Molecular and Cellular Pathobiology

Distinct Functions of Epidermal andMyeloid-Derived VEGF-A in Skin TumorigenesisMediated by HPV8Xiaolei Ding1, Tina Lucas1, Gian P. Marcuzzi2, Herbert Pfister2, and Sabine A. Eming1,3,4

Abstract

Beta humanpapillomaviruses (HPV) havebeen suspected to becarcinogenic in nonmelanoma skin cancers (NMSC), but the basisfor potential viral contributions to these cancers is poorly under-stood. In particular, it is unresolved how HPV-infected keratino-cytes escape cell-cycle control and whether their cross-talk withimmune cells is critical for tumorigenesis. In nonviral preclinicalmodels, the angiogenic cytokine VEGF-A has been identified as acritical regulator of NMSC. In this study, we dissected the contri-bution of epidermal versusmyeloid cell–derived VEGF-A inHPV-mediated skin cancer by interbreeding anHPV8 transgenicmousemodel with a conditional disruption of VEGF-A restricted to

either epidermal or myeloid cells. Although only epidermal-derived VEGF-A was essential for initiation of skin tumor devel-opment, both spontaneously and UV-light triggered, both epi-dermal and myeloid cell–derived VEGF-A contributed to regen-eration-induced tumorigenesis uponHPV8overexpression, partlynot only through a paracrine effect on endothelial cells, but alsomost probably through an additional autocrine effect on epider-mal cells.Ourfindings offer newmechanistic insights into distinctfunctions of epidermal versus myeloid cell–derived VEGF-A dur-ing HPV-mediated tumorigenesis, with possible implications forpreventing this disease. Cancer Res; 75(2); 330–43. �2014 AACR.

IntroductionHuman papillomaviruses (HPV) belong to a family of small

DNA viruses that can infect the epithelia of skin and mucosa.Viral persistence in the epithelium can lead to the developmentof benign or malignant proliferative lesions. To date, over 170different types of HPVs have been identified (1). The causalrelationship between a-genus HPV infection of mucosa(e.g., HPV16) and the development of squamous cell carcino-ma (SCC) of the genital tract is well established (2). HPVsubtypes of the b-genus are suspected of playing a role innonmelanoma skin cancer (NMSC), but their direct causal rolein this process has been difficult to establish using epidemio-logic approaches due to the ubiquitous prevalence of theviruses in the general population and their absence in somecancers (3–5). In 2009, the International Agency for Researchon Cancer classified HPV5 and 8 as "possibly carcinogenic" inpatients with the rare inherited skin disease epidermodysplasiaverruciformis (EV; ref. 6). Patients with EV are characterized by

the life-long occurrence of multiple flat warts and macularlesions and develop SCC later in life, mostly in sun-exposedskin areas (7, 8).

The mechanisms of HPV-mediated skin tumor developmentare still under debate. In contrast with mucosal high-risk HPVtypes (e.g., HPV16), the primary oncoproteins E6 and E7 ofHPV5andHPV8 do not or only weakly interfere directly with the centralregulators of cell cycle and apoptosis p53 and pRB (3). Someactivities of HPV5 and 8 E6 clearly act to enhance the carcinogenicpotential of sun exposure. The E6 proteins of several beta pap-illomavirus types, including HPV5 and 8, target the proapoptoticprotein Bak for degradation and thus prevent UV-induced apo-ptosis (9). They furthermore bind to p300, resulting in inhibitionof p53-induced apoptosis (10), reduction of ATR mRNA andprotein levels, and finally delayed repair of UV-damaged DNA(11). These activities may result in chromosomal instabilityparticularly in the context of sun exposure and may eventuallycontribute to cancer development. Independent oncogenic effectsmay be expected from E6–MAML1 interactions, which repress theNotch signaling pathway,whichhas tumor suppressor function inthe skin (12). HPV8 E7-induced upregulation of the MT1 matrixmetalloproteinase may explain invasion of E7-positive keratino-cytes into the dermis of organotypic keratinocyte cultures (13).

To gain insight into the molecular mechanisms underlyingHPV8-mediated skin tumor development in vivo, we previouslydeveloped HPV8 transgenic mouse models that recapitulate theHPV8-induced SCC pathology and have been proven to be avaluable in vivomodel tounravel themolecular pathologyofHPV-induced skin cancer (14–18). Transgenic expression of the com-plete early genome region (CER) of HPV8 under the humankeratin14 (K14) promoter in mice (FVB/N background) is suffi-cient to inducemultifocal or single premalignant skin papillomaswith varying degrees of epidermal dysplasia in over 90% ofthe mice within the first year of age (14). SCCs developed in

1Department of Dermatology, University of Cologne, Cologne,Germany. 2Institute of Virology, University of Cologne, Cologne, Ger-many. 3Center for Molecular Medicine Cologne (CMMC), University ofCologne, Cologne, Germany. 4Cologne Excellence Cluster on CellularStress Responses in Aging-Associated Diseases (CECAD), Universityof Cologne, Cologne, Germany.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

X. Ding and T. Lucas contributed equally to this article.

Corresponding Author: Sabine A. Eming, Department of Dermatology, Univer-sity of Cologne, Kerpener Street 62, 50937 Cologne, Germany. Phone: 49-221-4783196; Fax: 49-221-4785949; E-mail: sabine.eming@uni-koeln.de

doi: 10.1158/0008-5472.CAN-13-3007

�2014 American Association for Cancer Research.

CancerResearch

Cancer Res; 75(2) January 15, 2015330

approximately 6%of the transgenicmice. Early hyperproliferativeskin lesions in HPV8 mice are characterized by a subepidermalimmune cell infiltrate (14). Papillomatosis in HPV8 transgenicmice could be significantly accelerated by exposure to UVA/UVBlight or mechanical skin injury (15, 16).

The growth of new blood vessels is essential for the malignantphenotype in most human cancers (19), and most tumor cellsinduce the expression of angiogenic factors at an early point ofdevelopment, such as the homodimeric glycoprotein VEGF-A(hereafter abbreviated with VEGF; ref. 20). In several carcinomasincluding SCCs, increased VEGF expression has been detected inepithelial tumor cells, and has been proposed to be linked totumor progression and invasion (21–25). In addition, inflam-matory cell types within the tumor stroma particularly cells of themyeloid lineage have been shown to be important sources ofVEGF and to functionally affect tumorigenesis (26, 27). VEGFmediates its activities primarily through binding to two receptortyrosine kinases, known as VEGFR1 and VEGFR2, as well as tocoreceptors including neuropilins (Nrp; ref. 20). The classicalconcept that paracrine VEGF contributes to tumor growth throughangiogenesis has been challenged by recent reports, proposingthat VEGF may act in addition as an autocrine survival factor ontumor cells (24, 28, 29). Thus, the precise role of VEGF signalingand of neovascularization in HPV8-mediated skin tumor devel-opment still remains elusive. Here, we investigated the in vivofunction of epidermis- andmyeloid cell–specific VEGF expressionin HPV8-induced skin carcinogenesis by combining our estab-lished model of HPV8-induced skin cancer with cell type-restrict-ed VEGF ablation. Our findings provide evidence for distinctfunctions of epidermis- and myeloid cell–specific VEGF inHPV8-mediated tumor formation.

Materials and MethodsAnimals

Transgenic mice expressing CER of HPV8 (HPV8 mice) fromthehumankeratin14(K14)promoter (14), VEGFfl/flK14-Cre (VEG-FEKO), andVEGFfl/flLysM-Cre (VEGFMKO)mice (30)were generatedand genotyped as described. HPV8 mice were mated to VEGFmutants to obtain VEGFEKOHPV8 and VEGFMKOHPV8 mice, andadditional genotypes as specified throughout the manuscript.Individual expression of Cre or the heterozygous/homozygousloxP-flanked alleles of VEGF did not affect the phenotype of HPV8transgenic orwild-typemice. Allmice were backcrossed to the FVB/N background for at least six generations. VEGF-lacZ knock-inreporter mice were mated with HPV8 mice to obtain VEGF-lacZ-HPV8mice and genotyped as described (31). Mice were examinedevery3days tomonitor thedevelopmentof skin tumors.All animalexperimentswere approved by the national animal care committeeand the University of Cologne (Cologne, Germany).

UV-light exposure and woundingUV-light exposure, wounding, and tissue preparation for his-

tology were performed as described previously (15). Briefly, forUV irradiation, mice were anesthetized, shaved, and an area of2 cm2 on the back was UV irradiated. UV radiation was generatedby a UV device (UV801; Waldmann) equipped with 8 PUVAlamps (UVA: 320–400 nm) and 4 UV21 lamps (UVB: 280–360nm). The energy output of the UV device as measured at the levelof the skin surface using a radiometer was 1.41 mW/cm2 for theUVB region and 6.34 mW/cm2 for the UVA region (UV Radiom-eter Variocontrol; Waldmann). Animals received one single

UV-light exposure (UVA 2.5 J/cm2; UVB 0.25 J/cm2) to inducetumor formation (15). Forwounding,micewere anesthetized andshaved as described above, and four full-thickness punch biopsieswere created on the back (30). For histologic analysis, tissues wereeither fixed in 4% formaldehyde or embedded in optimal cuttingtemperature compound (Tissue Tek) .

Inhibition of VEGFR2Inhibition of VEGFR2 was performed by intraperitoneal injec-

tions of anti-VEGFR2 blocking antibody (DC101, BioXcell) or thecorresponding isotype-matched control antibody (IgG1, BioXcell;40 mg/kg body weight in PBS). Injections started at day 5 afterwounding and were continued for three times/week until day 26after injury. This antibodyhas been shown tobe apotent inhibitorof VEGFR2 (32, 33).

ImmunohistochemistryImmunohistochemistry (IHC) stainings were performed as

described previously (30). Briefly, cryosections (10 mm) werefixed (4% PFA or in methanol), blocked (10% NGS in PBS,5% BSA, 0.2% Triton-X-100), and incubated (minimum 1 hourat room temperature – maximum overnight at 4�C) with theprimary antibodies diluted in blocking buffer. Primary antibodiesused were: CD31 (1:1000, BD Pharmingen); Ki-67 (1:50, Dako);active caspase-3 (1:100, Cell Signaling Technology); F4/80(1:200, Dianova); Gr1 (1:100, BD Pharmingen); VEGF-A(1:100, Santa Cruz Biotechnology); VEGFR1 (1:200, Santa CruzBiotechnology); VEGFR2 (1:50, BD Pharmingen); and Nrp1(1:100, R&D). Bound primary antibody was detected by incuba-tion with Alexa-Fluor 488- or Alexa Fluor 594-conjugated (Invi-trogen) or peroxidase-conjugated (EnVision System, Dako) sec-ondary antibodies (1 hour, room temperature), followed bycounterstaining with DAPI (Invitrogen), propidium iodide, orhematoxylin. Specificity of primary antibodies was demonstratedby replacing them with irrelevant isotype-matched antibodies.

X-Gal stainingFor X-Gal staining, cryosections (20 mm) were prepared from

VEGF-lacZ wound tissues (fixed in 0.5% glutaraldehyde for 30minutes at room temperature) and washed in washing buffer(PBS, 0.02% NP-40, and 0.2 mmol/L MgCl2). Sections wereincubated in X-Gal staining solution containing 0.5 mg/mL ofX-Gal (Fermentas), 10 mmol/L K3[Fe(CN)6], 10 mmol/L K4[Fe(CN)6], and 0.2 mmol/L MgCl2 for 6 hours at 37�C andcounterstained with a nuclear fast red aluminum sulfatesolution.

Morphometric analysisMorphometric analysis was performed on hematoxylin and

eosin (H&E) or Giemsa-stained paraffin tissue sections. Papillo-ma/tumor formation was quantified by measuring epidermaland dermal compartments in a defined area of the skin lesion.Immunofluorescence images were acquired using a fluorescencemicroscope (BZ-9000 Fluorescence Microscope, KEYENCE). Forinvestigation of the average blood vessel density, staining forCD31 and VEGFR2 was quantified in six representative high-power fields (HPF, 200 � 300 mm2)/section in at least twoindependent sections of one lesion using ImageJ software; thedata are expressed as the percentage of CD31 or VEGFR2-positiveareas within the dermal tumor stroma. Ki-67–positive keratino-cytes were quantified as described previously (34, 35). Briefly,

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 331

Ki-67–positive cells were determinedwithin the stratumbasale oflesional skin over a distance of 100 mm in five representativefields/section in at least two independent sections of one lesion;the data are expressed as the percentage of Ki-67–positive basalcells. Numbers of F4/80þ, Gr1þ, and mast cells were determinedby counting positively stained cells in six representative HPFs(200� 300 mm2)/section of one lesion. Analyses were performedin a blinded manner by two independent investigators. Imageswere processed with Adobe Photoshop Version 7.0 software.

Cell cultureKeratinocytes were isolated from the skin of newbornmice and

cultured using a mitomycin-C–treated 3T3-J2 feeder layer aspreviously described (36). Briefly, keratinocytes were expandedin growthmedium [DMEM/Ham's F-12 (Biochrom), 10% chelex-treated FBS, 20 ng/mL EGF, 10 mg/mL insulin (R&D; 32�C, 5%CO2)] and used for experiments at passage 3. For the cell prolif-

eration assay, feeder cells were removed [0.1X Trypsin/EDTA(GIBCO)], keratinocytes were starved in starvation medium[DMEM/Ham's F-12 (Biochrom), 20 ng/mL EGF, 10 mg/mLInsulin (R&D), 4 hours] and subsequently stimulated in starva-tion medium complemented with increasing concentrations ofserum, with or without recombinant murine VEGF-A [(rmVEGF)100 ng/mL, R&D, 20 hours]. After bromodeoxyuridine (BrdUrd)pulse (10 mmol/L, 4 hours), BrdUrd incorporation was analyzedby FACS analysis (FACSCanto II flow cytometer) using the APCBrdU Flow Kit (BD).

Peritoneal cells were seeded in 6-well plates (0.5 � 106 cells/cm2), macrophages enriched by plastic adhesion, cultured overnight, and stimulated with lipopolysaccharide (LPS; 1 mg/mL;Sigma Aldrich) and recombinant IFNg (0.1 mg/mL; R&DSystems) in DMEM (1% FCS) as previously described (30).RNA was isolated and used for qRT-PCR analysis as describedbelow.

Figure 1.Multistage squamous carcinogenesis in HPV8transgenic mice is associated withneoangiogenesis. A and B, CD31 (green)/VEGFR2 (red) double immunofluorescentstaining (A) and morphometric quantification(B) during progressing stages of spontaneoustumor development in HPV8 transgenic mice(K14-HPV8-CER); nuclear DNA was stainedwith DAPI; each dot represents the analysisof one lesion per mouse; data are expressedas means � SD; ��� , P < 0.001; �� , P < 0.01;� , P < 0.05. e, epidermis; d, dermis; h, hornpearls outlined by hatched line; sf,subcutaneous fat tissue; sm, skeletal muscle;dotted line, the dermoepidermal junction.

Ding et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research332

Flow-cytometric analysis and cell sortingKeratinocytes were isolated from tumor tissues by a combi-

nation of enzymatic digestion (Liberase Blendzyme, RocheApplied Science) and mechanical disruption (MedimachineSystem, BD Biosciences) as previously described (30). For FACSsorting, keratinocytes were stained as described previously(37). Briefly, cells were passed through a 40-mm cell strainer,blocked for 15 minutes with BSA (2% in PBS), washed withPBS, and incubated with the following antibodies: FITC-con-jugated anti-CD45 (clone 30-F11, eBiosciences), APC-conju-gated anti-CD140a (clone APA5, eBiosciences) and anti-CD31(clone MEC13.3, BD Pharmingen), and PE-Cy7-conjugatedanti-Epcam (clone G8.8, Biolegend) in FACS buffer. Dead cellswere excluded using 7-AAD (BD Biosciences). EpcamþLin�

(CD45�CD31�CD140�) cells were sorted by FACSAria cellsorting system (BD Biosciences).

Real-time PCR analysisRNA was isolated from tissues and reverse transcription PCR

was performed as previously described (30). The hyperprolifera-tive epidermis was separated from the dermis (0.5 mol/L ammo-

nium thiocyanate solution) as described (38) and tissues werestored in RNAlater (Ambion). Primer sequences for VEGF-A andNrp1 were described previously (30) or will be available uponrequest; for HPV8-E6 and HPV8-E7 expression analysis, thefollowing forward and reverse primers (50 to 30) were used:E6, GCGGCTTTAGGTATTCCATTGC, GCTACACAACAACAAC-GACAACACG; E7 CCTGAAGTGTTACCAGTTGACCTGC, CAGT-TGCGTTGACAAAAAGACG. Amplification reactions (triplicates)were set up using PowerSYBR Green PCR Master Mix (AppliedBiosystems) and qRT-PCR was validated with the StepOnePlusReal-Time PCR system (Applied Biosystems). The comparativemethod of relative quantification (2�DDCt) was used to calculateexpression level of the target gene normalized to S26.

Statistical analysisSignificance of difference was analyzed using ANOVA one-

way test analysis with Bonferroni multiple comparison test. Alldata are presented as mean � SD, a P value of �0.05 wasconsidered significant. The Kaplan–Meier method and the log-rank test were used to compare tumor-free states betweenmouse genoyptes.

Figure 2.Epidermis-specific VEGF deletionprevents spontaneous skin tumordevelopment in HPV8 transgenicmice, whereas myeloid cell–derivedVEGF is dispensable. A, genotype ofmouse models used in this study. B,scheme illustrating the VEGF geneconstruct and the two loxP sitesflanking exon 3, and the PCR fragmentlength before and after successfulCre-mediated recombination.C, representative PCR of genomicDNA isolated from epidermis andperitoneal macrophages from variousmouse mutants (n ¼ 2 mice/genotype). D, relative expression ofVEGF in epidermis of lesional skin afterUV-light induced papilloma formationat day 34 and in activated peritonealmacrophages (LPS/IFNg) isolatedfrom mouse mutants (n ¼ 3 K14-HPV8-CER; n ¼ 3 VEGFEKOHPV8;n¼ 3VEGFMKOHPV8). E, Kaplan–Meierplot illustrating the progressivetumor-free state in diverse mousemutants (n ¼ 10 K14-HPV8-CER;n ¼ 10 VEGFEKOHPV8; n ¼ 10VEGFMKOHPV8); the indicatedP values refer to the log-rank test.

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 333

Figure 3.Epidermis-specific VEGF deletion prevents UV light-triggered skin tumor development in HPV8 transgenic mice. A, top, X-Gal staining of UV irradiated skin inVEGF-lacZ-HPV8 mice (n ¼ 15); bottom, representative VEGF (red)/F4/80 (green) double immunofluorescent stainings of skin sections from HPV8 mice atday 10 (n¼ 5) and 34 (n¼ 5) after UV irradiation. B–D, macroscopic appearance (B), H&E staining (C), and morphometric quantification of skin lesions at differenttime points after UV irradiation (n ¼ 8–14 mice/genotype; D). E, qRT-PCR analysis of E6 and E7 expression in epidermis of untreated skin and skin 10 and 34 daysafter UV irradiation in differentmousemutants; each dot represents onemouse. Data are expressed asmeans� SD. ��� , P <0.001; � , P <0.05. e, epidermis; d, dermis;h, horn pearls outlined by hatched line; sf, subcutaneous fat tissue; sm, skeletal muscle; dotted line, the dermoepidermal junction. Genotypes used, control:n ¼ 6 VEGFfl/fl, n ¼ 4 LysMCre, n ¼ 4 K14Cre; HPV8: n ¼ 12 K14-HPV-CER/VEGFfl/fl, n ¼ 4 K14-HPV-CER/LysMCre, n ¼ 4 K14-HPV-CER/K14Cre.

Ding et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research334

ResultsMultistage squamous carcinogenesis in HPV8 transgenic miceis associated with increased angiogenesis

To examine whether tumor progression in HPV8 transgenicmice is associatedwith neoangiogenesis, IHCdouble stainings forCD31 and VEGFR2 were performed in tissue sections at variousstages of tumor development. Although CD31 is constitutivelydetectable on the vascular endothelium, VEGFR2 expression isprimarily upregulated in endothelial cells during neoangiogen-esis. In macroscopically and histologically nonlesional back skinof HPV8 transgenic mice (8–10 weeks of age), few small CD31þ

/VEGFR2� capillaries were detectable in the dermis comparablewith wild-type mice (Fig. 1). At the age of 20 to 25 weeks, HPV8transgenic mice developed hyperplastic lesions with acanthosis,hyperkeratosis, and increased density of CD31þ/VEGFR2þ dou-ble positive, elongated vessels within papillomas (Fig. 1). Epi-thelial tumors characterized by epidermal dysplasia showed a

high-density vessel network, with elongated and tortuous vesselsin close apposition to the overlaying neoplastic epithelium(Fig. 1). Infiltration of epidermal tumor cells into the dermaltissue in SSCs was associated with large vessel lumen spreadingthroughout the dermal tissue (Fig. 1). The remarkable alterationof the vasculature in vessel number, distribution, and morphol-ogy during tumor progression was confirmed by increased num-bers of CD31þ/VEGFR2þ double positive vascular structuresparticularly detected in close proximity to the neoplastic epider-mis (Fig. 1). VEGFR2-positive stainingwas exclusively detected onvascular structures.

Epidermis-specific VEGF deletion prevents long-termspontaneous skin tumor development in HPV8 transgenicmice, whereas myeloid cell–derived VEGF is dispensable

To analyze the functional impact of epidermal-derived versusmyeloid cell–derived VEGF synthesis in HPV8-mediated skin

Figure 4.Attenuated epidermal proliferationand angiogenic responses in UV-triggered papilloma development inVEGFEKOHPV8 mice. A and B, Ki-67IHC (brown nuclear stain; hematoxylincounterstain) and CD31immunofluorescence staining (red;DAPI counterstain, blue; A) andmorphometric quantification of skinlesions at day 34 after UV irradiation(B); each dot represents the analysisof one lesion per mouse; data arerepresented as means � SD;��� , P < 0.001. e, epidermis; d, dermis;h, horn pearls outlined by hatched line;sf, subcutaneous fat tissue; dottedline, the dermoepidermal junction.Genotypes used, control: n ¼ 6VEGFfl/fl; HPV8: n ¼ 5 K14-HPV-CER/VEGFfl/fl.

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 335

tumor development, mice with epidermis-restricted VEGF dele-tion (VEGFEKO) and myeloid cell–restricted VEGF deletion(VEGFMKO) were mated with transgenic mice in which HPV8early gene expression is driven by the human K14 promoter(HPV8 mice; refs. 14, 30; Fig. 2A and B). High efficiency of

Cre-mediated vegf-a gene deletion in epidermis andmacrophagesin VEGFEKOHPV8 and VEGFMKOHPV8 mice, respectively, wasverified by PCR and qRT-PCR analyses (Fig. 2C and D) and wasconsistent with earlier results in VEGFEKO and VEGFMKO mice(30).Wemonitoredover a periodof 52weeks spontaneous tumor

Figure 5.Epidermis- and myeloid cell–derived VEGFs are essential for mechanical injury-induced papilloma formation in HPV8 mice. A–C, macroscopic analysis (A),H&E staining (B), andmorphometric quantification of papilloma formation in different mousemutants (n¼ 6–10mice/genotype; C). D, morphometric quantificationof Ki-67–positive basal cells and CD31-positive area in skin lesions at day 34 after injury; each dot represents one wound. Data are presented as means � SD;��� , P < 0.001. e, epidermis; d, dermis; st, scar; h, horn pearls outlined by hatched line; dotted line, the dermoepidermal junction. Genotypes used, control:n ¼ 4 wt, n ¼ 6 VEGFfl/fl, n ¼ 3 LysMCre, n ¼ 4 K14Cre; HPV8: n ¼ 5 K14-HPV8-CER/VEGFfl/fl, n ¼ 4 K14-HPV8-CER/K14Cre, n ¼ 3 K14-HPV8-CER/LysMCre.

Ding et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research336

development in cohorts of VEGFEKOHPV8 (n ¼ 10), VEGFMKO

HPV8 (n¼ 10), and HPV8 (n¼ 10) mice. The Kaplan–Meier plotillustrating the tumor-free state as a function of time showedstriking differences between the three genotypes (Fig. 2E). In linewith our previous observation, skin tumor development initiatedin more than 50% of HPV8mice at the age of 20 to 25 weeks andnearly all mice presented tumors at 52 weeks of age (Fig. 2E;ref. 14). The lesions were irregularly distributed and characterizedby hair loss, ulcerations, and irregularly shaped papillomas; theyoccurred most often at the laterodorsal trunk (14). Notably, inVEGFEKOHPV8 mice, spontaneous tumor development wasentirely abrogated during 1 year of assessment (the oldestVEGFEKOHPV8 mice are now 1.5 years old and are still tumorfree; Fig. 2E). In contrast, no significant difference was observed inspontaneous tumor development in VEGFMKOHPV8 mice whencomparedwithHPV8mice (Fig. 2E). Themacroscopic appearance

and the localization of skin tumors developing in VEGFMKOHPV8and HPV8 mice were similar.

Epidermis-specific VEGF deletion prevents UV-light triggeredskin tumor development in HPV8 transgenic mice, whereasmyeloid cell–derived VEGF is dispensable

We previously showed that a single UV-light exposure led toa synchronized induction of papilloma development inapproximately 3 weeks in HPV8 transgenic mice (15). To studythe role of VEGF in this rapid UV-triggered HPV8-mediated skintumor development, VEGFEKOHPV8, VEGFMKOHPV8, HPV8,and control mice were subjected to UV irradiation. Yet, firstwe identified VEGF-expressing cells in tumors of UV-lightexposed HPV8 transgenic mice, and we mated VEGF-lacZreporter mice with HPV8 transgenic mice. As revealed byb-galactosidase staining, skin not exposed to UV-light showed

Figure 6.Blocking VEGFR2 preventsmechanical injury-induced papillomaformation in HPV8 transgenic mice.A–C, macroscopic appearance (A),H&E staining (B), and morphometricquantification of wound-inducedpapilloma formation in HPV8 micetreated with VEGFR2 neutralizingantibody or isotype control antibody(IgG1; n ¼ 4 mice/group; C). D and E,Ki-67 (green; counterstain propidiumiodide, red) and CD31 (red;counterstain DAPI, blue) staining inskin lesions of antibody-treated HPV8mice at day 26 after wounding (D) andmorphometric quantification (E); eachdot represents one mouse (HPV8,n ¼ 4 K14-HPV8-CER/VEGFfl/fl percondition). Data are presented asmeans � SD. � , P < 0.05; �� , P < 0.01.e, epidermis; d, dermis; h, horn pearlsoutlined by hatched line; dotted line,the dermoepidermal junction.

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 337

Ding et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research338

only individual VEGF-expressing cells within the subcutaneousfat tissue (data not shown), whereas irradiation resulted in arobust increase of VEGF expression within the hyperplasticepithelium and infiltrating inflammatory cells by days 5 and24 after irradiation (Fig. 3A). In addition, costaining for VEGFand the macrophage antigen F4/80 in day 10 and day 34papillomas of UV-treated HPV8 mice identified VEGF-positivekeratinocytes and macrophages (Fig. 3A); in keratinocytes, theVEGF signal was present in both nuclei and cytoplasm.

By day 7 after UV exposure, all mice showed ulcerative lesionswithin the light exposed area (Fig. 3B). Control mice reachedcomplete restoration of skin integrity over a time course of 34 days,whereas HPV8 and VEGFMKOHPV8 mice showed pronouncedpapilloma formation at the ulcer edges (Fig. 3B). In contrast, ulcersinVEGFEKOHPV8mice closedwith a similar rate as in controlmice,without developing any sign of papilloma. H&E-stained tissuesections at day 34 after UV irradiation confirmed the absence ofpapillomas and complete epithelialization of erosions in controlandVEGFEKOHPV8mice (Fig. 3C andD). The irradiated area at day34was inbothmouse strains characterized by amild acanthosis. Incontrast, UV-exposed skin in HPV8 and VEGFMKOHPV8 miceshowed profound squamous papilloma development, includinga significant increase in hyperplasia, hyperkeratosis, paraker-atosis, and characteristic horn pearls (Fig. 3C and D). Thepapillary dermis in HPV8 and VEGFMKOHPV8 mice was char-acterized by a mixed inflammatory cell infiltrate dominated bymononuclear cells/macrophages and mast cells (Fig. 3C anddata not shown). Expression of viral E6 and E7 in HPV8 andVEGFEKOHPV8 mice was confirmed by qRT-PCR analysis inepidermis separated from untreated and lesional skin (Fig. 3E).

To analyze the proliferation status of epidermal cells in UVexposed skin after 34 days of irradiation, tissue sections werestained for the presence of Ki-67 (Fig. 4). Although in control andVEGFEKOHPV8mice, about 29.81%� 3.38 and 23.55%� 2.32 ofcells within the stratum basale stained positive for Ki-67, respec-tively, about 78.47%� 1.41 and 79%� 11.51 basal cells stainedKi-67 positive within the acanthotic epidermis in HPV8 andVEGFMKOHPV8 mice, respectively (Fig. 4). In addition, in HPV8and VEGFMKOHPV8mice Ki67þ cells scattered throughout supra-basal layers (Fig. 4). Furthermore, as shown byCD31 staining, thedermal compartment of papillomas inHPV8 and VEGFMKOHPV8mice showed a significant increase of elongated, enlarged, andtortuous blood vessels characteristic for tumor vasculature, whencompared with few and small capillaries detected in the dermis ofcontrol and VEGFEKOHPV8 mice (Fig. 4).

Epidermis- and myeloid cell–derived VEGF is critical formechanical injury-induced papilloma formation in HPV8transgenic mice

We next examined whether the opposing contribution ofepidermal- and myeloid cell–derived VEGF in spontaneous or

UV-light triggered tumor induction in HPV8 transgenic mice maybe related to the nature of cell and/or tissue injury. We inflictedfull-thickness excision skin wounds on the backs of control andmutant mice and monitored the wound-healing response over34 days. After approximately 8 days, wound edges in HPV8 mice(here we included also K14-HPV8-CER/K14Cre and K14-HPV8-CER/LysMCre mice as controls to exclude interference of Creexpression in K14Cre or LysMCre mice with tumor formation)showed signs of developing papillomatosis, which were macro-scopically clearly visible, and gave raise to full papillomas by day34 after injury (Fig. 5A). Notably, wounds in control, VEGFMKO

HPV8, and VEGFEKOHPV8 mice did not develop signs of papil-lomatosis and healed completely by day 34 after injury (Fig. 5A).H&E-stained tissue sections of wounded skin at day 34 confirmedsquamous papilloma formation inHPV8mice (Fig. 5B and C). Incontrast, tissue sections in control, VEGFMKOHPV8, and VEGFEKO

HPV8 mice all displayed a closed, slightly hyperproliferativeepithelium and scar tissue, typical for late stage, closed excisionalskinwounds in healthymice (Fig. 5B). Papilloma development inwounds of HPV8 mice was paralleled by a significant increase inKi-67þ basal cells, which in addition scattered throughout thehyperplastic epidermis, and a highly vascularized tumor stromacharacterized by increased numbers of F4/80þ cells (macro-phages), Gr1þ cells (polymorphonuclear cells), and mast cells(Fig. 5D and Supplementary Fig. S1). To examine whether mye-loid cell–derived VEGF controls tumor initiation by a direct effecton the inflammatory response (e.g., through chemotaxis), weanalyzed the inflammatory cell infiltrate at the early stage oftumor development (day 8 after injury) in control, HPV8, andVEGFMKOHPV8 mice. Although polymorphonuclear cells andmast cells were virtually absent in the papillary dermis of epithe-lialized wounds in all three genotypes, macrophages predomi-nated the inflammatory cell infiltrate and their numbers weresimilar in different genotypes. Hence, based on these findings, it isunlikely that myeloid cell–derived VEGF-A promotes tumor ini-tiation by a direct effect on the inflammatory response.

To investigate whether the vasculature is a target of VEGF inHPV8-mediated skin tumor development, we used an antibody-based strategy to inhibit angiogenesis by blocking VEGFR2 sig-naling in endothelial cells. As shown by IHC, VEGFR2 stainingwas exclusively detected on tumor vasculature inHPV8 transgenicmice, and not in the tumor epithelium (Fig. 1). Starting on day 5until day 25 after wounding, anti-VEGFR2 antibody (DC101) orisotype control antibody was repetitively administrated by intra-peritoneal injections. In HPV8 mice receiving isotype controlantibody, papilloma formation was macroscopically visible atwound edges at day 15 after injury and progressed in size by day23 after injury (Fig. 6A). In contrast, in HPV8 mice receiving anti-VEGFR2 antibody (DC101), papilloma development was notdetected until day 26 after injury (Fig. 6A). H&E-stained tissuesections of wounded skin in HPV8mice receiving isotype control

Figure 7.Epidermal expression of Nrp1 and VEGFR1 in papillomas in HPV8 mice. A, keratinocytes isolated from newborn wild-type or HPV8 mice (K14-HPV8-CER) werecultured in different conditions as indicated; BrdUrdþ cells are expressed as fraction of all keratinocytes in the culture dish (n¼ 3 mice/genotype; two independentexperiments were performed). B, qRT-PCR analysis for genes as indicated in epidermis isolated from skin of diverse mouse mutants at different time pointsafter UV irradiation (control: n¼ 3 VEGFfl/fl, HPV8: n¼ 4 K14-HPV8-CER/VEGFfl/fl, n¼ 3 VEGFEKOHPV8); expressionwas normalized to gene expression in epidermisof untreated skin in control mice; data are presented asmeans� SD. ��� , P < 0.001; ��, P < 0.01; � , P < 0.05. C, qRT-PCR analysis for VEGF, Nrp1, VEGFR1, and VEGFR2expression in Epcamþ Lin� keratinocytes isolated from untreated skin and day 30 UV-light–treated skin of HPV8 mice (n ¼ 4 K14-HPV8-CER/VEGFfl/fl).D, representative IHC staining for Nrp1 (brown; hematoxylin counterstain) and immunofluorescence double staining for VEGFR1 (green) and CD31 (red; counterstainDAPI, blue) in skin sections of HPV8 mice (n ¼ 4 K14-HPV8-CER/VEGFfl/fl) 34 days after irradiation.

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 339

antibody showed all hallmarks of papilloma formation, whereasHPV8 mice receiving anti-VEGFR2 antibody showed a mildhyperplasia by day 26 after injury (Fig. 6B and C). Furthermore,as shownby IHC stainings for Ki-67 andCD31, proliferating basalcells and tumor vasculature were significantly increased inwounded skin in HPV8 mice receiving isotype control antibody,when compared with HPV8 mice receiving blocking antibody(Fig. 6D and E). Staining for Ki-67 and caspase-3 in nonlesionalskin adjacent to tumors or wound tissue in IgG1 or aVEGFR2-treated mice was comparable, indicating that aVEGFR2 has nodirect (neoangiogenesis independent) effect on epidermal pro-liferation or apoptosis (Supplementary Fig. S2).

Epidermal expression of Nrp1 and VEGFR1 in papillomas inHPV8 mice

We next explored whether in addition to VEGF-mediatedangiogenesis-dependent mechanisms of papilloma developmentin HPV8-transgenic mice also angiogenesis independent, auto-crine signaling of VEGF in keratinocytes may contribute to rapidtumor development. Recent studies reported on unexpected andimportant autocrine functions of VEGF in epidermal tumor cells(24, 28, 33, 35). Therefore, we first addressed the questionwhether HPV8 transgenic keratinocytes acquire a cell-autono-mous growth advantage when compared with wild-type kerati-nocytes, and whether external rmVEGF acts as possible mitogenon HPV8 transgenic keratinocytes. For this purpose, primarykeratinocytes were isolated from newborn (nonlesional) wild-type and HPV8 transgenic mice and cell proliferation wasanalyzed by BrdUrd incorporation in the presence of growthmedium (DMEM/Ham's F-12, EGF 20 ng/mL, insulin 10 mg/mL) complemented with increasing serum concentrations andrmVEGF (100 ng/mL; Fig. 7A). Both wild-type and HPV8transgenic cells showed robust BrdUrd incorporation inresponse to increasing serum concentrations (Fig. 7A), which,however, was independent of externally added rmVEGF (Fig.7A). Thus, transgenic expression of CER of HPV8 in newborn(nonlesional) keratinocytes is not sufficient to convey a cell-autonomous growth advantage over wild-type keratinocytes,and neither renders the transgenic cells sensitive to externalrmVEGF as mitogen. These findings suggest that additionalfactors are required to deregulate keratinocyte proliferationduring carcinogenesis in HPV8 mice in vivo.

To gain further insight into potential autocrine VEGF effects inkeratinocyte during tumorigenesis in HPV8 mice, next, weexplored the expression status of epidermal VEGF receptors andVEGF during tumor development in HPV8mice. For this purposeat days 10 and 34 after UV exposure, the epidermis was separatedfrom dermis and epidermal expression for VEGF, VEGFR2,VEGFR1, and Nrp1 was determined by qRT-PCR analysis (Fig.7B and C). Epidermis of untreated skin in control mice served asbaseline for gene expression. Although papilloma formation inUV-light–exposed skin ofHPV8micewas associatedwith a robustinduction of epidermal VEGF and Nrp1 expression, absence oftumor formation in VEGFEKOHPVmice was paralleled by absenceof epidermal VEGF (as expected) and Nrp1 expression (Fig. 7B).In epidermal tissue, transcripts for VEGFR1 or VEGFR2 werehardly detectable in both UV-treated and untreated skin of the3 mouse strains and we are reluctant to draw any conclusion onthese findings with regards to expression of VEGFRs (DCt valuesfor qRT-PCR analysis were in the range of 29–34). Therefore, tofurther refine the analysis of VEGF receptor expression in kerati-

nocytes and to exclude contamination of epidermal tissue withdermal cells, EpcamþLin� (C45�CD31�CD140a�) cells wereisolated from untreated skin and papilloma tissues of HPV8mice(day 30 after UV-light–treated skin of HPV8mice; SupplementaryFig. S3) and purified cells were subjected to qRT-PCR analysis.These studies confirmed strong expression of VEGF and Nrp1 inpapilloma-derived keratinocytes and virtually no VEGFR2 (Fig.7C). In contrast, enrichment of EpcamþLin� cells revealed anupregulation of VEGFR1 in papilloma-derived keratinocytesalthough at low levels. Epidermal expressionofNrp1 andVEGFR1in papillomas was confirmed by IHC staining (Fig. 7D). Stainingfor Nrp1 was shown in the cytoplasm (partially also the cellmembrane), whereas VEGFR1 staining localized particularly(peri)nuclear. VEGFR2 staining was not detected in the epidermis(Fig. 1).

DiscussionOur study reveals a close functional interplay between the onset

of HPV8-induced progressive squamous tumor formation andneovascularization inmice. Thus, our findings corroborate earlierstudies reporting on vascular alterations during neoplastic pro-gression in both human cervix carcinoma and the correspondingtransgenic K14-HPV16 mouse model, speculating on a causalrelationship of virus-induced carcinogenesis andneoangiogenesis(39, 40). Although those early studies on HPV-induced mucosalcarcinogenesis hypothesized that VEGF released from epidermaltumor cells and/or recruited inflammatory cellsmay contribute tovascular alterations critical for the essential vascular switch intumor progression (19), here we provide clear evidence for afunctional link between epidermal andmyeloid cell–type-restrict-ed VEGF-mediated angiogenesis and HPV8-induced skin tumordevelopment.

However, emerging evidence in preclinical and clinical studiesof various epithelial cancers suggests that epidermal synthesizedVEGF in addition to regulating paracrine effects on the vascula-ture, mediates angiogenesis-independent, autocrine growth pro-moting, and survival effects on epidermal tumor cells (24, 28). Ina transgenic mouse model of non-viral K5-SOS/EGFR-mediatedepidermal tumorigenesis, Lichtenberger and colleagues reportedthat epidermal-derived VEGF induces cell-autonomous tumorcell proliferation via an autocrine mechanism, whereby VEGFcould act intra- and/or extracellularly mediated by VEGFR1 and/orNrp1 (35). Furthermore, recently, Beck and colleagues reportedin the DMBA/TPA two-stage chemical skin carcinogenesis model,that CD34þ epidermal cancer stem cells express Nrp1 and thatVEGF controls initiation and stemness through Nrp1 in an auto-crine loop; in vitro studies showed that externally added VEGFcontrols proliferation of isolated CD34þ epithelial tumor cells(33). Yet, in this study, the question remained unclear throughwhich additional VEGFRs, the coreceptor Nrp1 may direct VEGF-mediated signals in keratinocytes.

Here, we showed that the systemic inhibition of VEGFR2 byan antibody-based strategy effectively prevented wound-inducedtumor formation in HPV8 mice. These findings strongly suggestan angiogenesis-mediated effect of paracrine/exogenous VEGFin HPV8-induced tumor development, because robust levels ofVEGFR2 expression could only be detected in the tumor vas-culature but not in keratinocytes. Currently, there is somecontroversy about the expression of VEGFR2 in keratinocytes(33, 35, 41–43), which may be explained by different

Ding et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research340

experimental conditions and/or keratinocyte sources. However,as shown by IHC and in part also gene expression analysis,Nrp1 and VEGFR1 were not only expressed in vascular struc-tures, but also in the hyperplastic epidermis of tumor lesions,proposing in addition an angiogenesis-independent, potential-ly autocrine effect mediated by epidermal VEGF. AlthoughNrp1 staining revealed a cytoplasmatic pattern primarily inbasal layer cells, VEGFR1 staining was characterized by a (peri)nuclear staining pattern predominantly in basal and few supra-basal cell layers. Interestingly, a (peri)nuclear staining patternfor VEGFR1 has been recently reported in breast cancer cells andhas been shown to be critical for VEGF-mediated internalautocrine survival signals (28). Although at this stage, we arecautious to convey a mechanism of action based on our find-ings, we are tempted to speculate on an autocrine role ofepidermal VEGF in our model. Furthermore, in context withour in vitro findings that externally added rmVEGF does not actas mitogen neither on wild-type keratinocytes nor on HPV8-transfected cells, we propose an intracrine mechanism of VEGFaction in keratinocytes that promotes tumor development. Infact, intracellular activity of VEGF and unresponsiveness toexternal recombinant VEGF have been previously proposed invarious cell types (35, 44–46), including carcinoma cells(24, 28, 29); however, up to date, underlying functionalmechanisms are poorly investigated. Therefore, at this stageour findings in HPV8-mediated tumorigenesis support recentreports on potential intracrine, autocrine VEGF signaling incarcinomas and provide the basis for further studies examiningthe specific role of an intracrine, autocrine loop mechanism forVEGF signaling in epidermal carcinogenesis.

Tumor progression in HPV8 mice was paralleled by a stronginflammatory response within the tumor stroma, dominated bymacrophages. In various tumor entities, infiltrating macro-phages have been identified as critical cellular componentspromoting tumor initiation and progression (47). Specifically,in preclinical tumor models, a dual role of myeloid cell–derived VEGF has been described including both a tumor-promoting activity (48) and a negative function as tumorsuppressor (26). So far, it is unclear whether HPV8-positivekeratinocytes initiate a cross-talk with infiltrating immune cellsthat may promote epithelial tumorigenesis. Our findings ofabrogation of tumor initiation in VEGFMKOHPV8 mice afterexcision skin injury corroborate a critical role of myeloid cell–restricted VEGF in HPV8-induced tumor formation, most likelya proangiogenic function. This concept of an angiogenic-dependent mechanism is consistent with a recent report of ourgroup, demonstrating an essential role of myeloid cell–derivedVEGF for the initiation of wound angiogenesis (30). In addi-tion, beside its well-known proangiogenic potency, VEGF hasalso been described as monocyte chemoattractant (49, 50), andsomeone may argue that myeloid cell–restricted VEGF acts asimmunomodulator to promote carcinogenesis in HPV8 mice.Here, we could not detect any significant difference in theinflammatory cell response in HPV8 and VEGFMKOHPV8 micefew days after injury, the time point of tumor initiation. Thus,these findings argue against a major role of a direct action ofmyeloid cell–restricted VEGF on immune cells during HPV8-induced epidermal carcinogenesis. Our findings are consistentwith earlier experimental in vivo studies that could also notdetect a critical effect of myeloid cell–restricted VEGF onmonocyte recruitment in skin inflammation (30, 51).

For UV-light–triggered tumor formation in HPV8 transgenicmice, myeloid cell–derived VEGF was dispensable, whereas epi-dermis-derived VEGFwas essential. Different reasonsmay explaintheprevailing role of epidermal VEGFunder these conditions.UV-light has been identified as potent inducer of VEGF expression inkeratinocytes (52), and in irradiated VEGFMKOHPV8 mice,increased epidermal VEGF synthesis may compensate for the lackof myeloid cell–derived VEGF. Furthermore, UV-light induces astrong cellular inflammatory response that may provide addi-tional cellular sources for VEGF and/or other growth stimulatingand/or proangiogenic factors (40). In addition, the UV-light–induced DNA damage response may increase expression of otherproangiogenic factors that rescue the lack of myeloid cell–derivedVEGF in VEGFMKOHPV8 mice (53). Finally, as discussed above, apotential autocrine–intracrine VEGF loop mechanism may directgrowth promoting and survival effects on epidermal tumor cells,so that epidermal VEGF deletion protects effectively fromtumorigenesis.

In conclusion, our analysis reveals a strong functional interactionbetween neoangiogenesis and the progression of squamous carci-nogenesis inHPV8 transgenicmice. Inaddition,we showthatVEGFserves cell-type specific functions in HPV8-mediated premalignantpapilloma formation in context with the nature of the tumor-promoting condition. Furthermore, in addition to a paracrineeffect on the vasculature, epidermal VEGF may control HPV8-induced tumor development through an autocrine–intracrine loopmechanism, independent from angiogenesis. Collectively, thesefindings propose that blocking VEGF signaling both externally andintracellularly, may represent a promising therapeutic target for thetreatment of HPV8-mediated skin carcinogenesis.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: X. Ding, T. Lucas, S.A. EmingDevelopment of methodology: X. Ding, T. Lucas, S.A. EmingAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): X. Ding, G.P. Marcuzzi, S.A. EmingAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): X. Ding, T. Lucas, S.A. EmingWriting, review, and/or revision of the manuscript: X. Ding, T. Lucas,G.P. Marcuzzi, H. Pfister, S.A. EmingAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): G.P. MarcuzziStudy supervision: S.A. Eming

AcknowledgmentsThe authors thank Napeoleone Ferrara (VEGF-Aflox mouse line), Irmgard

F€orster (LysMCre mouse line), and Andras Nagy (VEGF-lacZ mouse line) forgenerously providing diverse mouse lines, Michael Piekarek, Sebastian W€ust,Margot Junker for excellent technical support, Gunter Rappel for cell sorting,and Sebastian Willenborg, Martin Hufbauer, and Dirk Weßler for helping onexperimental animal work.

Grant SupportThis work was supported by the German Research Society (SFB829 to S.A.

Eming) and Center of Molecular Medicine Cologne (S.A. Eming).The costs of publication of this articlewere defrayed inpart by the payment of

page charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received October 18, 2013; revised October 2, 2014; accepted October 15,2014; published OnlineFirst November 20, 2014.

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 341

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice

References1. de Villiers EM. Cross-roads in the classification of papillomaviruses.

Virology 2013;445:2–10.2. zurHausenH. Papillomaviruses in the causation of human cancers - a brief

historical account. Virology 2009;384:260–5.3. Akgul B, Cooke JC, Storey A. HPV-associated skin disease. J Pathol

2006;208:165–75.4. Pfister H. Chapter 8: Human papillomavirus and skin cancer. J Natl Cancer

Inst Monogr 2003:52–6.5. Weissenborn SJ, Nindl I, Purdie K, Harwood C, Proby C, Breuer J, et al.

Human papillomavirus-DNA loads in actinic keratoses exceed those innon-melanoma skin cancers. J Invest Dermatol 2005;125:93–7.

6. Bouvard V, Baan R, Straif K, Grosse Y, Secretan B, El Ghissassi F, et al. Areview of human carcinogens-Part B: biological agents. Lancet Oncol2009;10:321–2.

7. Gul U, Kilic A, Gonul M, Cakmak SK, Bayis SS. Clinical aspects ofepidermodysplasia verruciformis and review of the literature. Int J Derma-tol 2007;46:1069–72.

8. Patel T, Morrison LK, Rady P, Tyring S. Epidermodysplasia verruciformisand susceptibility to HPV. Dis Markers 2010;29:199–206.

9. UnderbrinkMP,HowieHL, Bedard KM, Koop JI, GallowayDA. E6 proteinsfrom multiple human betapapillomavirus types degrade Bak and protectkeratinocytes from apoptosis after UVB irradiation. J Virol 2008;82:10408–17.

10. Muench P, Probst S, Schuetz J, Leiprecht N, Busch M, Wesselborg S, et al.Cutaneous papillomavirus E6 proteins must interact with p300 and blockp53-mediated apoptosis for cellular immortalization and tumorigenesis.Cancer Res 2010;70:6913–24.

11. Wallace NA, Robinson K, Howie HL, Galloway DA. HPV 5 and 8 E6abrogate ATR activity resulting in increased persistence of UVB inducedDNA damage. PLoS Pathog 2012;8:e1002807.

12. Tan MJ, White EA, Sowa ME, Harper JW, Aster JC, Howley PM. Cutaneousbeta-human papillomavirus E6 proteins bind Mastermind-like coactiva-tors and repress Notch signaling. Proc Natl Acad Sci U S A 2012;109:E1473–80.

13. Akgul B, Garcia-Escudero R, Ghali L, Pfister HJ, Fuchs PG, Navsaria H, et al.The E7 protein of cutaneous human papillomavirus type 8 causes invasionof human keratinocytes into the dermis in organotypic cultures of skin.Cancer Res 2005;65:2216–23.

14. Schaper ID, Marcuzzi GP, Weissenborn SJ, Kasper HU, Dries V, Smyth N,et al. Development of skin tumors in mice transgenic for early genes ofhuman papillomavirus type 8. Cancer Res 2005;65:1394–400.

15. Marcuzzi GP, Hufbauer M, Kasper HU,Weissenborn SJ, Smola S, Pfister H.Spontaneous tumour development in human papillomavirus type 8 E6transgenicmice and rapid induction byUV-light exposure andwounding. JGen Virol 2009;90:2855–64.

16. Hufbauer M, Lazic D, Akgul B, Brandsma JL, Pfister H, Weissenborn SJ.Enhanced human papillomavirus type 8 oncogene expression levels arecrucial for skin tumorigenesis in transgenic mice. Virology 2010;403:128–36.

17. De Andrea M, Ritta M, Landini MM, Borgogna C, Mondini M, Kern F, et al.Keratinocyte-specific stat3 heterozygosity impairs development of skintumors in human papillomavirus 8 transgenic mice. Cancer Res 2010;70:7938–48.

18. Rolfs F,HuberM,Gruber F, BohmF, PfisterHJ, Bochkov VN, et al.Dual roleof the antioxidant enzyme peroxiredoxin 6 in skin carcinogenesis. CancerRes 2013;73:3460–9.

19. Bergers G, Benjamin LE. Tumorigenesis and the angiogenic switch. Nat RevCancer 2003;3:401–10.

20. Ferrara N. Vascular endothelial growth factor: basic science and clinicalprogress. Endocr Rev 2004;25:581–611.

21. Sauter ER, Nesbit M, Watson JC, Klein-Szanto A, Litwin S, Herlyn M.Vascular endothelial growth factor is a marker of tumor invasion andmetastasis in squamous cell carcinomas of the head and neck. Clin CancerRes 1999;5:775–82.

22. Detmar M, Velasco P, Richard L, Claffey KP, Streit M, Riccardi L, et al.Expression of vascular endothelial growth factor induces an invasive phe-notype inhuman squamous cell carcinomas. Am J Pathol 2000;156:159–67.

23. Bowden J, Brennan PA, Umar T, Cronin A. Expression of vascular endo-thelial growth factor in basal cell carcinoma and cutaneous squamous cellcarcinoma of the head and neck. J Cutan Pathol 2002;29:585–9.

24. Samuel S, Fan F, Dang LH, Xia L, Gaur P, Ellis LM. Intracrine vascularendothelial growth factor signaling in survival and chemoresistance ofhuman colorectal cancer cells. Oncogene 2011;30:1205–12.

25. Rossiter H, Barresi C, Pammer J, Rendl M, Haigh J, Wagner EF, et al. Loss ofvascular endothelial growth factor a activity in murine epidermal kerati-nocytes delays wound healing and inhibits tumor formation. Cancer Res2004;64:3508–16.

26. Stockmann C, Doedens A, Weidemann A, Zhang N, Takeda N, GreenbergJI, et al. Deletion of vascular endothelial growth factor in myeloid cellsaccelerates tumorigenesis. Nature 2008;456:814–8.

27. Qian BZ, Li J, Zhang H, Kitamura T, Zhang J, Campion LR, et al. CCL2recruits inflammatory monocytes to facilitate breast-tumour metastasis.Nature 2011;475:222–5.

28. Lee TH, Seng S, Sekine M, Hinton C, Fu Y, Avraham HK, et al. Vascularendothelial growth factor mediates intracrine survival in human breastcarcinoma cells through internally expressed VEGFR1/FLT1. Plos Med2007;4:1101–16.

29. Cao Y, E GQ, Wang EF, Pal K, Dutta SK, Bar-Sagi D, et al. VEGF exerts anangiogenesis-independent function in cancer cells to promote their malig-nant progression. Cancer Res 2012;72:3912–8.

30. Willenborg S, Lucas T, van Loo G, Knipper JA, Krieg T, Haase I, et al. CCR2recruits an inflammatory macrophage subpopulation critical for angio-genesis in tissue repair. Blood 2012;120:613–25.

31. Miquerol L, Gertsenstein M, Harpal K, Rossant J, Nagy A. Multiple devel-opmental roles of VEGF suggested by a LacZ-tagged allele. Dev Biol1999;212:307–22.

32. Witte L, Hicklin DJ, Zhu ZP, Pytowski B, Kotanides H, Rockwell P, et al.Monoclonal antibodies targeting the VEGF receptor-2 (Flk1/KDR) asan anti-angiogenic therapeutic strategy. Cancer Metastasis Rev 1998;17:155–61.

33. Beck B, Driessens G, Goossens S, Youssef KK, Kuchnio A, Caauwe A, et al. Avascular niche and a VEGF-Nrp1 loop regulate the initiation and stemnessof skin tumours. Nature 2011;478:399–403.

34. Rashel M, Alston N, Ghazizadeh S. Protein kinase D1 has a key rolein wound healing and skin carcinogenesis. J Invest Dermatol 2014;134:902–9.

35. Lichtenberger BM, Tan PK, Niederleithner H, Ferrara N, Petzelbauer P,Sibilia M. Autocrine VEGF signaling synergizes with EGFR in tumor cells topromote epithelial cancer development. Cell 2010;140:268–79.

36. Pasparakis M, Courtois G, Hafner M, Schmidt-Supprian M, Nenci A,Toksoy A, et al. TNF-mediated inflammatory skin disease in mice withepidermis-specific deletion of IKK2. Nature 2002;417:861–6.

37. Lapouge G, Beck B, Nassar D, Dubois C, Dekoninck S, Blanpain C. Skinsquamous cell carcinoma propagating cells increase with tumour progres-sion and invasiveness. EMBO J 2012;31:4563–75.

38. Stachelscheid H, IbrahimH, Koch L, Schmitz A, TscharntkeM,WunderlichFT, et al. Epidermal insulin/IGF-1 signalling control interfollicular mor-phogenesis and proliferative potential through Rac activation. EMBO J2008;27:2091–101.

39. SmithMcCune K, Zhu YH, Hanahan D, Arbeit J. Cross-species comparisonof angiogenesis during the premalignant stages of squamous carcinogen-esis in the human cervix and K14-HPV16 transgenic mice. Cancer Res1997;57:1294–300.

40. Coussens LM, Raymond WW, Bergers G, Laig-Webster M, Behrendtsen O,Werb Z, et al. Inflammatory mast cells up-regulate angiogenesis duringsquamous epithelial carcinogenesis. Gene Dev 1999;13:1382–97.

41. Wilgus TA, Matthies AM, Radek KA, Dovi JV, Burns AL, Shankar R, et al.Novel function for vascular endothelial growth factor receptor-1 on epi-dermal keratinocytes. Am J Pathol 2005;167:1257–66.

42. Kurschat P, Bielenberg D, Rossignol-Tallandier M, Stahl A, Klagsbrun M.Neuron restrictive silencer factor NRSF/REST is a transcriptional repressorof neuropilin-1 and diminishes the ability of semaphorin 3A to inhibitkeratinocyte migration. J Biol Chem 2006;281:2721–9.

43. Zhu JW, Wu XJ, Luo D, Lu ZF, Cai SQ, Zheng M. Activation of VEGFR-2signaling in response to moderate dose of ultraviolet B promotessurvival of normal human keratinocytes. Int J Biochem Cell Biol2012;44:246–56.

44. Gerber HP,Malik AK, Solar GP, ShermanD, Liang XH,Meng G, et al. VEGFregulates haematopoietic stem cell survival by an internal autocrine loopmechanism. Nature 2002;417:954–8.

Ding et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research342

45. Lee S, Chen TT, Barber CL, Jordan MC, Murdock J, Desai S, et al. AutocrineVEGF signaling is required for vascular homeostasis. Cell 2007;130:691–703.

46. Liu Y, Berendsen AD, Jia S, Lotinun S, Baron R, Ferrara N, et al. IntracellularVEGF regulates the balance between osteoblast and adipocyte differenti-ation. J Clin Invest 2012;122:3101–13.

47. Qian BZ, Pollard JW. Macrophage diversity enhances tumor progressionand metastasis. Cell 2010;141:39–51.

48. Pollard JW. Macrophages define the invasive microenvironment in breastcancer. J Leukoc Biol 2008;84:623–30.

49. Linde N, Lederle W, Depner S, van RooijenN, Gutschalk CM,Mueller MM.Vascular endothelial growth factor-induced skin carcinogenesis dependson recruitment and alternative activation of macrophages. J Pathol 2012;227:17–28.

50. Clauss M, Gerlach M, Gerlach H, Brett J, Wang F, Familletti PC, et al.Vascular permeability factor: a tumor-derived polypeptide that inducesendothelial cell andmonocyte procoagulant activity, and promotesmono-cyte migration. J Exp Med 1990;172:1535–45.

51. Cramer T, Yamanishi Y, Clausen BE, Forster I, Pawlinski R, Mackman N,et al. HIF-1 alpha is essential for myeloid cell-mediated inflammation (vol112, pg 648, 2003). Cell 2003;113:419-.

52. Mildner M, Weninger W, Trautinger F, Ban J, Tschachler E. UVA and UVBradiation differentially regulate vascular endothelial growth factor expres-sion in keratinocyte-derived cell lines and in human keratinocytes. Photo-chem Photobiol 1999;70:674–9.

53. Rodust PM, Stockfleth E, Ulrich C, Leverkus M, Eberle J. UV-inducedsquamous cell carcinoma–a role for antiapoptotic signalling pathways.Br J Dermatol 2009;161 Suppl 3:107–15.

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 343

VEGF Deletion Inhibits Tumorigenesis in HPV8 Mice