Cancer Cell

Article

The Immunoreceptor TIGIT Regulates Antitumorand Antiviral CD8+ T Cell Effector Function

Robert J. Johnston,1 Laetitia Comps-Agrar,2 Jason Hackney,3 Xin Yu,1 Mahrukh Huseni,4 Yagai Yang,5 Summer Park,6

Vincent Javinal,5 Henry Chiu,7 Bryan Irving,1 Dan L. Eaton,2 and Jane L. Grogan1,*1Department of Cancer Immunology2Department of Protein Chemistry3Department of Bioinformatics and Computational Biology4Department of Oncology Biomarker Development5Department of Translational Oncology6Department of Translational Immunology7Department of Biochemical and Cellular Pharmacology

Genentech, 1 DNA Way, South San Francisco, CA 94080, USA

*Correspondence: grogan.jane@gene.com

http://dx.doi.org/10.1016/j.ccell.2014.10.018

SUMMARY

Tumors constitute highly suppressive microenvironments in which infiltrating T cells are ‘‘exhausted’’ by

inhibitory receptors such as PD-1. Here we identify TIGIT as a coinhibitory receptor that critically limits anti-

tumor and other CD8+ T cell-dependent chronic immune responses. TIGIT is highly expressed on human and

murine tumor-infiltrating T cells, and, in models of both cancer and chronic viral infection, antibody coblock-

ade of TIGIT and PD-L1 synergistically and specifically enhanced CD8+ T cell effector function, resulting in

significant tumor and viral clearance, respectively. This effect was abrogated by blockade of TIGIT’s comple-

mentary costimulatory receptor, CD226, whose dimerization is disrupted upon direct interaction with TIGIT in

cis. These results define a key role for TIGIT in inhibiting chronic CD8+ T cell-dependent responses.

INTRODUCTION

Endogenous immune responses are often unable to reject tu-

mors, and strategies to elicit de novo antitumor responses by

vaccination have so far evinced relatively limited therapeutic

efficacy (Chen and Mellman, 2013; Kalos and June, 2013; Pal-

ucka and Banchereau, 2013; van den Boorn and Hartmann,

2013). These limitations are likely due to the immunosuppressive

nature of tumor microenvironments in which infiltrating tumor-

specific T cells become ‘‘exhausted’’ or otherwise suppressed

so that proliferative capacity and effector function are severely

impaired (Crespo et al., 2013, Schietinger and Greenberg,

2014). Recent efforts to reactivate immune responses by antag-

onizing the inhibitory signals utilized by tumors have shown

promise, with antibody blockade of T cell coinhibitory receptors

such as programmed death 1 (PD-1)/PD-L1 or cytotoxic

T-lymphocyte antigen 4 (CTLA-4) achieving response rates of

up to 50% in bladder, lung, and renal cancers and melanoma

(Chen and Mellman, 2013; Powles et al., 2014, J. Clin. Oncol.,

abstract). These results have generated substantial interest in

identifying additional inhibitory receptors that are expressed by

T cells in tumor microenvironments and that may contribute to

the suppression of cancer immunosurveillance.

T cell coinhibitory receptors were first identified by their re-

straint of autoimmunity inmice (Krummel and Allison, 1995; Nish-

imura et al., 1999; Tivol et al., 1995) and have subsequently been

found to be critical regulators of T cell exhaustion in the context of

chronic viral infections (Wherry, 2011). Mice acutely infectedwith

the Armstrong strain of lymphocytic choriomeningitis virus

(LCMV) mount a robust immune response that rapidly clears

the infection. In contrast, the use of the Clone 13 strain of

LCMV results in a chronic infection, eliciting the progressive

Significance

Strategies to reactivate exhausted antitumor immune responses with antibody blockade of key T cell coinhibitory receptors

such as PD-1/PD-L1 or CTLA-4 have demonstrated transformational potential in the clinic. However, the coexpression of

multiple coinhibitory receptors on chronically activated T cells suggests that many inhibitory pathways may synergistically

modulate antitumor and other chronic immune responses. Here we report that TIGIT collaborates with PD-1/PD-L1 to

potently limit the effector function of chronically stimulated CD8+ T cells. We also characterize a mechanism of action in

which TIGIT disrupts the homodimerization and function of its complementary costimulatory receptor, CD226. These results

expand the scope of TIGIT’s role in immune responses and identify TIGIT as a potential target for tumor immunotherapy.

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 923

exhaustion of effector CD8+ T cells, particularly in the absence of

adequate CD4+ T cell help (Ahmed et al., 1984; Gallimore et al.,

1998; Sullivan et al., 2011; Zajac et al., 1998). PD-1 is a central

regulator of this process, and antibody blockade of PD-1/PD-

L1 can partially restore antiviral T cell responses and enhance

viral clearance (Barber et al., 2006). Subsequent studies revealed

that these so-called exhausted CD8+ T cells expressmultiple co-

inhibitory receptors, including PD-1, CTLA-4, LAG-3, and TIM-3,

that converge to keep chronically activated effector T cells in

check (Wherry et al., 2007). These receptors can function syner-

gistically, particularly with PD-1, suggesting that individual

coinhibitory receptors contribute distinct functions to collectively

limit T cell responses (Blackburn et al., 2009; Curran et al., 2010;

Nakamoto et al., 2009). Moreover, tumor-infiltrating T cells

resemble exhausted antiviral T cells in their expression of many

of the same coinhibitory receptors, which, when inhibited, can

reactivate antiviral responses (Mellman et al., 2011).

To more fully assess the array of immune-modulatory recep-

tors on tumor-infiltrating lymphocytes, we interrogated human

tumor samples using immune cell-specific gene sets. The

T cell coinhibitory receptor TIGIT (T cell immunoglobulin and im-

munoreceptor tyrosine-based inhibitory motif [ITIM] domain)

was identified as consistently highly expressed across multiple

solid tumor types. TIGIT is an immunoglobulin superfamily mem-

ber expressed on subsets of activated T cells and natural killer

(NK) cells (Yu et al., 2009), although its role in CD8+ T cells is un-

known. Genetic ablation or antibody blockade of TIGIT has been

shown to enhance NK cell killing and CD4+ T cell priming in vitro

and in vivo and can exacerbate the severity of CD4+ T cell-

dependent autoimmune diseases such as experimental autoim-

mune encephalitis (Goding et al., 2013; Joller et al., 2011; Levin

et al., 2011; Lozano et al., 2012; Stanietsky et al., 2009, 2013;

Stengel et al., 2012; Yu et al., 2009). Conversely, administration

of TIGIT-Fc fusion proteins or agonistic anti-TIGIT antibodies

suppressed T cell activation in vitro and CD4+ T cell-dependent

delayed-type hypersensitivity in vivo (Yu et al., 2009). Coinhibi-

tory receptors can suppress immune responses by direct

signaling in cis, by inducing ligand signaling in trans, and by indi-

rect competition with complementary costimulatory receptors

(Chen and Flies, 2013). In the case of TIGIT, ligation of its high-

affinity cognate receptor, poliovirus receptor (PVR) (Stengel

et al., 2012), induces dendritic cells to acquire a tolerogenic

phenotype characterized by increased expression of IL-10 and

decreased expression of IL-12 (Yu et al., 2009). PVR also inter-

acts with CD226 (DNAM-1), the costimulatory counterreceptor

to TIGIT, but at a lower affinity. Analogous to the complementary

costimulatory and coinhibitory receptor pair CD28 and CTLA-4,

CD226 competes with TIGIT for ligand (Stengel et al., 2012). The

expression kinetics of the two receptor pairs are also similar in

that expression of both TIGIT and CTLA-4 is induced upon

activation, whereas CD226 and CD28 are expressed by both

naive and effector T cells. CD226 has been shown to support

antiviral and antitumor immune responses (Cella et al., 2010;

Welch et al., 2012; Ramsbottom et al., 2014). The molecular

and functional relationships between TIGIT and CD226 are

poorly characterized, and TIGIT’s role in CD8+ T cell responses

and their exhaustion has been unknown. Here we show that

TIGIT is a critical regulator of antitumor and antiviral CD8+

T cell responses.

RESULTS

TIGIT Is Highly Expressed in Human Tumors and

Correlated with CD8+ T Cell Infiltration

To identify genes associated with tumor-infiltrating T cells, we

used a gene signature-based approach to interrogate gene

expression data from the Cancer Genome Atlas lung squamous

cell carcinoma (LUSC) collection (Network, 2012b). Using im-

mune cell-specific gene sets defined by the Immune Response

In Silico (IRIS) project (Abbas et al., 2005), we identified a highly

specific 15-gene signature for tumor-associated T cells (see

Experimental Procedures and Figure S1 available online). Within

this T cell signature, we identified several coinhibitory receptors

associated with T cell dysfunction in tumors, particularly PD-1

(Figure S1).We also identified TIGIT, a T cell coinhibitory receptor

not previously associated with antitumor responses (Figure S1

and Table S1). TIGIT expression was also highly correlated with

CD3ε expression (Spearman’s rank correlation coefficient (r) of

0.82; Figure 1A). Similarly, in colon adenocarcinoma (COAD),

uterine corpus endometroid carcinoma (UCEC), breast carci-

noma (BRCA), and kidney renal clear cell carcinoma (KIRC)

(Network, 2012a, 2012c, 2013; Kandoth et al., 2013), TIGIT and

CD3ε were highly correlated (r = 0.83–0.94; Figures 1B–1E).

Furthermore, the TIGIT:CD3ε expression ratio was increased

significantly in LUSC, COAC, UCEC, and BRCA tumor samples

compared with matched normal tissue (116%–419% increase;

Figures 1A–1D). The TIGIT:CD3ε expression ratio in KIRC sam-

ples was unchanged, although expression of both TIGIT and

CD3ε was much higher in tumor samples than in normal tissue

samples (Figure 1E). Thesedata indicated that TIGITwasupregu-

lated by T cells in a broad range of solid tumors.

TIGIT has been described previously as an inhibitor of CD4+

T cell priming with no known function in CD8+ T cells. TIGIT

expression in LUSC samples was highly correlated with CD8A

and to a lesser extent with CD4 (r = 0.77 and 0.48 respectively;

Figure 1F). Expression of TIGIT was also correlated with expres-

sion of its complementary costimulatory receptor, CD226, as

well as with expression of PD-1, a key mediator and marker of

T cell dysfunction in tumors and during other chronic immune

responses (r = 0.64 and 0.82 respectively; Figures 1G–1H).

These data suggested that tumor-infiltrating lymphocyte (TIL)

T cells, particularly exhausted CD8+ T cells, may express high

levels of TIGIT.

TIGIT and PD-1 Are Coordinately Expressed by Human

and Murine Tumor-Infiltrating Lymphocytes

To confirm protein expression of TIGIT by tumor-infiltrating

T cells, we assessed surface TIGIT expression in human non-

small-cell lung carcinomas (NSCLC) by fluorescence-activated

cell sorting (FACS). TIGIT was expressed by a large percentage

of NSCLC-infiltrating CD8+ T cells (58% TIGIT+; Figure 2A).

Consistent with the correlation between TIGIT and PD-1 RNA

expression found in Figure 1, nearly all CD8+ T cells expressing

TIGIT also expressed high levels of PD-1 (Figure 2B; Figure S2).

A smaller subset of NSCLC-infiltrating CD4+ T cells similarly

coexpressed TIGIT and PD-1 (28%TIGIT+; Figure 2C; Figure S2).

Interestingly, peripheral CD8+ and CD4+ T cells from NSCLC

tumor donors also expressed higher levels of TIGIT than cells

from healthy donors (Figures 2A and 2C). Similar results were

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

924 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

A

C

E

F G H

D

B

Figure 1. TIGIT Expression Is Elevated in Human Cancer and Strongly Correlated with CD8 and PD-1

Gene expression analyses of human cancers were performed as described (see Experimental Procedures). Scatter plots show per-gene count data normalized

by library size. Box and whisker plots show the variance-stabilized expression ratio of TIGIT and CD3ε.

(A) Correlation of TIGIT and CD3ε RNA expression in LUSC (red) and normal lung (black). r = 0.86. Quantification of TIGIT/CD3ε expression ratios is also shown.

LUSC ratio increase = 372%. ***p = 1.46 3 10�46.

(B) Correlation of TIGIT and CD3εRNA expression in COAD (red) and normal colon (black). r = 0.83. Quantification of TIGIT/CD3ε expression ratios is also shown.

COAD ratio increase = 116%. ***p = 3.66 3 10�6.

(C) Correlation of TIGIT andCD3εRNA expression in UCEC (red) and normal uterine endometrium (black). r = 0.87. Quantification of TIGIT/CD3ε expression ratios

is also shown. UCEC ratio increase = 419%. ***p = 7.41 3 10�5.

(D) Correlation of TIGIT andCD3εRNA expression in BRCA (red) and normal breast (black). r = 0.82. Quantification of TIGIT/CD3ε expression ratios is also shown.

BRCA ratio increase = 313%. ***p = 4.6 3 10�44.

(E) Correlation of TIGIT and CD3ε RNA expression in kidney renal clear cell carcinoma (red) and normal kidney (black). r = 0.94. Quantification of TIGIT/CD3ε

expression ratios is also shown.

(F) Correlation of TIGIT and CD8A (left) or TIGIT and CD4 (right) RNA expression in lung squamous cell carcinoma (red) and normal lung (black). r = 0.77 and 0.48,

respectively.

(G) Correlation of TIGIT and PD-1 (PDCD1) RNA expression in lung squamous cell carcinoma (red) and normal lung (black). r = 0.82.

(H) Correlation of TIGIT and CD226 RNA expression in lung squamous cell carcinoma (red) and normal lung (black). r = 0.64.

See also Figure S1 and Table S1.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 925

A

C

E

G H

F

D

B

Figure 2. TIGIT and PD-1 Are Coordinately Expressed by Human and Murine Tumor-Infiltrating Lymphocytes

(A–D) Analysis of lymphocytes from resected human NSCLC tumors, tumor-matched peripheral blood, and normal donor peripheral blood. Data are pooled from

three independently acquired sets of samples.

(A) Representative FACS plots of TIGIT expression by peripheral and tumor-infiltrating CD8+ T cells. Shown is the quantitation of TIGIT+ cells as a percentage of all

CD8+ T cells. *p < 0.05.

(B) Flow cytometry histogram representative of TIGIT expression by PD-1high (red) and PD-1low (blue) tumor-infiltrating CD8+ T cells.

(C) Representative FACS plots of TIGIT expression by peripheral and tumor-infiltrating CD4+ T cells. Shown is the quantitation of TIGIT+ cells as a percentage of all

CD4+ T cells.

(D) Flow cytometry histogram representative of TIGIT expression by PD-1high (red) and PD-1low (blue) tumor-infiltrating CD4+ T cells.

(E and F) BALB/c mice were inoculated with syngeneic CT26 colorectal carcinoma cells. Splenocytes and TILs were analyzed 14 days after inoculation when

tumors had reached approximately 200 mm3 in size. Data are representative of more than three independent experiments (n = 5-6). Shown is a representative

FACS plot of TIGIT expression by tumor-infiltrating (E) CD8+ T cells and (F) CD4+ T cells, with TIGIT+ cells boxed. Also shown is the quantitation of the frequency of

TIGIT+ T cells as a percentage of all T cells. *p = 0.0134, ***p < 0.0001.

(G) Flow cytometry histogram representative of TIGIT expression by PD-1high and PD-1low tumor-infiltrating CD8+ T cells and by splenic CD8+ T cells. Quantitation

of TIGIT MFI is also shown. **p = 0.0023.

(H) Flow cytometry histogram representative of TIGIT expression by PD-1high and PD-1low tumor-infiltrating CD4+ T cells and by splenic CD4+ T cells. Quantitation

of TIGIT MFI is also shown. ***p = 0.0002.

Error bars depict SEM. See also Figure S2.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

926 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

obtained in a set of matched colorectal carcinoma (CRC) and pe-

ripheral blood mononuclear cell (PBMC) samples (Figure S2).

TIGIT was not expressed by myeloid cells, tumor cells, or other

nonhematopoietic cells (Figure S2).

We then characterized TIGIT expression in the murine synge-

neic CT26 and EMT6 tumor models. Twoweeks postinoculation,

when CT26 and EMT6 tumors had grown to 150–200mm3 in size

and were robustly infiltrated by T cells, TIGIT was expressed by

approximately 50% of CD8+ TILs and 25%of CD4+ TILs at levels

similar to those of primary CD8+ T cells stimulated in vitro (Fig-

ures 2D and 2E; Figure S2). In both CD8+ and CD4+ murine

TILs, CD226 was constitutively expressed, and TIGIT and PD-1

expression remained highly correlated (Figures 2F and 2G; Fig-

ure S2). TIGIT was not expressed by tumor cells or other nonhe-

matopoietic cells (Figure S2).

These results confirmed that TIGIT was highly expressed by

TILs and that expression of TIGIT occurred in parallel with

expression of other coinhibitory receptors, most notably PD-1.

Given the central role of PD-1 in mediating T cell exhaustion,

we hypothesized that TIGIT could similarly act as a checkpoint

inhibitor of chronically stimulated or exhausted tumor-infiltrating

T cells.

TIGIT and PD-1 Blockades Synergistically Elicit CD8+

T Cell-Mediated Tumor Rejection

To test the hypothesis that TIGIT regulates antitumor responses,

mice were inoculated with CT26 tumor cells and, when tumors

A

B

C

Figure 3. TIGIT andPD-L1BlockadesSyner-

gistically Elicit Tumor Rejection

(A–C) BALB/c mice were inoculated subcutane-

ously with CT26 colorectal carcinoma cells in their

right thoracic flanks. When tumors reached

approximately 200 mm3 in size, mice were treated

with isotype control (black), anti-PD-L1 + control

(red), anti-TIGIT + control (blue), or anti-PD-L1 +

anti-TIGIT (purple) antibodies for 3 weeks. Data are

representative of more than three independent

experiments (A and B) or two independent exper-

iments (C) [n = 10 (A and B) or 7–10 (C)].

(A) Median (left) and individual (right) CT26 tumor

volumes over time. On day 12, ***p < 0.001 be-

tween mice treated with anti-PD-L1 + anti-TIGIT

and all other groups.

(B) Mouse survival over time.

(C) Approximately 60 days after initial inoculation,

mice that received anti-TIGIT + anti-PD-L1 and

reached CR, as well as naive control mice, were

(re)inoculated with CT26 cells in their left thoracic

flanks and inoculated with EMT6 breast carcinoma

cells in their mammary fat pads. Median (left) and

individual (right) tumor volumes for CT26 (squares)

and EMT6 (triangles) in CRmice (purple and green)

and naive mice (black and orange) are shown.

See also Figure S3.

were 150–200 mm3 in volume, treated

with blocking antibodies against TIGIT

(Yu et al., 2009), PD-L1, or a combination

of both for 3 weeks. Tumor growth was

assessed during and after treatment. In

this model, single-agent treatment with anti-TIGIT or anti-PD-

L1 was insufficient to decrease the tumor burden and increased

median survival only by approximately 3 days (Figures 3A and

3B). However, coblockade of TIGIT and PD-L1 led to a striking

reversal of tumor growth (75% decrease in mean tumor volume

after 16 days of treatment, p < 0.0001; Figure 3A). Themajority of

mice receiving the combination treatment achieved a complete

response (CR) for the duration of the study, even in the absence

of further treatment (Figures 3A and 3B). These effects were also

observed in CT26 tumor-bearing mice treated with a combina-

tion of blocking antibodies against TIGIT and PD-1 (data not

shown) and in mice inoculated with syngeneic EMT6 breast car-

cinoma cells (Figure S3). Furthermore, mice that previously

cleared CT26 tumors mounted a protective antitumor response

to reinoculated CT26 tumors but not to novel EMT6 tumors (Fig-

ure 3C). These data indicated the induction of tumor antigen-

specific immunity in treated mice.

Next, CT26-tumor bearing mice were subjected to temporary

CD8+ T cell ablation using depleting antibodies prior to treatment

with anti-TIGIT and anti-PD-L1. Peripheral CD8+ T cell depletion

was confirmed 5 days after treatment by FACS (data not shown).

Mice treated with anti-TIGIT and anti-PD-L1 were unable to

reject CT26 tumors when depleted of CD8+ T cells at the start

of treatment (1,532% increase in mean tumor volume after

17 days of treatment, p = 0.0004; Figures 4A and 4B). Addition-

ally, CD8+ T cell depletion impaired the ability of CR mice to

control reinoculated CT26 tumors (Figure 4C). These data

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 927

demonstrated that anti-TIGIT and anti-PD-L1 acted through

CD8+ T cells to elicit effective primary and secondary antitumor

immune responses.

TIGIT Regulates Tumor-Infiltrating CD8+ T Cell Effector

Function

Consequently, we assessed the functional consequences of

TIGIT and PD-L1 inhibition on TILs and tumor-draining lymph

node-resident T cells. CT26 tumor-bearing mice were treated

with anti-TIGIT and/or anti-PD-L1 antibodies as described

above. After 7 days of treatment, when tumors had grown to

approximately 500 mm3 in size, tumors and tumor-draining

lymph nodes were analyzed by FACS. Tumor-draining lymph

node-resident CD8+ T cells from mice treated with anti-TIGIT

alone, anti-PD-L1 alone, or anti-TIGIT plus anti-PD-L1 all ex-

hibited enhanced interferon g (IFNg) production relative to con-

trol mice (75%–113% increase, p < 0.001; Figures 5A and 5B).

However, only tumor-infiltrating CD8+ T cells from mice treated

with both anti-TIGIT and anti-PD-L1 showed a significant in-

crease in IFNg production (274% relative to isotype-treated co-

horts, p = 0.0001; Figures 5A and 5C), consistent with the current

understanding that tumor microenvironments are highly immu-

nosuppressive and can induce more profound levels of T cell

dysfunction. The frequencies of IFNg/tumor necrosis factor a

A

B

C

Figure 4. TIGIT/PD-L1 Coblockade Efficacy

Is Dependent on CD8+ T Cells

(A and B) Wild-type BALB/c mice were inoculated

with CT26 tumors. When tumors reached 100–

150mm3 in size,micewere temporarily depleted of

CD8+ T cells and treated with anti-TIGIT + anti-PD-

L1 or control antibodies as described in Figure 3.

Data are representative of two independent ex-

periments (n = 10).

(A) Median (left) and individual (right) CT26 tumor

volumes over time.

(B) Quantitation of CT26 tumor volumes 17 days

after the start of treatment. ***p = 0.0004.

(C) Wild-type BALB/c mice were inoculated with

CT26 tumors and treated with anti-TIGIT + anti-

PD-L1 and subsequently rechallenged with CT26

tumors (as described in Figure 3) with temporary

depletion of CD8+ T cells at the time of rechallenge.

Data are representative of two independent ex-

periments (n = 5). Median (left) and individual (right)

CT26 tumor volumes over time are shown.

Error bars depict SEM.

(TNF-a) dual-producing CD8+ T cells in

tumors and tumor-draining lymph nodes

followed similar patterns (Figures 5D

and 5E). TIGIT/PD-L1 coblockade also

induced slightly higher frequencies of

CD8+ TILs but not tumor-draining lymph

node-resident CD8+ T cells (Figure S4).

Unlike CD8+ T cells, effector CD4+ T cell

activation and cytokine production were

unaffected by treatment (Figures 5B–5E;

Figure S4). Some immunoreceptor-tar-

geting antibodies, such as anti-CTLA-4

and anti-GITR, have been shown to elicit tumor rejection primar-

ily by mediating regulatory T cell (Treg) cell depletion (Bulliard

et al., 2013; Simpson et al., 2013). In contrast, treatment with

anti-TIGIT and anti-PD-L1 did not reduce tumor-infiltrating Tregfrequencies (Figure S4).

These data suggested that inhibition of TIGIT and PD-L1

selectively and synergistically enhanced CD8+ T cell effector

function, in agreement with the requirement for CD8+ T cells in

anti-TIGIT/PD-L1 treatment efficacy (Figure 4).

TIGIT Enforces CD8+ T Cell Exhaustion during Chronic

Viral Infection

To determine whether TIGIT had a broader role in other sup-

pressed T cell responses, we used models of viral infection

(Barber et al., 2006; Blackburn et al., 2009; Wherry et al.,

2007). Mice acutely infected with the Armstrong strain of

LCMV mount a robust immune response and rapidly clear the

virus. In this model, TIGIT was highly expressed by approxi-

mately 80% of effector (Teff) CD8+ T cells and 40% of CD4+ Teff

cells (Figure 6A). Interestingly, TIGIT expression was highest on

CD8+ Teff cells that also expressed high levels of PD-1, consis-

tent with TIGIT/PD-1 coexpression in tumors (Figure 6B).

Furthermore, in mice chronically infectedwith the Clone 13 strain

of LCMV, TIGIT was highly expressed by exhausted Teff cells, as

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

928 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

marked by high PD-1 expression, but not by PD-1low Teff cells,

central memory T cells, or naive T cells, consistent with a recent

report (Figure 6C; Doering et al., 2012).

We then tested mice in which TIGIT expression was selec-

tively ablated in T cells (TIGITfl/fl;CD4cre; Figure S5). The

absence of TIGIT had no appreciable effects on T cell develop-

ment or the overall response to acute viral infection with the

Armstrong strain of LCMV (Figure S5). However, upon chronic

infection with the Clone 13 strain of LCMV, viral loads were

reduced significantly, and CD8+ and CD4+ T cell cytokine pro-

duction was increased significantly in TIGITfl/fl CD4-cre+ mice

relative to wild-type littermates (Figure S5). These data sug-

gested that TIGIT deficiency could enhance T cell responses

to chronic viral infection, although were unable to distinguish

between potential roles for TIGIT during the early stages of

T cell priming and/or the later stages of chronic T cell stimula-

tion and dysfunction.

To distinguish these two possibilities, we treatedmice with es-

tablished chronic LCMV infection with anti-TIGIT and/or anti-PD-

L1. Blockade of the PD-1/PD-L1 pathway alone significantly

reduced viral titers in mice compared with mice treated with iso-

type-matched control antibodies (68% decrease, p = 0.0004;

Figure 6D), consistent with previous reports (Barber et al.,

2006). Treatment with anti-TIGIT alone did not significantly affect

viral titers, but anti-TIGIT did act synergistically with anti-PD-L1

to enhance viral clearance (74% decrease in viral titers over

treatment with anti-PD-L1 treatment alone, p = 0.0045; Fig-

ure 6D). Functional characterization of T cell effector function

in chronically infected mice confirmed that PD-L1 blockade

alone robustly enhanced CD8+ T cell activation and moderately

increased IFNg production relative to treatment with matched

isotype control antibodies (88% increase and 37% increase,

respectively, p < 0.05; Figure 6E). Mice treated with anti-TIGIT

alone did not have enhanced CD8+ or CD4+ T cell activation or

cytokine competency, suggesting that TIGIT blockade alone

was insufficient to restore effector function in T cells that were

already exhausted (Figure 6E). However, as in tumor models,

TIGIT/PD-L1 coblockade significantly enhanced CD8+ T cell

effector function, but not CD4+ T cell effector function, in mice

compared with mice treated with anti-PD-L1 alone (93% in-

crease, p = 0.0050; Figure 6F; Figure S5). Similar effects on

T cell expansion and effector function were observed in LCMV

gp33 antigen-specific T cells (Figure S5). Taken together with

our findings in tumor-bearing mice, these data demonstrate a

A

B

D E

C

Figure 5. TIGIT Regulates Tumor-Infiltrating

CD8+ T Cell Effector Function

BALB/C mice were inoculated subcutaneously

with CT26 colorectal carcinoma cells in their right

thoracic flanks and treated with control, anti-PD-

L1 + control, anti-TIGIT + control, or anti-PD-L1 +

anti-TIGIT, as described in Figure 3. Tumor-drain-

ing lymph node (dLN)-resident and tumor-infil-

trating T cells were analyzed by flow cytometry

7 days after the start of treatment. Data are

representative of more than three independent

experiments (n = 5).

(A) Representative FACS plots of dLN-resident and

tumor-infiltrating CD8+ T cells after stimulation

ex vivo, with IFNg+ cells boxed.

(B) Quantitation of IFNg-producing dLN-resident

CD8+ and CD4+ T cells as percentages of total

dLN-resident CD8+ and CD4+ T cells, respectively.

IFNg production by unstimulated (no stim.) T cells

is also shown. ***p < 0.001.

(C) Quantitation of IFNg-producing tumor-infil-

trating CD8+ and CD4+ T cells as percentages of

total tumor-infiltrating CD8+ and CD4+ T cells,

respectively. IFNg production by unstimulated

T cells is also shown. ***p = 0.0003.

(D) Quantitation of IFNg/TNFa dual-producing

dLN-resident CD8+ and CD4+ T cells as percent-

ages of total dLN resident CD8+ and CD4+ T cells,

respectively. Dual cytokine production by un-

stimulated T cells is also shown. **p = 0.002, 0.003,

and 0.001, respectively.

(E) Quantitation of IFNg/TNFa dual-producing

tumor-infiltrating CD8+ and CD4+ T cells as per-

centages of total tumor-infiltrating CD8+ and CD4+

T cells, respectively. Dual cytokine production by

unstimulated T cells is also shown. ***p < 0.0001.

Error bars depict SEM. See also Figure S4.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 929

A

C

E

F

D

B

Figure 6. TIGIT Enforces CD8+ T Cell Exhaustion during Chronic Viral Infection

(A and B) C57BL6/J mice were infected with the Armstrong strain of LCMV, and splenocytes were analyzed 7 days after infection. Data are representative of two

independent experiments (n = 5).

(A) Flow cytometry histogram representative of TIGIT expression by naive (CD44lowCD62Lhigh) and effector (CD44highCD62Llow) CD4+ and CD8+ T cells.

Quantitation of TIGIT+ cells as a percentage of naive and effector memory (Teff) CD4+ and CD8+ T cells is also shown. ***p < 0.001.

(B) Flow cytometry histogram representative of TIGIT expression by PD-1high and PD-1low effector CD8+ T cells. Quantitation of TIGIT MFI is also shown.

***p < 0.001.

(C) C57BL6/J mice were briefly depleted of CD4+ T cells and infected with the Clone 13 strain of LCMV. Splenocytes were analyzed 42 days after infection.

Shown is a flow cytometry histogram representative of TIGIT expression by naive (CD44lowCD62Lhigh), central memory (CD44highCD62Lhigh), and effector

memory (CD44highCD62Llow) CD8+ T cells. Quantitation of TIGIT MFI is also shown. ***p < 0.001. Data are representative of two independent experiments

(n = 5).

(D–F) C57BL6/J mice were briefly depleted of CD4+ T cells and infected with the Clone 13 strain of LCMV. Mice were treated with isotype-matched control, anti-

PD-L1 + control, anti-TIGIT + control, or anti-PD-L1 + anti-TIGIT antibodies starting 28 days after infection. Splenocytes and liver viral titers were analyzed 42 days

after infection. Data are representative of two independent experiments (n = 10).

(D) Quantitation of liver LCMV titers. *p = 0.0106, **p = 0.0047.

(legend continued on next page)

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

930 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

strong synergy between PD-1 and TIGIT on exhausted CD8+

T cells and indicate that TIGIT specifically regulates CD8+

T cell cytokine competency and effector function.

TIGIT Impairs CD226 Function by Directly Disrupting

CD226 Homodimerization

TIGIT contains an ITIM-like motif. However, unlike PD-1 or

CTLA-4, there is no direct biochemical evidence of a T cell inhib-

itory signaling cascade initiated by endogenous TIGIT (Joller

et al., 2011; Lozano et al., 2012; Yu et al., 2009). Some inhibitory

receptors also function to suppress T cell responses by limiting

complementary costimulatory receptor activation, as is the

case with CTLA-4-mediated suppression of CD28 signaling

(Pentcheva-Hoang et al., 2004). Consequently, we asked

whether TIGIT induced T cell exhaustion indirectly via suppres-

sion of its complementary costimulatory receptor, CD226, which

is highly expressed by peripheral and tumor-infiltrating CD8+

T cells (Figure S6). TIGIT-deficient murine CD8+ T cells were

stimulated with suboptimal levels of anti-CD3 with or without re-

combinant murine PVR-Fc fusion protein. In the absence of

TIGIT, T cells responded more robustly to PVR costimulation

than T cells from wild-type littermates, and this enhanced

response was dependent on CD226 (46% increase in prolifera-

tion, p = 0.0061; Figure 7A). Consistent with these data, the

proliferation of wild-type murine CD8+ T cells stimulated with

anti-CD3 and PVR was enhanced significantly by addition of

anti-TIGIT blocking antibodies relative to isotype-matched con-

trol antibodies (p = 0.0010; Figure 7B). This effect was again

dependent on CD226 (Figure 7B). To test the relevance of TIGIT

to primary human CD8+ T cells, we purified CD8+ T cells from

healthy donor blood and stimulated them with suboptimal levels

of anti-CD3 and recombinant human PVR-Fc fusion proteins. In

the presence of isotype-matched control antibodies, PVR costi-

mulation moderately enhanced T cell stimulation and prolifera-

tion. Again, addition of anti-TIGIT blocking antibodies signifi-

cantly enhanced the effects of PVR costimulation (69%

increase in proliferation, p = 0.0071; Figure 7C). These data

demonstrated a cell-intrinsic role for TIGIT inhibition of CD226

function on primary murine and human CD8+ T cells.

To determine the molecular mechanism by which TIGIT

impaired CD226 activity, we utilized time-resolved fluorescence

resonance energy transfer (TR-FRET) (Bazin et al., 2002; Keppler

et al., 2004; Maurel et al., 2008). First, we expressed and labeled

human ST-CD226 with nonpermeant donor and acceptor fluoro-

phores. These cells yielded a strong FRET signal, confirming the

ability of CD226 to homodimerize (Figure 7D). To monitor CD226

and TIGIT interactions on the cell surface, we expressed ST-

CD226 in the absence or presence of human hemagglutinin

(HA)-TIGIT and labeled the two molecules with the SNAP tag

substrate and anti-HA antibodies, respectively. Strikingly,

expression of TIGIT (monitored by ELISA; Figure S6) attenuated

the CD226/CD226 FRET signal in a dose-dependent manner,

indicating that TIGIT could disrupt CD226 homodimerization

(Figure 7E). Indeed, the use of acceptor CD226 and donor TIGIT

also resulted in a significant FRET signal, revealing a direct inter-

action between the two proteins (Figure 7F). This interaction was

further confirmed by coimmunoprecipitation (Figure 7G). To test

the effects of TIGIT antibody blockade on TIGIT-CD226 interac-

tion, we coexpressed human ST-CD226 and HA-TIGIT in the

presence or absence of blocking antibodies against human

TIGIT. The addition of anti-TIGIT to the cell cultures significantly

reduced the ability of TIGIT and CD226 to associate (Figure 7H),

consistent with ability of anti-TIGIT to enhance CD226 costimu-

lation (Figure 6). Next we confirmed the capacity of endogenous

TIGIT and CD226 to interact. Primary human T cells were stimu-

lated in vitro with anti-CD3 and anti-CD28, sorted on the basis of

TIGIT expression, rested, restimulated, and labeled with

antibodies against endogenous TIGIT and CD226 that were con-

jugated to fluorophores compatible with TR-FRET. TIGIT-ex-

pressing T cells labeled with donor-conjugated anti-TIGIT and

acceptor-conjugated anti-CD226 antibodies yielded a strong

FRET signal. In contrast, only a negligible FRET signal was de-

tected on T cells that did not express TIGIT or that were labeled

with donor-conjugated anti-TIGIT and acceptor-conjugated

anti-herpes virus entry mediator (HVEM) antibodies, confirming

the specificity of the detected interaction between endogenous

TIGIT and CD226.

These data demonstrated that TIGIT and CD226 could directly

interact at the cell surface and that this interaction impaired

CD226 homodimerization. Given the role of CD226 as a costimu-

lator of T cell responses in vivo, we hypothesized that suppres-

sion of CD226may be a keymechanism of action by which TIGIT

enforces CD8+ T cell exhaustion during chronic viral infection

and cancer.

TIGIT Suppression of CD8+ T Cell Responses

Is Dependent on CD226

To assess the requirement of CD226 for the anti-TIGIT mediated

rescue of effector T cell responses, we treated CT26 tumor-

bearing mice as before with the addition of either blocking

anti-CD226 antibodies or control antibodies. Treatment with

anti-CD226 alone slightly accelerated tumor growth relative to

control mice, resulting in a decreased median survival of

2 days (anti-CD226 alone versus control, p = 0.0118; Figures

8A and 8B). Strikingly, the addition of anti-CD226 blocking anti-

bodies to mice treated with anti-TIGIT and anti-PD-L1 coblock-

ade reversed the effects of TIGIT/PD-L1 coblockade on tumor

regression and survival (Figures 8A and 8B). A similar effect

was observed on LCMV titers in chronically infectedmice treated

with anti-TIGIT, anti-PD-L1, and/or anti-CD226 (Figure S7),

suggesting that TIGIT limits CD226 activity during both antitumor

and antiviral T cell responses.

Next we assessed T cells from CT26 tumors and tumor-drain-

ing lymph nodes inmice treated as above for 7 days. Blockade of

CD226 alone had no effect on IFNg production by tumor-infil-

trating and tumor-draining lymph node-resident CD8+ T cells,

(E) Representative FACS plots gated on CD8+ T cells, with activated cells (CD44highCD62Llow) boxed. Shown is the quantitation of activated cells as a percentage

of total CD8+ T cells. ***p < 0.0001.

(F) Representative FACS plots gated on activated CD8+ T cells after stimulation in vitro, with IFNg+ cells boxed. Quantitation of IFNg-producing cells as a

percentage of activated CD8+ T cells.*p = 0.0352, **p = 0.0047.

Error bars depict SEM. See also Figure S5.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 931

A

D

G H I

E F

B C

Figure 7. TIGIT Impairs CD226 Function by Directly Disrupting CD226 Homodimerization

(A) CD8+ T cells were MACS-enriched from TIGITfl/fl CD4cre (CKO) and TIGITfl/fl CD4wt (WT) littermates and stimulated in the presence of anti-CD226 or isotype-

matched control antibodies as indicated. H3-thymidine uptake is shown as a ratio relative towild-type cells culturedwithout stimulation or treatment. **p = 0.0061,

***p < 0.0001. Data are representative of two independent experiments (n = 5).

(B) Wild-type C57BL6/J CD8+ T cells were MACS-enriched and stimulated in the presence of anti-TIGIT, anti-CD226, and/or isotype-matched control antibodies

as indicated. H3-thymidine uptake is shown as a ratio relative to cells cultured without stimulation or treatment. ***p < 0.001 in paired t tests.

(C) Primary human CD8+ T cells were MACS-enriched from blood and stimulated with suboptimal levels of plate-bound anti-CD3 in the presence or absence of

human recombinant PVR-Fc. Anti-TIGIT antibodies or isotype-matched control antibodies were added as indicated. Shown is the quantitation of 3H-thymidine

uptake. **p = 0.0071 and 0.0014, respectively.

(D) CHO cells were transiently transfected with increasing concentrations of acceptor (green ‘‘A’’) and donor (gray ‘‘D’’) FLAG-ST-CD226 as indicated. Shown is

the quantification of FRET intensity relative to donor emission. Data are representative of three independent experiments (n = 3).

(E and F) CHO cells were transiently transfected with human FLAG-ST-CD226 and with increasing concentrations of human HA-TIGIT as indicated. Data are

representative of two or more independent experiments (n = 4). Data are normalized to the maximal signal.

(E) Quantification of the CD226:CD226 FRET ratio.

(F) Quantification of the TIGIT:CD226 FRET ratio.

(legend continued on next page)

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

932 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

indicating that the contribution of CD226 costimulation in already

exhausted T cells is limited (Figures 8C-8F). As before, coblock-

ade of PD-L1 and TIGIT enhanced IFNg production by both tu-

mor-infiltrating and tumor-draining lymph node-resident CD8+

T cells (130% and 99% increase, respectively, p < 0.001; Figures

8C and 8D). CD226 blockade impaired the effector function and

frequency of CD8+ TILs in mice treated with anti-TIGIT and anti-

PD-L1 (57% decrease, p = 0.0015; Figure 8C) but had no effect

on tumor-draining lymph node-resident CD8+ T cells in these

mice (Figures 8D and 8F). Because anti-PD-L1 alone was suffi-

cient to enhance CD8+ T cell effector function in tumor-draining

lymph nodes but both anti-TIGIT and anti-PD-L1 were needed to

restore CD8+ T cell effector function in tumors, these data sug-

gested that anti-CD226 selectively abrogated the effects of

anti-TIGIT and not those of anti-PD-L1.

Taken together, these results reveal critical and antagonistic

roles for TIGIT and CD226 in regulating the effector function of

chronically stimulated CD8+ T cells in vivo.

DISCUSSION

The induction of an effective anti-tumor T cell response requires

passage through several checkpoints: antigen-specific priming,

effector cell differentiation, trafficking to the tumor bed, and,

finally, killing of the tumor cells by cytotoxic CD8+ T cells (Chen

and Mellman, 2013). However, the immunosuppressive nature

of tumor microenvironments often preempts this last step by

rendering the infiltrating T cells dysfunctional and unable to elab-

orate their full effector functions. Remarkably, this so-called

exhaustion can be rapidly reversed in at least some T cells

upon antagonism of key coinhibitory receptors. These receptors

are expressed by CD8+ and CD4+ T cells and by innate immune

cells, and their effects on each lineage can potently and coordi-

nately suppress immune responses. TIGIT has been described

as a modest inhibitor of CD4+ T cell priming and NK cell killing.

However, its importance to CD8+ T cells and to antitumor and

other chronic immune responses has been untested. Here we

show that TIGIT is highly expressed by both human and murine

tumor-infiltrating CD8+ T cells. TIGIT was expressed in parallel

with PD-1, a key T cell checkpoint inhibitor during chronic

immune responses. In CT26 tumor-bearing mice, antibody

blockade of both TIGIT and PD-L1 was required to elicit tumor

rejection and antigen-specific protection against tumor rechal-

lenge. Indeed, although blockade of either TIGIT or PD-L1 alone

was sufficient to enhance CD8+ T cell effector function within

tumor-draining lymph nodes, blockade of both receptors was

necessary to restore the potency of CD8+ T cells within highly

suppressive tumor microenvironments. Taken together, these

results identify TIGIT as a critical and specific regulator of the

effector function of chronically stimulated CD8+ T cells in addi-

tion to its roles in other immune cell lineages.

Similarly, T cells in the canonical LCMV model of chronic viral

infection and T cell exhaustion coordinately expressed TIGIT and

PD-1. Blockade of both receptors synergistically hastened viral

clearance and greatly enhanced CD8+ T cell effector function.

Genetic deletion of TIGIT was sufficient to enhance both CD8+

and CD4+ T cell effector function and to improve viral clearance.

However, mice with established chronic infections required

treatment with both anti-TIGIT and anti-PD-L1 to reveal a

specialized role for TIGIT in limiting the effector function of

chronically stimulated CD8+ T cells. These results distinguish

two functions for TIGIT: one to limit T cell priming, which impacts

both CD4+ and CD8+ T cell responses, and a second, more

potent and specialized role in the suppression of effector func-

tion in chronically stimulated CD8+ T cells. This second role

may be subordinate to PD-1, consistent with PD-1’s status as

a key limiter of acute and chronic T cell responses. Indeed, the

absence of TIGIT in and of itself was not sufficient to perturb

the T cell response to acute viral infection, in marked contrast

to the lethal consequences of PD-L1 ablation in similar settings

(Barber et al., 2006; Mueller et al., 2010).

Coinhibitory receptors can suppress immune responses by

direct signaling in cis, by inducing ligand signaling in trans, and

by competitionwith costimulatory receptors. The firstmechanism

bywhich TIGITwas found to inhibit T cell responseswas signaling

through ligation of its ligand PVRon dendritic cells, resulting in the

conversion of those dendritic cells to a tolerogenic phenotype

characterized by increased IL-10 anddecreased IL-12production

(Yuetal., 2009).Theabsenceof IL-10wassufficient to fully reverse

the effects of TIGIT signaling in a model of T cell hypersensitivity,

suggesting that thiswas the predominantmechanismof action by

which TIGIT suppressed acute CD4+ T cell responses (Yu et al.,

2009). AlthoughTIGIT signaling in cis has been described in trans-

fectedNKcell lines (Liu et al., 2013), only indirect evidenceofTIGIT

signaling in T cells has been reported (Joller et al., 2011; Lozano

et al., 2012; Yu et al., 2009). Further studies are needed to clarify

the capacity of TIGIT to signal in T cells and the effects of this

signaling on CD4+ and CD8+ T cell responses. TIGIT may also

act to directly compete with costimulatory receptors, particularly

CD226, for ligand binding (Lozano et al., 2012; Stengel et al.,

2012; Yuet al., 2009). CD226deficiency hasbeen shown to impair

antiviral and antitumor T cell effector function (Cella et al., 2010;

Ramsbottomet al., 2014;Welch et al., 2012); however, themolec-

ular relationship between TIGIT and CD226 has not been fully

elucidated. Here we showed that TIGIT directly interacts with

CD226 and that this interaction impaired CD226 homodimeriza-

tion and function. CD226 costimulation was enhanced in primary

human andmurine CD8+ T cells treated with an antibody that dis-

rupted TIGIT-CD226 interaction in vitro, and treatment of tumor-

bearingmicewith anti-CD226 antibodies was sufficient to abolish

the effects of anti-TIGIT in vivo. These data depict a relationship

between TIGIT and CD226 that echoes that of CTLA-4 and

(G) Anti-FLAG (left) and anti-HA (right) immunoblots (IB) performed on either anti-FLAG or anti-HA immunoprecipitates (IP) prepared fromCOS-7 cells transfected

with either an empty pRK vector or a combination of FLAG-CD226 and HA-TIGIT. Data are representative of two independent experiments.

(H) CHO cells were transfected as in (F), and the TIGIT:CD226 FRET ratio was quantified after treatment with PBS (white) or anti-TIGIT (red). ***p < 0.001. Data are

representative of four independent experiments (n = 3).

(I) Primary human T cells were MACS-enriched from blood and stimulated with anti-CD3 and anti-CD28. TIGIT+ and TIGIT� cells were sorted, rested,

restimulated, and labeled for FRET with the indicated antibodies. Data are representative of two independent experiments. Shown is the quantification of FRET

ratios. ***p < 0.001.

Error bars depict SEM. See also Figure S6.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 933

CD28 in that suppression ofCD226activity is a keymechanismby

which TIGIT inhibits CD8+ T cell responses.

Recently, antibody blockade of individual coinhibitory recep-

tors involved in tumor immunosuppression has proven to be

remarkably effective at reversing T cell exhaustion and eliciting

immune rejection of tumors in the clinic. Moreover, targeting a

combination of receptors holds the potential to be even more

effective (Wolchok et al., 2013), making it essential to fully eluci-

date the individual and synergistic functions of the coinhibitory

receptors that regulate antitumor responses. Our findings here

suggest that TIGIT is key checkpoint inhibitor of chronic antiviral

and antitumor responses and, consequently, may represent a

target for future immunotherapies.

EXPERIMENTAL PROCEDURES

Bioinformatics

To derive a T cell-specific gene signature, wemanually curated the T cell genes

identified by the IRIS project, removing genes associated with cell cycle pro-

cesses, genes highly expressed in other tissues, and known coactivating

and coinhibitory receptors. We then ranked the genes by their correlation

with the T cell signature in our linear model, choosing only genes positively

correlated with the T cell signature.

A

B

E F

C D

Figure 8. TIGIT Suppression of CD8+ T Cell Responses Is Dependent on CD226

(A–F) BALB/c mice were inoculated subcutaneously with CT26 colorectal carcinoma cells in their right thoracic flanks. When tumors reached approximately

200 mm3 in size, mice were treated with isotype control (black), anti-CD226 + control (orange), anti-PD-L1 + control (red), anti-TIGIT + anti-PD-L1 + control

(purple), or anti-TIGIT + anti-PD-L1 + anti-CD226 (green) antibodies for 3 weeks. Data are representative of two independent experiments (n = 10 [A and B] or

5 [C–F]).

(A) Median (left) and individual (right) CT26 tumor volumes over time.

(B) Mouse survival over time.

(C–F) After 7 days of treatment, tumor-infiltrating lymphocytes and tumor-draining lymph node-resident lymphocytes were assessed by flow cytometry.

(C) Quantitation of IFNg-producing CD8+ TILs as a percentage of total CD8+ TILs after stimulation in vitro. **p < 0.01.

(D) Quantitation of IFNg-producing cells as a percentage of tumor-draining lymph node CD8+ T cells after stimulation in vitro. *p < 0.05.

(E) Quantitation of CD8+ TILs as a percentage of total TILs. **p < 0.01.

(F) Quantitation of CD8+ T cells as a percentage of all tumor-draining lymph node-resident lymphocytes.

Error bars depict SEM. See also Figure S7.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

934 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

Human Tumor and PBMC Samples

Matched whole blood and fresh surgically resected tumor tissues were ob-

tained from Conversant Biosciences or Foundation Bio. All specimens were

obtained with written informed consent and collected using a protocol

approved by the Hartford Hospital Institutional Review Board (IRB) (NSCLC

patient 1, depicted in Figure 2) or the Western IRB (NSCLC patient 2 and

CRC patient 1, depicted in Figure S2).

Syngeneic Tumor Studies

BALB/cmicewere inoculatedsubcutaneously in the right thoracicflankwith13

105 syngeneic CT26 colon carcinoma cells in Matrigel (BD Biosciences) or in

the fourth mammary fat pad with 1 3 105 syngeneic EMT6 breast carcinoma

cells in Matrigel. After 2 weeks, mice bearing tumors of 150–200 mm3 were

randomized into treatment groups and treated with anti-PD-L1 (10 mg/kg),

anti-TIGIT (25 mg/kg), or isotype control antibodies (to total 35 mg/kg overall

antibody dosing) by intraperitoneal injection three times per week for 3 weeks.

Where indicated, mice received an additional 10 mg/kg of anti-CD226 anti-

bodies or an equivalent amount of control antibodies. For depletion of CD8+

T cells, mice were injected with 250 mg of CD8-depleting antibodies (clone

53.6.7) 1 day before and 3days after tumor inoculation. Tumorsweremeasured

two times per week by caliper, and tumor volumes were calculated using the

modified ellipsoid formula 1/2 3 (length 3 width2). Animals whose tumors

shrank to 32 mm3 or smaller were considered to be in CR. Animals whose

tumors grew to larger than 2000 mm3 were considered to have progressed

and were euthanized. Animals whose tumors became ulcerated prior to pro-

gression or complete response were euthanized and removed from the study.

Tumor Rechallenge Studies

Naive BALB/c mice and mice inoculated previously with CT26 colon carci-

noma cells and treated as described above were inoculated with CT26 cells

into the left (not inoculated previously) unilateral thoracic flank. Where indi-

cated, some mice were also inoculated with 1 3 105 EMT6 breast carcinoma

cells in Matrigel into the fourth mammary fat pad. Where indicated, some mice

were also depleted of CD8+ T cells by injection of 250 mg of CD8-depleting

antibodies 1 day before and 3 days after tumor reinoculation. Tumors were

measured two times per week as described above. Animals whose tumors

became ulcerated/necrotic or whose combined tumor burden exceeded

2,000 mm3 were euthanized.

Viral Infection Studies

For acute infections, C57BL6/J mice were infected intravenously with 23 106

plaque-forming units (pfu) of the Armstrong strain of LCMV. For chronic infec-

tions, C57BL6/J mice, TIGITfl/fl mice, and TIGITfl/fl;CD4cre mice were infected

intravenously with 2 3 106 pfu of the Clone 13 strain of LCMV and treated

with intraperitoneal injections of 500 mg and 250 mg of depleting anti-CD4

antibodies (clone GK1.5) 3 days before and 4 days after infection, respectively.

Where indicated, mice infected with the Clone 13 strain of LCMV were

randomly recruited into experimental groups and treated with intraperitoneal

injections of 200 mg of isotype control antibodies, 200 mg of anti-PD-L1 anti-

bodies, and/or 500 mg of anti-TIGIT antibodies three times per week from

days 28–42 postinfection.

TR-FRET with Transfected Cell Lines

Chinese hamster ovary (CHO) cells were transfected with N terminus SNAP-

tagged (ST) CD226 and N terminus HA-TIGIT and then labeled to measure

TR-FRET either between the SNAP donor and the SNAP acceptor or between

the SNAP acceptor and the anti-HA donor.

TR-FRET with Human T Cells

Human anti-TIGIT (Genentech clone 1F4), anti-CD226 (Santa Cruz Biotech-

nology), and anti-HVEM (eBioscience) were conjugated with fluorophores

compatible with TR-FRET (Cisbio). Primary human T cells were magnetic acti-

vated cell sorting (MACS)-enriched from blood and stimulated in vitro with

plate-bound anti-CD3 and anti-CD28 for 72 hr. TIGIT-expressing and nonex-

pressing T cells (all expressing CD226) were then sorted, rested without stim-

ulation for 72 hr, and restimulated for 48 hr before incubation with FRET

antibodies.

Animal Study Oversight

All animal studieswere approved byGenentech’s Institutional Animal Care and

Use Committee. The group sizes used for in vivo studies were those estimated

to be the smallest necessary to generate meaningful data. Studies were not

conducted in a blinded fashion. Mice were monitored regularly, and those

requiring medical attention were provided with appropriate care and excluded

from the studies.

Additional details are provided in the Supplemental Experimental

Procedures.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

seven figures, and one table and can be found with this article online at

http://dx.doi.org/10.1016/j.ccell.2014.10.018.

AUTHOR CONTRIBUTIONS

R.J.J. developed the hypothesis, designed and performed experiments,

analyzed the data, and wrote the manuscript. L.C.A. and D.L.E. designed

and performed protein interaction and FRET experiments and edited the

manuscript. J.A.H. designed and conducted bioinformatics analyses and edi-

ted the manuscript. X.Y. contributed to reagent generation, experimental

design, and in vitro assays. M.H. conducted the flow cytometry analysis of hu-

man TILs. Y.Y. contributed to experimental design and tumor experiments.

S.P. contributed to experimental design and LCMV experiments. V.J. contrib-

uted to experimental design and tumor experiments. H.C. contributed to re-

agent generation. B.I. contributed to experimental design. J.L.G. conceived

and supervised the project, analyzed the data, and wrote the manuscript.

ACKNOWLEDGMENTS

We thank Kristin Harden-Bowles, Jill Calemine-Fenaux, Joshua Tanguay,

Merone Roose-Girma, Jeanne Cheung, and Janice Kim for expert technical

assistance. We thank Ira Mellman, Min Xu, Klara Totpal, Eugene Chiang, Jill

Schartner, and Stephen Gould for valuable discussions. We thank Allison

Bruce for the design and construction of our graphical abstract. All authors

are employees of Genentech, a corporation that develops and markets drugs

for profit.

Received: January 28, 2014

Revised: July 17, 2014

Accepted: October 28, 2014

Published: November 26, 2014

REFERENCES

Abbas, A.R., Baldwin, D., Ma, Y., Ouyang, W., Gurney, A., Martin, F., Fong, S.,

van Lookeren Campagne, M., Godowski, P., Williams, P.M., et al. (2005).

Immune response in silico (IRIS): immune-specific genes identified from a

compendium of microarray expression data. Genes Immun. 6, 319–331.

Ahmed, R., Salmi, A., Butler, L.D., Chiller, J.M., and Oldstone, M.B. (1984).

Selection of genetic variants of lymphocytic choriomeningitis virus in spleens

of persistently infected mice. Role in suppression of cytotoxic T lymphocyte

response and viral persistence. J. Exp. Med. 160, 521–540.

Barber, D.L., Wherry, E.J., Masopust, D., Zhu, B., Allison, J.P., Sharpe, A.H.,

Freeman, G.J., and Ahmed, R. (2006). Restoring function in exhausted CD8

T cells during chronic viral infection. Nature 439, 682–687.

Bazin, H., Trinquet, E., and Mathis, G. (2002). Time resolved amplification of

cryptate emission: a versatile technology to trace biomolecular interactions.

J. Biotechnol. 82, 233–250.

Blackburn, S.D., Shin, H., Haining, W.N., Zou, T., Workman, C.J., Polley, A.,

Betts, M.R., Freeman, G.J., Vignali, D.A., and Wherry, E.J. (2009).

Coregulation of CD8+ T cell exhaustion by multiple inhibitory receptors during

chronic viral infection. Nat. Immunol. 10, 29–37.

Bulliard, Y., Jolicoeur, R., Windman, M., Rue, S.M., Ettenberg, S., Knee, D.A.,

Wilson, N.S., Dranoff, G., and Brogdon, J.L. (2013). Activating Fc g receptors

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 935

contribute to the antitumor activities of immunoregulatory receptor-targeting

antibodies. J. Exp. Med. 210, 1685–1693.

Cella, M., Presti, R., Vermi, W., Lavender, K., Turnbull, E., Ochsenbauer-

Jambor, C., Kappes, J.C., Ferrari, G., Kessels, L., Williams, I., et al.; CHAVI

Clinical Core B; NIAID Center for HIV/AIDS Vaccine Immunology (2010).

Loss of DNAM-1 contributes to CD8+ T-cell exhaustion in chronic HIV-1 infec-

tion. Eur. J. Immunol. 40, 949–954.

Chen, L., and Flies, D.B. (2013). Molecular mechanisms of T cell co-stimulation

and co-inhibition. Nat. Rev. Immunol. 13, 227–242.

Chen, D.S., and Mellman, I. (2013). Oncology meets immunology: the cancer-

immunity cycle. Immunity 39, 1–10.

Crespo, J., Sun, H., Welling, T.H., Tian, Z., and Zou, W. (2013). T cell anergy,

exhaustion, senescence, and stemness in the tumor microenvironment.

Curr. Opin. Immunol. 25, 214–221.

Curran, M.A., Montalvo, W., Yagita, H., and Allison, J.P. (2010). PD-1 and

CTLA-4 combination blockade expands infiltrating T cells and reduces regula-

tory T and myeloid cells within B16 melanoma tumors. Proc. Natl. Acad. Sci.

USA 107, 4275–4280.

Doering, T.A., Crawford, A., Angelosanto, J.M., Paley, M.A., Ziegler, C.G., and

Wherry, E.J. (2012). Network analysis reveals centrally connected genes and

pathways involved in CD8+ T cell exhaustion versus memory. Immunity 37,

1130–1144.

Gallimore, A., Glithero, A., Godkin, A., Tissot, A.C., Pluckthun, A., Elliott, T.,

Hengartner, H., and Zinkernagel, R. (1998). Induction and exhaustion of lym-

phocytic choriomeningitis virus-specific cytotoxic T lymphocytes visualized

using soluble tetrameric major histocompatibility complex class I-peptide

complexes. J. Exp. Med. 187, 1383–1393.

Goding, S.R., Wilson, K.A., Xie, Y., Harris, K.M., Baxi, A., Akpinarli, A., Fulton,

A., Tamada, K., Strome, S.E., and Antony, P.A. (2013). Restoring immune func-

tion of tumor-specific CD4+ T cells during recurrence of melanoma.

J. Immunol. 190, 4899–4909.

Joller, N., Hafler, J.P., Brynedal, B., Kassam, N., Spoerl, S., Levin, S.D.,

Sharpe, A.H., and Kuchroo, V.K. (2011). Cutting edge: TIGIT has T cell-intrinsic

inhibitory functions. J. Immunol. 186, 1338–1342.

Kalos, M., and June, C.H. (2013). Adoptive T cell transfer for cancer immuno-

therapy in the era of synthetic biology. Immunity 39, 49–60.

Kandoth, C., Schultz, N., Cherniack, A.D., Akbani, R., Liu, Y., Shen, H.,

Robertson, A.G., Pashtan, I., Shen, R., Benz, C.C., et al.; Cancer Genome

Atlas Research Network (2013). Integrated genomic characterization of endo-

metrial carcinoma. Nature 497, 67–73.

Keppler, A., Pick, H., Arrivoli, C., Vogel, H., and Johnsson, K. (2004). Labeling

of fusion proteins with synthetic fluorophores in live cells. Proc. Natl. Acad. Sci.

USA 101, 9955–9959.

Krummel, M.F., and Allison, J.P. (1995). CD28 and CTLA-4 have opposing ef-

fects on the response of T cells to stimulation. J. Exp. Med. 182, 459–465.

Levin, S.D., Taft, D.W., Brandt, C.S., Bucher, C., Howard, E.D., Chadwick,

E.M., Johnston, J., Hammond, A., Bontadelli, K., Ardourel, D., et al. (2011).

Vstm3 is a member of the CD28 family and an important modulator of T-cell

function. Eur. J. Immunol. 41, 902–915.

Liu, S., Zhang, H., Li, M., Hu, D., Li, C., Ge, B., Jin, B., and Fan, Z. (2013).

Recruitment of Grb2 and SHIP1 by the ITT-like motif of TIGIT suppresses

granule polarization and cytotoxicity of NK cells. Cell Death Differ. 20,

456–464.

Lozano, E., Dominguez-Villar, M., Kuchroo, V., and Hafler, D.A. (2012). The

TIGIT/CD226 axis regulates human T cell function. J. Immunol. 188, 3869–

3875.

Maurel, D., Comps-Agrar, L., Brock, C., Rives, M.L., Bourrier, E., Ayoub, M.A.,

Bazin, H., Tinel, N., Durroux, T., Prezeau, L., et al. (2008). Cell-surface protein-

protein interaction analysis with time-resolved FRET and snap-tag technolo-

gies: application to GPCR oligomerization. Nat. Methods 5, 561–567.

Mellman, I., Coukos, G., and Dranoff, G. (2011). Cancer immunotherapy

comes of age. Nature 480, 480–489.

Mueller, S.N., Vanguri, V.K., Ha, S.J., West, E.E., Keir, M.E., Glickman, J.N.,

Sharpe, A.H., and Ahmed, R. (2010). PD-L1 has distinct functions in hemato-

poietic and nonhematopoietic cells in regulating T cell responses during

chronic infection in mice. J. Clin. Invest. 120, 2508–2515.

Nakamoto, N., Cho, H., Shaked, A., Olthoff, K., Valiga, M.E., Kaminski, M.,

Gostick, E., Price, D.A., Freeman, G.J., Wherry, E.J., and Chang, K.M.

(2009). Synergistic reversal of intrahepatic HCV-specific CD8 T cell exhaustion

by combined PD-1/CTLA-4 blockade. PLoS Pathog. 5, e1000313.

Network, C.G.A.; Cancer GenomeAtlas Network (2012a). Comprehensive mo-

lecular portraits of human breast tumours. Nature 490, 61–70.

Network, T.C.G.A.; Cancer Genome Atlas Network (2012c). Comprehensive

molecular characterization of human colon and rectal cancer. Nature 487,

330–337.

Network, T.C.G.A.; Cancer Genome Atlas Research Network (2012b).

Comprehensive genomic characterization of squamous cell lung cancers.

Nature 489, 519–525.

Network, C.G.A.; Cancer Genome Atlas Research Network (2013).

Comprehensive molecular characterization of clear cell renal cell carcinoma.

Nature 499, 43–49.

Nishimura, H., Nose, M., Hiai, H., Minato, N., and Honjo, T. (1999).

Development of lupus-like autoimmune diseases by disruption of the PD-1

gene encoding an ITIM motif-carrying immunoreceptor. Immunity 11,

141–151.

Palucka, K., and Banchereau, J. (2013). Dendritic-cell-based therapeutic can-

cer vaccines. Immunity 39, 38–48.

Pentcheva-Hoang, T., Egen, J.G., Wojnoonski, K., and Allison, J.P. (2004).

B7-1 and B7-2 selectively recruit CTLA-4 and CD28 to the immunological syn-

apse. Immunity 21, 401–413.

Ramsbottom, K.M., Hawkins, E.D., Shimoni, R., McGrath, M., Chan, C.J.,

Russell, S.M., Smyth, M.J., and Oliaro, J. (2014). Cutting edge: DNAX acces-

sorymolecule 1-deficient CD8+ T cells display immunological synapse defects

that impair antitumor immunity. J. Immunol. 192, 553–557.

Schietinger, A., and Greenberg, P.D. (2014). Tolerance and exhaustion:

defining mechanisms of T cell dysfunction. Trends Immunol. 35, 51–60.

Simpson, T.R., Li, F.,Montalvo-Ortiz,W., Sepulveda,M.A.,Bergerhoff, K., Arce,

F.,Roddie,C.,Henry, J.Y.,Yagita,H.,Wolchok, J.D.,et al. (2013). Fc-dependent

depletion of tumor-infiltrating regulatory T cells co-defines the efficacy of anti-

CTLA-4 therapy against melanoma. J. Exp. Med. 210, 1695–1710.

Stanietsky, N., Simic, H., Arapovic, J., Toporik, A., Levy, O., Novik, A., Levine,

Z., Beiman, M., Dassa, L., Achdout, H., et al. (2009). The interaction of TIGIT

with PVR and PVRL2 inhibits human NK cell cytotoxicity. Proc. Natl. Acad.

Sci. USA 106, 17858–17863.

Stanietsky, N., Rovis, T.L., Glasner, A., Seidel, E., Tsukerman, P., Yamin, R.,

Enk, J., Jonjic, S., and Mandelboim, O. (2013). Mouse TIGIT inhibits NK-cell

cytotoxicity upon interaction with PVR. Eur. J. Immunol. 43, 2138–2150.

Stengel, K.F., Harden-Bowles, K., Yu, X., Rouge, L., Yin, J., Comps-Agrar, L.,

Wiesmann, C., Bazan, J.F., Eaton, D.L., and Grogan, J.L. (2012). Structure of

TIGIT immunoreceptor bound to poliovirus receptor reveals a cell-cell adhe-

sion and signaling mechanism that requires cis-trans receptor clustering.

Proc. Natl. Acad. Sci. USA 109, 5399–5404.

Sullivan, B.M., Emonet, S.F., Welch, M.J., Lee, A.M., Campbell, K.P., de la

Torre, J.C., and Oldstone, M.B. (2011). Point mutation in the glycoprotein of

lymphocytic choriomeningitis virus is necessary for receptor binding, dendritic

cell infection, and long-term persistence. Proc. Natl. Acad. Sci. USA 108,

2969–2974.

Tivol, E.A., Borriello, F., Schweitzer, A.N., Lynch, W.P., Bluestone, J.A., and

Sharpe, A.H. (1995). Loss of CTLA-4 leads to massive lymphoproliferation

and fatal multiorgan tissue destruction, revealing a critical negative regulatory

role of CTLA-4. Immunity 3, 541–547.

van den Boorn, J.G., and Hartmann, G. (2013). Turning tumors into vaccines:

co-opting the innate immune system. Immunity 39, 27–37.

Welch, M.J., Teijaro, J.R., Lewicki, H.A., Colonna, M., and Oldstone, M.B.

(2012). CD8 T cell defect of TNF-a and IL-2 in DNAM-1 deficient mice delays

clearance in vivo of a persistent virus infection. Virology 429, 163–170.

Wherry, E.J. (2011). T cell exhaustion. Nat. Immunol. 12, 492–499.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

936 Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc.

Wherry, E.J., Ha, S.J., Kaech, S.M., Haining, W.N., Sarkar, S., Kalia, V.,

Subramaniam, S., Blattman, J.N., Barber, D.L., and Ahmed, R. (2007).

Molecular signature of CD8+ T cell exhaustion during chronic viral infection.

Immunity 27, 670–684.

Wolchok, J.D., Kluger, H., Callahan, M.K., Postow, M.A., Rizvi, N.A., Lesokhin,

A.M., Segal, N.H., Ariyan, C.E., Gordon, R.A., Reed, K., et al. (2013).

Nivolumab plus ipilimumab in advanced melanoma. N. Engl. J. Med. 369,

122–133.

Yu, X., Harden, K., Gonzalez, L.C., Francesco, M., Chiang, E., Irving, B., Tom,

I., Ivelja, S., Refino, C.J., Clark, H., et al. (2009). The surface protein TIGIT sup-

presses T cell activation by promoting the generation of mature immunoregu-

latory dendritic cells. Nat. Immunol. 10, 48–57.

Zajac, A.J., Blattman, J.N., Murali-Krishna, K., Sourdive, D.J., Suresh, M.,

Altman, J.D., and Ahmed, R. (1998). Viral immune evasion due to persis-

tence of activated T cells without effector function. J. Exp. Med. 188,

2205–2213.

Cancer Cell

TIGIT Limits Antitumor CD8+ T Cell Responses

Cancer Cell 26, 923–937, December 8, 2014 ª2014 Elsevier Inc. 937

Cancer Cell, Volume 26

Supplemental Information

The Immunoreceptor TIGIT Regulates Antitumor

and Antiviral CD8+ T Cell Effector Function

Robert J. Johnston, Laetitia Comps-Agrar, Jason Hackney, Xin Yu, Mahrukh Huseni, Yagai Yang, Summer Park, Vincent Javinal, Henry Chiu, Bryan Irving, Dan L. Eaton, and Jane L. Grogan

Gene Gene Name

PIK3IP1 phosphoinositide-3-kinase interacting protein 1

CD3E CD3ε molecule, epsilon (CD3-TCR complex)CD8B CD8β molecule

CD3D CD3δ molecule, delta (CD3-TCR complex)

CD3G CD3γ molecule, gamma (CD3-TCR complex)

CD5 CD5 molecule

CD8A CD8a molecule

GZMK granzyme K (granzyme 3; tryptase II)

DGKA diacylglycerol kinase, alpha 80kDa

TCF7 transcription factor 7 (T-cell specific, HMG-box)

LEF1 lymphoid enhancer-binding factor 1

TRAT1 T cell receptor associated transmembrane adaptor 1

DUSP2 dual specificity phosphatase 2

LDLRAP1 low density lipoprotein receptor adaptor protein 1

CAMK4 calcium/calmodulin-dependent protein kinase IV

A

LEF1

DGKA

TRAT1

CAMK4

CD3E

CD3G

CD3D

GZMK

CD5

DUSP2

TCF7

PIK3IP1

LDLRAP1

CD8B

CD8A

-4

-2

0

2

4

B cell

T cell

NK cell

Monocyte

Dendritic cell

Neutrophil

Supplemental Data

B

C

CD5

TRAT1

GZMK

CD3D

CD3E

CD3G

CD8A

CD8B

CAMK4

TCF7

DUSP2

PIK3IP1

LDLRAP1

DGKA

LEF1

PTPRCAPITGALCD48CCL5SCML4CTLA4

PCED1B-AS1GZMHIL10RATESPA1

LOC100652927CCR5CD6IL2RGGVINP1PDCD1CRTAMSLAMF1GIMAP4GIMAP7SASH3

ARHGAP15PYHIN1SIT1

GIMAP1GIMAP5PTPN22PTPRCCST7

LINC00426NKG7

TBC1D10CLCK

IL12RB1TIGITCD96

SLAMF6GPR174SH2D1A

LOC100506776GPR171UBASH3ACXCR3ITK

P2RY10CD247CXCR6SIRPGSLA2CD2

Signature

Figure S1. Related to Figure 1. T cell gene signature and analysis of T cell-associated gene

expression in human lung squamous cell carcinoma (LUSC). (A) The T cell-specific genes

constituting the signature used to analyze gene expression in breast cancer (left) and normalized

expression data for the T cell signature genes in IRIS cell subsets (right). (B-C) Gene expression

in LUSC and normal lung tissue samples was analyzed as described in the Methods. (B)

Normalized expression data for the T cell signature genes in normal lung tissue (black) and

LUSC (red) samples. Data were centered and scaled to have unit variance. (C) Heat map of the

genes best correlated with the T cell gene signature in LUSC samples (genes are also listed in

Supplemental Table 1). Genes and samples were both clustered using hierarchical clustering

using Ward linkage on the Euclidean distance matrix for the centered and scaled expression data.

Table S1. Related to Figure 1. Genes associated with T cell infiltration in human LUSC. A

list of the 200 genes best correlated with the T cell gene signature in LUSC samples (top genes

are also depicted graphically in Figure S1). Provided as a Microsft Excel file (xls).

53% 47%

CD8

TIG

IT

CD4

TIG

IT

BCRC CD8+ TIL

C

CD8

TIG

IT

CD4

TIG

IT

CRC CD8+ PBMC

27% 18%

TIGIT

A

TIGIT TIGIT

CD45+

Lymphocytes

CD45+

Myeloid Cells CD45- Cells

D

CD8

PD

-1

CD4

PD

-1

NSCLC CD8+ TIL NSCLC CD4+ TILE

TIGIT

PD

-1

CRC CD4+ TIL CRC CD4+ PBMC

NSCLC CD8+ TIL NSCLC CD4+ TIL

TIGIT

PD

-1

I

TIGIT

Resting 24 hr

TIGIT

48 hr

TIGIT

isot

ype

no stim

.

24 h

r

48 h

r0

50

100

150

200

250

300 G H

CD8

TIG

IT76%

CD4

TIG

IT

44%

TIGIT

CD8+

Spleen

TIL PD-1low

TIL PD-1high

TIGIT

CD4+

Spleen

TIL PD-1low

TIL PD-1high

Splee

n

Tumor

0

20

40

60

80

% T

IGIT

+ o

f C

D8

+ T

cells

***

Splee

n

Tumor

0

10

20

30

40

50

% T

IGIT

+ o

f C

D4

+ T

cells

***

Splee

n

TIL

PD-1

low

TIL P

D-1

high

0

500

1000

1500

2000

TIG

IT M

FI

***

**0

1000

2000

3000

4000***

***

TIG

IT M

FI

J

CD8

PD

-1

CD4

PD

-1

CT26 CD8+ TIL CT26 CD4+ TIL

CD45- CD4+ CD8+

TIGIT TIGIT TIGIT

K

TIGIT

PD

-1

CD8+ TIL CD4+ TIL

F

L M

******

TIG

IT M

FI

Splee

n

TIL

PD-1

low

TIL P

D-1

high

Figure S2. Related to Figure 2. Further characterization of TIGIT expression in human

and murine tumors. (A) Representative flow cytometry histograms of TIGIT expression by

human NSCLC tumor-resident lymphocytes (red, CD45+ FSC

low), myeloid cells (blue, CD45

+

FSChigh

), and non-hematopoietic cells (green, CD45-) relative to subset-matched isotype staining

(gray). (B-C) FACS plots of TIGIT expression by human colorectal carcinoma (CRC) tumor-

infiltrating CD8+ and CD4

+ T cells (B) and by matched PBMC CD8

+ and CD4

+ T cells (C), with

TIGIT+ cells boxed. (D) Gating strategy for human PD-1

high and PD-1

low NSCLC tumor-

infiltrating CD8+ and CD4

+ T cells. (E) Representative FACS plots of TIGIT and PD-1 co-

expression by human NSCLC tumor-infiltrating CD8+ and CD4

+ T cells. (F) Murine splenic

C57BL6/J CD8+ T cells were enriched by MACS and cultured with plate-coated anti-CD3 and

anti-CD28 agonist antibodies. Representative histograms of TIGIT (red) and isotype-matched

control (solid gray) staining over time. Quantitation of TIGIT MFI. ***, p < 0.001. Stimulated

cells inducibly expressed PD-1 and constitutively expressed CD226 (data not shown). Data are

representative of two independent experiments; n = 5. (G-J) Wildtype BALB/c mice were

subcutaneously inoculated with syngeneic EMT6 breast carcinoma cells. Tumors were allowed

to grow without intervention until they reached 150-200 mm3 in size. Data are representative of

two independent experiments; n = 5. (G) Representative FACS plot of tumor-infiltrating CD8+ T

cells, with TIGIT+ cells boxed. Quantitation of the frequency of TIGIT

+ cells as a percentage of

all tumor-infiltrating or splenic CD8+ T cells. ***, p < 0.0001. (H) Representative FACS plot of

tumor-infiltrating CD4+ T cells, with TIGIT

+ cells boxed. Quantitation of the frequency of

TIGIT+ cells as a percentage of all tumor-infiltrating or splenic CD4

+ T cells. ***, p < 0.0001.

(I) Representative histogram of TIGIT expression by PD-1high

and PD-1low

tumor-infiltrating

CD8+ T cells (red and blue, respectively) and by splenic CD8

+ T cells (gray). Quantitation of

TIGIT MFI. ***, p < 0.0001. **, p = 0.0046. (J) Representative histogram of TIGIT expression

by PD-1high

and PD-1low

tumor-infiltrating CD4+ T cells and by splenic CD4

+ T cells.

Quantitation of TIGIT MFI. ***, p < 0.0001. (K-M) Wildtype BALB/c mice were inoculated

with CT26 tumors and analyzed as described in Figure 2. (K) Gating strategy for PD-1high

and

PD-1low

CT26 tumor-infiltrating CD8+ and CD4

+ T cells. (L) Representative FACS plots of

TIGIT and PD-1 co-expression by CT26 tumor-infiltrating CD8+ and CD4

+ T cells. (M)

Representative histograms of TIGIT expression comparing CT26 tumor and non-hematopoietic

cells (left) to CD4+ and CD8

+ T cells (center and right, respectively). Isotype-matched controls

are shown in gray. Error bars depict the standard error of the mean.

Complete Response (CR)

0 10 20 30 40 5010

100

1000

10000M

ed

ian

Tu

mo

r V

olu

me

(m

m3)

Day

Control

anti-TIGITanti-PD-L1

anti-TIGIT

+ anti-PD-L1

0 10 20 30 40 5010

100

1000

10000

0 10 20 30 40 5010

100

1000

10000

0 10 20 30 40 5010

100

1000

10000

0 10 20 30 40 5010

100

1000

10000

Figure S3. Related to Figure 3. Efficacy of TIGIT/PD-L1 co-blockade in mice inoculated

with EMT6 tumors. EMT6 tumor-bearing mice were generated as described in Figure

S2 and treated as described in Figure 3 with blocking antibodies against PD-L1 (red), TIGIT

(blue), TIGIT and PD-L1 (purple) or isotype-matched control antibodies (black) for three weeks.

n = 10 (control, anti-PD-L1 alone, anti-TIGIT alone) or 20 (anti-TIGIT + anti-PD-L1). Median

(left) and individual (right) EMT6 tumor volumes over time.

C

0

20

40

60

D

0

20

40

60

80

100

E F

0

15

30

45

60

0

20

40

60

A B

0

10

20

30

0

20

40

60

80

100

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT0

10

20

30

0

20

40

60

80

100

G H I J

CD3+ Isotype

CD3+FoxP3+

CD3+FoxP3-

TIGIT

0

15

30

45

60

0

15

30

45

60

Figure S4. Related to Figure 4. Effects of TIGIT/PD-L1 co-blockade on tumor-infiltrating

and tumor-draining lymph node-resident T cell frequency and phenotype. . (A-H) Naïve

BALB/c mice were inoculated with CT26 tumor cells and treated with anti-PD-L1 and/or anti-

TIGIT or isotype-matched control antibodies, as described in Figure 4. Tumor-infiltrating and

tumor-draining lymph node-resident T cells were analyzed by FACS after 7 days of treatment.

Data are representative of more than three independent experiments; n = 5/group. (A) Frequency

of tumor-infiltrating CD8+ T cells as a percentage of all tumor-infiltrating lymphocytes. **, p <

0.01. (B) Frequency of activated (CD44high

CD62Llow

) tumor-infiltrating CD8+ T cells as a

percentage of all tumor-infiltrating CD8+ T cells. *, p < 0.05. (C) Frequency of tumor-draining

lymph node-resident CD8+ T cells as a percentage of all tumor-draining lymph node-resident

lymphocytes. (D) Frequency of activated tumor-draining lymph node-resident CD8+ T cells as a

percentage of all tumor-draining lymph node-resident CD8+ T cells. *, p < 0.05. (E) Frequency

of tumor-infiltrating CD4+ T cells as a percentage of all tumor-infiltrating lymphocytes. *, p <

0.05. (F) Frequency of activated tumor-infiltrating CD4+ T cells as a percentage of all tumor-

infiltrating CD4+ T cells. (G) Frequency of tumor-draining lymph node-resident CD4

+ T cells as

a percentage of all tumor-draining lymph node-resident lymphocytes. (H) Frequency of activated

tumor-draining lymph node-resident CD4+ T cells as a percentage of all tumor-draining lymph

node-resident CD4+ T cells. (I-J) Naïve BALB/c mice were inoculated with CT26 tumor cells

and treated with anti-PD-L1 + anti-TIGIT or isotype-matched control antibodies, as described in

Figure 4. Tumor-infiltrating and tumor-draining lymph node-resident T cells were analyzed by

FACS after 7 days of treatment. Data are representative of two independent experiments; n = 3-

5/group. (I) Representative histogram of TIGIT expression by FoxP3+ Treg cells (red) and FoxP3

-

T cells (blue) within CT26 tumors. (J) Frequencies of FoxP3+ Treg cells as a percentage of all

tumor-infiltrating CD4+ T cells (left) or as a percentage of all tumor-draining lymph node-

resident CD4+ T cells (right). **, p = 0.01. Error bars depict the standard error of the mean.

% C

D8

+ T

ILs

** *

% A

ctiva

ted

CD

8+ T

ILs

% C

D8

+ T

ce

lls

in t

um

or-

dra

inin

g L

Ns

*

*

% A

ctiva

ted

CD

8+ T

ce

lls

in t

um

or-

dra

inin

g L

Ns

% C

D4

+ T

ILs

% A

ctiva

ted

CD

4+ T

ILs

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

% C

D4

+ T

ce

lls

in t

um

or-

dra

inin

g L

Ns

% A

ctivate

d C

D4

+ T

ce

lls

in t

um

or-

dra

inin

g L

Ns

% F

oxP

3+ o

f tu

mo

r-in

filtra

tin

g

CD

4+ T

ce

lls

% F

oxP

3+ o

f tu

mo

r-d

rain

ing

LN

-resid

ent

CD

4+ T

ce

lls

**

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1

anti-

TIGIT

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1+an

ti-TIG

IT

Con

trol

anti-

PD-L

1+an

ti-TIG

IT

3’extra for PCR

Total length homology arms = 6030 bp 5’homology arm = 2405 bp Floxed region contain exon1 = 937 bp 3’homology arm = 3175 bp Selection Marker = Frt- PGK-em7-NEO- Frt

Control vector

Screening by PCR followed by sequencing

5’extra genomic region = 556 bp 3’extra genomic region = 284 bp

Floxed region

LINE LTR SINE Other

annotations

RNA

UTR Exon CNS

5’arm 3’arm 3’extra for PCR

1

loxP1 loxP2

Frt-PGK-em7-NEO-Frt

A

C

WT

CKO

0

20

40

60

80

60% 60%

WT CKO

CD8

IFNγ

WT

CKO

0

20

40

60

80

100

84% 84%

WT CKO

CD8

CD

44

CD4

IFNγ

C

WT

CKO

0

5

10

15

20WT CKO

12% 10%

D

WT

CKO

0

20

40

60WT CKO

38% 42%

CD4

CD

44

E

WT

CKO

0

15

30

45

20% 22%

CD62L

CD

44

a

WT

CKO

0

10

20

30

40F

21% 24%

CD62L

CD

44

G

WT

CKO

0

2

4

6

8

WT

CKO

0

1

2

3

4

4% 9%

CD8

IFNγ

2% 4%

CD4

IFNγ

WT

CKO

1000

10000

100000H I J

WT CKOWT CKO

WT CKOWT CKO

% A

ctivate

dC

D8

+ T

cells

% I

FNγ+

of

CD

8+ T

cells

% A

ctivate

dC

D4

+ T

cells

% I

FNγ+

of

CD

4+ T

cells

% A

ctivate

dC

D8

+ T

cells

% A

ctivate

dC

D4

+ T

cells

% IF

Nγ+

of C

D8

+ T

cells

**

% IF

Nγ+

of C

D4

+ T

cells

**

Liv

er

LC

MV

Titer

(PF

U/m

L)

***

K

CD62L

CD

44

Control anti-PD-L1 anti-TIGIT anti-PD-L1

+ anti-TIGIT

25% 27% 21% 25%

L

CD4

IFNγ

0

4

8

12

16

20

*

0

10

20

30

40

50

11% 11%8%7%

M

0

3

6

9

% T

ota

l C

D8

+

T c

ells

0

1

2

3

% G

P33-s

pecific

of

CD

8+ T

cells **

N O

IFNγ

Penta

mer

Control

24% 25% 59%

anti-PD-L1 anti-TIGIT

anti-PD-L1

+ anti-TIGIT

19%

0

20

40

60

% I

FNγ+

of

GP

33

-

specific

CD

8+ T

ce

lls **

*

anti-PD-L1anti-TIGIT

_ _

_ _+ +

+ +

% I

FNγ+

of

CD

4+ T

ce

lls

% A

ctiva

ted

CD

4+ T

ce

lls

Control anti-PD-L1 anti-TIGIT anti-PD-L1

+ anti-TIGIT

anti-PD-L1anti-TIGIT

_ _

_ _+ +

+ +

anti-PD-L1anti-TIGIT

_ _

_ _+ +

+ +anti-PD-L1anti-TIGIT

_ _

_ _+ +

+ +anti-PD-L1anti-TIGIT

_ _

_ _+ +

+ +

Figure S5. Related to Figure 6. Further characterization of TIGIT’s role in regulating T cell

responses to acute and chronic LCMV infection. (A) Exon 1 of TIGIT was flanked by loxP

sites to generate TIGITfl/fl

mice, which were then crossed to CD4cre

mice. (B-D) TIGITfl/fl

;CD4cre

(CKO) and TIGITfl/fl

littermates (WT) were infected with Armstrong strain LCMV. Splenocytes

were analyzed 7 days after infection. Data are representative of two independent experiments; n

= 5. (B) Representative FACS plots gated on CD8+ T cells, with activated (CD44

high) cells

boxed. Quantitation of activated CD8+ T cells as a percentage of total CD8

+ T cells. (C)

Representative FACS plots gated on CD8+ T cells after stimulation in vitro, with IFN -producing

cells boxed. Quantitation of IFN -producing cells as a percentage of total CD8+ T cells. (D)

Representative FACS plots gated on CD4+ T cells, with activated (CD44

high) cells boxed.

Quantitation of activated CD4+ T cells as a percentage of total CD4

+ T cells. (E) Representative

FACS plots gated on CD4+ T cells after stimulation in vitro, with IFN -producing cells boxed.

Quantitation of IFN -producing cells as a percentage of total CD4+ T cells. (F-J) TIGIT

fl/fl ;CD4

(CKO) and TIGITfl/fl

(WT) littermates were briefly depleted of CD4+ T cells and

infected with Clone 13 strain LCMV. Splenocytes and liver viral titers were analyzed 42 days

after infection. Data are representative of 2 independent experiments, and n = 6-9 per group. (F)

Representative FACS plots gated on CD8+ T cells, with activated cells (CD44

high CD62L

low)

boxed. Quantitation of activated cells as a percentage of total CD8+ T cells. (G) Representative

FACS plots gated on CD4+ T cells, with activated cells (CD44

high CD62L

low) boxed. Quantitation

of activated cells as a percentage of total CD4+ T cells. (H) Representative FACS plots gated on

CD8+ T cells after stimulation in vitro, with IFN

+ cells boxed. Quantitation of IFN -producing

cells as a percentage of CD8+ T cells. **, p = 0.010. (I) Representative FACS plots gated on

CD4+ T cells after stimulation in vitro, with IFN

+ cells boxed. Quantitation of IFN -producing

cells as a percentage of CD4+ T cells. **, p = 0.010. (J) Quantitation of liver LCMV titers. ***,

p < 0.0001. (K-O) C57BL6/J mice were briefly depleted of CD4+ T cells and infected with

Clone 13 strain LCMV as described in Figure 6. Mice were treated with isotype-matched control,

anti-PD-L1, anti-TIGIT, or anti-PD-L1 + anti-TIGIT antibodies starting 28 days after infection,

as in Figure 6. Splenocytes and liver viral titers were analyzed 42 days after infection. (K)

Representative FACS plots gated on CD4+ T cells, with activated cells (CD44

high CD62L

low)

boxed. Quantitation of activated cells as a percentage of total CD4+ T cells. (L) Representative

FACS plots gated on activated CD4+ T cells after stimulation in vitro, with IFN

+ cells boxed.

Quantitation of IFN -producing cells as a percentage of activated CD4+ T cells. *, p = 0.0497.

(M) Quantitation of CD8+ T cells as a percentage of all splenocytes. (N) Quantitation of gp33

Pentamer+ cells as a percentage of all splenic CD8

+ T cells. **, p = 0.0040. (O) Representative

FACS plots gated on gp33 pentamer+ CD8

+ T cells after stimulation in vitro, with IFN

+ cells

boxed. Quantitation of IFN -producing cells as a percentage of all gp33 pentamer+ CD8

+ T cells.

*, p = 0.0319. **, p = 0.0030. Error bars depict the standard error of the mean.

cre

A

CD8+ T

cells

CD4+ T

cells

non

-T cells

CD8+ T

cells

CD4+ T

cells

non

-T cells

0

20

40

60

80

100

CD226 CD226

Spleen Tumor

non-T cells

CD8+ T cells

CD4+ T cells

B

Figure S6. Releated to Figure 7. Expression of endogenous CD226 by murine tumor-

infiltrating T cells and expression of exogenous FLAG-ST CD226 and HA-TIGIT by CHO cells.-

(A) Wildtype BALB/c mice were inoculated with CT26 tumor cells as described in Figure 3.

After tumors have grown to approximately 150-200 mm3 in size, tumors and spleens were

analyzed by flow cytometry. (A) Quantitation of CD226+ CD8

+ T cells, CD4

+ T cells, and non-T

cells, as a percentage of all CD8+ T cells, CD4

+ T cells, and non-T cells respectively.

Representative histograms of CD226 expression in tumor and spleen. Data are representative of

two independent experiments; n = 5. (B) Wildtype CHO cells were transiently transfected with

FLAG-ST-CD226 and HA-TIGIT as indicated. Quantification of anti-FLAG (black) and anti-

HA (red) ELISAs is shown. Data are representative of three independent experiments; n = 4.

Error bars depict the standard deviation of the mean.

% C

D2

26

+

TumorSpleen

Fla

g lum

inescence (

a.u

.)

1000

10000

100000

1000000

anti-CD226anti-TIGIT

_ _

_ _+ +

+ +anti-PD-L1

_ _ + +

Figure S7. Related to Figure 8. TIGIT/PD-L1 blockade efficacy is dependent on CD226

during chronic viral infection. C57BL6/J mice were briefly depleted of CD4+ T cells and

infected with Clone 13 strain LCMV as described in Figure 6. Mice were treated with isotype

control, anti-CD226, and/or anti-PD-L1 + anti-TIGIT antibodies between days 28 and 42 post-

infection. Liver viral titers were analyzed 42 days after infection. n = 10. ***, p < 0.001. Error

bars depict the standard error of the mean.

Liv

er

LC

MV

Tite

r (P

FU

/mL

) ***

Supplemental Experimental Procedures

Bioinformatics. Processing and analysis of RNA-sequencing data was performed using the R

programming language (http://www.r-project.org) along with several packages from the

Bioconductor project (http://www.bioconductor.org). RNA-sequencing data for cancer and

matched normal samples were obtained from the TCGA for five different indications: breast

cancer (Network, 2012a), colon adenocarcinoma (Network, 2012c), renal clear cell

carcinoma(Network, 2013), lung squamous cell carcinoma (Network, 2012b), and endometrial

carcinoma (Network, 2012d). Raw RNA-seq reads were processed using the HTSeqGenie

Bioconductor package. Briefly, reads were aligned to the human genome (NCBI build 37) using

the GSNAP algorithm (Wu and Nacu, 2010). Uniquely aligned read pairs that fell within exons

were counted to give an estimate of gene expression level for individual genes. We used the

library size estimation from the edgeR package (Robinson et al., 2010) to normalize across

different samples for their respective sequencing depths.

To calculate the T cell gene expression signature score in the lung squamous cell

carcinoma data, we first performed a variance stabilizing transform on the raw count data using

the voom function from the limma Bioconductor package. We then calculated the first

eigenvector of the centered and scaled variance-stabilized data from the 15-gene T cell signature.

This approach yields a robust per-sample estimate of relative T cell abundance. A linear model

including the T cell signature score was then fit for each gene, again using the limma package.

We then ranked the genes by their correlation with the T cell signature in our linear model,

choosing only genes positively correlated with the T cell signature. For visualizing T cell-

associated genes as a heatmap, we centered and scaled the variance-stabilized data to unit

variance, allowing for comparison of genes with different average expression levels.

To determine the correlation between expression of TIGIT and other genes, we

normalized RNA-sequencing count data to account for differences in library size, using the

method from the edgeR Bioconductor package (Robinson et al., 2010). We then calculated

Spearman’s rank correlation coefficient on the normalized counts. We consider rho > 0.75 to be

indicative of strong correlation, rho ≤ 0.75 but > 0.5 to be indicative of moderate correlation, and

rho ≤ 0.5 but > 0.25 to be indicative of weak correlation.

For calculation of TIGIT/CD3ε ratios across each indication, we first calculated the

variance-stabilized data for each RNA-sequencing data set. We then calculated the log2 ratio of

the variance-stabilized data for TIGIT and CD3ε. To calculate the difference between tumor and

normal samples, we performed standard linear model analysis using standard R functions. We

accepted a p-value of <0.01 as evidence of a significant difference between tumor and normal.

Flow cytometry. Single cell suspensions of mouse spleen, lymph node, and tumor were prepared

with gentle mechanical disruption. Surface staining was performed with commercial antibodies

against CD4, CD8, CD44, CD62L, PD-1 (eBiosciences) and CD226 (Biolegend). TIGIT

antibodies were generated at Genentech as previously described (Yu et al., 2009) and conjugated

to Alexa Fluor 647 according to the manufacturer’s directions (Molecular Probes). LCMV gp33-

specific pentamer staining (Proimmune) was performed using the manufacturer’s instructions.

For intracellular cytokine staining (ICS), cells were stimulated for 4 hours with 20 ng/mL

Phorbol 12-myristate 13-acetate (PMA, Sigma) and 1 µM Ionomycin (Sigma) in the presence of

3 µg/mL Brefeldin A (eBiosciences). After stimulation, cells were stained for surface markers as

described and fixed and permeabilized with eBioscience FoxP3 fixation buffer set according to

the manufacturer’s directions. Fixed cells were stained with antibodies against IFNγ and TNFα

(eBiosciences).

Human tumor and PBMC samples were prepared as described above. Surface staining

was performed with a viability dye (Molecular Probes), commercial antibodies against CD45

(eBiosciences), CD3, CD4, CD8, PD-1 (BD Biosciences), and with anti-TIGIT antibodies

prepared as described above.

All samples were acquired on LSR-II or LSR-Fortessa instruments (BD Biosciences) and

analyzed using FlowJo software (Treestar).

Mice. C57BL/6J and BALB/c mice were purchased from the Jackson Laboratory and Charles

River Laboratories, respectively. CD4cre

mice and TIGITfl/fl

mice were generated on a C57BL/6J

background and housed in specific pathogen-free conditions at Genentech. Crossing CD4cre

and

TIGITfl/fl

mice reduced the frequency of TIGIT expressing T cells by 96% in the spleen.

Cell lines. CT26, EMT6, COS-7, and CHO cell lines were obtained from the American Type

Culture Collection (ATCC). All cell lines were banked at Genentech and subsequently tested to

be free of mycoplasma contamination.

Therapeutic antibodies and recombinant proteins. A blocking anti-TIGIT IgG2a monoclonal

antibody (clone 10A7, reactive against both mouse and human TIGIT) was generated as

previously described(Yu et al., 2009) and cloned onto a murine IgG2a isotype. A blocking anti-

PD-L1 IgG2a monoclonal antibody (clone 25A1) was generated by immunization of Pdl1-/-

mice

with a PD-L1-Fc fusion protein and cloned onto a murine IgG2a isotype. Clone 25A1 was

modified with previously described mutations abolishing binding to Fcγ receptors (Shields et al.,

2001). A blocking anti-CD226 IgG2a monoclonal antibody (clone 37F6) was generated by

immunization of hamsters with recombinant murine CD226 and cloned onto a murine IgG2a

backbone. CD8+ T cell depletion was performed with a depleting rat IgG2a monoclonal antibody

against CD8α (clone 53-6.72, BioXcell) (Hathcock, 2001). Recombinant human PVR-Fc fusion

protein was generated as previously described (Yu et al., 2009). Recombinant murine PVR-Fc

fusion protein was generated as described, with murine PVR amino acids 1-319.

Viral titer assay. Monolayers of MC57 cells were cultured with an overlay of 1%

methylcellulose and infected with serially diluted liver homogenates from LCMV-infected mice.

72 hours after infection, the cells were fixed with 4% paraformaldehyde and permeabilized with

0.5% Triton-X. Viral plaques were stained with anti-LCMV NP (clone VL-4) and HRP-

conjugated anti-rat IgG and visualized with O-phenylenediamine (OPD, Sigma)

In vitro murine CD8+ T cell cultures. Splenic CD8

+ T cells were enriched by negative magnetic

selection (Miltenyi) from TIGITfl/fl

;CD4cre

and TIGITfl/fl

littermate mice, or from wild-type

C57BL6/J mice, and stimulated with sub-optimal levels of plate-bound anti-CD3 and plate-

bound recombinant murine PVR-Fc. Anti-CD226 blocking antibodies were added as indicated.

Proliferation was measured via 3H-Thymidine incorporation performed in triplicate.

In vitro human CD8+ T cell cultures. CD8

+ T cells were enriched from whole blood by

negative magnetic selection (RosetteSep) and stimulated with sub-optimal levels of plate-bound

anti-CD3 and plate-bound recombinant human PVR-Fc. Proliferation was measured via 3H-

Thymidine incorporation performed in triplicate.

TR-FRET with transfected cell lines. CHO cells were transfected with N-terminus

SNAP-tagged (ST) CD226 and N- terminus HA-TIGIT using Lipofectamine 2000 (Life

Technologies) and seeded in a white 96-well plate (Costar) at 100,000 cells per well. 24

hours later, cells were labeled to measure TR-FRET either between SNAP-donor /

SNAP-acceptor or between SNAP-acceptor / anti-HA donor. 1) SNAP-donor / SNAP-

acceptor labeling: Cells were incubated with 100 nM of donor-conjugated benzyl-

guanine SNAP-Lumi-4Tb (Cisbio) and 1 µM acceptor-conjugated benzyl-guanine SNAP-

A647 (New England Biolabs) diluted in DMEM 10% FCS for lh at 37°C, 5% C02. Cells

were then washed three times in PBS before reading of the FRET signal. 2 ) S N A P -

a c c e p t o r / a n t i - H A d o n o r : C e l l s w e r e i n c u b a t e d w i t h 1 µM acceptor-

conjugated benzyl-guanine SNAP-A647 diluted in DMEM 10% FCS for lh at 37°C, 5%

C02. After three washes in PBS, cells were incubated for 2 hours with 2 nM anti-HA

Lumi-4Tb (Cisbio) in PBS + 0.2% BSA at room temperature. The FRET signal was

t h e n recorded at 665 nm for 400 µs after a 60 µs delay following laser excitation at 343

nm using a Safire2 plate reader (Tecan). When anti-TIGIT was tested at 10µg/ml final, the

FRET signal was also recorded after a 15 min incubation. For the Flag-ST-CD226/Flag-ST-

CD226 interaction, the FRET ratio was calculated as the FRET intensity divided by the donor

emission at 620 nm, which is proportional to the CD226 expression. The FRET intensity being:

(signal at 665 nm from cells labeled with SNAP-donor and acceptor) - (signal at 665 nm from

the same batch of transfected cells labeled with SNAP-donor only). For the Flag-ST-

CD226/HA-TIGIT interaction, the FRET ratio represents the FRET intensity divided by the

Flag-ST-CD226 expression as measured by an anti-Flag ELISA. In that case, the FRET

intensity = (signal at 665 nm from cells labeled with SNAP-acceptor and anti-HA donor) -

(signal at 665 nm from mock transfected cells labeled with SNAP-acceptor and anti-HA donor).

ELISA. CHO cells were fixed with 4% paraformaldehyde, washed twice, and blocked in

phosphate-buffered saline + 1% fetal calf serum (FCS). Cells were then incubated with an anti-

HA monoclonal antibody (clone 3F10, Roche applied science) or or anti-Flag-M2 monoclonal

antibody (Sigma), both conjugated with horseradish peroxidase. After washes, cells were

incubated with a SuperSignal ELISA substrate (Pierce) and chemoluminescence was detected

on a Safire2 plate reader (Tecan). Specific signal was calculated by subtracting the signal

recorded on mock transfected cells.

Co-immunoprecipitation. Briefly, COS 7 Cells in 15 cm plates were co-transfected with

expression plasmids containing the cDNA for either TIGIT-HA (5ng) or CD226-Flag (10ng)

tagged proteins, or a control plasmid (pRK). 23 hrs after transfection the cells were washed

with PBS and harvested in 4 ml of ice cold PBS and centrifuged at 300xg for 5min and cell

pellets were re-suspended in 2 ml of Lysis buffer at 4°C. The cells were lysed over 50 min

with vortexing every 15 min and subsequently centrifuged at 10,00xg for 15 min at 4C.

The resultant supernatant was pre-cleared with 160 ul of CL6B sepahrose slurry by

rotating for 30 min at 4°C, and centrifuged for 2 min at 3000xg. The supernatant was

equally split into two tubes and immuno-precipitated with either an anti-HA or an anti-flag

using standard procedures. The immune-precipitated proteins were subjected to SDS-

PAGE and western blotted. Western blots were probed with either anti-Flag-HRP or anti-

HA-HRP.

TR-FRET with human T cells. Human anti-TIGIT (Genentech clone 1F4), anti-CD226 (Santa

Cruz Biotechnology) and anti-HVEM (eBioscience) were conjugated fluorophores compatible

with TR-FRET (Cisbio). Primary human T cells were MACS-enriched from blood, stimulated

in vitro with plate bound anti-CD3 and anti-CD28 for 72 hours. TIGIT-expressing and non-

expressing T cells (all expressing CD226) were then sorted, rested without stimulation for 72

hours, and re-stimulated for 48 hours. Each population was then washed once with Tris-KREBS

buffer (20mM Tris pH 7.4, 118mM NaCl, 5.6mM glucose, 1.2mM KH2PO4, 1.2mM MgSO4,

4.7mM KCl, 1.8mM CaCl2) and cultured under the following conditions, in triplicate: 1) Anti-

TIGIT Ab-Lumi4-Tb (5µg/ml), 2) Anti-TIGIT Ab-Lumi4-Tb (5µg/ml) + anti-HVEM-d2

(10µg/ml), 3) Anti-TIGIT Ab-Lumi4-Tb (5µg/ml) + anti-CD226-d2 (10µg/ml), 4) Anti-TIGIT

Ab-Lumi4-Tb (5µg/ml) + anti-CD226-d2 (10µg/ml) + cold anti-TIGIT Ab (clone 1F4)

(50µg/ml). The indicated concentrations were optimized to ensure the highest FRET signal.

Cells were incubated for 2 hours at room temperature on a rotator and then washed 3 times in

Tris-KREBS buffer. T cells were then seeded at 400,000 cells/well in a white 96-well plate

(Costar) and TR-FRET was recorded at 665 nm for 400 µs after a 60 µs delay following laser

excitation at 343 nm using a PHERAstar plate reader (BMG Labtech). FRET intensity was

expressed as the signal at 665 nm from cells labeled with Ab-Lumi4-Tb + Ab-d2 minus the

signal at 665 nm from the same batch of cells labeled with Ab-Lumi4-Tb alone. The non-

specific FRET signal was given by the T cells incubated with Lumi4Tb + d2 + an excess of

cold Ab.

Statistics. Statistical tests were conducted using unpaired (or paired where specified) 2-tailed

Student’s t-tests. Mantel-Cox log-rank tests were used for survival analyses. For differential

expression analysis, moderated t-statistics were calculated with the limma package, as previously

described (Smyth, 2004). To evaluate correlation, Pearson’s correlation coefficients were used.

Error bars depict the standard error of the mean.

Supplemental References

Hathcock, K. S. (2001). T cell depletion by cytotoxic elimination. Current Protocols in

Immunology. Edited by John E Coligan et al. Chapter 3, Unit 3 4.

Network, C. G. A. (2012a). Comprehensive molecular portraits of human breast tumours. Nature

490, 61-70.

Network, C. G. A. (2013). Comprehensive molecular characterization of clear cell renal cell

carcinoma. Nature 499, 43-49.

Network, T. C. G. A. (2012b). Comprehensive genomic characterization of squamous cell lung

cancers. Nature 489, 519-525.

Network, T. C. G. A. (2012c). Comprehensive molecular characterization of human colon and

rectal cancer. Nature 487, 330-337.

Network, T. C. G. A. (2012d). Integrated genomic characterization of endometrial carcinoma.

Nature 497, 67-73.

Robinson, M. D., McCarthy, D. J., and Smyth, G. K. (2010). edgeR: a Bioconductor package for

differential expression analysis of digital gene expression data. Bioinformatics (Oxford,

England) 26, 139-140.

Shields, R. L., Namenuk, A. K., Hong, K., Meng, Y. G., Rae, J., Briggs, J., Xie, D., Lai, J.,

Stadlen, A., Li, B., et al. (2001). High resolution mapping of the binding site on human IgG1 for

Fc gamma RI, Fc gamma RII, Fc gamma RIII, and FcRn and design of IgG1 variants with

improved binding to the Fc gamma R. J. Biol. Chem. 276, 6591-6604.

Smyth, G. K. (2004). Linear models and empirical bayes methods for assessing differential

expression in microarray experiments. Statistical applications in genetics and molecular biology

3, Article 3.

Wu, T. D., and Nacu, S. (2010). Fast and SNP-tolerant detection of complex variants and

splicing in short reads. Bioinformatics (Oxford, England) 26, 873-881.

Yu, X., Harden, K., Gonzalez, L. C., Francesco, M., Chiang, E., Irving, B., Tom, I., Ivelja, S.,

Refino, C. J., Clark, H., et al. (2009). The surface protein TIGIT suppresses T cell activation by

promoting the generation of mature immunoregulatory dendritic cells. Nat. Immunol. 10, 48-57.