Endothelial cells increase the radiosensitivity oforopharyngeal squamous carcinoma cells incollagen gel

Lorenzo Rossia,b,*, Francesco Boccardob,c, Renzo Corvod

aLaboratory of Comparative Oncology, Instituto Nazionale per la Ricerca sul Cancro, Largo Rosanna Benzi,10-16132 Genoa, ItalybDepartmentofOncology,BiologyandGenetic,UniversityofGenoa,LargoRosannaBenzi,10-16132Genoa,ItalycProfessorial Unit of Medical Oncology, Instituto Nazionale per la Ricerca sul Cancro, Largo Rosanna Benzi,10-16132 Genoa, ItalydDivisionofRadiotherapy, InstitutoNazionaleper laRicercasulCancro,LargoRosannaBenzi,10-16132Genoa,Italy

Received 23 June 2003; accepted 7 August 2003

Summary We assessed the radiosensitivity of human HSCO oropharyngeal squamouscarcinoma cells in the presence of paracrine factors produced by human HECV umbilicalvein endothelial cells. To this end the cells were embedded in separate collagen gels atthe density of 1�106 cells/ml each, and the two gels were placed in a well of a six-wellplate, sharing the same medium but without physical contact (two-gel model). Themedium was not changed during the observation period to ensure the accumulation ofsoluble factors. On day 2 of culture the gels were irradiated with 0, 0.5, 1, 2 and 8 Gray(Gy) and on day 7 of culture they were disaggregated and cell survival evaluated by theMTT assay. Results were compared with proper untreated and irradiated controlgroups. Under these experimental conditions it was found that: (1) HSCO and HECVcells influenced reciprocally their behaviour in the two-gel model, in terms that cellsurvival was enhanced and inhibited, respectively; (2) as expected, HSCO cells weremore radioresistant in collagen gel than in monolayer; (3) on the average the survivalof HECV cells was enhanced at low radiation doses, irrespective of whether they werecultured alone or with HSCO cells in the two-gel model and (4) HSCO cells displayed ahigh radioresistance when irradiated alone at doses from 0.5 to 8 Gy. However, whenco-cultured with HECV cells in the two-gel model, they become highly radiosensitivealready at the dose of 2 Gy, while none of them survived at the dose of 8 Gy. Thisradiosensitizing effect was specifically induced by paracrine factors circulating in themedium, supporting the notion that stromal endothelial cells may be essential compo-nents of a metabolic circuitry supplying solid tumors with radiosensitizing factors.# 2003 Elsevier Ltd. All rights reserved.

KEYWORDSEpithelial-mesenchymal

interactions;

Tissue remodeling;

Collagen gel model;

Ex-vivo radiotherapy;

Cellular co-cultures;

Angiogenic factors;

Oral tumors

1368-8375/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.

doi:10.1016/j.oraloncology.2003.08.009

Oral Oncology (2004) 40 214–222

www.elsevier.com/locate/oraloncology

* Corresponding author. Tel.: +39-010-5600200; fax: +39-010-5600208.E-mail address: lorenzo.rossi@istge.it (L. Rossi).

Introduction of this model system to radiation effects was asses-

During the natural history of carcinomas, at leasttwo phases of cellular aggregation are distinguished,an initial phase represented by small, uniformclones of epithelial transformed cells, and a moreadvanced phase characterized by cellular aggre-gates made up of an heterogeneous mix of normaland transformed epithelial cells at various stagesof malignancy, and of stromal cells, includingfibroblasts, myofibroblasts, and endothelial andinflammatory cells.1 These diverse mesenchymaland epithelial cell populations are connected by anintricate, largely altered network, destined to sus-tain tumor progression. This tumor—stroma rela-tionship emphasizes the importance of acontinuous dialogue between tumor cells and thesurrounding microenvironment in order to exploitits energies. There is now ample evidence thattumor cell survival depends on a supportivestroma.2

The role of angiogenesis in the growth and survi-val of solid tumors is two-edge. For one aspect,angiogenesis appears as a necessary requisite inorder for small aggregates of transformed cells tomature into fully grown tumors.3 In vitro studieshave indicated that carcinoma cells promote theproliferation of endothelial cells.4 On the otherhand, a non-angiogenesis-dependent pathway oftumor growth has been reported,5,6 while otherauthors have indicated that the reduction orabsence of endothelial cells in the stromal com-partment may increase the aggressivity of malig-nant cells.7

Radiation exposure may damage tissue integrityby altering the flow of information at the epithe-lial-mesenchymal interaction. The growth of avariety of tumors transplanted in SCID and nudemice was found to primarily depend on the radio-sensitivity of the host-stromal components.8 Murinenontumorigenic mammary epithelial cells developedinto tumors in irradiated but not in unirradiated fatpads, an effect ascribed to radiation-induced dis-ruption of solid tissue interactions.9 In mouse intes-tine, ionizing radiation induced apoptosis ofmicrovascular endothelial cells, and only seconda-rily, damaged the epithelial cells.10 On the contrary,an increased angiogenesis expression was reportedin human patients exposed to radiation therapy.11

In this study we present a new in vitro co-culturemodel intended to evaluate the functional inter-action between head and neck squamous carci-noma cells (H&NSCC) and endothelial cells grown inseparate collagen gels, in conditions in which thetwo cell types shared the same culture medium butwere precluded from physical contact. The response

sed in terms of cell proliferation and survival. Themetabolic cooperation derived from the exchange ofsoluble factors in this experimental setting impac-ted on the growth of both cell types, and it resultedin the increased radiosensitivity of the tumor cells.

Materials and methods

Cell culture

The human HSCO86 oropharyngeal squamouscarcinoma cell line was isolated from a post-irradiation squamous carcinoma of the oropharynxand retained epithelial morphology after morethan 25 passages in culture. The cells were main-tained in RPMI-1640 supplemented with 10% fetalbovine serum and enriched with 0.5% gentamycin.The human HECV umbilical vein endothelial cells(ATCC, Rockville, MD) were maintained in DMEMsupplemented with 10% fetal bovine serum andenriched with 2% l-glutamine and 0.5% gentamy-cin. Both cell lines were incubated at 37 �C in ahumidified incubator (5% CO2, 95% air). Confluentcultures were harvested with trypsin-EDTA andcells were counted and centrifuged. The pelletswere mixed with collagen or were seeded inmonolayer (all the above reagents were fromEuroClone, Wetherby, UK). Both cell types werecharacterized immunocytochemically as reportedin Table 1.

Table 1 Immunocytochemical detection of selectedproteins in HSCO and HECV cells grown in collagen gel

Ab anti-

Cells

HECV

HSCO

Vimentin

++ — Laminin + + Factor VIII + ++ VEGF � ++ VEGFR (Flk-1) ++ NT Integrin b1 +++ �

FGF-2

+ — E-cadherin — — EGFR ++ — EGF ++ — Pan-cytocheratins — +

The frequency and intensity of staining were as fol-lows:—, no staining; �, staining obvious at X100 in 1—20% of the cells; +, staining obvious at X40 in 21—70%of the cells; ++, staining obvious at X10 in 21—70% ofthe cells; +++, staining obvious at X10 in 100% of thecells. NT, not tested.

Endothelial cells increase the radiosensitivity of oropharyngeal squamous carcinoma cells 215

Collagen gel preparation

The stock solution was prepared by dissolving 100mg acid-soluble Type I collagen from calf skin (ICNBiomedicals, Irvine, CA) in 40 ml sterile 0.1% aceticacid in distilled water. The final ready-to-use col-lagen solution was obtained from 1 volume of a 2:1mixture of 10� DMEM and 0.34 M NaOH mixed with4 volumes of the stock solution. This solution, thecomplete collagen mixture to serve as the3-dimensional substrate for cell growth, will notgel if kept in ice. After harvest, HSCO and HECVcells were resuspended in culture medium, coun-ted and adjusted to the chosen experimental cellnumber. Following centrifugation, the pellets wereresuspended in 1 ml of the complete collagensolution kept at 4 �C. Experiments were performedusing two distinct cell culture models outlined inFig. 1. In the first model, called the mixed-cellmodel, HSCO and HECV cells were mixed in thesame gel. A volume of 50 ml of this mixed cell sus-pension was dropped into each well of a six-wellplate (Corning, Corning, NY). In the second model,called the two-gel model, 50 ml of a cell suspensioncontaining HSCO cells was dropped into a well of asix-well plate. Another 50 ml of a cell suspensioncontaining HECV cells was carefully dropped in thesame well, making sure that the gels were physi-cally distanced. In the matched control groups, onegel was empty. In any case, the collagens wereallowed to gelify in the incubator and after 1 h, 4ml of medium was added per well. The media werenot changed for the entire experimental period,

generally fixed in 7 days of culture. Appropriatecontrols were used throughout.

Irradiation

Cell-containing gels and monolayer cell cultureswere irradiated with 6 Mv photons using a LinearAccelerator (Varian, Palo Alto, CA). Irradiation wasdelivered with the gantry set at 180� and dosed todmax using a source-to-axis distance of 100 cm.Irradiation occurred at a dose rate of 2.5 Gy/min ina flask placed on top of 1.5 cm of plexiglass. Thisarrangement was used to prevent underdosing ofcells as a result of the physical ‘‘buildup dose’’effect. The gels were kept at room temperature forless than 5 min during the irradiation period.Depending on experimental group, cells receivedsingle irradiation doses of 0.5, 1, 2, 4, 8 and 16 Gy ondays 1 or 2 of culture. The gels assigned to theuntreated control groups were exposed to the samemanipulations, except that they were not irradiated.

MTT assay

To measure cell proliferation in collagen gels, weused the colorimetric method based on the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bro-mide (MTT) assay (Sigma Diagnostics). Briefly, 1 mgMTT (200% of a stock solution in PBS) was added totube, each containing a single gel in 1 ml serum-free medium, and the tubes were incubated at37 �C for 4 h. The medium was then removed bycentrifugation, and 1 ml of collagenase (200 U/mlin serum-free M199) (Sigma Diagnostics, S. Louis,Missouri) was added to each tube to allow digestionfor 1 h at 37 �C. The collagen gel was then dis-sociated by vortexing. After being rinsed with PBS,the digested samples were centrifuged at 5140 RPMfor 15 min. The supernatants were discarded, thepellets were dissolved in DMSO, and the solutionswere read in a Beckman DU-70 spectrophotometerat 570 nm. A standard curve was prepared utilizinga known concentration of cells before each experi-ment. Production of formazan crystals, and there-fore the intensity of color after their dissolution, isproportional to the number of viable cells. Thenotations HECV (HSCO) and its reverse HSCO(HECV), meant that the MTT assay was performedon either cell line co-cultured with the cells inbrackets in the two-gel model.

Histology and immunocytochemistry

Cell cultures using Lab Tek 2-chamber slides(Nalge, Nunc, Naperville, IL) were fixed with 3.5%paraformaldehyde in PBS and were immunostained

Figure 1 Schematic representation of the cell culturemodels designed to study the interaction of different celltypes in collagen gel. The aim of the mixed-cell model isto evaluate the effects of direct cell-cell contact,including physical contact and autocrine stimuli. That ofthe two-gel model is to evaluate the effects of paracrinefactors, given that each cell type is cultured in separategels nurtured by the same culture medium.

216 L. Rossi et al.

using the avidin-biotin complex (ABC) amplifica-tion system and anti-mouse and anti-rabbit sec-ondary antibodies (Dako, Glostrup, DK). Thefollowing primary antibodies were used: anti-Vimentin, polyclonal anti-Factor VIII (Dako);monoclonal anti-pan-cytokeratins and polyclonalanti-laminin (Sigma Diagnostics); monoclonal anti-EGFR, polyclonal anti-FGF-2, polyclonal anti-VEGFand polyclona anti-VEGFR (Flk-1) (Santa Cruz,Santa Cruz, CA) and monoclonal anti-b1-integrin(Chemicon International, Temecula, CA).

Morphometric parameters and statisticalanalysis

Values are given as means�SEM. There werethree to four independent set of gels per group,and all experiments were repeated at least twice.P<0.05 was considered statistically significant.Statistical analysis was performed with Graph-PadPrism version 3 (GraphPad Software, SanDiego, CA). Immunostaining was evaluatedaccording to staining intensity and percentage asfollows:—, no staining; �, staining obvious at X100in 1—20% of the cells; +, staining obvious at X40 in21—70% of the cells; ++, staining obvious at X10 in21—70% of the cells; +++, staining obvious at X10in 100% of the cells.

Results

Immunocytochemical characterization ofHSCO and HECV cells

Immunocytochemical staining confirmed themesenchymal and epithelial nature of HECV andHSCO cells, respectively (Table 1). In particular,Factor VIII and VEGF proteins were abundantlyexpressed by HSCO cells while VEGFR, integrin b1and EGF and EGFR were pronounced on HECV cellsbut practically absent on HSCO cells.

Reciprocal influence of HSCO and HECV cells

We first assessed the consequences of mixingHSCO and HECV cells in collagen gel in terms of cellsurvival as determined by the MTT assay, and com-pared the outcome with that obtained by growingthe cells in separate gels in the same well. In bothapproaches the gels were incubated for 7 days,during which time the culture medium was notchanged, to allow the accumulation of cell-derivedsoluble factors. The results are summarized inFig. 2. When HECV and HSCO cells were mixedtogether in the same gel at the initial density of

5�105 and 1�106 cells/ml, respectively, cell survi-val was 4.5�106 cells/ml, as compared to 3.5�106

and 5.6�106 cells/ml in the groups with HECV andHSCO cells alone, respectively (Fig. 2A). While cellsurvival in the mixed-cell group was roughly halfthe sum of the values detected in the controlgroups, this approach was clearly unsuitable toquantify the relative proportion of surviving cells.To do this HSCO and HECV cells were embedded inseparate gels at the density of 1�106 cells/mleach, and one gel per cell type was placed in thesame well of a six-well plate, making sure that thetwo gels were physically separated. The controlgroups still contained two gels/well, but one gelwas empty and the other was embedded witheither HECV or HSCO cells. As shown in Fig. 2B, theresults of this approach were quite stricking inthat, when both cell types were incubated in thesame well, the growth of HECV cells was inhibitedand that of HSCO cells was enhanced, compared tothe respective controls (in both cases P<0.0001).Specifically, the survival of HECV cells fell to avalue as low as 1.3�106 cells/ml compared to anaverage of 3.7�106 cells/ml found in the controlgroup, and that of HSCO cells increased to a valueof 5.2�106 cells/ml, up from 3.9�106 cells/mldetected in the control group (Fig. 2B). Thesefindings were confirmed by extending the assay to 6

Figure 2 Comparative survival of HSCO and HECV cellsin the two collagen gel models. (A) In the mixed-cellmodel cells were embedded in collagen gel, alone (den-sity of 1�106 and 5�105 cells/ml, respectively) or mixedtogether; (B) in the two-gel model cells were embeddedin separate collagen gels in a well of a six-well plate. Onegel contained HSCO cells and the other gel containedHECV cells, both at the density of 1�106 cell/ml. In thecontrol groups one gel was empty. The MTT assay wasperformed on day 7 of culture. Values are means�SEM(bars) of triplicate incubations.

Endothelial cells increase the radiosensitivity of oropharyngeal squamous carcinoma cells 217

and 9 days of incubation (data not shown). Theseresults established firmly the two-gel model as areliable approach to analyze the interaction of dif-ferent cell types in a three-dimensional context,with specific reference to the role of paracrinefactors in experimental carcinogenesis.

Cellular reaggregation increases theradioresistance of HSCO cells

Since culture conditions influence strongly thebehaviour of HSCO cells,12 we determined radia-tion effects on HSCO cells grown separately inmonolayer and in collagen gel exposed to singledoses of 0, 2, 4, 8 and 16 Gy. First, cells were see-ded in six-well dishes at the density of 1�104 cells/well, irradiated 24 h later and maintained up to 7days in culture. The results, as obtained by thetrypan blu-exclusion test, are reported in Fig. 3.On day 6 of culture it was apparent a dose—response effect (P40.01 compared to control) inthat the dose of 2 Gy caused growth inhibition in44% of the cells compared to untreated control, avalue that increased to 68% in the group with 4 Gyand was more than 98% in the groups with 8 and 16Gy (Fig. 3). Second, collagen gels were embeddedwith 1�106 cells/ml and irradiated on day 9 ofculture, a time when cells were already reag-gregated into spheroids. The results with the MTTassay, performed on day 23 of culture, are sum-marized in Fig. 4. More cells survived in the groupsexposed to 2 and 4 Gy than in control (5.3�106

cells/ml and 5.5�106 cells/ml, compared to

4.3�106 cells/ml, respectively, P<0.0001). Thedose of 8 Gy did not inhibit the growth of HSCOcells that remained at the density of 4.2�106 cells/ml, a value comparable to that of the controlgroup, while more than 6�105 cells/ml were stillalive in the high dose of 16 Gy (Fig. 4). Theseresults confirmed that in a three-dimensional set-ting the response of epithelial cells to radiationeffects is diversified, and on the average radio-resistance is higher, compared to growth in mono-layer. This was mainly ascribed to the fact that incollagen gel HSCO cells reaggregated as multi-cellular spheroids, a structure usually containing acentral necrosis and found to be resistant to arange of toxic agents.13

Paracrine factors modulates theradiosensitivity of HSCO and HECV cells

Next, we determined the radiosensitivity ofHSCO and HECV cells co-cultured in the mixed- andtwo-gel models described above, following thesame procedure and dosing schedules. We firsttested radiation effects in the mixed-cell modelconsisting of gels embedded with HECV and HSCOcells, mixed together at the densities of 5�105 and1�106 cells/ml, respectively. Control groups con-tained either one or the other cell type. The gelswere irradiated with 0, 1 and 2 Gy on day 2 of cul-ture, and on day 7 of culture they were submittedto the MTT assay. The results are reported in Fig. 5.The survival of HECV cells alone was stimulatedfrom a value of 1.9�106 cells/ml in the controlgroup to values of 2.5�106 and 2.3�106 cells/ml inthe 1 and 2 Gy dose groups, respectively (P<0.005in both cases). Comparable figures in the groups

Figure 3 Radiation effects on HSCO cells cultured inmonolayer. Cells were seeded in six-well plates at thedensity of 1�104 cells/ml. On day 1 of culture they wereirradiated with single doses of 0, 2, 4, 8 and 16 Gy. cellswere counted in a Thoma’s chamber. Values are mean-s�SEM (bars) of six wells per group.

Figure 4 Response of HSCO cells to radiation effects incollagen gel. Cells were seeded at the initial density1�106 cells/ml. On day 9 of culture gels were irradiatedwith 0. 2, 4, 8 and 16 Gy. The MTT assay was performedon day 23 of culture. Values are means�SEM (bars) oftriplicate incubations.

218 L. Rossi et al.

with HSCO cells alone were 4.2�106, 4.0�106 and4.3�106 cells/ml, respectively. When HECV andHSCO cells were mixed together in the same gel,the number of cells surviving radiation effects was3.6�106 cells/ml and 2.6�106 cells/ml (P<0.0001)in the groups exposed to 1 and 2 Gy, respectively,down from the value of 3.9�106 cells/ml detectedin the untreated control (Fig. 5).To document radiation effects on HSCO and HECV

cells co-cultured in the two-gel model, cells wereembedded into separate collagen gels at the den-sity of 1�106 cells/ml each. On day 2 of culturethey were irradiated with a single dose of 0, 0.5, 1,2 and 8 Gy and on day 7 the gels were submitted tothe MTT assay. The results are summarized inFig. 6. The average number of HECV cells alone was3.9�106 cells/ml in the control group and it was3.6�106, 3.8�106, 4.3�106 and 1.9�106 cells/mlin the groups with 0.5, 1, 2 and 8 Gy, respectively(this last group was significantly lower, atP<0.0001, compared to control). When HECV cellswere co-cultured with HSCO cells, only 1�106

cells/ml were found alive in the control group.However, their survival rate rose to 1.6�106,1.9�106, 2.5�106 cells/ml in the groups exposed to0.5, 1 and 2 Gy, respectively, while it dropped to avalue of 0.6�106 cells/ml in the group with 8 Gy (inall groups P<0.0001 compared to control). Theaverage survival rate of HSCO cells alone was2.9�106 cells/ml in the control group and it was2.9�106, 2.8�106, 2.5�106 and 1�106 cells/ml in

the groups irradiated with 0.5, 1, 2 and 8 Gy,respectively (for the last two groups P<0.0002compared to control). Following the co-culturewith HECV cells in the two-gel model, the survi-val of untreated HSCO cells rose to a value of5�106 cells/ml. Comparable figures in the groupstreated with 0.5 and 1 Gy were 4.7�106 and4.8�106 cells/ml, respectively, which were notsignificantly different from the untreated con-trol. However, this situation changed dramati-cally in the two highest dose groups, in thatalready at the dose of 2 Gy cell survival declinedto a value as low as 0.7�106 cells/ml (P<0.0001compared to control), and practically no cellswere found alive in the group treated with 8 Gy(Fig. 6).To characterize the expression of selected fac-

tors in the two-gel model, HECV and HSCO cellswere cultured as indicated and the gels exposed to0, 1 and 2 Gy on day 2 of culture. On day 4 of cul-ture cells were immunostained for the expressionof VEGF and EGFR proteins. As shown in Table 2, inthe control groups VEGF was expressed by HSCOcells and very poorly by HECV cells, while EGFR wasexpressed by HECV cells and not at all by HSCOcells. Irradiation changed this immunocytochem-ical profile in that it increased the expression ofVEGF in HECV and HSCO cells cultured in the samewell, while leaving unaffected the expression ofEGFR.

Figure5 Radiosensitivity of HSCO and HECV cells in themixed-cell model. The cells were embedded in collagengel, alone at the density of 1�106 and 5�105 cells/ml,respectively, or mixed together. On day 2 of culture thegels were irradiated with a single dose of 0, 1 and 2 Gyand on day 7 of culture cells were disaggregated andsubmitted to the MTT assay. Values are means�SEM(bars) of triplicate incubations.

Figure 6 Modulation of the radiosensitivity of HSCOand HECV cells in the two-gel model. Cells were embed-ded in separate collagen gels at the density of 1�106

cells/ml. One gel with HSCO cells and one gels with HECVcells were placed in a well of a six-well plate. In thecontrols one gel was empty. On day 2 of culture the gelswere irradiated with 0, 0.5, 1, 2 and 8 Gy. The MTT assaywas performed on day 7 of culture. Values are mean-s�SEM (bars) of triplicate incubations. (asterisk indi-cates absence of cell survival).

Endothelial cells increase the radiosensitivity of oropharyngeal squamous carcinoma cells 219

Discussion

The present investigation provides compellingevidence that paracrine factors regulated thebehaviour of endothelial and squamous carcinomacells co-cultured in collagen gel. The contributionof metabolic factors to this effect became quanti-tatively manifest when HECV and HSCO cells wereco-cultured in the two-gel model, where thegrowth of HECV cells was strongly inhibited and, onthe contrary, the growth of HSCO cells was greatlyenhanced, compared to matched controls.This growth inhibition of HECV cells in the pre-

sence of HSCO cells is at odd with most publishedreports stressing that paracrine factors released bycancer cells, including head and neck squamouscarcinoma cells, support endothelial cell survivaland growth.4 While we are presently unable toexplain this discrepancy, the possibility that fac-tors intrinsic to the two-gel model dictated thisresult was ruled out in a study in which HSCO cellswere replaced by A431 skin squamous carcinomacells. In this instance, in fact, no reciprocal inter-ference in the proliferation rate of both A431 andHECV cells occurred (data not shown). Given theexperimental conditions established in the two-gelmodel, one or more secreted proteins may havebeen involved in this inhibition. In particular, theexpression of the endothelial growth factor VEGFwas higher in the malignant cells as compared tothe endothelial cells. Possibly, not all the availableVEGF protein ligated to its receptor VEGFR2/Flk-1,a VEGF-specific tyrosine kinase receptor highly

expressed in HECV cells. A sizable fraction mayhave remained soluble in the culture medium, andit has been reported that an excess of free unli-gated VEGF in the microenvironment may be det-rimental to the survival of endothelial cells.14,15 Inany event, there are emerging indications that theprocess of tumor angiogenesis is much more com-plicated than thought before.16 In particular, insolid tumours such as melanomas and head andneck carcinomas, epithelial malignant cells cancontribute to tumor angiogenesis by a processcalled ‘‘vasculogenic mimicry’’, in addition todirecting the aggregation of endothelial cells fromthe stromal compartment.17 In keeping with thisphenomenon, retinoic acid was found to induceHSCO cells to undergo an epithelial to mesenchy-mal transition into endothelial-like cells, shaping acomplex network resembling intricate capillaryneovascularization.12 This plasticity suggests thatHSCO cells may have been able to regulate thehomeostasis of HECV cells, thus accounting fortheir growth inhibition observed in the presentinvestigation.While HSCO cells inhibited the growth of HECV

cells, HECV cells greatly enhanced the survival ofHSCO cells in the same paracrine environment.Numerous reports have stressed the importance ofangiogenesis in the development of solid tumors,and in vitro studies have documented the cap-ability of endothelial cells to enhance the pro-liferation of malignant epithelial cells.3,18 In thepresent context this effect was possibly, but by nomeans exclusively, triggered by b1 integrins, func-tionally ubiquitous adhesion molecules abundantlyexpressed by HECV cells and scarsely present onHSCO cells (Table 1). These proteins were collec-tively reported to play an important role in thesurvival and invasive behaviour of squamous carci-noma cells.19 In our system b1 integrins may havefavored HSCO cell survival by sparing them fromentering anoikis, the process by which apoptoticsignals are triggered by the loss of cellular atta-chement to the supporting scaffold.20 This wouldbe in keeping with our observation that more than90% of HSCO cells did not survive the spatial stressrepresented by collagen fibers.HSCO cells were much more resistant to radiation

effects in collagen gel than they were in monolayercultures, as indicated by the observation that asingle dose of 8 Gy completely inhibited theirgrowth in monolayer, but was much less effectivein collagen gel. This different outcome illustratesthe importance of a spatial environment, i.e., two-dimensional versus three-dimensional, in theresponse of cells to exogenous toxic agents. How-ever, when HSCO cells were grown under the direct

Table 2 Immunocytochemical detection of VEGF andEGFR proteins in HSCO and HECV cells grown in the 2-gel model

Growth factors andexperimental groups

Radiation dose (Gy)

0

1 2

VEGF

HECV � � �

HECV (HSCO)a

� ++ ++ HSCO ++ ++ ++ HSCO (HECV) � ++ ++ EGFR HECV ++ ++ ++ HECV (HSCO) ++ ++ ++ HSCO — — — HSCO (HECV) — — —

The gels were irradiated with 0, 1 and 2 Gy on day 2 ofculture and stained on day 4 of culture. Immuno-staining evaluation: see legend to Table 1.

a Symbol adopted for the 2-gel model.

220 L. Rossi et al.

influence of HECV-derived paracrine factors in thetwo-gel model, the results changed dramatically inthat HSCO cells became extremely radiosensitive,as shown by the finding that nearly 80% of the cellswere growth inhibited following the exposure to aradiation dose as low as 2 Gy, and inhibition rose toalmost 100% with 8 Gy (see Table 2). Multiplecomponents of the extracellular microenvironmentmay have been involved in determining this out-come, including growth factors. Of these, EGF, agrowth factor produced by HECV cells (but not byHSCO cells) was reported to enhance the radio-sensitivity of epithelial malignant cells.21 In thepresent context HSCO cells reaggregated as multi-cellular spheroids, which were centrally necrotic,and previous reports have shown that EGF canreduce hypoxia in comparable experimental mod-els.22 Reduced hypoxia can rescue cells fromapoptosis, while in the meantime increasing cel-lular radiosensitivity,23 thus suggesting that micro-environmental factors other than inherent tumorcell radiosensitivity are important determinants ofradiocurability.The increased radiosensitivity displayed by HSCO

cells in the presence of endothelial cells raises thequestion of the target cells in radiation oncology.We speculate that this result was implemented byan apoptosis-dependent mechanism, in that HSCOcells were possibly rescued from apoptosis byparacrine factors produced by HECV cells. Theseapoptosis-rescued cells could have been a favoritetarget of ionizing radiation, thus explaining thestrong inhibition induced by a radiation dose as lowas 2 Gy on the survival of HSCO cells in the presenceof HECV cells. A prospected player in the radio-sensitization of HSCO cells could have been thebystander effect, a kind of cell-cell interaction bywhich irradiated cells can modulate the biologicalresponse of surrounding cells not directly targetedby radiation.24 Albeit pertinent to the two-gelmodel, however, this mechanism can hardlyexplain the present results because: (1) As far asthe technical procedure is concerned, both HSCOand HECV cell populations received the sameamount of radiation, and (2) whereas at the dose of2 Gy the survival of HECV cells was enhanced, thatof HSCO cells was inhibited, a result in apparentcontrast with the bystander effect.In general, the response of HECV cells to radia-

tion effects was dual (see Figs. 5 and 6) in that theywere inhibited at the dose of 8 Gy, while their sur-vival was significantly increased at lower radiationdoses, this depending on the initial proliferationrate. This trend was particularly pronounced fol-lowing the co-culture with HSCO cells in the two-gel model, where a dose-dependent manner of

growth was manifest. Presently, no explanation isavailable to account for this behaviour of HECVcells, although our data confirm previous reportsthat low-dose ionizing radiation can promote thegrowth of tissue vasculature.25,26

In summary, the capability of HECV cells to lowerthe amount of radiation needed to destroy HSCOcells, concur with the suggestion that stromal-derived endothelial cells may be essential compo-nents of the metabolic circuitry supplying solidtumors with radiosensitizing factors.27 The three-dimensional model described in this paper mayhelp to analyze the role played by these and otherstromal cells in head and neck carcinogenesis andradiotherapy.

Acknowledgements

We thank Mr. Daniele Reverberi for technical skillin cell culture, Mrs. Anna Calabresi for executingsome immunocytochemical preparations, Dr. Gior-gio Tanara for helping reading the immunocyto-chemical slides and Dr. Giovanni De Pascalis forhelping in statistical analysis. Funds for this workwere provided in part by the Department ofOncology, Biology and Genetic of the University ofGenoa.

References

1. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell2000;100:57—70.

2. Elambaas B, Weinberg RA. Heterologous signals betweenepithelial tumor cells and fibroblasts in carcinoma form-ation. Exp Cell Res 2001;264:169—184.

3. Hanahan D, Folkman J. Patterns and emerging mechanismsof the angiogenic switch during tumorigenesis. Cell 1996;86:353—364.

4. Shemirani B, Crowe DL. Head and neck squamous cell car-cinoma lines produce biologically active angiogenic factors.Oral Oncol 2000;36:61—66.

5. Pezzella F, Pastorino U, Tagliabue E, Andreola S, Sozzi G,Gasparini G, et al. Non-small-cell lung carcinoma tumorgrowth without morphological evidence of neo-angiogen-esis. Am J Pathol 1997;151:1417—1423.

6. Holash J, Maisonpierre PC, Compton D, Boland P, AlexanderCR, Zagzag D, Yancopoulos GD, Wiegand SJ. Vessel coop-tion, regression, and growth in tumors mediated by angio-poietins and VEGF. Science 1999;284:1994—1998.

7. Shirakawa K, Tsuda H, Heike Y, Kato K, Asada R, Inomata M,et al. Absence of endothelial cells, central necrosis, andfibrosis are associated with aggressive inflammatory breastcancer. Cancer Res 2001;61:445—451.

8. Budach W, Taghian A, Freeman J, Gioioso D, Suit HD.Impact of stromal sensitivity on radiation response oftumors. J Natl Cancer Inst 1993;85:988—993.

9. Barcellos-Hoff MH, Ravani SA. Irradiated mammary glandstroma promotes the expression of tumorigenic potential byunirradiated epithelial cells. Cancer Res 2000;60:1254—1260.

Endothelial cells increase the radiosensitivity of oropharyngeal squamous carcinoma cells 221

10. Paris F, Fuks Z, Kang A, Capodieci P, Juan G, Ehleiter D, etal. Endothelial apoptosis as the primary lesion initiatingintestinal radiation damage in mice. Science 2001;293:293—297.

11. Riekki R, Jukkola A, Oikarinen A, Kallioinen M. Radiationtherapy induces tenascin expression and angiogenesis inhuman skin. Acta Derm Venereol 2001;81:329—333.

12. Rossi L, Corvo R. Retinoic acid modulates the radio-sensitivity of head-and-neck squamous carcinoma cellsgrown in collagen gel. Int J Rad Oncol Biol Phys 2002;53:1319—1327.

13. Guirado D, Aranda M, Vilches M, Villalobos M, Lallena AM.Dose dependence of the growth rate of multicellulartumour spheroids after irradiation. Br J Radiol 2003;76:109—116.

14. Millauer B, Wizigmann-Voos S, Schnurch H, Martinez A,Moller N, Risa W, et al. High affinity VEGF binding anddevelopment expression suggest Flk-1 as a major regulatorof vasculogenesis and angiogenesis. Cell 1993;72:835—846.

15. Gorski D H, Beckett M A, Jaskowiak N T, Calvin D P, MauceriH J, Salloum R M, et al. Blockage of the vascular endothelialgrowth factor stress response increases the antitumor effectsof ionizing radiation. Cancer Res 1999;59:3374—3378.

16. Chang YS, di Tomaso E, McDonald DM, Jones R, Jain RK,Munn LL. Mosaic blood vessels in tumors: frequency of can-cer cells in contact with flowing blood. Proc Natl Acad SciUSA 2000;97:14608—14613.

17. Folberg R, Hendrix MJC, Maniotis AJ. Vasculogenic mimicryand tumor angiogenesis. Am J Pathol 2000;156:361—381.

18. Rak J, Filmus J, Kerbel RS. Reciprocal paracrine interac-tions between tumour cells and endothelial cells: the‘‘angiogenesis progression’’ hypothesis. Eur J Cancer 1996;32A:2438—2450.

19. Koivisto L, Grenman R, Heino J, Larjava H. Integrinsalpha5beta1, alphavbeta1, and alphavbeta6 collaborate insquamous carcinoma cell spreading and migration on fibro-nectin. Exp Cell Res 2000;255:10—17.

20. Bonfoco E, Chen W, Paul R, Cheresh DA, Cooper NR. Beta1integrin antagonism on adherent, differentiated humanneuroblastoma cells trigger an apoptotic signaling pathway.Neuroscience 2000;101:1145—1152.

21. Lammering G, Hewit TH, Hawkins WT, Contessa JN, Rear-don DB, Lin P-S, et al. Epidermal growth factor receptor asa genetic therapy target for carcinoma cell radio-sensitization. J Natl Cancer Inst 2001;93:921—929.

22. Levy R, Smith SD, Chandler K, Sadowsky Y, Nelson DM.Apoptosis in human cultured trophoblasts is enhanced byhypoxia and diminished by epidermal growth factor. Am JPhysiol Cell Physiol 2000;278:C982—C988.

23. Lee CG, Heijn M, di Tomaso E, Griffon-Etienne G, Ancu-kiewicz M, Koike C, et al. Anti-vascular endothelialgrowth factor treatment augments tumor radiationresponse under normoxic or hypoxic conditions. CancerRes 2000;60:5565—5570.

24. Mothersill C, Seymour CB. Bystander and delayed effectsafter fractionated radiation exposure. Radiat Res 2002;158:626—633.

25. Levesque L, Lam MH, Allaire P, Mondat M, Houle S, Beau-doin G, et al. Effects of radiation therapy on vascularresponsiveness. J Cardiovasc Pharmacol 2001;37:381—393.

26. Denham JW, Hauer-Jensen M. The radiotherapeutic injury—a complex wound. Radiother Oncol 2002;63:129—145.

27. Geng L, Donnelly E, McMahon G, Lin PC, Sierra-Rivera E,Oshinka H, et al. Inhibition of vascular endothelial growthfactor receptor signaling leads to reversal of tumor resis-tance to radiotherapy. Cancer Res 2001;61:2413—2419.

222 L. Rossi et al.