O

Dpi

MTa

Nb

Sc

a

ARRA

KPPOCEC

1

nc(csm

Ra

h0

Pathology – Research and Practice 212 (2016) 426–436

Contents lists available at ScienceDirect

Pathology – Research and Practice

jou rn al hom epage: www.elsev ier .com/ locate /prp

riginal article

ifferential immunohistochemical expression profiles oferlecan-binding growth factors in epithelial dysplasia, carcinoma

n situ, and squamous cell carcinoma of the oral mucosa

ayumi Hasegawaa,b, Jun Chenga, Satoshi Maruyamac, Manabu Yamazakia,atsuya Abéa,c, Hamzah Babkaira, Chikara Saitob, Takashi Sakua,c,∗

Division of Oral Pathology, Department of Tissue Regeneration and Reconstruction, Niigata University Graduate School of Medical and Dental Sciences,iigata, JapanDivision of Reconstructive Surgery for Oral and Maxillofacial Region, Department of Tissue Regeneration and Reconstruction, Niigata University Graduatechool of Medical and Dental Sciences, Niigata, JapanOral Pathology Section, Department of Surgical Pathology, Niigata University Hospital, Niigata, Japan

r t i c l e i n f o

rticle history:eceived 7 October 2015eceived in revised form 15 January 2016ccepted 14 February 2016

eywords:erlecanerlecan-binding growth factorsral squamous cell carcinomaarcinoma in situpithelial dysplasiaell proliferating zone

a b s t r a c t

The intercellular deposit of perlecan, a basement-membrane type heparan sulfate proteoglycan, is con-sidered to function as a growth factor reservoir and is enhanced in oral epithelial dysplasia and carcinomain situ (CIS). However, it remains unknown which types of growth factors function in these perlecan-enriched epithelial conditions. The aim of this study was to determine immunohistochemically whichgrowth factors were associated with perlecan in normal oral epithelia and in different epithelial lesionsfrom dysplasia and CIS to squamous cell carcinoma (SCC). Eighty-one surgical tissue specimens of oral SCCcontaining different precancerous stages, along with ten of normal mucosa, were examined by immuno-histochemistry for growth factors. In normal epithelia, perlecan and growth factors were not definitelyexpressed. In epithelial dysplasia, VEGF, SHH, KGF, Flt-1, and Flk-1were localized in the lower half of reteridges (in concordance with perlecan, 33–100%), in which Ki-67 positive cells were densely packed. InCIS, perlecan and those growth factors/receptors were more strongly expressed in the cell proliferatingzone (63–100%). In SCC, perlecan and KGF disappeared from carcinoma cells but emerged in the stromal

space (65–100%), while VEGF, SHH, and VEGF receptors remained positive in SCC cells (0%). Immuno-fluorescence showed that the four growth factors were shown to be produced by three oral SCC celllines and that their signals were partially overlapped with perlecan signals. The results indicate that per-lecan and its binding growth factors are differentially expressed and function in specific manners before(dysplasia/CIS) and after (SCC) invasion of dysplasia/carcinoma cells.

© 2016 Elsevier GmbH. All rights reserved.

. Introduction

It remains a challenge to make objective histopathological diag-oses of oral borderline malignancies from epithelial dysplasia andarcinoma in situ (CIS) to microinvasive squamous cell carcinomasSCC) only on hematoxylin and eosin (HE) stained sections, as the

onventional grading systems are too heavily dependent on theubjectivity of pathologists, which leads to considerable disagree-ent [1–3]. Recently, we have proposed that the characteristic

∗ Corresponding author at: Division of Oral Pathology, Department of Tissueegeneration and Reconstruction, Niigata University Graduate School of Medicalnd Dental Sciences, 2-5274 Gakkocho-dori, Chuo-ku, Niigata 951-8514, Japan.

E-mail address: tsaku@dent.niigata-u.ac.jp (T. Saku).

ttp://dx.doi.org/10.1016/j.prp.2016.02.016344-0338/© 2016 Elsevier GmbH. All rights reserved.

two-phase appearance, which results from a sharp and contrastivelayering of the upper keratinized cell layer and the lower halfbasaloid cells, is recognized in some particular histological typesof epithelial dysplasia or CIS [4–12], and it could be an impor-tant histopathological hallmark of potentially malignant epitheliallesions even on HE sections. In the lower half of the two-phaseepithelial dysplasia, composed of basaloid cells which are immuno-histochemically positive for Ki-67 [5,6] as well as podoplanin[12,13], there are enriched intercellular deposits of extracellularmatrix (ECM) molecules such as perlecan, a basement-membranetype heparan sulfate proteoglycan [14–18]. In addition, the basaloid

cells in the lower half showed simultaneous loss of E-cadherin andnuclear translocation of �-catenin from the cell membrane, whichindicates that those basaloid cells form a cell proliferating centerin the lower half [6].

search

tttwfialfc[otipstagcbtwtl[

hncV

2

2

ssUslfswmantaedatswhmN(

2

S

M. Hasegawa et al. / Pathology – Re

To further confirm our hypothesis that the lower half of thewo-phase epithelial dysplasia is a cell proliferation center andhat its histopathological recognition is of considerable help forhe objective differential diagnosis of oral borderline malignancies,e now consider it necessary to investigate the expression pro-les of perlecan-binding growth factors in oral epithelial dysplasiand CIS comparatively in normal epithelia and SCC because per-ecan has been known to be an important extracellular reservoiror several kinds of growth factors or cytokines [19] including vas-ular endothelial growth factor (VEGF) [20], sonic hedgehog (SHH)21], or keratinocyte growth factor (KGF) [22,23]. VEGF, which actsn endothelial cells to promote angiogenesis, is also required forumor cells to proliferate in a cell-autonomous and angiogenesis-ndependent manner [24]. It is known that the SHH signalingathway regulates cell migration, proliferation, and apoptosis ineveral cancer cells from the skin, oral cavity, gastrointestinalracts, urinary bladder, and lung [25]. KGF has been recognized as

mesenchymal cell-derived paracrine mediator of epithelial cellrowth [26], but it is also known to stimulate various carcinomaells from the biliary tract [27] and breast [28], though it has noteen immunolocalized in SCC of the head and neck [29]. Thus,he expression modes of these molecules in oral SCC are some-hat controversial. Their pathophysiological functions also remain

otally unknown during the oral precancerous stages, though per-ecan has been suggested to function in epithelial dysplasia and CIS6,9,14,16].

In this study, our aim was to determine comparative immuno-istochemical profiles in oral mucosal epithelia ranging fromormal to SCC among the following molecules: perlecan; Ki-67, aell cycle marker; such perlecan-binding factors as KGF, SHH andEGF; as well as VEGF receptors Flt-1 and Flk-1.

. Materials and methods

.1. Tissue materials

Eighty-one surgical specimens of SCC or CIS and 10 biopsypecimens of epulis of the oral mucosa were selected from theurgical pathology files in the Division of Oral Pathology, Niigataniversity Graduate School of Medical and Dental Sciences. Each

pecimen simultaneously contained histopathologically differentesions ranging from frankly invasive and well-differentiated SCCoci and foci of CIS, epithelial dysplasia, and epithelial hyperpla-ia to definitely normal epithelial parts. From these specimens,e selected 30 foci of normal and hyperplastic epithelia, 50 ofoderate epithelial dysplasia with the characteristic two-phase

ppearance [4–6], 45 of CIS, and 30 of SCC, all of which were diag-osed on hematoxylin and eosin (HE) stained sections as well as onheir immunohistochemically stained sections for keratin 13 (K13),

prickle cell marker; K19, a basal cell marker; Ki-67, a cell prolif-ration marker; and K17/K16, carcinoma cell markers, as we haveescribed elsewhere [5–9]. The diagnostic criteria used in this studyre described in a separate section. All the specimens were rou-inely fixed in 10% formalin and embedded in paraffin. Serial 3-�mections were cut from paraffin blocks, and one set of the sectionsas stained with HE while the other sets were used for immuno-istochemistry. The experimental protocol for analyzing surgicalaterials was reviewed and approved by the Ethical Board of theiigata University Graduate School of Medical and Dental Sciences

Oral Life Science).

.2. Cells

SCC cell systems (ZK-1, ZK-2, and MK-1) were established fromCC arising in the tongue (ZK-1 and ZK-2) and gingiva (MK-1)

and Practice 212 (2016) 426–436 427

[13]. SCC cells were cultured in Dulbecco’s modified Eagle medium(DMEM) (Gibco, Invitrogen, Thermo Fisher Scientific, Waltham,MA, USA), which contained 10% fetal bovine serum (FBS) (Gibco),50 �g/ml streptomycin, and 50 IU/ml penicillin (Gibco). They wereincubated at 37 ◦C in a humidified 5% carbon dioxide/95% air atmo-sphere.

2.3. Antibodies

Polyclonal antibodies against the mouse basement membrane-type perlecan core protein were raised in rabbits as describedelsewhere (diluted at 50 �g/ml) [14,16]. Mouse monoclonal anti-bodies against VEGF (clone C-1, IgG2a, 1:200), Flk-1 (A-3, IgG1,1:300) and rabbit polyclonal antibodies against KGF (IgG, 1:50)were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz,CA, USA). Rabbit polyclonal antibodies against Flt-1 (IgG, 1:2000)were obtained from Oncogene Research Products (La Jolla, CA, USA)and those against SHH (IgG, 1:100) were obtained from Abcam Inc.(Cambridge, UK). A mouse monoclonal antibody against humanKi-67 (MIB-1, IgG1, 1:50) was obtained from Dako (Glostrup,Denmark).

2.4. Immunohistochemistry

Paraffin sections were subjected to immunohistochemical stain-ings for perlecan core protein, VEGF, KGF, SHH, Flt-1, Flk-1, andKi-67 by using the Envision+/HRP system (Dako). For VEGF, sec-tions were treated with 0.15% trypsin (type II, Sigma ChemicalCo., St Louis, MO, USA) in 10 mM Tris–HCl (pH 7.6) for 30 min at37 ◦C. For SHH, Flt-1, Flk-1 and Ki-67, sections were autoclavedin citric acid buffer (pH 6.0) at 120 ◦C for 10 min. After that, thesections were rinsed in 0.01 M PBS containing 0.5% milk pro-tein (Morinaga Milk Industry Co. Ltd., Tokyo, Japan) and 0.05%Triton X-100 (T-PBS) and treated with 0.3% hydrogen peroxidein methanol for 30 min at room temperature to block endoge-nous peroxidase activities. After rinsing in T-PBS, sections wereincubated with 5% milk protein in T-PBS for 1 h at room temper-ature to block non-specific protein-binding sites. They were thenincubated with the primary antibodies overnight at 4 ◦C. After incu-bation, the sections were rinsed in T-PBS and incubated with thepolymer-immune complexes (EnVision+peroxidase, rabbit/mouse,Dako) for 1 h at room temperature. After rinsing with T-PBS, theywere treated with 0.02% 3,3′-diaminobenzimine (Dohjindo Lab-oratories, Kumamoto, Japan) in 0.05 M Tris–HCl buffer (pH 7.6)containing 0.005% hydrogen peroxide to visualize the reactionproducts. Finally, the sections were counterstained with hema-toxylin. For control studies on antibodies, the primary antibodieswere replaced with pre-immune rabbit IgG or mouse IgG subclasses(Dako).

Following HE staining and immunohistochemistry examina-tions for K13, K19, K17, K16, and Ki-67, performed as describedelsewhere [5–12], all of the focus samples were classified as (i)normal or hyperplastic epithelia, (ii) mild and moderate epithe-lial dysplasia, (iii) CIS, or (iv) SCC. We did not use the category ofsevere dysplasia because we considered that there was no objectivedistinction between so-called severe dysplasia and CIS [5].

2.5. Immunohistochemical evaluation

Foci of SCC, CIS, dysplasia, and normal epithelial parts wereevaluated by extension and intensity of the immunohistochemi-cal reactions for the three perlecan-binding molecules, VEGF, SHH,

and KGF, and compared with those for perlecan. The staining wasevaluated in four epithelial zones—basal, parabasal, lower prickle,and upper prickle layers as indicated in Figs. 1–3—for positiveratios. Each layer was considered positive (+) or not positive (−)

428 M. Hasegawa et al. / Pathology – Research and Practice 212 (2016) 426–436

Fig. 1. Normal oral squamous epithelium. (a) Hematoxylin and eosin (HE) stain; (b) immunoperoxidase stain for extracellular matrix protein perlecan; (c) cell proliferationmarker Ki-67; perlecan-binding factors: (d) KGF; (e) VEGF; (f) SHH; and VEGF receptors: (g) Flt-1; (h) Flk-1, hematoxylin counterstain. (A–H) ×160. In normal (a) or hyperplasticepithelia, perlecan was localized faintly in the parabasal cells layer (b) in which Ki-67 positive (+) cells were located (c). While KGF was not positive in the epithelial layer(d), VEGF was definitely positive in the cytoplasm of basal or parabasal cells, in addition to vascular endothelial cells and other stromal cells in the lamina propria (e). SHHw layersw ition

wliwtIwuReccd

as faintly positive in nuclei of epithelial cells from the basal to lower prickle cell

ithin the nuclei of epithelial cells from the basal to lower prickle cell layers in add

hen a particular molecule was or was not expressed in epithe-ial cells located in those layers without consideration of stainingntensities or extensions. In terms of SCC foci, the four layers

ere basically separated corresponding to normal ones withinhe range from the periphery (basal) to the center (keratinized).n the epithelial zone, both nuclear and cytoplasmic stainings

ere equally counted positive. The +/− judgments were agreedpon by three examiners who were experienced pathologists.esults were expressed as the ratio of positive foci to all the

xamined ones. In addition, rates of immunolocalization in con-ordance between the three growth factors and perlecan werealculated for better understanding their colocalization. Statisticalifferences were determined by a Student’s t-test or a chi-square

as well as in round-shaped stromal cells (f). Flt-1 and Flk-1 were mainly localizedto vascular endothelial cells (g, h).

test for independence. Differences with P < 0.05 were consideredsignificant.

2.6. Immunofluorescence

Immunofluorescence experiments were performed usingNuncTM Lab-TekTM II Chamber Slide System (Thermo Fisher).Cells were plated at the concentration at 1.2 × 104 cells/well andcultivated for 7 days. The cells were washed with PBS and fixed

with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) for30 min on ice. To prevent non-specific protein binding, they wereincubated with 5% milk protein in PBS containing 0.05% TritonX-100 overnight at 4 ◦C. The cells were then incubated with the

M. Hasegawa et al. / Pathology – Research and Practice 212 (2016) 426–436 429

Fig. 2. Oral epithelial dysplasia with a characteristic two-phase appearance. (a) HE stain; immunoperoxidase stain for perlecan (b), Ki-67 (c), KGF (d), VEGF (e), SHH (f), Flt-1(g), and Flk-1 (h), hematoxylin counterstain. (a–h) ×260. In two-phase dysplasia (a), perlecan was positive mainly on the cell border of basaloid cells densely packed in thelower half of the epithelial layer (b) in which Ki-67+ cells were stratified from the first basal layer (c). KGF was faintly positive on the cell border and in the cytoplasm of thel f (e). Sl

p1bfcFaFaFFstm

ower half (d), while VEGF was strongly positive in the cytoplasm in the lower halocalized mainly in the nuclei of the lower-half cells.

rimary antibodies (perlecan, diluted at 50 �g/ml in PBS; VEGF,:100; KGF, 1:50; SHH, 1:50) and further with secondary anti-odies. For double immunofluorescence, cells were firstly stainedor perlecan using secondary goat antibodies against rabbit IgGonjugated with Alex FluorTM 488 (Molecular ProbesTM, Thermoisher), and then sequentially for VEGF using a secondary goat IgGgainst mouse IgG conjugated with Alexa FluorTM 568 (Thermoisher). When secondarily stained for SHH, or KGF, the rabbitntibodies against SHH and KGF were directly labeled with AlexaluorTM 568 without using dye-conjugated secondary antibodies.

inally, cells were counterstained with CellstainTM Hoechst-33258olution (Dojindo) diluted at 1:100 in PBS. For control studies,he primary antibodies were replaced with pre-immune rabbit or

ouse IgGs.

HH was mainly positive in nuclei in the lower half (f). Flt-1 (g) and Flk-1 (h) were

3. Results

Immunohistochemical staining results are separately describedin each category as follows. Table 1 summarizes the ratios of pos-itive foci by layers for VEGF, SHH and KGF (upper row) and theirconcordance rates with perlecan (lower row) in each category.

3.1. Normal/hyperplastic epithelia

In normal and hyperplastic epithelia from SCC/CIS or epulis

specimens (Fig. 1a), perlecan was faintly positive in and above theparabasal cells layer (Fig. 1b), where Ki-67 positive (+) cells weresporadically located (Fig. 1c). Perlecan was occasionally positive inthe lamina propria connective tissue. Such immunohistochemical

430 M. Hasegawa et al. / Pathology – Research and Practice 212 (2016) 426–436

Fig. 3. Oral carcinoma in situ (CIS). (a) HE stain; immunoperoxidase stain for perlecan (b), Ki-67 (c), KGF (d), VEGF (e), SHH (f), Flt-1 (g), and Flk-1 (h), hematoxylin counterstain.(a–h) ×240. In CIS (a), perlecan was localized on the cell border in the whole epithelial layer except for a few cell layers just beneath the surface keratinized layer (b). Ki-67+cells were spread in the perlecan+ rete ridge area (c). KGF was positive on the cell border as well as in the cytoplasm in the perlecan+ area (d). VEGF was expressed mainlyi four

l nized

r

pitbopcpipltei

n the cytoplasm in the perlecan+ area (e). SHH+ areas were nearly the same as theocalized in the nuclei and cytoplasm in the perlecan+ area including surface keratiete ridges (h).

rofiles were stably observed even in hyperplastic epithelia cover-ng the epulis (not shown). While KGF was not definitely positive inhe epithelial layer (Fig. 1d), VEGF was positive in the cytoplasm ofasal or parabasal cells, in addition to vascular endothelial cells andther stromal cells in the lamina propria (Fig. 1e). SHH was faintlyositive in nuclei of epithelial cells from the basal to lower prickleell layers as well as in round-shaped stromal cells in the laminaropria (Fig. 1f). Flt-1 and Flk-1, VEGF-receptors, were mainly local-

zed within the nuclei of epithelial cells from the basal to lowerrickle cell layers, and Flt-1 was also localized in vascular endothe-

ial cells (Fig. 1g, h). The immunohistochemical positivities for thehree growth factors were compared with those of perlecan by fourpithelial layers in every category of epithelial lesions as shownn Table 1. Their colocalizations with perlecan were not observed

molecules mentioned above, though SHH was localized in the nuclei (f). Flt-1 waslayer (g), while Flk-1 was mainly positive in the nuclei of the lower half zone of the

because perlecan was not definitely expressed either. It was thussuggested that VEGF and SHH signals mainly expressed in the basaland parabasal layers of normal/hyperplastic epithelia were medi-ated by ligands other than perlecan.

3.2. Epithelial dysplasia

In epithelial dysplasia with the characteristic two-phase appear-ance (Fig. 2a), perlecan was positive mainly on the cell border (inthe intercellular space) of the lower half of the epithelial layer, in

addition to the subepithelial connective tissue (Fig. 2b). Ki-67+ cellswere stratified up to the fifth layers from the bottom (Fig. 2c).The perlecan localization showed a meshwork-like appearance.In the same manner, KGF was faintly positive on the cell border

M. Hasegawa et al. / Pathology – Research

Tab

le

1C

omp

arat

ive

imm

un

ohis

toch

emic

al

pos

itiv

e

rati

os

for

per

leca

n

vs. i

ts

bin

din

g

grow

th

fact

ors

by

fou

r

epit

hel

ial l

ayer

s

in

oral

squ

amou

s

epit

hel

ial l

esio

ns.

Posi

tive

rati

os

of

per

leca

n-b

ind

ing

grow

th

fact

ors

by

fou

r

epit

hel

ial l

ayer

s

(%, l

eft)

and

thei

r

rati

o

for

con

cord

ance

wit

h

per

leca

n

loca

liza

tion

(%, r

igh

t,

in

par

enth

eses

)

Lesi

ons

Focu

sn

um

ber

VEG

F

SHH

KG

F

Perl

ecan

(%)

Bas

al

Para

basa

l

Low

erp

rick

leU

pp

erp

rick

leB

asal

Para

basa

l

Low

erp

rick

leU

pp

erp

rick

leB

asal

Para

basa

l

Low

erp

rick

leU

pp

erp

rick

leB

asal

Para

basa

l Lo

wer

pri

ckle

Up

per

pri

ckle

Nor

mal

/hyp

erp

last

icep

ith

elia

30

53

(–)a

40

(100

)b13

(100

)

3

(–)

65

(–)

41

(100

)

12

(100

)

0

(100

)

0

(100

)

0

(–)

0

(–)

0

(100

)

0

8 8

0

Epit

hel

ial

dys

pla

sia,

mod

erat

e

50

100

(100

)

100

(100

)

86

(90)

36

(41)

89

(93)

86

(90)

71

(74)

29

(33)

62

(65)

62

(65)

62

(65)

62

(71)

96

96

96

88

Car

cin

oma

in

situ

45

100

(100

)62

(63)

62

(63)

98

(100

)10

0

(100

)10

0

(100

)94

(96)

91

(98)

95

(97)

95

(97)

95

(97)

95

(100

) 98

98

98

93Sq

uam

ous

cell

carc

inom

a30

100

(–)

100

(–)

100

(–)

97

(–)

100

(–)

100

(–)

100

(–)

100

(–)

0

(100

)

0

(100

)

0

(100

)

0 (1

00)

0

0

0

0

Tota

l nu

mbe

r

155

aN

ot

iden

tica

l to

per

leca

n

loca

liza

tion

.b

Iden

tica

l to

per

leca

n

loca

liza

tion

(max

imu

m

valu

e:

100%

).

and Practice 212 (2016) 426–436 431

of the lower half (Fig. 2d). VEGF was more strongly positive inthe lower half (Fig. 2e), in which SHH was mainly positive in thenuclei of epithelial cells (Fig. 2f). Flt-1 (Fig. 2g) and Flk-1 (Fig. 2h)were localized mainly in the nuclei of the lower half. The rates ofimmunolocalization in concordance between the three growth fac-tors and perlecan were 33–100% in the lower two layers, thoughthose in the upper prickle cell layer were not so stable betweenthe growth factors (Table 1). The differences on the positive ratiosfor the three molecules in epithelial dysplasia were significantlyhigher than normal/hyperplastic epithelia (P < 0.001).

3.3. CIS

In CIS (Fig. 3a), perlecan was localized on the cell border in thewhole epithelial layer except for a few cell layers just beneath thekeratinized surface layer, in addition to the subepithelial connec-tive tissue (Fig. 3b). The meshwork-like appearance of perlecan wasmore intensive and expansive than that in dysplasia. Since Ki-67+cells were spread in perlecan+ rete ridge parts in CIS (Fig. 3c), itwas confirmed that perlecan was expressed in the cell proliferat-ing zone in every category of oral squamous epithelia from normaland dysplasia up to CIS. KGF was positive on the cell border aswell as in the cytoplasm in the perlecan+ area (Fig. 3d). VEGF wassimilarly expressed mainly in the cytoplasm in the perlecan+ area(Fig. 3e). Although SHH was localized in the nuclei, SHH+ areaswere nearly the same as the areas positive for the four moleculesmentioned above (Fig. 3f). Flt-1 was localized in both the nucleiand cytoplasm in the perlecan+ area including the keratinized sur-face layer (Fig. 3g), while Flk-1 was mainly positive in nuclei of thelower half of rete ridges (Fig. 3h). VEGF, its receptors, and SHH werealso localized in vascular endothelial cells in the stroma (Fig. 3e–h).The concordance rates between the three growth factors and per-lecan were 63–100% in the whole layers (Table 1). The concordancesrates for KGF and SHH in CIS were significantly higher than thosein epithelial dysplasia, and the differences were statistically sig-nificant in each layer (P < 0.001), while there were no significantdifferences for VEGF.

3.4. SCC

In SCC (Fig. 4a), perlecan (Fig. 4b) and KGF (Fig. 4d) were focallylocalized in the stromal connective tissue space around invadingSCC foci but not in SCC cells (Fig. 4b), most of which were posi-tive for Ki-67 (Fig. 4c). In contrast, VEGF was strongly positive inthe cytoplasm of SCC cells and of stromal cells, including vascularendothelial cells (Fig. 4e). SHH was also positive in the nuclei andin the cytoplasm of SCC cells as well as in those of stromal cells(Fig. 4f). Flt-1 was positive in the cytoplasm of SCC cells (Fig. 4g),and Flk-1 was localized both in the nuclei and cytoplasm of SCCcells (Fig. 4h). Both of the VEGF receptors were positive in vas-cular endothelial cells in the stroma (Fig. 4g, h). The concordancerates between VEGF/SHH and perlecan were nearly 0% in the wholelayers of SCC foci, while those between KGF and perlecan were65–100% not only within SCC foci but also in the stromal space(Table 1).

3.5. Immunofluorescence in oral SCC cells in culture

The immunofluorescence signals for perlecan and three growthfactors were compared in the three SCC cell systems. At day 3 afterseeding, when ZK-1 cells formed small colonies (Fig. 5a–c), per-lecan was localized both in the perinuclear zone in the cytoplasm

as well as in the peripheral cell border of external ends (Fig. 5a).KGF showed almost similar localizations in the perinuclear zone aswell as in the external cell border (Fig. 5b). Merged images for thetwo molecules obviously showed their colocalizations especially in

432 M. Hasegawa et al. / Pathology – Research and Practice 212 (2016) 426–436

Fig. 4. Oral squamous cell carcinoma (SCC). (a) HE stain; immunoperoxidase stain for perlecan (b), Ki-67 (c), KGF (d), VEGF (e), SHH (f), Flt-1 (g), and Flk-1 (h), hematoxylincounterstain. (a–h) ×160. In invading fronts of SCC (a), perlecan (b) and KGF (d) were localized in the stromal connective tissue space but not in trabecular SCC cell nests,w in SCa (g), an

tcafisbwtKcb(ts

hich were packed with Ki-67+ SCC cells (c). In contrast, VEGF was strongly positivelso positive in SCC cells (f). Flt-1 was weakly positive in the cytoplasm of SCC cells

he external ends (Fig. 5c, arrows). Signals for VEGF were also colo-alized with those for perlecan (Fig. 5d, f) in the perinuclear spaces well as in the nuclei in addition to on the cell border (Fig. 5e,, arrow, external end; arrowhead, intercellular). Similar colocal-zation patterns including those in the external ends suggestive ofuch cellular processes as filopodia or lamellipodia were obtainedetween perlecan (Fig. 5g, i) and SHH (Fig. 5h, i). In MK-1 cells,hich were taller than ZK-1 and formed more condensed aggrega-

ion within colonies, thick dot-like signals for perlecan (Fig. 5j) andGF (Fig. 5 k) were colocalized in the perinuclear to intercellularell border (Fig. 5l, arrowhead). Similar tendencies in colocalization

etween perlecan (Fig. 5m) and VEGF (Fig. 5n) or between perlecanFig. 5p) and SHH (Fig. 5q) were observed in the central zone ofhe colonies (Fig. 5o, r, arrowheads). ZK-2 showed signal patternsimilar to ZK-1 (not shown). The three cell types showed signals

C cells, in addition to stromal cells including vascular endothelial cells (e). SHH wasd Flk-1 was apparently localized in both the nuclei and cytoplasm of SCC cells (h).

for those four molecules at day 5 when cells reached their conflu-ency (not shown). Thus, different from immunoperoxidase stainingresults in tissue sections, the biosynthesis of KGF, VEGF, and SHHwere confirmed in all of the three oral SCC cell types, and theirimmunofluorescence signals were focally colocalized with perlecansignals.

4. Discussion

We have for the first time demonstrated the immunohis-tochemical profiles of perlecan-binding growth factors in the

developmental process of oral epithelial malignancies from epithe-lial dysplasia to SCC. Before invasion or up to the stage of CIS,the expressions of perlecan and perlecan-binding growth fac-tors were overlapped, and they spread within the epithelial

M. Hasegawa et al. / Pathology – Research and Practice 212 (2016) 426–436 433

Fig. 5. Double immunofluorescence for perlecan and its binding growth factors in oral SCC cells in culture. (a–i) ZK-1 cells, (j–r) MK-1 cells, (a, d, g, j, m, p) perlecan, (b, k) KGF,(e, n) VEGF, (h, q) SHH, (c, f, i, l, o, r) merges of perlecan and KGF, VEGF, and SHH, nuclear counterstain with Hoechst-33258, no counterstain in single immunofluorescence,(a–r) ×680. At day 3 after seeding, when ZK-1 cells formed small colonies (a–c), perlecan was localized both in the perinuclear zone in the cytoplasm as well as in theperipheral cell border of external ends (a). KGF showed almost similar localizations in the perinuclear zone as well as in the external cell border (b). Merged images for thetwo molecules obviously showed their colocalizations especially in the external ends (c, arrows). VEGF signals were also colocalized with those for perlecan in the perinuclearspace (e) as well as in the nuclei in addition to on the cell border (f, arrow, external end; arrowhead, intercellular). Similar colocalization patterns were obtained betweenp , arrow( d VEGt

lztc

erlecan (g) and SHH (h, i, arrow). In MK-1 cells, thick dot-like signals for perlecan (jl, arrowhead). Similar tendencies in colocalization between perlecan (m, arrow) anhe central zone of the colonies (o, r, arrowheads).

ayer, though the growth factors were confined to the basalone in normal epithelia. Once carcinoma cells started to invade,he expressions of perlecan and KGF were converted from car-inoma cells to stromal cells, while VEGF and SHH remained

) and KGF (k, arrow) were colocalized in the perinuclear to intercellular cell borderF (n, arrow) or between perlecan (p, arrow) and SHH (q, arrow) were observed in

only in the SCC foci, as these three growth factors were con-firmed to be biosynthesized in oral SCC cells in culture. Thepresent histological study indicates that perlecan plays importantroles in oral epithelial dysplasia, CIS, and SCC by differentially

4 search

mi

pStulaKttflfuBte[afef

solislscnbrsbTtPiatcptsaitlctsogaf

hhKapk

34 M. Hasegawa et al. / Pathology – Re

odulating the perlecan-bound growth signals before and afternvasion.

In the present study, the upper half and the lower half of two-hase dysplasia were well contrasted by the presence of VEGF,HH, and KGF in the lower half, which was in accordance withhe intercellular deposit of perlecan, and by their absence in thepper half. The results clearly indicate that the perlecan-enriched

ower half is also enriched with perlecan-binding growth factors,nd that the lower half is therefore optimized for proliferation ofi-67+ basaloid cells [4,5,11]. In other words, the lower half of

he two-phase appearance, which is characterized by the simul-aneous loss of E-cadherin and nuclear translocation of �-cateninrom the cell membrane [6], can be regarded as a distinct cell pro-iferating center in epithelial dysplasia. These molecular devicesor cellular proliferation seem to be correlated from each othernder the circumstance of the intercellular deposits of perlecan.ased on such molecular crosstalk via perlecan, we have proposedhe concept of the intraepithelial stroma [30] not only in oralpithelial lesions but also in odontogenic organs [31] or tumors32]. With the increase in dysplastic grades, the Ki-67+/perlecan+rea expanded together with the areas which were also positiveor VEGF, SHH and KGF. Finally in the stage of CIS, the wholepithelial layer became positive for Ki-67, perlecan, and the growthactors.

The roles of perlecan in cancer cell growth, invasion, metasta-is and angiogenesis have been well documented in various typesf tumors including human oral [33], salivary [34], breast [35], andiver [36] carcinomas or melanoma [37]. However, most of the stud-es on the function of perlecan were performed in single cell cultureystems, namely in circumstances in which cancer cells are iso-ated from or not in contact with any other types of cells, includingtromal fibroblasts. Therefore, these experimental conditions mustorrespond at tissue levels with CIS in which carcinoma cells areot exposed to stromal cells [14,16,38]. When perlecan is secretedy parenchymal (carcinoma) cells before invasion [16,17,33], it iseasonable to expect perlecan-binding molecules function at theame time within the parenchymal space. A variety of perlecan-inding molecules, including VEGF, SHH, FGF, EGF, PDGF, andGF-�, are basically regarded as tumor cell growth factors, thoughheir perlecan-biding modes are different from each other [39,40].erlecan consists of a core protein with a molecular mass of approx-mately 500 kDa and three major heparan sulfate (HS) chains whichre attached to domain I of the core protein [18,41]. VEGF bindso HS chains [19], while SHH binds to both HS chains and theore protein [20], and KGF binds to domains III and V of the corerotein of perlecan [21]. However, their perlecan-association situa-ions have never been investigated at the tissue levels. The presenttudy has revealed that perlecan and its binding growth factorsre at least colocalized in epithelial dysplasia and CIS foci beforenvasion, but after that, those growth factors do not always behaveogether with carcinoma cells because the biosynthesis of per-ecan has been demonstrated to be switched over from carcinomaells to stromal cells in co-culture experiments [33]. Similar tohe perlecan-binding growth factors, perlecan receptors have beenhown to switch from �-dystroglycan to integrin �1 on invasion ofral SCC [16]. In prostate carcinomas, KGF and VEGF of stromal ori-in have been shown to mediate interactions between tumor cellsnd stromal cells to induce secretion of ECM-degrading enzymesor invasion [37].

Similar to our present results, the absence of KGF expressions inead and neck SCC cells and their presence in stromal fibroblastsas already been reported [27]. We have also demonstrated that

GF is enriched in the stromal space with spindle cells, which arective in proliferation, in the invading front of salivary pleomor-hic adenomas [42]. Since KGF [43], VEGF [44], and SHH [45] arenown to be produced by both epithelial and mesenchymal cells,

and Practice 212 (2016) 426–436

their biosynthesis switching from carcinoma cells to stromal cellson invasion is not so surprising, though it is unknown at presentwhat sorts of molecular mechanisms regulate VEGF and SHH toremain in carcinoma cells, or KGF to move to stromal cells afterinvasion.

As to VEGF, Iozzo and his group have shown that angiogenesisin the zebrafish embryonic development as well as prolifera-tion of human endothelial cells, both of which were dependenton VEGF-VEGF-receptors, were modulated by perlecan [46]. Inprostate carcinomas, heparin-binding growth factors, includingVEGF or FGF-2, are reduced in the absence of perlecan, suggest-ing that the VEGF signaling is controlled by perlecan [47]. Inthe present study, the immunohistochemical expressions of Flt-1 and Flk-1 were similarly related to VEGF/perlecan expressionsin oral epithelial dysplasia, CIS, and SCC. There have been sev-eral studies reporting differential expressions between these twoVEGF receptors in malignancies, in addition to angiogenetic func-tions [48]. An autocrine manner of VEGF signaling via Flt-1 hasalready been shown to function in carcinogenesis and prolifera-tion of epidermal tumor cells [23], in proliferation of pleomorphicadenoma cells [49], or in migration and invasion of pancreaticcarcinoma cells [50]. Since Flk-1 and VEGF were found in dys-plastic nodules of the liver [51], the Flk-1 expressions could berelated to cell proliferation. Thus, the VEGF signaling via the tworeceptors may be different from tumor to tumor or from organto organ.

From the present results, it is now obvious that varieties of per-lecan signaling play important roles in the SCC growth both beforeand after invasion. Since the functional modes are differentially reg-ulated before and after invasion at the tissue level, it is necessaryto confirm the phenomena in in vitro studies using co-culture sys-tems before the whole molecular mechanism of oral SCC invasionmediated by perlecan is fully understood.

5. Conclusions

The present study demonstrated the significance of intercellu-lar deposit of perlecan, a basement-membrane type heparan sulfateproteoglycan in oral precancerous lesions and SCC. Before invasion,namely in oral epithelial dysplasia and CIS, VEGF, SHH, KGF, Flt-1,and Flk-1were colocalized in the lower half of rete ridges whereperlecan is enriched and Ki-67+ proliferating cells are condensed.After invasion, perlecan and KGF disappeared from SCC cells butemerged in the stromal space, while VEGF, SHH, and VEGF recep-tors remained in SCC cells. Since we have reported that perlecanbiosynthesis was switched from CIS cells to stromal fibroblasts onand after invasion of SCC [14,17,33], and that the switching wasdifferentially correlated with that of such perlecan receptors as dys-troglycan and integrin �1 [16]. Immunohistochemistry must be oneand only tool to demonstrate such switching phenomena in tissuesamples. It is now reasonably explained that perlecan is requiredfor recruiting growth factors to oral SCC/CIS/dysplasia cells for theirproliferation and invasion.

6. Conflicts of interest

We declare that we have no conflicts of interest.

Acknowledgments

This work was supported in part by Grants-in-Aid for ScientificResearch from the Japan Society for the Promotion of Science (JSPSKAKENHI grant nos. 25305035 and 25462849 to J.C.; 23406038,26305032, and 15K15693 to T.S.).

search

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

M. Hasegawa et al. / Pathology – Re

eferences

[1] J.J. Pindborg, P.A. Reichart, C.J. Smith, I. van der Waal, Precancerous lesions, in:J.J. Pindborg, P.A. Reichart, C.J. Smith, I. van der Waal (Eds.), HistologicalTyping of Cancer and Precancer of the Oral Mucosa, World HealthOrganization International Histological Classification of Tumours, 2nd ed.,Springer-Verlag, Berlin, 1997, pp. 24–26.

[2] I.R. Kramer, R.B. Lucas, J.J. Pindborg, L.H. Sobin, Definition of leukoplakia andrelated lesions: an aid to studies on oral precancer, Oral Surg. Oral Med. OralPathol. 46 (1978) 518–539.

[3] N. Gale, B.Z. Pilch, D. Sidransky, A. El-Naggar, W. Westra, J. Califano, N.Johnson, D.G. MacDonald, Epithelial precursor lesions, in: L. Barnes, J.W.Eveson, P. Reichart, D. Sidransky (Eds.), World Health OrganizationClassification of Tumours, Pathology & Genetics, Head and Neck Tumours,IARC, Lyon, 2005, pp. 177–179.

[4] M. Syafriadi, J. Cheng, K.Y. Jen, H. Ida-Yonemochi, M. Suzuki, T. Saku,Two-phase appearance of oral epithelial dysplasia resulting from focalproliferation of parabasal cells and apoptosis of prickle cells, J. Oral Pathol.Med. 34 (2005) 140–149.

[5] T. Kobayashi, S. Maruyama, J. Cheng, H. Ida-Yonemochi, M. Yagi, R. Takagi, T.Saku, Histopathological varieties of oral carcinoma in situ: diagnosis aided byimmunohistochemistry dealing with the second basal cell layer as theproliferating center of oral mucosal epithelia, Pathol. Int. 60 (2010) 156–166.

[6] C.G. Alvarado, S. Maruyama, J. Cheng, H. Ida-Yonemochi, T. Kobayashi, M.Yamazaki, R. Takagi, T. Saku, Nuclear translocation of �-catenin synchronizedwith loss of E-cadherin in oral epithelial dysplasia with a characteristictwo-phase appearance, Histopathology 59 (2011) 283–291.

[7] T. Mikami, J. Cheng, S. Maruyama, T. Kobayashi, A. Funayama, M. Yamazaki,H.A. Adeola, L. Wu, S. Shingaki, C. Saito, T. Saku, Emergence of keratin 17 vs.loss of keratin 13: their reciprocal immunohistochemical profiles in oralcarcinoma in situ, Oral Oncol. 47 (2011) 497–503.

[8] A. Funayama, J. Cheng, S. Maruyama, M. Yamazaki, T. Kobayashi, M. Syafriadi,S. Kundu, S. Shingaki, C. Saito, T. Saku, Enhanced expression of podoplanin inoral carcinomas in situ and squamous cell carcinomas, Pathobiology 78(2011) 171–180.

[9] H. Ida-Yonemochi, S. Maruyama, T. Kobayashi, M. Yamazaki, J. Cheng, T. Saku,Loss of keratin 13 in oral carcinoma in-situ: a comparative study of proteinand gene expression levels using paraffin sections, Mod. Pathol. 25 (2012)784–794.

10] A. Funayama, S. Maruyama, M. Yamazaki, K. Al-Eryani, S. Shingaki, C. Saito, J.Cheng, T. Saku, Intraepithelially entrapped blood vessels in oral carcinomain-situ, Virchows Arch. 460 (2012) 473–480.

11] T. Kobayashi, S. Maruyama, T. Abe, J. Cheng, R. Takagi, C. Saito, T. Saku, Keratin10-positive orthokeratotic dysplasia: a new leucoplakia-type precancerousentity of the oral mucosa, Histopathology 61 (2012) 910–920.

12] T. Mikami, S. Maruyama, T. Abe, T. Kobayashi, M. Yamazaki, A. Funayama, S.Shingaki, T. Kobayashi, J. Cheng, T. Saku, Keratin 17 is co-expressed with14-3-3 sigma in oral carcinoma in-situ and squamous cell carcinoma andmodulates cell proliferation and size but not cell migration, Virchows Arch.466 (2015) 559–569.

13] M. Tsuneki, M. Yamazaki, S. Maruyama, J. Cheng, T. Saku,Podoplanin-mediated cell adhesion through extracellular matrix in oralsquamous cell carcinoma, Lab. Invest. 93 (2013) 921–932.

14] T. Ikarashi, H. Ida-Yonemochi, K. Ohshiro, J. Cheng, T. Saku, Intraepithelialexpression of perlecan, a basement membrane-type heparan sulfateproteoglycan reflects dysplastic changes of the oral mucosal epithelium, J.Oral Pathol. Med. 33 (2004) 87–95.

15] W.M. Tilakaratne, T. Kobayashi, H. Ida-Yonemochi, W. Swelam, M. Yamazaki,T. Mikami, C.G. Alvarado, A.M. Shahidul, S. Maruyama, J. Cheng, T. Saku,Matrix metalloproteinase 7 and perlecan in oral epithelial dysplasia andcarcinoma in situ: an aid for histopathologic recognition of their cellproliferation centers, J. Oral Pathol. Med. 38 (2009) 348–355.

16] M.S. Ahsan, M. Yamazaki, S. Maruyama, T. Kobayashi, H. Ida-Yonemochi, M.Hasegawa, A.H. Ademola, J. Cheng, T. Saku, Differential expression of perlecanreceptors, �-dystroglycan and integrin �1, before and after invasion of oralsquamous cell carcinoma, J. Oral Pathol. Med. 40 (2011) 552–559.

17] S. Maruyama, Y. Shimazu, T. Kudo, K. Sato, M. Yamazaki, T. Abé, H. Babkair, J.Cheng, T. Aoba, T. Saku, Three-dimensional visualization of perlecan-richneoplastic stroma induced concurrently with the invasion of oral squamouscell carcinoma, J. Oral Pathol. Med. 43 (2014) 627–636.

18] S. Maruyama, M. Itagaki, H. Ida-Yonemochi, T. Kubota, M. Yamazaki, T. Abé, H.Yoshie, J. Cheng, T. Saku, Perlecan-enriched intercellular space of junctionalepithelium provides primary infrastructure for leukocyte migration throughsquamous epithelial cells, Histochem. Cell Biol. 142 (2014) 297–305.

19] J.M. Whitelock, J. Melrose, R.V. Iozzo, Diverse cell signaling events modulatedby perlecan, Biochemistry 47 (2008) 11174–11183.

20] A. Muthusamy, C.R. Cooper, R.R. Gomes Jr., Soluble perlecan domain Ienhances vascular endothelial growth factor-165 activity and receptorphosphorylation in human bone marrow endothelial cells, BMC Biochem. 11(2010) 43.

21] S. Datta, M. Pierce, M.W. Datta, Perlecan signaling: helping hedgehog

stimulate prostate cancer growth, Int. J. Biochem. Cell Biol. 38 (2006)1855–1861.

22] M. Mongiat, K. Taylor, J. Otto, S. Aho, J. Uitto, J.M. Whitelock, R.V. Iozzo, Theprotein core of the proteoglycan perlecan binds specifically to fibroblastgrowth factor-7, J. Biol. Chem. 275 (2000) 7095–7100.

[

and Practice 212 (2016) 426–436 435

23] M. Mongiat, J. Otto, R. Oldershaw, F. Ferrer, J.D. Sato, R.V. Iozzo, Fibroblastgrowth factor-binding protein is a novel partner for perlecan protein core, J.Biol. Chem. 276 (2001) 10263–10271.

24] B.M. Lichtenberger, P.K. Tan, H. Niederleithner, N. Ferrara, P. Petzelbauer, M.Sibilia, Autocrine vegf signaling synergizes with EGFR in tumor cells topromote epithelial cancer development, Cell 140 (2010) 268–279.

25] P.A. Beachy, S.S. Karhadkar, D.M. Berman, Tissue repair and stem cell renewalin carcinogenesis, Nature 432 (2004) 324–331.

26] T. Yamayoshi, T. Nagayasu, K. Matsumoto, T. Abo, Y. Hishikawa, T. Koji,Expression of keratinocyte growth factor/fibroblast growth factor-7 and itsreceptor in human lung cancer: correlation with tumour proliferative activityand patient prognosis, J. Pathol. 204 (2004) 110–118.

27] R. Amano, N. Yamada, Y. Doi, M. Yashiro, M. Ohira, A. Miwa, K. Hirakawa,Significance of keratinocyte growth factor receptor in the proliferation ofbiliary tract cancer, Anticancer Res. 30 (2010) 4115–4121.

28] Y. Hishikawa, N. Tamaru, K. Ejima, T. Hayashi, T. Koji, Expression ofkeratinocyte growth factor and its receptor in human breast cancer: itsinhibitory role in the induction of apoptosis possibly through theoverexpression of bcl-2, Arch. Histol. Cytol. 67 (2004) 455–464.

29] B. Knerer, M. Formanek, B. Schickinger, H. Martinek, J. Kornfehl, Different KGFexpression in squamous cell carcinomas of the head and neck and in normalmucosa, Acta Otolaryngol. 118 (1998) 438–442.

30] H. Ida-Yonemochi, M. Nakajima, T. Saku, Heparanase, heparan sulfate andperlecan distribution along with the vascular penetration during stellatereticulum retraction in the mouse enamel organ, Arch. Oral Biol. 55 (2010)778–787.

31] H. Ida-Yonemochi, K. Ohshiro, W. Swelam, H. Metwaly, T. Saku, Perlecan, abasement membrane-type heparan sulfate proteoglycan, in the enamelorgan: its intraepithelial localization in the stellate reticulum, J. Histochem.Cytochem. 53 (2005) 763–772.

32] H. Ida-Yonemochi, M.S. Ahsan, T. Saku, Differential expression profilesbetween �-dystroglycan and integrin �1 in ameloblastoma: two possibleperlecan signalling pathways for cellular growth and differentiation,Histopathology 58 (2011) 234–245.

33] H. Metwaly, S. Maruyama, M. Yamazaki, M. Tsuneki, T. Abe, K.Y. Jen, J. Cheng,T. Saku, Parenchymal-stromal switching for extracellular matrix productionon invasion of oral squamous cell carcinoma, Hum. Pathol. 43 (2012)1973–1981.

34] S. Kimura, J. Cheng, K. Toyoshima, K. Oda, T. Saku, Basement membraneheparan sulfate proteoglycan (perlecan) synthesized by ACC3, adenoid cysticcarcinoma cells of human salivary gland origin, J. Biochem. 125 (1999)406–413.

35] V.I. Guelstein, T.A. Tchypysheva, V.D. Ermilova, A.V. Ljubimov, Myoepithelialand basement membrane antigens in benign and malignant human breasttumors, Int. J. Cancer 53 (1993) 269–277.

36] T. Roskams, R. De Vos, G. David, B. Van Damme, V. Desmet, Heparan sulphateproteoglycan expression in human primary liver tumours, J. Pathol. 185(1998) 290–297.

37] I.R. Cohen, A.D. Murdoch, M.F. Naso, D. Marchetti, D. Berd, R.V. Iozzo,Abnormal expression of perlecan proteoglycan in metastatic melanomas,Cancer Res. 54 (1994) 5771–5774.

38] A. Kaminski, J.C. Hahne, el-M. Haddouti, A. Florin, A. Wellmann, N. Wernert,Tumour-stroma interactions between metastatic prostate cancer cells andfibroblasts, Int. J. Mol. Med. 18 (2006) 941–950.

39] H. Ida-Yonemochi, I. Satokata, H. Ohshima, T. Sato, M. Yokoyama, Y. Yamada,T. Saku, Morphogenetic roles of perlecan in the tooth enamel organ: ananalysis of overexpression using transgenic mice, Matrix Biol. 30 (2011)379–388.

40] J. Kruegel, N. Miosge, Basement membrane components are key playersin specialized extracellular matrices, Cell. Mol. Life Sci. 67 (2010)2879–2895.

41] R.V. Iozzo, I.R. Cohen, S. Grassel, A.D. Murdoch, The biology of perlecan: themultifaceted heparan sulphate proteoglycan of basement membranes andpericellular matrices, Biochem. J. 302 (Pt. 3) (1994) 625–639.

42] S. Maruyama, J. Cheng, M. Yamazaki, A. Liu, T. Saku, Keratinocyte growthfactor colocalized with perlecan at the site of capsular invasion and vascularinvolvement in salivary pleomorphic adenomas, J. Oral Pathol. Med. 38 (2009)377–385.

43] J.S. Rubin, D.P. Bottaro, M. Chedid, T. Miki, D. Ron, G.R. Cunha, P.W. Finch,Keratinocyte growth factor as a cytokine that mediatesmesenchymal–epithelial interaction, EXS 74 (1995) 191–214.

44] M. Klagsbrun, P.A. D’Amore, Vascular endothelial growth factor and itsreceptors, Cytokine Growth Factor Rev. 7 (1996) 259–270.

45] J. Mao, B.M. Kim, M. Rajurkar, R.A. Shivdasani, A.P. McMahon, Hedgehogsignaling controls mesenchymal growth in the developing mammaliandigestive tract, Development 137 (2010) 1721–1729.

46] J.J. Zoeller, J.M. Whitelock, R.V. Iozzo, Perlecan regulates developmentalangiogenesis by modulating the vegf-vegfr2 axis, Matrix Biol. 28 (2009)284–291.

47] C. Savorè, C. Zhang, C. Muir, R. Liu, J. Wyrwa, J. Shu, H.E. Zhau,L.W. Chung, D.D. Carson, M.C. Farach-Carson, Perlecan knockdown in

metastatic prostate cancer cells reduces heparin-binding growth factorresponses in vitro and tumor growth in vivo, Clin. Exp. Metastasis 22 (2005)377–390.

48] G. Neufeld, T. Cohen, S. Gengrinovitch, Z. Poltorak, Vascular endothelialgrowth factor (VEGF) and its receptors, FASEB J. 13 (1999) 9–22.

4 search

[

[

36 M. Hasegawa et al. / Pathology – Re

49] W. Swelam, H. Ida-Yonemochi, S. Maruyama, K. Ohshiro, J. Cheng, T. Saku,

Vascular endothelial growth factor in salivary pleomorphic adenomas: one ofthe reasons for their poorly vascularized stroma, Virchows Arch. 446 (2005)653–662.

50] J.S. Wey, F. Fan, M.J. Gray, T.W. Bauer, M.F. McCarty, R. Somcio, W. Liu, D.B.Evans, Y. Wu, D.J. Hicklin, L.M. Ellis, Vascular endothelial growth factor

[

and Practice 212 (2016) 426–436

receptor-1 promotes migration and invasion in pancreatic carcinoma cell

lines, Cancer 104 (2005) 427–438.

51] K. Nakamura, Y. Zen, Y. Sato, K. Kozaka, O. Matsui, K. Harada, Y. Nakanuma,Vascular endothelial growth factor, its receptor flk-1, and hypoxia induciblefactor-1alpha are involved in malignant transformation in dysplastic nodulesof the liver, Hum. Pathol. 38 (2007) 1532–1546.