SATB1 Expression Governs Epigenetic Repression of PD-1 in Tumor-Reactive T Cells (original) (raw)

Immunity. Author manuscript; available in PMC 2018 Jan 17.

Published in final edited form as:

PMCID: PMC5336605

NIHMSID: NIHMS840382

Tom L. Stephen,1,6,7 Kyle K. Payne,1,7 Ricardo A. Chaurio,1 Michael J. Allegrezza,1 Hengrui Zhu,2 Jairo Perez-Sanz,1 Alfredo Perales-Puchalt,1 Jenny M. Nguyen,1 Ana E. Vara-Ailor,1 Evgeniy B. Eruslanov,3 Mark E. Borowsky,5 Rugang Zhang,2 Terri M. Laufer,4 and Jose R. Conejo-Garcia1,8,*

Tom L. Stephen

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Kyle K. Payne

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Ricardo A. Chaurio

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Michael J. Allegrezza

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Hengrui Zhu

2Gene Expression and Regulation Program The Wistar Institute, Philadelphia, PA 19104, USA

Jairo Perez-Sanz

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Alfredo Perales-Puchalt

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Jenny M. Nguyen

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Ana E. Vara-Ailor

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

Evgeniy B. Eruslanov

3Department of Surgery, University of Pennsylvania, Philadelphia, PA 19104, USA

Mark E. Borowsky

5Helen F. Graham Cancer Center, Christiana Care Health System, 4701 Ogletown-Stanton Road, Newark, DE 19713, USA

Rugang Zhang

2Gene Expression and Regulation Program The Wistar Institute, Philadelphia, PA 19104, USA

Terri M. Laufer

4Department of Rheumatology Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA

Jose R. Conejo-Garcia

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

1Tumor Microenvironment and Metastasis Program, The Wistar Institute, Philadelphia, PA 19104, USA

2Gene Expression and Regulation Program The Wistar Institute, Philadelphia, PA 19104, USA

3Department of Surgery, University of Pennsylvania, Philadelphia, PA 19104, USA

4Department of Rheumatology Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA

5Helen F. Graham Cancer Center, Christiana Care Health System, 4701 Ogletown-Stanton Road, Newark, DE 19713, USA

6Present address: AstraZeneca Pharmaceuticals PLC, Oncology Bioscience, 35 Gatehouse Drive, Waltham, MA 02451, USA

7Co-first author

8Lead Contact

SUMMARY

Despite the importance of programmed cell death-1 (PD-1) in inhibiting T cell effector activity, the mechanisms regulating its expression remain poorly defined. We found that the chromatin organizer special AT-rich sequence-binding protein-1 (Satb1) restrains PD-1 expression induced upon T cell activation by recruiting a nucleosome remodeling deacetylase (NuRD) complex to Pdcd1 regulatory regions. Satb1 deficienct T cells exhibited a 40-fold increase in PD-1 expression. Tumor-derived Transforming Growth Factor β (Tgf-β) decreased Satb1 expression through binding of Smad proteins to the Satb1 promoter. Smad proteins also competed with the Satb1-NuRD complex for binding to Pdcd1 enhancers, releasing Pdcd1 expression from Satb1-mediated repression, _Satb1_-deficient tumor-reactive T cells lost effector activity more rapidly than wild-type lymphocytes at tumor beds expressing PD-1 ligand (CD274), and these differences were abrogated by sustained CD274 blockade. Our findings suggest that Satb1 functions to prevent premature T cell exhaustion by regulating Pdcd1 expression upon T cell activation. Dysregulation of this pathway in tumor-infiltrating T cells results in diminished anti-tumor immunity.

Graphical abstract

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0001.jpg

INTRODUCTION

Costimulatory and co-inhibitory signaling mechanisms control T cell activation to prevent excessive immune responses. T cell activation induces the expression of Pdcd1, encoding the inhibitory receptor PD-1. In acute responses, PD-1 expression on T cells decreases rapidly; however, chronic exposure to antigenic stimulation, such as in the case of chronic viral infections, results in high levels of expression of PD-1, leading to impaired effector responses through a distinctive transcriptional program termed exhaustion (Barber et al., 2006; Pauken et al., 2016; Sen et al., 2016; Wherry et al., 2007). Tumor-infiltrating T cells also express high levels of PD-1. Accordingly, antibodies to PD-1 or its ligand, PD-L1, mediate durable cancer regression in a growing number of human cancers (Brahmer et al., 2012; Garon et al., 2015; Topalian et al., 2012, 2015; Zou et al., 2016). There is much interest in understanding the mechanisms underlying response or resistance to anti-PD-1-PD-L1 immunotherapy. Recent studies in melanoma indicate that the mutational load and CD274 (PD-L1) mRNA expression levels of tumors responsive to anti-PD-1 treatment are not higher than that of non-responsive tumors (Hugo et al., 2016). The frequency of PD-L1+ tumor cells appears to be a better predictor of anti-PD-1 effectiveness in lung cancer (Garon et al., 2015).

The genetic and epigenetic mechanisms that govern PD-1 expression in T cells have been primarily studied in settings of chronic viral infection (Austin et al., 2014; Bally et al., 2016; Kao et al., 2011; Lu et al., 2014; McPherson et al., 2014; Oestreich et al., 2008; Pauken et al., 2016; Sen et al., 2016; Young-blood et al., 2011). Multiple inhibitory mechanisms identified in chronic infections also play a role in the suppression of anti-tumor immunity; however, the tumor microenvironment has unique metabolic restrictions and orchestrates specific tolerogenic networks that differ from those seen in the context of viral diseases (Cui et al., 2013; Stephen et al., 2014; Zou, 2005). For instance, epigenetic changes induced by tumor-driven metabolic restrictions suppress T-cell activity (Zhao et al., 2016). In addition, TGF-β, an immunosuppressive mediator expressed at high levels in most established malignancies, does not play a crucial role in the pathogenesis of many viral infections (Flavell et al., 2010; Furuya et al., 2015; Garidou et al., 2012; Lewis et al., 2015; Wrzesinski et al., 2007).Thus, tumors induce dendritic cells (DCs), as well as conventional and regulatory T (Treg) cells to secrete TGF-β, which impairs effector T-cell activity through multiple mechanisms that include Foxp1-mediated transcriptional repression (Stephen et al., 2014) and suppression of effector cytokines (Ahmadzadeh and Rosenberg, 2005).

PD-1 expression is regulated through epigenetic mechanisms that involve two critical regulatory regions upstream of the Pdcd1 promoter (Lu et al., 2014; McPherson et al., 2014; Oestreich et al., 2008; Youngblood et al., 2011). Epigenetic changes in T cell precursors are controlled throughout thymic development by the chromatin organizer Satb1 (Alvarez et al., 2000; Dickinson et al., 1992). In thymocytes, Satb1 organizes a highly looped, transcriptionally active chromatin structure that regulates coordinated expression of multiple genes at a single temporal time point (Cai et al., 2003). In _Satb1_−/− mice, T cell development arrests at the CD4+CD8+ stage (Alvarez et al., 2000). Little is known, however, about the role of Satb1 in mature T cells since the function of this molecule is highly cell context dependent (Yasui et al., 2002). Thus, Satb1 expression has been associated with Th2 lineage commitment in CD4+ T cells through Wnt-beta-catenin signaling (Notani et al., 2010), but the role of Satb1 in CD8+ T cells remains elusive.

In pursuing whether Satb1 contributes to mature CD8+ T cell function, we found that T cell receptor (TCR) activation increases Satb1 expression; conversely, Tgf-β induces reduced Satb1 levels. Unexpectedly, we found that incubation with Tgf-β also elicits higher PD-1 expression levels in activated T cells. We therefore hypothesized that Satb1 could contribute to restrain PD-1 expression. In this study, we show that Satb1 recruits a NuRD repressor complex to Pdcd1 enhancers, limiting the increase of Pdcd1 expression that ensues upon T cell activation. In contrast, Tgf-β signaling decreases the expression of Satb1, which contributes to T cell dysfunction in the tumor microenvironment by inhibiting the repression of PD-1.

RESULTS

Tgf-β Inhibits TCR-Activation-Dependent Increased Expression of Satb1 in Mature T Cells

To define how TCR activation influences Satb1 expression, we activated mature T cells of mouse and human origin with CD3-CD28 agonistic antibodies. T cell activation increased the expression of Satb1 in both CD8+ and CD4+ T cells (Figures 1A and 1B); Satb1 proteins amounts were higher in CD4+ T cells. To gain insight into potential negative regulators of Satb1 expression, we focused on Tgf-β, an immunosuppressive cytokine that is present in high amounts in the tumor microenvironment (Massagué, 2008; Stephen et al., 2014). TCR stimulation in the presence of Tgf-β resulted in reduced levels of Satb1 in both mouse and human CD8+ and CD4+ T cells (Figures 1A and 1B), suggesting that Satb1-mediated epigenetic regulation could play a role in T cell effector function, which would be counteracted Tgf-β. Supporting this proposition, OT1 TCR transgenic CD8 T cells also increased Satb1 expression upon recognition of specific antigen (SIINFEKL) presented by bone marrow dendritic cells (BMDCs), while addition of Tgf-β on day 0 or day 2 inhibited Satb1 increased expression (Figure 1C). Accordingly, chromatin immunoprecipitation (ChIP) using Smad2/3-specific antibodies resulted in an enrichment of a fragment ~600 bp upstream of the transcription initiation site of Satb1 promoter (around –600) when T cells were activated in the presence of Tgf-β, compared to control pull-downs with an irrelevant IgG (Figure 1D).

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0002.jpg

TCR-Activation-Induced Satb1 Expression Is Impaired by Tgf-β

(A) Expression levels of Satb1 in negatively purified mouse CD4 or CD8 T cell splenocytes stimulated or not with plate bound CD3 (5 μg/mL) and CD28 (1 μg/mL) for 30 hours, with or without Tgf-β1 (5 ng/mL). Representative of three independent experiments.

(B) Human CD4 and CD8 T cells at rest or CD3-CD28-activated with beads for 30 hr in the presence or the absence of Tgf-β1 (5 ng/mL).

(C) OT1 T cells activated for ~65 hr with BMDCs previously pulsed for 15 hr with 1 μg/mL of ovalbumin (Sigma-Aldrich), where Tgf-β (5 ng/mL) was added on day 0 or day 2. Representative of three independent experiments.

(D) Smad2/3 binding to the Satb1 promoter region. Chromatin was immunoprecipitated (IPed) with anti-Smad2/3 or control IgG from negatively immunopurified mouse CD8 T cells activated for 24 hr. Enrichment of the Satb1 promoter sequence in IPed chromatin was quantified by real-time qPCR. Pooled from two independent experiments with similar results.

(E) Proliferation of CD4+ and CD8+ T cells from CD4CreSatb1f/f versus Satb1f/f mice, activated for 3 days with CD3-CD28 beads. Representative of three independent experiments.

(F) Analysis of CD3 T cells in the spleen of 5- to 10-week-old CD4CreSatb1f/fOT1 versus Satb1f/fOT1 TCR transgenic mice.

(G) Mature Vβ5+ OT1 T cells in the periphery of CD4CreSatb1f/f OT1 mice. Representative of three independent experiments.

(H) T cells from indicated mice were labeled with Cell Trace Violet and incubated for ~65 hours at a 1:10 ratio with BMDCs pulsed with ovalbumin. Representative of three independent experiments.

(I) IFN-γ production by CD4 and CD8 T cell splenocytes from CD4CreSatb1f/f versus Satb1f/f mice stimulated with plate-bound CD3 (5 μg/mL) and CD28 (1 μg/mL) for 30 hr. Representative of three mice.

(J) Expression of different cytokine receptors in CD8 T cell splenocytes from CD4CreSatb1f/f versus Satb1f/f mice stimulated with plate-bound CD3 (5 μg/mL) and CD28 (1 μg/mL) plus 30 U of IL-2 for 30 hr. Representative of three mice.

Bar graphs represent mean ± SEM.

To define the role of Satb1 in effector T cells, we conditionally ablated Satb1 (Tesone et al., 2016) using CD4Cre transgenic transgene (CD4CreSatb1f/f mice). Proliferative responses were not impaired in _Satb1_-deficient and control CD4 and CD8 T cells from 5- to 10-week-old mice (Figure 1E), with only a slight reduction in the proportions of T cell splenocytes in CD4CreSatb1f/f mice (Figure 1F). In addition, lingering Satb1 expression in thymocytes from CD4CreSatb1f/f mice allowed the selection of OT1 T cells (Figure 1G), which effectively responded to OVA peptide stimulation (Figure 1H). In contrast, T cell selection was compromised in Vav1CreSatb1f/f mice (Figures S1A and S1B). Finally, no obvious differences in the production of IFN-γ (Figure 1I) or the expression of multiple cytokine receptors was observed in activated T cells from CD4CreSatb1f/f mice (Figure 1J). Therefore, deletion of Satb1 in double-positive thymocytes allows the selection and maturation of functional CD8+ and CD4+ T cells, which we subsequently used to determine the role of Satb1 in T cell function.

Tgf-β Induces Increased Expression of PD-1 in Activated T Cells by Decreasing Satb1 Expression

As expected, TCR stimulation with plate-bound agonistic CD3 antibodies increased surface PD-1 expression in both CD8+ and CD4+ T cells. However, identical activation stimuli induced ~40-fold higher PD-1 expression (MFI) levels in previously negative _Satb1_-deficient peripheral CD8 T cells, indicating that Satb1 is required for PD-1 repression (Figure 2A). Accordingly, OVA-antigen-activated OT1 transgenic T cells from CD4CreSatb1f/f mice expressed much higher levels of PD-1 compared to OT1 T cells from control (Satb1f/f) mice (Figure 2B), and PD-1 expression was most pronounced in activated CD44+ or CD69+ T cells (Figure 2C). Increased PD-1 expression in _Satb1_-deficient T cells was not the result of thymic selection or pre-activation artifacts because mature activated T cells from _Satb1_f/fCre-ERT2+RosaYFP mice, in which tamoxifen-induced Cre recombinase induces excision of loxP-flanked Satb1 alleles and expression of yellow fluorescent protein (YFP) as a reporter of ablation, also showed increases in PD-1 expression (Figure 2D). No other co-stimulatory marker or checkpoint inhibitory molecule analyzed was affected to the same extent as PD-1 (Figure 2E). Of note, unlike the strong increase in mRNA expression reported in cells from constitutive _Satb1_-deficient mice, which rarely survive for more than 2 weeks (Alvarez et al., 2000), we did not observe increased PD-1 expression in CD4CreSatb1f/f or VavCreSatb1f/f double-positive thymocytes (Figure 2F), suggesting that Satb1 ablation could have different consequences in adult mice.

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0003.jpg

Satb1-Deficient T Cells Overexpress PD-1 upon TCR Activation

Negatively immunopurified CD4CreSatb1f/f and control Satb1f/f T cell splenocytes were activated for 60 hr with CD3-CD28 beads.

(A) Expression levels of PD-1 in activated versus unstimulated cells in both populations.

(B) Negatively purified CD8 T cell splenocytes from CD4CreSatb1f/fOT1 and control Satb1f/fOT1 transgenic mice were incubated for ~65 hours at a 1:10 ratio with BMDCs, which were pulsed with 50 ug/mL full-length ovalbumin for 3–15 hr. PD-1 expression in resting vs. activated T cells in both groups is shown.

(C) PD-1 expression in antigen-experienced (CD44+) versus CD44− (top), and recently activated (CD69+) versus CD69− (bottom) splenic T cells from Satb1-deficient or WT mice.

(D) Expression of PD-1 in CD44+CD69+_Satb1_f/fCre-ERT2+RosaYFP T cells activated with plate-bound CD3 and soluble CD28 antibodies for 36 hr in the presence of 1 μg/mL of tamoxifen compared to identically activated T cells incubated with vehicle.

(E) Negatively purified CD8 T cell splenocytes from CD4CreSatb1f/fOT1 and control Satb1f/fOT1 transgenic mice were incubated for ~65 hours at a 1:10 ratio with OVA-pulsed BMDCs, and surface expression of the indicated inhibitory and co-stimulatory molecules were compared.

(F) PD-1 expression in ≥ 6-week-old CD4CreSatb1f/f, VavCreSatb1f/f, and control Satb1f/f double positive thymocytes. All representative of ≥ 3 independent experiments.

Because Tgf-β inhibits Satb1 in activated T cells, we next investigated whether Tgf-β could unleash PD-1 expression by decreasing the expression of its Satb1 repressor. Supporting this proposition, the presence of Tgf-β during TCR stimulation was sufficient to induce an increase in PD-1 expression in mouse and human CD8+ and CD4+ T cells (Figures 3A and 3B). In contrast, tamoxifen-treated YFP+ (_Satb1_-deficient) Satb1f/fCre-ERT2+RosaYFP T cells did not show further PD-1 increased expression in the presence of Tgf-β, unlike control T cells activated without tamoxifen (Figure 3C). Furthermore, although some surviving resting T cells spontaneously showed increased PD-1 expression in the absence of exogenous cytokines, the addition of Tgf-β had no effect on PD-1 expression (Figure 3D).

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0004.jpg

Tgf-β Regulates PD-1 Expression in Activated T Cells

(A) Expression levels of PD-1 in negatively purified WT mouse CD4 and CD8 T cell splenocytes at rest or CD3-CD28-activated (beads) for 48 hr with 50 U of IL-2/mL in the presence or the absence of Tgf-β1 (5 ng/mL). Representative of two independent experiments.

(B) PD-1 expression in CD4 and CD8 T cells negatively purified from human peripheral blood, at rest or CD3-CD28-activated for 60 hr, in the presence or the absence of Tgf-β1 (5 ng/mL). Representative of ≥ 3 independent experiments.

(C) Expression of PD-1 in negatively immunopurified Satb1f/fCre-ERT2+RosaYFP T cell splenocytes activated CD3-CD28-coated beads for 32 hr in the presence of 1 μg/mL of tamoxifen (bottom) ± 5 ng/mL of Tgf-β compared to identically activated Satb1f/fCre-ERT2+RosaYFP T cells without tamoxifen treatment (top). Representative of three independent experiments.

(D) Expression levels of PD-1 in negatively purified WT mouse CD4 and CD8 T cell splenocytes at rest without cytokines for 48 hr in the presence or the absence of Tgf-β1 (5 ng/mL). Representative of two mice.

(E) T cell splenocytes from WT and dnTGF-βRII mice were primed for 9 days against BMDCs pulsed with UV+gamma-irradiated ID8-Defb29/Vegf-a tumor cells. Tumor-antigen-primed T cells were then i.p. transferred into congenic day 35 ID8-Defb29/Vegf-a tumor-bearing mice, and peritoneal wash cells were analyzed by flow cytometry 3 days later. Representative of three different mice.

(F) Smad2/3 binding to the CR-C regulatory region upstream of PD-1. Chromatin was IPed with anti-Smad2/3 or control IgG from negatively immunopurified mouse CD8 T cells activated for 24 hr. Enrichment was quantified by real-time qPCR. Pooled from two independent experiments with comparable results.

(G) Expression of PD-1 in negatively immunopurified CD4 and CD8 T cell splenocytes from Satb1f/f (top) versus CD4CreSatb1f/f(bottom) mice activated CD3-CD28-coated beads for 30 hr in the presence or the absence of 5 ng/mL of Tgf-β. Representative of two independent experiments.

(H) _Rag1_−/− mice received different combinations of negatively immunopurified, aged-matched control Satb1f/f versus CD4CreSatb1f/f CD4 and CD8 T cell splenocytes I.V. (106 and 500,000, respectively). Expression of PD-1 in CD3+CD8+ T cells was analyzed in peripheral blood after 7 days (four mice per group, with similar results in each mouse). * p <0.05; ** p <0.01, unpaired t test.

Bar graphs represent mean ± SEM.

To demonstrate that Tgf-β signaling at tumor beds is sufficient to decrease Satb1 expression in tumor-reactive T cells, we enriched naïve T cell splenocytes from wild-type (WT) mice versus mice carrying a dominant-negative Tgf-βR type II (dnTGF-βRII; in which Tgf-β signaling is blocked [Chen et al., 2005; Gorelik and Flavell, 2000]) for tumor reactivity by priming them against BMDCs pulsed with double (UV+gamma) irradiated ID8-Defb29-Vegf-a ovarian epithelial tumor cells. As reported, this system elicits increases in the expansion of tumor-reactive T cells (Nesbeth et al., 2009, 2010). As shown in Figure 3E, Tgf-β-sensitive T cells, but not their equally primed Tgf-β-resistant counterparts, exhibited decreased Satb1 expression after only 3 days in the peritoneal cavity of orthotopicID8-Defb29-Vegf-a congenic (CD45.1) tumor-bearing mice.

A recent report identified that Tgf-β-driven PD-1 expression is associated with binding of Smad3 to PD-1 promoter regions (Park et al., 2016). Our ChIP-PCR experiments confirmed that, in response to Tgf-β, Smad2/3 bind to CR-C, one of the conserved upstream regulatory regions hypersensitive to DNase I that control PD-1 expression in response to CD8 T cell activation (Lu et al., 2014; Oestreich et al., 2008) (Figure 3F). However, Tgf-β-driven increased PD1 expression was Satb1 dependent because _Satb1_-deficient T cells did not exhibit higher levels of PD-1 upon incubation with Tgf-β unlike their counterparts expressing endogenous Satb1 (Figures 3A, 3C, and 3G). Taken together, these data indicate that Satb1 is required for inhibition of PD-1 during T cell activation. By decreasing Satb1 expression, and also by driving the binding of Smad proteins to PD-1 regulatory regions, Tgf-β releases Satb1-mediated repression of PD-1, leading to PD-1 increased expression, which has obvious implications to understand T cell exhaustion in the tumor microenvironment.

Satb1 Recruits a NuRD Complex to PD-1 Enhancers to Repress PD-1 Expression through Histone Deacetylation

Because both CD4 and CD8 T cells undergo Satb1 ablation, we next aimed to rule out that differences in PD-1 expression in CD8 T cells are the result of Satb1-dependent differences in CD4 help. For that purpose, we transferred Rag KO mice with different combinations of Satb1+ versus _Satb1_-deficient CD4 and CD8 T cells from aged-matched mice. As shown in Figure 3H, Satb1-dependent differences in PD-1 expression and activation markers are maintained in the presence of both _Satb1_-competent versus _Satb1_-deficient CD4 T cells, further supporting that PD-1 deregulation in Satb1-deficient T cells results from an intrinsic lymphocyte defect.

To understand how Satb1 regulates PD-1 expression at a molecular level, we next used a mass-spectrometry-based approach to define the Satb1 interactome in primary CD3-CD28-activated murine CD8 T cells transduced with a Flag-tagged Satb1. Immunoprecipitation of flagged Satb1 after crosslinking with anti-Flag-coated beads and posterior elution with Flag peptides showed that Satb1physically interacts with the histone deacetylase Hdac2, Gatad2a (p66α), Gatad2b (p66β), Chd4, and Mta2, typical components of a chromatin remodeling, histone deacetylase-containing NuRD repressor complex (Kloet et al., 2015) (Tables S1 and S2). Of note, similar interactions have been reported in immature DP thymocytes (Yasui et al., 2002). In addition, Satb1 co-immunoprecipated with Bcl11b, also known to recruit repressive NuRD complexes to promoter regions in T cells (Avram and Califano, 2014). Confirming the interaction of Satb1 with components of the NuRD complex, further immunoprecipitation of p66α (Figure 4A), Hdac2 (Figure 4B), and Cdh4 (Figure 4C) in activated primary CD8 T cells from Satb1-competent mice pulled down a protein of ~103kD specifically detected by an anti-Satb1 antibody, while no signal was detected using a control IgG or, as expected, in equally treated CD8 T cells from CD4CreSatb1f/f mice in any condition (Figures 4A–4C). Accordingly, Satb1 immunoprecipitation under identical conditions pulled down a ~260kD protein detected with specific anti-Cdh4 antibodies (Figure 4D).

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0005.jpg

Satb1 Physically Interacts with Other Members of the NuRD Repressor Complex

Negatively immunopurified CD8 T cell splenocytes from CD4CreSatb1f/f and control Satb1f/f mice were activated for 60 hr with CD3-CD28 beads.

(A) p66α was IPed from the extracted proteins, followed by immunoblotting for Satb1. Irrelevant rabbit IgG was used as an additional control.

(B) IP using anti-Hdac2 antibodies under identical conditions.

(C) Same using anti-Cdh4 antibodies.

(D) Reverse IP with Satb1 and immunoblotting for Cdh4. All representative of two independent experiments.

See also Tables S1 and S2.

To define whether Satb1 recruitment of a NuRD complex specifically to PD-1 enhancers explains Satb1-mediated repression of PD-1 during T cell activation, we next performed ChIP of Satb1 using WT and _Satb1_-deficient activated CD8 T cells. We again focused on CR-B and CR-C, two upstream regulatory regions that govern PD-1 expression in response to CD8 T cell activation (Lu et al., 2014; Oestreich et al., 2008), in particular in the sequence of CR-C amplified with DNA ChIP-ed with Smad2/3 antibodies upon Tgf-β signaling (Figure 3F), located immediately downstream of putative Smad and NFATc1 binding sites (Figure 5A). Confirming Satb1 binding to PD-1 enhancers, Satb1 precipitation with two different antibodies specifically pulled down sequences of both CR-B and CR-C regions (Figures 5B and S2). Furthermore, ChIP of other components of the NuRD complex that physically interact with Satb1, including Gatd2a (p66α) and Cdh4, specifically pulled down both enhancers in a Satb1-dependent manner (Figures 5C and 5D). In contrast, precipitation of Hdac2, which along with Hdac1 typically binds to the scaffold generated by other elements of canonical NuRD complexes (Dege and Hagman, 2014), only pulled down the CR-C region in a Satb1-dependent manner, while only a non-significant trend was found for CR-B occupancy (Figure 5E). Accordingly, activation of _Satb1_-competent CD8 T cells in the presence of specific HDAC1-2 inhibitors increased histone acetylation at CR-C, but not at CR-B (Figure 5F), indicating that Satb1 recruits a classical NuRD complex with Hdac1/2 activity primarily to the distal enhancer. Nevertheless, histone acetylation at both CR-B and CR-C, which controls PD-1 expression (Lu et al., 2014), was still completely dependent on Satb1 occupancy because ChIP with specific acetylated histone 3 (H3) antibodies from activated Satb1-deficient CD8 T cells showed an enrichment of both enhancers compared to identically activated (Satb1+) control lymphocytes (Figure 5G).

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0006.jpg

Satb1 Recruits a NuRD Complex to CR-B and CR-C Regulatory Elements Upstream of Pdcd1 Transcription Start Site, Resulting in Histone De-acetylation

(A) Schematic depiction of the CR-C regulatory region specifically amplified with DNA ChIP-ed with both anti-Smad2/3 and anti-Satb1 antibodies, but not control IgGs.

(B) Chromatin was IPed with anti-Satb1 (clone#EPR3895) or control IgGs from negatively immunopurified control Satb1f/f (black) versus CD4CreSatb1f/f (red) mouse CD8 T cells, previously CD3-CD28-activated for 30 hr. Percent of input before IP (1% gel input values) was quantified by real-time qPCR in chromatin IPed with anti-Satb1 antibodies (clone#P472) versus irrelevant IgGs using primers specific for CR-B and CR-C sequences.

(C–E) ChIp-PCR quantification of the same CR-B and CR-C sequences using Satb1f/f/CD4CreSatb1f/f mouse CD8 T cells handled as above IPed with specific antibodies for the Gatd2a (C), Cdh4 (D), or Hdac2 (E) elements of the NuRD complex.

(F) ChIP PCR quantification of CR-B and CR-C in chromatin IPed with anti-H3Ac antibodies versus irrelevant IgGs using Satb1f/fCD8 T cells activated in the presence of 1 μM of the HDAC1/2 inhibitor CI-994.

(G) Same experiment using activated Satb1f/f versus CD4CreSatb1f/f CD8 T cells.

(H and I) ChIP PCR quantification of CR-B and CR-C in chromatin IPed with anti-H3Ac (H) or anti-Satb1 (I) antibodies versus irrelevant IgGs from with Satb1f/f CD8 T cells activated in the presence versus the absence of Tgf-β1 (5 ng/mL). All pooled from ≥ three independent experiments with ≥ three replicates. * p <0.05, unpaired t test.

Bar graphs represent mean ± SEM. See also Figure S2.

Furthermore, consistent with the Satb1 repressive activity of Tgf-β and its role in PD-1 increased expression, incubation with Tgf-β during TCR activation reproducibly increased histone acetylation at both CR-B and CR-C regions upstream of the PD-1 promoter locus (Figure 5H). In addition, Tgf-β signaling decreased Satb1 occupancy at both CR-B and CR-C (Figure 5I). Because Tgf-β-dependent Smad proteins do not bind to CR-B, these results suggest that Tgf-β-induced Satb1 decreased expression, rather than competition with Smad proteins for binding to CR-C, is the main mechanism of PD-1 de-repression driven by Tgf-β, consistent with the fact that Tgf-β does not increase PD-1 expression in the absence of Satb1 (Figure 3C). Therefore, Satb1 occupancy of the enhancers that govern PD-1 expression upon T cell activation regulates histone acetylation. At CR-C, Satb1 specifically recruits a NuRD repressor complex, while at CR-B, Satb1 occupancy drives histone deacetylation independently of Hdac1-2 activity. By inhibiting TCR activation-mediated Satb1 expression, Tgf-β releases the PD-1 transcriptional machinery from Satb1-mediated histone de-acetylation, leading to PD-1 increased expression.

_Satb1_-Deficient CD8 T Cells Elicit Weaker Anti-tumor Effector Responses

Our results so far imply that the repressive activity of Tgf-β on Satb1 could modulate anti-tumor T cell responses by unleashing Pdcd1 from Satb1-mediated repression. To support the relevance of this hypothesis, we first investigated the expression of SATB1 in previously activated (CD45RA−) human CD8 T cells. As expected, CD45RA−CD8+ T cells infiltrating six different freshly dissociated human ovarian carcinoma specimens expressed lower levels of SATB1 compared to CD8+CD45RA+ T cells in the peripheral blood of healthy donors (Figure 6A). More importantly, human ovarian cancer-infiltrating CD45RA−CD8+ T cells expressing the lowest levels of PD-1 (Figure 6B, left), or the highest levels of IFN-γ (Figure 6B, right), also exhibited higher expression of SATB1 in ~50% of patients. Supporting that tumor microenvironmental signals decrease SATB1 expression to unleash PD-1 expression, SATB1 levels were lower in previously activated CD8 T cells infiltrating tumor beds compared to their autologous counterparts in the peripheral blood of five additional patients using a different anti-SATB1 antibody (Figure 6C). Corresponding differences in PD-1 expression were found between activated CD8 T cells in the tumor microenvironment versus matching peripheral blood (Figure 6D), consistent with decreased SATB1 levels playing a role in the hyporesponsiveness of tumor-infiltrating lymphocytes (Stephen et al., 2014).

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0007.jpg

Satb1 Expression Is Lower in Human Ovarian Cancer-Infiltrating CTLs Compared to Peripheral Blood

(A) SATB1 expression quantified through intracellular staining in CD45RA− (previously activated) CD8 T cells from buffy coats from the peripheral blood of four healthy donors, as well as in activated CD8 T cells infiltrating six freshly dissociated stage III-IV human serous ovarian carcinoma specimens. Gated on Aqua Live Dead (viability probe) negative, CD45RA−CD8+ T cells. Anti-SATB1, clone 14/SATB1.

(B) Expression of intracellular SATB1 and IFN-γ, and extracellular PD-1, in CD45RA−CD8+ T cells from freshly dissociated stage III-IV serous ovarian cancer patients. Gated on Zombie yellow (viability probe) negative, CD45RA−CD8+ T cells. Anti-SATB1, clone EPR3895.

(C) Intracellular SATB1 expression in CD45RA−CD8+ T cells from buffy coats from the peripheral blood of five additional advanced serous ovarian cancer patients compared to their counterparts in matching freshly dissociated tumors. Gated on Zombie yellow (viability probe) negative, CD45RA−CD8+ T cells. Anti-SATB1, clone EPR3895.

(D) Comparison of PD-1 expression in the same tumor-infiltrating versus autologous peripheral blood CD45RA−CD8+ T cells.

To experimentally demonstrate that Satb1 decreased expression in tumor-reactive T cells indeed contributes to accelerated malignant progression, we treated ID8-Defb29-Vegf-a tumor-bearing mice with tumor-antigen-primed naïve _Satb1_-deficient or WT T cells. We enriched naïve _Satb1_-deficient and WT T cells for tumor reactivity by priming them against BMDCs pulsed with double (UV+gamma) irradiated ID8-Defb29-Vegf-a ovarian epithelial tumor cells, a system that results in increased expansion of tumor-specific T cells (Nesbeth et al., 2009; Nesbeth et al., 2010). As shown in Figures 7A and 7B, the absence of Satb1 did not impair in vitro priming of T cells against cognate tumor antigen presented by BMDCs, as determined by both Granzyme and IFN-γ ELISPOT analysis. However, when these lymphocytes enriched for polyclonal tumor reactivity were transferred into the peritoneal cavity of congenic (CD45.1+) mice bearing orthotopic, syngeneic ID8-Defb29-Vegf-a ovarian cancer—a model in which increased PD-L1 expression drives malignant progression (Cubillos-Ruiz et al., 2009)—the anti-tumor activity of _Satb1_-deficient (PD-1high) T cells was suppressed in vivo in the tumor microenvironment to a much greater extent than the anti-tumor activity of their Satb1+ counterparts, as demonstrated by ELISPOT analysis using re-stimulated transferred (CD45.2+) T cells isolated through FACS-sorting from peritoneal wash (Figures 7A and 7B). Correspondingly, _Satb1_-deficient tumor-antigen-primed T cells accelerated, rather than delayed, the progression of established orthotopic ovarian tumors, while identically activated and adoptively transferred WT T cells elicited measurable protection against malignant progression (Figure 7C). Notably, Satb1-dependent differences in T cell anti-tumor activity were completely abrogated upon PD-L1 blockade in ovarian cancer-bearing hosts, supporting that differences in Satb1-controlled PD-1 expression are sufficient to explain our observed changes in effector function in vivo (Figures 7D and S3). Further illustrating differences in anti-tumor activity between _Satb1_-deficient and control T cells, comparable differences in outcome were found in mice bearing Lewis Lung Carcinoma treated with tumor-antigen-enriched Satb1+ versus _Satb1_−/− T cells (Figure 7E). To demonstrate intrinsic differences in the anti-tumor activity of CD8 T cells, we finally transferred _Rag1_−/− mice with Satb1+ CD4 T cells and either Satb1+ or _Satb1_-deficient CD8 T cells and challenged them with Lewis Lung Carcinoma. As shown in Figure 7F, _Satb1_-deficient CD8 T cells again showed decreased protective activity compared to their Satb1+ counterparts. Together, these data indicate that loss of Satb1-mediated PD-1 repression regulates the suppression of tumor-reactive T cells in vivo at PD-L1+ tumor beds, which contributes to blunting their anti-tumor activity, resulting in accelerated malignant progression.

An external file that holds a picture, illustration, etc. Object name is nihms-840382-f0008.jpg

Tumor-Induced Satb1 Decreased Expression Results in Anti-tumor T Cell Paralysis at Tumor Beds

(A and B) T cell splenocytes from Satb1f/f and CD4CreSatb1f/f mice were primed for 6 days against BMDCs pulsed with UV+gamma-irradiated ID8-Defb29-Vegf-a tumor cells. Tumor antigen-primed T cells were then i.p. transferred into congenic day 35 ID8-Defb29-Vegf-a tumor-bearing mice, FACS-sorted from peritoneal wash 3 days later, and subjected to Granzyme B (A, right) and Ifn-γ (B, right) ELISPOT analysis, again using ID8-Defb29-Vegf-a tumor-pulsed-BMDCs. An aliquot of day 6 primed T cells was kept in vitro for 3 days as a control and subjected to identical ELISPOT analysis for Granzyme B (A, left) and Ifn-γ (B, left). All representative of ≥ two independent experiments.

(C) T cells from Satb1f/f and CD4CreSatb1f/f mice were primed for 6 days as in (A) and (B) and transferred (106 per mouse) into the peritoneal cavity of ID8-Defb29-Vegf-a tumor-bearing mice at days 7 and 10 after tumor challenge. Differences in survival are shown and compared to PBS-treated mice. Representative of two independent experiments with 5 mice per group.

(D) Satb1-dependent differences in anti-tumor T cell activity disappeared when PD-L1 is neutralized with 100 mg of specific antibodies every 5 days, starting at day 7 after tumor challenge (representative of two additional independent experiments, each with five mice per group; with similar results).

(E) T cells from Satb1f/f and CD4CreSatb1f/f mice were primed as in (A–C) and transferred (106 per mouse) into the peritoneal cavity of Lewis Lung Carcinoma-bearing mice at days 7 and 10 after tumor challenge. Differences in survival are shown. Representative of two independent experiments, with five mice/group per experiment.

(F) _Rag1_−/− mice were challenged with Lewis Lung Carcinoma 10 days after receiving (I.V.) 106 negatively immunopurified CD4 T cells from Satb1f/f mice, admixed with 500,000 CD8 T cells from either CD4CreSatb1f/f or Satb1f/f mice. Differences in survival are shown (five mice per group). * p <0.05, unpaired t test or long-rank (survival).

Bar graphs represent mean ± SEM. See also Figure S3.

DISCUSSION

Here, we show that the expression of the genomic organizer Satb1 increases upon T cell activation in mature CD8+ and CD4+ T cells, driving histone de-acetylation at genomic regions that regulate the expression of PD-1, thus restraining the expression of PD-1 on the cell surface. Correspondingly, _Satb1_-deficient T cells failed to repress the elevation of inhibitory PD-1 upon TCR activation and quickly became hyporesponsive in vivo in the presence of PD-L1 in the tumor microenvironment.

A myriad of transcription factors (Austin et al., 2014; Kao et al., 2011; Lu et al., 2014), DNA methylation mechanisms (McPherson et al., 2014; Youngblood et al., 2011), and histone modifications (Lu et al., 2014; Oestreich et al., 2008) have been associated with high PD-1 expression in effector lymphocytes, collectively contributing to attenuate their effector activity in different inflammatory conditions (Bally et al., 2016). In addition, PD-1 expression levels, at least in CD4+ T cells, are elevated when lymphocytes are forced to use “energy-saving” oxidative metabolism instead of aerobic glycolysis, which is also associated with diminished effector activity (Chang et al., 2013). This begs the question of whether decreased expression of Satb1 in tumor-infiltrating lymphocytes, in addition to Tgf-β signaling, could be caused by glucose restrictions at tumor beds, collectively driving T cell hyporesponsiveness in solid tumors. In turn, it is also possible that changes in Satb1 levels also govern T cell metabolic re-programming. The role of Satb1 in metabolic demands that T cells face in the tumor microenvironment is currently under investigation.

Besides expressing high levels of PD-1, the global chromatin landscape of exhausted T cells differs substantially from that of effector lymphocytes (Pauken et al., 2016; Sen et al., 2016). Although our results demonstrate that Satb1 mediates PD-1 transcriptional repression directly through histone de-acetyla tion, it is tempting to speculate that the loop-forming activity of Satb1 could also play a role in differences in chromatin accessibility, contributing to the exhaustion phenotype. Elucidating whether changes in chromatin architecture in exhausted T cells are Satb1 dependent could be important to extend the success of anti-tumor immunotherapies. For instance, objective responses upon PD-L1 or PD-1 neutralization are only achieved in 12%–15% of ovarian cancer patients (Gaillard et al., 2016; Hamanishi et al., 2015). Satb1 transduction in tumor-reactive T cells could prevent the rapid transcriptional rewiring experienced by exhausted T cells after PD-L1 blockade, as recently reported (Pauken et al., 2016), allowing lymphocytes to acquire effector and memory attributes.

Besides governing histone de-acetylation at PD-1 regulatory regions, we cannot rule out that Satb1 regulates DNA or histone methylation. In fact, a transient loss of DNA methylation of the PD-1 regulatory locus occurs during T cell activation, and these regions remain completely de-methylated in exhausted CD8 T cells in chronic viral infections (Youngblood et al., 2011). However, recent studies identified that aging is associated with strong inverse correlation between methylation and SATB1 expression levels in CD8+ T cells (Tserel et al., 2015). Nevertheless, because SATB1 regulates multiple epigenetic mechanisms, it is theoretically possible that SATB1 loss or decreased expression (e.g., in response to TGF-β) could contribute to specifically de-methylating PD-1 enhancers.

Other mechanisms involving histone de-acetylation at both CR-B and CR-C regions have been previously associated with repression of PD-1 expression by Boss and colleagues (Lu et al., 2014). These seminal studies demonstrate the role of the transcription factor Blimp as a PD-1 repressor. However, our data suggest that Blimp and Satb1 act through different mechanisms: First, Satb1 deficiency results in much higher PD-1 expression upon TCR activation, compared to _Blimp_-deficient T cells (Lu et al., 2014). Second, we did not find Blimp in our Satb1 pull-downs in activated T cells, suggesting that they participate in different repressive mechanisms. In contrast, we found that Satb1 physically interacts with Bcl11b, previously associated with the recruitment of repressive NuRD complexes to promoterregionsin T-cells (Avram and Califano, 2014). Future studies should determine whether Bcl11b also co-localizes with Satb1 and NuRD elements at PD-1 regulatory regions or whether its repressive activity depends on Satb1 in any immune compartment.

Consistent with a recent report (Park et al., 2016), we confirmed that translocated Smad proteins occupy crucial regulatory regions upstream of the Pdcd1 transcriptional start site in response to Tgf-β. Interestingly, we found that the segment of CR-C occupied by Satb1 overlaps with, or is adjacent to, the Smad protein binding site. Accordingly, Satb1 occupancy was reduced at those locations upon Tgf-β signaling. Interestingly, we found that Tgf-β also induced a reduction in Satb1 occupancy at CR-B, a sequence that is not co-occupied by Smad2/3. It is therefore likely that decreased levels of Satb1 protein resulting from Smad-mediated transcriptional repression at the Satb1 promoter, rather than competition between Satb1 and Smad proteins for binding to CR-C DNA regions, is the main mechanism driving Tgf-β-dependent PD-1 increases. Nonetheless, Tgf-β-mediated PD-1 increase is completely dependent on Satb1 because Tgf-β had no effect on PD-1 expression in _Satb1_-deficient T cells.

TGF-β is a lymphocyte inhibitor almost universally hyper-secreted in aggressive cancers by tumor cells, fibroblasts, dendritic cells, and conventional, as well as regulatory, T cells (Flavell et al., 2010; Wrzesinski et al., 2007), but it is not always hyper-secreted in, or crucial for the pathophysiology of, all acute or chronic infections (Furuya et al., 2015; Lewis et al., 2015). Further supporting the role of TGF-β in modulating PD-1 expression, a recent study identified that blocking TGF-β decreases both PD-1 and PD-L1 levels in a transplantation setting, precipitating graft rejection (Baas et al., 2016). In addition, another report found that HDAC inhibition also increased the expression of PD-1 ligands in melanoma and augmented the effectiveness of PD-1 blockade (Woods et al., 2015). Converging evidence therefore supports the potential of epigenetic interventions or genetic manipulation of primary T cells to prevent PD-1-mediated inhibition of protective anti-tumor immunity against multiple malignancies. Understanding the full spectrum of activities of the genomic organizer SATB1 in the process of T cell exhaustion could be required for these advances.

EXPERIMENTAL PROCEDURES

Animals and Human Samples

Genetically deficient Satb1 mice were generated in the Wistar Institute's transgenic facility on a C57BL/6 background (Tesone et al., 2016). Satb1f/f mice were crossed with CD4Cre mice (Taconic, 4196M). Satb1f/f mice were bred with Cre-ERT2+ mice and RosaYFP mice (both from Jackson Labs) to generate Satb1f/fCre-ERT2+RosaYFP mice. WT C57BL/6 mice used for in vivo tumor challenge were purchased from Charles River. All animals were maintained and used in accordance with the Institutional Care and Use Guidelines of the Wistar Institute.

ID8 cells (Roby et al., 2000) were provided by K. Roby (Department of Anatomy and Cell Biology, University of Kansas) and retrovirally transduced to express Defb29 and Vegf-a (Conejo-Garcia et al., 2004). 106 per mouse ID8-Defb29-Vegf-a cells or 2 × 105 Lewis Lung carcinoma cells (ATCC) were intraperitoneally (i.p.) administered to induce peritoneal tumors.

Human ovarian carcinoma tissues were procured through Research Pathology Services at Dartmouth Medical Center and through Helen F. Graham Cancer Center under IRB regulations. Tumors were freshly dissociated and cryopreserved, along with buffy coats from peripheral blood. Viable cells in thawed samples were gated for FACS analysis using a Zombie Yellow viability kit (Bio-Legend).

Peripheral blood lymphocytes were obtained by leukapheresis-elutriation by the Human Immunology Core at the University of Pennsylvania, and Miltenyi bead purified.

Human and mouse T cells were labeled with Cell Trace Violet (Invitrogen) and stimulated with either agonistic CD3-CD28 beads (Dynabeads, Life Technologies).

Generation of DCs, Tumor Antigen Priming, and Tumor Studies

Mononuclear bone marrow cells were collected from femurs and tibias of adult mice (5–8 weeks) and incubated in complete RPMI (10% heat-inactivated Fetal Bovine Serum, 0.5 mM Sodium Pyruvate, 48,6 uMβ-Mercaptoethanol, 2 mM L-glutamine, 100 I.U/mL Penicillin, 100 ug/mL Streptomycin) with recombinant mouse GM-CSF (20 ng/mL; Peprotech) or 10% conditioned medium from B78H1-GM-CSF-producing cell line, with periodic media renewal every 3 days. Non-adherent cells were harvested at the indicated time points for further analysis.

In vitro tumor antigen priming of CD45.2+CD4CreSatb1f/f and control CD4CreSatb1wt/f littermate control T cells was performed as reported (Nesbeth et al., 2009, 2010; Stephen et al., 2014). Briefly, day 6 BMDCs from C57BL/6 mice were pulsed overnight with γ-irradiated (10,000 rad) and UV-treated (30 minutes) ID8-Defb29-Vegf-a mouse ovarian tumor cells or Lewis Lung Carcinoma cells at a 10:1 (DC:tumor cell) ratio. Tumor-antigen-pulsed BMDCs were then co-cultured with T cells enriched from CD4CreSatb1f/f mice or CD4CreSatb1wt/f littermates at 10:1 (T cell:DC) ratio with 10 ng/ml IL-2 and 0.5 ng/ml IL-7 (both from Peprotech) for 5 days in ELISPOT plate or 7 days in a 6-well plate.

ELISPOT assay were carried out using mouse Granzyme B (R&D Systems) and mouse IFN-γ (eBioscience) assay kits.

For in vivo tumor studies, 2 × 106 in vitro tumor antigen primed T cells were transferred into the peritoneal cavity of ID8-Defb29-Vegf-a tumor-or Lewis Lung carcinoma-bearing CD45.1+ congenic mice at the indicated days post-tumor challenge, as described previously (Stephen et al., 2014). Mice were either euthanized and adoptively transferred CD45.2+ T cells were FACS sorted and restimulated with tumor antigen primed BMDCs and assayed for IFN-γ and Granzyme B using ELISPOT or evaluated for disease progression and survival.

Immunoprecipitation

Immunopurified CD8 T cell splenocytes from CD4CreSatb1f/f and CD4CreSatb1f/f mice were stimulated in vitro with CD3-CD28 microbeads for 60 hrs. Cells pellets were resuspended in lysis buffer (50 mM Tris-HCl pH8.0, 1 mM EDTA, 0.5% NP-40, 300 mM NaCl, 10 mM NaF, protease inhibitors, PMSF) and rotated at 4°C for 20 min. The lysate was centrifuged at 12,000g at 4°C for 15 min, and supernatant was used for immunoprecipitation with the indicated antibodies (Satb1, rabbit clone#P472, Cell Signalling; Cdh4, rabbit clone#D8B12, Cell Signalling; Hdac2, Polyclonal, A300-705, Bethyl Laboratories; p66α, Abcam, ab87663; or control normal Rabbit IgG, #2729, Cell Signalling). Proteins were incubated overnight at 4°C and subsequently with Protein G Dynabeads (Life Technologies) for 1 hr. Beads were washed with NETN buffer three times, boiled in Laemmli sampling buffer, and subjected to western blot.

​

Highlights

Supplementary Material

1

2

ACKNOWLEDGMENTS

Support for Shared Resources was provided by Cancer Center Support Grant CA010815 to The Wistar Institute. This study was supported by R01CA157664, R01CA124515, R01CA178687, CA187392, The Jayne Koskinas & Ted Giovanis Breast Cancer Research Consortium at Wistar, Ovarian Cancer Research Fund Program Project Development awards, and the Department of Defense Ovarian Cancer Academy. M.J.A. was supported by T32CA009171. K.K.P. was supported by T32CA009140-39. A.P.-P. was supported by the Ann Schreiber Mentored Investigator Award (OCRF).

Footnotes

AUTHOR CONTRIBUTIONS

T.L.S. and K.K.P. design, performed, and analyzed most of the experiments and co-wrote the manuscript; R.A.C. performed the Satb1 regulation experiments; M.J.A., J.P.-S., and A.P.-P. performed in vivo experiments and provided support for ChIP experiments; H.R. performed co-IP experiments; M.E.B. provided clinical samples and expertise; E.B.E., R.Z., and T.M.L. contributed to the design of the study and provided intellectual support and technical guidance; J.M.N. and A.E.V.-A. processed and stored clinical specimens and supported in vivo experiments; and J.R.C.-G. oversaw and designed the study and experiments, analyzed data, and co-wrote the manuscript.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures, three figures, and two tables and can be found with this article online at http://dx.doi.org/10.1016/j.immuni.2016.12.015.

REFERENCES