Báo cáo khoa học: The dipeptidyl peptidase IV family in cancer and cell biology pot

Although DPIV, FAP, DP8 and DP9 possess similar enzyme activity and are structurally conserved, they have varying patterns of expression and localization Keywords cancer; dipeptidyl pept

The dipeptidyl peptidase IV family in cancer and cell

biology

Denise M T Yu1, Tsun-Wen Yao1, Sumaiya Chowdhury1, Naveed A Nadvi1,2, Brenna Osborne1,

W Bret Church2, Geoffrey W McCaughan1and Mark D Gorrell1

1 A.W Morrow Gastroenterology and Liver Centre, Royal Prince Alfred Hospital, Centenary Institute and Sydney Medical School, University

of Sydney, Australia

2 Pharmaceutical Chemistry, Faculty of Pharmacy, University of Sydney, Australia

Introduction

Proteases are heavily involved in specialized biological

functions and thus often play important roles in

patho-genesis The dipeptidyl peptidase IV (DPIV⁄ CD26)

gene family has attracted ongoing pharmaceutical

interest in the areas of metabolic disorders and cancer

Four of its members – DPIV (EC 3.4.14.5),

fibro-blast activation protein (FAP), DP8 and DP9 – are

characterized by a rare enzyme activity, namely

hydro-lysis of a prolyl bond two residues from the

N-termi-nus DPIV is the best-studied member of the family

and has a variety of roles in metabolism, immunity,

endocrinology and cancer biology DPIV is a new and successful type 2 diabetes therapeutic target, and FAP

is under investigation as a cancer target Although the exact functions of the newer members, DP8 and DP9, are yet to be elucidated, thus far they have been found

to have interesting biological properties and, like DPIV and FAP, are likely to be multifunctional and employ both enzymatic and extra-enzymatic modes of action Although DPIV, FAP, DP8 and DP9 possess similar enzyme activity and are structurally conserved, they have varying patterns of expression and localization

Keywords

cancer; dipeptidyl peptidase; distribution;

enzyme; extracellular matrix; immune

function; liver fibrosis; structure

Correspondence

M D Gorrell, Molecular Hepatology,

Centenary Institute, Locked Bag No 6,

Newtown, NSW 2042, Australia

Fax: +61 2 95656101

Tel: +61 2 95656156

E-mail: m.gorrell@centenary.usyd.edu.au

(Received 23 October 2009, revised 25

November 2009, accepted 30 November

2009)

doi:10.1111/j.1742-4658.2009.07526.x

Of the 600+ known proteases identified to date in mammals, a significant percentage is involved or implicated in pathogenic and cancer processes The dipeptidyl peptidase IV (DPIV) gene family, comprising four enzyme members [DPIV (EC 3.4.14.5), fibroblast activation protein, DP8 and DP9] and two nonenzyme members [DP6 (DPL1) and DP10 (DPL2)], are inter-esting in this regard because of their multiple diverse functions, varying patterns of distribution⁄ localization and subtle, but significant, differences

in structure⁄ substrate recognition In addition, their engagement in cell bio-logical processes involves both enzymatic and nonenzymatic capabilities This article examines, in detail, our current understanding of the biological involvement of this unique enzyme family and their overall potential as therapeutic targets

Abbreviations

bFGF, basic fibroblast growth factor; DP, dipeptidyl peptidase; ECM, extracellular matrix; FAP, fibroblast activation protein; gko, gene knockout; HSC, hepatic stellate cell; IL, interleukin; IP, interferon-c-inducible protein; IRAK-1, IL-1 receptor-associated serine ⁄ threonine kinase I; ITAC, interferon-inducible T-cell chemo-attractant; MAPK, mitogen-activated protein kinase; MMP, matrix metalloproteinase; NK, natural killer; NPY, neuropeptide Y; SDF-1, stromal cell-derived factor-1; Th, T helper; uPAR, urokinase plasminogen activator receptor.

and are therefore likely to play diverse roles Some of

the functional significance placed on DPIV research

over decades is now being credited to the whole family,

particularly the newer members, DP8 and DP9, and so

modern selective DPIV pharmaceutical inhibitor design

has placed value on the structure and function of the

other DPs The development and application of DPIV

inhibitors as successful type 2 diabetes therapeutics has

occurred over a relatively short period of time, in the

span of about a decade Thus, careful consideration of

the biological properties of each DP is required in the

application of DP inhibitors to treat other disorders

The DPIV gene family

The DPIV gene family is a subgroup of the prolyl

oligopeptidase family of enzymes (Table 1), which are

specialized in the cleavage of prolyl bonds As most

peptide hormones and neuropeptides comprise one or

more proline residues, this family of enzymes is useful

for processing and degrading such peptides [1,2]

DPIV and FAP (also known as seprase) are closely

related cell-surface enzymes, with DPIV-like enzyme

activity FAP is also a narrow-specificity endopeptidase

[3–5] FAP endopeptidase activity includes a type I

col-lagen-specific gelatinase [6,7] activity and seems

restricted to Gly-Pro-containing substrates [4], which is

interesting because the DP activity of FAP is greater

on H-Ala-Pro than on H-Gly-Pro-derived artificial

substrates [3]

DP8 and DP9 are dimers with DPIV-like enzyme

activity [8–11] Although DP8 and DP9 are very

clo-sely related to each other and share similar distribution

patterns [12], there are some differences in their cell

biological effects (discussed later), perhaps related to their cytoplasmic localization, and so may play differ-ent roles [13]

The nonenzymatic members of the family – DP6 (DPL1⁄ DPX) and DP10 (DPL2 ⁄ DPY) – are modula-tors of voltage-gated potassium channels in neurons and are primarily expressed in brain [14–18] Although structurally similar to DPIV [19,20], they lack the cata-lytic serine and other residues necessary for enzyme activity [17,21] Thus, they are likely to exert effects via protein–protein interactions [20], similarly to the enzyme DP members that also have extra-enzymatic abilities (Fig 1)

Distribution of DPs in normal and pathogenic tissue

DPIV distribution DPIV is expressed by epithelial cells of a large number

of organs, including liver, gut and kidney; by endothe-lial capillaries; by acinar cells of mucous and salivary glands and pancreas; by the uterus; and by immune organs such as thymus, spleen and lymph node [22–25] (Fig 2) Our recent study using the DPIV selective inhibitor, sitagliptin, on wild-type and DPIV gene knockout (gko) mouse tissue homogenates has con-firmed the presence of DPIV enzyme activity in a large number of organs [12]

DPIV is a potential marker for a number of cancers, but with variability among different types of cancers DPIV is upregulated in a number of aggressive types

of T-cell malignancies, such as T-lymphoblastic lymphomas, T-acute lymphoblastic leukaemias and

Cell type?

ECM environment?

Pro

DPi

DP ligand

Mechanism?

Enzymatic, extraenzymatic? Effect of inhibitor?

Effect?

Protumorigenic or Anti-tumorigenic?

Regulatory processes?

Fig 1 Dynamics involved in DP biology The DPs have both enzy-matic and extra-enzyenzy-matic properties, and the outcomes of their action may lead to anti-tumorigenic or tumorigenic effects, depend-ing on factors such as cell type, regulation and microenvironment.

Table 1 Criteria and subclans of the prolyl oligopeptidase (POP;

EC 3.4.21.26) family of enzymes (MEROPS –the Peptidase

Data-base; merops.sanger.ac.uk; [166]).

Criteria for POP family members

DNA sequence homology to prolyl endopeptidase (PEP ⁄ POP)

Subclans of the POP family

S9A Prolyl endopeptidase (PEP ⁄ POP)

S9B Dipeptidyl peptidase IV (DPIV)

S9C Acylaminoacyl peptidase

S9D Glutamyl endopeptidase (plant)

S9B subclan - DPIV gene family

DPIV

Fibroblast activation protein (FAP)

DP8

DP9

Nonenzyme DPIV-related POP family members

DP6 (DPX, DPL1)

DP10 (DPY, DPL2)

T-anaplastic large cell lymphomas [26], and is a

mar-ker of poor prognosis for T-large granular lymphocyte

lymphoproliferative disorder [27] DPIV is also

upregu-lated in lung adenocarcinoma [28], oesophageal

adeno-carcinoma [29], thyroid adeno-carcinoma [30–32], prostate

cancer [33] and B-cell chronic lymphocytic leukaemia

[34, 35], and dysregulated in liver cirrhosis [36]

How-ever, DPIV expression is progressively downregulated

in endometrial adenocarcinoma [37] Thus, some care

needs to be taken in the use of DPIV as a target for

different cancers Further understanding of the

biologi-cal anti-invasive effect of DPIV in vitro could be of

importance in the control of certain carcinomas

FAP distribution

The unique tissue distribution of FAP has made it a

potential marker and target for certain epithelial

can-cers FAP is generally absent from normal adult tissues

[38] In silico electronic northern blot analysis shows

that normal tissues generally lack FAP mRNA

expres-sion, with the exception of endometrium [39] Also,

typing of cancers by electronic northern blotting

reveals predominant FAP signals in tumour types

marked by desmoplasia [39] In vivo, FAP is generally

absent in normal adult epithelial, mesenchymal, neural

and lymphoid cells [40], or in nonmalignant tumours,

such as fibroadenomas, and in nonproliferating

fibro-blasts [38] Nevertheless, a soluble form of FAP has

been isolated from bovine serum [41] and from human

plasma [42,43] FAP expression is highly induced dur-ing inflammation, for example, within fibroblast-like synovial cells in rheumatoid arthritis and osteoarthritis [44,45] FAP is also significantly upregulated at sites of tissue remodelling, such as the resorbing tadpole tail [46], during scar formation in wound healing [38] and

at sites of tissue remodelling during mouse embryogen-esis, including somites and perichondrial mesenchyme from cartilage primordia [47]

In addition, FAP is preferentially expressed by acti-vated, but not by resting, hepatic stellate cells (HSCs)

of cirrhotic liver, but not in normal human liver [6] FAP-immunopositive cells are present in the early stages of liver injury, and the expression level of FAP mRNA correlates with the histological severity of fibrosis in chronic liver diseases [48] FAP co-localizes with fibronectin and collagen in cirrhotic liver, with collagen fibrils present alongside activated HSCs [49,50] FAP is expressed only by myofibroblasts and activated HSCs at sites of tissue remodelling, which is the portal–parenchymal interface of cirrhotic liver [6] FAP is upregulated in most human cancers [51] FAP is highly expressed by fibroblasts at the remodel-ling interface in human idiopathic pulmonary fibrosis [52] Interestingly, FAP is highly upregulated on reac-tive stromal fibroblasts of over 90% of human epithe-lial tumours, but not in benign tumours [38] As stromal fibroblasts are a common feature of epithelial cancers, including breast, colorectal, ovarian and lung carcinomas, FAP is a potential therapeutic target for

Fig 2 Overview of the distribution of the DPs in normal and pathogenic tissue and cell types.

multiple human epithelial cancers [38] Previous

FAP-specific cancer therapeutic investigations have included

antibody targeting [53–55], FAP DNA vaccination

[56], immunotherapy [57] and inhibitor therapies [58]

It is not yet clear whether inhibiting FAP enzyme

activity alone can lead to anti-tumorigenic effects

(Fig 1), although the use of FAP enzyme-inactive

mutants in tumour growth studies have supported this

concept [59] Recent alternative approaches that utilize

or localize FAP enzyme activity have shown potential,

including a FAP-activated promelittin protoxin that

reduces tumour growth in mice [60], and a

FAP-trig-gered photodynamic molecular beacon for the

detec-tion and treatment of epithelial cancers [61]

DP8 and DP9 distribution

The distribution of DP8 and DP9 has been studied by

our group in some depth Ubiquitous DP8 and DP9

mRNA expression was previously shown by a Master

RNA dot-blot [62] and a multiple-tissue northern blot

[8,9] More recently, we confirmed ubiquitous

expres-sion using enzyme assays in the presence of a DP8⁄ 9

inhibitor, in situ hybridization and

immunohistochem-istry, particularly in immune cells, epithelia, brain,

testis and muscle [12] DP8 and DP9 enzyme levels are

predominant over DPIV in mouse testis and brain

In situhybridization and immunohistochemistry

analy-ses on baboon and human tissues detected DP8 and

DP9 in lymphocytes and epithelial cells in the

gastroin-testinal tract, skin, lymph node, spleen, liver and lung,

as well as in pancreatic acinar cells, adrenal gland,

spermatogonia and spermatids of testis, and in

Pur-kinje cells and in the granular layer of cerebellum The

results of other studies are in agreement with these

findings [63–66] The significance of the three DP8

splice variants is not known [8,67]; however, one of the

splice variants is upregulated in human adult testis

compared with fetal testis [67] There are two known

DP9 transcripts – a ubiquitously expressed transcript

of 863 amino acids [9,68,69] and a larger 971-amino

acid transcript in muscle, spleen and peripheral blood

leukocytes [9] The larger form appears to be expressed

in tumours [9]

There is some early evidence suggesting that DP8

and DP9 expression may be associated with disease

pathogenesis The DP9 mRNA levels are elevated in

testicular tumours [12] and DP9 has also been shown

to be upregulated in DNA arrays comparing

nontu-mour and normal liver tissue [70] In diseased liver,

DP8 and DP9 mRNA has been detected in infiltrating

lymphocytes [12] DP8⁄ 9 expression is higher in

inflamed lung, probably also associated with activated

lymphocytes [63] These distribution patterns, as well

as DP inhibitor studies (see a later section), support possible roles for DP8 and DP9 in inflammation and

in the immune system

Biological functions of DPIV, FAP and DP8/9

The DPs have interesting roles in cell biology and in pathogenic processes DPIV and FAP have been iden-tified both as potential cancer markers and as prote-ases with anti-tumorigenic properties [71] Their mechanisms of action generally fall into two catego-ries, namely enzymatic and extra-enzymatic (protein– protein interactions) (Fig 1) The enzymatic roles relate to the substrates of the DPs, whereas the extra-enzymatic roles relate to their ligand-binding properties

DPIV

Enzymatic activity of DPIV and its role in type 2 diabetes

The ubiquitously expressed enzymatic action of DPIV covers a large range of physiological substrates involved in varied functions DPIV is best known for its enzymatic ability to inactivate the incretin hormones glucagon-like peptide-1 and glucose insuli-notropic peptide In the treatment of type 2 diabetes, DPIV inhibitors extend incretin action, resulting in improved glucose metabolism via prolonged insulin release and trophic beta cell effects [72,73] We have discussed this therapeutic application of DPIV inhibi-tors elsewhere [74]

Other physiological substrates of DPIV include neu-ropeptide Y (NPY), substance P and the chemokine stromal cell-derived factor-1 (SDF-1⁄ CXCL12) NPY

is involved in the control of appetite, energy homeosta-sis and blood pressure [75] DPIV-truncated NPY is unable to bind to its Y1 receptor, instead binding to its Y2 and Y5 receptors, which promote angiogenesis [75] and inflammation [76] Substance P, involved in pain perception and nociception, is inactivated by DPIV enzyme activity [77] DPIV enzyme activity is effective on a number of chemokines in vitro (Table 2)

DPIV in cell biology

A number of DPIV-binding proteins have been identi-fied, including adenosine deaminase [78,79], CD45 (protein tyrosine phosphatase) [80], caveolin-1 [81], CARMA1 [82], fibronectin III [83,84], plasminogen 2

[85], Na+-H+exchanger isoform 3 [86] and glypican-3

[87] Adenosine deaminase, CD45, caveolin-1 and

CARMA1 are involved in the costimulation of T cells

by DPIV, whereas fibronectin III, plasminogen 2 and

glypican-3 may have roles in cancer biology Binding

of DPIV to fibronectin III is important for metastasis

and colonization of breast cancer cells, implicating the

role of DPIV in tumour progression [88] In the

human prostate tumour 1-LN cell line, direct binding

of plasminogen 2 with cell-surface DPIV induces a

sig-nal transduction cascade that produces a rapid increase

in the calcium ion concentration, subsequently

result-ing in the expression of matrix metalloproteinase 9

(MMP-9), which enhances the invasiveness of cells

[89]

Overexpression of DPIV in cell lines results in

inter-esting cell-behavioural effects Our studies have found

that 293T epithelial cells transfected with DPIV

exhibited less cell migration on extracellular matrix

(ECM)-coated plastic, and exhibited increases in both

spontaneous and induced apoptosis [50] Wesley et al

[90–93] found that DPIV overexpression in a number

of cell lines (melanocytes, nonsmall cell lung, prostate

and neuroblastoma cancer lines) caused anti-tumori-genic effects, such as inhibition of in vitro cell migra-tion and cell growth, increased apoptosis and inhibition of anchorage-independent growth Other studies have confirmed similar findings in melanoma cells and in ovarian carcinoma cells [94,95] In vivo, nude mice injected with DPIV-overexpressing cancer cells showed inhibition of tumour progression compared with control cancer cells [91,93]

A number of signalling pathways have been associ-ated with DPIV ECM interactions, including the basic fibroblast growth factor (bFGF) pathway, which is involved in cell proliferation, migration, cell survival, wound healing, angiogenesis and tumour progression Overexpression of DPIV in prostate cancer cells blocks the nuclear localization of bFGF, lowers bFGF levels and subsequently affects downstream components of the bFGF pathway [mitogen-activated protein kinase (MAPK)-extracellular signal-regulated kinase 1⁄ 2 and urokinase plasminogen activator] These changes are accompanied by the induction of apoptosis, cell cycle arrest and the inhibition of in vitro cell migration [92] Sato et al [96] have shown that DPIV mediates cell

Table 2 Potential downstream effects of DPIV on chemokines.

neutrophilic granulocytes

[167] MIG c (CXCL9) Expressed by stimulated monocytes,

macrophages and endothelial cells.

Acts on Th1 d lymphocytes

ablates chemotaxis of activated Th1 lymphocytes

[168]

IP10e(CXCL10) Expressed by neutrophils, hepatocytes,

endothelial cells, keratinocytes Acts on CD4 + T cells, haematopoetic progenitor cells, lymphocytes

less chemoattraction CD4+T cells inhibits haematopoetic progenitor proliferation

ablates chemotaxis of activated Th1 lymphocytes

[169,170]

ITAC f (CXCL11) Expressed by leucocytes, fibroblasts,

endothelial cells, pancreas, liver astrocytes Acts on activated T cells

loss of Ca 2+ flux via CXCR3 less chemotaxis of activated Th1 and NK cells

[168,171]

SDF-1 g (CXCL12) Acts on lymphocytes, dendritic cells,

haematopoetic cells

less tumour growth less lymphocyte chemotaxis ablates antiviral activity more chemoattraction of monocytes regulation of haematopoietic stem cell recruitment

[131,172,173]

LD78b (CCL3 ⁄ L1) Expressed by T cells, B cells and monocytes more chemoattraction of monocytes [174] Eotaxin (CCL11) Acts on eosinophils, basophils, Th2 lymphocytes less chemotaxis of eosinophils

less binding ⁄ signalling via CCR3

[170,175,176] MDC h (CCL22) Acts on NK cells, T-cell subsets, monocytes,

dendritic cells

ablates chemotactic activity for lymphocytes less Ca 2+ mobilization via CCR4

[168,177,178]

a

Gro b, growth regulated protein b;bGCP, granulocyte chemotactic protein;cMIG, monokine-induced interferon-c;dTh, T helper;eIP10, interferon-c-inducible protein 10; f ITAC, interferon-inducible T-cell chemo-attractant; g SDF-1, stromal cell-derived factor 1; h MDC, macro-phage-derived chemokine; i NK, natural killer.

adhesion to the ECM via p38 MAPK-dependent

phos-phorylation of integrin b1 Inhibition of DPIV

expres-sion using small interfering RNA in the T-anaplastic

large cell lymphoma cell line Karpas 299 causes

reduc-tion of adhesion to fibronectin and collagen I Also,

DPIV-depleted Karpas 299 cells have reduced

tumori-genicity compared with control Karpas 299 cells when

injected into severe combined immunodeficient mice

[96] This finding contrasts that observed by Wesley

et al., possibly reflecting cell-type differences In the

Burkitt B-cell lymphoma line, Jiyoye, DPIV

overex-pression results in increased phosphorylation of p38

MAPK but no accompanying increase in cell adhesion

[96,97] DPIV overexpression in neuroblastoma lines

leads to induction of apoptosis mediated by caspase

activation, and downregulation of the chemokine

SDF-1 and its receptor CXCR4 SDF-1

downregula-tion, in turn, leads to induced cell migration and to

decreased levels of phospho-Akt and active MMP-9

[93] Other molecules that are upregulated by DPIV

overexpression include p21, CD44 [50,91],

topoisomer-ase IIa [97] and the known cell-adhesion molecules

E-cadherin and tissue inhibitor of matrix

metallopro-teinases [50,98]

DPIV in immune function

Also known as CD26 T-cell differentiation marker,

DPIV plays vital roles in immunology and

autoimmu-nity [99] It is expressed at detectable levels by some

resting T cells, but the cell-surface expression increases

by 5 to 10-fold following stimulation with antigen,

anti-CD3 plus interleukin (IL)-2 or mitogens such as

phytohaemagglutinin [25,100–105] The strongest

lym-phocytic CD26 expression is found on cells

co-express-ing high densities of other activation markers, such

as CD25, CD71, CD45RO and CD29 [106–108] The

CD26brightCD4+ population of T cells is the

CD45RO+CD29+ memory⁄ helper subset, which

responds to recall antigens, induces B-cell IgG

synthe-sis and activates cytotoxic T cells [102,107,109] In

addition, CD26brightCD4+ memory T cells

preferen-tially undergo transendothelial migration [108,110]

CD26 has a costimulatory role in T-cell activation

and proliferation CD26 is mainly expressed on T

helper (Th) 1 cells and its expression is induced by

stimuli favouring the development of Th1 responses

[111–113] Through its expression on T cells, CD26 is

able to provide a costimulatory signal in lipid rafts

[80,114,115] to augment the T-cell response to foreign

antigens [109,116,117] (Fig 3) Crosslinking of CD26

with antibody increases the recruitment of CD26 and

CD45 to these rafts [80] and induces T-cell activation

[113,118] The signal transduced by CD26 overlaps with the T-cell receptor⁄ CD3 pathway, increasing tyro-sine phosphorylation of p56lck, p59fyn, ZAP-70, phos-pholipase C-c, MAPK and c-Cb1 in that pathway [119,120] CD26 on activated memory T cells binds to caveolin-1 on antigen-presenting cells at the immuno-logical synapse for T cell⁄ antigen-presenting cell inter-actions [81] Stimulation of CD26 also causes IL-1 receptor-associated serine⁄ threonine kinase I (IRAK-1) and Toll-interacting protein to disengage from caveo-lin-1 IRAK is phosphorylated in this process, which leads to the upregulation of CD86 expression [121] The interaction between CD26 and caveolin-1 also leads to the recruitment of lipid rafts, which are impor-tant for modulating signal transduction Additional recruitment of a complex including CARMA1 in lipid rafts leads to events downstream of the T-cell receptor complex to activate the nuclear factor-jB pathway [82]

Thus, the role of CD26 in lymphocyte activation is probably attributable to its extra-enzymatic ligand-binding properties, as further evidenced by in vitro studies Stimulation with a combination of anti-CD3 and anti-CD26 IgGs induces more IL-2 production by

Fig 3 Features of the effects of DP upon T-cell activation ⁄ func-tion Cell-surface CD26 ⁄ DPIV interacts with adenosine deaminase, CARMA-1 and caveolin-1 Stimulation of CD26 causes IRAK-1 and Toll-interacting protein to disengage from caveolin-1, resulting in the phosphorylation of IRAK-1 and the upregulation of CD86 expression Interaction with caveolin-1 also results in the recruit-ment of lipid rafts, leading to events downstream that activate the nuclear factor-jB pathway The roles of intracellular DP8 and DP9

in lymphocyte proliferation are likely to be enzymatic, in contrast to the role of DPIV, and they may modulate signalling molecules.

CD26-transfected Jurkat cells than by CD26

enzyme-negative mutant transfected cells [116], suggesting that

the proteolytic activity of CD26 is not essential in the

costimulatory function of CD26 Moreover, soluble

CD26 enhances the proliferation of activated

periph-eral blood lymphocytes by decreasing strong responses

and increasing weak responses of T cells [122,123]

Studies involving CD26 enzyme-deletion mutants have

shown that the costimulatory activity of CD26

involves the ligand-binding domain [124,125] A recent

in vivostudy using a DPIV-selective inhibitor in DPIV

gko mice has further shown independence of immune

functions of CD26 from its enzyme activity [126]

CD26 has a role in the development of effector

func-tions by CD8+T cells [109,127,128] and is also

upreg-ulated in activated B cells [34,129,130] CD26

overexpression in a B-cell line enhances p38

phosphor-ylation, suggesting that as in T cells, CD26 in B cells

could be involved in the MAPK p38 signalling

pathway to activate signaling molecules such as

extra-cellular signal-regulated kinase, p56lck, p59fyn, ZAP-70,

c-Cbl and phospholipase C-c [97,119]

CD26 has nine chemokine substrates in vitro:

eotaxin (CCL11), macrophage-derived chemokine

(CCL22), growth-regulated protein b (CXCL2),

LD78b (CCL3⁄ L1), granulocyte chemotactic protein 2

(CXCL6), monokine-induced interferon-c (CXCL9),

interferon-c-inducible protein (IP-10⁄ CXCL10),

inter-feron-inducible T-cell chemo-attractant (ITAC⁄ CXCL

11) and SDF-1 (CXCL12) (Table 2) Of these, SDF-1

is the only verified chemokine substrate in vivo [131]

By enzyme cleavage, CD26 reduces the inflammatory

properties of these chemokines by altering or

abrogat-ing the ability to trigger a signal via the cognate

recep-tors, and in some cases the cleaved chemokine also

blocks binding by the corresponding intact chemokine

molecule

FAP

The endopeptidase activity of soluble FAP cleaves

a2-antiplasmin [42,43, 132], which is involved in blood

clotting The gelatinase activity of FAP is likely to be

associated with its expression in ECM remodelling

One putative FAP ligand, urokinase plasminogen

acti-vator receptor (uPAR), has been reported in LOX

malignant melanoma cells [133,134] Because the

uPAR ligand, urokinase plasminogen activator, is able

to convert plasminogen to plasmin, which degrades

fibrin and certain ECM proteins, formation of the

heterogeneous proteolytic complex between FAP and

uPAR might enhance the invasive and metastatic

abili-ties of tumour cells [133,134]

FAP expression is associated with normal or exces-sive wound healing, and with malignant tumour growth and chronic inflammation [38], including human liver cirrhosis [6] All of these processes involve ECM degradation Proteolytic degradation of ECM components facilitates angiogenesis and⁄ or tumour cell migration Many proteases, including secreted and cell-surface metalloproteinases, and some serine peptidases, have roles in these processes The gelatinase activity of FAP, specifically collagenolytic activity towards type I collagen fragments, suggests that FAP could in this way contribute to ECM degradation [6,7,135]

Like DPIV, overexpression of FAP frequently leads to anti-tumorigenic effects In overexpression studies using the HEK293T epithelial cell line, FAP had decreased adhesion on collagen I, fibronectin and Matrigel, and exhibited increases in both sponta-neous and induced apoptosis [50] Overexpression of FAP in melanoma cells leads to suppression of the malignant phenotype in cancer cells, specifically cell cycle arrest at the G0⁄ G1 phase, increased suscepti-bility to stress-induced apoptosis and restoration of contact inhibition [136] Overexpression of FAP abrogates tumorigenicity in nude mice; surprisingly, enzymatically inactive FAP further abrogates tumorigenicity [136]

In contrast to the above described anti-tumorigenic effects, FAP-overexpressing HEK293 cells, when xeno-grafted into severe combined immunodeficient mice, result in a significantly greater incidence of tumour development and growth compared with controls, including an enzyme-inactive mutant [59,137] FAP overexpression in the hepatic stellate cell line, LX-2, enhances adhesion and migration on collagen and fibronectin on ECM substrata in vitro [50] These data suggest that FAP has a critical role in liver fibrosis, probably by influencing the functions of activated hepatic stellate cells and⁄ or by interacting with the ECM FAP expression is stimulated by transforming growth factor-b and retinoic acid, which also stimulate HSC and myofibroblasts [134] Moreover, transforming growth factor-b1 is a major stimulus for epithelial–mesenchymal transition, a contributor of myofibroblasts in chronic liver injury [138]

DP8 and DP9

DP8 and DP9 have no confirmed physiological substrates, but do have the ability to cleave the DPIV substrates glucagon-like peptide-1, glucagon-like peptide-2, NPY and peptide YY in vitro [11,65] In addition, DP8 can cleave four chemokines [139] No ligands of DP8 and DP9 have been reported

DP9 overexpression studies in the HEK293T

epithe-lial cell line have revealed roles for DP9 in cell

adhe-sion, in in vitro wound healing, in cell migration, and

in proliferation and apoptosis, and roles for DP8 have

been found in wound healing, in cell migration and in

apoptosis enhancement [13] DP9 overexpression

impaired cell behaviour on a wider range of ECM

components than for DP8 Despite their close sequence

relatedness, DP8 and DP9 exert some differences in

their cellular effects Therefore, these two proteins are

likely to have different functions and ligands

The mechanism of action of intracellular DP8 and

DP9 remains unknown Many cytoplasmic events are

involved in cell–ECM interactions, causing changes to

cell behaviour, so it is difficult to predict which events

are influenced by cytoplasmic DP8 and DP9 The

observed decreases in DP9-overexpressing cells of the

ECM-interacting molecules discoidin domain receptor

1, a kinase activated by collagen binding, and the

MMP inhibitor, tissue inhibitor of matrix

metallopro-teinase-2, suggest possible DP9 target pathways [13]

The discovery of DP8 and DP9 as reactive oxygen

species-responsive molecules may provide an indication

of a cytoplasmic function [140] DP8 and DP9 might

be mammalian H2O2-sensing proteins that are

impor-tant in intracellular processes where H2O2 is regulated,

such as phosphorylation, signaling pathways,

apopto-sis, cancer and immune function [141–144] While

DP9, but not DP8, overexpression is associated with

spontaneous apoptosis, both elevate induced apoptosis

Apoptosis is an important process in tissue

remodel-ling, including recovery from liver injury [145] At a

biochemical level, apoptosis is a complex cellular event

involving the coordinated action of proteins, several

different peptidases, nucleases and

membrane-associ-ated ion channels and phospholipid translocases As

DP8 and DP9 activities are dependent on the redox

state of their cysteines, the redox states of DP8 and

DP9 may be a molecular switch in the regulation of

apoptotic pathways [140] In addition, as cytoplasmic

DPIV can be phosphorylated [146], DP8 and DP9 may

also be phosphorylated in signalling pathways, and, in

fact, phosphorylation sites in DP8 and DP9 are

identi-fiable using the NetPhos server [147]

Studies involving the use of nonselective CD26

inhibitors in CD26-deficient systems suggest that DP8

and DP9 are likely to play immune roles previously

attributed to CD26, for example, in in vitro

prolifera-tion [148], arthritis [149] and haematopoiesis [150]

Recently, there has been more direct evidence of DP8

and DP9 immune function, and their potential as

tar-gets for inflammatory diseases As previously outlined,

DP8 and DP9 are present in leucocytes and leucocyte

cell lines [8,12,151]; DP8 mRNA is upregulated in asthma-induced lung [63]; and an inhibitor of DP8 and DP9 attenuates T-cell proliferation [152] and sup-presses DNA synthesis in mouse splenocytes from both wild-type and DPIV gko mice [153]

The use of inhibitors in these studies has suggested that while the CD26 immune system function appears

to be extra-enzymatic, DP8 and DP9 immune func-tions appear to be enzymatic, although the mecha-nisms are yet to be elucidated We have reported four chemokine substrates of DP8 in vitro, namely SDF-1a, SDF-1b, IP10 and ITAC [139], although it is unclear whether intracellular DP8 makes physical contact with chemokines in vivo DP8 could potentially be released

to the extracellular space upon cell death in inflamma-tory lesions, whereby it could retain its activity and process chemokines involved in these pathological lesions [139] In addition, IP10 and ITAC have crucial roles in hepatitis C virus infection, and DP8 is expressed in B-cell chronic lymphocyte leukaemia, var-ious tumours and activated T cells [9] This selective chemokine inactivation might have implications for cancer biology and immunobiology The reactive oxygen species responsiveness of DP8 and DP9 enzyme activities may have an involvement in apoptosis induction of activated T cells [141]

Insights of DP biological functions from DP-deficient animals

DPIV gko and FAP gko mice are healthy Moreover, DPIV gko mice have increased glucose clearance after

a glucose challenge, compared with wild-type mice [154] The same effect is found with tor-treated wild-type mice, but not with DPIV-inhibi-tor-treated DPIV-deficient mice, showing that the mechanism is DPIV enzyme-activity dependent DPIV gko mice resist diet-induced obesity and associated insulin resistance, probably through the activation of peroxisome proliferator-activated receptor-a, which is involved in fatty acid oxidation, downregulation of ste-rol regulatory element-binding protein-1c (which is involved in lipid synthesis) and reduced appetite [155] DPIV-deficient animals also appear to have a mildly altered lymphocyte phenotype DPIV gko mice have a decreased number of natural killer (NK) T lympho-cytes in peripheral blood, suggesting that DPIV may

be involved in the development, maturation and migra-tion of NK T cells [130] Moreover, NK cell cytotoxic-ity against breast adenocarcinoma cells has been found

to be decreased in CD26-deficient rats, suggesting that DPIV activity is associated with NK cytotoxicity [156] Studies on the DPIV-deficient Fischer rat have shown

age-dependent alteration of thymic emigration patterns

and leucocyte subsets [157] A recent report in DPIV

gko mice treated with a DPIV-selective inhibitor

dem-onstrated that DPIV selective inhibition does not

impair T-dependent immune responses to antigenic

challenge [126]

The FAP gko mouse has a normal phenotype in

his-tological and haemahis-tological analysis [158], and the

lack of FAP expression does not impair development

or tissue remodelling in embryos [47] Therefore, other

compensatory pathways are likely to exist involving

molecules with functions similar to FAP, which could

include other DPIV family members or MMPs Studies

on tissue-remodelling models of FAP gko mice may

help to elucidate its roles, in greater detail, in

extra-cellular matrix interactions, liver fibrosis and cancer

Our studies have indicated that FAP gko mice have

reduced fibrosis in a liver injury model [159]

Care should be taken when interpreting the results

of studies on the DPIV and FAP gko mice because

these mice have dual ablation of enzymatic and

extra-enzymatic activities, and therefore the results may not

accurately reflect the effect of DP inhibitors, which

only inhibit enzymatic functions The gko mice could

still be useful to prove that the effect of an inhibitor is

DPIV or FAP specific and not caused by the nature of

the compound In any case, it is essential to carefully

distinguish the enzymatic roles of a DP from its

extra-enzymatic roles in any given cell type The apparently

normal phenotype of the DPIV gko and FAP gko

mice suggests that targeting either or both enzymatic

and extra-enzymatic functions of DPIV and FAP is

likely to produce few, if any, additional off-target

effects It appears that either all roles of DPIV and

FAP in vivo are not critical, or perhaps that

compensa-tory upregulation of another DP occurs in their

absence, or both Our enzyme distribution study

sug-gests that in some DPIV gko mouse organs, a DP

activity was detected that is probably not DP8⁄ 9

derived, but is present at low levels [12] The adjacent

position of the DPIV and FAP genes causes a DPIV⁄

FAP double knockout mouse to be very difficult to

generate, and DP8 and DP9 knockout animals have

not been reported

Implications for DPs in cell biology and

cancer targeting

Overall, there appears to be evidence for both

extra-enzymatic and extra-enzymatic functions of DPs in cell

biol-ogy The two functions may work synergistically, in

opposition or even independently, depending on the

microenvironment and cell type In many

overexpres-sion studies, similar effects have been found with both enzyme-inactive DP mutants and wild-type DP However, a change in expression level of a DP in response to a stimulus is likely to have downstream enzymatic effects; for instance, in neuroblastoma cell lines, SDF-1-mediated migration is attenuated in the presence of overexpressed DPIV [93]

There has been some interest in the use of DP inhib-itors for cancer therapy The nonselective DP inhibi-tor, PT100 (Val-boro-Pro), slows growth of syngeneic tumours derived from fibrosarcoma, lymphoma, mela-noma and mastocytoma cell lines to the same extent in both wild-type and DPIV gko mice [58], and reduces myeloma growth and bone disease [160] In these cases

it is not clear which DP(s), when inhibited, have anti-tumorigenic effects, or whether inhibiting multiple DPs has a synergistic effect Further study using specific inhibitors is required to understand the mechanisms involved It may be that in some cancers DP inhibition attenuates tumour growth, while extra-enzymatic DP functions have no effect, or extra-enzymatic and enzymatic activities are synergistic

There is some evidence in the literature that DPIV-and FAP-exerted effects on cell behaviour are cell-type dependent For example, while FAP overexpression in 293T cells was associated with decreased cell adhesion and cell migration, contrasting effects, namely increased cell adhesion and migration, were found with the LX-2 human stellate cell line [50] In other instances, while anti-tumorigenic effects were associ-ated with increased DPIV expression in melanocytes, nonsmall cell lung carcinoma, prostate cancer and neu-roblastoma cell lines [90–92,95], anti-tumorigenic effects were conversely associated with decreased DPIV expression in the Karpas 299 T-anaplastic cell lym-phoma line [96] Another comparative study found that activation of DPIV in hepatic carcinoma cell lines induces cell apoptosis, but DPIV in Jurkat T cells con-versely plays a role in cell survival [161] DPIV expres-sion is variable in cancers, being upregulated in certain cancer types and downregulated in others As DPIV and FAP have multifunctional properties, their expres-sion levels and mechanisms or sites of action in various cell types may depend on the particular requirement of the cell and on the surrounding environmental factors

at any given time (Fig 1) This seems to be the case for a number of proteases [71]

Structure of the DPIV gene family proteins and therapeutic considerations

There are a number of favourable factors in consider-ing the design and application of DP inhibitors in

therapeutics First, the relatively small size of the

enzyme family can make it easier to specifically target

the enzyme of interest and distinguish it from other

members of the family Second, as indicated by the

phenotype of the gko mice, neither the enzymatic nor

the extra-enzymatic roles of DPIV and FAP appear to

play critical survival roles, which reduces the likelihood

of side effects Third, although similar in structure, the

DPs have some differences at their active site

[3,140,162], so it is likely that individual DPs can be

specifically targeted through careful drug design

Fourth, although they have overlapping properties, the

DPs appear to play different roles to each other

in vivo This is apparent by differences in their

distri-bution [12] and in vitro biological effects [13,50], and

in the absence of compensatory upregulation of DP8

and DP9 in the DPIV gko mouse [12] Yet another

potential advantage is that in certain cell

micro-environments, extra-enzymatic functions could be

anti-tumorigenic, while enzymatic functions may have

pro-tumorigenic properties Thus, targeting DP enzyme

activity may be useful in certain therapies without

disrupting the beneficial effect of extra-enzymatic

functions

The crystal structures of DPIV and FAP [3,163],

and the predicted structures of DP8 and DP9

[140,162], at first glance reveal almost identical fold

and general topology amongst the family members

These proteins are composed of an N-terminal

b-pro-peller domain and a C-terminal a⁄ b-hydrolase domain

The a⁄ b-hydrolase domain, containing the catalytic

triad, is highly conserved throughout this protein

fam-ily The b-propeller domain, which is associated with

extra-enzymatic functions, is variable The active site,

buried deep within the protein, is formed by amino

acids from each domain and includes the catalytic

triad and both conserved and variable residues These

proteins are hollow and accommodate the substrate,

which is stabilized within the active site by these

struc-tures (Fig 4A)

For development of inhibitors selective for each DP,

it is helpful to carefully consider structural differences,

particularly around the active site In order to develop

selective inhibitors, current research has therefore

focused on these variable regions and on the diversity

of each DPIV gene family protein (Fig 4B) Variable

regions around the active site include two loop regions

– one forming the P2 pocket (P2-loop) and the other

forming a substrate-binding region connected to the

glutamate-rich region (EE-helix) and stabilized by salt

bridges (R-loop) [162,164] Analysis of the crystal

structures of DPIV, FAP and DP6, and of the models

of DP8 and DP9, indicate that the R-loop is ideal for

selectivity and provides a structural basis for the design of enzyme-selective inhibitors [162] At present

a number of DP8⁄ 9 selective inhibitors have been

A

B

Fig 4 Ribbon representation of the sitagliptin-bound DPIV mono-mer (PDB ID 1X70) (A) Variable and conserved structural features

of the DPIV gene family proteins The C-terminal a ⁄ b-hydrolase domain and the b-propeller domain are coloured blue and grey respectively The DPIV inhibitor sitagliptin (shown as a green stick) denotes the location of the active site Some residues in front of the figure, which would otherwise obscure the active site, have been omitted to indicate the hollow cavity found in this protein family The regions of the active site are represented as follows: the catalytic triad Ser, Asp and His (conserved region) as magenta, grey and blue spheres, respectively; the P2-loop (variable region) in cyan; the R-loop (variable region) in dark green; and the glutamate-rich EE-helix (conserved) in red [162] The double-glutamate motif

is shown as red spheres The yellow sphere denotes the acidic region caused by the presence of Asp663 in DPIV (conserved in DP8 and DP9, equivalent to Ala657 in FAP) (B) Close-up view of the sitagliptin-bound active site of DPIV Sitagliptin is shown in stick representation, with carbon in green, nitrogen in blue, oxygen in red and fluorine in grey Polar interactions between sitagliptin and the surrounding structural motifs are denoted by black dotted lines Rational drug design for the DPIV gene family proteins focuses on the variable regions presented by the P2-loop (cyan) and the R-loop (dark green), and on the acidic pocket presented by the Asp (yellow sphere) [162] The image was generated using PYMOL (DeLano WL: http: ⁄ ⁄ www.pymol.org).