward, 5′-GCAGTGAAAGCTGGAGGGATGCG-3′ and reverse, 5′-TGTGCCGTGTGAACTGTCGCTG-3′; Igfbp4 forward, 5′-TTCATCATCCCCATTCCAAAC-3′ and reverse, 5′-ACCCCTGTCTTCCGATCCA-3′; Swap70 forward, 5′-TGCACAGGTTATGGGAAAGG-3′ and reverse, 5′-ACGAACTGCTCAAAGCCATT-3′; and Helios forward, 5′-ACACCTCAGGACCCATTCTG-3′ and reverse, 5′-TCCATGCTGACATTCTGGAG-3′. A Stratagene Mx3005P Real-Time PCR system (Agilent Technologies) was used for all reactions and detection. Changes in the expression of a particular gene in sorted GFP⁺Nrp1^high versus GFP⁺Nrp1^low cells (normalized to β-actin) were calculated using the 2^−△△C_T method (19).

Significance in pairwise comparisons was determined using paired Student t test. A p value ≤0.05 was considered significant. For all figures: *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001.

Sustained intratumoral delivery of IL-12 and GM-CSF promotes effective reversal of tumor immune suppression (15). The restoration of immune activity, however, is short-lived due to a rapid rebound of tumor Treg (12). We undertook a detailed characterization of posttherapy Treg to obtain further insight into the biology of the counterresponse. Quantitative monitoring of TDLN and tumors revealed a 3-fold decline in Treg numbers between days 0 and 2 followed by a 4- to 5-fold increase between days 3 and 10 (Fig. 1A). To determine whether the expanding Treg were phenotypically similar to their pretherapy cohort, we performed qualitative assessment of Treg populations between days 0 and 10. Analysis of CD4⁺Foxp3⁺ Nrp-1^low pTreg and CD4⁺Foxp3⁺ Nrp-1^high tTreg subsets revealed that the pTreg diminished at a faster rate than tTreg between days 0 and 2, but dominated the expansion phase, resulting in a 5-fold increase in pTreg to tTreg ratio between days 2 and 10 (Fig. 1B). Similar findings in two other tumor models confirmed the generality of the observed phenomenon (Supplemental Fig. 1). These data demonstrated that the two Treg subsets responded differentially to immune perturbation.

Although high surface levels of Nrp-1 have been associated with tTreg, its expression is not unique to this subset (10). In contrast, methylation status of the CpG islands found in the conserved noncoding sequences that are located proximal to the basal Foxp3 promotor region have been shown to effectively distinguish tTreg from pTreg (20, 21). To this end, we analyzed the methylation status of the TSDR in sorted TDLN Treg populations from pre- and posttreatment mice. Purified tTreg and pTreg populations (Fig. 2A) displayed 0–2% versus 90–100% methylation in the TSDR, respectively (Fig. 2B). Intriguingly, the degree of methylation associated with the TSDR in our pTreg population was essentially identical to levels seen in iTreg (90–100%) (21, 22) as opposed to that associated with conventional pTreg (10–30%) (10, 23).

The CpG islands in several Foxp3-dependent genes including CTLA-4 and GITR have also been found to be hypomethylated in tTreg compared with conventional T cells or iTreg (21). To confirm that our Nrp-1–sorted cells displayed a similar profile, methylation status of the CpG elements in exon 2 of CTLA-4 and exon 5 of GITR was examined. The results again demonstrated a pattern of near complete demethylation in the Nrp-1^high population consistent with a tTreg phenotype (Fig. 2B). Interestingly, these regions were hypermethylated in the Nrp-1^low population (80–100% for CTLA-4 and 50–80% for GITR), similar to those found in iTreg (21).

Finally, expression levels of several transcripts that are known to correlate with tTreg or pTreg phenotype were assessed in sorted subsets. Quantitative PCR analysis demonstrated 4–20-fold higher levels of Nrp-1, Helios, and Swap70 in the Nrp-1^high subset and, conversely, 2–6-fold higher levels of Dpl1 and Igfbp4 expression in the Nrp-1^low subset (Fig. 2C), consistent with previous reports (10). Collectively, these data confirmed that membrane Nrp-1 represented a reliable surface marker for identification and purification of tTreg versus pTreg in tumors.

The above studies, although confirming the utility of Nrp-1 in distinguishing tumor tTreg from pTreg, revealed hypermethylation of the CpG islands in the TSDR region of both pre- and posttherapy tumor pTreg populations. In fact, the degree of methylation was closer to that found in uncommitted iTreg as opposed to conventional pTreg. Because iTreg, pTreg, and tTreg populations can display different biological properties (24, 25), we analyzed selected physiological properties of tumor and/or TDLN pTreg and tTreg populations.

We first determined whether the superior numbers of tumor pTreg were associated with enhanced proliferative capacity. TDLN Treg subsets were assessed for BrdU uptake following treatment. The data demonstrated that tTreg proliferated at a 2-fold higher rate than pTreg at steady state prior to treatment (Fig. 3A). This trend, however, reversed in posttreatment TDLN, and pTreg proliferated with progressively faster kinetics between days 2 and 10, with peak activity occurring on day 10 (Fig. 3A).

Next, the stability of Foxp3 expression in the expanding pTreg was assayed in an in vivo adoptive cell-transfer system. Nrp-1^high and Nrp-1^low Treg subsets were sorted from day 10 TDLN of tumor-bearing Foxp3-GFP mice and were adoptively transferred to SCID recipient mice. The donor cells were then analyzed for the expression of Foxp3 as well as Nrp-1 over a 4-wk period. The data shown in Fig. 3 demonstrate that the expanding pTreg were significantly less stable than tTreg with regard to Foxp3 expression (Fig. 3B). In contrast, Nrp-1 expression was similarly unstable in both populations (Fig. 3B).

A number of reports suggest that tTreg and pTreg display similar suppressive activity, whereas iTreg are functionally inferior (10, 21, 26, 27). We evaluated the suppressive capacity of sorted day 10 subsets in the in vivo setting. Adoptive transfer of pTreg or tTreg with naive responder CD4⁺ T cells into lymphodeficient hosts revealed that tumor tTreg were significantly more effective than pTreg in suppressing the homeostatic expansion and activation of the responders in the 4T1 (Fig. 3C), Line-1, and CT-26 (Supplemental Fig. 1) tumor models.

These results confirmed the correlation between hypermethylated CpG in TSDR, CTLA-4 exon 2, and GITR exon 5 and the uncommitted Treg phenotype and provided further support for the notion that tumor pTreg are phenotypically and functionally more similar to iTreg than to conventional pTreg.

The ability to direct the differentiation of naive CD4⁺ T cells toward a specific phenotype via stimulation with selected cytokine combinations is well established (28). In contrast, the plasticity of fully committed subsets has been more controversial (29). We speculated that the iTreg-like properties of posttherapy pTreg population may predispose these cells to phenotypic and functional modulation. To this end, purified day 10 pTreg and tTreg populations were cultured under Th1-promoting conditions and tested for the production of IFN-γ. Persistent exposure of cells to IFN-γ followed by IL-12 stimulation resulted in enhanced production of IFN-γ by a significant proportion of day 10 pTreg compared with tTreg in three different tumor models (Fig. 4A, Supplemental Fig. 1). This change was accompanied with an upregulation of the IL-12Rβ2 chain on the pTreg, suggestive of a direct role for IL-12 signaling in the conversion (Fig. 4B). These findings further supported the notion that posttherapy pTreg represented an iTreg-like uncommitted Treg population.

Generation of pTreg in the periphery is mediated primarily by tolerogenic DC (30). Multiple molecular factors including coinhibitory molecules, cytokines, and/or physiological metabolites contribute to this process. Specifically, different studies have shown that TGF-β (31), all-trans retinoic acid (32–34), tryptophan catabolites (35, 36), and PD-L1 (37, 38) can promote pTreg development in a tissue-specific manner. We investigated whether the above candidates contributed to the preferential expansion of pTreg in our model. Ab or small molecule inhibitor-based blockade of IDO, PD-L1, and TGF-β was performed in treated mice, and Treg expansion was monitored. Expansion of pTreg was strongly inhibited in the presence of 1-MT, an inhibitor of both paralogues of IDO, IDO-1, and IDO-2 (Fig. 5A). Specific targeting of IDO-1 and IDO-2 with L-1MT and D-1MT, respectively (39), revealed that both enzymes contributed to the pTreg expansion. D-1MT had a slight but significant advantage over L-1MT, suggesting that IDO-2 might play a more dominant role in driving the pTreg expansion. In contrast, blockade of PD-L1 or TGF-β did not result in significant suppression of pTreg proliferation (Fig. 5B, 5C), although a slight trend toward reduced pTreg expansion was observed in the presence of anti–TGF-β Ab (Fig. 5C). These data are consistent with our previous finding that the IFN-γ–IDO axis was a major contributor to the posttreatment Treg rebound in IL-12–treated tumors (13, 17).

The ability of IDO⁺ DC subsets to expand Treg is well established (36). To determine whether the pTreg expansion was ultimately associated with IDO-expressing DC in our model, we first analyzed IDO levels in posttreatment TDLN DC via intracellular staining. The data shown in Fig. 6A establish that treatment resulted in rapid accumulation of IDO-1 in both the TDLN and the tumor DC. A similar pattern for IDO-2 could not be demonstrated, as an IDO-2–specific Ab is not yet available for intracellular staining. Next, the requirement for DC in pTreg expansion was determined by depletion of DC in posttherapy mice. DC were depleted via administration of diphtheria toxin to tumor-bearing CD11c-DTR mice beginning on day 3 posttreatment, and tumors and TDLN were analyzed for pTreg and tTreg numbers on day 10. Elimination of DC greatly diminished the preferential expansion of pTreg in treated mice (Fig. 6B). Collectively, these data are consistent with the notion that IDO-expressing DC promote the expansion of pTreg in posttreatment mice.

Our data demonstrate a dominant quantitative role for pTreg over tTreg in restoring suppressive homeostasis in tumors following Th1-directed therapeutic intervention. The superior proliferative capacity of pTreg, driven by IDO⁺ DC, underlay the observed dominance. Importantly, in an effort to better characterize the tumor pTreg versus tTreg populations, we found that the TSDR, CTLA-4 exon 2, and GITR exon 5 methylation profiles of both pre- and posttherapy tumor pTreg were essentially identical to that of uncommitted iTreg rather than conventional pTreg. Consistent with this phenotype tumor pTreg displayed unstable Foxp3 expression, inferior suppressive capacity, and increased functional plasticity. These findings revealed that as a whole, pre- and posttherapy tumor Treg populations were qualitatively distinct and that posttreatment Treg, despite increased numbers, represented a population that is less stable.

In this study, the CD4⁺Foxp3⁺ pTreg and tTreg populations were sorted using a single phenotypic marker (i.e., Nrp-1. Nrp-1 expression is not unique to tTreg in that it can also be detected on activated effector T cells as well as on pTreg that are found in inflammatory microenvironments) (10). Therefore, its exclusive use for subset identification and purification in tumors can be questioned. Analysis of sorted Nrp-1^high versus Nrp-1^low populations for molecular markers of pTreg versus tTreg demonstrated that known subset-specific DNA methylation patterns and transcriptional markers cosegregated with purified tTreg and pTreg populations. Therefore, our findings establish that Nrp-1 can be effectively used for identification and analysis of tumor-associated Treg subsets.

CpG methylation profiles and functional characteristics of pre- and posttreatment tumor pTreg were essentially identical. It is therefore likely that the observed iTreg-like properties were engendered by factors associated with the tumor microenvironment and that treatment simply induced the expansion of these pre-existing cells. One such tumor microenvironment-associated factor is TGF-β, which is critical to the generation of both iTreg and pTreg (25). In fact, 4T1 cells produce significant quantities of TGF-β (data not shown). However, blockade of TGF-β during treatment resulted in only a modest reduction in the pTreg rebound, which was not statistically significant. This finding is consistent with a previous report in which TGF-β was shown to be upstream of IDO-mediated Treg expansion (40). Taken together, these observations suggest a sequelae in which TGF-β mediates the initial generation of tumor pTreg, whereas the posttherapy expansion is driven by the IL-12–IFN-γ–IDO axis.

IDO has two paralogues, IDO-1 and -2, both of which can mediate immune-regulatory activity. Specific targeting of IDO-1 versus IDO-2 with the L and D enantiomers of 1-MT resulted in similar loss of pTreg expansion, suggesting that both enzymes contributed to the regulatory rebound in our model. This finding is in line with a recent report, which demonstrated that the two paralogues act in concert (41). At the same time, it should be noted that the above interpretation is not unequivocal in view of the controversy that exists in the literature on the in vitro versus in vivo specificity and the mechanism of action of the L and D enantiomers (42).

Proliferation kinetics of tumor tTreg and pTreg changed following treatment. Specifically, tTreg were dominant at steady state, but treatment reversed this pattern. Both tTreg and pTreg expansion are dependent on DC, and it has been suggested that different DC subsets may be responsible for driving tTreg versus pTreg proliferation (30). We found that the expansion of pTreg was strictly dependent on IDO⁺ DC. IDO has been shown to induce Treg via production of the tryptophan catabolite kynurenine, a ligand for the transcription factor Aryl hydrocarbon receptor, which directly promotes Foxp3 expression (40). Whereas the role of IDO–kynurenine/Aryl hydrocarbon receptor–Foxp3 axis in Treg expansion is known, the mechanistic basis for the selective expansion of pTreg (versus tTreg) by IDO⁺ DC remains to be defined. It has been proposed that pTreg are primarily tumor-associated Ag specific, whereas tTreg recognize self-antigens (9). Therefore, one possible explanation is that the preferential expansion of pTreg is driven by the high-affinity tumor-associated Ags that are released from post–IL-12 necrotic tumors independent of DC phenotype and/or composition.

The finding that tumor pTreg display hypermethylation of loci that are associated with Treg commitment and stability raises the possibility that they may be more susceptible to functional manipulation. Because it is now well recognized that homeostatic regulatory mechanisms represent a major impediment to achieving lasting antitumor immune responses (14), the iTreg-like characteristics of tumor pTreg may provide an opportunity to intervene in favor of antitumor effector activity.

N.K.E. has ownership interest in TherapyX, Inc. The other authors have no financial conflicts of interest.