Tài liệu Báo cáo khoa học: Structural and functional specificities of PDGF-C and PDGF-D, the novel members of the platelet-derived growth factors family docx

The PDGFs have been classified as members of the superfamily of growth factors characterized by the strongly conserved pattern of six cysteine residues making up intra- and intermonomer d

Structural and functional specificities of PDGF-C and

PDGF-D, the novel members of the platelet-derived growth factors family

Laila J Reigstad1,2, Jan E Varhaug2,3and Johan R Lillehaug1

1 Department of Molecular Biology, University of Bergen, Norway

2 Department of Surgical Sciences, University of Bergen, Norway

3 Haukeland University Hospital, Bergen, Norway

Introduction

The platelet-derived growth factors PDGF-A and -B

have since the late 1970s been recognized as important

factors regulating embryonic development,

differenti-ation, cell growth and many diseases including

malig-nancies The PDGFs have been classified as members

of the superfamily of growth factors characterized by

the strongly conserved pattern of six cysteine residues

making up intra- and intermonomer disulfide bridges,

the cystine knot family of proteins [1–3] Examples of cystine knot subfamilies are the glycoprotein hormone family [4], the cyclotide family [5,6], and the TGFb family and NGF family [2] Extended information about subfamilies can be obtained in the Cystine Knot Database (http://hormone.stanford.edu/cystine-knot) This review focuses on the structure and function of the two novel members, PDGF-C and -D, of the PDGF subfamily of the cystine knot superfamily The PDGFs show high sequence identity with the vascular

Keywords

PDGF; cystine knot; CUB; growth factor

domain

Correspondence

J R Lillehaug, Department of Molecular

Biology, University of Bergen, Post Box

7800, 5020 Bergen, Norway

Fax: +47 55 58 96 83

Tel: +47 55 58 64 21

E-mail: johan.lillehaug@mbi.uib.no

(Received 15 July 2005, revised 19

September 2005, accepted 22 September

2005)

doi:10.1111/j.1742-4658.2005.04989.x

The platelet-derived growth factor (PDGF) family was for more than

25 years assumed to consist of only PDGF-A and -B The discovery of the novel family members PDGF-C and PDGF-D triggered a search for novel activities and complementary fine tuning between the members of this fam-ily of growth factors Since the expansion of the PDGF famfam-ily, more than

60 publications on the novel PDGF-C and PDGF-D have been presented, highlighting similarities and differences to the classical PDGFs In this paper we review the published data on the PDGF family covering struc-tural (gene and protein) similarities and differences among all four family members, with special focus on PDGF-C and PDGF-D expression and functions Little information on the protein structures of PDGF-C and -D

is currently available, but the PDGF-C protein may be structurally more similar to VEGF-A than to PDGF-B PDGF-C contributes to normal development of the heart, ear, central nervous system (CNS), and kidney, while PDGF-D is active in the development of the kidney, eye and brain

In adults, PDGF-C is active in the kidney and the central nervous system PDGF-D also plays a role in the lung and in periodontal mineralization PDGF-C is expressed in Ewing family sarcoma and PDGF-D is linked to lung, prostate and ovarian cancers Both PDGF-C and -D play a role in progressive renal disease, glioblastoma⁄ medulloblastoma and fibrosis in several organs

Abbreviations

CNS, central nervous system; CUB, Clr ⁄ Cls, urchin EGF-like protein and bone morphogenic protein 1; CVB3, coxsackievirus B3;

EGF, endothelial growth factor; PDGF, platelet-derived growth factor; VEGF, vascular endothelial growth factor.

endothelial growth factors (VEGF) and the family is

therefore often referred to as the PDGF⁄ VEGF

family

The PDGF family of growth factors

The PDGF family consists of PDGF-A, -B, -C and -D

[7–12] The cystine knot motif of the four PDGFs

contains two disulfide bridges linking the antiparallel

strands of the peptide chain forming a ring penetrated

by the third bridge [3] This forces the protein to adapt

a three-dimensional arrangement that partly exposes

hydrophobic residues to the aqueous surroundings,

leading to the formation of either homo- or

hetero-dimers (PDGF-AA, -AB, -BB, -CC, -DD) [13,14] In

addition to a conserved cystine knot motif, these four

growth factors show a high sequence identity The four

PDGFs are inactive in their monomeric forms They

share the same receptors; the PDGF receptor-a and -b

These receptors dimerize when the dimeric PDGF

binds The receptors may combine to generate

homo-or heterodimers, resulting in three possible

combina-tions, PDGFR-aa, -ab and -bb, having different

affinities towards the four PDGFs

All PDGFs play important roles in embryogenesis

and adult maintenance, in addition to participating in

the phenotypes of various diseases and malignancies

The novel PDGFs are both involved in progressive

renal diseases, glioblastomas, medulloblastomas and

fibrosis of a variety of tissues PDGF-C appears to

play an important role in Ewing family sarcomas,

while PDGF-D is linked to lung, prostate and ovarian

cancers

Discovery of the PDGFs

PDGF-A and PDGF-B have been extensively studied

since the 1970s, while PDGF-C and PDGF-D were

discovered recently The PDGF-c gene was published

in 2000 by three independent groups and named

PDGF-C [8], Fallotein [15] and SCDGF [7] In each

case, the discovery was based on the identification of a

cDNA sequence showing similarity to members in the

PDGF⁄ VEGF family of growth factors PDGF-c and

PDGF-C are now the accepted name for this gene and

protein, respectively

The last member of the PDGF family was published

in 2001, again by three independent groups, and

named PDGF-D and SCDGF-B [10–12] The gene was

identified by BLAST search in the EST database for

homologues of the PDGF⁄ VEGF family PDGF-d and

PDGF-D are now the accepted names for this gene

and protein, respectively

The pdgf genes PDGF-c and PDGF-d were named and placed in the PDGF⁄ VEGF family because they encode the highly conserved cystine knot motif characteristic of the growth factor family While the classical PDGF-a and PDGF-b mainly encode the growth factor domain, PDGF-c and PDGF-d encode a unique two-domain structure with an N-terminal ‘Clr⁄ Cls, urchin endo-thelial growth factor (EGF)-like protein and bone morphogenic protein 1’ (CUB) domain [16] in addition

to the C-terminal growth factor domain (Fig 1A) The pdgf genes are located on four different chromo-somes; PDGF-a and -b on chromosomes 7 and 22 [17,18], and PDGF-c and -d on chromosomes 4 and 11 [19], respectively The genomic organization of the pdgf genes is quite similar, although PDGF-c and -d genes are significantly longer due to large intron sizes and cover about 200 kb compared to approximately 20 kb for PDGF-a and -b [19–21]

Each of the four pdgf genes contains a long 5¢ untranslated region and a verified (PDGF-a and -b) or putative (PDGF-c and -d) signal peptide in exon 1 (Fig 1B) In the PDGF-c and -d genes, exons 2 and 3 encode the CUB domains, while in PDGF-a and -b these exons encode precursor sequences residing 5¢ to the cystine knot encoding sequence The hinge regions

of PDGF-c and PDGF-d connecting the CUB and the cystine knot domains are encoded by exons 4 and 5, respectively These hinge region sequences encode con-served basic motifs and similar motifs are found in PDGF-A and -B The motifs are identified as proteo-lytic cleavage sites for proteases used in post-transla-tional protein processing Exon 4 of PDGF-d encodes

a unique sequence not present in the other PDGFs So far, no function has been assigned to this sequence The exons in PDGF-c (exons 5 and 6) and PDGF-d (exons 6 and 7) encoding the cystine knot motifs resemble the corresponding exons in the PDGF-a and -b(exons 4 and 5) genes Both in PDGF-a and -b, exon

6 encodes a C-terminal basic retention sequence that may be removed during the maturing and release of the proteins Analysis on the 3¢ untranslated mRNA region of PDGF-C identified five adenylate⁄ uridylate-rich elements, these being the best characterized mam-malian determinant for highly unstable RNAs [15]

The pdgf promoters PDGF-A and -C share common mechanisms of gene regulation Their expression is controlled by the zinc finger transcription factors Egr1 and Sp1, which have affinity for overlapping GC-rich binding sites in the

proximal region of the PDGF-a and PDGF-b

promot-ers [22,23] So far, no information on the PDGF-d

promoter has been published, but functional

character-istics of the human PDGF-c promoter has been

repor-ted [24] Comparison of the PDGF-c promoter of

human smooth muscle cells with the characterized

human PDGF-a promoter identified a GC-rich

sequence ()35 to )1) in PDGF-c, with high similarity

to the )76 to )47 sequence of the PDGF-a promoter

Both Egr1 and Sp1 were shown to bind the 35 bp

sequence of the PDGF-c promoter FGF-2 stimulates

Egr1 expression through the Erk⁄ MAPK pathway,

and Egr1 translocates to the nucleus where it binds to

the proximal PDGF-c promoter resulting in increased

PDGF-C expression

Alternative splicing of the PDGFs

No alternative splicing of PDGF-c mRNA has been

demonstrated However, alternative splicing is

sugges-ted because two shorter PDGF-C cDNAs have been obtained [25] Based on the variant PDGF-c sequences isolated by PCR, the splice donor⁄ acceptor sites are located to nucleotides 719⁄ 720 and 988 ⁄ 989, resulting

in two alternative proteins; one short variant encom-passing almost only the CUB domain, and the longer variant containing the CUB domain and the final 30 residues in the C-terminal end of the growth factor domain These splice variants are also present in human thyroid papillary carcinomas (L J Reigstad,

J E Varhaug and J R Lillehaug, unpublished results) Based on mRNA analysis of PDGF-d, splice variants have been reported to be present in mouse heart, liver and kidney [26] Interestingly, deletion of exon 6 cau-ses a frame shift and an early stop codon in exon 7, resulting in a protein lacking the growth factor domain and without mitogenic activity (Fig 1B) The

PDGF-D protein encoded by this splice variant could only be detected in mouse tissues and not in human cell lines

or tissues A second PDGF-d RNA splice variant lacks

A

B

Fig 1 PDGF protein and gene structure.

(A) Schematic drawing of the four full-length

PDGF proteins In PDGF-C and -D, the

hydrophobic putative N-terminal signal

pep-tide (black) is separated from the N-terminal

CUB domain (110 residues, red) by a short

region (orange) A hinge region (blue)

separ-ates the CUB domain from the C-terminal

growth factor domain containing the cystine

knot motif (115 residues, yellow) PDGF-A

can be alternatively spliced and carries two

stop codons resulting in proteins of 198 and

211 amino acid residues The numbers

show residue numbering in the PDGFs.

(B) Schematic drawing of gene structures

encoding the four PDGF polypeptide chains.

Exons are coloured and numbered: CUB

domain (red), hinge region (blue) and growth

factor domain (yellow) The introns are

shown in white Start codons (ATG), stop

codons (Stop) and the proteolytic cleavage

sites (black arrows) are denoted Exons and

introns are not drawn in scale The PDGF-A

and -B genes cover approximately 20 kb and

the PDGF-C and -D cover approximately

200 kb Alternative splicing has been shown

for PDGF-A and PDGF-D, in which exon 6 is

missing (see text).

18 bp within the CUB domain of both mouse and

human PDGF-D mRNA [26,27] In the case of PDGF-a,

exon 6 may be present or not resulting in two splice

variants encoding a long and a short PDGF-A protein

(Fig 1B) [28] PDGF-B mRNA has not been reported

to be alternatively spliced

The PDGF proteins

The PDGF-A and -B proteins contain only the growth

factor domains whereas PDGF-C and -D have a

unique two-domain structure containing the

N-ter-minal CUB domain separated from the C-terN-ter-minal

growth factor domain by a hinge region (Fig 1A)

PDGF-C and -D share an overall sequence identity of

42% with highest similarity in the CUB and cystine

knot-containing growth factor domain, whereas the

hinge region and the N-terminal region show less

identity [10,12]

While PDGF-A and -B can form both homo- and

heterodimers (PDGF-AA, -AB, -BB), PDGF-C and -D

exist only as homodimers (PDGF-CC and -DD) The

full-length PDGF-C and -D monomers are 54–55 kDa

and 49–56 kDa, respectively, differing from their

theo-retical sizes of 39 and 43 kDa based on their amino

acid sequences [8,12,29] The divergence from the

theoretical values indicates that PDGF-C and -D may

be post-translationally modified In addition to being

secreted to the extracellular space, the PDGF-C

protein is shown to be constitutively expressed in the

cytoplasm in rat smooth muscle cells residing in

arteries and arterioles [30] Additionally, our data

(L J Reigstad, J E Varhaug and J R Lillehaug, unpublished results) show full-length PDGF-C to be present in both the cytoplasm and nucleus, a feature also described for PDGF-B [31,32] The function of these two PDGFs in the nucleus is unclear

The region N-terminal of the CUB domain in PDGF-C and PDGF-D

The N-terminal ends of both PDGF-C and -D contain

a hydrophobic sequence predicted to be a signal pep-tide with a putative peptidase cleavage site between residues 22 and 23 [8,10,12] When part of the putative PDGF-C signal sequence was deleted no mitogenic activity was detected, suggesting that PDGF-C is secre-ted by the aid of this N-terminal region [7] The sequence between residues 23 and 50 of PDGF-C, and residues 23–56 of PDGF-D, contain no known motifs

or domains (Fig 1)

The CUB domain of PDGF-C and PDGF-D The CUB domain was first identified in complement subcomponents Clr⁄ Cls, urchin EGF-like protein and bone morphogenic protein 1 [16] These proteins are often referred to as the prototype CUB domains Like the prototype CUB domains, the PDGF-C and -D CUB domains span approximately 110 residues and show 27–37% and 29–32% sequence identity to the prototypic CUB domains, respectively (Fig 2) [8] The CUB domains of PDGF-C and -D share  55%

sequence identity CUB domains are assembled as a

Fig 2 The CUB domain The CUB domains of PDGF-C and PDGF-D aligned with the prototype CUB domains of human neuropilin (acces-sion no CAI40251) and human bone morphogenic protein-1 (BMP-1, acces(acces-sion no CAA69974) Red, squared areas highlight sequence iden-tity among the sequences The four cysteines conserved in the prototypical CUB domains are labelled in yellow The two cysteines missing

in the PDGF-C and PDGF-D are marked by red circles while the two cysteines present are marked in blue circles The two cysteines of PDGF-C (accession no AAF80597) correspond to Cys104 and 124, whereas in PDGF-D (accession no AAK38840) these cysteines are Cys109 and 131.

compact ellipsoidal b-sandwich, with a hydrophobic

core essential for the overall domain folding The

b-sandwich is built up of two five-stranded b-sheets of

antiparallel b-strands [33–36] Most CUB domains are

reported to contain four conserved cysteines that form

two disulfide bridges between nearest-neighbour

cyste-ines, resulting in disulfide bridges located on opposite

edges of the domain As both the PDGF-C and -D

CUB domains contain only two cysteines [7] which,

compared to the classical CUB domains, are the two

most-C-terminally located cysteines, the CUB domains

of PDGF-C and -D may have only one disulfide

bridge At present it is unclear how this may influence

their 3D structure In the two crystallised CUB

domains of a serine protease associated with serum

mannose-binding proteins (MAPS), the N-terminally

located CUB domain contains only one disulfide

bridge, while the second CUB domain of MAPS has

two bridges One disulfide bridge instead of two may

result in a slightly less tight b-sandwich in the

N-ter-minally located CUB domain, but the structural and

functional significance of only one bridge remains

unknown [35]

The CUB domain is found in several extracellular

proteins, many involved in development, and they are

thought to mediate protein–protein and

protein–carbo-hydrate interactions, in addition to binding to

low-molecular-mass ligands [16,37] Several reports state

that the CUB domains of PDGF-C and -D have to be

cleaved extacellularly to make the C-terminal growth

factor domains active [8,10,12] The CUB domains are

believed to prevent PDGF-C and -D binding to their

receptors by structurally blocking receptor-binding

res-idues of the growth factor domain In contrast, two

reports state that full-length PDGF-C and -D exhibit

in vitro mitogenic activity towards coronary artery

smooth muscle cells and fibroblasts [25,26]

Addition-ally, the CUB domain of PDGF-C exhibits mitogenic

activity on human coronary artery cells independent of

the presence of its growth factor domain, suggesting a

possible biological activity of the CUB domains

them-selves [25] Interestingly, when Cys124 of the PDGF-C

CUB domain was mutated to serine, the mitogenic

activity of CUB was reduced by approximately 50%

The mitogenic CUB activity could not be confirmed in

transgenic mouse hearts overexpressing CUB and

over-expression gave no pathological effect in the heart [38]

CUB domains may facilitate unique, undiscovered

functions of full-length PDGF-C and -D This is

reflec-ted in a report on full-length PDGF-D of eye lens

tis-sue, in which the secreted PDGF-D does not appear to

be proteolytically cleaved [39] The CUB may mediate

interactions between PDGF-C or PDGF-D and

elements of the extracellular or pericellular matrix Furthermore, a role for the PDGF CUB domains in receptor binding is suggested based on studies of the transmembrane receptor, neuropilin-1, which consists

of two CUB domains and a coagulation factor domain, acting as coreceptors for VEGF-A and sem-aphorins (reviewed in [40]) Crystallography studies of the MAPS protein containing two CUB domains sug-gests that CUB domains may also participate in pro-tein heterodimer formations [35]

The hinge region and proteolytic cleavage for growth factor activation

The hinge regions of PDGF-C and -D, separating the CUB and the growth factor domains (Fig 1), show no homology to known sequences [41] but contain dibasic cleavage sites for proteolytic removal of the CUB domains and thereby activation of the growth factor domains PDGF-C and -D contain both the CUB and growth factor domains when they are secreted and pro-teolytic cleavage is therefore suggested to take place extracellularly Plasmin cleaves PDGF-C at RKSR234 [8,41], and PDGF-D at RKSK257 [12] Tissue plasmi-nogen activatior (tPA) cleaved PDGF-C at RKSR234

in vivo [42,43] and urokinase plasminogen activator (uPA) was found to cleave PDGF-D at RGRS250, thereby activating this growth factor [44] PDGF-A is cleaved by furin at RRKR86 [45], while PDGF-B is cleaved at RGRR81 by a still unidentified protease [46]

The PDGF growth factor domain The determination of the crystal structure of nerve growth factor [47], transforming growth factor b2 [48], PDGF-BB [49] and chorionic gonadotropin [4] revealed unexpected topological similarities among these four proteins belonging to separate families of growth factors Despite very little sequence similarity, they all contain an unusual arrangement of cysteines linked in disulfide bridges to form a conserved cystine knot motif [1,2,50] The cystine knot is located in a conserved b-sheet structure referred to as the growth factor domain Although the four growth factor super-families have a common topology, they differ in the number of disulfide bonds, the interfaces used to form the dimers, and the way in which the monomers dimer-ize [1,2]

In the PDGF⁄ VEGF family, the crystal structures

of PDGF-BB [49], VEGF-AA [51,52], VEGF-AA together with elements of its Flt1 receptor [53], and PlGF-1 dimer [54] have been solved at 3.0, 1.9, 1.7 and 2.0 A˚, respectively Characteristic for these growth

factor domains are two long, highly twisted

antiparal-lel pairs of b-strands in an antiparalantiparal-lel side-by-side

mode They all contain eight (I–VIII) highly conserved

cysteines The monomeric antiparallel b-strands are

connected by three loops, referred to as loops 1, 2 and

3 (Fig 3A), and due to the head-to-tail arrangement,

loop 2 of one monomer will be close to loops 1 and 3

of the other monomer when the dimer is formed (Fig 3B) Six of the conserved cysteines are engaged in three intrachain disulfide bonds (Cys I-VI, III-VII, V-VIII) stabilizing the cystine knot structure, while two cysteines (Cys II and IV) are involved in inter-chain disulfide bonds (Fig 3C,D) [55,56] The three intrachain disulfide bonds makes the cystine knot very

Fig 3 The PDGF-C growth factor domain Ribbon presentations of the proposed PDGF-C model [57] displaying the twisted b-sheets and the N-terminal a-helix of the PDGF-C monomer (A) and the PDGF-CC dimer (B) The N-terminal (N) and C-terminal (C) ends for the monomer are marked The three loops (loop 1-2-3) connecting the b-strands are labelled (C, D) Sequence alignments of the growth factor domains of PDGF-A (accession no P15692), PDGF-B (accession no 1109245 A), PDGF-C (acces-sion no AAF80597), PDGF-D (acces(acces-sion

no AAK38840), VEGF-A (accession no NP003367) and PIGF-1 (accession no 1FZV) Red, squared areas show sequence identity among the sequences The eight conserved cysteines are shown in yellow The extra cysteines of PDGF-C and -D are labelled in blue The green squares highlight the area of disagreement in sequence align-ment (see text) (C) Sequence alignalign-ment of PDGF ⁄ VEGF family members where the green square highlights the area containing the insert of three residues between con-served cysteines III and IV in both PDGF-C and -D, and that PDGF-D is missing the con-served cysteine V (D) Sequence alignment

of PDGF ⁄ VEGF family members showing PDGF-C and -D to contain all eight con-served cysteines and have an insert of three residue between cysteine V and VI (green area).

stable as the first (Cys I-VI) and second (Cys III-VII)

disulfide bonds link two adjacent b-strands, making a

ring which is penetrated by the third (Cys V-VIII)

disulfide bond, covalently connecting two further

b-strands [49,54] Additional stability to the dimer

structure is the extensive hydrophobic core formed by

residues from both monomers

The PDGF-C growth factor domain shares 27–35%

sequence identity with the rest of the PDGF⁄ VEGF

family [8], and with specific reference to the other three

PDGFs the identity is nearly 25% [20] The growth

factor domains of PDGF-A and -B show

approxi-mately 50% sequence identity, while the identity

between the domains of PDGF-C and -D is also nearly

50%

Compared to the eight conserved cysteines in

PDGF-A and -B, PDGF-C contains four and PDGF-D

two additional cysteines These extra cysteines and the

lack of solved PDGF-C and -D 3D structures makes

identification of the cysteines that participate in the

conserved disulfide bridges difficult Because of this,

two different sequence alignments of PDGF-C and -D

covering the area of conserved cysteines III to VI are

included here (Fig 3C,D) Figure 3C shows the

align-ment of PDGF-C and -D to allow three residues

(NCA and NCG, respectively) between conserved

cysteines III and IV, an insert not present in the other

members of the PDGF⁄ VEGF family This alignment

also indicates that PDGF-D lacks the conserved

cys-teine V [7,8,10,12,15,25,57] The alignment in Fig 3D

has a different three-residue insert, which is located

between conserved cysteines V and VI In this

align-ment, PDGF-D contains all eight conserved cysteines

of the cystine knot motif [20,21,41] Crystallization or

NMR studies of PDGF-C and -D proteins will resolve

this debate, but our published 3D model of the

PDGF-C growth factor domain indicates the disulfide

bridges in PDGF-C to consist of Cys250 and 294,

Cys280 and 335, and Cys287 and 337, and the

inter-monomeric bonds to consist of Cys274 and 286 [57]

At present, analysis of PDGF⁄ VEGF domains show

that PDGF-C is more similar to VEGFs than PDGFs

[8,15,57], all in all favouring the alignment in Fig 3C

The region C-terminal of the growth factor

domain

In PDGF-A and -B, the C-terminal regions contain a

basic sequence with a dual function First, the

sequence mediates electrostatic interactions with

com-ponents of the extracellular matrix such as heparin [58]

and collagens [59] Second, the basic sequence may

cause retention of the growth factors within the

produ-cer cell [60] PDGF-C is shown to have a heparin-binding domain in the C-terminal region of the growth factor domain but the exact residues responsible for the binding have not been identified [25] PDGF-D has not been shown to bind heparin

Post-translational modifications and regulations

of PDGFs Several members of the PDGF family are predicted to have potential N-glycosylation sites PDGF-C has three predicted N-glycosylation sites (N25, 55, 254), the last residing in the growth factor domain [8] Due

to the difference in expected (39 kDa) and observed (55 kDa) relative molecular mass as determined by SDS⁄ PAGE electrophoretic mobility, it has been sug-gested that PDGF-C may be glycosylated One report gives experimental indication of glycosylation, recom-binant full-length PDGF-C protein, secreted from insect cells, slightly changed mobility when treated with N-glycosidase F, but the mobility change did not result in the expected 39 kDa protein size [25]

PDGF-D has one predicted N-glycosylation site at N276 located in the growth factor domain [12] and PDGF-B has a verified N-glycosylation site at N63 [61]

PDGF-A has one predicted N-glycosylation site at N134 but

is not reported to be N-glycosylated [62], despite a report from 1981 in which PDGF was stated to be glycosylated [63] In the case of PDGF-C, the post-translational modification remains unidentified but SUMOylation or ubiquitinylation may be candidates

Receptor binding of PDGFs The PDGFs bind to the protein tyrosine kinase receptors PDGF receptor-a and -b These two recep-tor isoforms dimerize upon binding the PDGF dimer, leading to three possible receptor combina-tions, namely -aa, -bb and -ab The extracellular region of the receptor consists of five immunoglo-bulin-like domains while the intracellular part is a tyrosine kinase domain The ligand-binding sites of the receptors are located to the three first immuno-globulin-like domains (reviewed in [64]) The residues

in PDGF-A and -B responsible for receptor binding reside in loop 2, in addition to RKK161 in

PDGF-AA and R27 and I30 in PDGF-BB The residues involved in PDGF-CC and -DD receptor binding remain to be identified, but our published 3D model

of PDGF-C suggests, when compared to the crystal structure of VEGF-AA complexed to domain 2 of its receptor, that the region containing residues W271 and LR312 might be involved [57]

PDGF-CC specifically interacts with PDGFR-aa and

-ab, but not with -bb, and thereby resembles

PDGF-AB [8,41] PDGF-DD binds to PDGFR-bb with high

affinity, and to PDGFR-ab to a markedly lower extent

and is therefore regarded as PDGFR-bb specific [10,12]

PDGF-AA binds only to PDGFR-aa, while PDGF-BB

is the only PDGF that can bind all three receptor

combinations with high affinity [65] Both PDGF-CC

and -DD activate PDGFRs resulting in downstream

phosphorylation of extracellular signal-regulated

pro-tein kinase⁄ mitogen-activated protein kinase (Erk ⁄

MAPK) and Akt⁄ PKB pathways [57,66,67]

Fine tuning of PDGF and PDGFR isoform

expression and regulation

Expression of both receptors and each of the four

PDGFs is under independent control, giving the

PDGF⁄ PDGFR system a high degree of combinatorial

flexibility To understand how the four PDGFs may

generate different biological signals, five observations

may be relevant First, different cell types vary greatly

in the ratio of PDGF isoforms and PDGFRs expressed

Second, the PDGFR expression levels are not constant

Different external stimuli such as inflammation,

embry-onic development or differentiation modulate cellular

receptor expression allowing binding of some PDGFs

but not others Additionally, some cells display only one

of the PDGFR isoforms while other cells express both

isoforms, simultaneously or separately Third, different

splice forms of the PDGFs appear to be expressed

dif-ferently, as shown for the two PDGF-A proteins in

rest-ing and active monocytes [68] and as indicated for the

two PDGF-D proteins identified in mouse but not

humans [26] Fourth, regulation of the classical PDGFs

after secretion includes covalent binding to the

extracel-lular secreted protein, acidic and rich in cysteine

(SPARC), which only binds PDGF-AB and -BB,

decreasing their reactive concentrations and favouring

PDGF-AA signalling [69] Data on possible PDGF-CC

or -DD binding to SPARC have not been reported The

major reversible PDGF-A and -B binding to

extracellu-lar protein is a2-macroglobulin [70] The PDGF–a2

-macroglobulin complex serves multiple functions It

makes PDGF-AA, -AB and -BB unable to bind their

receptors, it protects the PDGFs against proteolytic

degradation, and may remove the PDGFs from

circula-tion via a2-macroglobulin receptors There are no data

currently available about interactions between the novel

PDGFs and a2-macroglobulin, but several other growth

factors, such as FGF-2, TGF-b and TNF-a, also bind

a2-macroglobulin Fifth, the expression of highly

speci-fic proteases that proteolytically activate the PDGFs

will also influence the availability and activity of the dif-ferent isoforms This can be exemplified by the proteo-lytic cleavage of PDGF-D While the human prostate carcinoma cell line LNCaP produces a specific protease

to process the full-length PDGF-D [66], there is no pro-tease capable of cleaving the full-length PDGF-D secre-ted by cells and tissues in the eye [39]

Gene knockout studies on the PDGFs For the PDGF-a gene knockout mouse, there are two restriction points concerning animal survival; one pren-atally at E10 and one postnpren-atally at about two weeks [71] The postnatally surviving mice had a symmetrical reduction of the size of most organs, developed lung emphysema due to lack of alveolar myofibroblasts, resulting in the loss of parenchymal elastin fibres and

no formation of alveolar septa The mice died about two weeks old due to respiratory problems The phe-notype reveals a role for PDGF-A in embryonic devel-opment, as well as a highly specific and critical role for PDGF-A in lung alveolar myofibroblast differentiation and lung development

When the PDGF-b gene is knocked out the mice die perinatally, displaying several anatomical and histolog-ical abnormalities [72] The glomerular tufts of the kidneys do not form as there is complete absence of mesangial cells, and instead one single or a few disten-ded capillary loops fill the glomerular space Further-more, the heart and some large arteries dilate in late-stage embryos and fatal haemorrhages occur just prior to birth Based on these findings, PDGF-B is assigned a crucial role in establishing certain renal and circulatory functions

Comparing the PDGF-a and PDGF-b knockout mice, there are similarities in the resulting phenotypes Both the alveolar myofibroblasts and the mesangial cells express a-smooth muscle actin and have a con-tractile phenotype, functioning as anchors for an involuted epithelial sheet, the alveolar sac or the glom-erulus By losing this anchor in the knockout mice, there is a failure of involution and the physiological functions are impaired, as a result of decreased surface area for gas exchange or glomerular filtration, in PDGF-A and PDGF-B mutants, respectively

Knockout studies on PDGF-c in mice clearly dem-onstrate a role for PDGF-C in embryonal development [73] The knockout of PDGF-c results in mice dying perinatally owing to difficulties in feeding and brea-thing, as they have a complete cleft of the secondary palate because the palate bones do not meet Addition-ally, the dorsal spinal cord was deformed in the lower spine The null mutant PDGF-C embryos had

subcutaneous oedema in the flank of the body between

the limbs lacking connective tissue, and showed several

blood-filled blisters in frontnasal and lateral forehead

In the early embryo development, the features of the

knockout PDGF-c mice largely overlap with knockout

PDGF-a mice PDGF-c⁄ PDGF-a mice showed growth

retardation, pericardial effusion, a wavy neural tube

and subepidermal blisters, dying before E17 In total,

PDGF-C has specific roles in palatogenesis and in

morphogenesis of the skin tissue PDGF-d knockouts

have not been reported

Functions of the PDGF-C and PDGF-D proteins

The role of PDGF-A and -B proteins in normal

pro-cesses, malignancies and diseases have been

character-ized in a wide diversity of cells, organs and species

(reviewed in [62]) This part of the review will therefore

focus on the PDGF-C and PDGF-D proteins, as their

functions are starting to be revealed

PDGF-C in normal processes

The expression of PDGF-C mRNA in embryonic

mouse tissue is located in the kidney, lung, brain,

heart, spinal cord and several other tissues, and

partic-ularly at sites of developing epidermal openings such

as the mouth, nostrils, ears and eyelids [7–9,74] In the

adult mouse, PDGF-C is mainly expressed in kidney,

testis, liver, brain and heart Adult humans

addition-ally express PDGF-C in the pancreas, adrenal gland,

skeletal muscles, ovary, prostate, uterus and placenta

[8,10,15,20,27,74] PDGF-C mRNA and protein

expression is also detected in normal human thyroid

tissue (L J Reigstad, J E Varhaug and J R

Lilleh-aug, unpublished results)

PDGF-C in tissue remodelling

PDGF-C appears to be involved in all three phases

(inflammation, proliferation and remodelling⁄

matur-ing) of wound healing [75] Extensive expression and

secretion of full-length PDGF-C from a-granules of

isolated platelets indicate that it plays a role in the

inflammatory phase [76] In the proliferative phase of

wound healing capillary growth is triggered by low

oxygen, and PDGF-C was recently shown to

revascu-larize ischemic mouse heart and limb in vivo as

effi-ciently as VEGF and PlGF-1 [77] PDGF-C mediates

increased mRNA and protein levels of

metalloprotein-ase-1 (MMP-1) and its inhibitor (TIMP-1), both being

important in the remodelling phase of tissues [78]

These results are further verified by in vivo experiments

showing that PDGF-C enhanced the repair of a full-thickness skin excision in a delayed diabetic wound healing mouse model by stimulation of fibroblast pro-liferation, epithelial migration, extensive vasculariza-tion and neutrophil infiltravasculariza-tion [41]

PDGF-C in angiogenesis The high PDGF-C expression in the angiogenic tissues

of placenta, ovary and embryo has led to several in vitro and in vivo experiments defining PDGF-C as a potent angiogenic factor, similar to VEGF and the classical PDGFs The underlying mechanisms are still to be understood In the aortic ring outgrowth assay,

PDGF-C mediated significant increased outgrowth of fibro-blasts and smooth muscle cells, to a degree comparable

to that of VEGF, PDGF-AA and -BB [41] PDGF-C efficiently stimulated the formation of new blood vessels with high vessel density growing towards the implanted dish of the chorioallantoic membrane (CAM) assay [79] In addition, PDGF-C stimulated formation of new branches and vessel sprouts from those initially formed Several reports show in vivo angiogenic PDGF-C effects When PDGF-C-coated micropellets were added

to mouse corneal micropockets, PDGF-C potently induced neovascularization of the avascular corneal tis-sue In these experiments, PDGF-C was as potent as BB and more potent than AA

PDGF-C affects the endothelial cells lining the blood vessels

by mobilizing the endothelial progenitor cells, promo-ting the differentiation of bone marrow progenitor cells into mature endothelial cells, and by stimulating the chemotaxis of different mature endothelial cells in ischemic heart and limb muscles In these experiments, PDGF-C also gave enhanced vessel maturing (arterio-genesis) by inducing the differentiation of bone mar-row cells into smooth muscle cells which coat the endothelial cell layer of the vessels

PDGF-C in embryonic development and adults: kidney, central nervous system (CNS) and ears PDGF-C has important functions both in embryonic development and in adult tissues (Fig 4) and there appears to be expression differences between species High constitutive PDGF-C expression is present in the adult kidneys of mouse, rat and humans [8,30,80,81]

In human adult kidneys, PDGF-C is detected in vascu-lar endothelial cells and smooth muscle cells of arter-ies, in parietal glomerular cells but not in the glomerular tuft In the tubulointerstitium, PDGF-C is located in collecting ducts and the loop of Henle [81] PDGF-C protein localization in adult rodent kidneys

was similar to the adult human kidney, but the rodent

kidneys did not contain PDGF-C protein in the

pari-etal glomeruli cells [30,74,76] In contrast to kidney

development in rodents, the developing human

glo-meruli express PDGF-C in the metanephric epithelial

mesenchyme and in the parietal epithelial cells [30,81]

During kidney development, PDGFR-a is expressed

in the glomerular epithelial mesenchyme, suggesting a

paracrine signalling pathway for both PDGF-A and

-C in kidney vascular and interstitial development [74]

In the embryonal rat CNS, PDGF-C mRNA was

expressed in the notochord (prestage of the spinal

cord) and subsequently in the maturing spinal cord,

while the adult spinal cord does not express

PDGF-C [27] The presence of PDGF-PDGF-C in the developing

spinal cord has also been shown in chicken [7]

PDGF-C mRNA is detected in the floor plate and

the ventricular zones of cortex and adjacent to the

floor plate of the embryonic brain, whereas in the

adult brain weak PDGF-C expression was observed

only in the olfactory nucleus and pontine nuclei [27]

Quantitative RT-PCR analysis did not detect

PDGF-C in human embryonic or adult brain tissues [82],

although this has been shown through northern blot

analyses [8,74]

PDGF-C mRNA has been detected in the

develop-ing ears of mouse and rat [9,74,83] Durdevelop-ing rat

embryonic development, significant mRNA levels of

PDGF-C, PDGF-A and both PDGFRs are expressed

in cochlear progenitor hair cells of the inner ear [83]

PDGF-D in normal processes Since its discovery four years ago, PDGF-D has been linked to important functions both in embryogenesis and in adult tissues (Fig 4) In human adult tissue, PDGF-D is highly expressed in heart, kidney, pan-creas, ovary, adipose tissue, stomach, bladder, trachea, testis and mammary gland [10,12,84] In organs such

as the kidney and lung, there are several differences in expression patterns between species

PDGF-D in embryonic development and adults: kidney, lung, CNS and eye

Most information on PDGF-D biology has been obtained from studies using the kidney as a model Starting with embryogenesis, PDGF-D protein in the human kidney is expressed mainly in visceral glomer-ular epithelial cells and in smooth muscle cells in renal arteries and also in some fibroblast-like intersti-tial cells but not in the fibrous capsule surrounding the embryonic kidney [29] In mouse, PDGF-D is expressed in the highly vascularized fibrous capsule, the most peripheral part of the cortex metanephric mesenchyme, and in the basal aspect of the branch-ing ureter [12] PDGF-D colocalizes with the PDGFR-b in the differentiating metanephric mesen-chyme, whereas PDGF-B expression is restricted to endothelial cells, indicating the possibility of

PDGF-D⁄ PDGFR-b constituting an autocrine loop, and PDGF-B acting in a paracrine manner to promote proliferation and migration of mesangial and intersti-tial cells in the kidney In the developing human kidney, PDGF-D expression does not colocalize with PDGFR-b, as PDGF-D is expressed in the visceral epithelial cells and PDGFR-b in the mesangial cells Thus here a paracrine role for PDGF-D in prolifer-ation and migrprolifer-ation of the mesangial cells can be indicated [29] In the human adult kidney, PDGF-D protein expression was also detected in smooth mus-cle cells of arteries, arterioles and vasa rectae In contrast to human and mouse adult kidney, the rat adult kidney shows no PDGF-D protein in the glomeruli [85] As in the kidneys, lungs show spe-cies-different PDGF-D expression Cells in normal human lungs do not express PDGF-D protein at detectable levels [10], while in murine lungs PDGF-D mRNA is constitutively expressed [84]

In the embryo, PDGF-D mRNA is hardly detect-able in the spinal cord, but in the adult spinal cord prominent expression is located to the motor neurons [27] In the brain, PDGF-D mRNA was registered

in the thalamus and in a ventricular zone of the

Kidney (9,30,74,81)

CNS (5,27,74)

Heart (74,100)

Ear (9,74,83)

Kidney (30,80,81) CNS (27,74)

Embryonic development Adult tissue

PDGF-C

PDGF-D

Kidney (27,29,80,85) Eye (27,39)

CNS and brain (27) Lung (84,101) Peri mineral (86)

Kidney (27,29)

Eye (27)

Brain (27,110)

Fig 4 Defined functions of PDGF-C and PDGF-D in specific organs

during embryonic development and adult tissue See text and

refer-ences (numbers given) for detailed descriptions Peri mineral,

peri-odontal mineralization.