� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 67

Tsvetelina Pentcheva-Hoang

Emily Corse

James P. Allison

Negative regulators of T-cellactivation: potential targets fortherapeutic intervention in cancer,autoimmune disease, and persistentinfections

Authors’ address

Tsvetelina Pentcheva-Hoang1, Emily Corse1, James P. Allison1

1Department of Immunology, Memorial Sloan-Kettering

Cancer Center, New York, NY, USA.

Correspondence to:

James P. Allison

Department of Immunology

Howard Hughes Medical Institute, Ludwig Center for

Cancer Immunotherapy

Memorial Sloan-Kettering Cancer Center

1275 York Avenue

New York, NY 10021, USA

Tel.: +1 646 888 2332

Fax: +1 646 422 0618

e-mail: allisonj@mskcc.org

Immunological Reviews 2009

Vol. 229: 67–87

Printed in Singapore. All rights reserved

� 2009 John Wiley & Sons A/SImmunological Reviews

0105-2896

Summary: The generation of productive adaptive immune responsesdepends on the antigen-specific activation of T and B cells. The outcomeof T-cell receptor engagement is influenced by signals from both positiveand negative regulatory molecules that can either activate or inhibit T-cellfunction. CD28 and cytotoxic T-lymphocyte antigen-4 are the prototypi-cal members of an immunoglobulin domain-containing protein familythat play important roles in the control of T-cell responses against infec-tion, cancer, and in autoimmune disease. Although the precise moleculardetails of their functions are still under active investigation, tumors andchronic pathogens seem to have exploited these pathways to achieveimmune evasion. Furthermore, malfunction of the inhibitory arm of theimmune response appears responsible for the development of multipleautoimmune pathologies. As a result, the negative regulators of T-cellactivation have become attractive targets for therapeutic intervention incancer, chronic infection, and autoimmune disease. The application offindings from basic research has provided insight into the manipulationof these pathways in the clinic and offers promising strategies for thetreatment of disease.

Keywords: CTLA-4, PD-1, BTLA, LAG-3, immunotherapy

Introduction

The activation of ab T cells requires the engagement of unique

T-cell receptors (TCRs) by antigenic peptides bound to major

histocompatibility complex (MHC) molecules on the surface of

antigen-presenting cells (APCs). This interaction provides the

so-called ‘signal 1’, which by itself is not only insufficient for

T-cell activation but can lead to apoptosis or a state of antigen-

specific non-responsiveness (anergy) (1). Optimal T-cell

proliferation and acquisition of effector functions require an

additional costimulatory ‘signal 2’,which is provided by surface

molecules on specialized professional APCs (1).

CD28 is the founding member of a growing family of

costimulatory molecules and has been shown to enhance

interleukin-2 (IL-2) secretion and to prevent the induction of

anergy in both T-cell clones and primary T cells (2, 3). Upon

binding to its ligands, B7-1 (CD80) and B7-2 (CD86), CD28

enhances T-cell proliferation by increasing the transcription

and the mRNA stability of IL-2 (4) as well as by upregulating

the anti-apoptotic protein Bcl-XL (5). Although the precise

mechanism of CD28-mediated costimulation is unknown,

some of the possibilities include driving the formation of a

mature immunological synapse (6), assembly of an indepen-

dent signaling complex (7), and augmentation of the TCR

signals, potentially via the recruitment of lipid rafts to the

synapse (8).

A closely related molecule, cytotoxic T-lymphocyte antigen-

4 (CTLA-4) (CD152) (9), binds to the same ligands as CD28

with much higher affinity (10, 11). In contrast to CD28,

CTLA-4 restricts T-cell activation, as indicated by the fact that

CTLA-4 ligation inhibits IL-2 production, IL-2 receptor

expression, and cell cycle progression (12, 13) and that

CTLA-4-deficient mice die of a lymphoproliferative disorder

within 3–4 weeks of age (14–16).

The more recent discovery of new homologues of

CD28 ⁄ CTLA-4 and their ligands has gradually expanded thisprotein family to include additional costimulatory [inducible

T-cell costimulator (ICOS) ⁄CD278 and its ligand B7h ⁄B7-RP ⁄ ICOS-L ⁄CD275 (17–19)], as well as inhibitory[programmed death-1 (PD-1) ⁄CD279 and its ligands PD-L1 ⁄B7-H1 ⁄ CD274 and PD-L2 ⁄B7-DC ⁄CD273 (20–23)] members.Furthermore, the identification of B7-H3 ⁄CD276 (24, 25)

and B7x ⁄B7-H4 (26–28) as novel homologues of B7-1 andB7-2 that can inhibit T cells by binding to currently unknown

receptors extends this family even further. An additional

receptor of the immunoglobulin superfamily (IgSF) that

negatively regulates T-cell activation is the B and T-lympho-

cyte attenuator (BTLA) ⁄CD272 (29), which binds the herpes-virus entry mediator (HVEM) (30). HVEM has also recently

been shown to interact with another negative regulator of T

cells, CD160 (31). Finally, recent studies of the lymphocyte

activation gene 3 (LAG-3) ⁄CD223 suggest that it plays animportant role in the regulation of T-cell responses (32–34).

This review focuses on recent advances in our understanding

of these negative regulators of T-cell activation (Fig. 1).

CTLA-4, B7-1, and B7-2

Protein structure

Mouse CTLA-4 is a 223-amino acid (a.a.) type I trans-

membrane glycoprotein that contains an unpaired cysteine

residue at position 122 (Cys157 if the signal sequence is

included) that is responsible for the formation of a disulfide-

linked homodimer (9, 35). There are several CTLA-4 splice

variants that include a soluble form (36, 37) as well as a

ligand-independent molecule, found only in the mouse,

which may be associated with resistance to autoimmune dis-

ease (38, 39). The solution and crystal structures of both

mouse and human CTLA-4, either alone or in complexes with

B7-1 and B7-2, have been solved (40–43) and show that the

B7-1B7-1PD-1PD-1

LAG-3LAG-3TCRTCR

CD160CD160

BTLABTLA

PD-L1PD-L1

CTLA-4CTLA-4LAG-3LAG-3

TCRTCRCD28CD28

PD-1PD-1

TCRTCR

TCRTCR

??

BTLABTLACD160CD160B7-1B7-1

PD-1PD-1LAG-3LAG-3

CD28CD28BTLABTLA

CD160CD160BTLABTLAPD-1PD-1

B7-1B7-1

PD-L1PD-L1

HVEMHVEMPD-L1PD-L1

CTLA-4CTLA-4

PD-1PD-1B7-1B7-1

CTLA-4CTLA-4

pMHCpMHCB7-1B7-1

B7-1B7-1

HVEMHVEM

pMHCpMHCpMHCpMHC

PD-L2PD-L2B7-2B7-2

PD-L1PD-L1PD-L1PD-L1

B7xB7x

B7-H3B7-H3

HVEMHVEMEffector T cellEffector T cellproliferationproliferation

and/or functionand/or function

B7-2B7-2

HVEMHVEM

PD-L2PD-L2

pMHCpMHC

PD-L1PD-L1

DendriticDendriticcellcell

PeripheralPeripheraltissuestissues

RegulatoryRegulatoryT cellT cell

TissueTissuemacrophagemacrophage

Fig. 1. T-cell inhibitors of the immunoglobulin superfamily.

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

68 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

extracellular domain consists of a two-layer b-sandwich that

resembles an Ig variable (IgV) domain (40, 41). The two

ligand-binding sites lie distally to the homodimer interface,

allowing the simultaneous binding of two B7-1 or two B7-2

molecules to a single CTLA-4 (41–43).

The mouse B7-1 protein is a 306 aa type I transmembrane

glycoprotein that has an extracellular region consisting of two

Ig-like domains: a membrane-distal IgV-like domain and a

membrane-proximal Ig constant region (IgC)-like domain

(44). The transmembrane segment contains two cysteine resi-

dues that might be involved in lipid modifications or covalent

interactions with other membrane proteins (44). Because of

alternative splicing, there is a mouse B7-1 variant with a

different cytoplasmic domain, which might contain distinct

signaling modules (45). The crystal structure of human B7-1

shows that it forms non-covalently linked homodimers, both

when free and when bound to CTLA-4 (43, 46). Each mono-

mer consists of two anti-parallel b-sandwich Ig domains, one

characteristic of adhesion molecules and the other typical for

antigen receptors (46).

The mouse B7-2 protein is a type I transmembrane glyco-

protein that consists of 309 aa (47). The extracellular regions

of both mouse and human B7-2 contain highly homologous

IgV-like and IgC-like domains, while the transmembrane

regions and the cytoplasmic tails are much less conserved (47,

48). The crystal structures of the IgV-like receptor-binding

region of human B7-2 alone and in complex with CTLA-4

have been solved (42, 49) and show that this domain is a

two-layer b-sandwich that upon binding to CTLA-4 may form

a non-covalently linked homodimer. Unlike B7-1, where the

amino acids at the homodimer interface are mostly hydropho-

bic, the residues mediating the B7-2 homodimerization are

highly hydrophilic (49) and might account for the tendency

of B7-2 to be a monomer in solution (50). This conclusion is

supported by data from over-expression studies in Chinese

hamster ovary (CHO) cells using fluorescence energy

resonance transfer (FRET) measurements (51).

Expression of CTLA-4, B7-1, and B7-2

The mRNA for CTLA-4 is not detectable in resting T cells but

is upregulated within 1 h of TCR stimulation and peaks

between 24 and 48 h after activation (52, 53). The protein is

detectable on the surface of T cells within 24 h of stimulation

(53); however, the majority of the cellular CTLA-4 is present

in intracellular compartments (54, 55). CTLA-4 has been

observed in endosomes (56), lysosomes (57), and the trans

Golgi network (TGN) (58). This intracellular localization

seems to be predominantly due to the presence of a tyrosine-

based internalization motif in the cytoplasmic tail of CTLA-4

that binds the l2 subunit of the clathrin adapter protein-2

(AP-2) (59–61). Phosphorylation or mutations of Tyr201 inhi-

bit the binding of l2 and lead to increased levels of CTLA-4

on the surface (59, 61). As the phosphorylated Tyr201 is also

part of a Src homology 2 (SH2)-binding motif that mediates

association with a number of signaling molecules (see below),

there appears to be an inverse relationship between CTLA-4

internalization and its ability to generate inhibitory signals

into T cells.

TCR ligation and increased intracellular calcium levels cause

an upregulation of surface CTLA-4 expression, possibly via the

secretion of CTLA-4-containing secretory lysosomes (56, 57).

The same Tyr201 is also part of another intracellular trafficking

motif that binds to the clathrin adapter protein AP-1 and

mediates transport from the Golgi complex to lysosomes or

possibly to the plasma membrane (62). More recently, the

TCR-interacting molecule (TRIM) has been described as a

chaperone that facilitates CTLA-4 transport from the TGN to

the surface (58). The observations that the levels of TRIM

expression correlate with the amount of surface CTLA-4

suggest that TRIM is an important regulator of CTLA-4 export

from the TGN to the plasma membrane (58).

During T-cell migration, the majority of CTLA-4 is present

in vesicles near the microtubule-organizing center (MTOC) at

the back of the cell (63). Upon engagement of the TCR, these

intracellular vesicles re-orient towards the APC with a fraction

of CTLA-4 eventually localizing to the synapse (63). Interest-

ingly, the amount of CTLA-4 that translocates to the synapse is

proportional to the strength of the TCR signal, suggesting that

CTLA-4 might preferentially restrict T-cell responses to strong

TCR signals (63).

The expression of CTLA-4 in effector T cells is under tight

regulation at the level of synthesis and degradation: in the

absence of new protein synthesis, the half-life of CTLA-4 is

about 2 h (63). In contrast, T-regulatory (Treg) cells constitu-

tively express high levels of CTLA-4 on their surface (64, 65).

This expression could be due to the fact that these cells appear

to carry self-reactive TCRs of high affinity (66, 67) that might

trigger the constant synthesis of CTLA-4. However, the

molecule appears to be more than a marker of chronic TCR

stimulation, because the targeted deletion of CTLA-4 specifi-

cally in mouse Treg cells leads to spontaneous systemic lym-

phoproliferation and fatal T-cell-mediated autoimmunity (68).

As already mentioned, CTLA-4 shares its ligands, B7-1 and

B7-2, with CD28. Initial studies of B7-1 and B7-2 seemed to

indicate that the molecules have redundant functions in both

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 69

costimulation and inhibition (69–71). B7-1 and B7-2 are

expressed on the same types of cells, including B cells,

activated T cells, macrophages, and dendritic cells (DCs) (72),

although B7-2 is expressed earlier than B7-1 on B cells after

lipopolysaccharide (LPS) stimulation (73) and appears to be

the major costimulatory ligand on in vitro cultured DCs (74).

Interestingly, the expression of B7-1 on in vitro generated

immature DCs appears to be higher than that of B7-2, and this

pattern correlates with a more tolerogenic phenotype (75).

This association is noteworthy in light of the fact that B7-1 is

the major ligand mediating CTLA-4 accumulation at the

synapse (76), presumably because of the high affinity and

avidity of this interaction (11).

Mechanisms of CTLA-4 inhibition

There have been numerous models proposed for the way in

which CTLA-4 inhibits T-cell activation. These mechanisms

can be generally divided into cell-intrinsic (affecting only the

effector T cell expressing CTLA-4) and cell-extrinsic (affecting

other cells besides the effectors). A summary of the suggested

models is below.

Competition for ligand and inhibitory lattice formation

The basis for ligand competition as a mechanism for CTLA-4-

mediated inhibition of T-cell activation lies with the observa-

tion that the affinities of B7-1 and B7-2 for CTLA-4 are 10- to

100-fold greater than their affinities for CD28 (10, 11). This

finding suggests that when CTLA-4 is expressed at the plasma

membrane after activation and under conditions where ligand

expression is limiting, CTLA-4 might be able to act as a B7

sink and effectively inhibit CD28-mediated costimulation by

ligand sequestration. Support for this model comes from stud-

ies showing that a CTLA-4 mutant lacking the cytoplasmic tail

is capable of preventing the multiorgan lymphocytic infiltra-

tion and early death observed in CTLA-4-deficient mice (77).

However, the tailless CTLA-4 cannot correct the lymphade-

nopathy or the accumulation of large numbers of activated

T cells in older mutant mice, suggesting that CTLA-4 signaling

or proper cellular trafficking are required for complete inhibi-

tion of T-cell activation (77). In contrast, experiments using

Jurkat cells expressing a panel of CTLA-4 mutants have shown

that a tailless molecule, incapable of delivering a negative signal

in these cells, is fully able to inhibit IL-2 secretion, and this

inhibition is directly proportional to the amount of CTLA-4

on the surface, suggesting that ligand sequestration is a major

mechanism for CTLA-4-mediated inhibition (50). More recent

work from our laboratory has led to additional support for the

idea that CTLA-4 restricts T-cell activation through a CD28-B7-

dependent mechanism (78). TCR transgenic T cells that over-

express CTLA-4 show diminished antigen-specific responses

in vitro; however, this decrease depends on the expression of

CD28 by the transgenic T cells. In the absence of CD28, the

CTLA-4-over-expressing T cells respond similarly to littermate

controls, suggesting that in this model, CTLA-4 functions pre-

dominantly by competing with CD28 for binding to B7-1 (78).

A related mechanism for CTLA-4-mediated inhibition was

suggested by the crystal structures of CTLA-4 and B7-1

described above. As already mentioned, CTLA-4 is a covalently

linked homodimer, and the two ligand-binding sites (consti-

tuted by the M99YPPPY104 motif also present in CD28) were

observed to lie distally to the homodimer interface (41–43).

Because B7-1 has been reported to homodimerize in solution

(50) as well as in the X-ray crystals (46), a molecular model

has been proposed where a single CTLA-4 dimer can bind two

B7-1 dimers that can in turn bind two more CTLA-4 dimers,

and so on, allowing the formation of an extended protein

lattice that could in theory assemble at the T-cell ⁄APC inter-face (the ‘immunological synapse’) and may be able to disrupt

its function (43). One study has attempted to assess the

importance of this putative CTLA-4 lattice for the inhibition

of Jurkat T-cell activation by generating a ‘monomeric’ CTLA-

4 by mutating Cys122 that mediates the interchain disulfide

bond and both N-glycosylation sites in human CTLA-4 (79).

This mutant molecule is able to traffic to the plasma

membrane, concentrates at the immunological synapse, binds

anti-CTLA-4 antibodies and B7-2-Ig, and inhibits IL-2 secre-

tion. Although compelling, the importance of the results is

limited by the observation that monomeric B7-2-Ig binding

can induce the dimerization of the mutant CTLA-4 and by the

fact that the cell lines expressing this mutant have significantly

lower amounts of CTLA-4 (79). Consequently, it is possible

that small numbers of mutant dimers (below the resolution of

the biochemical experiments) are able to assemble and traffic

to the immunological synapse, where they can bind ligand,

form an extended protein lattice, and inhibit T-cell activation

identically to the wildtype (WT) protein (79).

Inhibitory signaling

The idea that CTLA-4 delivers a negative signal in T cells

can be traced to early experiments showing that anti-CTLA-4

antibodies are capable of inhibiting the proliferation and IL-2

secretion of polyclonal T cells stimulated with a combination

of anti-CD3 and anti-CD28 antibodies in the absence of

B7-1 or B7-2 (13, 80). In addition, B7-1 can inhibit CD28-

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

70 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

deficient mouse T cells, and anti-CTLA-4 antibodies can

reverse this inhibition (81), similarly to their effect in the

accelerated rejection of cardiac allografts in CD28-deficient

mice due to CTLA-4 blockade (82). The fact that CTLA-4 can

restrict T-cell activation in the absence of CD28 suggests that

ligand competition cannot be its only mechanism of function.

Furthermore, the cytoplasmic tails of mouse and human CTLA-4

are 100% conserved at the amino acid level (83), suggesting

that they might bind proteins critical for their function.

The molecular basis for the CTLA-4-mediated inhibitory

signaling appears to lie in the association of the CTLA-4

cytoplasmic tail with the tyrosine phosphatase SH2-domain-

containing phosphatase 2 (SHP-2) ⁄ SH2-domain-containingtyrosine phosphatase SYP that can dephosphorylate TCR proxi-

mal signaling molecules like the tyrosine kinases fyn, lck, and

f chain-associated protein kinase of 70 kDa (ZAP-70) (84).

Subsequent studies have shown that CTLA-4 can associate with

the CD3f chain of the TCR in T cells, and the over-expression

of lck, CD3z and CTLA-4 in transfected human embryonic

kidney 293 cells improves the association between the last two

(85). SHP-2 also associates with the CD3f ⁄CTLA-4 complex inprimary T cells, and its over-expression in 293 cells eliminates

the binding of CTLA-4 and CD3f, as well as the tyrosine phos-

phorylation of CD3f (85). These data suggest a model where

TCR engagement leads to the recruitment of lck and the phos-

phorylation of CD3f, ZAP-70, and linker for activation of

T cells (LAT), followed by the binding of CTLA-4 and the asso-

ciated SHP-2, the subsequent dephosphorylation of signaling

molecules and termination of signaling (85). One caveat of

these experiments is that SHP-1, SHP-2, and the p85 subunit

of phosphatidylinositol 3-kinase (PI3K) can associate not only

with the phosphorylated Tyr201 in the cytoplasmic tail of

CTLA-4 but also with the tyrosine-phosphorylated tail of

CD28, making it difficult to envision a way in which the same

proteins can lead to completely opposite outcomes when

bound to CD28 versus CTLA-4 (60).

Another signaling molecule that associates with both CD28

and CTLA-4 is the serine ⁄ threonine protein phosphatase 2A(PP2A) (86, 87). This protein binds a conserved three-lysine

motif in the membrane-proximal region of the cytoplasmic

tail of CTLA-4 and represses its inhibitory function by a

currently unknown mechanism (87). As PP2A also binds to

CD28 and inhibits its function (86), it is possible to speculate

that perhaps one way in which CTLA-4 inhibits T-cell activa-

tion is by bringing the phosphatase in proximity to CD28,

where it can repress CD28-mediated costimulatory signaling.

More recent reports that CTLA-4 and CD28 have opposite

effects on the surface expression of lipid rafts (88) have added

yet another facet to the inhibitory signals potentially triggered

by CTLA-4 engagement. Because the rafts are enriched in

signaling molecules and are critical for early TCR signal

transduction (89), small changes in their expression can have

profound effects on T-cell activation. Furthermore, CTLA-4 is

found in lipid rafts at the plasma membrane, and this associa-

tion is required for its inhibitory function (90), presumably

because it negatively regulates the amount of phosphorylated

TCR present in these domains (91). The raft localization of

CTLA-4 depends on its cytoplasmic tail (90), suggesting that

the tailless mutants described earlier might have additional

defects, besides their inability to associate with signaling

components.

The importance of ligand sequestration in CTLA-4 function

has been recently questioned by the discovery of a mouse CTLA-

4 isoform that lacks the extracellular domain and thus functions

in a ligand-independent manner (38, 39). The reduced expres-

sion of this isoform correlates with increased susceptibility to a

mouse model of type I diabetes (38), and the protein itself

appears capable of generating negative signals in T cells by bind-

ing and dephosphorylating CD3f (39). Similarly, a CTLA-4

mutant that lacks B7-1 and B7-2 binding has been shown to

inhibit proliferation, cytokine production, and TCR-mediated

activation of the extracellular signal regulated kinase (ERK) in

T cells lacking endogenousCTLA-4 (92). Thismutant also delays

the rampant lymphoproliferation and early death of CTLA-

4-deficient mice, suggesting that some negative signals can

occur in the absence of ligand binding; however, the fact that

100% of the mice die at 14 weeks of age suggests that ligand

binding is critical for CTLA-4 function (92).

Reversal of the TCR-mediated STOP signal

A more recent model for CTLA-4 function has emerged from

in vitro migration and in vivo imaging experiments showing that

the antibody-mediated ligation of CTLA-4 increases the

motility of polyclonal CD4+ T cells and CTLA-4-positive TCR

transgenic T cells fail to stop in response to antigen, unlike

their CTLA-4-negative counterparts (93). These results suggest

that CTLA-4 might be able to reverse the STOP signal gener-

ated by TCR engagement and thus prevent T-cell activation.

One caveat of the in vivo experiments is that the CTLA-4-

positive population probably contains Treg cells that might

have higher mobility than effector T cells, especially because

they have been shown to inhibit stable contacts between

effector T cells and DCs (94, 95).

In an attempt to demonstrate that CTLA-4 expression by

itself is sufficient for the reversal of the TCR STOP signal, the

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 71

same group investigated the motile behavior of a mouse

hybridoma that was transfected with human CTLA-4 (96).

In these experiments, the CTLA-4-positive cells had shorter

contact times with APCs, formed smaller contact areas in the

immunological synapse, and had reduced Ca2+ influx and IL-2

secretion. The data also showed that CTLA-4 disrupted

ZAP-70 microcluster formation, suggesting that CTLA-4

might reduce the interactions between T cells and APCs as a

way of inhibiting T-cell activation (96).

Treg function

The idea that CTLA-4 is important for the function of Treg

cells has been a subject of controversy due to conflicting

results in different model systems. Early studies showed that

recombination-activating gene (RAG)-deficient mice reconsti-

tuted with a mixture of CTLA-4-deficient and normal bone

marrow cells were healthy, even though similar chimeras

containing only CTLA-4) ⁄ ) cells died 10 weeks after reconsti-

tution (97). These results were originally interpreted to mean

that a factor secreted by normal T cells could compensate for

the defect of CTLA-4-deficient cells; however, the discovery

that Treg cells can suppress effectors lacking CTLA-4 and anti-

CTLA-4 antibodies abrogate this suppression (64) provides an

alternative explanation for the results. The findings suggest

that CTLA-4-sufficient Treg cells can inhibit the proliferation

of CTLA-4-deficient effectors and that CTLA-4 is crucial for

this process (64). The importance of CTLA-4 for Treg func-

tion is reinforced by a study of a mouse model of colitis,

where Treg cells can prevent the intestinal inflammation

mediated by the transfer of CD4+CD45RBhigh T cells, and this

inhibition is abrogated by CTLA-4 blockade (65).

In contrast, the significance of CTLA-4 for the function of

Treg cells is brought into question by the observation that

CTLA-4-deficient mice have normal Treg development and

homeostasis and are capable of suppressing effector T cells

in vitro (98). The fact that the CTLA-4) ⁄ ) Treg cells produce

increased amounts of inhibitory cytokines like IL-10 and

transforming growth factor b (TGFb) suggests that in the

absence of CTLA-4, these cells can develop compensatory

suppressive mechanisms (98). Similarly, in a mouse model

of colitis, CTLA-4-deficient Treg cells are able to suppress

inflammation, predominantly via the secretion of IL-10 (99).

However, CTLA-4-sufficient Treg cells are able to suppress the

colitis induced by CTLA-4-deficient CD4+CD45RBhigh cells

from mice lacking B7-1 and B7-2, and anti-CTLA-4 antibodies

abrogate this inhibition, suggesting that CTLA-4 is critical for

the function of WT Treg cells (99).

Most recently, experiments using mice that have a condi-

tional deletion of CTLA-4 specifically in Treg cells have shown

that the molecule is required for Treg function (68). The

mutant mice spontaneously develop systemic lymphoprolifer-

ation, hyperproduction of IgE, and die of T-cell-mediated

autoimmunity 10 weeks after birth, while also exhibiting

some tumor immunity (68). The CTLA-4) ⁄ ) Treg cells

develop and survive normally but are highly proliferative and

comprise up to half of the CD4+ T cells in these mice, suggest-

ing that CTLA-4 has an important cell-autonomous inhibitory

function in Treg cells, similarly to its function in effector

T cells (68). The CTLA-4-deficient Treg cells are impaired in

their ability to suppress the proliferation of effector T cells

in vitro, potentially because they are unable to downregulate the

expression of B7-1 and B7-2 on APCs (68). The latter finding

provides a possible alternative mechanism for the way in

which Treg cells can inhibit effector T-cell proliferation

besides the induction of indoleamine 2,3-dioxygenase (IDO),

discussed below. While compelling, the study does not address

the possibility that CTLA-4 signaling in Treg cells might be also

important for the suppressive function of these cells.

Recent work from our laboratory has attempted to assess

the relative significance of CTLA-4 on effector versus Treg

cells through the use of antibodies that block the function of

the molecule selectively in the two types of cells (Peggs KS,

Quezada SA, Chambers CA, Korman AJ, JP Allison, unpub-

lished data). Compartment-specific blockade is achieved by

the generation of mice that express human CTLA-4 under the

control of the mouse promoter and the use of antibodies

specific to the mouse protein. The data show that in a mouse

model of melanoma, the blockade of CTLA-4 on the Treg

compartment alone is insufficient for tumor protection, while

the blockade of CTLA-4 on effector T cells is able to protect

about 40% of the mice; however, the simultaneous blockade

of CTLA-4 function on both effector and Treg cells is most

effective and cures 70% of the mice. These results suggest that

CTLA-4 has important inhibitory roles in both effector and

Treg cells, and successful immunotherapies should target both

its cell-autonomous inhibition of effector T-cell activation, as

well as its cell-extrinsic function via Treg cells.

CTLA-4-mediated ‘reverse’ signaling, IDO, and tryptophan

metabolism

The idea that IDO is important for peripheral tolerance comes

from experiments showing that treatment of pregnant mice

with the IDO inhibitor 1-methyltryptophan (1-MT) triggers

the rejection of allogeneic concept (100). It has been

suggested that a similar mechanism might be responsible for

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

72 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

the survival of pancreatic islet allografts after CTLA-4-Ig treat-

ment of diabetic mice. This hypothesis has been supported by

experiments showing that the tolerogenic effects of CTLA-4-

Ig administration are reversed by 1-MT treatment and that

B7-1 and B7-2 expression on DCs is required for the success

of the CTLA-4-Ig therapy (101). In this model, CTLA-4-Ig

appears to bind B7-1 or B7-2 on DCs and triggers a signaling

cascade involving signal transducer and activator of transcrip-

tion 1 (STAT1), p38 mitogen-activated protein kinase

(MAPK), and nuclear factor jB (NF-jB) that leads to inter-

feron-c (IFN-c) secretion by the DCs. IFN-c then elicits the

production of IDO, which degrades tryptophan and results in

the inhibition of T-cell proliferation and the induction of

apoptosis (101). The additional observation that IL-10 pre-

vents the IFN-c-induced downregulation of IDO in human

DCs (102) and the fact that Treg cells secrete IL-10 (103) sug-

gest the existence of a regulatory mechanism where the

engagement of B7-1 on DCs by CTLA-4 on Treg cells may

trigger IDO production by tolerogenic DC and the possibility

that this secretion may be further enhanced by Treg-generated

IL-10. This hypothesis has been supported by a study showing

that mouse Treg cells elicit IDO activity in DCs, and this pro-

cess is abrogated by anti-CTLA-4 antibodies, the absence of

B7-1 and B7-2 on the DCs, or the inability of the DCs to make

IFN-c (104).

The idea that CTLA-4 might be able to ‘reverse’ signal

into APCs via B7-1 is reinforced by experiments showing

that CTLA-4-Ig administration extends the long-term cardiac

allograft survival in mice, and the success of the treatment

depends on the expression of B7-1 by the allograft donor

but not by the recipient (105, 106). Although the initial

interpretation of these results consisted of a speculation that

CTLA-4-Ig is an incomplete blocking agent leaving some

unbound B7-1 in the WT donors that can engage CTLA-4

and inhibit the effector T cells mediating the graft rejection,

an alternative hypothesis is that CTLA-4-Ig binds B7-1 on

the DCs in the allograft, triggers IDO secretion, and induces

tolerance.

Another example of reverse signaling by CTLA-4 has been

reported by a study showing that B7-1 and B7-2 expression

on effector T cells is required for their suppression by Treg

cells in vitro and in vivo (107). The lentiviral expression of full

length B7-1 or B7-2 but not mutant molecules lacking the

transmembrane and cytoplasmic domains in B7-1) ⁄ )B7-2) ⁄ )

effector cells restores their susceptibility to Treg-mediated

inhibition, suggesting that signals originating from B7-1 or

B7-2 engagement on the effectors, presumably by CTLA-4 on

Treg cells, are responsible for this suppression (107).

CTLA-4 involvement in tolerance, autoimmunity, and

cancer progression

The early demonstration that CTLA-4-Ig potently inhibits

in vitro T-cell activation (108) suggested that it might be

useful as a therapeutic agent in autoimmune disease and

in transplantation. Subsequent studies have shown that

CTLA-4-Ig can block autoantibody production in a mouse

model of systemic lupus erythematosus and can extend

the life span of the affected animals even when treatment

is delayed until the most advanced stages of disease

(109). Furthermore, CTLA-4-Ig induces long-term survival

of xenogeneic pancreatic islet grafts in mice (110) and

suppresses T-cell-dependent antibody responses to multiple

antigens in vivo (111). Strikingly, CTLA-4-Ig treatment of

CTLA-4-deficient mice from birth prevents the fatal lym-

phoproliferation observed in these animals, even though

disease develops shortly after cessation of therapy (112).

These results are consistent with a study showing that the

genetic ablation of B7-1 and B7-2 abrogates the lympho-

proliferative disorder in CTLA-4-deficient mice and admin-

istration of anti-CD28 antibodies to these animals initiates

disease, despite having no effect on CTLA-4-sufficient

mice (113). These studies suggest that the main mecha-

nism by which CTLA-4-Ig inhibits T-cell immune

responses is by blocking CD28-mediated costimulatory

signals triggered by B7-1 and B7-2 engagement, although

it is also possible that it stimulates IDO production by

DCs in grafts, as discussed above. Most relevantly to

patients, CTLA-4-Ig has been approved for treatment of

moderate-to-severe rheumatoid arthritis (RA) by the United

States Food and Drug Administration (FDA). Randomized

clinical trials have shown that CTLA-4-Ig (abatacept) is

well-tolerated, reduces the signs and symptoms of RA,

slows the progression of joint damage, and improves the

physical function of adults that have had inadequate

responses to at least one other anti-rheumatic drug

(reviewed in 114).

The potent inhibitory effects that CTLA-4 exerts on T-cell

activation render it an attractive target for disease settings

where immune responses are limited or missing. One such

example is the case of tumors that do not provide suffi-

cient T-cell costimulatory signals and remain invisible to

the immune system. Early studies from our laboratory have

shown that treatment with anti-CTLA-4 antibodies leads to

the rejection of established colon carcinoma and fibrosar-

coma in mice, and this rejection creates immunity to

subsequent challenges with tumor cells, suggesting the

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 73

existence of long-term immunological memory (115).

Following these initial experiments, multiple studies have

shown that the anti-CTLA-4 antibody therapy is effective

for treatment of other cancers including prostatic carci-

noma (116), ovarian carcinoma (117), and B and T lym-

phomas (118, 119). However, CTLA-4 blockade is

unsuccessful as a single-agent therapy in cases involving

poorly immunogenic tumors like B16 melanoma (120)

and SM1 mammary carcinoma (121). For effective treat-

ment of established tumors in both cases, CTLA-4 blockade

has to be combined with an irradiated cellular vaccine

engineered to produce granulocyte-macrophage colony-

stimulating factor (GM-CSF), which is thought to enhance

antigen presentation (120, 121). A similar combination of

a cellular vaccine and anti-CTLA-4 antibody therapy is also

successful at preventing spontaneous prostate cancer devel-

opment in TRAMP (transgenic adenocarcinoma of mouse

prostate) mice (122).

A recent analysis of the effects of the combination therapy

of GM-CSF and anti-CTLA-4 antibodies on the effector and

regulatory T-cell compartments in the B16 mouse model of

melanoma has shown that the treatment acts via a cell-autono-

mous mechanism on both kinds of T cells, allowing them to

expand without depleting or permanently impairing the func-

tion of Treg cells (123). The major change in the lymphocytic

infiltration of the tumors induced by the combination therapy

consists of an increase in the ratio of T effector-to-Treg cells,

and this increase directly correlates with tumor rejection,

suggesting that Treg cells restrict the anti-tumor function of

both CD8+ and CD4+ effectors (123). The recent demonstra-

tion that this combination therapy synergizes with prophylac-

tic but not therapeutic Treg depletion and the finding that the

success of the treatment is limited by the inability of the effec-

tor cells to infiltrate the tumor, despite the generation of

strong systemic anti-tumor immune responses, has important

implications for the current clinical strategies used for treat-

ment of established tumors in patients (124).

In addition to the successful combination of anti-CTLA-4

antibodies and GM-CSF for treatment of poorly immunogenic

tumors, CTLA-4 blockade has been shown to synergize with

a number of therapies that enhance antigen presentation or

activate the innate immune system like peptide and DNA

vaccination, chemotherapy, radiation, Toll-like receptor (TLR)

engagement, and high-dose IL-2 (reviewed in 125). A

detailed summary of the available data from pre-clinical

studies and clinical trials of CTLA-4 as a monotherapy or

in combination with other treatments has been recently

presented and is beyond the scope of this review (125).

PD-1, PD-L1 (B7-H1), and PD-L2 (B7-DC)

PD-1 is a relatively new member of the extended

CD28 ⁄ CTLA-4-family of T-cell regulatory molecules that wasoriginally identified as a gene that was highly expressed by cell

lines undergoing programmed cell death (20). It has two

ligands, PD-L1 (B7-H1) and PD-L2 (B7-DC), which are

distantly related to B7-1 and B7-2 (21–23).

Protein structure

PD-1 is a 288 a.a. type I transmembrane glycoprotein that

contains a single IgV-like domain in its extracellular region

(20, 126). The cytoplasmic tail contains an immunoreceptor

tyrosine-based inhibitory motif (ITIM) (20) and an immuno-

receptor tyrosine-based switch motif (ITSM) (127). The

protein lacks the membrane-proximal cysteine that mediates

the inter-chain disulfide bond in the CD28 and CTLA-4

homodimers (126), and as a result, PD-1 is monomeric both

in solution and on the surface of transfected cells, as deter-

mined by FRET (128). Multiple splice isoforms have been

reported for human PD-1, some of which encode soluble

PD-1 molecules that are elevated in the peripheral blood and

synovial fluid of RA patients (129), and others may code for

ligand-independent variants (130). X-ray crystallography has

shown that the monomer is a two-layer b-sandwich that

forms a structure closely resembling the IgV-like fold of

CTLA-4, despite the fact that the sequence identity between

the two extracellular domains is only 20% (128).

PD-L1 is a 290 aa type I transmembrane glycoprotein that

contains a membrane-distal IgV-like and a membrane-

proximal IgC-like domain in its extracellular region (21, 22).

The cytoplasmic tail is 30 aa long and has no known signaling

motifs but is highly conserved between humans and mice

(22). The crystal structures of human PD-L1 alone and in

complex with mouse PD-1 have been solved (131). PD-L1

binds to PD-1 via its IgV-like domain in a 1:1 receptor ⁄ ligandstoichiometry (131). The binding interface is different from

that of the CTLA-4 ⁄B7-1 complexes and predominantlyinvolves the faces of the IgV-like domains, similarly to the

antigen-binding regions of antibodies and TCRs (131).

Mouse PD-L2 is a 248 aa type I transmembrane glycopro-

tein that contains a membrane-distal IgV-like and a mem-

brane-proximal IgC-like extracellular domain and a

cytoplasmic tail of only 4 aa (23). The human PD-L2 protein

is 26 aa longer, because it lacks the premature STOP codon

present in the mouse (23). The crystal structures of mouse

PD-L2 alone and in complex with mouse PD-1 have shown

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

74 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

that PD-L2 binds to PD-1 via its IgV-like domain in a 1:1

receptor ⁄ ligand stoichiometry (132). The ‘face-to-face’ bind-ing interface is similar to the one between PD-1 and PD-L1,

and the acute angle between PD-1 and its ligands may explain

the need for the relatively long ‘stalk’ regions separating the Ig

domains and the transmembrane segments relative to the

homologous structures in CTLA-4 and B7-1 (132).

Expression of PD-1, PD-L1, and PD-L2

PD-1 is not detectable on naive T cells but its expression goes

up in T cells, B cells, and myeloid cells after activation (22,

133). PD-1 expression is also upregulated by the treatment of

purified human T cells with the common c chain cytokines

IL-2, IL-7, IL-15, and IL-21 in the absence of TCR ligation

(134). PD-1 has been also found on the surface of TCR-stimu-

lated Treg cells, as well as in an intracellular compartment in

these cells prior to activation (135). In activated primary

T cells, PD-1 is present at the plasma membrane and in

Golgi-proximal intracellular vesicles that do not appear to be

recycling endosomes or lysosomes (136). The expression of

PD-1 is particularly high on the surface of functionally

exhausted CD8+ effector T cells in multiple persistent viral

infections in both mice and humans (137–140).

The two PD-1 ligands differ significantly in their expression

patterns. PD-L1 is broadly expressed on both hematopoietic

and non-hematopoietic cells. It has been detected on resting

mouse T cells, B cells, DCs, macrophages, natural killer (NK)

cells, and bone marrow-derived mast cells, and its levels

increase after activation (141, 142). PD-L1 is also expressed

by cells in the heart endothelium and the small intestine, by

syncyciotrophoblasts in the placenta, as well as by islet cells in

the pancreas of naive mice (143). In mouse macrophages,

PD-L1 expression is upregulated by LPS and IFN-c, as well as

the addition of T-helper type 1 (Th1) cells in a process that

requires TLR4 and STAT1 (144).

Unlike PD-L1, the expression of PD-L2 is restricted to

hematopoietic cells. Various cytokines can induce PD-L2

on DCs, macrophages, bone marrow-derived mast cells, and

B1 B cells (142, 143, 145, 146). In mouse macrophages,

PD-L2 expression is upregulated by IL-4 and the addition of

Th2 cells in a process that requires the IL-4 receptor a and

STAT6 (144).

Molecular mechanisms of PD-1 inhibitory signaling

The initial indications that PD-1 may engage an inhibitory

pathway upon binding to its ligands came from studies show-

ing that PD-L1-Ig and PD-L2-Ig can restrict the proliferation

and cytokine secretion of WT but not PD-1-deficient T cells

(22, 23). The fact that the inhibition is greatest at lower

amounts of antigen suggests that PD-1 might be more effec-

tive at attenuating weak TCR signals rather than strong ones

(22, 23). The formal demonstration that PD-1 might in fact

deliver inhibitory signals in the cells that express it came from

studies of B-cell lines transfected with chimeric molecules that

fused the extracellular region of the IgG Fc receptor type IIB

to the PD-1 cytoplasmic tail (147). The simultaneous ligation

of the B-cell receptor (BCR) and the fusion protein results in

growth inhibition, impairs Ca2+ mobilization, and reduces

tyrosine phosphorylation of several effector molecules (147).

Mutagenesis studies have shown that these inhibitory effects

do not require the N-terminal tyrosine in the cytoplasmic tail

but depend on the C-terminal tyrosine because upon

phosphorylation, that amino acid recruits the tyrosine phos-

phatase SHP-2 to the BCR and leads to the dephosphorylation

of proximal signaling molecules (147).

The second cytoplasmic tyrosine has subsequently been

identified as part of an ITSM, while the first one appears part

of an ITIM (127), and similarly to the results from B cells,

only mutations of the C-terminal tyrosine reverse the PD-1-

mediated inhibition of proliferation and cytokine secretion of

primary human T cells (148, 149). Both SHP-1 and SHP-2

co-immunoprecipitate with the cytoplasmic tail of PD-1 upon

stimulation of these cells with pervanadate, but only SHP-2

appears to associate with PD-1 in the absence of activation

(148). Despite this constitutive interaction, the inhibition of

T-cell functions requires the simultaneous engagement of the

TCR and PD-1 (148). The engagement of PD-1 by PD-L1-Ig

in conjunction with TCR ligation has been shown to decrease

ERK and protein kinase C h (PKCh) phosphorylation and IL-2

secretion from human T-cell blasts, as well as to inhibit the

phosphorylation of ZAP-70 and CD3f in Jurkat cells (150).

Sequencing of the proteins that are bound to the two

phosphotyrosines in the cytoplasmic tail of PD-1 from Jurkat

cell lysates has shown that lck can bind to the first but not the

second tyrosine, SHP-1 and the c-src tyrosine kinase (Csk) can

bind to the second but not the first, and SHP-2 can bind to

both (150). Biochemical studies using primary human T cells

have shown that PD-1 and CTLA-4 inhibit T-cell activation by

distinct mechanisms. While ligation of both PD-1 and CTLA-4

reduces the phosphorylation of the serine ⁄ threonine kinaseAkt in response to TCR ⁄CD28 stimulation, the inhibition viaCTLA-4 but not via PD-1 depends on the function of PP2A,

because it is sensitive to okadaic acid, while PD-1 inhibits Akt

phosphorylation by preventing the CD28-mediated activation

of PI3K (149). In general, the data from these studies suggest

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 75

that PD-1 can associate with SHP-2 and perhaps more weakly

with SHP-1, and upon ligand engagement, it can recruit these

phosphatases to the immunological synapse (136), where

they can interfere with T-cell activation by inhibiting PI3K

and the subsequent phosphorylation and activation of Akt.

The relatively recent discovery that PD-L1 can bind not only

PD-1 but also B7-1 in both mouse and human cells (151,

152) introduces an additional level of complexity in the

interaction between the costimulatory molecules and the

inhibitory pathways that they control. The affinity of the asso-

ciation of PD-L1 and B7-1 is higher than that between B7-1

and CD28 and lower than that between B7-1 and CTLA-4 and

between PD-L1 and PD-1 (151, 152). The fact that upon

engagement by the other molecule, both PD-L1 and B7-1

appear able to deliver inhibitory signals in the T cells that

express them (151) blurs the current definitions of receptors

and ligands and demands the characterization of the signaling

pathways mediated by these two proteins.

PD-1 involvement in tolerance, autoimmunity, cancerprogression, and viral immunity

The importance of PD-1 for the prevention of autoimmunity

was initially indicated by the phenotype of PD-1-deficient

mice. C57 ⁄ Bl6 mice lacking PD-1 develop lupus-like arthritisand glomerulonephritis with predominant deposition of IgG3

antibodies (153), while PD-1-deficient BALB ⁄ c mice developautoimmune-dilated cardiomyopathy that leads to congestive

heart failure and premature death (154). The heart disease is

caused by the accumulation of antibodies specific for cardiac

troponin I that causes the chronic stimulation of Ca2+ influx

in cardiomyocytes (155). PD-1-deficient animals on the

non-obese diabetic (NOD) background develop autoimmune

disease with accelerated onset and increased frequency

because of strong Th1 polarization of the T cells infiltrating

the affected islets (156). Similarly, PD-L1-deficient animals

display enhanced T-cell responses and are more susceptible to

experimental autoimmune encephalomyelitis (EAE) and

experimental autoimmune hepatitis (157, 158). In contrast,

one study of PD-L2-deficient mice has shown that these mice

have diminished Th1 and CTL responses (159), while another

line of PD-L2) ⁄ ) mice displays increased T-cell activation after

in vivo immunization and is unable to generate tolerance in

response to an oral antigen (160). These results suggest that

PD-1 and both its ligands are important for the establishment

and ⁄or maintenance of peripheral tolerance.The role of PD-1 and its ligands in regulating autoimmune

disease has been investigated in multiple mouse models. In

EAE induced by immunization with myelin oligodendrocyte

glycoprotein (MOG), anti-PD-1, and anti-PD-L2 but not anti-

PD-L1 antibodies cause more severe disease as a result of

increased frequencies of IFN-c-producing T cells, higher lev-

els of serum anti-MOG antibodies, and elevated delayed-type

hypersensitivity responses (161). The effects of anti-PD-1,

anti-PD-L1, and anti-PD-L2 antibody blockade on EAE suscep-

tibility and disease progression differ among mouse strains,

suggesting that all three molecules can limit autoimmune

responses, and their relative significance for disease manifesta-

tion may be influenced by differences in the cytokine milieu

under the various experimental conditions (162, 163).

The importance of PD-1 and PD-L1 but not PD-L2 for the

regulation of autoimmune diabetes in NOD mice has been

emphasized by multiple reports. Studies using blocking anti-

bodies to the molecules have shown that PD-1 and PD-L1 are

required for prevention of diabetes, presumably because

PD-L1 but not PD-L2 is expressed on inflamed islets of NOD

mice (164). Furthermore, the expression of PD-L1 by paren-

chymal cells rather than hematopoietic cells in the pancreas is

required for protection against immunopathology after trans-

plantation of syngeneic islets in diabetic recipients (165). The

PD-L1-mediated inhibition of autoreactive T cells appears to

involve PD-1, because PD-1-deficient ovalbumin-specific OT-I

CD8+ T cells cause diabetes in mice expressing ovalbumin in

the pancreas, while WT OT-I T cells do not (166). Finally, the

importance of the PD-1 ⁄ PD-L1 inhibitory pathway is empha-sized by a report showing that insulin-coupled APCs can be

successfully used as a therapy after diabetes onset, and anti-

body-mediated blockade of PD-1 and PD-L1 reverses these

therapeutic effects in a process that seems to involve the

re-activation of pathogenic effector cells without affecting

Treg activity (167). In contrast, the constitutive ectopic

expression of PD-L1 on pancreatic islets leads to spontaneous

diabetes that can be reversed by low doses of anti-CD3

antibodies, suggesting that the disease is T-cell mediated

(168). Further experiments are needed to determine the

reasons for this discrepancy.

PD-1 and its ligands have also been shown to be involved in

fetomaternal tolerance (169), the regulation of alloimmune

responses (170–172), graft-versus-host disease (173), asthma

(174, 175), and experimental allergic conjunctivitis (176). In

addition, the early observation that multiple tumor lines

express PD-L1 (177, 178) has prompted the investigation of

the potential role that PD-1 and its ligands might have in the

regulation of tumor immunity. In humans, high PD-L1

expression by tumor cells correlates with poor prognosis in

ovarian cancer (179), renal cancer (180–182), urothelial

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

76 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

carcinoma (183, 184), pancreatic cancer (185), non-small-

cell lung cancer (186), and gastric carcinoma (187). The

mechanism by which tumor-associated PD-L1 might protect

tumors from T-cell-mediated immune destruction has been

alternatively suggested to involve the induction of T-cell death

(177) or by making the tumor cells resistant to T-cell-medi-

ated apoptosis (188). In contrast, PD-L2 has been suggested

to promote tumor immunity via a PD-1-independent mecha-

nism (189), which may potentially involve the PD-L2 engage-

ment of a currently unknown costimulatory receptor on

T cells that might also bind PD-L1 (21, 159, 190, 191).

PD-1 or PD-L1 blockade has been shown to mediate some

tumor protection in mice (178, 192–194), and a humanized

anti-PD-1 antibody, CT-011, is currently in phase I clinical

trials for treatment of patients with advanced hematologic

malignancies (195). The antibody appears to be well tolerated

and has had clinical benefit in 33% of patients with one

complete remission (195). These results suggest that targeting

of the PD-1 ⁄ PD-L1 pathway is a promising new strategy fortumor therapy that might be able to synergize with existing

anti-cancer treatments, including CTLA-4 blockade, especially

in light of the distinct inhibitory mechanisms employed by

these molecules in T cells.

The relatively recent discovery that PD-1 is upregulated on

functionally exhausted CD8+ effector T cells during persistent

viral infections renders it an attractive target for therapeutic

vaccination in chronic infections. Treatment of mice that

harbor chronic lymphocytic choriomeningitis virus (LCMV)

infection with anti-PD-L1 antibodies restores the ability of the

exhausted CD8+ T cells to proliferate, secrete cytokines, kill

infected targets, and decrease viral load in the animals (137).

Similarly, PD-1 is expressed at high levels on non-functional T

cells during human immunodeficiency virus (HIV) infections,

and anti-PD-1 or anti-PD-L1 antibodies appear able to restore

their proliferation and effector functions, at least in vitro (138–

140). Comparable findings have been observed during chronic

infections with hepatitis B and C viruses (196–200), herpes

simplex virus (201), Helicobacter pylori (202), and Mycobacterium

tuberculosis (203). These reports suggest that PD-1 and its ligands

are potentially attractive therapeutic targets for the treatment of

persistent viral and bacterial infections in human patients.

B7-H3 and B7x

B7-H3 and B7x (B7-H4) are distantly related to B7-1 and

B7-2 and appear to bind to a receptor on activated but not

naive T cells (24, 26–28). B7-H3 has been shown to

alternatively activate (24) or inhibit (25) T-cell function,

presumably via distinct receptors, while B7x is a negative reg-

ulator of T-cell activation (26–28). The costimulatory recep-

tor for B7-H3 has been recently identified as the triggering

receptor expressed on myeloid cells (TREM)-like transcript 2

(TLT-2 ⁄TREML2), which is present constitutively on CD8+

T cells and on activated CD4+ T cells (204). The inhibitory

receptors that B7-H3 and B7x bind are currently not known.

Protein structure and expression of B7-H3 and B7x

B7-H3 is a 316 a.a. type I transmembrane glycoprotein that

contains an IgV-like and an IgC-like domain in its extracellular

region and a cytoplasmic tail of 45 a.a. (24, 205, 206). In

humans, there is also an alternatively spliced 4Ig-B7-H3

variant of 534 a.a. containing two IgV-like and two IgC-like

domains that appear to have arisen via gene duplication

(205–207). This isoform may protect tumor cells from NK

cell-mediated lysis (208). B7x is 282 a.a. glycoprotein that

contains a single IgV-like and a single IgC-like domain in its

extracellular region and has a very short cytoplasmic tail

(26–28) or may be attached to the membrane via a glyco-

phosphatidylinositol (GPI) linkage (27).

Unlike B7-1 and B7-2, B7-H3 and B7x are ubiquitously

expressed in most organs, although the exact cell types express-

ing the molecule in these tissues are currently unclear. B7-H3

mRNA has been found in the prostate, testes, uterus, small and

large intestine, placenta, heart, lung, liver, kidney, pancreas, as

well as in the spleen, thymus, and lymph nodes (24, 205). The

B7-H3 protein is found on monocytes and DCs and is induced

on T cells, B cells, and NK cells (24, 207). B7x mRNA has been

found in the lung, liver, prostate, testes, pancreas, placenta,

uterus, brain, kidney, as well as in the thymus, and spleen

(26–28, 209). The B7x protein has been detected on in vitro

activated T cells, B cells, peritoneal macrophages, monocytes,

and DCs (26–28). B7x has been also suggested to be a marker

for a novel suppressive macrophage population in human ovar-

ian carcinoma (210), and its expression may be induced by

Treg-mediated IL-10 secretion (211), although this proposed

mechanism is somewhat controversial (212).

Analysis of B7-H3-deficient mice has shown that the protein

has a role in limiting Th1-mediated immune responses,

because the animals show enhanced Th1-mediated airway

hypersensitivity, they accumulate higher concentrations of

anti-DNA antibodies and are more susceptible to EAE (205).

In contrast, B7x-deficient mice appear to have mildly elevated

Th1 responses against Leishmania major leading to decreased

parasite burden but otherwise have unaltered hypersensitivity

reactions in the airway and the skin and mount normal T-cell

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 77

responses against LCMV (213). These results suggest that

while B7x has a negative regulatory function in vivo, the

existence of multiple inhibitory mechanisms for the control of

T-cell activation may limit the absolute significance of each

individual inhibitory pathway.

B7-H3 and B7x in cancer progression

Consistent with the inhibitory effects of B7x on T-cell activa-

tion (26–28) and despite the conflicting descriptions of

B7-H3 as a costimulatory or an inhibitory ligand for T-cell

function (24, 25, 214), multiple studies have found that B7x

and B7-H3 expression in tumors correlates with poor progno-

sis. In non-small-cell lung cancer, high B7x and B7-H3 levels

correlate with decreased number of tumor-infiltrating lym-

phocytes (TILs) and increased lymph node metastasis (215).

In prostatic carcinoma, high levels of B7x and B7-H3 are

associated with increased risk of cancer recurrence, tumor

progression after surgery, and cancer-related death (216,

217). In clear cell renal cell carcinoma (RCC), B7x and B7-H3

are expressed by the tumor or the tumor vasculature, and this

expression is significantly correlated with increased risk of

death from cancer (218, 219). Similarly, serum-soluble B7x is

elevated in RCC patients and correlates with advanced stage of

disease (220). B7x is also highly expressed in ductal and lobu-

lar breast cancer (221), and soluble B7x is a potential biomar-

ker for ovarian carcinoma (222). These findings suggest that

while the molecular details of the inhibitory pathways that

B7x and B7-H3 engage in T cells are currently not known, the

proteins can be useful prognostic tools in human cancer

and potential targets for antibody-mediated tumor immuno-

therapy.

BTLA, HVEM, and CD160

Protein structure

BTLA is type I membrane glycoprotein of the IgSF that was

first identified as an expressed sequence tag (EST) in

Th1-polarized TCR transgenic T cells (29) and independently

as an EST expressed after activation of double-positive thymo-

cytes (223). Crystal structures of human and mouse BTLA

reveal a single b-sandwich Ig domain containing two b-sheets

(224, 225). BTLA is related to IgV receptors such as CD28 and

CTLA-4, but the structural data highlight notable differences.

On one of the b-sheets BTLA lacks a strand present in CTLA-4,

and thus belongs in the Ig intermediate (IgI) rather than the

IgV subset of the IgSF (224). BTLA binds to the transmem-

brane receptor HVEM (30, 226) via a region on the IgI

domain that is distinct from the MYPPPY motif-containing

surface through which CD28 and CTLA-4 bind to their B7-1

and B7-2 ligands (42, 43, 224).

Expression of BTLA, HVEM, and related ligands

HVEM is a member of the tumor necrosis factor receptor super-

family (TNFRSF), and the BTLA–HVEM interaction is the first

demonstration of crosstalk between IgSF and TNF receptors

(227). As its name would suggest, HVEM binding to herpes

simplex virus glycoprotein D (HSV gD) is required for viral

entry into host cells (228), and the HVEM binding site on

gD is a structural mimic of that on BTLA (224). Prior to the

discovery of its interaction with BTLA, HVEM was also known

as a costimulatory receptor which binds to the TNF ligands

homologous to lymphotoxins, exhibits inducible expression,

and competes with HSV glycoprotein D for HVEM, a receptor

expressed by T lymphocytes (LIGHT) and lymphotoxin a

(LTa) (229), although the phenotype of HVEM-deficient mice

is consistent with a role for HVEM in the negative regulation of

T-cell activation (230). A complex of purified recombinant

HVEM, BTLA, and LIGHT has been isolated chromatographi-

cally (224), suggesting that the binding of LIGHT and BTLA to

HVEM is not mutually exclusive (30, 231), but the functional

relevance of such a ternary interaction is unknown.

BTLA is constitutively expressed on most hematopoietic

cells, including T cells, B cells, and DCs, with the highest

levels present on B cells (29, 223, 232). The relatively low

levels of BTLA on naive CD4+ T cells increase substantially

after activation (232). Similarly to BTLA, HVEM is expressed

widely throughout lymphoid compartments, with constitutive

levels on naive T cells that decrease after activation (30, 233).

The broad expression patterns of BTLA and HVEM have made

it hard to define the particular cell–cell interactions for which

BTLA–HVEM binding is relevant. The production of inducible

cell type-specific BTLA- and HVEM-deficient mice will likely

be necessary to gain insight into this problem.

Initial evidence that BTLA negatively regulates T-cell activa-

tion came from in vitro assays using BTLA-deficient mouse

T cells and antibody engagement of BTLA during anti-CD3

T-cell stimulation (29, 223). The addition of anti-BTLA

antibodies to naive primary T cells has been shown to inhibit

both division and IL-2 production (234). BTLA ligation on

human T cells has similar consequences (226). CD4+ T cells

were shown to be more sensitive than CD8+ T cells to inhibi-

tion by anti-BTLA antibody engagement, which may be due

to their faster induction of higher levels of surface BTLA after

activation (234). Finally, CHO cells expressing HVEM and

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

78 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

MHC class II were shown to inhibit the antigen-specific divi-

sion of WT but not BTLA-deficient T cells (30).

BTLA inhibitory signaling

The approximately 100 a.a. cytoplasmic tail of mouse BTLA

contains three conserved tyrosine-based signaling motifs

(235). The most membrane proximal motif (YDND) can

interact with the adapter protein Grb-2 when phosphorylated

and also binds to the p85 subunit of PI3K. The other two

motifs, surrounding tyrosine residues 257 and 282 in human

BTLA, are ITIM sequences that mediate BTLA binding to the

protein tyrosine phosphatases SHP-1 and SHP-2 when phos-

phorylated (236). Thus, TCR crosslinking, which induces

phosphorylation of BTLA cytoplasmic tyrosines and phospha-

tase recruitment (29, 223), could result in the dephosphoryla-

tion of signaling intermediates and attenuation of T-cell

activation. The functional relevance of this mechanism, which

has also been proposed for CTLA-4 and PD-1 (84, 85, 147), is

in question (237), and, as mentioned above, it is unclear how

specific phosphatase binding only to phosphotyrosines on

inhibitory receptors is regulated (60).

A growing body of work using mouse models of disease

supports the conclusion that BTLA and HVEM have a negative

influence on T-cell mediated inflammatory responses. Early

data showed that BTLA-deficient mice are somewhat more

sensitive to MOG-peptide-induced EAE (29), and, like

PD-1-deficient mice, take longer to resolve acute eosinophilic

airway inflammation induced by protein sensitization and

subsequent intranasal challenge (238). In contrast to PD-1,

BTLA expression was induced on partially MHC-mismatched

cardiac allografts, and BTLA-deficient recipients, but not

PD-1-deficient recipients, rejected normally tolerated MHC

class II mismatched grafts (239). Graft rejection was also seen

in recipients lacking the BTLA ligand HVEM and in those that

received anti-BTLA antibodies. Interestingly, in a more recent

study, HVEM has been found on stimulated Tregs, and HVEM

expression on Tregs and BTLA expression on effector T cells

seems to be important for optimal in vitro Treg suppression as

well as cardiac allograft acceptance (240). Finally, administra-

tion of anti-BTLA antibodies was reported to synergize with

CTLA-4-Ig to prolong survival of islet allografts in mice

(241, 242).

BTLA, HVEM, and autoimmunity

Two recent studies indicate that negative regulation of T-cell

responses by BTLA and HVEM is important in the prevention

of autoimmunity. Careful examination of older (approxi-

mately 12 months) BTLA-deficient mice revealed develop-

ment of a spontaneous autoimmune hepatitis-like disease,

which was characterized by infiltrates similar to those seen in

human autoimmune hepatitis and increased serum levels of

liver transaminases (243). The contribution of HVEM expres-

sion on various cell types to intestinal inflammation was eval-

uated using a model of autoimmune colitis involving the

transfer of CD4+CD45RB+ T cells into RAG-deficient recipi-

ents (244). Bone marrow chimeric mice in which HVEM was

deficient only on radioresistant compartments showed rapid

acceleration of colitis, suggesting that HVEM expression on

tissue-specific APCs may be involved in attenuating intestinal

inflammation caused by BTLA-expressing T cells.

Recent data with BTLA-deficient mice indicate that the

greater proliferation of total T-cell populations seen in the

absence of BTLA is not a cell-intrinsic defect but is rather due

to greater numbers of antigen-experienced (CD44hi) CD8+

T cells in the lymphoid compartments of these mice (245),

which explains the lack of an effect of BTLA deficiency on

proliferation of purified CD4+ T cells (234). Interestingly,

the same study (245) points to a T-cell-intrinsic role for BTLA

in regulating the generation of CD8+CD44hiCD62Lhi central

memory phenotype cells, rather than inhibiting initial prolif-

eration per se. The authors proposed that BTLA may limit cell

survival during the contraction phase of memory generation,

and this idea is supported by recent data from a mouse model

of asthma (246). Prolonged airway inflammation mediated by

adoptively transferred BTLA-deficient T cells in response to

inhaled antigen was correlated with enhanced survival of

the BTLA-deficient T cells in the lung when compared with

WT cells. Although increased apoptosis was not detected in

primary T-cell cultures inhibited by anti-BTLA antibodies

(234), it is likely that this type of assay does not realistically

model the later stages of T-cell activation where BTLA may

function, and regulation of apoptotic pathways during the

contraction phase remains an attractive mechanism of BTLA

regulation of effector T cells.

Human CD160 was recently recognized to be another IgSF

member that functions as an inhibitory HVEM ligand

expressed on T cells (31). Like BTLA, CD160 contains a single

Ig domain and binds to the same region of HVEM (31).

Crosslinking antibodies to CD160 (in conjunction with anti-

CD3 ⁄CD28 stimulation) was shown to inhibit both prolifera-tion and cytokine production of purified CD4+ T cells, as well

as the phosphorylation of CD3f and other proteins. In contrast

to BTLA, CD160 is GPI-anchored and so the direct recruitment

of intermediates that interfere with TCR signaling seems

unlikely. Further study of this newly appreciated inhibitory

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 79

pathway is likely to yield insight into the biology of TNFRSF–

IgSF interactions and the negative regulation of T-cell activa-

tion in general.

LAG-3

LAG-3 (CD223) is a type I membrane glycoprotein of the IgSF

first cloned from an NK cell line and shown to be expressed in

activated human NK and T cells (32) as well as in activated

B cells (247). It has genomic and structural similarity to CD4,

with four extracellular Ig-like domains and conserved struc-

tural motifs in between these domains. Both human and

mouse LAG-3 bind to MHC class II more strongly than CD4

(248–250), and this interaction has been reported to modu-

late cytokine production by APCs (251).

The initial reports of LAG-3 knockout mice demonstrated a

deficiency in NK killing (252), but no T-cell phenotypes were

found. On naive CD4+ and CD8+ T cells, LAG-3 is present at

low levels that are upregulated after antigen-specific activation

(250). Several reports have suggested that mouse LAG-3

negatively regulates T-cell activation (33, 253, 254) and that

signaling mediated by the cytoplasmic domain is involved in

this process (255). Human LAG-3 is also postulated to inhibit

T-cell activation (256). LAG-3 was shown to be upregulated

on adoptively transferred CD4+ T cells undergoing an abortive

response to antigen expressed on pulmonary epithelium

(257), but whether the non-functional LAG-3-positive CD4+

cells correspond to bona fide induced CD4+Foxp3+ Treg cells is

debatable. The relevance of MHC class II binding to the func-

tion of LAG-3 in inhibition of T-cell activation and differentia-

tion is unclear.

Studies of adoptively transferred antigen-specific CD8+ cells

in mice which express hemagglutinin (HA) as a self or tissue-

specific tumor antigen provide the most direct evidence that

LAG-3 inhibits T-cell function in a cell-intrinsic manner (34).

Injection of anti-LAG-3 antibodies or transfer of LAG-3-

deficient T cells results in greater tissue-specific accumulation

and IFN-c production by the antigen-specific CD8+ T cells,

suggesting that anti-LAG-3 antibody administration blocks

LAG-3 function. In addition, the cells display enhanced cyto-

lytic function after vaccination with HA when this is com-

bined with anti-LAG-3 injection. Finally, in mice that develop

autochthonous prostate cancer and express HA specifically in

the prostate (proHA · TRAMP) (258), this type of vaccina-tion results in reduced tumor grade. The effects on tumor pro-

gression, although modest, suggest that LAG-3 may be a

potential immunotherapeutic target, and this possibility is

supported by clinical reports of LAG-3 expression on TILs.

Examination of TILs from RCC has shown LAG-3 expression

on CD8+ but not CD4+ T cells in all eight patients examined

(259). In addition, a study of 94 Hodgkin’s lymphoma

patients demonstrated a correlation between Foxp3) TILs that

expressed high levels of LAG-3 and decreased Epstein-Barr

virus-specific IFN-c production (260). These data, along with

the studies in the proHA · TRAMP mice (34), are consistentwith a role for LAG-3 in the regulation of tumor-reactive

CD8+ T cells. Additional examination of the cell-intrinsic

and ⁄or -extrinsic mechanisms by which LAG-3 inhibits T-cellfunction and the contextual dependence of this on MHC class

II ligand binding is warranted and may yield insight into

potential immunotherapeutic strategies that involve LAG-3

blockade.

Concluding remarks

The importance of CD28-mediated costimulation and CTLA-

4-mediated inhibition for the activation of T cells has been

known for 30 years. Since their discovery in the late 1980s, it

has become apparent that other receptors of the IgSF inhibit

T-cell activation. Mechanistic descriptions of T-cell inhibition

commonly invoke the recruitment of activated tyrosine phos-

phatases to the TCR; however, the molecular details of the

regulation of specific binding to inhibitory receptors are

unclear and require further studies. Despite this, efforts have

long been made to manipulate the signaling pathways down-

stream of T-cell inhibitory receptors for treatment of both

cancer and autoimmune disease. The fact that CTLA-4-Ig is

currently an FDA-approved therapy for RA and that anti-

CTLA-4 antibodies are used for treatment of melanoma, pros-

tate cancer, colon cancer, and RCC either as a monotherapy or

in combination with other treatments emphasize the clinical

importance of these molecules. The more recent discovery of

additional members of the extended CD28 ⁄CTLA-4 familyand their ligands offers potentially novel targets for diagnosis

or therapeutic intervention in cancer, autoimmunity, and

chronic infections. As the ligands for CD28 and CTLA-4 are

predominantly expressed by professional APCs, the main

impact of immune modulators of this pathway is probably

limited to the initial priming phase of autoimmune disease

and immune responses against cancer. In contrast, the expres-

sion of PD-L1, B7-H3, and B7x on cells in the peripheral

tissues suggests that these molecules may act as gatekeepers

for lymphocytic infiltration of these organs. This putative

function as part of a protective mechanism against autoim-

mune destruction has probably made them subversion targets

for tumors and pathogens to achieve immune escape and offer

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

80 � 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009

additional opportunities for therapeutic intervention. The

expression of BTLA and HVEM throughout hematopoietic

compartments necessitates cell type-specific analysis of their

function, but the available data suggest that this pathway is

important in the regulation of T-cell responses in multiple

tissues. The negative regulation of T-cell activation involves

numerous receptors and ligands expressed in separate cellular

compartments at different times in the immune response. This

observation suggests that there may be synergy between

blockade of the various inhibitory pathways and encourages

the experimental testing of combinatorial strategies for

treatment of disease.

References

1. Mueller DL, Jenkins MK, Schwartz RH.Clonal expansion versus functional clonal

inactivation: a costimulatory signallingpathway determines the outcome of T cell

antigen receptor occupancy. Annu RevImmunol 1989;7:445–480.

2. Jenkins MK, Taylor PS, Norton SD, UrdahlKB. CD28 delivers a costimulatory signal

involved in antigen-specific IL-2 productionby human T cells. J Immunol

1991;147:2461–2466.3. Harding FA, McArthur JG, Gross JA, Raulet

DH, Allison JP. CD28-mediated signalling

co-stimulates murine T cells and preventsinduction of anergy in T-cell clones. Nature

1992;356:607–609.4. Lindsten T, June CH, Ledbetter JA, Stella G,

Thompson CB. Regulation of lymphokinemessenger RNA stability by a surface-

mediated T cell activation pathway. Science1989;244:339–343.

5. Boise LH, et al. CD28 costimulation canpromote T cell survival by enhancing the

expression of Bcl-XL. Immunity 1995;3:87–98.

6. Wulfing C, Sumen C, Sjaastad MD, Wu LC,Dustin ML, Davis MM. Costimulation and

endogenous MHC ligands contribute toT cell recognition. Nat Immunol 2002;

3:42–47.7. Rudd CE, Raab M. Independent CD28 sig-

naling via VAV and SLP-76: a model for intrans costimulation. Immunol Rev

2003;192:32–41.8. Viola A, Schroeder S, Sakakibara Y,

Lanzavecchia A. T lymphocyte costimulationmediated by reorganization of membrane

microdomains. Science 1999;283:680–682.9. Brunet JF, et al. A new member of the

immunoglobulin superfamily–CTLA-4.Nature 1987;328:267–270.

10. Linsley PS, Greene JL, Brady W, Bajorath J,Ledbetter JA, Peach R. Human B7-1 (CD80)

and B7-2 (CD86) bind with similar aviditiesbut distinct kinetics to CD28 and CTLA-4

receptors. Immunity 1994;1:793–801.

11. Collins AV, et al. The interaction propertiesof costimulatory molecules revisited. Immu-

nity 2002;17:201–210.12. Walunas TL, Bakker CY, Bluestone JA.

CTLA-4 ligation blocks CD28-dependent

T cell activation [erratum appears in J ExpMed 1996;184:301]. J Exp Med

1996;183:2541–2550.13. Krummel MF, Allison JP. CTLA-4 engage-

ment inhibits IL-2 accumulation and cellcycle progression upon activation of resting

T cells. J Exp Med 1996;183:2533–2540.14. Waterhouse P, et al. Lymphoproliferative

disorders with early lethality in micedeficient in Ctla-4. Science 1995;270:

985–988.15. Tivol EA, Borriello F, Schweitzer AN, Lynch

WP, Bluestone JA, Sharpe AH. Loss of

CTLA-4 leads to massive lymphoprolifera-tion and fatal multiorgan tissue destruction,

revealing a critical negative regulatory roleof CTLA-4. Immunity 1995;3:541–547.

16. Chambers CA, Cado D, Truong T, Allison JP.Thymocyte development is normal in

CTLA-4-deficient mice. Proc Natl Acad SciUSA 1997;94:9296–9301.

17. Hutloff A, et al. ICOS is an inducible T-cellco-stimulator structurally and functionally

related to CD28. Nature 1999;397:263–266.

18. Swallow MM, Wallin JJ, Sha WC. B7h, anovel costimulatory homolog of B7.1 and

B7.2, is induced by TNFalpha. Immunity1999;11:423–432.

19. Yoshinaga SK, et al. T-cell co-stimulationthrough B7RP-1 and ICOS. Nature

1999;402:827–832.20. Ishida Y, Agata Y, Shibahara K, Honjo T.

Induced expression of PD-1, a novel mem-ber of the immunoglobulin gene superfam-

ily, upon programmed cell death. EMBO J1992;11:3887–3895.

21. Dong H, Zhu G, Tamada K, Chen L. B7-H1,a third member of the B7 family, co-stimu-

lates T-cell proliferation and interleukin-10secretion. Nat Med 1999;5:1365–1369.

22. Freeman GJ, et al. Engagement of the PD-1immunoinhibitory receptor by a novel B7

family member leads to negative regulationof lymphocyte activation. J Exp Med

2000;192:1027–1034.

23. Latchman Y, et al. PD-L2 is a second ligandfor PD-1 and inhibits T cell activation. Nat

Immunol 2001;2:261–268.24. Chapoval AI, et al. B7-H3: a costimulatory

molecule for T cell activation and

IFN-gamma production. Nat Immunol2001;2:269–274.

25. Suh WK, et al. The B7 family memberB7-H3 preferentially down-regulates T

helper type 1-mediated immune responses.Nat Immunol 2003;4:899–906.

26. Zang X, Loke P, Kim J, Murphy K, Waitz R,Allison JP. B7x: a widely expressed B7

family member that inhibits T cell activa-tion. Proc Natl Acad Sci USA 2003;100:

10388–10392.27. Prasad DV, Richards S, Mai XM, Dong C.

B7S1, a novel B7 family member that nega-

tively regulates T cell activation. Immunity2003;18:863–873.

28. Sica GL, et al. B7-H4, a molecule of the B7family, negatively regulates T cell immu-

nity. Immunity 2003;18:849–861.29. Watanabe N, et al. BTLA is a lymphocyte

inhibitory receptor with similarities toCTLA-4 and PD-1. Nat Immunol

2003;4:670–679.30. Sedy JR, et al. B and T lymphocyte attenua-

tor regulates T cell activation through inter-action with herpesvirus entry mediator. Nat

Immunol 2005;6:90–98.31. Cai G, Anumanthan A, Brown JA, Greenfield

EA, Zhu B, Freeman GJ. CD160 inhibits acti-vation of human CD4+ T cells through

interaction with herpesvirus entry mediator.Nat Immunol 2008;9:176–185.

32. Triebel F, et al. LAG-3, a novel lymphocyteactivation gene closely related to CD4. J Exp

Med 1990;171:1393–1405.33. Workman CJ, Cauley LS, Kim IJ, Blackman

MA, Woodland DL, Vignali DA. Lympho-cyte activation gene-3 (CD223) regulates

the size of the expanding T cell populationfollowing antigen activation in vivo.

J Immunol 2004;172:5450–5455.34. Grosso JF, et al. LAG-3 regulates CD8+ T

cell accumulation and effector function inmurine self- and tumor-tolerance systems.

J Clin Invest 2007;117:3383–3392.35. Linsley PS, et al. Binding stoichiometry of

the cytotoxic T lymphocyte-associated

molecule-4 (CTLA-4). A disulfide-linkedhomodimer binds two CD86 molecules.

J Biol Chem 1995;270:15417–15424.36. Magistrelli G, et al. A soluble form of

CTLA-4 generated by alternative splicing is

Pentcheva-Hoang et al Æ T-cell regulation in cancer and autoimmunity

� 2009 John Wiley & Sons A/S • Immunological Reviews 229/2009 81

expressed by nonstimulated human T cells.Eur J Immunol 1999;29:3596–3602.

37. Oaks MK, Hallett KM, Penwell RT, StauberEC, Warren SJ, Tector AJ. A native soluble

form of CTLA-4. Cell Immunol2000;201:144–153.

38. Ueda H, et al. Association of the T-cellregulatory gene CTLA4 with susceptibility

to autoimmune disease. Nature2003;423:506–511.

39. Vijayakrishnan L, et al. An autoimmunedisease-associated CTLA-4 splice variant

lacking the B7 binding domain signalsnegatively in T cells. Immunity

2004;20:563–575.

40. Metzler WJ, et al. Solution structure ofhuman CTLA-4 and delineation of a

CD80 ⁄ CD86 binding site conserved inCD28. Nat Struct Biol 1997;4:527–531.

41. Ostrov DA, Shi W, Schwartz JC, Almo SC,Nathenson SG. Structure of murine CTLA-4

and its role in modulating T cell responsive-ness. Science 2000;290:816–819.

42. Schwartz JC, Zhang X, Fedorov AA,Nathenson SG, Almo SC. Structural basis for

co-stimulation by the human CTLA-4 ⁄ B7-2complex. Nature 2001;410:604–608.

43. Stamper CC, et al. Crystal structure of theB7-1 ⁄ CTLA-4 complex that inhibits humanimmune responses. Nature 2001;410:608–611.

44. Freeman GJ, et al. Structure, expression,and T cell costimulatory activity of the

murine homologue of the human Blymphocyte activation antigen B7. J Exp

Med 1991;174:625–631.45. Borriello F, Freeman GJ, Edelhoff S, Disteche

CM, Nadler LM, Sharpe AH. Characteriza-tion of the murine B7-1 genomic locus

reveals an additional exon encoding analternative cytoplasmic domain and a chro-

mosomal location of chromosome 16, bandB5. J Immunol 1994;153:5038–5048.

46. Ikemizu S, et al. Structure and dimerizationof a soluble form of B7-1. Immunity

2000;12:51–60.47. Freeman GJ, et al. Murine B7-2, an

alternative CTLA4 counter-receptor that

costimulates T cell proliferation andinterleukin 2 production. J Exp Med

1993;178:2185–2192.48. Freeman GJ, et al. Cloning of B7-2: a CTLA-

4 counter-receptor that costi