Dual Roles for Regulatory T-cell Depletion and ... · Rat anti-mouse DTA-1 GITR antibody (S....

Microenvironment and Immunology

Dual Roles for Regulatory T-cell Depletion andCostimulatory Signaling in Agonistic GITRTargeting for Tumor ImmunotherapyAshley E. Mahne, Smita Mauze, Barbara Joyce-Shaikh, Jane Xia,Edward P. Bowman, Amy M. Beebe, Daniel J. Cua, and Renu Jain

Abstract

Agonistic monoclonal antibodies (mAb) targeting the T-cellreceptor coregulatory molecule GITR exert potent therapeuticactivities in preclinical tumor models. Although anti-GITRmAb are thought to act by depleting and destabilizing theintratumoral T regulatory cell (Treg) population, the precisemechanism of action is obscure. Here, we addressed this issueusing a Treg fate-mapping approach, which revealed that Tregloss was primarily due to cell depletion, with minimal evidenceof Treg conversion to a non–Foxp3-expressing population.Further characterization of persisting Tregs following anti-GITRmAb treatment showed that a highly activated subpopulationof CD44hiICOShi intratumoral Tregs were preferentially tar-geted for elimination, with the remaining Tregs exhibiting aless suppressive phenotype. With these changes in the Treg

population, intratumoral CD8þ T cells acquired a more func-tional phenotype characterized by downregulation of theexhaustion markers PD-1 and LAG-3. This reversal of CD8þ

T-cell exhaustion was dependent on both agonistic GITR sig-naling and Treg depletion, as neither mechanism by itself couldfully rescue the exhaustion phenotype. Tests of anti-humanGITR antibody MK-4166 in a humanized mouse model ofcancer mimicked many of the effects of anti-mouse GITR mAbin syngeneic tumor models, decreasing both Treg numbers andimmune suppressor phenotype while enhancing effectorresponsiveness. Overall, our results show how anti-GITR mAbshifts Treg populations to enable immune attack on tumors,with clinical implications for molecular markers to modifyemerging treatments. Cancer Res; 77(5); 1108–18. �2016 AACR.

IntroductionGlucocorticoid-induced TNFR family-related protein (GITR,

TNFRSF18, CD357) is a costimulatory immune modulatingreceptor expressed on various immune cell subsets, including Tcells, natural killer (NK) cells, and B cells, with particularly highexpression on regulatory T cells (Tregs; refs. 1–3). Awidely studiedanti-GITR agonistic monoclonal antibody (mAb), clone DTA-1,has been shown tohave costimulatory effects on T cells in vitro andin vivo (4) and potent antitumor efficacy in multiple mousesyngeneic tumor models (5, 6).

Previous studies investigating the mechanism of actionbehind the antitumor efficacy of DTA-1 have focused heavilyon the impact on Tregs, a key population involved in theinhibition of antitumor immune responses in a range of solidtumor types (reviewed in ref. 7). These studies showed thatTregs are greatly reduced in number following DTA-1 admin-

istration. Multiple explanations have been proposed for thisreduction in Treg number, including lineage destabilization(8, 9). Previous work in autoimmune and inflammatory set-tings has shown that Tregs under highly inflammatory condi-tions can become destabilized, lose expression of their mastertranscription factor Foxp3, and convert to an effector cellphenotype (10–12). In the setting of cancer, the possibilityof converting immune-suppressive Tregs within the tumormicroenvironment into antitumor effector cells is an attractivetherapeutic aim. Alternatively to the Treg destabilizationhypothesis, Treg depletion mediated by FcgR activity has alsobeen proposed as the mechanism of Treg reduction followingDTA-1 treatment (6, 13). We sought to assess the relativecontributions of these two mechanisms using a genetic lineagetracing approach.

In addition to its impact on Tregs, GITR signaling directlyon effector cells likely contributes to the antitumor efficacy ofDTA-1. This is partially shown by the attenuated antitumorresponses of DTA-1–treated Rag�/� mice reconstituted withGITR�/� effector T cells and GITRþ/þ Tregs (8). In a tumorvaccination model, coadministration of GITRL increased CD8þ

T-cell resistance to Treg suppression in vitro (14). However, howGITR signaling might affect intratumoral CD8þ T-cell exhaus-tion status in the presence or absence of Treg reduction remainsto be investigated.

Immune-deficient mice reconstituted with human immunestem cells, commonly referred to as humanized mice, are anincreasingly accessible model system that can be used to helpbridge the divide between mouse and man. Mechanistic

Merck Research Laboratories, Palo Alto, California.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Current address for A.E. Mahne: Pfizer, South San Francisco, California; currentaddress for R. Jain: Bristol-Myers Squibb, Redwood City, California.

Corresponding Author: Renu Jain, Bristol-Myers Squibb, 700 Bay Road, Red-wood City, CA 94063. Phone: 650-260-9643; Fax: 650-260-9898; E-mail:[email protected]

doi: 10.1158/0008-5472.CAN-16-0797

�2016 American Association for Cancer Research.

CancerResearch

Cancer Res; 77(5) March 1, 20171108

on May 16, 2020. © 2017 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst October 20, 2016; DOI: 10.1158/0008-5472.CAN-16-0797

http://cancerres.aacrjournals.org/

preclinical data for a clinical candidate anti-human mAbare generally limited to in vitro or ex vivo studies. Humanizedmice are a valuable system to probe the function of immunecell-targeting clinical candidate antibodies in an in vivo setting(15). In this study, we describe such testing for MK-4166, ananti-human GITR mAb currently in clinical testing for cancerimmunotherapy.

Materials and MethodsMice and reagents

Wild-type C57BL/6, Foxp3-GDL, Foxp3-GFP-Cre, and ROSA26-loxP-Stop-loxP-Tomato mice were obtained from The JacksonLaboratory and housed and bred under specific pathogen-freeconditions in theMerck PaloAlto animal facility.Hu-CD34þNSGmice were purchased from The Jackson Laboratory after recon-stitution and verification of human immune cell engraftment. Allmouse experiments were performed according to Merck IACUC-approved protocols.

MC38 mouse colon carcinoma cell line was obtained from theDevelopmental Therapeutics Program Tumor Repository (Freder-ick National Laboratory), and SK-MEL-5 human melanoma cellline was purchased from ATCC in 2015. Both cell lines wereauthenticated using genomic profiling (IDEXXRADILCell Check)and tested to be Mycoplasma free (IMPACT I PCR Profile). MC38was expanded and maintained in our laboratory in DMEMsupplemented with 10% FBS and SK-MEL-5 in MEM supplemen-ted with 10% FBS. Both cell lines were stored in liquid nitrogen toensure that cells used for experiments were passaged for fewerthan 6 weeks.

Rat anti-mouse DTA-1 GITR antibody (S. Sakaguchi, KyotoUniversity), was murinized as previously described (16). TheN297A-mutant IgG2aDTA-1was generated in-house bymutationof mIgG2a DTA-1.

Tumor challenge and treatmentFor syngeneic tumor experiments, 8- to 10-week-old mice were

subcutaneously (s.c.) injected with 106 MC38 cells on the rightflank. For humanized mouse experiments, hu-CD34þ NSG micedisplaying greater than 25%human immune cell chimerism at 12or more weeks post-reconstitution were s.c. injected with 106

SKMEL-5 cells. Tumor diameter was measured by electroniccalipers and tumor volume was calculated by length � width �width � 1/2. Treatments were started when tumors reachedapproximately 100 mm3. For mDTA-1 experiments, mice weretreated with a single s.c. dose of antibody at 5 mg/kg unlessotherwise noted. For MK-4166 experiments, a single s.c. dose of10 mg/kg was given when tumors reached 130 mm3. Diphtheriatoxin (Sigma)was given i.p. at a dose of 15 ng/g bodyweight everyother day for a total of three doses.

Bioluminescence imagingMice were imaged 10 minutes following i.p. injection of

3 mg D-luciferin (Perkin Elmer) using an IVIS Spectrum imagingsystem (Xenogen). Analysis was performed with Living Imagesoftware (Xenogen), and radiance was quantified in units ofphotons/s/cm2/steradian.

Cell isolationSpleens and tumor draining lymph nodes (dLN) were har-

vested and mechanically disrupted to obtain a single-cell suspen-

sion. For tumor-infiltrating lymphocyte (TIL) isolation, tumorswere mechanically disrupted and digested for 45 minutes at 37degrees in the presence of collagenase 1 (300 Collagenase Diges-tion Units/mL; Sigma), DNase 1 (400 Domase Units/mL; Cal-biochem), and Dispase II (1 mg/mL; Roche). The digested tumormaterial was spun in 40% Percoll Plus (GE Healthcare) to furtherenrich leukocytes.

Flow cytometryIsolated cells were surface stained and analyzed on an LSRII

or LSR Fortessa flow cytometer (BD). Foxp3 intracellularstains were done using the eBioscience transcription factorfixation/permeabilization kit. For intracellular cytokine stains,cells were stimulated in vitro for 2.5 hours with PMA andionomycin in the presence of Brefeldin A. Following stimu-lation, cells were surface stained and then fixed in 1% PFA(Thermo Scientific Pierce) for 15 minutes at room tempera-ture. Fixed cells were permeabilized for 20 minutes (BD) andstained for IFNg .

Cytokine detectionIsolated TILs were cultured overnight in RPMI 10% FCS at 37

degrees. Supernatants were collected the following day and cyto-kine levels were measured with a Luminex kit following themanufacturer's instructions.

Statistical analysisStatistical tests were run with the aid of Graph Pad Prism

software. Two sample comparisons were made by unpairedt test. Comparisons of 3 or more samples were made by one-way ANOVA followed by Dunnett multiple comparisons test.Simultaneous comparison of treatment groups across multipletissues was done by two-way ANOVA followed by the Sidakmultiple comparisons test. For all tests, �, P < 0.05; ��, P < 0.01;��, P < 0.001; ��, P < 0.0001.

ResultsmDTA-1 depletes rather than converts intratumoral Tregs

Murinized anti-GITR antibody (mDTA-1) is highly effica-cious in the MC38 colon carcinoma model, where a single doseas low as 3 mg/kg provided complete tumor clearance in themajority of mice (Fig. 1A). GITR was expressed on peripheralTregs and was upregulated on all T-cell subsets within thetumor, with intratumoral Tregs displaying the highest levelsof GITR expression (Fig. 1B). Similar to previous observations(13), a rapid loss of intratumoral Tregs was observed upontreatment with mDTA-1 (Fig. 1C). This loss was highly specificto Tregs in the tumor microenvironment as the number andfrequency of dLN Tregs and effector CD4þ and CD8þ T cellswithin the tumor were unchanged (Fig. 1D; Supplementary Fig.S1). This decline in intratumoral Tregs resulted in an increase inthe ratio of CD8þ T cells to Tregs, an indicator of enhancedantitumor immune responses (Fig. 1E).

Previous work has suggested that the loss of intratumoral Tregsfollowing DTA-1 is due in part to Treg destabilization. Destabi-lized Tregs that lose their suppressive capacity could becomepotent antitumor effector cells due to their responsiveness totumor antigens. To address the question of whether mDTA-1treatment causes conversion of intratumoral Tregs into antitumoreffector T cells, we used a previously described Treg lineage tracing

Impact of GITR Targeting on Tregs and CD8þ T Cells in Tumor

www.aacrjournals.org Cancer Res; 77(5) March 1, 2017 1109




mouse, Foxp3-GFP-Cre � ROSA26-loxP-Stop-loxP-tdTomato(10). Foxp3-expressing Tregs in these mice express both GFP andCre, the latter of which removes a stop codon downstream of theROSA26 promoter to drive expression of the red fluorescentprotein Tomato. In this way, cells stably expressing Foxp3 aremarked by both GFP and Tomato (Fig. 2A). Cells that expressFoxp3 long enough to allow Cre-mediated activation of theTomato locus are permanently marked, even if Foxp3 expressionis subsequently lost. Such TomatoþGFP� cells are referred to as"exFoxp3 cells." These cells are present at low frequency in steady-state mice and can be expanded in settings of inflammation(10–12). Similar to a previous report in the B16F10 syngeneictumormodel (17), exFoxp3 cells were present at low frequency inMC38 tumors. These exFoxp3 cells were not enhanced in fre-quency in tumors as compared with other tissues, suggesting thatthe tumor microenvironment supports Treg stability (Fig. 2B).ExFoxp3 cells in the tumor microenvironment phenotypicallyresemble non-Tregs more so than Tregs, as shown by their lowexpression of the Treg markers CD25 and GITR (Fig. 2C).

When MC38 tumor–bearing Treg lineage tracing mice weretreated with mDTA-1, no increase in the numbers of intratu-moral exFoxp3 cells occurred, despite the reduction observed inintratumoral Tregs (Fig. 2D). Similar results were observedin the B16 melanoma tumor model (Supplementary Fig. S2).These data indicate that the loss of intratumoral Tregs followingtreatment with mDTA-1 is not due to conversion of these cellsto an exFoxp3 lineage.

mDTA-1 preferentially reduces highly activated intratumoralTregs

The lack of evidence for mDTA-1–mediated tumor Treg con-version to an alternate lineage, along with previous data linkingFcgR function to Treg reduction (13), suggested that the observeddecrease in Treg numbers is primarily due to cell depletion.However, the specificity of this depletion and the potential impacton persisting Tregs remained unknown. GITR expression waselevated on intratumoral Tregs relative to dLN Tregs and GITRhi

Tregs coexpressed elevated levels of Treg activation markers

Figure 1.

A single dose of mDTA-1 to MC38 tumor–bearing mice depletes intratumoral Tregs. MC38 tumor–bearing mice received a single dose of the indicatedantibody at 5 mg/kg (unless otherwise noted) when tumors reached approximately 100 mm3 in size. A, Tumor growth over time following a single s.c.treatment with isotype control (mIgG2a) or mouse DTA-1 IgG2a (mDTA-1) at 0.3, 3, or 30 mg/kg. Data represent the mean � SEM of 9 mice pergroup. B, Representative flow plots depicting GITR expression on CD4þ T cells (left) and CD8þ T cells (right) in tumor (top) and dLN (bottom).C, Representative flow plots gated on CD4þ T cells depicting intratumoral Foxp3þCD25þ Treg frequency at 1 and 5 days after treatment. D, Absolute counts ofFoxp3þCD25þCD4þ Tregs, Foxp3�CD4þ Tconv, and CD8þ T cells in tumor (top) and dLN (bottom) of MC38 tumor–bearing mice at 1 and 5 days aftertreatment. Intratumoral absolute numbers were normalized to tumor weight. E, The ratio of CD8þ T cells to Tregs in tumor (top) and dLN (bottom) at1 and 5 days after treatment. Each point represents an individual animal and lines represent the mean (D–E).

Mahne et al.

Cancer Res; 77(5) March 1, 2017 Cancer Research1110




including CD44, ICOS, and TIGIT (Fig. 3A). This correlation ledus to hypothesize that mDTA-1 treatment might preferentiallyaffect highly activated Tregs. To address this, we performedextensive phenotypic profiling of persisting tumor Tregs at earlytime points following mDTA-1 treatment. Indeed, highly activat-ed Tregs were the primary target of mDTA-1, as evidenced by thereduced expression of activation and functionality markers by theremaining Tregs (Fig. 3B). This change in the activation and

functional status of tumor Tregs occurred within one day oftreatment, corresponding with the rapid decline in Treg numbers.

To test whether this change in Treg phenotype was due to apreferential loss of highly activated Tregs or to a downregulationof suppressive markers induced by agonistic GITR signaling, wecompared tumor Treg phenotype between mice treated witheither mDTA-1 or with a variant containing an N297A mutationin the Fc domain that lacks FcgR binding (18, 19). N297A-mutant

Figure 2.

Intratumoral Treg loss following mDTA-1 treatment is not due to Treg conversion. MC38 tumor-bearing Treg lineage tracing mice (Foxp3-GFP-Cre x ROSA26-loxP-Stop-loxP-Tomato) were treated as in Fig. 1. A, Representative flow plot of splenic CD4þ T cells depicting the gating of GFP�Tomato� Tconvcells, GFPþTomatoþ Tregs, and GFP�Tomatoþ exFoxp3 cells. B, Representative histograms depicting CD4þ Tomatoþ T cells (red) and CD4þ Tomato� T cells(gray) in thymus, blood, spleen, nontumor dLN, tumor dLN, and tumor. Gates show the percentage of Foxp3þGFPþ cells among the Tomatoþ population.C, Representative flow plots showing CD25 (left) and GITR (right) expression on intratumoral CD4þ T cells. Tomatoþ CD4þ T cells are shown in redand Tomato� CD4þ T cells in gray. Corresponding graphs depict median fluorescence intensity (MFI) of these markers for the CD4þ T-cell subsets. D,Representative flow plots gated on intratumoral CD4þ T cells 5 days after treatment. Numbers on gates represent the percentage of each populationamong total CD4þ T cells. Graphs depict the absolute numbers of Tregs (left) and exFoxp3 cells (right) in isotype and mDTA-1–treated tumors normalizedto tumor weight. Each point represents an individual animal and lines represent the mean (C–D).






Mahne et al.





mDTA-1 showed agonist activity equivalent to mDTA-1 in acti-vation ofNF-kB signaling in splenic and tumor T cells and in T-cellproliferation assays (Supplementary Fig. S3). This suggested thatthe agonist signaling capacity of these two antibodies is equiva-lent and is not dependent on FcgR binding, unlike anti-CD40antibodies that require FcgR binding for their agonist activity(20).When administered in vivo, N297A-mutantmDTA-1 showedsimilar binding characteristics on intratumoral T cells and hadequivalent pharmacokinetics properties as non-mutant mDTA-1(Supplementary Fig. S4). Using bioluminescence as a readout ofTregs in Foxp3-GDL mice, in which Foxp3 drives expression ofGFP, human diphtheria toxin receptor, and luciferase (21), weobserved no reduction in intratumoral Treg numbers in N297A-treated mice (Fig. 3C). Flow cytometry, however, did reveal amodest reduction in Treg percentage (Fig. 3D). Importantly,tumor Tregs of N297A-mutant–treated mice also showed littleto no reduction in their expression of suppressive markers (Fig.3E). This suggested that the reduced suppressive phenotype ofremaining Tregs inmDTA-1–treatedmicewasdue to apreferentialdepletion of themost highly activated Tregs rather than signaling-induced downregulation of suppressive markers on a subset ofcells. Furthermore, pan-Treg depletion by diphtheria toxinadministration in Foxp3-GDL mice reduced intratumoral Tregsto a similar extent as that seen with mDTA-1 (Fig. 3C and D).Unlike mDTA-1, however, persisting Tregs in DT-depleted micedidnot showas large a reduction in their expressionof suppressivemarkers (Fig. 3E). This further supported the hypothesis that bytargeting GITRhi Tregs, mDTA-1 preferentially reduces a highlysuppressive subpopulation. Taken together, these data suggestthat in addition to the impact ofmDTA-1 on tumor Treg numbers,diminished functionality of persisting Tregs could also contributeto the efficacy of mDTA-1.

BothGITR agonismandTreg depletion are required for reversalof CD8þ T-cell exhaustion and antitumor efficacy of mDTA-1

In the MC38 tumor model, treatment with the N297A variantof mDTA-1 showed intermediate efficacy (Fig. 4A). This was incontrast to previously published results in the CT26 tumormodelwhere N297A showed no efficacy (13), and suggested that theMC38 model might be ideal for further dissecting the relativecontributions of GITR agonism versus Treg depletion in theefficacy of mDTA-1. As above, we used Foxp3-GDL mice to testthe relative impact of Treg depletion in the absence of GITRsignaling. DT-mediated Treg depletion significantly inhibitedtumor growth; however, unlike mDTA-1–treated animals, veryfew DT-treated mice showed a complete regression (CR; 9/10 CRfor mDTA-1 vs. 1/10 CR for DT).

In addition to its impact on intratumoral Tregs, one effectof mDTA-1 treatment, which we and others have observed, is

increased immune cell activation in the tumor dLN (22). This isnoted most simply in the increased size of the dLN and wasrecapitulated by treatment with N297A mutant mDTA-1 and byDT-mediated systemic Treg depletion (Fig. 4B). Enhanced dLNpriming was further observed specifically in CD8þ T cells, whichexhibited increased IFNg production upon stimulation followingall treatments (Fig. 4C). Together, these data suggest that bothGITR agonism and systemic Treg depletion alone are sufficient toenhance priming of CD8þ T cells in the tumor dLN.

In addition to its effect in the dLN, mDTA-1 treatment alsoaffected the activation status of CD8þ T cells within the tumor.Compared with their dLN counterparts, CD8þ T cells in estab-lished MC38 tumors display an exhausted phenotype, as evidentby their expressionof high levels of inhibitory receptors, includingPD-1 and LAG-3 (Fig. 4D; ref. 23). Upon treatment withmDTA-1,expression of these receptors decreased significantly. This changein CD8þ T-cell phenotype was apparent at day 5 followingtreatment but not at day 1 (data not shown), making it delayedrelative to the rapid changes observed in tumor Tregs. Notably,neither DT-mediated Treg ablation nor treatment with N297Amutant mDTA-1 reproduced the reversal in intratumoral CD8þ

T-cell exhaustion observed with mDTA-1 (Fig. 4D and E).In further support of an antigen-experienced exhausted phe-

notype, intratumoral CD8þ T cells from control mice were com-petent producers of IFNg and granzyme B when given a strong exvivo stimulus (Fig. 4F). As a result, no further enhancement ofthese effectormoleculeswas evident in treatedmice under these exvivo stimulation conditions (Fig. 4F). Taken together, these dataindicate that the combination of both GITR agonism and areduction in Treg numbers is required for reversal of intratumoralCD8þ T-cell exhaustion and concomitantly for full antitumorefficacy.

Anti-human GITR clinical candidate mimics the effects ofmDTA-1 in a humanized mouse tumor model

Lastly, we wished to examine how our observations on themechanism of action of the anti-mouse GITR antibody mDTA-1might translate to human cancer patients. One way to circumventthe lack of compatibility of anti-human antibodies in mice isthrough the use of immunodeficient mice reconstituted withhuman immune cells, collectively known as humanized mice(reviewed in ref. 24). One suchmodel involves the engraftment ofhuman CD34þ hematopoietic stem cells into immunodeficientNOD.Cg-PrkdcscidIl2rgtm1Wjl (NSG) mice. Several weeks followingengraftment, these mice possess human T cells, B cells, and somemyeloid cells.

MK-4166 is a humanized anti-GITR agonist mAb that is cur-rently being evaluated in a phase I clinical study in patients withcancer. In order to assess the effects ofMK-4166 in an in vivo tumor

Figure 3.mDTA-1 preferentially reduces highly activated intratumoral Tregs. A, Representative flow plots depicting the correlation between expression of GITRand other activation markers on intratumoral Tregs (blue) and dLN Tregs (gray). B, Intratumoral Tregs were profiled for expression of markers related toTreg function and activation status. Histograms depict representative examples of tumor Tregs at 1 day after isotype (blue) and mDTA-1 (red) treatment.Shaded gray histogram depicts dLN isotype-treated Tregs. Graphs depict intratumoral Treg median fluorescence intensity (MFI) for the indicated markers. C,MC38 tumor–bearing Foxp3-GDL mice were treated with mIgG2a, mDTA-1, mDTA-1 N297A-mutant IgG2a, or diphtheria toxin (DT) when tumors reachedapproximately 100 mm3 in size. Representative bioluminescent images of individual mice at 4 days after treatment, with red circles showing regionused for determining photon flux in tumors (top row) and ventral view (bottom row). Graphs depict quantification of multiple mice. D, Graphs depict thefrequency of Foxp3þCD25þCD4þ Tregs among CD4þ T cells at 5 days after treatment in the tumor (top) and dLN (bottom). E, Intratumoral Tregs fromMC38-bearing Foxp3-GDL mice were assayed by flow cytometry as in B at 5 days after treatment. Each point represents an individual mouse andlines represent the mean (B–E).






model, we tested it in hu-CD34þ NSG mice bearing SK-MEL-5human melanoma tumors. Human CD25þFoxp3þ Tregs werereadily detectable in both spleens (NSG mice lack most tissue-draining LNs) and tumors of humanizedmice, and were enrichedin the tumor (Fig. 5A), similar to the syngeneicMC38model (Fig.1C). Furthermore, these human Tregs expressed high levels ofGITR in both sites (Fig. 5B). In contrast to the tumor specificity ofthe Treg impact seen in the mDTA-1–treated syngeneic model,humanized mice treated with MK-4166 showed significantlyreduced frequency of Tregs in the spleen and a trend toward

reduced Treg frequency in the tumor (Fig. 5A). This reduced Tregfrequency resulted in an increased CD8þ T cell to Treg ratio,especially in the spleen (Fig. 5C). Although not highly impactednumerically, intratumoral Tregs were affected at the functionallevel, as exhibited by reduced expression of the activation markerICOS upon treatment with MK-4166 (Fig. 5D). These effectscorresponded with increased production of antitumor effectorcytokines within the tumor (Fig. 5E), which originated from bothCD4þ and CD8þ tumor T cells (data not shown). Lastly, whenfollowedover time,MK-4166–treatedmice exhibited significantly

Figure 4.

Both GITR agonism and Treg depletion are required for reversal of CD8þ T-cell exhaustion and the antitumor efficacy of mDTA-1. MC38 tumor–bearingFoxp3-GDL mice were treated with the indicated agent when tumors reached approximately 100 mm3 in size. A, Left graph shows mean tumor volume � SEMover time of 7 to 10 mice per group. Right graph shows individual tumor volumes at 22 days after treatment. B–F, Mice were sacrificed 5 days aftertreatment for analysis. Graphs depict dLN mass (B) and the percentage of dLN CD8þ T cells that produced IFNg upon ex vivo restimulation (C). D,Representative flow plots show PD-1 and LAG-3 expression on CD8þ T cells in the tumor (top) and dLN (bottom). Graph summarizes the frequency ofPD-1þ LAG3þ CD8þ tumor T cells for multiple mice. E, Graphs depict the median fluorescence intensity (MFI) of LAG-3 on LAG-3þ CD8þ T cells (left) and of PD-1 on PD-1þ CD8þ T cells (right) within the tumor as shown in D. F, Graphs depict the percentage of tumor CD8þ T cells that produced IFNg (left) or granzyme B(right) upon ex vivo restimulation. Each point represents an individual mouse and lines represent the mean (A–F).

Mahne et al.





Figure 5.

Clinical candidate MK-4166 mimics the effects of mDTA-1 when tested in a humanized mouse tumor model. Hu-CD34þ NSG mice bearing SK-MEL-5human melanoma tumors were treated with either MK-4166 or an IgG1 isotype control at 10 mg/kg when tumors reached a volume of approximately130 mm3. A–E, 4 days later, spleens and TILs were harvested for analysis of the human immune compartment. A, Representative flow plots showgating of human CD4þ and CD8þ T cells among CD45þCD3þ cells, and of Tregs among CD4þ T cells in the spleen (top) and TILs (bottom).Graphs depict the summarized frequencies of Tregs among total CD3þ T cells. B, Representative histograms showing GITR expression on humanTregs (blue line), non-Treg CD4þ T cells (dashed line), and CD8þ T cells (filled gray line) in the spleen (top) and TILs (bottom) of SKMEL-5 tumor–bearinghumanized mice. C, Graph depicts CD8þ T cell to Treg ratio in spleen and TILs as determined by flow cytometry. D, Representative histogramshowing ICOS expression on intratumoral Tregs from isotype- (blue) and MK-4166-treated (red) mice. Graph summarizes median fluorescence intensity(MFI) on intratumoral Tregs for multiple mice. E, Following overnight culture of isolated TILs, levels of IL2 (left) and IFNg (right) were measuredin the culture supernatants. F, Tumor growth was monitored over time following treatment. Data represent the mean � SEM of 7 mice per group.A and C–E, Each graph point represents an individual mouse and lines represent the mean.






attenuated tumor growth (Fig. 5F). In total, anti-human GITRmAb MK-4166 exhibited more systemic effects in SK-MEL-5tumor–bearing hu-CD34þ NSG mice than the mouse surrogatemDTA-1 did in WT MC38 tumor–bearing mice with regard toTregs; however, the overall effects were similar with increasedimmune activation, attenuated Treg frequency and suppressivephenotype, and overall tumor growth inhibition.

DiscussionIn this study, we have demonstrated that mDTA-1 treatment in

theMC38 tumormodel leads to a tumor-specific reduction in notonly Treg numbers but also in the functional status of those Tregsthat persist. The observed reduction in intratumoral Tregs did notcorrespond with any observable increase in Treg conversion to anexFoxp3 lineage. Upon examination of intratumoral CD8þ T cellsfollowing mDTA-1 treatment, we observed a reversal of anexhaustion phenotype that could not be reproduced with eitherTreg depletion or agonistic GITR signaling alone. Finally, exam-ination of an anti-human GITR clinical candidate antibody MK-4166 in a humanized mouse tumor model showed similarity tothe mouse surrogate in terms of attenuating Treg frequency andsuppressive phenotype.

While reduction in intratumoral Treg numbers appears to be amajor contributing factor to the efficacy of mDTA-1, multiplelines of evidence suggest that this reduction alone is insufficient toaccount for efficacy. In support of a function of GITR signalingdirectly on CD8þ T cells, mice in which effector T cells alone lackGITR expression exhibited incomplete responses to DTA-1 (8).Additionally, previous studies comparing DTA-1 treatment toTreg depletion alone found that Treg depletion was insufficientto reproduce the efficacy of DTA-1 (6, 25, 26). A major caveat ofprevious studies, however, is that the methods used to depleteTregs were imprecise and also depleted either effector T cells byusing anti-CD25 antibody PC61 (6, 25), or total CD4þ T cells viaanti-CD4 antibody (26). In our study, we compared mDTA-1treatment with Treg depletion in Foxp3-GDL mice using DTadministration for specific and systemic deletion of Tregs (21).We have shown that while DT-mediated Treg depletion promotesa strong antitumor response, these responses are not as potent asthose observed withmDTA-1 treatment. This was the case in spiteof comparable levels of intratumoral Treg depletion.One possibleexplanation for the lesser efficacy of DT-mediated Treg depletionversus mDTA-1 is the selectivity of mDTA-1 in targeting the mosthighly activated Tregs. While nearly all intratumoral Tregs displaya highly activated phenotype, the correlation of GITR expressionwith other functional and activation markers may play a role inthe selective loss of the most highly activated Tregs. This was notthe case with DT-mediated depletion, and the differences in thefunctional status of remaining Tregs following either treatmentcould contribute to the differential efficacy seen between the two.An additional explanation for the differential efficacy could be therole of costimulatory GITR signaling directly on effector T cells.The lack of complete tumor regression in DT-mediated Treg-depleted animals corresponded with a persistent exhaustionphenotype in intratumoral CD8þ T cells that was reversed bytreatment with mDTA-1. We believe our approach more defini-tively demonstrates that Treg ablation alone does not account forthe full effects of mDTA-1 and adds further support for thehypothesis that GITR agonism contributes to the efficacy ofmDTA-1.

Although robust Treg depletion alone is insufficient to reca-pitulate the full antitumor effects of mDTA-1, it is necessary.N297A-mutant mDTA-1 is incapable of mediating FcgR-depen-dent Treg depletion that has been shown to be the primarymechanism of Treg depletion for both DTA-1 and CTLA-4 IgG2aantibodies (13, 27, 28). Without this robust depletion, N297A-mutant mDTA-1 efficacy was greatly attenuated, and no reversalof the intratumoral CD8þ T-cell exhaustion phenotype wasobserved. This suggests that the reversal in CD8þ TIL exhaustionseen with mDTA-1 is due to a combination of intratumoral Tregremoval as well as antibody-mediated GITR agonism. Interest-ingly, N297A-mutant mDTA-1 treatment did lead to a partialbut significant reduction in intratumoral Tregs. This suggeststhat an additional mechanism other than FcgR-mediated cellremoval could contribute to DTA-1–mediated Treg depletion,which has not been observed with Fc-mutant CTLA-4 antibody(27). Similarly to nonmutated mDTA-1, N297A mutantmDTA-1 treatment did not enhance Treg conversion in ourlineage tracing system (data not shown). However, agonist GITRsignaling mediated by N297A mutant mDTA-1 could lead toTreg loss via a signaling mechanism that warrants future study.We believe that the partial reduction in tumor Tregs, along withthe enhanced CD8þ T cell priming in the dLN, is the primarymechanism behind the tumor growth inhibition exhibited byN297A mutant mDTA-1; however, other mechanisms not exam-ined in the current study may also contribute to the observedefficacy.

In this study, we have followed up on previous reports thatDTA-1may destabilize Treg lineage commitment. We felt this wasan important undertaking as previous reports of Treg lineageinstability have primarily arisen in the context of highly inflam-matory microenvironments that are quite different from theimmune-suppressive tumor microenvironment that is generallybelieved to promote Treg induction and function. Previous stud-ies reporting Treg instability following DTA-1 treatment reliedprimarily on the adoptive transfer of Foxp3þ Tregs and found thatthese cells weremore prone to loss of Foxp3 expression inDTA-1–treated recipients (9). This approach comes with the potentialcaveat that even a highly pure population of sorted Tregs maycontain rare contaminating Foxp3� cells that can outgrowcotransferred Tregs and skew results. Indeed, the outgrowth ofsuch a population would be predicted to be enhanced by thecostimulatory effects of DTA-1. We have circumvented this caveatin the present study by using Foxp3-lineage tracing mice. By thisapproach, we have confirmed the findings of others that Tregswithin the tumor microenvironment are relatively stable com-pared with those found in other peripheral tissues (17). However,we did not find any evidence for enhanced conversion of Tregs toexFoxp3 cells upon treatment with mDTA-1. Although this doesnot confirm previous reports of DTA-1–mediated intratumoralTreg instability (9), perhaps for the reasons outlined above, ourdata on the reduced functional phenotype of those Tregs thatpersist following mDTA-1 treatment strongly agree with previ-ously published data on the emergence of a Treg population withreduced suppressive properties following DTA-1 treatment (9).Despite a lack of full conversion of tumor Tregs to an alternateT-cell lineage, this weakening of tumor Treg suppressive capabil-ities, coupled with an overall reduction in their numbers, is likelycritical to the antitumor efficacy of DTA-1.

As with nearly all novel disease therapeutics, it is difficult topredict with confidence how efficacy in mouse models will

Mahne et al.





translate to human patients. Previous work by Sanmamed andcolleagues has shown efficacy of anti–PD-1 and anti-CD137clinical candidate antibodies in a humanizedmousemodelwhereimmunodeficient mice were reconstituted with either allogeneicor patient-derived PBMCs at the time of tumor transfer (15). Inour experiments testing an anti-human GITR mAb in tumor-bearing humanized mice reconstituted with CD34þ hematopoi-etic stem cells, we observed both similarities and differences interms of impact on Tregs as compared with mDTA-1 treatment ofconventional tumor-bearing mice. The most obvious differencebetween the two was the difference in tissue specificity. While weand others have shown that mDTA-1 specifically affects tumorTregs and leaves systemic Tregs relatively unaltered in syngeneicmousemodels,MK-4166 greatly reduced peripheral Treg frequen-cy in humanizedmice. The exact reasons for this difference remainunknown. One hypothesis is the potential differential distribu-tion of myeloid populations capable of carrying out FcgR-medi-ated cell depletion in the two models (13). Humanized micecontain both mouse and human derived myeloid populations.Detailed characterization of these populations in both the periph-ery and tumor remains to be done in order to more fully under-stand their likely interactions with a humanized IgG1 antibody.Although not highly impacted numerically, intratumoral Tregswere significantly affected at the functional level, as exhibited byreduced expression of the activation marker ICOS upon treat-ment. These effects directly correlated to increased production ofantitumor effector cytokines within the tumor. These resultssuggest that MK-4166 depletes Tregs in settings where FcgR-effector functions are intact while the effects of GITR agonismon Tregs are more apparent in settings where they are deficient.While subcutaneous tumor engraftment in humanized miceprovides a means to examine the impact of a humanized mAbon human immune cells in an in vivo setting, it remains farremoved from a trial in human patients. Nonetheless, theobserved similarities between mDTA-1 in conventional mice andanti-human GITR mAb in humanized mice are notable. Both

resulted in increased ratios of CD8þ T cells to Tregs; be it in thetumor or in the periphery, and both resulted in a reducedfunctional phenotype of remaining intratumoral Tregs andenhanced immune activation that corresponded with inhibitionof tumor growth.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: A.E. Mahne, A.M. Beebe, R. JainDevelopment of methodology: A.E. Mahne, S. Mauze, B. Joyce-Shaikh, J. Xia,R. JainAcquisition of data (provided animals, acquired and managed pati-ents, provided facilities, etc.): A.E. Mahne, S. Mauze, B. Joyce-Shaikh,J. Xia, E.P. Bowman, D.J. Cua, R. JainAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): A.E. Mahne, S. Mauze, B. Joyce-Shaikh, J. Xia, R. JainWriting, review, and/or revision of the manuscript: A.E. Mahne, S. Mauze,A.M. Beebe, R. JainAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): S. Mauze, R. JainStudy supervision: S. Mauze, D.J. Cua, R. Jain

AcknowledgmentsWe are grateful to the MRL Postdoctoral Research Fellows Program for

financial fellowship support (A.E. Mahne).

Grant SupportThis work was supported by Merck Research Laboratories (MRL), Merck &

Co. Inc. The MRL Postdoctoral Research Program provided financial fellowshipsupport to A.E. Mahne.

The costs of publication of this article were defrayed in part by the paymentof page charges. This article must therefore be hereby marked advertisementin accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received March 21, 2016; revised September 14, 2016; accepted September27, 2016; published OnlineFirst October 20, 2016.

References1. Nocentini G, Giunchi L, Ronchetti S, Krausz LT, Bartoli A,Moraca R, et al. A

new member of the tumor necrosis factor/nerve growth factor receptorfamily inhibits T cell receptor-induced apoptosis. Proc Natl Acad Sci U S A1997;94:6216–21.

2. Kwon B, Yu KY, Ni J, Yu GL, Jang IK, Kim YJ, et al. Identification of a novelactivation-inducible protein of the tumor necrosis factor receptor super-family and its ligand. J Biol Chem 1999;274:6056–61.

3. Gurney AL, Marsters SA, Huang RM, Pitti RM, Mark DT, Baldwin DT,et al. Identification of a new member of the tumor necrosis factor familyand its receptor, a human ortholog of mouse GITR. Curr Biol 1999;9:215–8.

4. Shimizu J, Yamazaki S, Takahashi T, Ishida Y, Sakaguchi S. Stimulation ofCD25(þ)CD4(þ) regulatory T cells through GITR breaks immunologicalself-tolerance. Nat Immunol 2002;3:135–42.

5. Ko K, Yamazaki S, Nakamura K, Nishioka T, Hirota K, Yamaguchi T, et al.Treatment of advanced tumorswith agonistic anti-GITRmAb and its effectson tumor-infiltrating Foxp3þCD25þCD4þ regulatory T cells. J Exp Med2005;202:885–91.

6. Coe D, Begom S, Addey C, White M, Dyson J, Chai JG. Depletion ofregulatory T cells by anti-GITR mAb as a novel mechanism for cancerimmunotherapy. Cancer Immunol Immunother 2010;59:1367–77.

7. Nishikawa H, Sakaguchi S. Regulatory T cells in tumor immunity. Int JCancer 2010;127:759–67.

8. Cohen AD, Schaer DA, Liu C, Li Y, Hirschhorn-Cymmerman D, Kim SC,et al. Agonist anti-GITR monoclonal antibody induces melanoma tumor

immunity in mice by altering regulatory T cell stability and intra-tumoraccumulation. PLoS One 2010;5:e10436.

9. Schaer DA, Budhu S, Liu C, Bryson C, Malandro N, Cohen A, et al. GITRpathway activation abrogates tumor immune suppression through loss ofregulatory T cell lineage stability. Cancer Immunol Res 2013;1:320–31.

10. Zhou X, Bailey-Bucktrout SL, Jeker LT, Penaranda C, Martinez-Llordella M,Ashby M, et al. Instability of the transcription factor Foxp3 leads to thegeneration of pathogenic memory T cells in vivo. Nat Immunol 2009;10:1000–7.

11. Bailey-Bucktrout SL, Martinez-Llordella M, Zhou X, Anthony B, RosenthalW, Luche H, et al. Self-antigen-driven activation induces instability ofregulatory T cells during an inflammatory autoimmune response. Immu-nity 2013;39:949–62.

12. Komatsu N, Okamoto K, Sawa S, Nakashima T, Oh-hora M, Kodama T,et al. Pathogenic conversion of Foxp3þ T cells into TH17 cells in auto-immune arthritis. Nat Med 2014;20:62–8.

13. Bulliard Y, Jolicoeur R, Windman M, Rue SM, Ettenberg S, Knee DA, et al.Activating Fc gamma receptors contribute to the antitumor activities ofimmunoregulatory receptor-targeting antibodies. J Exp Med 2013;210:1685–93.

14. Nishikawa H, Kato T, Hirayama M, Orito Y, Sato E, Harada N, et al.Regulatory T cell-resistant CD8þT cells inducedby glucocorticoid-inducedtumor necrosis factor receptor signaling. Cancer Res 2008;68:5948–54.

15. Sanmamed MF, Rodriguez I, Schalper KA, Onate C, Azpilikueta A, Rodri-guez-Ruiz ME, et al. Nivolumab and Urelumab Enhance Antitumor






Activity of Human T Lymphocytes Engrafted in Rag2-/-IL2RgammanullImmunodeficient Mice. Cancer Res 2015;75:3466–78.

16. Murphy JT, Burey AP, Beebe AM, Gu D, Presta LG, Merghoub T, et al.Anaphylaxis caused by repetitive doses of a GITR agonist monoclonalantibody in mice. Blood 2014;123:2172–80.

17. Wang C, Lee JH, Kim CH.Optimal population of FoxP3þ T cells in tumorsrequires an antigen priming-dependent trafficking receptor switch. PLoSOne 2012;7:e30793.

18. Shields RL, Namenuk AK, Hong K, Meng YG, Rae J, Briggs J, et al. Highresolution mapping of the binding site on human IgG1 for Fc gammaRI, Fc gamma RII, Fc gamma RIII, and FcRn and design of IgG1 variantswith improved binding to the Fc gamma R. J Biol Chem 2001;276:6591–604.

19. ChaoDT,MaX, LiO, ParkH, LawD. Functional characterizationofN297A,a murine surrogate for low-Fc binding anti-human CD3 antibodies.Immunol Invest 2009;38:76–92.

20. White AL, Dou L, ChanHT, Field VL,Mockridge CI,Moss K, et al. Fcgammareceptor dependency of agonistic CD40 antibody in lymphoma therapycan be overcome through antibody multimerization. J Immunol2014;193:1828–35.

21. Suffner J, Hochweller K, Kuhnle MC, Li X, Kroczek RA, Garbi N, et al.Dendritic cells support homeostatic expansionof Foxp3þ regulatory T cellsin Foxp3.LuciDTR mice. J Immunol 2010;184:1810–20.

22. Zhou P, L'Italien L, Hodges D, Schebye XM. Pivotal roles of CD4þeffector T cells in mediating agonistic anti-GITR mAb-induced-immune

activation and tumor immunity in CT26 tumors. J Immunol 2007;179:7365–75.

23. Woo SR, Turnis ME, Goldberg MV, Bankoti J, Selby M, Nirschl CJ, et al.Immune inhibitory molecules LAG-3 and PD-1 synergistically regulate T-cell function to promote tumoral immune escape. Cancer Res 2012;72:917–27.

24. Shultz LD, Brehm MA, Garcia-Martinez JV, Greiner DL. Humanized micefor immune system investigation: progress, promise and challenges. NatRev Immunol 2012;12:786–98.

25. Ramirez-Montagut T, Chow A, Hirschhorn-Cymerman D, Terwey TH,Kochman AA, Lu S, et al. Glucocorticoid-induced TNF receptor familyrelated gene activation overcomes tolerance/ignorance to melanoma dif-ferentiation antigens and enhances antitumor immunity. J Immunol2006;176:6434–42.

26. Kim YH, Shin SM, Choi BK, Oh HS, Kim CH, Lee SJ, et al. Authentic GITRsignaling fails to induce tumor regression unless Foxp3þ regulatory t cellsare depleted. J Immunol 2015;195:4721–9.

27. Selby MJ, Engelhardt JJ, Quigley M, Henning KA, Chen T, Srinivasan M,et al. Anti-CTLA-4 antibodies of IgG2a isotype enhance antitumor activitythrough reduction of intratumoral regulatory T cells. Cancer Immunol Res2013;1:32–42.

28. Simpson TR, Li F, Montalvo-Ortiz W, Sepulveda MA, Bergerhoff K, Arce F,et al. Fc-dependent depletion of tumor-infiltrating regulatory T cells co-defines the efficacy of anti-CTLA-4 therapy against melanoma. J Exp Med2013;210:1695–710.


Mahne et al.




2017;77:1108-1118. Published OnlineFirst October 20, 2016.Cancer Res Ashley E. Mahne, Smita Mauze, Barbara Joyce-Shaikh, et al. Signaling in Agonistic GITR Targeting for Tumor ImmunotherapyDual Roles for Regulatory T-cell Depletion and Costimulatory

Updated version

10.1158/0008-5472.CAN-16-0797doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2016/10/20/0008-5472.CAN-16-0797.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/77/5/1108.full#ref-list-1

This article cites 28 articles, 17 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/77/5/1108.full#related-urls

This article has been cited by 10 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/77/5/1108To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-16-0797

http://cancerres.aacrjournals.org/content/suppl/2016/10/20/0008-5472.CAN-16-0797.DC1

http://cancerres.aacrjournals.org/content/77/5/1108.full#ref-list-1

http://cancerres.aacrjournals.org/content/77/5/1108.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/77/5/1108


Dual Roles for Regulatory T-cell Depletion and ... · Rat anti-mouse DTA-1 GITR antibody (S....

Documents

Transcript of Dual Roles for Regulatory T-cell Depletion and ... · Rat anti-mouse DTA-1 GITR antibody (S....