ALTERED TUMOR VESSEL MATURATION AND PROLIFERATION INPLACENTA GROWTH FACTOR-PRODUCING TUMORS: POTENTIALRELATIONSHIP TO POST-THERAPY TUMOR ANGIOGENESIS ANDRECURRENCEAlice P. TAYLOR*, Marisol RODRIGUEZ, Kelly ADAMS, David M. GOLDENBERG and Rosalyn D. BLUMENTHAL

Garden State Cancer Center, Center for Molecular Medicine and Immunology, Belleville, NJ, USA

Cells in human tumor xenografts express similar levels ofangiogenic growth factors before treatment. After radioim-munotherapy (RAIT) surviving tumor cells upregulate angio-genic growth factors, including placenta growth factor(PlGF), in a tumor-specific pattern. To determine the role ofpost-treatment PlGF expression on blood vessel recovery,tumor xenografts were assayed for post-RAIT vessel density(CD34�), proliferation (PCNA�) and maturity (SMA� peri-cytes/mural cells). To further analyze the role of PlGF inblood vessel formation, PlGF-containing Matrigel implantswere also assessed in a similar manner. The xenografts pro-ducing post-treatment PlGF increased CD34� microvesseldensity 2- to 4-fold over untreated controls (p < 0.05) within3 weeks of RAIT treatment. The proportion of mature mi-crovessels (SMA�) decreased. Pericyte coverage and densityof microvessels remained stable in the tumor that expressedneither PlGF nor VEGF after treatment. Hemoglobin con-tent of PlGF-containing Matrigel implants was 5.7-fold that ofanti-PlGF/anti-VEGF treated controls (Day 6, p < 0.03 ). Thevessel density in PlGF-implants averaged 36.8 � 10.6/mmcompared to 4.9 � 6.5/mm2 in controls (p < 0.001). Vesselsof PlGF-implants were lined by vWF� cells, which weremostly flt-1�. These findings point to a role for PlGF in rapidrestoration of tumor blood supply after treatment and thus,to enhanced likelihood of tumor regrowth. Likewise, the cellsof primary human tumors that upregulate PlGF after treat-ment may be more likely to survive and form recurringtumors. Prevention of this angiogenic response to treatmentmay require administration of anti-angiogenic therapy dur-ing, rather than after, treatment.© 2003 Wiley-Liss, Inc.

Key words: angiogenesis; neoplasms; carcinogenesis; PlGF

The efficacy of radioimmunotherapy (RAIT), which typicallydistributes in the perivascular space,1 may be due, in part, todestruction of tumor endothelium, even though radiolabeled anti-bodies target tumor cell antigens. RAIT-mediated damage to tumorblood supply is a phenomenon predicted mostly from externalbeam radiation studies;2 in vivo histologic evidence for RAIT-induced endothelial damage is limited.1 Because tumors vary intheir susceptibility to RAIT and other treatments and becausetumor growth is dependent on adequate vascularization, the extentof vessel damage and the post-treatment angiogenic potential ofindividual tumors may predict whether the tumor will regrow aftertreatment.

One mechanism crucial to recurrence may be induction ofangiogenesis, driven by growth factors produced by survivingtumor cells.3 Growth factors that promote recovery may include,but are not limited to, vascular endothelial growth factor (VEGF),placenta growth factor (PlGF) and the angiopoietins-1 and -2(ang-1 and -2). It is already known that the abnormal over-expression of VEGF is responsible, in part, for the tumor vesselpathology of primary malignancies,4,5 but its role in promotingmetastases or regrowth of tumors after treatment remains unclear.Even though PlGF is ectopically over-expressed in a few primaryhuman tumors, its role in tumor vessel morphology or density isnot established.6 Post-RAIT expression of ang-1 and its tightlyregulated antagonist, ang-2, may play a role in restoring circulationby promoting vessel sprouting and tight bonds between endothelial

and mural cells (ang-1), or vessel remodeling by destabilizingendothelial-mural cell bonds and promoting endothelial cell pro-liferation (ang-2 in synergy with VEGF7).

In a prior study using xenograft tumor models of varied suscep-tibility to RAIT, 3 of 4 tumors upregulated PlGF (HT-29, GW-39[both colon carcinomas] and Calu-3 [lung carcinoma]), whereasonly 1 of 4 over-expressed VEGF after RAIT (HT-29).8 One of the4 tumor models, LoVo (colon carcinoma), the most susceptible toRAIT, expressed neither VEGF nor PlGF, but upregulated onlyang-1. Post-RAIT upregulation of VEGF or PlGF, based on pre-vious growth inhibition studies using the same tumor models,9correlated with relative resistance to RAIT. Our study extendsthese results by investigating the relationship between these tumor-specific differences in angiogenic growth factor expression andangiogenic response to RAIT. Microvessels were evaluated be-cause the microenvironment of the tumor microvessel, isolatedfrom other vessels and connective tissue, differs from that of largervessels. In addition, microvessel density is a prognostic indicatorof disease progression in patients with a variety of tumors.10

Mural cell contact with endothelial cells is a measure of mi-crovessel maturity. SMA� mural cells associated with arteries andveins are defined as smooth muscle cells; SMA� cells associatedwith capillaries and tumor microvessels are pericytes.11,12 In nor-mal tissue, mural cells are a source of growth factors for repair andmaintenance of the underlying endothelial cells.11,12 Vascular mal-formations have an abnormally high frequency of mural cellssurrounding the endothelium, but in tumors the mural cell coveringis often interrupted.1,13 If mural cell coverage is interrupted orlacking, growth signals from surrounding tissue are more likely toinfluence the endothelium.7,14 Prolonged lack of mural cell contactcan signal endothelial cell death and regression of vessels.14 Be-sides being an indicator of vascular functional maturity, smoothmuscle cells and pericytes are also major producers of VEGF.11,12

To analyze vessel recovery after RAIT, we reasoned that endo-thelial cell density, the proportion of endothelial cells that prolif-erate and mural cell covering of vessels would highlight anychanges in the tumor’s post-treatment angiogenic response. Endo-thelial cells and proliferating nuclei associated with vessels werevisualized by immunohistochemistry (IHC) for CD34 or prolifer-ating nuclear antigen (PCNA).15–17 The frequency of smooth mus-

Grant sponsor: New Jersey Commission for Cancer Research; Grantnumber: 01-1079-CCR-N-0 and NJCRR 02/03-1105-CMM-NO; Grantsponsor: USPHS; Grant number: R01 CA60764.

*Correspondence to: Garden State Cancer Center, 520 Belleville Ave-nue, Belleville, NJ 07109. Fax: �973-844-7020 or 844-7120.E-mail: ataylor@gscancer.org.

Received 22 July 2002; Revised 21 November 2002; Accepted 15January 2003

DOI 10.1002/ijc.11059

Int. J. Cancer: 105, 158–164 (2003)© 2003 Wiley-Liss, Inc.

Publication of the International Union Against Cancer

cle actin-positive (SMA�) mural cells, either smooth muscle cellsassociated with arteries and veins, or pericytes associated withmicrovessels, was also determined by IHC for SMA in treated anduntreated tumors.

To test the hypothesis that PlGF stimulates vessel formation, weimplanted PlGF containing Matrigel plugs subcutaneously on theflanks of nude mice.18 This assay allows quantitative estimation ofvessel density in the matrix by measurement of the amount ofhemoglobin and direct observation of vessels by immunohisto-chemical analysis. Vessel density, morphology and type of cellssurrounding vessels were assessed.

MATERIAL AND METHODS

Tumors and treatment of miceHuman tumor lines HT-29, LoVo and Calu-3 were purchased

from American Type Tissue Collection (Manassas, VA). LoVoand HT-29 are colon adenocarcinomas; Calu-3 is a lung adeno-carcinoma. LoVo, 107 HT-29 or Calu-3 cells were implanted s.c.onto the flanks of nude mice. When tumors were about 0.5 cm3 involume, RAIT was commenced with 1 i.v. injection of 240 �Ci of131I labeled monoclonal antibody MN-14 anti-carcinoembryonicantigen (anti-CEA murine IgG) for HT-29 and LoVo19 or 295 �Ci131I-RS-7-3G11 murine IgG for Calu-3 anti-epithelial glycopro-tein-1(EGP-1),20 representing the maximum tolerated dose foreach tumor. Untreated (UT) (n � 3–11/time point/experiment) andRAIT-treated tumor samples (n � 3–6/time point/experiment)were harvested at Weeks 1, 3 and 5 after RAIT, before tumorstypically show signs of regrowth (i.e., at 6–10 weeks after RAIT).Untreated 3- and 5-week samples were assayed for all tumorsexcept Calu-3, for which UT, 1-, 3- and 5-week samples wereassayed.

Determination of vessel count and proliferating cell index oftreated and untreated tumor xenografts

To measure regional expression of vascular markers, IHC wascarried out on 5-�m paraffin-embedded sections. Sections weredeparaffinized, then rehydrated, blocked with normal serum andincubated with either anti-PCNA (mAb, Santa Cruz BiotechnologyInc., Santa Cruz, CA), anti-SMA (mAb, Sigma, St. Louis, MO), oranti-CD34 (mouse reactive, goat polyclonal, Santa Cruz) at aconcentration of 10 �g/ml. Appropriate negative antibody controlswere run for each sample. Biotinylated secondary antibody wasapplied, non-specific peroxidases were quenched with H2O2 andthen slides were exposed to avidin-biotin-HRP conjugate. Signalwas developed with HRP substrate, 3,3�-diaminobenzidine tetra-hydrochloride (DAB, from Sigma or Santa Cruz Biotechnology).Slides were lightly counterstained with hematoxylin (Sigma).

Stained slides were examined at 100� to determine consistencyof staining and overall distribution of stained vessels. Positivelystained endothelial cells (CD34�) and nuclei (PCNA�) or peri-cyte/smooth muscle cells (SMA�) (all associated with a lumen)were counted at 400�. The first field was generally located adja-cent to a necrotic area and subsequent contiguous fields werecounted across the tumor toward the outer border of the tumor.Only microscopic fields with viable tumor cells were included inthe counts. Two researchers examined at least 5 fields/slide (each);the separate researcher’s results were combined and averaged foreach sample.

Slides were evaluated independently for total vessels and mi-crovessels. Microvessels were defined as any lumen that wasseparate from adjacent vessels and tumor cells.21 To determine thelarger, non-microvessel fraction, the number of microvessels foreach sample was subtracted from the total number of vessels; thisfraction is referred to as larger vessels.

Vessel counts are reported as the mean � SD of the number ofpositively stained vessels/mm2 (1 400� field � 0.196 mm2), from5–10 400� fields/sample/time point, 4–6 samples/time point, 2separate experiments for all tumors.

Proliferating cell index (PCI) was calculated as the (mean num-ber of PCNA� vessels/mm2 divided by mean number of CD34�vessels/mm2) � 100. PCI was calculated for each sample and thenall PCIs for that time point were averaged; values shown are�SD.11,21 Proportion of SMA� vessels was determined by calcu-lating: (mean number of SMA� vessels/mm2 � mean number ofCD34� vessels/mm2) � 100. Percentages for each individualtumor sample were determined separately, then averaged for PCI.

ImplantsUse of Matrigel implants was based on the method of Passaniti

et al.18 with some variation. Growth factor-depleted Matrigel(Becton-Dickinson, Bedford, MA) was mixed with 250 ng/mlrecombinant human PlGF or VEGF (R&D Systems, Minneapolis,MN and Genentech, San Francisco, CA, respectively). Controlswere implants containing PBS or both anti-PlGF/anti-VEGF(2,000 ng/ml, each antibody) (Santa Cruz). Further control im-plants also contained 250 ng/ml PlGF plus anti-PlGF (2,000 ng/ml)or VEGF plus anti-VEGF (2,000 ng/ml). Nude mice were lightlyanesthetized and 500 �l of the Matrigel mixture was implantedsubcutaneously on their flanks. Three or 6 days later, implantswere removed, washed 2� in PBS and flash frozen in pre-weighedtubes. Implants with visible hematomas were eliminated. Frozenimplants were spin-dried and their dry weight determined. Theimplants were re-hydrated in 0.75 ml 0.1% Triton X-100 (Sigma)for 1 hr and then disrupted by centrifugation and vigorous pipet-ting, followed by centrifugation at 14,000g for 15 min. The he-moglobin content of the resulting supernatants was assessed byabsorbance at 405 nm and compared to standards.18 Two experi-ments were carried out using 3–5 mice/experimental group.

A portion of the implants was fixed in formalin, H&E stained orprobed by IHC as described; sections were not counterstained.Number of positively stained vessels was calculated as described.The percentage of vessels positive for CD34, SMA, flk-1 and flt-1were calculated as the number of marker-positive vessels/numbervWF positive vessels.

StatisticsRAIT-treated vessel counts were subjected to 1-way (unpaired)

ANOVA analysis vs. untreated controls (p � 0.05 was consideredsignificant).

RESULTS

Results of a previous study suggest that surviving tumor cellsmobilize angiogenic growth factors after initial treatment in atumor-specific manner.8 To investigate how RAIT impacts vesselstructure and regrowth, we determined larger vessel and microves-sel (CD34�) density, mural cell (SMA�) density and endothelialcell proliferation (PCNA�) frequency in UT and Week 3 and 5post-RAIT treated tumor xenografts, 1 lung (Calu-3) and 2 coloniccancers (HT-29, LoVo). HT-29 and Calu-3 are the most resistantto RAIT, as indicated by shortened growth arrest and earliercommencement of tumor regrowth.9

Pericyte/smooth muscle cells after treatmentWe assessed RAITs effect on smooth muscle cell/pericyte den-

sity by staining for SMA.22 As shown in Figure 1a–d, largerSMA� tumor vessels were visible at low power (100�). By Week3 after RAIT, regression of these vessels was most obvious inCalu-3 and HT-29 (Fig. 1b, Calu-3), but was not as evident inLoVo (Fig. 1d). Table I shows that the decline in the density oflarge vessels is also apparent at high power (400�). When thepercentage of SMA� vessels was calculated (see Material andMethods), only HT-29 showed a significant change in pericyte/mural cell coverage after RAIT (100% for UT declining to 70% byWeek 5; p � 0.05).

As shown in Table II, the percentage of SMA� microvesselsdecreased in both HT-29 and Calu-3 at Weeks 3 and 5 after RAIT(p � 0.05). Interestingly, the proportion of SMA� microvessels

159VESSEL MATURATION IN P1GF-PRODUCING TUMORS

calculated for the untreated colon xenografts, HT-29 and LoVo, issimilar to the 65% reported for primary colon carcinomas21 (TableII). The similarity points to constant microvessel restructuring andsprouting in both primary tumors and in xenografts. DecreasedSMA� cell association with microvessels in the xenografts HT-29and Calu-3 suggest enhanced stimulation of microvessel sproutingby Weeks 3–5, accompanied by diminished pericyte coverage.

Vessel density after RAITCD34-immunoreactive cells associated with a lumen were con-

sidered endothelial cells of larger vessels or microvessels, accord-ing to their location and presence within vessel complexes (seeMaterial and Methods). HT-29 CD34� large vessel density de-creased significantly by Week 3, but recovered by Week 5 (TableIII). LoVo displayed a slight decrease that was non-significant atWeek 3 (Table III).

In contrast to larger vessels, CD34� microvessel density as-sessed in a similar manner increased 4-fold in HT-29 (p � 0.05)and 2.5-fold in Calu-3 (p � 0.05) by Week 3, whereas LoVoremained near untreated levels (Table III, Figure 1i–l). Theseresults suggest that the angiogenic environment of treated HT-29and Calu-3 stimulates microvessel cell proliferation, whereas thatof LoVo is less active.

Proliferation of vessel asociated cells and PCITo assess the proliferation of tumor vessel cells, tumors were

also probed for PCNA, a nuclear marker of cellular proliferation(see Material and Methods). Mean density of PCNA� nucleiassociated with larger vessels in UT tumors were not significantlydifferent from tumor to tumor (20 � 14/mm,2, 41 � 5 and 52 �22/mm2 for Calu-3, HT-29 and LoVo, respectively; Table III).Large-vessel-associated PCNA� nuclei increased 2.1-fold atWeek 3 in Calu-3 (p � 0.05 UT) (Fig. 1e,f; Table III). HT-29proliferating larger vessel cells lagged at 55% of UT at Week 3,but by Week 5, proliferation was significantly higher than un-treated (p � 0.05). LoVo large vessel-associated proliferationdeclined slightly at Week 3 (Figure 1g,h; Table III).

FIGURE 1 – IHC of untreated (UT) and RAIT-treated tumors at Week 3 for SMA, CD34 andPCNA. (a–c) Regression of SMA� larger ves-sels after treatment: UT (a) and treated (b) HT-29; LoVo UT (c) and treated (d). (e–h) PCNA,UT (e) and 3-week treated (f) Calu-3; LoVo UT(g), treated (h). (i–l) CD34, Calu-3 UT (i) andtreated (j) and LoVo UT (k) and treated (l). (m,n)Double-stained Calu-3 showing mural cell cov-erage of vessels before and after RAIT. DAB(brown) SMA; Vector VIP (purple) CD34; nocounterstain. Magnification � 100� (a–d,m,n);400� (e–l).

TABLE I – DENSITY OF SMA � LARGE VESSELS1

TumorDensity of SMA � Large Vessels � SD

UT 3 Week 5 Week

LoVo 63 � 27 40 � 10 49 � 8HT-29 79 � 9 75 � 25 63 � 9Calu-3 97 � 39 60 � 38 47 � 18*1Number of SMA positive large vessels/mm2 at 400 � magnifica-

tion. *p � 0.05 for UT vs. RAIT-treated (ANOVA).

TABLE II – PERCENT SMA � MICROVESSELS1

Tumor% SMA � microvessels � SD

UT 3 Week 5 Week

LoVo 66 � 42% 72 � 24% 60 � 40%HT-29 61 � 30% 34 � 8% 29 � 2%Calu-3 120 � 60% 50 � 20%* 110 � 40%1Number shown is the (no. SMA positive microvessels/mm2 divided

by no. CD34 positive microvessels/mm2) � 100 � SD (400 �magnification). Value 100% indicates more SMA � vessels thanCD34�. *p � 0.05 UT vs. RAIT-treated (ANOVA).

TABLE III – MICROVESSELS POSITIVE FOR CD34�AND PCNA� ENDOTHELIAL CELLS1

Large vessels Microvessels

CD34 PCNA CD34 PCNA

LoVo UT 82 � 24 52 � 22 18 � 10 11 � 73 Weeks 61 � 11 35 � 17 26 � 11 16 � 65 Weeks 75 � 12 51 � 10 24 � 13 16 � 10

HT-29 UT 69 � 13 41 � 5 13 � 2 7 � 33 Weeks 32 � 132 23 � 162 52 � 92 26 � 62

5 Weeks 91 � 7 69 � 132 43 � 172 25 � 62

Calu-3 UT 46 � 25 20 � 14 19 � 11 13 � 83 Weeks 51 � 31 42 � 192 48 � 212 33 � 52

5 Weeks 39 � 21 36 � 9 57 � 272 23 � 72

1Values are mean number of CD34� cells bordering a lumen or(PCNA) vessel-associated nuclei/mm2 � SD, using samples from 2independent experiments (except for Calu-3), n � 5–6 samples/timepoint, 400� magnification–2Indicates p � 0.05 vs. UT controls(ANOVA).

160 TAYLOR ET AL.

Tumor microvessel cell proliferation differed from that of largervessels. Both HT-29 and Calu-3 significantly increased proliferat-ing nuclei associated with microvessels by Week 3 (2.5- and3.8-fold) for Calu-3 and HT-29, respectively) (Table III, p � 0.05for both). Elevated proliferation persisted into Week 5 for bothCalu-3 and HT-29, tumor lines that upregulates angiogenic growthfactors PlGF. LoVo microvessel proliferation stayed slightly aboveuntreated through Week 5.

Even though PCNA� density increased in treated HT-29 andCalu-3 at Weeks 3 or 5, the percentage of PCNA� nuclei associ-ated with a vessel lumen (PCI) did not change significantly, due toa concomitant and proportionate rise in CD34� cell density (TableIV). The PCI for Calu-3 rose from 53% in UT to 74% at Week 1(not shown), suggesting that increased proliferation rate may occurearlier than Week 3.

In vivo activity of P1GF in matrigel implantsMatrigel implants containing 250 pg/ml recombinant human

PlGF or VEGF were subcutaneously implanted on the flanks ofnude mice. H&E stained sections from PlGF- and VEGF-implantsdemonstrated vessels containing red blood cells throughout. Ves-sels of the PBS and anti-PlGF/anti-VEGF controls were locatedalmost entirely around the periphery (Fig. 2a). The vessels ofPlGF-implants were of plieomorphic morphology and larger di-ameter than were those in VEGF-implants. Most of the vessels inthe PlGF-implants were thin walled with what appeared to be asingle layer of surrounding cells (Fig. 2a). VEGF-stimulated ves-sels were smaller in diameter, with vessel walls made up ofmultiple cell layers (Fig. 2a). The vessel density and phenotype ofvessel associated cells in the interior of the implants were assessed.

The density of interior vessels of PlGF-containing implantsdiffered markedly from PBS controls. PlGF- and VEGF-implantsdemonstrated similar vessel density at Day 6, but both weresignificantly higher than PBS controls (PlGF, 36.8 � 10.6/mm;2;VEGF, 27.6 � 9.4/mm2 PBS controls, 4.9 � 6.5/mm,2, p � 0.001)(Fig. 2). At 3 days after implantation, the PlGF- and VEGF-treatedimplants contained migrating cells and vessels with smaller diam-eter, but also with similar density (per mm2): PlGF, 23 � 8;VEGF, 20 � 8.8; control 3.8 � 5.8). These results suggest thatPlGF is able to stimulate the formation of blood vessels at a ratesimilar to VEGF. When anti-PlGF or anti-VEGF was included inPlGF- or VEGF-implants, the density of interior vessels dimin-ished 23% and 25%, respectively.

The phenotype of cells bordering interior implant vessels wasassessed for von Willebrand Factor (vWF), CD34, SMA PlGFreceptor fit-1, and VEGF receptor flk-1. A vessel was defined as alumen surrounded by vWF� cells. Figure 2b summarizes theresults. The proportion of CD34� endothelial cells associated withvWF� interior vessels in PlGF-containing implants was 44%,compared to 35% for VEGF implants (see Material and Methods).SMA� mural cell association with vessels was only 10% in PlGFimplants, whereas in VEGF implants, 51% of vessels wereSMA�. One hundred percent of vessels were associated withflt-1� cells in PlGF implants, contrasted with 31% in VEGFimplants. Anti-PlGF diminished the number of flt-1 positive ves-sels in PlGF implants by 87% (not shown).

Hemoglobin (Hb) content was higher in PlGF- and VEGF-implants when compared to anti-PlGF/anti-VEGF controls (2.72ug/mg, 2.32 ug/mg and 0.48 ug/mg, respectively, p � 0.03, bothPlGF and VEGF, 2 experiments). When compared to PBS controls(1.73 ug/mg), however, the difference was no longer significant.Inclusion of anti-PlGF or anti-VEGF in PlGF or VEGF implants(respectively) brought Hb content close to the PBS control levels(2.01 ug/ml and 1.89 ug/ml, respectively).

DISCUSSION

The purpose of our study was to develop a better understandingof the tumor revascularization process after cytotoxic therapy. Weused 3 tumor models that differentially regulate angiogenic growthfactors8 and that vary in sensitivity to RAIT9 to analyze bloodvessel damage and recovery after RAIT. Although vessel damagewas evident, RAIT did not eradicate all tumor vessels (Fig. 1,Table I). Therefore, tumor blood vessels could rebuild by exten-sion from live vessel tissue or by migration of endothelial andpericyte cells into the tumor parenchyma to form new microves-sels. We observed that LoVo, which produces ang-1 only,8 lackedmicrovessel expansion after RAIT. In the LoVo tumor the princi-pal mechanism of vessel rebuilding may be ang-1-mediated exten-sion of endothelium from pre-existing vessels. In contrast, RAITstimulated rapid and significant sprouting of CD34� microvesselendothelium by Week 3 in HT-29 and Calu-3, tumors that upregu-late VEGF, PlGF, or both (Tables II,III). The role of VEGF intumor vascularization is fairly well established,4,5 but its functionin rebuilding vessels after treatment is not. Our observations sug-gest that VEGF plays a role in post-treatment proliferation ofmicrovessels when it is upregulated. At least 1 study, however,suggests that anti-tumor treatment abrogates the expression ofVEGF.23 In the absence of VEGF, renewal of vessels may dependon an angiogenic growth factor that shares redundant functions,such as PlGF. We have shown that the kinetics of post-RAITmicrovessel renewal is similar in the xenograft, Calu-3, that up-regulates PlGF only and HT-29 that produces VEGF (and PlGF).

In assessing the formation of new vessels, the recruitment ofmural cells must be considered,11 because mural cell (pericyte)association with endothelial cells is crucial to endothelial cells’survival, stability and maturation. Pericytes typically use pre-existing endothelial cells as migration contacts to extend coverageof new endothelium.14,24 When mural cell binding is disruptedduring normal physiological vessel remodeling, the endothelialcells without pericyte coverage are ‘immature’ and are subject togrowth signals from the surrounding tissue.14,22,25,26 As shown inTable II, HT-29 and Calu-3 displayed this immature microvesselphenotype and, thus, endothelial cells from these tumors weremost likely responsive to growth and to survival and proliferationsignals from VEGF/PlGF produced by these tumor cells.

If ang-1, a mediator of tight endothelial cell–pericyte contacts,predominates after treatment, mural cell coverage of new endo-thelium would most likely accompany angiogenesis. Correspond-ingly, LoVo, which expresses ang-1 post-RAIT, maintained pre-treatment levels of mural cell coverage of both microvessels andlarger vessels after treatment. In contrast to ang-1, VEGF, workingsynergistically with ang-2, stimulates endothelial cell proliferation

TABLE IV – PROLIFERATING CELL INDEX1

TumorProliferating cell index for large vessels and microvessels

UT Week 3 Week 5

LoVo 50 � 28 (60 � 23) 61 � 23 (66 � 29) 65 � 8 (70 � 8)HT-29 62 � 14 (54 � 25) 64 � 38 (50 � 10) 78 � 22 (65 � 39)Calu-3 40 � 20 (50 � 20) 100 � 60 (80 � 30) 90 � 30 (40 � 30)1Proliferating cell index equals the (#-PCNA�/mm2/divided by # CD34�/mm2) � 100. PCI for each

tumor sample was determined separately, and then separate PCI’s were averaged. Microvessel values arein parentheses. Results shown are the mean � SD, n � 5–6 for each tumor. p 0.05 vs. UT controls(ANOVA).

161VESSEL MATURATION IN P1GF-PRODUCING TUMORS

and detachment from mural cells7 and therefore the formation ofimmature microvessels. Both HT-29 and Calu-3 displayed in-creased immature microvessel formation, despite the absence of

VEGF in Calu-3. This observation provides more evidence of thefunctional redundancy of VEGF and PlGF in the post-treatmenttumor. Lagging pericyte association with microvessels persisted

FIGURE 2 – H&E-stained Matrigel implants, 100� original magni-fication. (a) PBS-containing implant shows very few vessels in theinterior of the implant. PIGF-containing implant has numerous, largevessels lined mostly by a single layer of cells. VEGF-containingimplant vessels were fewer, with a smaller diameter than PIGF-simulated vessels. (b) Phenotype of vessel-associated cells in implants.Values shown are average number of vessels with positively-stainedcells/mm2 � SD, from 2 separate experiments. Arrow denotes p �0.001 vs. PBS controls by ANOVA analysis.

162 TAYLOR ET AL.

through Week 5 in both HT-29, possibly because of lower ang-2levels in this tumor (Table II).8

A previous report indicated localized collapse of vessels, in-creased basement membrane deposition and decreased vascularpermeability after RAIT (Weeks 2–5).2 At present, there is verylittle information on the effect of ionizing radiation on pericytefunction or activation status. It is possible, however, that pericytesare responsible for these post-RAIT changes in vessel function,because pericytes activated by vessel injury mediate similar effectson endothelium by depositing thickened basement membranes andchanging vessel permeability.27,28

Not all structures resembling blood vessels were positive forCD34. Our purpose in choosing CD34 as an endothelial cellmarker was to highlight non-quiescent endothelial cells and mi-crovessels.26 In treated tumors, we observed basement membrane-lined structures with a lumen, but without detectable endothelialcells. These structures may be RAIT-damaged vessels that havesloughed off dead endothelial cells and are therefore regressing orare inactivated endothelium. It is also possible that vascular mim-icry, the formation of matrix-lined channels first found in primaryuveal melanoma, is operating in these tumors.29 Correspondingly,we observed what appeared to be SMA-positive tumor cells adja-cent to matrix-lined channels at Week 5 (Calu-3). If these channelsare not remnants of damaged vessels, vascular mimicry may bedue to tumor cells functioning like mural cells (e.g., laying downbasement membrane material), rather than as endothelial cells.

We have shown that PlGF is able to stimulate angiogenesis inMatrigel implants (Fig. 2). Formation of functioning blood vesselswas detected in the interior of implants by Day 3. Vessels fromPlGF-implants were generally of larger diameter than VEGF, butthey more closely resembled immature microvessels or capillaries.In this way, the PlGF-containing implants resemble some humancolon carcinomas for which larger microvessel diameter and de-creased VEGF production are predictors of metastasis.30 The in-creased microvessel diameter detected in these primary coloncancers may result from undetected PlGF expression in the tumor.

The cells surrounding vessels in the PlGF-implants differedmarkedly from the VEGF-implants (Fig. 2b). In contrast to VEGF,PlGF-stimulated blood vessels were overwhelmingly flt-1 positive.We have observed increased flt-1 expression when PlGF is in-creased in vivo8 and in our study (Fig. 2b), but it is unclear whethera causal relationship exists between the 2 observations. PlGF-containing implants were also relatively devoid of pericytes/muralcells. Lack of pericyte/mural cell coverage resembles the immaturevessels we detected in xenografts that express post-treatment PlGFand the morphology of microvessels in primary malignanciesreported elsewhere.13 Tumor expression of PlGF may promotemicrovessel immaturity and weakened morphology, facilitatingtumor cell invasion into blood vessels. Therefore, it may be apredictor of metastatic potential.

The results of our study document that PlGF mediates theformation of a vessel network that resembles microvessels de-tected in primary human tumors. Even though microvessel densityand maturity were quite similar in all 3 xenografts before treat-ment, their post-treatment angiogenic phenotype varied consider-ably. In addition, treatment-stressed tumor cells that initiated pro-duction of PlGF (or both PlGF and VEGF) were able toaggressively replenish their microvessels.

Our results raise the question of whether, clinically, humantumors may also vary in their vascular growth factor expressionand angiogenic potential after tumoricidal therapy. To prevent thisangiogenic response, prophylactic administration of multiple anti-angiogenic reagents during, rather than after, primary treatmentcould be effective in blocking tumor cell-mediated angiogenesisand tumor recurrence.

ACKNOWLEDGEMENTS

The guidelines set by Garden State Cancer Center ResearchAnimal Resource Center were followed in experiments usinganimals. We gratefully acknowledge the technical assistance of L.Osorio, R. Craig, B.S., L. Catalano-Gizas, M.S. and M. Hernan-dez, B.S.

REFERENCES

1. Blumenthal RD, Kashi R, Sharkey RM, Goldenberg DM. Quantitativeand qualitative effects of experimental radioimmunotherapy on tumorvascular permeability. Int J Cancer 1995;61:557–66.

2. Mauceri HJ, Hanna NN, Beckett MA, Gorski DH, Staba M-J, StellatoKA, Bigelow KA, Heimann R, Gately S, Dhanabal M, Soff GA,Sukhatme VP, et al. Combined effects of angiostatin and ionizingradiation in antitumour therapy. Nature 1998;394:287–91.

3. Ts’ao C, Molteni A, Taylor JM. Injury-specific cytotoxic response oftumor cells and endothelial cells. Pathol Res Pract 1996;192:1–9.

4. White FC, Carroll SM, Kamps MP. VEGF mRNA is reversiblystabilized by hypoxia and persistently stabilized in VEGF-overex-pressing human tumor cell lines. Growth Factors 1995;12:289–301.

5. Gannon G, Mandiroita SJ, Cui L, Baetens D, Pepper MS, ChristoforiG. Overexpression of vascular endothelial growth factor-A165 en-hances tumor angiogenesis but not metastasis during beta-cell carci-nogenesis. Cancer Res 2002;62:603–8.

6. Donnini S, Machein MR, Plate KH, Weich HA. Expression andlocalization of placenta growth factor and PlGF receptors in humanmeningiomas. J Pathol 1999;189:66–71.

7. Korff T, Kimmina S, Martiny-Baron G, Augustin HG. Blood vesselmaturation in a 3-dimensional spheroidal coculture model: directcontact with smooth muscle cells regulates endothelial cell quiescenceand abrogates VEGF responsiveness. FASEB J 2001;15:447–57.

8. Taylor AP, Osorio L, Craig R, Raleigh JA, Ying Z, Goldenberg DM,Blumenthal RD. Tumor-specific regulation of angiogenic growth fac-tors and their receptors during recovery from cytotoxic therapy. ClinCancer Res 2002;8:1213–22.

9. Blumenthal RD, Sharkey RM, Kashi R, Sides K, Stein R, GoldenbergDM. Changes in tumor vascular permeability in response to experi-mental radioimmunotherapy: a comparative study of 11 xenografts.Tumor Biol 1997;18:367–77.

10. Bochner BH, Esrig D, Groshen S, Dickinson M, Weidner N, NicholsPW, Skinner DG, Cote RJ. Relationship of tumor angiogenesis andnuclear p53 accumulation in invasive bladder cancer. Clin Cancer Res1997;3:1615–22.

11. Eberhard A, Kahlert S, Goede V, Hemmerlein B, Plate KH, AugustinHG. Heterogenicity of angiogenesis and blood vessel maturation inhuman tumors: implications for anti-angiogenic tumor therapy. Can-cer Res 2000;60:1388–1393 [Erratum] 3668.

12. Tsurumi Y, Murohara T, Krasinski K, Chen D, Witzenbichler B,Kearney M, Couffinhal T, Isner JM. Reciprocal relation betweenVEGF and NO in the regulation of endothelial integrity. Nat Med1997;8:879–86.

13. Vikkula M, Boon LM, Mulliken JB. Molecular genetics of vascularmalformations. Matrix Biol 2001;20:327–35

14. Benjamin LE, Hemo I, Keshet E. A plasticity window for blood vesselremodeling is defined by pericyte coverage of the preformed endo-thelial network and is regulated by PDGF-B and VEGF. Development1998;125:1591–8.

15. Peichev M, Naiyer AJ, Pereira D, Zhu Z, Lane WJ, Williams M, OzMC, Hicklin DJ, Witte L, Moore MA, Rafii S. Expression ofVEGFR-2 and AC133 by circulating human CD34 (�) cells identifiesa population of functional endothelial precursors. Blood 2000;95:952–8.

16. Wood HB, May G, Healy L, Enver T, Morriss-Kay GM. CD34expression patterns during early mouse development are related tomodes of blood vessel formation and reveal additional sites of hema-topoiesis. Blood 1997;90:2300–11.

17. Ito A, Nomura S, Hirota S, Suda J, Suda T, Kitamura Y. Enhancedexpression of CD34 messenger RNA by developing endothelial cellsof mice. Lab Invest 1995;72:532–8.

18. Passaniti A, Taylor RM, Pili R, Guo Y, Long PV, Haney JA, PaulyRR, Grant DS, Martin GR. A simple, quantitative method for assess-ing angiogenesis and antiangiogenic agents using reconstituted base-ment membrane, heparin and fibroblast growth factor. Lab Invest1992;67:519–28.

19. Blumenthal RD, Sharkey RM, Natale AM, Kashi R, Wong G, Gold-enberg DM. Comparison of equitoxic radioimmunotherapy and che-motherapy in the treatment of human colonic xenografts. Cancer Res1994;54:142–51.

163VESSEL MATURATION IN P1GF-PRODUCING TUMORS

20. Stein R, Chen S, Sharkey RM, Goldenberg DM. Radioimmuno-therapy of human non-small cell carcinoma of the lung xenograftswith 131I-labeled monoclonal antibody RS7-3G11. Antibody Immun-conj Radiopharm 1991;4:703–11.

21. Fontanini G, Vignatti S, Boldrini L, Chine S, Silvestri V, Lucchi M,Mussi A, Angeletti A, Bevilacqua G. Vascular endothelial growthfactor is associated with neovascularization and influences progres-sion of non-small lung carcinoma. Clin Cancer Res 1997;3:861–5.

22. Stratman A, Risqu W, Plate KH. Cell type-specific expression ofangiopoietin-1 and angiopoietin-2 suggests a role in glioblastomaangiogenesis. Am J Pathol 1998;153:1459–66.

23. Myoung H, Hong SD, Kim YY, Hong SP, Kim MJ. Evaluation of theanti-tumor and anti-angiogenic effect of paclitaxel and thalidomide onthe xenotransplanted oral squamous cell carcinoma. Cancer Lett 2001;163:191–200.

24. Nicosia RF, Villaschi S. Rat aortic smooth muscle cells becomepericytes during angiogenesis in vitro. Lab Invest 1995;73:658–66.

25. Siemeister G, Schirner M, Weindel K, Reusch P, Menrad A, MarmeD, Mariny-Baron G. Two independent mechanisms essential for tu-mor angiogenesis: inhibition of human melanoma xenograft growth

by interfering with either the vascular endothelial growth factor re-ceptor pathway or the tie-2 pathway. Cancer Res 1999;59:3185–91.

26. Augustin HG, Braun K, Telemenakis I, Modlich U, Kuhn W. Pheno-typic characterization of endothelial cells in a physiological model ofblood vessel growth and regression. Am J Pathol 1995;147:339–51.

27. Benjamin LE, Hemo I, Keshet E. A plasticity window for blood vesselremodeling is defined by pericyte coverage of the preformed endo-thelial network and is regulated by PDGF-B and VEGF. Development1998;125:1591–8.

28. Ivarsson M, Sundberg C, Farrokhnia N, Pertoft K, Rubin K, Gerdin B.Recruitment of type I collagen producing cells from the microvascu-lature in vitro. Exp Cell Research 1996;229:335–49.

29. Maniotis AJ, Folberg R, Hess A, Seftor EA, Gardner LM, Pe’er J,Trent JM, Meltzer PS, Hendrix MJ. Vascular channel formation byhuman melanoma cells in vivo and in vitro: vasculogenic mimicry.Am J Pathol 1999;155:739–52.

30. Tsuji T, Sasaki Y, Tanaka M, Hanabata N, Hada R, Munakata A.Microvessel morphology and vascular endothelial growth factor ex-pression in human colonic carcinoma with or without metastasis. LabInvest 2002;82:555–62.

164 TAYLOR ET AL.