Comparison of gene expression in squamous cell carcinoma and adenocarcinoma of the uterine cervix

Gynecologic Oncology 95 (2004) 610 – 617 www.elsevier.com/locate/ygyno

Comparison of gene expression in squamous cell carcinoma and adenocarcinoma of the uterine cervix Stephen A. Contaga, Bobbie S. Gostouta,*, Amy C. Claytonb, Melanie H. Dixona, Renee M. McGoverna, Eric S. Calhounc a

Section of Gynecologic Surgery, Mayo Clinic, Rochester, MN 55905, USA b The Division of Anatomic Pathology, Mayo Clinic, Rochester, MN, USA c The Division of Experimental Pathology and Laboratory Medicine, Mayo Clinic, Rochester, MN, USA Received 10 May 2004 Available online 22 October 2004

Abstract Objectives. Microarray expression analysis of cervical tumors has revealed differential expression of genes that may be useful as markers or targets for treatment. We question the application of array findings across the major categories of cervical cancer. We sought to identify differences between normal squamous epithelium (NSQ) and glandular epithelium (NGL) of the uterine cervix and their malignant variants: squamous cell cancer (SCC) and adenocarcinoma (ACA). Methods. Eight genes were selected: 12-lipoxygenase (12-LOX), keratin 4, trypsinogen 2 (TRY2), Rh glycoprotein C (RhGC), collagen type V alpha 2, integrin alpha 5, integrin alpha 6, and c-myc. Ten cases each of SCC and ACA of the cervix were selected from our tumor bank. NSQ and NGL epithelia were obtained from consecutive patients undergoing surgery for benign disease. RNA extraction, cDNA synthesis, and DNA amplification of all samples were performed according to an established protocol. Electrophoresis of the multiplexed polymerase chain reaction (PCR) products was performed under standard conditions, followed by digital image capture. A ratio of target to control gene (h-actin) was obtained for each sample. Analysis of variance was applied to the mean ratios for each tissue to establish significant differences. Individual pairwise comparisons were made by Student t tests and verified with the TukeyKramer test. Results. Clinically valid comparisons are NSQ to NGL, NSQ to SCC, NGL to ACA, and SCC to ACA. Various expression patterns were observed between the epithelia and their malignant phenotypes. Significant differences in gene expression were observed between benign squamous and glandular epithelium in four of the eight genes and between malignant squamous and glandular epithelium in three of the eight genes. Significant differences in gene expression between benign and malignant tissues were demonstrated in four of the eight genes. Conclusions. We have defined significant differential expression changes between the two principal cervical tumor types. Differences in genes are demonstrated and must be considered if array technology is applied to the study of the biologic behavior of these tumors as well as their screening and management. The observed differential expression should be a compelling argument to perform type-specific expression analysis for other tumors with histological variants. D 2004 Elsevier Inc. All rights reserved. Keywords: Normal squamous epithelium; Squamous cell carcinoma; Adenocarcinoma

Introduction

* Corresponding author. Section of Gynecologic Surgery, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. E-mail address: [email protected] (B.S. Gostout). 0090-8258/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ygyno.2004.08.021

Using microarray expression analysis, researchers have been able to discern the expression changes characterizing tumors, including cervical carcinoma. The output from these arrays provides a tumor bfingerprintQ composed of RNA expression patterns for thousands of genes. For tumors with

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617

histological variants, the relevance of the fingerprint from one histologic type to another is as yet uncertain. This is important because molecular diagnostic or therapeutic measures that might be derived from the information gleaned from array technologies will be applicable to tumors only with expression changes that match those predicted by the reference array. Cancer of the uterine cervix is an example of a tumor with a clinically important histological variant. Squamous cell carcinoma (SCC) accounts for the majority of tumors. Adenocarcinomas (ACAs) comprise 10% to 18% of all cervical tumors and appear to be increasing in absolute number of cases as well as in relative percentage of invasive cervical tumors diagnosed [1,2]. ACAs pose significant clinical challenges compared with their squamous counterparts. They are less likely to be detected by Papanicolaou tests, more likely to be locally advanced when diagnosed, and therefore, less responsive to radiation therapy, and, stage for stage, more likely to result in death [2–5]. In conjunction with scientists at Millennium Pharmaceuticals, Inc., we participated in an extensive analysis of gene expression in cervical carcinomas. The findings from this analysis have been published [6]. A 30-K gene expression pattern was established from a pool of nine normal squamous (NSQ) epithelial specimens, three normal endocervical epithelial specimens, five low-grade squamous intraepithelial lesions (LSILs), five high-grade squamous intraepithelial lesions (HSILs), nine SCCs, and three ACAs. Sixty-two genes of interest were identified as being significantly overexpressed or underexpressed compared with normal tissue. We describe our work exploring the degree to which the described expression pattern is valid for cervical tumors originating in the glandular rather than squamous epithelium of the cervix. Prior work to validate the 30-K array results in our laboratory suggested that some overexpressed or underexpressed genes have a different pattern of expression in cervical ACAs than cervical SCCs. The majority of

611

documented differential expression observed between normal cervical epithelium and SCC involved genes coding for extracellular matrix (ECM) proteins as well as their cell membrane receptors [6]. ECM proteins, including vitronectin, fibronectin, collagen, and matrix metalloproteinase, have been studied extensively, and their essential role in cell-to-cell signaling, tumor growth, and invasiveness has been established [7]. The microarray and subsequent profiling have demonstrated down-regulation of arachidonate 12-lipoxygenase (12-LOX), cytokeratin 4 type II (KRT4), gap-junction h2-protein or connexin 26 (CX26), Homo sapiens connexin 30 (CX30), trypsinogen II precursor 9 (TRY2), and Rh type C glycoprotein (RhGC) in SCC compared with normal cervical squamous epithelium. Fibronectin was overexpressed in squamous tumor cells. This finding was further corroborated by increased fibronectin immunohistochemical staining in the ECM of the stroma surrounding squamous cancer cells compared with normal cervical epithelium. Subsequent validation of SCC antigen 1 (SCCA1) and GINGIVA, which were downregulated in the initial microarray experiment, was not observed in a later semiquantitative reverse transcription– polymerase chain reaction (RT–PCR) performed on tumor tissue. All of the above gene products, which interact with the ECM, are known to be important in autocrine and paracrine signaling between cells. Additional genes were selected for study on the basis of ongoing research in our laboratory. We chose to analyze 12-LOX, KRT-4, integrin alpha 5 (ITGA-5), ITGA-6, TRY2, RhGC, and COL5A because of visible differences in band intensity between SCC and normal squamous (NSQ) epithelium. The protooncogene c-myc was included because of our interest in understanding its possible role as a promoter site in viral integration (Table 1). The purpose of this study is to further characterize expression differences between NSQ and normal glandular (NGL) epithelium and the respective malignant derivatives: SCC and ACA of the cervix.

Table 1 Genes selected for comparison and their function Gene

Function

Cytokeratin 4 (IF)

An intermediate filament involved in such processes as cell division, motility, and plasticity [8]. Decreased expression may lead to cellular immaturity, fewer differentiated cells, diminished keratohyaline granules, and disturbances in the maturation of cell layers [9] Collagen type V is widely distributed in all connective tissues other than cartilage and bone. Collagen IV, a major component of the basal lamina, has also been shown to modulate angiogenesis [10,11] Arachidonic acid metabolism in endothelial cells, especially the 12-LOX pathway, has a critical role in angiogenesis [12–14] Integrins form cell surface receptors. Most integrins bind to extracellular matrix components, and mediated signals are necessary in normal cells to block apoptosis and to stimulate cell cycle progression [15] Epithelial cells predominantly express integrin a6/h4. Integrin a6 on viruslike particles has been shown to be a receptor for HPV-VLP 16 [16] The MYC proteins appear to be important in transcription and DNA replication and are important in regulating cellular proliferation and retarding differentiation [17] A defect in trypsinogen precursor II causes hereditary pancreatitis [18] It has been found to be present in cervical as well as other tissues and is down-regulated with development of malignancy in esophageal epithelium [19]

Collagen precursor a2 (V) Lipoxygenase 12 Integrin receptor a5 Integrin receptor a6 c-myc Trypsinogen precursor II Rh C glycoprotein

612

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617

Materials and methods

cDNA synthesis

Tissue samples

Five micrograms total RNA was used for cDNA synthesis for each sample. We added 4 Al 5 first-strand buffer (250mM Tris–HCl [pH 8.3] 375mM KCl, 15mM MgCl2; Invitrogen, Carlsbad, CA), 2 Al 0.1M dithiothreitol (Invitrogen), 1 Al 10mM dNTP mix (Roche, Mannheim, Germany), 1 Al 0.5 Ag/Al oligo dT (Integrated DNA Technologies, Coralville, IA), 1 Al RNaseOUT at 40 U/Al (Invitrogen), and 1 Al Superscript II at 200 U/Al (Invitrogen) for a total volume of 21 Al. A reverse transcriptase negative control was included with each cDNA sample. Synthesis reactions were incubated at 428C for 60 min, then 998C for 5 min, then cooled to 48C, all in a Perkin Elmer 9600 Gene Amp PCR System (Perkin Elmer Co.). cDNA was stored at 208C.

Institutional Review Board approval was received for this study. Tumor samples were available from banked specimens previously frozen at 808C. Fourteen SCC tumors and 12 ACA tumors were identified. Tumor grade and stage were noted when identifying the samples. The first 10 samples from each group were selected for expression analysis. Samples of normal cervical epithelium were obtained from consecutive patients undergoing hysterectomy for benign disease. Samples (50 to 150 mg) were excised separately from the ectocervix and the endocervix. All normal tissue was removed to a depth of 1 to 2 mm, carefully avoiding the underlying stromal tissue. The normal samples were immediately placed in vials containing 2 ml RNAlater (Ambion, Austin, TX), stored at 48C for up to 12 h, and then frozen at 808C. RNA extraction The RNA extraction process involved placing 50 to 75 mg of each tissue in a 12  75-mm tube containing 1 ml RNA-Bee (Tel-test, Inc., Friendswood, TX) and kept on ice. Tissues were homogenized by using a Pro 200 postmounted tissue homogenizer with a 7  120-mm generator at 30,000 rpm (Pro Scientific Inc., Oxford, CT). RNA extraction proceeded according to the RNA-Bee protocol. Total RNA was quantified by a Perkin Elmer Lambda 3B UV/VIS Spectrophotometer (Perkin Elmer Co., Norwalk, CT). The mean 260/280 ratio was 1.83 (1.78–1.93) for all the samples extracted. Total RNA was kept at 808C until cDNA synthesis was performed.

Primer design The genomic DNA and cDNA sequences for the selected study genes were downloaded from www.ensembl.org [20] as well as from the National Center for Biotechnology Information Web site at www.ncbi.nlm.nih.gov [21]. Oligo primer analysis software (Molecular Biology Insights, Inc., Cascade, CO) was used to design primer pairs for KRT-4, COL-5, 12-LOX, ITGA-6, RhGC, TRY2, ITGA-5, and cmyc (Table 2). b-Actin was used as a reference gene in all multiplex PCRs. The primers for b-actin were selected from published primer sequences [22]. Multiplex PCR Optimal conditions for gene amplification and multiplexing were developed [23]. b-Actin PCR with all

Table 2 Primer sets used for gene amplification Gene

Ensembl or GenBank gene accession number

Product size, bp

Primer sequence

KRT-4

ENSG00000110815, X07695, X61028, X67683

643

COL5a2

ENSG00000179877, Y14690, BC015705, M58529

462

12-LOX

ENSG00000108839, D12638, M35418, M58704, M62982, M87004, S68587 ENSG00000161638, BC008786, M13918, X06256

445

1214D 5V-TTTgTggTCCTAAAgAAggACgTg-3VY 1857U 5V-TggCATTCTCCAgACATTCTgTAC-3Vp 2966D 5V-TgCTgTTggAgAACgTggTg-3VY 3427U 5V-CCAAgAgCAgCTgTAAggTg-3Vp 1181D 5V-ACCCCATCTTCAAgTTCCTg-3VY 1625U 5V-ATggTgAggAAATggCAgAg-3Vp 2832D 5V-CATTTCCgAgTCTgggCCAA-3VY 3155U 5V-TggAggCTTgAgCTgAgCTT-3Vp 5V-CTCCCTgAACCTAACggAgTC-3VY 5V-AAACACAgTCACTCgAACCTg-3Vp 227-21D 5V-CACTgCTACAAgTCCCgCATC-3VY 649-21D 5V-AgggCCACCAgAATCACCCTg-3Vp 996-25D 5V-gggTTTTgTATACCTgACCCCATTC-3VY 865-21D 5V-AAgCTggACATggTgCACATC-3VY 859-21D 5V-AAgggCAAgCTggACATggTg-3VY 1368-21U 5V-gCTgTTCCCTTCAggCATCTC-3VY 524-20D 5V-gggTAgTggAAAACCAgCAg-3VY 971-21U 5V-gAgAAgCCgCTCCACATACAg-3Vp 1D 5V-AgCCTCgCCTTTgCCgA-3VY 2U 5V-CgCCCCAggCACCAg-3Vp

ITGA-5 ITGA-6 TRY2

ENSG00000091409, S52135, S66196, S66213, X53586, X59512 ENSG00000173636, M27602

295 342 422

RhGC

ENSG00000140519, AF193809, AF21998, AF219982, AF219983, AF219984, AF219985, AF219986, AF284446, BC030965

372

c-myc

ENSG00000136997, M13929, V00568, X00196, X00198 ENSG00000075624, D28354, X00351

447

b-ACTIN

175

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617

613

Table 3 Human papillomavirus types in normal epithelium and in tumor samples Sample NSQ NGL SCC ACA

Human papillomavirus type 1

2

3

4

5

6

7

8

9

10

Neg Neg 16 45

16 16 35 16

Neg Neg 16 Neg

Neg Neg 16 16

Neg Neg 16 16

Neg Neg 16 16

Neg Neg 16 Neg

Neg Neg 39 16

Neg Neg 16 18

Neg Neg 16 18, 33

ACA, adenocarcinoma; NGL, normal glandular epithelium; NSQ, normal squamous epithelium; SCC, squamous cell carcinoma.

samples confirmed adequate cDNA quality and comparable amplification. All PCR gene products were sequenced and identified. Multiplex PCR was performed in a 25-Al reaction containing 1 Perkin Elmer buffer I, 200AM 10mM dNTP mix (Roche), 2 U AmpliTaq Gold polymerase (Perkin Elmer Applied Biosystems), and 1 Al of each target and reference primer. All target primers were at 15 pmol/Al. The h-actin primer concentration used with targets KRT-4, COL-5, 12-LOX, and ITGA-5 was 3.2 pmol/Al, and with targets ITGA-6, RhGC, TRY2, and cmyc, it was 5 pmol/Al. One microliter cDNA was added to the aliquot mix. RT negative controls, genomic DNA controls, and negative DNA controls were included with each PCR set. PCR reactions were cycled in a Perkin Elmer 9600 Gene Amp PCR System (Perkin Elmer Co.) at 958C for 10 min, followed by 35 cycles at 958C for 30 s, 608C for 30 s, and 728C for 90 s.

at 958C, followed by 40 cycles of 1 min at 958C, 1 min at 558C, 1 min at 728C, and ending with a final 728C elongation step of 5 min. All samples negative for HPV with the MY primers were amplified in a nested PCR [25] with the GP5+ and GP6+ [26] primer set to detect any low-level HPV infection. The nested PCR reaction consisted of 1 Perkin Elmer buffer II, 2.5mM MgCl, 200AM dNTP, 2.5 U AmpliTaq Gold polymerase, and 0.5AM of primers GP5+ and GP6+. The first-round PCR product (1 Al) was added before cycling with the conditions listed above. Samples positive for HPV were sequenced by using the ABI PRISM 377 system (Perkin Elmer Applied Biosystems). Sequences were identified by using the FASTA program (Accelrys Inc., San Diego, CA).

Semiquantitative expression measurement

Statistical analysis was performed with JMP 5.0.1.2 software (SAS Institute Inc., Cary, NC) on the 10 samples each representing NSQ epithelium, NGL epithelium, SCC, and ACA of the uterine cervix. The Levene test was used to check for equality of group variances and used the absolute value of the difference of each observation from the group mean. Analysis of variance was applied to the ratios of each tissue group to control for a normal distribution of the results. Individual pairwise comparisons were made by using Student t tests and verified with the Tukey-Kramer test. Alpha was set at 0.05. The samples for NSQ epithelium and NGL epithelium were paired except for one set of samples. A multivariate correlation was used to observe the effect among the tissue samples for a given

After electrophoresis, digital images of the PCR products were captured by using the Kodak Electrophoresis Documentation and Analysis System 290 (Eastman Kodak Co., Rochester, NY). Each gel image was captured once immediately after completing the electrophoresis, and the image intensity was measured three times. We used the mean intensity of three measurements. A ratio of the net intensities of the target to control was obtained for each sample, and the mean value for each tissue group was used for comparison. HPV typing The human papillomavirus (HPV) status for the tumors was recorded in the tumor bank registry (Table 3). In each case, the previous designation of HPV type was accepted after reviewing the prior PCR results. For the normal tissue samples, HPV testing was accomplished by using the genomic DNA. HPV amplification was performed with a single or nested PCR with L1 consensus primers [24]. A 50-Al firstround PCR reaction for the amplification of HPV contained 1 Perkin Elmer buffer II, 2.5mM MgCl, 200AM dNTP, 0.5% bovine serum albumin, 2.5 U AmpliTaq Gold polymerase, and 0.5AM of primers MY09 and MY11 [24]. Cycling conditions were 10 min

Statistics

Table 4 Trends in gene expression between malignant and benign cervical epithelium Gene

SCC versus NSQ

ACA versus NGL

Prior RT–PCR

Prior microarray

ITGA-6 ITGA-5 COL-Va KRT-4 RhGC 12-LOX TRY-II c-myc

A z A A A A A z

A z A A A z A A

z

z

z A A A A

z

614

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617

Table 5 Summary of statistical comparisons across all tissue types within each gene group Statistics

Gene group result KRT-4

COL-5

12-LOX

ITGA-5

ITGA-6

TRY2

c-myc

RhGC

R2 ANOVA Prob N F

0.347 0.0014

0.393 0.0004

0.433 0.0001

0.242 0.0178

0.423 0.0002

0.194 0.0487

0.596 b0.0001

0.655 b0.0001

ANOVA, analysis of variance.

did not find any association between the gene expression in the paired NSQ and NGL epithelium samples. All of the correlations between tissue pairs were less than 1. The highest correlation was 0.74, but that was not statistically significant (Table 6). Average expression of the individual genes was compared across the tissue groups. The clinically valid comparisons were NSQ to NGL, NSQ to SCC, NGL to ACA, and SCC to ACA tissue. Diverse expression patterns were observed between the normal tissues and the cancerous tissues. Statistically significant expression differences were observed between NSQ and NGL tissue for four of the eight genes and between malignant squamous and malignant glandular tissue for three of the eight genes (Table 7). A difference in expression of selected genes in benign versus corresponding malignant tissue was demonstrated four times: twice involving squamous tumor compared to benign squamous tissue and twice involving glandular tumor compared to benign glandular tissue (Fig. 1). Trends toward overexpression and underexpression were distinctly different between squamous and glandular cancers when malignant tissues were compared with the corresponding benign cervical epithelium.

gene. These were all tested for significance by a matched pair’s response.

Results The mean patient age for the 10 subjects in each of the four tissue groups was as follows: 49.7 (SD, 16.1) years for the patients with NSQ and NGL epithelium, 45.6 (SD, 13.3) years for the patients with SCC, and 40.4 (SD, 11.5) years for the patients with ACA. The SCC subset included 10 pure squamous tumors, including eight grade 3 tumors and two grade 4 tumors [27]. Staging for the SCCs included four tumors at stage IB1, four at stage IB2, one at stage 2A, and one at stage 2B. The ACA set included 10 pure ACAs, including 6 grade 2, 1 grade 3, and 1 grade 4 tumors. Staging for the ACAs included three at stage IA and seven with stage IB1. The HPV types observed for the NSQ, NGL, SCC, and ACA tissues are listed in Table 3. One case of high-risk HPV 16 viral infection was observed in the normal squamous tissue group and in its glandular counterpart. No histological evidence of dysplasia was noted in any of the normal tissue samples. The relative expression of genes in these two samples did not vary significantly from the gene expression in epithelium without evidence of HPV infection. For each of the target genes, cDNA was successfully amplified in each of the 40 samples. The range of expression varied widely among the target genes and the target tissues. Overexpression of the gene in SCCs compared with normal squamous tissues confirmed the previous trend toward overexpression or underexpression of the genes predicted by the microarray analysis or the previous RT–PCR on SCC cell lines in four of six genes. For the ACAs, the trend toward overexpression or underexpression corresponded to the microarray prediction in three of six cases (Table 4). Analysis of variance between the tissue groups indicated significant expression ( P b 0.05) differences for every target gene analyzed (Table 5). The multivariate correlation

Discussion Our results confirm that expression differences are associated with malignant transformation of the cervical epithelium. We have further defined expression patterns unique to cervical squamous and glandular epithelium as well as cervical SCC and ACA. These differences reflect those seen in the biologic behavior of these tumors and may have implications for their screening and management. Our hypothesis was that biologically important differences exist in the gene expression patterns between NSQ and NGL epithelium and between SCC and ACA of the uterine cervix. Diverse expression patterns were observed between the glandular and squamous tissues. No significant

Table 6 Multivariate correlations and significance of matched pairs of normal epithelium Gene results KRT-4 Correlation P

0.03 0.0004

12-LOX 0.03 0.0007

COL-5 0.74 0.8206

ITGA-5 0.52 0.0674

c-myc 0.06 0.0765

TRY2 0.25 0.4967

ITGA-6 0.22 0.0074

RhGC 0.32 0.0017

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617 Table 7 Trends in gene expression between squamous and glandular cervical epithelium Gene

SCC versus ACA

NSQ versus NGL

SCC versus NSQ

ACA versus NGL

ITGA-6 ITGA-5 COL-Va KRT-4 RhGC 12-LOX TRY-II c-myc

A A A z* z* z A z*

A* A

A z A* A A* A A z

A z A* A A z A A*

z* z* z* A A

* P b 0.01.

correlation was observed by multivariate analysis in the expression patterns between NSQ and NGL epithelium. These assertions support the concept that these different

615

tissues have different expression patterns that reflect their different transformation pathways and biologic behavior. The concept that cervical squamous and glandular epithelium behave in distinct manners at the transcriptional level may have important clinical implications. In a comparison of the expression patterns between NSQ and NGL epithelium, significant differences in gene expression could be seen for KRT-4, which has been identified in squamous epithelium. Cytokeratin 4 forms the intermediate filaments in the suprabasal cells of stratified, noncornified epithelia, which is consistent with our findings [28]. KRT-4 was expressed preferentially in NSQ compared with glandular epithelium. This same pattern was observed for 12-LOX and RhGC. The inverse pattern was seen with ITGA-6; NGL expressed ITGA-6 at significantly higher levels than those observed with NSQ. The expression patterns of SCC and ACA revealed a consistently higher expression of KRT-4,

Fig. 1. Relative intensity by tissue type and gene (mean and SD). The scale of the net intensity ratios was not the same for all the genes. Significant differences are noted beneath each figure ( P b 0.05). ACA, adenocarcinoma; NGL, normal glandular epithelium; NSQ, normal squamous epithelium; SCC, squamous cell carcinoma.

616

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617

RhGC, and c-myc in SCC compared with ACA. For two of these genes, the difference was consistent with that seen between the corresponding normal epithelial tissues. However, for c-myc, an expression difference was seen between the malignant tissues but not between the normal tissues. When comparing the normal tissues to the cancerous tissues, COL-5 was down-regulated in gene expression in both malignant squamous and malignant glandular tissue. For cmyc and RhGC, there was statistically significant downregulation observed in malignant glandular and malignant squamous tissue, respectively. A review of the literature on gene expression analysis for cervical cancer yielded limited information regarding the expression patterns for cervical carcinoma. We are aware of one microarray that used cervical tissue to analyze expression patterns. In that array, squamous tissue was used preferentially. This would not allow adequate discrimination of the expression patterns for glandular tissue or ACA [6]. One study of ACA compared expression patterns to endometrial carcinoma but not to SCC [29]. A review of the expression studies of uterine cervix tissue revealed frequent use of SCC tissue for analysis [30–32]. The PCR method used to estimate original differences in abundance might vary according to the number of cycles used [33]. When testing for the relative intensity of h-actin by using different numbers of cycles, the intensity of the signal increased with increased cycle number but the ratio between the target signal and the control gene remained relatively stable (Table 8). If there was an effect of cycle number on our results, it would be to decrease the ratio of the relative intensities between the samples. The accuracy of this experiment is high given the small standard deviations observed in our gene expression ratios. A valid argument against the precision of gene expression analysis by RT– PCR is the inherent variability at the time of RNA extraction and PCR. A prior microarray of normal cervical epithelium compared with LSIL and HSIL showed a differential expression between normal and low-grade dysplasia compared with high-grade dysplasia and SCC. It did not show a differential gene expression pattern between SCC and ACA of the cervix. It showed that virtually all genes overexpressed in cervical high-grade dysplasias or SCCs were also up-regulated in ACA. This finding could have been due to an underrepresentation of ACAs within the microarray [6]. Use of b-actin as a control gene against which to quantify our target genes was selected because it is a noncompetitive standard. Its use in situations in which there is extensive remodeling of the tissues or cells is not advised. This could be significant in cells that have undergone carcinomatous transformation and would explain the larger standard deviation and variability observed in the KRT-4/hactin ratio among the SCC tissues. The same variability was also seen in the NSQ epithelium. When the samples were initially analyzed with single target PCR reactions for hactin, less variability was observed. h-Actin is usually

Table 8 Comparison of polymerase chain reaction cycle number on net intensity ratios for KRT-4 and ITGA-5 with h-actin Gene

Tissue

Cycle no.

Mean

F ratio

KRT-4

NSQ

27 31 35 27 31 35 27 31 35 27 31 35

11.48 7.19 9.40 4.93 4.11 3.44 2.08 2.91 6.07 1.54 2.70 2.98

1.05

0.37

0.19

0.83

2.44

0.11

0.69

0.51

NGL

SCC

ACA

ITGA-5

NSQ

NGL

SCC

ACA

27 31 35 27 31 35 27 31 35 27 31 35

0.99* 1.33 1.32 1.23 1.69 1.38 1.17 1.30 1.33 1.69 1.26 1.34

10.11

Prob N F

0.0005

1.64

0.21

0.46

0.64

0.39

0.68

ACA, adenocarcinoma; NGL, normal glandular epithelium; NSQ, normal squamous epithelium; SCC, squamous cell carcinoma. * P b 0.05.

expressed at higher levels than most genes. Increasing the number of PCR cycles to amplify the target gene can compensate for this higher expression [34]. Our results could be limited by the number of samples included in each group as well as by the importance that any of these particular genes may have in the transformation process that the tissues undergo. The expression differences established in this study did not denote up-regulation of the genes that are active in the ECM when comparing normal with cancerous tissue. The patterns observed reflect gene activity at a given moment in the cell cycle. They do not necessarily reflect the dynamic processes associated with cell transformation, immortalization, and invasiveness. In conclusion, the gene expression patterns observed within and between the normal and malignant epithelia for the genes evaluated do not necessarily reflect the expression patterns reported in a microarray of normal cervical epithelium compared with cervical carcinoma. When gene expression in squamous and glandular epithelium was compared, no relationship was found in the patterns of expression. These findings support our initial hypothesis that the two major benign and malignant histologic variants of uterine cervical tissue have important differences in gene expression patterns. The observed differential expression should be a compelling consideration to perform typespecific expression analysis for tumors with histological variants.

S.A. Contag et al. / Gynecologic Oncology 95 (2004) 610–617

References [1] Chan PG, Sung HY, Sawaya GF. Changes in cervical cancer incidence after three decades of screening US women less than 30 years old. Obstet Gynecol 2003;102:765 – 73. [2] Smith HO, Tiffany MF, Qualls CR, Key CR. The rising incidence of adenocarcinoma relative to squamous cell carcinoma of the uterine cervix in the United States: a 24-year population-based study. Gynecol Oncol 2000;78:97 – 105. [3] Schoolland M, Segal A, Allpress S, Miranda A, Frost FA, Sterrett GF. Adenocarcinoma in situ of the cervix. Cancer 2002; 96:330 – 7. [4] Lea JS, Sheets EE, Wenham RM, Duska LR, Coleman RL, Miller DS, et al. Stage IIB–IVB cervical adenocarcinoma: prognostic factors and survival. Gynecol Oncol 2002;84:115 – 9. [5] Shingleton HM, Gore H, Bradley DH, Soong SJ. Adenocarcinoma of the cervix: I. Clinical evaluation and pathologic features. Am J Obstet Gynecol 1981;139:799 – 814. [6] Chen Y, Miller C, Mosher R, Zhao X, Deeds J, Morrissey M, et al. Identification of cervical cancer markers by cDNA and tissue microarrays. Cancer Res 2003;63:1927 – 35. [7] Adams JC, Watt FM. Regulation of development and differentiation by the extracellular matrix. Development 1993;117:1183 – 98. [8] Kirfel J, Magin TM, Reichelt J. Keratins: a structural scaffold with emerging functions. Cell Mol Life Sci 2003;60:56 – 71. [9] Rugg EL, McLean WH, Allison WE, Lunny DP, Macleod RI, Felix DH, et al. A mutation in the mucosal keratin K4 is associated with oral white sponge nevus. Nat Genet 1995;11:450 – 2. [10] Adachi E, Hayashi T. Anchoring of epithelia to underlying connective tissue: evidence of frayed ends of collagen fibrils directly merging with meshwork of lamina densa. J Electron Microsc (Tokyo) 1994;43:264 – 71. [11] Petitclerc E, Boutaud A, Prestayko A, Xu J, Sado Y, Ninomiya Y, et al. New functions for non-collagenous domains of human collagen type IV: novel integrin ligands inhibiting angiogenesis and tumor growth in vivo. J Biol Chem 2000;275:8051 – 61. [12] Shappell SB, Olson SJ, Hannah SE, Manning S, Roberts RL, Masumori N, et al. Elevated expression of 12/15-lipoxygenase and cyclooxygenase-2 in a transgenic mouse model of prostate carcinoma. Cancer Res 2003;63:2256 – 67. [13] Pidgeon GP, Kandouz M, Meram A, Honn KV. Mechanisms controlling cell cycle arrest and induction of apoptosis after 12lipoxygenase inhibition in prostate cancer cells. Cancer Res 2002;62:2721 – 7. [14] Nie D, Tang K, Diglio C, Honn KV. Eicosanoid regulation of angiogenesis: role of endothelial arachidonate 12-lipoxygenase. Blood 2000;95:2304 – 11. [15] Hynes RO. Integrins: bidirectional, allosteric signaling machines. Cell 2002;110:673 – 87. [16] Yoon C-S, Kim K-D, Park S-N, Cheong S-W. a6 Integrin is the main receptor of human papillomavirus type 16 VLP. Biochem Biophys Res Commun 2001;283:668 – 73. [17] Freedman AS, Nadler LM. Non-Hodgkin’s lymphomas. In: Bast Jr RC, Kufe DW, Pollock RE, Weichselbaum RE, Holland JF, Frei III E, editors. Cancer medicine. 5th ed. Hamilton, Ontario7 BC Decker Inc; 2000. p. 2035 – 40.

617

[18] Ferec C, Raguenes O, Salomon R, Roche C, Bernard JP, Guillot M, et al. Mutations in the cationic trypsinogen gene and evidence for genetic heterogeneity in hereditary pancreatitis. J Med Genet 1999;36:228 – 32. [19] Chen BS, Xu ZX, Xu X, Cai Y, Han YL, Wang J, et al. RhCG is downregulated in oesophageal squamous cell carcinomas, but expressed in multiple squamous epithelia. Eur J Cancer 2002; 38:1927 – 36 [Erratum in: Eur J Cancer 2003;39:405]. [20] Hubbard T, Barker D, Birney E, Cameron G, Chen Y, Clark L, et al. The Ensembl genome database project. Nucleic Acids Res 2002; 30:38 – 41. [21] National Center for Biotechnology Information [homepage on the Internet]. Bethesda (MD): National Center for Biotechnology Information [updated 2004 Apr 7; cited 2004 Apr 13]. Available from: http://www.ncbi.nlm.nih.gov/. [22] Kreuzer K-A, Lass U, Landt O, Nitsche A, Laser J, Ellerbrok H, et al. Highly sensitive and specific fluorescence reverse transcription–PCR assay for the pseudogene-free detection of h-actin transcripts as quantitative reference. Clin Chem 1999;45:297 – 300. [23] Henegariu O, Heerema NA, Dlouhy SR, Vance GH, Vogt PH. Multiplex PCR: critical parameters and step-by-step protocol. BioTechniques 1997;23:504 – 11. [24] Ting Y, Manos MM. Detection and typing of genital human papillomaviruses. In: Innis MA, Gelfand DH, Sninsky JJ, White TJ, editors. PCR protocols: a guide to methods and applications. San Diego7 Academic Press; 1990. p. 356 – 67. [25] Berkhout RJ, Tieben LM, Smits HL, Bavinck JN, Vermeer BJ, ter Schegget J. Nested PCR approach for detection and typing of epidermodysplasia verruciformis-associated human papillomavirus types in cutaneous cancers from renal transplant recipients. J Clin Microbiol 1995;33:690 – 5. [26] de Roda Husman AM, Walboomers JM, van den Brule AJ, Meijer CJ, Snijders PJ. The use of general primers GP5 and GP6 elongated at their 3V ends with adjacent highly conserved sequences improves human papillomavirus detection by PCR. J Gen Virol 1995;76:1057 – 62. [27] Broders AC. Carcinoma of the mouth: types and degrees of malignancy. Am J Roentgenol 1927;17:90 – 3. [28] Fuchs E, Green H. Changes in keratin gene expression during terminal differentiation of the keratinocyte. Cell 1980;19:1033 – 42. [29] Dabbs DJ, Geisinger KR, Norris HT. Intermediate filaments in endometrial and endocervical carcinomas: the diagnostic utility of vimentin patterns. Am J Surg Pathol 1986;10:568 – 76. [30] Kulkarni S, Rader JS, Zhang F, Liapis H, Koki AT, Masferrer JL, et al. Cyclooxygenase-2 is overexpressed in human cervical cancer. Clin Cancer Res 2001;7:429 – 34. [31] Brummer O, Bohmer G, Hollwitz B, Flemming P, Petry KU, Kuhnle H. MMP-1 and MMP-2 in the cervix uteri in different steps of malignant transformation: an immunohistochemical study. Gynecol Oncol 2002;84:222 – 7. [32] Hazelbag S, Fleuren GJ, Baelde JJ, Schuuring E, Kenter GG, Gorter A. Cytokine profile of cervical cancer cells. Gynecol Oncol 2001; 83:235 – 43. [33] Spencer WE, Christensen, MJ. Multiplex relative RT–PCR method for verification of differential gene expression. BioTechniques. 1999;27: 1044–6, 1048–50, 1052. [34] Lion T. Current recommendations for positive controls in RT–PCR assays. Leukemia 2001;15:1033 – 7.