2013 SSAT PLENARY PRESENTATION

Clinical Significance of Serum COL6A3 in PancreaticDuctal Adenocarcinoma

Christopher Y. Kang & Jonathan Wang & Dierdre Axell-House & Pranay Soni &Mon-Li Chu & Galina Chipitsyna & Konrad Sarosiek & Jocelyn Sendecki &Terry Hyslop & Mazhar Al-Zoubi & Charles J. Yeo & Hwyda A. Arafat

Received: 18 May 2013 /Accepted: 19 August 2013 /Published online: 4 September 2013# 2013 The Society for Surgery of the Alimentary Tract

Abstract Type VI collagen (COL6) forms a microfibrillar network often associated with type I collagen and constitutes a majorcomponent of the desmoplastic reaction in pancreatic ductal adenocarcinoma (PDA). We have demonstrated recently that the 3chain of COL6, COL6A3, is highly expressed in PDA tissue and undergoes tumor-specific alternative splicing. In this study, weinvestigated the diagnostic value and clinical significance of circulating COL6A3 protein andmRNA in PDA. COL6A3 levels in serafrom patients with PDA (n=44), benign lesions (n=46) and age-matched healthy volunteers (n=30) were analyzed by enzyme-linked immunosorbent assays (ELISA). Predictive abilities of COL6A3were examined using receiver operating characteristic (ROC)curves from logistic regression models for PDA versus normal or benign serum levels. Expression levels were correlated withclinicopathological parameters. Real-time PCR was used to analyze the presence of COL6A3 mRNA containing alternative splicedexons E3, E4, and E6. Circulating COL6A3 protein levels were significantly elevated in PDA patients when compared to healthy sera(p =0.0001) and benign lesions (p =0.0035). The overall area under the ROC was 0.975. Log(COL6A3) alone provided gooddiscrimination between PDAand benign lesions (area under the curve (AUC)=0.817), but combinedwith CA19-9 provided excellentdiscrimination (AUC=0.904). Interestingly, high COL6A3 serum levels were significantly associated with perineural invasion andcigarette smoking. Combined E3, E4, and E6 serum RNAvalues provided good sensitivity but low specificity. Our data demonstratefor the first time the potential clinical significance of circulating COL6A3 in the diagnosis of pancreatic malignancy.

Keywords Pancreatic cancer . CollagenVI . Serummarker .

Diagnostic biomarkerIntroduction

Pancreatic ductal adenocarcinoma (PDA) is often a lethaldisease that is currently the fourth leading cause of cancerdeaths in the USA.1 In most cases at the time of diagnosis, thetumor has already spread locally or metastasized into nearbyorgans >85 %; therefore, the overall 5-year survival rate isonly

type I collagen. COL6 is composed of three alpha chains: 1,2, and 3. The 3 chain (3 chain of type VI collagen(COL6A3) protein) is made of a short triple helical regionflanked by N and C globular domains.6 We have shownrecently that the COL6A3 protein is significantly upregulatedin PDA when compared to the adjacent nonmalignant tissueand is localized to the malignant cells and fibroblasts of thedesmoplastic stroma.7 We have also shown that the COL6A3gene undergoes a tumor-specific alternative splicing processthat is associated with increasing its transcription and transla-tion in PDA. This process creates the tumor-specific tran-scripts containing alternatively spliced exons E3, E4, and E6.7

The aims of our study were to investigate the expression ofCOL6A3 protein in sera from patients with PDA, benignlesions and from age-matched healthy volunteers, and toexplore its discriminative ability between the different lesions,alone and in combination with serum CA19-9. Serum levelswere correlated with the different clinicopathological param-eters. We also examined the presence of COL6A3 mRNAcontaining E3, E4, and E6 exons in the same sera and evalu-ated their diagnostic potential.

Materials and Methods

Serum Acquisition Serum samples were retrospectivelyobtained from patients that underwent pancreatic resection atthe Thomas Jefferson University Hospital from 2006 to 2011.Sera were from patients with pathologically confirmed PDA(n =44) and benign lesions (n =46; intraductal papillary mu-cinous neoplasm (IPMN), n =23; cystic lesions, n =23). Se-rum was also obtained from age-matched normal healthyvolunteers (n =30). CA 19-9 levels and the clinical data forbenign and PDA patients were obtained from Thomas Jeffer-son University Hospital electronic medical records and frompatients' charts. All patients signed an appropriate consentform. The study was approved by the Institutional ReviewBoard at Thomas Jefferson University.

Enzyme-Linked Immunosorbent Assay Serum concentrationsof COL6A3 protein were measured in triplicate using com-mercially available human COL6A3-specific enzyme-linkedimmunosorbent assays (ELISA; USCN Life Science Inc, Bur-lington, NC, USA) according to the manufacturer's protocol.Spectrophotometric evaluation of COL6A3 levels was ana-lyzed by a Synergy HT Multidetection Microplate Reader(BioTeck, Winooski, VT, USA). Clinical data, tumor charac-teristics, and AJCC staging of PDA and benign patients andtheir COL6A3 levels are shown in Tables 1 and 2.

RNA Extraction and Real-Time Reverse TranscriptionPCR RNA from serum samples was prepared usingPureYieldTM RNA Midiprep System (Promega Corporation,

Madison, WI, USA). The RNAwas then made into comple-mentary DNA (cDNA) using the Versa cDNA synthesis kit(Thermo Scientific) and following the manufacturer's proto-col. Using specific primers for exons E3, E4, E6, and GAPDHas an internal control, the cDNA was subjected to RT-PCRusing TaqMan technology (7500 Sequence Detector; AppliedBiosystems). The relative mRNA levels were presented as

Table 1 Clinicopathological characteristics of PDA patients and theirCOL6A3 levels

PDA (n =44) % COL6A3 (pg/mL)MeanSEM

Average age (years)

unit values of 2^[CT(GAPDH )CT(COL6A3-E3, E4, E6)],where CT is the threshold cycle value defined as the fractionalcycle number at which the target fluorescent signal passes afixed threshold above baseline.

Statistical Analysis COL6A3 protein expression was analyzedfirst as raw levels then in a logarithmic scale in order to achievea symmetric distribution. The difference in COL6A3 levelsbetween PDA, benign, and normal sera was calculated usingpairwise t tests adjusting the p value using Hochberg's method.A box plot of log-transformed COL6A3 by group was alsoproduced. COL6A3 levels in PDA by clinicopathologic factorssuch as invasion, diabetes, obesity, and smoking were exam-ined. No adjustment for multiple comparisons was made to thep values. The predictive ability of COL6A3 alone for PDA andnormal patients was examined using receiver operating charac-teristic (ROC) curves from a logistic regression model. TheSpearman correlation was used to assess the relationship be-tween ln(COL6A3) and CA 19-9, or ln(CA 19-9) within benignand tumor samples. The predictive value of COL6A3 alone andin combination with CA19-9 was examined using ROC curvesfrom logistic regressionmodels for PDAversus benign patients.Probabilities 0.05 were considered significant. Ln(COL6A3)and CA 19-9 were categorized into high and low groups bytheir respective medians and survival in the tumor patients wasanalyzed using the KaplanMeier method. Additionally, Coxproportional hazards regression was carried out using continu-ous ln(COL6A3) and CA 19-9 as predictors alone and together.CA-19-9 was log-transformed for use in the Cox proportionalhazard model in order to meet the proportionality assumption.Analysis was completed using R version 2.9.2 and SAS version9.2 (SAS; Campus Drive, Cary, NC, USA).

Results

Patient Characteristics and COL6A3 Serum Levels COL6A3levels were determined and averaged in sera from 44 PDApatients by ELISA (26,6703,063 pg/ml) and 46 patients be-nign lesions (16,7781,045 pg/ml). Tables 1 and 2 show thedistribution of COL6A3 protein levels among age, sex, andBMI for patients with PDA and benign lesions, respectively.

COL6A3 in PDACompared to Benign and Normal CirculatingCOL6A3 levels were significantly elevated in PDA patientswhen compared to healthy normal sera (p =0.0001) and be-nign lesions (p =0.0035). When log values, ln(COL6A3),were used to analyze the data, there was a significant differ-ence in ln(Col6A3) between PDA, benign, and normal pa-tients (all adjusted pairwise differences p

significance in patients with perineural invasion compared topatients without perineural invasion (median, 21,015 versus13,121; p =0.0337) and in patients who are cigarette smokerscompared to patients who are nonsmokers (median, 53,386versus 19,390; p =0.0012).

Correlation Between COL6A3 and CA19-9 CA19-9 is themost widely used serum marker and is routinely measured inall pancreatic cancer patients. We obtained the CA19-9 valuesfor all benign and PDA patients from our hospital's electronicmedical records. The scatter plots (Fig. 2a,b) display the corre-lation between the two biomarkers using Spearman's rank test.The Spearman correlation coefficient for ln(COL6A3) andCA19-9 was 0.11617 with a p value of 0.31 (Fig. 2a). CA-19-9 was log-transformed in order to meet the proportionalityassumption, ln(CA-19-9). The correlation coefficient betweenln(COL6A3) and ln(CA19-9) was 0.116 with a p value of0.3176 (Fig. 2b). The Spearman rank test was also used to testfor correlation between Col6A3 and CA 19-9 and determinedthat while there was a positive relationship between the twobiomarkers, it was not significant (rho=0.1204; p =0.3138).

Survival and COL6A3 Expression Ln(ColA3) and CA 19-9were categorized into high and low groups by their re-spective medians and survival in PDA patients was analyzedusing the KaplanMeier method (Fig. 3a,b). Elevatedln(COL6A3) levels were not associated with overall survival

(log-rank; p =0.2227). However, there was a stronger associ-ation with overall survival than with CA 19-9 (log-rank; p =0.8611). Additionally, Cox proportional hazards regressionwas carried out using continuous ln(Col6A3) and CA 19-9as predictors alone and together. CA 19-9 was analyzed in alogarithmic scale, ln(CA 19-9) in order to achieve a symmetricdistribution. A Cox regression hazards ratio analysis usingln(COL6A3) and ln(CA19-9) values also did not yield signif-icant association with survival as seen in Table 4.

ROC Analysis for lnCol6A3 and CA19-9 In a multivariateregression model considering serum ln(COL6A3) and CA19-9 (Table 5), both ln(COL6A3) and CA19-9 revealed a signifi-cant influence on the odds of presenting with PDA versusbenign lesion (multivariate logistic regression, p =0.015 andp =0.21, respectively). In a model considering serumln(COL6A3) and ln(CA19-9) (Table 5), both ln(COL6A3)and ln(CA19-9) revealed a significant influence on the oddsof presenting with PDA versus benign lesion (multivariatelogistic regression, p =0.008 and p

Figure 4a,b present the ROC plots for the two bio-marker models, with an AUC of 0.817 for COL6A3alone, 0.817 for CA19-9 alone, and an overall AUCof 0.904. These data indicate that combining COL6A3with CA 19-9 gives excellent discrimination betweenPDA and benign lesions.

COL6A3 E3, E4, and E6 mRNA Levels in PDA CirculatingRNA in the serum is a relatively newly developed area formolecular diagnosis. We have shown previously that CO-L6A3 undergoes tumor-specific alternative splicing that re-sults in the inclusion of 3 exons, E3, E4, and E6, in anycombination.7 We examined the expression of serum RNAcontaining the three exons in the serum and evaluated theirsignificance in PDA diagnosis. RNAs were extracted fromserum samples that were used for ELISA assay and analyzedby real-time PCR using E3, E4, and E6 specific primers andGAPDH as an internal control. Analysis of serum samplesfrom PDA, benign lesions, and healthy volunteers showedthat E6 exon was consistently present and easily detected inthe serum. Exons E3 and E4 were expressed at much lowerlevels than E6. The three exons were present at undetectablelevels in the mRNA from normal sera. Serum values of E3,E4, and E6 were analyzed and sensitivity and specificityvalues were calculated for PDA and benign lesions. Com-bined evaluation of the three isoforms shows that their valuesshow high sensitivity (0.91), but very low specificity (0.46)for PDA (Table 6). These data indicate that although serumE3, E4, and E6 mRNA have value in differentiating betweennormal and PDA sera, their low specificity makes them oflesser utility than serum COL6A3 protein levels fordistinguishing between benign and malignant lesions.

Discussion

A hallmark of PDA is the strong desmoplastic reaction thatincludes an ECMheavywith collagens, proliferating fibroblasts,and pancreatic stellate cells.8 The tumor microenvironment is

Fig. 3 Survival in tumor patients was analyzed using log-transformedCOL6A3 and CA 19-9 and KaplanMeier curves. Each biomarker wascategorized into low and high groups by their medians. No signifi-cant associations were seen between a ln(COL6A3) and CA 19-9 (p =0.22) and b ln(COL6A3) and ln(CA19-9) (p =0.86)

Table 4 Cox proportional hazards regression model for prediction ofoverall survival showing hazards ratio for ln(COL6A3) and ln(CA19-9)

HR (95 % CI) p value aHR (95 % CI) p value

ln(Col6A3) 1.46 (0.84, 2.54) 0.1773 1.46 (0.80, 2.68) 0.2232

ln(CA19-9) 1.02 (0.86, 1.20) 0.8438 1.05 (0.88, 1.26) 0.5816

CI confidence interval, HR hazard ratio, aHR adjusted hazard ratio

Table 5 Odds ratios, confidence intervals, and p value for the logisticregression model containing both ln(Col6A3) and CA19-9

OR (95 % CI) p value

Ln(Col6A3) 10.94 (1.57, 76.17) 0.0157a

CA19-9 1.014 (1.002, 1.025) 0.0216a

Ln(Col6A3) 12.92 (1.92, 86.80) 0.0085a

Ln(CA19-9) 2.42 (1.59, 3.69)

responsible for a complex interplay of interaction and signalingthat results in tumor growth and metastases. Our laboratory haspreviously described an increased deposition of the COL6A3 inPDA.7 This was associated with tumor-specific alternative splic-ing of COL6A3 gene that result in the production of mRNAcontaining three exons: E3, E4, and E6. To our knowledge, this

is the first study to (a) evaluate levels of the serum COL6A3protein in PDA, benign, and healthy volunteer sera; (b) demon-strate a significant correlation between the high levels of CO-L6A3 protein and perineural invasion and smoking; (c) revealthe potential validity of COL6A3 protein as a noninvasivedifferentiation marker between healthy and malignant sera;and (d) combine COL6A3 protein results with CA19-9 andshow that it may be used as a powerful screening tool forPDA diagnosis in high-risk patients.

COL6 has been found to be responsible for cell anchoring,forming a filamentous network with collagen types I and III andECM remodeling, which creates an environment for the tumorto flourish.911 It is proposed that high serum collagen is theproduct of high levels of degradation, angiogenesis, and remod-eling in the ECM.12,13 Our results show significantly elevatedlevels of COL6A3 when compared to healthy sera (p =0.0001)and to benign lesions (p =0.0035). Serum levels of COL6A3can be the result of the PDA stroma being remodeled constantlyas the tumor progresses.12,13 With higher circulating serum

Fig. 4 Individual receiveroperator characteristics (ROC)curves comparing patients withPDA (n =44) and benign lesions(n =46) with respect to their aln(COL6A3) and CA19-9 with anAUC of 0.904 and bln(COL6A3) and ln(CA19-9)with an AUC of 0.895

Table 6 Serum values of E3, E4, and E6 mRNAwere evaluated in PDA,benign lesions, and healthy volunteer by real time PCR. Data wereanalyzed and sensitivity and specificity values were calculated and areseen in this table. Combined evaluation of the three isoforms shows highsensitivity (0.91), but low specificity (0.46)

Estimated value 95 % Confidence interval

Lower limit Upper limit

Prevalence 0.69 0.58 0.78

Sensitivity 0.91 0.80 0.96

Specificity 0.46 0.27 0.66

12 J Gastrointest Surg (2014) 18:715

COL6A3 levels, we would expect to see higher-grade tumors.Our results did not show a significant difference of expressionlevels between tumor stages; however, we do note that due tothe insidious nature of PDA, most if not all patients in our studywere either stage IIA or IIB.

Our results shown that perineural invasion has a significantlyhigh association with high expression levels of COL6A3(Table 3). Previous studies have shown the involvement ofCOL6 in the remodeling of the tumor microenvironment inovarian cancer leads to increase in their resistance tochemotherapy.11 Other studies show that collagen VI contrib-utes to drug resistance through ECM remodeling.14 Further-more, type I collagen was shown to play a role in pancreaticcancer cells, where it promotes the expression of membranetype 1 matrix metalloproteinase-dependent (MT1-MMP)invasion.15 It is clear that the high levels of collagens in thetumor microenvironment contribute to tumor growth, metasta-sis, and resistance to chemotherapy.10,16,17 Collagens play anintegral role with tumorstroma interaction that includeschemokines, cytokines, nerve growth factors, MMPs, and othergrowth factors.18 Nonetheless, it is yet to be determined whatrole exactly COL6A3 plays in PDA cell biological behaviors.Studies in this regard are currently ongoing in our laboratory.

We also demonstrate that there is a significant increase inCOL6A3 expression with patients that smoke tobacco.Smoking tobacco carries a greater than twofold risk of devel-oping PDA.19,20 There is also an increase in cancer risk withan increased duration of smoking.19 Nicotine is a major com-ponent of cigarette smoke and has been shown in manyexperimental models to induce proliferation and angiogenesis;however, its exact mechanism in pancreatic cancer isunclear.2123 We have previously shown a relationship be-tween nicotine and two factors: MMP-9 and VEGF.24 Numer-ous studies have correlated high levels of MMP-9 with tumorinvasion and progression in many cancers.13,2427 MMP-9also promotes cell survival and VEGF promotes survivaland angiogenesis. Nicotine has been studied to promote col-lagen breakdown in gingival tissues through the activation ofMMPs.2830 With the degradation of the ECM by decreasingthe collagen content, PDA can invade through the basementmembrane into adjacent tissues.

The absence of significant correlations between survivaland the circulating COL6A3 levels should be interpreted withcaution. The sample size was relatively small and our follow-up period was relatively short for association studies and mayhave mitigated against demonstrating statistical associations.

CA 19-9 is currently the only Food and DrugAdministration-approved biomarker available for pancreaticcancer. Although it is measured routinely in every pancreaticcancer patient undergoing surgery, it has been only beneficialas a postoperative prognostic marker. There have been numer-ous studies of CA 19-9, but a definitive cutoff value for PDAhas not been determined.31,32 From our ROC analysis, the

predictive value of CA 19-9 showed an accuracy of 81 %when distinguishing between malignant and benign lesions.Many studies show the sensitivity of CA19-9 at 80%, which issimilar to the levels we have seen in our samples.32 With theaddition of COL6A3 expression values with CA 19-9, thepredictive value accuracy increased to 90.4 %. Thus, we pro-pose that a panel composed of CA19-9 and COL6A3 wouldprovide additional excellent sensitivity for PDA diagnosis.

We and others have shown previously that COL6A3 un-dergoes tumor-specific alternative splicing in PDA and othercancers.7,33,34 As a complimentary study, we sought to deter-mine whether serum mRNA expression levels for COL6A3E3, E4, and E6 exons could be detected and used as potentialdiagnostic markers for PDA. Although the combined valuesfor the three isoforms were found to have high sensitivity, theyhave low specificity. At the individual levels, serum E6mRNAwas consistently found in PDA sera, while E3 and E4 mRNAswere found at much lower levels and not in all samples. Theideal diagnostic marker has to be consistent and reproducibleand that is why we believe that serum RNA is not a suitablemeans to analyze the presence of COL6A3 tumor-specifictranscripts. RNA is very labile and can very easily degradeover a short period of time.35 Also, it can degrade if not storedin the right conditions. The samples used in our study weresubjected tomany other experiments other thanRNA isolation.Studies have also shown that a single freeze and thaw cycle canhave no effect to the mRNA isolates but with multiple cycles,the mRNA can degrade and produce different results.36,37

Since COL6A3 ELISA assay produced excellent results, webelieve that synthesizing isoform-specific antibodies to exam-ine isoform protein expression would yield a more sensitiveand reliable results. We are currently in the process of produc-ing isoform-specific antibodies in collaboration with one of thespecialized biotechnology companies.

Study Limitations

Because the findings of this study are limited to preoperativeserum samples from surgically resected patients with PDA orIPMN, the temporal relationship between COL6A3 levelchanges in the serum and the progression of PDA remainsunknown. Thus, we still need to determine whether COL6A3stands as a good prognostic marker for PDA patients' followup. Studies to answer this question are in preparation. Sincethis was a pilot study, we focused only on our patients'population. A larger study that includes a larger cohort ofPDA patients, chronic pancreatitis, and patients with othermalignancies is currently under preparation to determine thespecificity of this marker to PDA. The inconsistency of serumRNA results can be attributed to serum isolation-related er-rors. Our serum samples were obtained retrospectively andwere not originally collected for the purpose of RNA isolation.Sera dedicated only to RNA isolation should be considered.

J Gastrointest Surg (2014) 18:715 13

Conclusion

Wehave demonstrated for the first time the value of COL6A3 asa novel noninvasive serum diagnostic marker for PDA that candifferentially discriminate between having a pancreatic malig-nancy or nomalignancy. It can also add sensitivity when used incombination with the current marker CA 19-9 in the differenti-ation between benign and malignant pancreatic lesions. Thepotential of using COL6A3 as a diagnostic marker alone or incombination with CA19-9 and as a predictive of tumor aggres-siveness must be explored further in a larger population tovalidate and demonstrate its ultimate clinical significance.

Acknowledgments This work was supported by a grant from theUniversity City Science Center QED award and the PA CommonwealthKeystone Innovation Grant.

References

1. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2012. CACancer J Clin 2012: 62:1029.

2. Hidalgo M. Pancreatic cancer. N Engl J Med 2010: 362:16051617.3. Maitra A, Hruban RH. Pancreatic cancer. Annu Rev Pathol 2008: 3:

157188.4. Winter JM, Yeo CJ, Brody JR. Diagnostic, prognostic, and predictive

biomarkers in pancreatic cancer. J Surg Oncol 2013: 107:1522.5. Mahadevan D, Von Hoff DD. Tumorstroma interactions in

pancreatic ductal adenocarcinoma. Mol Cancer Ther 2007: 6:11861197.

6. Chu ML, Zhang RZ, Pan TC et al. Mosaic structure of globulardomains in the human type VI collagen alpha 3 chain: similarity tovon Willebrand factor, fibronectin, actin, salivary proteins andaprotinin type protease inhibitors. EMBO J 1990:9(2): 385393.

7. Arafat H, Lazar M, Salem K et al. Tumor-specific expression andalternative splicing of the COL6A3 gene in pancreatic cancer. Sur-gery 2011: 150:306315.

8. Chang KC, Lu YC, LinMJ, Chen HY, Jin YT. Desmoplastic tumour-associated stroma versus neural tissue in central nervous systemmetastasis: effects of different microenvironments on tumour growth.Histopathology 2011: 59:3139.

9. Lohi J, Leivo I, Oivula J, Lehto V, Virtanen I. Extracellular matrix inrenal cell carcinomas. Histol Histopathol 1998: 13:785796.

10. Sethi T, Rintoul RC, Moore SM et al. Extracellular matrix proteinsprotect small cell lung cancer cells against apoptosis: a mechanismfor small cell lung cancer growth and drug resistance in vivo. NatMed 1999: 5:662668.

11. Sherman-Baust CA, Weeraratna AT, Rangel L et al. Remodeling ofthe extracellular matrix through overexpression of collagen VI con-tributes to cisplatin resistance in ovarian cancer cells. Cancer cell2003: 3:377386.

12. Ohlund D, Lundin C, Ardnor B, Oman M, Naredi P, Sund M. TypeIV collagen is a tumour stroma-derived biomarker for pancreascancer. Br J Cancer 2009: 101:9197.

13. Mazouni C, Arun B, Andre F et al. Collagen IV levels are elevated inthe serum of patients with primary breast cancer compared to healthyvolunteers. Br J Cancer 2008: 99:6871.

14. Ohlund D, Franklin O, Lundberg E, Lundin C, Sund M. Type IVcollagen stimulates pancreatic cancer cell proliferation, migration,and inhibits apoptosis through an autocrine loop. BMC Cancer(2013). DOI: 13:154-2407-13-154.

15. ShieldsMA, Dangi-Garimella S, Krantz SB, BentremDJ,Munshi HG.Pancreatic cancer cells respond to type I collagen by inducing snailexpression to promote membrane type 1 matrix metalloproteinase-dependent collagen invasion. J Biol Chem 2011: 286:1049510504.(duplicate of ref. 11)

16. Sherman-Baust CA, Weeraratna AT, Rangel L et al. Remodeling ofthe extracellular matrix through overexpression of collagen VI con-tributes to cisplatin resistance in ovarian cancer cells. Cancer cell2003: 3:377386.

17. Miyamoto H, Murakami T, Tsuchida K, Sugino H, Miyake H, TashiroS. Tumor-stroma interaction of human pancreatic cancer: acquiredresistance to anticancer drugs and proliferation regulation is dependenton extracellular matrix proteins. Pancreas 2004: 28:3844.

18. Korc M. Pancreatic cancer-associated stroma production. Am J Surg2007: 194:S84-S86.

19. Lynch SM, Vrieling A, Lubin JH et al. Cigarette smoking andpancreatic cancer: a pooled analysis from the pancreatic cancercohort consortium. Am J Epidemiol 2009: 170:403413.

20. Lowenfels AB, Maisonneuve P. Epidemiology and risk factors forpancreatic cancer. Best Practice and Research: Clinical Gastroenter-ology 2006: 20:197209.

21. Heeschen C, Jang JJ, Weis M et al. Nicotine stimulates angiogenesisand promotes tumor growth and atherosclerosis. Nat Med 2001: 7:833839.

22. Bose C, ZhangH,Udupa KB, Chowdhury P. Activation of p-ERK1/2by nicotine in pancreatic tumor cell line AR42J: effects on prolifer-ation and secretion. Am J Physiol Gastrointest Liver Physiol 2005:289:G926-34.

23. Chowdhury P, Bose C, Udupa KB. Nicotine-induced proliferation ofisolated rat pancreatic acinar cells: effect on cell signalling andfunction. Cell Prolif 2007: 40:125141.

24. Lazar M, Sullivan J, Chipitsyna G et al. Involvement of osteopontin inthe matrix-degrading and proangiogenic changes mediated by nicotinein pancreatic cancer cells. J Gastrointest Surg 2010: 14:15661577.

25. Mahadevan D, Von Hoff DD. Tumorstroma interactions in pancre-atic ductal adenocarcinoma. Mol Cancer Ther 2007: 6:11861197.

26. Makareeva E, Han S, Vera JC et al. Carcinomas contain a matrixmetalloproteinase-resistant isoform of type I collagen exerting selec-tive support to invasion. Cancer Res 2010: 70:43664374.

27. Stallings-Mann M, Radisky D. Matrix metalloproteinase-inducedmalignancy in mammary epithelial cells. Cells Tissues Organs2007: 185:104110.

28. Zhou J, Olson BL, Windsor LJ. Nicotine increases the collagen-degrading ability of human gingival fibroblasts. J Periodontal Res2007: 42:228235.

29. Heeschen C, Jang JJ, Weis M et al. Nicotine stimulates angiogenesisand promotes tumor growth and atherosclerosis. Nat Med 2001: 7:833839.

30. Tipton DA, Dabbous MK. Effects of nicotine on proliferation andextracellular matrix production of human gingival fibroblasts in vitro.J Periodontol 1995: 66:10561064.

31. Singh S, Tang SJ, Sreenarasimhaiah J, Lara LF, Siddiqui A. Theclinical utility and limitations of serum carbohydrate antigen (CA19-9) as a diagnostic tool for pancreatic cancer and cholangiocarcinoma.Dig Dis Sci 2011: 56:24912496.

32. Winter JM, Yeo CJ, Brody JR. Diagnostic, prognostic, and predictivebiomarkers in pancreatic cancer. J Surg Oncol 2013: 107:1522.

33. Yang S, Shin J, Park KH et al. Molecular basis of the differencesbetween normal and tumor tissues of gastric cancer. Biochimica etbiophysica acta. Molecular basis of disease 2007: 1772:10331040

34. Thorsen K, Sorensen KD, Brems-Eskildsen AS et al. Alternativesplicing in colon, bladder, and prostate cancer identified by exonarray analysis. Mol Cell Proteomics 2008: 7:12141224.

35. Farkas DH, Kaul KL, Wiedbrauk DL, Kiechle FL. Specimen collec-tion and storage for diagnostic molecular pathology investigation.Arch Pathol Lab Med 1996: 120:591596.

14 J Gastrointest Surg (2014) 18:715

36. Kopreski MS, Benko FA, Kwak LW, Gocke CD. Detection of tumormessenger RNA in the serum of patients with malignant melanoma.Clin Cancer Res 1999: 5:19611965.

37. Hunt JL. Molecular pathology in anatomic pathology practice: areview of basic principles. Arch Pathol LabMed 2008: 132:248260.

Discussant

Dr. Syed Ahmad (Cincinnati, Ohio) : Christopher, I want tocongratulate you, Dr. Arafat, Dr. Yeo and your co-authors on avery nice presentation and novel work. I also want to thankyou for sending me the manuscript in advance of the meeting.Your work suggests that the alpha 3 chain of the type VIcollagen may be utilized as a novel biomarker for diagnosingpancreatic cancer. With this background, I would like toremind everyone on the potential characteristics of a usefulbiomarker, they should: (1) have specific production by pre-malignant or malignant tissue early in the progression ofdisease; (2) be produced at detectable levels in all patientswith a specific malignancy; (3) they should be expressed in anorgan site-specific manner; (4) their levels should be relatedquantitatively to tumor volume, biological behavior, or dis-ease progression; (5) they should have a relatively short half-life, reflecting temporal changes in tumor burden and responseto therapy; (7) there should be a standardized and reproducibleassay for measuring the biomarker; and finally, for screening,(8) they should have relatively high sensitivity and very highdisease specificity, i.e., to yield a very low false-positive rate(FPR) and a reasonable low false-negative rate (FNR).

Based on these characteristics, I have a few questions:1. You evaluated 46 patients with IPMN and benign cystic

lesions, do you have any data on the production of COL6A3 inpremalignant tissue. Does the level rise from premalignant tomalignant disease or is COL6A3 only elevated in pancreascancer? Does COL6A3 correlate with prognosis or recurrence?

2. Are levels elevated in other types of cancer or benigninflammatory states?

3. How do you explain the lack of increase in levels ofCOL6A3 from less aggressive to more aggressive clinical-pathologic factors? Studies would suggest that more aggres-sive cancers have more of a desmoplastic response and Iwould have expected increasing levels?

4. Please give us insight into the false-positive and false-negative rate for this biomarker?

5. Your data would suggest that, by itself, COL6A3 isequivalent to CA 19-9 in predicting the presence of canceraround 81 %. This increases to 90 % when the two are

combined. How is this biomarker different than others thathave been reported to increase the PPVof CA 19-9?

Once again, I thank you for the opportunity to review thispaper and I look forward to hearing your answers.

Closing Discussant

Dr. Christopher Kang : Thank you Dr. Ahmad for the insight-ful and thought-provoking comments and questions. In regardsto the levels of COL6A3 from normal to premalignant andmalignant disease, the levels of serum COL6A3 were elevatedwith statistical significance in premalignant lesions compared tonormal as seen with our IPMN and cystadenomas. From pre-malignant to malignant lesions, there was also a statisticalsignificant increase in COL6A3 levels. Although there was noassociation with statistical significance between survival andlevels of COL6A3, there was a trend with patients with lowlevels of COL6A3 having better survival.

There have not been any studies from our lab or in theliterature review that show serum COL6A3 levels in othercancers or inflammatory states. We would like to examine thelevels in other cancers and inflammatory states in a biggerstudy. Chronic pancreatitis would be especially interesting.

Actually, our data show significant association of high CO-L6A3 levels with perineural invasion and with cigarettesmoking. This implies a role for COL6A3 in the metastaticbehavior of PDA cells. We are currently exploring the functionsof the individual isoforms that we showed to be high in themalignant lesions. However, the lack of increase in COL6A3levels inmore advanced stages or larger masses can be explainedby the fact that COL6A3 itself has no effect on cell proliferation(unpublished observation). With actual functional studies usingPDA cell lines, studies that are currently ongoing will be able toelucidate the functional aspects of this intriguing stromal protein.

As we have seen with CA 19-9, differing the cutoff levelsof expression can change false-positive and false-negativerate. We have not determined a true cutoff value for COL6A3as of yet, but with more samples and including the otherpathologies, we may achieve a true cutoff value.

In response to your final question, from my review of theliterature regarding CA19-9 and other biomarkers improvingPPV, a common theme is great initial results; however, there isa failure of reproducibility. We have reproduced our resultsand believe when we synthesize antibodies to COL6A3isoforms, we will have a true diagnostic biomarker that fol-lows the characteristics you have mentioned.

J Gastrointest Surg (2014) 18:715 15

Clinical Significance of Serum COL6A3 in Pancreatic Ductal AdenocarcinomaAbstractIntroductionMaterials and MethodsResultsDiscussionStudy Limitations

ConclusionReferences