Ž .Clinica Chimica Acta 312 2001 135–142www.elsevier.comrlocaterclinchim

Colorectal cancer metastatic phenotype stimulates production byfibroblasts of N-terminal peptide of type III collagen: clinical

implications for prognosis

Daniela Basso a, Claudio Belluco b, Saverio Mazza a, Eliana Greco a,Fosca Della Rocca c, Paolo Pauletto c, Donato Nitti b, Mario Lise b, Mario Plebani a,)

a Department of Laboratory Medicine, UniÕersity Hospital, UniÕersity of PadoÕa, Via Giustiniani 2, 35128 Padua, Italyb Department of Oncological and Surgical Sciences, UniÕersity of PadoÕa, Padua, Italy

c Department of Clinical and Experimental Medicine, UniÕersity of PadoÕa, Padua, Italy

Received 7 June 2001; accepted 25 June 2001

Abstract

Ž .In this study we assessed whether the serum levels of the N-terminal peptide of type III collagen PIIIP , an index of typeIII collagen synthesis, are influenced by colorectal cancer stage, and whether Ain vitroB fibroblast growth and PIIIPproduction could be altered by tumor tissues obtained from metastatic and nonmetastatic colorectal cancer. 208 colorectal

Ž .cancer patients 115 colon and 93 rectum were studied; 54 were stage I, 62 stage II, 37 stage III and 55 stage IV. PIIIPserum levels were significantly higher in stage IV as compared to all other patient groups. The 5-year survival of stage I,stage II, stage III and stage IV patients were 87%, 88%, 32% and 20%, respectively. In the subgroup of stage I and stage II

Ž . Ž .patients considered together, PIIIP )0.5 Urml , but not CEA )5 mgrl serum levels, were predictive for survival.Fibroblast growth was significantly inhibited, while PIIIP production was significantly enhanced, when these cells wereconditioned with colorectal cancer homogenates obtained from patients with distant metastases, than from those withoutdistant metastases. In conclusion, colorectal tumors, when metastatic, stimulate fibroblasts’ PIIIP synthesis and the serumlevels of this peptide might predict patients’ outcome after radical surgery. q 2001 Elsevier Science B.V. All rights reserved.

Keywords: Colorectal cancer; Fibroblasts; Type III collagen; Prognosis; CEA

1. Introduction

Colorectal cancer, the most common malignancyof the gastrointestinal tract, is one of the three mostfrequent malignancies in industrialized countries, theother two being breast and lung cancer. Reportedly,colorectal cancer is responsible for about 14% of all

) Corresponding author. Tel.: q39-0498212792; fax: q39-049663240.

Ž .E-mail address: pad08821@pd.nettuno.it M. Plebani .

cancer-related deaths in the US, and its incidencecontinues to be high.

In patients with colorectal cancer a prognosticfactor of critical importance is the absence or pres-

w xence of metastasis 1 , the development of whichdepends on complex interactions between tumor cells

w xon one hand and the host on the other 2,3 . Themetastatic potential is acquired by only some of theelements in the tumor cell mass, through the devel-opment of genetic and phenotypic modifications.Examples of genetic alterations associated with a

0009-8981r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.Ž .PII: S0009-8981 01 00621-0

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142136

Fig. 1. Serum levels of CEA in the studied colorectal cancerpatients subdivided according to tumor stage. Each square repre-sents five patients.

higher incidence of metastases in colorectal cancerare microsatellite instability, transforming growthfactor-b type II receptor gene mutation, k-ras trans-version, c-myc and stromelysin-3 gene over-expres-

w xsion and p53 mutations 4–6 . Furthermore, some ofthe phenotypic alterations occurring in highly ag-gressive colorectal cancer cells are variations in theexpression of MUC1 mucin, a-catenin, TGF-b1,

w xfibronectin and laminin receptors 7–13 . Metastatictumor cells, moreover, have the ability to produce

Ž .matrix degrading metalloproteinases MMPs , which,Ž .by disrupting the extracellular matrix ECM pro-

teins adjacent to the tumor mass, may favor tumorw xcell migration 6,14–16 . However, we and others

have previously suggested that metastatic tumor cellsmay not only cause the degradation of ECM, butmay also interfere with its synthesis, probably by

wreleasing soluble mediators targeting fibroblasts 17–x23 . Alteration in the fibroblastic synthesis of type I,

III and IV collagens seems to occur mainly in pa-w xtients with metastatic tumors 17,18,20,22,23 .

Ž .The aims of this study were therefore to: 1ascertain whether the serum levels of the N-terminal

Ž .peptide of type III collagen PIIIP , an index of typew xIII collagen synthesis 24 , are influenced by colorec-Ž .tal cancer stage; and 2 make an Ain vitroB analysis

of the effects on fibroblast growth and PIIIP produc-tion by tumor tissues obtained from metastatic andnonmetastatic colorectal cancer.

2. Material and methods

For the study we utilized sera obtained from 208Žpatients 117 men, 91 women; mean age 64 years,

Fig. 2. Serum levels of the N-terminal peptide of type III procolla-Ž .gen PIIIP in the studied colorectal cancer patients subdivided

according to tumor stage. Each square represents five patients.

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142 137

Table 1Ž .Sensitivity of CEA circulating values above 4 mgrl in classify-

ing colorectal cancer patients subdivided on the basis of tumorstage

Ž .Stage Sensitivity %

Ž .I ns54 10Ž .II ns62 41.9Ž .III ns37 8.9Ž .IV ns55 72.7

.range, 27–92 before they underwent surgery forŽ .histologically confirmed colorectal carcinoma CRC

at the Department of Surgery of the University ofPadova, Italy, between 1991 and 1995. The bloodsamples, obtained from fasting patients, had beencentrifuged at 3000=g for 10 min and the seraseparated and stored at y80 8C until biochemicaldeterminations were made.

Ž .The tumor site was the colon in 115 55% pa-Ž .tients and the rectum in 93 45% . The histologic

grade was assessed according to the World HealthŽ . w x Ž .Organization WHO criteria 25 . Fifty-two 25%

Ž .of the tumors were well differentiated, 134 64%Ž .moderately differentiated and 22 11% poorly dif-

ferentiated. According to the UICC classification

w x Ž .TNM staging system 26 , 54 26% of the tumorsŽ . Ž .were stage I, 62 30% stage II, 37 18% stage III

Ž . Žand 55 26% stage IV in 12 of these cases liverresection was performed for hepatic metastases, whilein the other 43 cases the disease was unresectable,

.and alternative therapy was given . Patients withstage III CRC underwent standard adjuvant chemo-

Ž 2therapy six cycles of 5-FU, 370 mgrm with leu-2 .covorin 100 mgrm rday for 5 days .

The following were measured in sera: theŽ .aminoterminal peptide of type III collagen PIIIP

Ž .using an IRMA procedure CIS Diagnostici, FranceŽ .and carcinoembryonic antigen CEA using an ELISA

procedure by means of a commercially availableŽmethod Immuno 1 System, Bayer, Tarry Town,

.NY .ŽFrom six patients without liver metastases group

. Ž .1 and from five with liver metastases group 2 , theŽ .following tissue samples were also obtained: a

Ž .tumor tissue, b macroscopically AnormalB mucosasampled at a distance of 3 cm from the primary

Ž . Žtumor, and c a metastasis sample from the five.group 2 patients . All tissue samples were immedi-

ately frozen in liquid nitrogen and stored at y80 8Cuntil used for Ain vitroB experiments with fibroblasts.

Fig. 3. Cumulative proportion surviving of stage I and stage II colorectal cancer patients subdivided according to pre-operative serum levelsof PIIIP below or above 0.5 Urml.

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142138

Fibroblasts were obtained from the saphena veinsof patients who underwent saphenectomy. After re-moval of the intima, the adventitia was cut into small

Žpieces, which were digested with collagenase 2.mgrml for 90 min at 37 8C. The digested material

Žwas cultured in DMEM FCS 12.5%qglutamine2%qgentamycin 0.1%, GibcoBRL, Life Technolo-

.giese, San Giuliano Milanese, Italy at 37 8C, in aŽ .CO incubator 5% CO . The tissue culture medium2 2

was replaced every 3 days and, when fibroblasts hadreached confluence, they were detached from thetissue culture flasks using mild trypsin digestionŽ0.25% trypsin solution, GibcoBRL, Life Technolo-

.giese , and then used for the experiments.

AIn vitroB experiments with fibroblasts: all tissuesamples were homogenized in a 0.05 molrl phos-

Ž .phate buffer, at pH 7.4 1:20, wrv , using the Ultra-ŽTurrax T25 homogenizer Janke and Kunkel, IKA-

. Ž .Labortechnik . The fibroblasts 2000rwell wereplated in a 96-well cell culture plate starting with acell suspension of 20,000 fibroblastsrml. To thedifferent wells we added 10 ml of tumor sample

Ž .homogenates tumor-conditioned fibroblasts, TCF ,Ž10 ml of normal mucosa homogenates normal mu-

.cosa-conditioned fibroblasts, NMCF or 10 ml ofŽmetastasis homogenates metastasis-conditioned fi-

.broblasts, MCF . Each condition was repeated inquadruplicate. After 1, 4 and 8 days of culture in the

Fig. 4. Fibroblasts cell growth, assessed by means of the XTT assay, when conditioned with 10% of tissue homogenates obtained fromŽ . Ž .tumor samples tumor-conditioned fibroblasts, TCF , from normal mucosa normal mucosa-conditioned fibroblasts, NMCF or from

Ž . Ž . Žmetastasis metastasis-conditioned fibroblasts, MCF of group 1 patients without liver metastases or group 2 patients with liver.metastases patients. Each experimental condition was performed in triplicate. Median values are reported.

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142 139

above conditions, an aliquot of conditioned fibroblastculture medium was collected and stored at y20 8Cfor up to 3 months, until PIIIP determination. Fibrob-last cell growth was assessed by means of the XTT

Ž .assay Roche Diagnostics, Milano, Italy : 50 ml ofthe reagent was added to each well and after 4-hincubation at 37 8C in the CO incubator, absorbance2

at 450 nm was read on a microplate spectrophotom-eter.

The statistical analysis of data was made usingŽ .analysis of variance Anova one-way , Bonferroni’s

test for pairwise comparisons and Kruskal–WallisAnova. Survival curves were plotted using the Ka-plan–Meier method and differences were assessed

w xby the log-rank test 27 .

3. Results

Figs. 1 and 2 show the individual data for circu-lating CEA and PIIIP in the colorectal cancer pa-tients, subdivided on the basis of tumor stage. Me-dian CEA values were significantly higher in stage IIand stage III than in stage I patients, and in stage IV

Žcompared with all the other patient groups Krus-kal–Wallis analysis of variance: Hs50.25, p-

.0.001 . PIIIP serum levels were significantly higherin stage IV patients than in all the other groupsŽ .Hs13.83, p-0.01 .

Table 1 reports the sensitivity of CEA at a cut-offvalue of 4 mgrl in diagnosing colorectal cancerpatients subdivided on the basis of tumor stage.

Neither PIIIP nor CEA levels were influenced bythe differentiation grade of colorectal tumors studiedŽKruskal–Wallis test: chi-squares0.065, p: ns for

.PIIIP and 0.522, p: ns for CEA .At analysis of survival, conducted in 198 of the

Ž208 patients two died perioperatively and eight were.lost to follow-up , the 5-year survival of stage I,

stage II, stage III and stage IV patients were 87%,88%, 32% and 20%, respectively. When patientswere considered overall, at univariate analysis an

Žassociation was found between elevated CEA )5. Ž .mgrl or PIIIP )0.5 Urml serum levels and a

Žshorter survival p-0.001 and p-0.01, respec-.tively . In the subgroup of stage I and stage II

patients considered together, CEA serum levels wereŽ .not predictive for survival p: ns ; in this same

subgroup, the 46 patients who had PIIIP serum levelsabove 0.5 Urml had a 5-year survival of 77%, whilethe remaining 64, who had PIIIP levels below 0.5

Ž .Urml, had a 5-year survival of 95% ps0.0182Ž .Fig. 3 .

Fig. 4 shows the findings of fibroblast cell growth,assessed using XTT assay, when conditioned with10% of tissue homogenates obtained from tumor

Ž .samples tumor-conditioned fibroblasts, TCF , fromŽnormal mucosa normal mucosa conditioned fibro-

. Žblasts, NMCF or from metastasis metastasis-.conditioned fibroblasts, MCF ; the patients were

Žsubdivided into two groups: group 1 absence of. Ž .metastases and group 2 presence of metastases .

Fibroblasts grew significantly over time when condi-tioned with normal mucosa or tumor tissues from

Žgroup 1 patients analysis of variance for repeatedmeasures: Fs5.7, p-0.05 and Fs7.82, p-0.01,

.respectively , whereas fibroblast cell growth wasinhibited by normal mucosa and tumor tissue from

Žgroup 2 patients Fs1.4, p: ns and Fs0.6, p: ns,.respectively . MCF growth was inhibited to a similar

Ž .extent Fs0.9, p: ns .Table 2 reports mean values, standard errors of

the mean and the results of a statistical analysis of

Table 2Ž .Mean values"standard deviation SD and statistical analysis

Ž . Ž .analysis of variance of PIIIP levels Urml measured in fibrob-last conditioned tissue culture medium after 1, 4 and 8 days

1st day, 4th day, 8th day,mean"SD mean"SD mean"SD

Group 2 NMCF 1.05"0.1 1.34"0.2 1.72"0.1Group 1 NMCF 1.15"0.2 1.37"0.2 1.67"0.2

)Group 2 TCF 1.11"0.1 1.35"0.2 2.14"0.1Group 1 TCF 1.13"0.2 1.38"0.1 1.70"0.1Group 2 MCF 1.12"0.1 1.30"0.1 1.89"0.3Anova one-way Fs0.47, Fs0.14, Fs5.95,

p: ns p: ns p-0.01

Fibroblasts were conditioned with 10% of tissue homogenatesŽobtained from tumor samples tumor-conditioned fibroblasts,

. ŽTCF , from normal mucosa normal mucosa-conditioned fibrob-. Žlasts, NMCF or from metastasis metastasis-conditioned fibrob-

. Žlasts, MCF tissues obtained from group 1 absence of liver. Žmetastases, ns6 and group 2 presence of liver metastases,

.ns5 patients.) Bonferroni’s test for pairwise comparisons: p-0.05 as

Ž . Ž . Ž .compared to: a group 2 NMCF, b group 1 NMCF, and cgroup 1 TCF.

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142140

Fig. 5. PIIIP production by fibroblasts after 8 days of culture conditioned with 10% of tissue homogenates obtained from tumor samplesŽ . Ž .tumor-conditioned fibroblasts, TCF , from normal mucosa normal mucosa-conditioned fibroblasts, NMCF or from metastasisŽ . Ž .metastasis-conditioned fibroblasts, MCF . Closed dotssgroup 1 patients absence of liver metastases ; open squaressgroup 2 patientsŽ .presence of liver metastases patients.

PIIIP measured in fibroblast-conditioned tissue cul-ture medium after 1, 4 and 8 days of culture.

Fig. 5 shows values for PIIIP measured in thefibroblast tissue culture medium after 8-day culture.

4. Discussion

In the first part of this study, we investigatedwhether the progression of colorectal cancer is ac-companied by a modification in fibrogenesis. In serafrom a large series of colorectal cancer patients wemeasured the N-terminal peptide of type III procolla-gen, a segment cleaved from the type III collagenprecursor molecule during the formation and growth

w xof collagen fibers 24 . In the present paper we didnot measure the serum levels of the C-terminal pep-

Ž . w xtide of type I collagen PIP since elsewhere 22 , wedemonstrated that PIIIP is a more sensitive indexthan PIP in detecting residual colorectal tumor afterapparently radical surgery. Serum levels of PIIIPwere significantly increased only in patients withdistant metastases, this pattern being close to that ofCEA, the sensitivity of which is known to be highlyunsatisfactory in the diagnosis of localized colorectal

w xtumors 28 .

We also ascertained whether preoperative circulat-ing levels of CEA and PIIIP were prognostic forpatients’ outcome. Considering the patients overall,increased levels of both indices were associated witha shorter 5-year survival. This may have dependedupon the association between CEA and PIIIP serumlevels and tumor stage. Importantly, in patients with-

Ž .out lymph-node or liver metastases stages I and II ,PIIIP, unlike CEA levels, were predictive for deathfrom recurrence after radical surgery. While theoverall 5-year survival of stage I and stage II patients

Ž .were quite similar 87% and 88%, respectively , asignificant difference in outcome was observed whenthese patients were subdivided on the basis of PIIIPlevels: survival were 95% and 77% when PIIIPvalues were below or above 0.5 Urml, respectively.The worse prognosis observed in the latter groupmight be explained by the presence of occult mi-crometastases, inducing fibrogenesis, which in turn,increases circulating PIIIP levels.

The association found between increased PIIIPserum levels and advanced tumor stage may have

Ž .been due to enhanced: 1 liver fibrogenesis sec-Ž .ondary to hepatic metastasis; 2 collagen synthesis

induced by aggressive colorectal cancer cells. Theformer phenomenon is feasible, as liver fibrogenesis,an important response mechanism, is induced by

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142 141

different pathogenetic noxae causing liver cell necro-sis. To test the latter hypothesis, we performed aseries of Ain vitroB experiments with isolated fibrob-

Žlasts. Tissue homogenates primary tumor, metastasisand apparently normal colon mucosa adjacent to the

.primary tumor obtained from patients with distantmetastases, compared with tissue homogenates ob-tained from patients without distant metastasesŽprimary tumor and apparent normal colon mucosa

.adjacent to the primary tumor , significantly inhib-ited fibroblast cell growth, but also significantlystimulated PIIIP production. These results indicatethat potentially metastatic cancer cells may not onlyadhere more avidly than their nonmetastatic counter-

w xparts to collagen 29,30 , but may also enhance thesynthesis of collagen, which might AguideB cancercell migration outside the primary tumor. These AinvitroB results are in line with our above-reportedclinical findings, but not with findings reported in a

w xprevious paper by us 22 , in which we reported thatpostoperative PIIIP levels were not increased in pa-tients who subsequently developed recurrence. How-ever, these results may be only apparently contradic-tory in that PIIIP level variations in the sera ofcolorectal cancer patients observed soon after surgerymay mainly reflect fibroblast growth in wound heal-ing; as demonstrated by us Ain vitroB, this growthmay be inhibited by residual foci of metastatic tumorcells.

The mechanism by which invasive tumor cellsmodify fibroblast growth and collagen synthesis isprobably correlated to the occurrence of molecularchanges during progressive transformation from earlycancer to the invasive and metastatic phenotype. Ithas been suggested that one of the changes involvedis TGF-b1 gene overexpression, accompanied byincreased plasma levels of this cytokine, which cor-

w xrelates with tumor stage 13 . This multifunctionalcytokine, which can trigger different intracellular

w xevents in the same cell type 31 , can induce collagensynthesis not only by fibroblasts, but also by tumor

w x w xcells 32 , inducing CTGF transcription 33 , and itmay be the putative colorectal cancer-associated fi-brogenetic pro-metastatic factor. However, further

Ž .studies are needed in order to elucidate which is areŽ .the colorectal cancer-associated factor s altering fi-

broblast growth and PIIIP production.

References

w x1 Hermanek P. pTNM and residual tumor classifications: prob-lems of assessment and prognostic significance. World JSurg 1995;19:184–90.

w x2 Jiang WG, Puntis MCA, Hallett MB. Molecular and cellularbasis of cancer invasion and metastasis: implications fortreatment. Br J Surg 1994;81:1576–90.

w x3 Liotta LA, Stetler-Stevenson WG. Tumor invasion andmetastasis: an imbalance of positive and negative regulation.

Ž .Cancer Res 1991;51:5054s–9s Suppl. .w x4 Iacopetta BJ, Welch J, Soong R, House AK, Zhou X-P,

Hamelin R. Mutation of the transforming growth factor-btype II receptor gene in right-sided colorectal cancer: rela-tionship to clinicopathological features and genetic alter-ations. J Pathol 1998;184:390–5.

w x5 Iniesta P, de Juan C, Caldes T, et al. Genetic abnormalities´and microsatellite instability in colorectal cancer. CancerDetect Prev 1998;22:383–95.

w x6 Porte H, Chastre E, Prevot S, et al. Neoplastic progression ofhuman colorectal cancer is associated with overexpression ofthe stromelysin-3 and BM-40rSPARC genes. Int J Cancer1995;64:70–5.

w x7 Dorudi S, Hanby AM, Poulsom R, Northover J, Hart IR.Level of expression of E-cadherin mRNA in colorectal can-cer correlates with clinical outcome. Br J Cancer 1995;71:614–6.

w x8 Ilyas M, Tomlinson IPM, Hanby A, Talbot IC, Bodmer WF.Allele loss, replication errors and loss of expression ofE-cadherin in colorectal cancers. Gut 1997;40:654–9.

w x9 Mafune K-I, Ravikumar TS, Wong JM, Yow H, Chen LB,Steele GD. Expression of a Mr 32,000 laminin-binding pro-tein messenger RNA in human colon carcinoma correlateswith disease progression. Cancer Res 1990;50:3888–91.

w x10 Nakamori S, Ota DM, Cleary KR, Shirotani K, Irimura T.MUC1 mucin expression as a marker of progression andmetastasis of human colorectal carcinoma. Gastroenterology1994;106:353–61.

w x11 Raftopoulos I, Davaris P, Karatzas G, Karayannacos P,Kouraklis G. Level of a-catenin expression in colorectalcancer correlates with invasiveness, metastatic potential, andsurvival. J Surg Oncol 1998;68:92–9.

w x12 Stallmach A, Von Lampe B, Orzechowski H-D, Matthes H,Riecken E-O. Increased fibronectin-receptor expression incolon carcinoma-derived HT 29 cells decreases tumorigenic-ity in nude mice. Gastroenterology 1994;106:19–27.

w x13 Tsushima H, Kawata S, Tamura S, et al. High levels oftransforming growth factor b1 in patients with colorectalcancer: association with disease progression. Gastroneterol-ogy 1996;110:375–82.

w x14 Hewitt RE, Leach IH, Powe DG, Clark IM, Cawston TE,Turner DR. Distribution of collagenase and tissue inhibitor of

Ž .metalloproteinases TIMP in colorectal tumors. Int J Cancer1991;49:666–72.

w x15 Watson SA, Morris TM, Robinson G, Crimmin MJ, BrownPD, Hardcastle JD. Inhibition of organ invasion by the

( )D. Basso et al.rClinica Chimica Acta 312 2001 135–142142

Ž .matrix metalloproteinase inhibitor batimastat BB-94 in twohuman colon carcinoma metastasis models. Cancer Res 1995;55:3629–33.

w x16 Werb Z, Vu TH, Rinkenberger JL, Coussens LM. Matrix-de-grading proteases and angiogenesis during development andtumor formation. APMIS 1999;107:11–8.

w x17 Jukkola A, Tahtela R, Tholix E, Vuorinen K, Blanco G,¨ ¨ ¨Risteli L, et al. Aggressive breast cancer leads to discrepantlevels of the type I procollagen propeptides PINP and PICP.Cancer Res 1997;57:5517–20.

w x18 Kauppila S, Stenback F, Risteli J, Jukkola A, Risteli L.¨Aberrant type I and type III collagen gene expression inhuman breast cancer in vivo. J Pathol 1998;186:262–8.

w x19 Kimoto T, Watanabe S, Hyodoh F, Saito T. Collagen fiberand proliferation as a mechanism of cancer prevention andregression induced by extract from mycobacterium tubercolo-sis: correlation between clinical observation and animal ex-periments. Cancer Detect Prev 1988;11:173–89.

w x20 Korenaga D, Funahashi S, Yano K, Maekawa S, Ikeda T,Sugimachi K. Relationship between peritoneal collagen typeIV concentrations and the presence of disseminated metas-tases in gastric cancer. Arch Surg 1995;130:769–73.

w x21 Matsui H, Kubochi K, Okazaki I, Yoshino K, Ishibiki K,Kitajima M. Collagen biosynthesis in gastric cancer: im-munohistochemical analysis of prolyl 4-hydroxylase. J SurgOncol 1999;70:239–46.

w x22 Plebani M, Basso D, Roveroni G, De Paoli M, Galeotti F,Corsini A. N-terminal peptide of type III procollagen: apossible predictor of colorectal carcinoma recurrence. Cancer1997;79:1299–303.

w x23 Yoshida K, Yokozaki H, Niimoto M, Ito H, Ito M, Tahara E.Expression of TGF-b and procollagen type I and type III inhuman gastric carcinomas. Int J Cancer 1989;44:194–8.

w x24 Plebani M, Burlina A. Biochemical markers of hepatic fibro-sis. Clin Biochem 1991;24:219–39.

w x25 Morson B.C., Sobin L.H. , Histologic typing of intestinaltumors: WHO technical report. Geneva, World Health Orga-nization, 1976.

w x26 Sobin LH, Wittekind C. UICC TNM classification of malig-nant tumors. Fifth edn. New York: Wiley-Liss, 1997.

w x27 Kaplan EL, Meier P. Non parametric estimation for incom-plete observations. J Am Stat Assoc 1958;53:457–81.

w x28 Plebani M, De Paoli M, Basso D, Roveroni G, Giacomini A,Galeotti F, Corsini A. Serum tumor markers in colorectalcancer staging, grading, and follow-up. J Surg Oncol 1996;62:239–44.

w x29 Danen EHJ, van Muijen GNP, van de Wiel-van Kemenade E,Jansen KFJ, Ruiter DJ, Figdor CG. Regulation of integrin-mediated adhesion to laminin and collagen in humanmelanocytes and in non-metastatic and highly metastatichuman melanoma cells. Int J Cancer 1993;54:315–21.

w x30 Stefani AL, Basso D, Panozzo MP, et al. Cytokines modulateMIA PaCa 2 and CAPAN-1 adhesion to extracellular matrixproteins. Pancreas 1999;19:362–9.

w x31 Fukami J, Tsuji K, Ueno A, Ide T. Transforming growthfactor-b1 has both promoting and inhibiting effects on induc-tion of DNA synthesis in human fibroblasts. Exp Cell Res1995;216:107–12.

w x32 Panozzo MP, Basso D, De Paoli M, Carraro P, Burighel D,Plebani M. Cytokines may influence tumor growth andspread. An in vitro study in two human cancer cell lines. IntJ Clin Lab Res 1996;26:240–4.

w x33 Wenger C, Ellenrieder V, Alber B, et al. Expression anddifferential regulation of connective tissue growth factor inpancreatic cancer cells. Oncogene 1999;18:1073–80.