Abnormalities of Fragile Histidine Triad Genomicand Complementary DNAs in Cervical Cancer:Association With Human Papillomavirus Type

Carolyn Y. Muller, John D. O’Boyle, Kwun M. Fong, Ignacio I. Wistuba,Eric Biesterveld, Mohsen Ahmadian, David S. Miller, Adi F. Gazdar,John D. Minna*

Background:Chromosome 3p14.2 contains FRA3B, the mostactive chromosome breakage site in the human genome. Thefragile histidine triad (FHIT) gene, a putative tumor sup-pressor gene, overlaps FRA3B. Human papillomavirus(HPV), a known cofactor in cervical carcinogenesis, can in-tegrate into FRA3B. We examined abnormalities in FHITand its RNA transcripts in cervical cancer cell lines andtumors. We also investigated the relationship between loss ofheterozygosity (LOH) in FHIT/FRA3B and the presence ofoncogenic HPV types.Methods:Eleven cell lines, 40 tumors(20 fresh and 20 archival), and 10 normal cervical epitheliawere examined. Two intragenic polymorphic markers(D3S1300 and D3S4103) and the polymerase chain reaction(PCR) were used to examine FHIT LOH. Reverse transcrip-tion–PCR (RT–PCR) analysis and single-strand conforma-tion polymorphism analysis of RT–PCR products were usedto characterize FHIT transcripts. Oncogenic HPV typeswere identified by PCR, using general and type-specificprimers. Results:All normal epithelia, 19 of 20 fresh tumorsand nine of 11 cell lines expressed wild-type and, occasion-ally, exon 8-deleted FHIT transcripts. Additional aberrantFHIT transcripts were seen in nine of 20 fresh tumors and inseven of 11 cell lines. DNA sequencing of the aberrant tran-scripts revealed a variety of insertions and deletions but nopoint mutations. Three cell lines also had homozygous FHITdeletions. Oncogenic HPV types (i.e., 16, 18, 31, and 33) weredetected in 18 of 20 archival tumors, and, in these tumors,LOH within FHIT was identified in nine of 16 informativecases. HPV 16 was found to be associated with LOH in theFHIT/FRA3B region ( P = .041).Conclusion:FHIT/FRA3B isfrequently altered in cervical cancer, demonstrating LOH,occasional homozygous deletions, and frequent aberranttranscripts not found in normal epithelia. However, the pres-ence of wild-type transcripts and the lack of protein-alteringpoint mutations raise questions about FHIT’s function as aclassic tumor suppressor gene in cervical tissue. [J Natl Can-cer Inst 1998;90:433–9]

The sequence of molecular events responsible for cervicalcarcinogenesis has not been elucidated. Although oncogenictypes of human papillomavirus (HPV) DNA have been found inmore than 90% of cervical squamous cell carcinomas, only afraction of the women who harbor oncogenic HPV in their lowergenital tract will develop cervical cancer. Loss of heterozygosity(LOH) at some loci on chromosome 3p has been observed and

occurs in both invasive (70%) and high-grade preinvasive (61%)cervical lesions, suggesting the involvement of a 3p tumor sup-pressor gene(1). Many tumors, including those of lung, breast,kidney, head and neck, ovary, and cervix, demonstrate 3p LOHinvolving at least four distinct chromosomal regions(2–8). Ofrecent interest is the 3p14.2 locus, which contains the aphidico-lin-inducible FRA3B fragile site(9). This site has also beenreported to be an integration site for HPV 16(9) and demon-strates frequent LOH in cervical carcinoma(1,7,8).Within thisregion and spanning FRA3B is the candidate tumor suppressorgene, the fragile histidine triad (FHIT) gene(10). In this study,we have conducted a detailed evaluation of FHIT/FRA3B, in-cluding mutation and transcript analysis, in cervical cancer celllines and primary cervical tumors, and we have examined theassociation between these changes and their oncogenic HPVstatus.

Materials and Methods

Cell lines.Nine human epidermoid cervical carcinoma cell lines (C33A, SiHa,HeLa, MS751, C4I, C4II, Caski, HT-3, and ME180) were obtained from theAmerican Type Culture Collection (ATCC) (Rockville, MD). Two additionalvials of HeLa and MS751 were obtained later in the study and are designatedHeLa B and MS751 B; the former cell lines are thus designated HeLa A andMS751 A, for a total of 11 cell lines studied. Cells were grown to near conflu-ence in Dulbecco’s minimum essential medium with 10% fetal bovine serum(Life Technologies, Inc. [GIBCO BRL], [Gaithersburg, MD]), antibiotic/antimycotic (10 000 U/mL penicillin G sodium, 10 000mg/mL streptomycinsulfate, and 25mg/mL amphotericin B) (Life Technologies, Inc.), and gentami-cin 10 mg/mL (Life Technologies, Inc.) at 37 °C under 5% CO2.

Primary cervical tumors. Twenty paraffin-embedded archival early-stagecervical tumor specimens were retrieved from hysterectomy specimens for pre-cision microdissection as previously reported (cases 1–20)(1). Additional freshhuman cervical tumor biopsy specimens (prefix X or UT) were obtained at thetime of staging surgery from patients treated by members of the Division of

*Affiliations of authors:C. Y. Muller, J. D. O’Boyle (Department of Obstet-rics and Gynecology, Hamon Center for Therapeutic Oncology Research); K. M.Fong, I. I. Wistuba, E. Biesterveld, M. Ahmadian (Hamon Center for Therapeu-tic Oncology Research); D. S. Miller (Department of Obstetrics and Gynecol-ogy); A. F. Gazdar (Hamon Center for Therapeutic Oncology Research andDepartment of Pathology); J. D. Minna (Hamon Center for Therapeutic Oncol-ogy Research and Departments of Internal Medicine and Pharmacology), TheUniversity of Texas Southwestern Medical Center, Dallas.

Correspondence to:Carolyn Y. Muller, M.D., Hamon Center for TherapeuticOncology Research, The University of Texas Southwestern Medical Center,5323 Harry Hines Blvd., Dallas, TX 75235–9032. E-mail: cmulle@mednet.swmed.edu

See‘‘Notes’’ following ‘‘References.’’

© Oxford University Press

Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998 ARTICLES 433

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from

Gynecologic Oncology at The University of Texas Southwestern Medical Cen-ter, Dallas. Fresh tumor samples were snap-frozen in liquid nitrogen and storedat −150 °C. Eighteen primary and one recurrent squamous cell carcinoma andone primary cervical small-cell carcinoma were evaluated. The mean age of thepatients was 45.1 years (range, 29.6–70.0 years). Six patients had stage I disease,seven had stage II disease, four had stage III disease, and three had stage IVdisease as defined by the 1985 classification of the Federation of GynecologicOncologists(11). In addition, epithelial scrapings of exocervix (n4 3), endo-cervix (n 4 7), or endometrium (n4 4) were obtained from normal uteri inpatients requiring hysterectomies for noncancerous medical indications. Scrap-ings were taken immediately after surgical removal of the uterus and placed inRNAzol (Tel-Test, Friendswood, TX) for total RNA isolation. Informed consentwas obtained from all patients involved in fresh normal epithelial or tumoracquistion after approval from the institutional review board.

DNA and RNA isolation. Total genomic DNA was isolated from trypsinizedcell lines by proteinase K digestion with phenol/chloroform extraction. Protein-ase K digestion was performed on the microdissected paraffin slides using theoptimized protocol described previously(2). Total RNA was isolated directlyfrom the frozen homogenized tumors or from trypsinized cell lines and epithelialscrapings by either the RNAzol (Tel-Test) method by the manufacturer’s pro-tocol or by the cesium chloride method(12).

Complementary DNA (cDNA) synthesis.First-strand cDNAs were synthe-sized from 6 to 8mg of total RNA by use of oligo-dT primers and a commercialprotocol (Superscript II; Life Technologies, Inc.). The quality of the cDNAs wastested by successful amplification of glyceraldehyde-3-phosphate dehydrogenase(GAPDH) cDNA using specific primers of either a 728 base-pair (bp) (forwardprimer: 58-AGTCAGCCGCATCTTCTT-38; reverse primer: 58-CGTTC-AGCTCAGGGATGA-38) or 1.0-kilobase (forward primer: 58-AGTCAGCC-GCATCTTCTT-38; reverse primer: 58-CTGGGGCTGGTGGTCCAGGG-38)products. Ten of 14 samples of normal epithelia demonstrated good qualitycDNA as determined by GAPDH PCR followed by electrophoresis in a 1%agarose gel.

Reverse transcription (RT)–PCR.One microliter of the confirmed cDNA(not further quantified) was used for FHIT RT–PCR amplification covering theopen-reading frame using nested and non-nested strategies. Non-nested PCR wasperformed with primers designed from FHIT exons 5 and 9 (amplifying nucleo-tides 378–788, Genbank U46922) and nested PCR (amplifying nucleotides 203–1038, and then reamplifying nucleotides 203–904) was performed as previouslydescribed using [32P]deoxycytidine triphosphate(13). Products were diluted innondenaturing gel-loading dye and were subjected to electrophoresis in 4%polyacrylamide gels at 30 W. The gels were then subjected to autoradiography.

Direct sequencing of transcripts.Selected FHIT RT–PCR products of wild-type size and aberrant size were excised from the gels. DNA was eluted andreamplified in a 40-mL reaction volume using appropriate original or M13 tailedprimers described by Fong et al.(13). The final products were separated in 1%agarose gels, excised, eluted, spin purified (Ultrafree-MC; Millipore Corp., Bed-ford, MD), and then ethanol precipitated. Purified PCR products were sequencedin both forward and reverse directions either by manual cycle sequencing (Se-quiTherm; Epicenter, Madison, WI) or automated (ABI Prism 373, AppliedBiosystems, Foster City, CA) sequencing.

Mutational analysis. FHIT single-strand conformation polymorphism(SSCP) analysis covering the open-reading frame was performed on nestedRT–PCR products of cell lines, primary tumors, and normal cervical epithelia asdescribed(13).

Genomic DNA analysis.Genomic DNA from cervical cancer cell lines weresubjected to eight separate PCR amplifications using primers designed to am-plify each of the FHIT exons 3–10 as previously described(13). In addition, themicrosatellite markers D3S1300 and D3S4103, which map within FHIT intron 5,were studied in the 11 cervical cancer cell lines and in the 20 archival earlyinvasive microdissected cervical cancers for heterogeneity or allele loss as pre-viously described(1).

HPV analysis. HPV status was determined by PCR analysis using generalprimers and positive cases were subtyped using type-specific primers as previ-ously described(1).

Statistical analysis.Comparisons between aberrant FHIT transcripts in cer-vical cancer versus normal epithelium and comparisons between LOH at mi-crosatellite markers D3S1300 and D3S4103 and HPV status were performedusing two-sided Fisher’s exact test. HPV-positive cases were categorized aseither containing HPV 16 DNA or other oncogenic HPV types (HPV 18, 31, and33).

Results and Discussion

The candidate tumor suppressor gene, FHIT, localizes to3p14.2 and has been found to be altered in many solid tumors(14–19). To study the role of FHIT and FRA3B in cervicalcancer, we examined cervical cancer cell lines, primary tumors,and normal epithelia for hemizygous and homozygous deletions,aberrant-sized FHIT transcripts, and point mutations. We alsoevaluated FHIT LOH and HPV status to identify any associa-tions between HPV subtype and allele loss.

Homozygous FHIT/FRA3B Deletions in Cervical CancerCell Lines

DNA from 11 epidermoid cervical cancer cell lines wastested for homozygous deletions by 10 separate PCRs, eachcovering one of the FHIT exons 3–10 and two intragenic FHITmarkers (D3S1300, D3S4103), to detect small deletions (Fig. 1;Table 1). Homozygous deletions were found in three of the 11cell lines (HeLa A, C33A, and MS751 A) (Table 1); however,our results support the notion that the large homozygous deletionin the HeLa A line tested is likely due to cell culturing pressuresand may not reflect the status of the original tumor. C33A wasfound to have an exon 5 deletion with retention of the intron 5regions tested. MS751 A had a consistent homozygous deletionby PCR of D3S4103 in intron 5. However, we also noted vari-able results when testing for homozygous deletions with othermarkers including, for example, D3S1300 within intron 5 (Fig.1). These data are consistent with a complete homozygous de-letion of the teleomeric end of intron 5 (D3S4103); however, thevariable results with the other markers also suggests geneticheterogeneity for the proximal intron 5 and exon 5 homozygousdeletions. Boldog et al.(20) using Southern blot techniques alsofound genetic heterogeneity of the homozygous deletions inMS751.

It is possible that the varying results seen between our studyand those of others (Table 1) could occur due to cross contami-nation of cell lines in culture. Since the genetic heterogeneitywithin MS751 was similar within the two separately obtainedsamples from ATCC (MS751 A and MS751 B), this wouldsuggest that if contamination exists it would have originated atthe time of preparation of the cell stocks. To further evaluate thispossibility to explain the variation in our HeLa results comparedwith those of other investigators(10), two independently ob-tained HeLa cell lines (HeLa A and HeLa B) were tested forhomozygous deletions. HeLa A demonstrated homozygous de-letions involving exons 3–5 and D3S1300 (Fig. 1). There wereno homozygous deletions seen in any FHIT exons (3–10) orintron 5 markers tested in HeLa B (data not shown). To confirmthat both samples were truly of HeLa origin and not due to crosscontamination or possible mislabeling, 10 additional microsat-ellite markers on chromosomes 3, 5, 8, 9, 13, and 17 were usedfor allelotyping. The alleles of eight markers were identical be-tween HeLa A and HeLa B while two of the 10 microsatellitemarkers show the identical bands plus a gain of a new band inHeLa B (data not shown). These results demonstrate that theHeLa samples A and B are from the same source but there wasgenetic heterogeneity of the FHIT homozygous deletions thatprobably occurredin vitro during propagation of the cell lines.Alternatively, it is possible that these varying results may be due

434 ARTICLES Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from

to minor contamination that exists in some HeLa cell lines andbecomes differentially propagated under differing growth con-ditions. Druck et al.(21) also found FHIT homozygous deletionheterogeneity in HeLa cells obtained from four different sources.Boldog et al.(20) have described homozygous deletions affect-ing up to 87.5% of cervical cancer cell lines when the entireFHIT/FRA3B region is evaluated by cosmid/phage-derivedprobes covering the entire genomic region. It would appear thatmany of the deletions they described in SiHa, Caski, and C41 didnot include the FHIT exons that we found to be intact (Table 1).

Aberrant FHIT Transcripts in Cervical Carcinoma CellLines and Primary Tumors

We used two methods to search for aberrant transcripts: anon-nested RT–PCR approach, covering coding region exons5–9 and yielding a 411-bp (nucleotides 378–788) wild-type re-verse transcript (Fig. 2, A), and a nested approach, amplifyingexons 3–10 (nucleotides 203–1038, and then nucleotides 203–904), and producing a 702-bp wild-type reverse transcript (Fig.3). In addition, we excised and sequenced eight wild-type sizedreverse transcripts and 36 aberrant sized reverse transcripts de-rived from tumors and normal tissues. Eleven cervical cancercell lines, 20 fresh cervical tumor samples, and 10 normal epi-thelial samples were tested. Representative non-nested RT–PCR

results are shown in Fig. 2, A, and representative nested RT–PCR results are shown in Fig. 3. The results combining bothmethods are summarized in Table 2.

In non-nested RT–PCR, all samples except two tumor celllines (HeLa A and C33A) expressed a wild-type transcript of theFHIT open-reading frame (Fig. 2, A; Table 2). This was ex-pected, for both the HeLa A and C33A cell lines because theydemonstrated genomic homozygous deletion of exon 5, whichcontains the site of priming for the forward oligonucleotideprimer in the non-nested assay. Since MS751 demonstrated aheterogeneous clonal population, a wild-type FHIT transcriptwas amplified from the population that maintained exon 5.

Although a wild-type FHIT transcript was detected in tumorX16 by non-nested RT–PCR, nested RT–PCR revealed mostlyaberrant transcripts (Fig. 3, B;see below). Twenty of the 41samples, including normal tissues, expressed an FHIT exon 8-deleted splice variant (four sequenced) in non-nested RT–PCR(Table 2). In addition, a normal cervical epithelial sample, threecell lines, and three tumors expressed aberrant transcripts withrepetitive AT sequences (two sequenced) spliced in betweenexons 5 and 8. Since portions of FRA3B have been found to berich in AT content and LINE and MER repeats,1 BlastN search(www.ncbi.nlm.nih.gov/Blast) of the nonredundant databasesdid not reveal any significant homologies that would suggestthat this repetitive AT-rich insertion is not derived from FHITintron 4 sequence (GenBank U66722)(21).

As noted above, a nested RT–PCR approach was also used.Other authors have cautioned about the use of nested RT–PCRfor evaluating low-level FHIT transcripts because of the possi-bility that aberrant bands may be due to PCR artifact(19). Wehave found that most of the aberrant-nested transcripts that wedetected (Fig. 3; Table 2) are reproducible on repeat testing,although there is a background of nonspecific bands, which hasalso been seen by others(13,18).Nevertheless, for complete-ness, particularly for comparison of the current work to theinitial description of FHIT abnormalities(10), we performed adetailed nested RT–PCR analysis with sequencing of aberrantbands. We were struck by easily finding wild-type-sized tran-scripts in the tumor cell lines using the non-nested RT–PCRapproach but their notable absence using the nested RT–PCRstrategy. Second, we saw many more aberrant transcripts usingthe nested approach. We cut out and sequenced 14 of the wild-

Table 1. Comparison of the genomic homozygous fragile histidine triad(FHIT) gene deletions in cervical cancer cell lines

Cell lines Present study* Boldog et al.(20)†

HeLa A‡ Exons 3–5, intron 5 Discontinuous deletion exon 5, intron 4HeLa B‡ Intact Not doneSiHa Intact Intron 4Caski Intact Intron 4, exon 5, intron 5C4I Intact Intron 5C4II Intact Not doneMS751 Exon 5, intron 5 Intron 4, exon 5, intron 5ME180 Intact IntactHT-3 Intact Not doneC33A Exon 5, intron 5 Exon 5, intron 5

*By polymerase chain reaction methodology using exon-specific and intron-specific primers used in this study.

†By Southern blot hybridization using mapped cosmid/phage clones.‡Although HeLa A showed a large region of homozygous deletion, HeLa B

was wild-type for all of FHIT.

Fig. 1. Analysis of genomic DNA from cervicalcancer cell lines for homozygous deletions of thefragile histidine triad (FHIT) gene, exons 3–5and the intron 5 marker D3S1300. C33A washomozygously deleted for D3S1300 and exon 5(nonspecific bands visible). MS751 showed ho-mozygous deletions in seven of 11 differentpolymerase chain reaction (PCR) assays forexon 5 and in five of eight assays for D3S1300in both radioactive labeled and cold PCR assays.Another intron 5 marker, D3S4103, was deletedin five of five assays (not shown). All reactionswere performed using the same batch of DNAmade from an identical cell culture passage.HeLa A showed homozygous deletions forD3S1300 and exons 3–5. HT-3 was intact forexons 3–5 and D3S1300 and is presented as apositive control.

Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998 ARTICLES 435

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from

type and aberrant transcripts from cell lines, 16 of the aberranttranscripts from tumors, and two wild-type transcripts from nor-mal tissues. While we saw the exon 8-deleted variant and the ATrepetitive insert in tumor cDNAs and a subset of normal cDNAs,we never found other aberrant transcripts in normal tissue. Incontrast, we found consistent aberrant bands in seven of 11tumor cell lines and in nine of 20 of the fresh tumor samples (P4 .003). The aberrant transcript detected in tumor X16 wasfound to have a 6-bp insertion between nucleotides 811 and 812,outside of the FHIT protein coding region. The wild-type tran-scripts in this tumor are likely due to stromal cell contamina-tion. Since we have shown that genetic heterogeneity exists inthe HeLa and MS751 cell lines tested and failed to demonstrateaberrant transcripts in one strain of each line (Table 2), wecannot conclude that the aberrant transcripts seen in the otherstrains are tumor specific and are represented as such in theFisher analysis.

We found a variety of different additional aberrant tran-scripts, all of which were similar to previously reported FHITcDNA abnormalities in lung and breast cancers(13,18). Se-quence analysis confirmed that most aberrant bands were due toFHIT alternate splice variants, deletions, or alternate Alu se-quence2 insertions between exons similar to those previouslydescribed(13,18).An 11-bp alternate splice from nucleotide 812to 823 was also commonly identified as described previously(13,18).HeLa A did not show any product as expected, since theexon 3 priming site was shown to be homozygously deleted. C4Idemonstrated multiple bands consisting of exon 4–7 deletions,while C4II demonstrated a single aberrant band consisting of anAlu repeat sequence spliced between exon 5 and 9. C4I and C4IIare distinct lines from the same cervical tumor, again suggestingthat aberrant FHIT splice variants possibly occur during cell line

propagation. Interestingly, tumor X7 had three distinct ALUsequences inserted between exons 4 and 8 and 4 and 9. In con-trast, none of the 10 normal cervical epithelial samples testeddemonstrated significant abnormal transcripts by nested RT–PCR, indicating that these aberrant transcripts may be tumorspecific. It is likely that these aberrant transcripts represent ab-normal splicing, since multiple variants exist in cell lines andtumors that had no genomic homozygous deletions. As previ-ously reported, we found that nested PCR was less sensitive thanthe non-nested RT–PCR for detecting the exon 8-deleted splicevariant.(18).

Other investigators have described varying numbers of aber-rant FHIT transcripts in cervical cancer cell lines(20).Ohta et al.(10) initially described both wild-type and aberrant FHIT tran-scripts in HeLa. Boldog et al.(20) found multiple identical ab-errant bands in C33A and SiHa and no wild-type transcripts. Wehave found multiple different aberrant transcripts in these celllines. In addition, repeat testing confirmed that an exon 4-deleted splice variant in SiHa was the predominant transcript,since the exon 5–7 deleted splice variant was not always ampli-fied. While it is possible that the different results between theinvestigators are indeed due to PCR artifact, we favor the notionthat these are authentic low- abundance transcripts that are vari-ably expressed, depending onin vitro cell culture selection pres-sures.

RT–PCR–SSCP: Polymorphisms but No Mutations in theCoding Region

No point mutations were found in any of the cell lines orprimary tumors on SSCP screening with overlapping primersspanning the entire FHIT open-reading frame. Frequent poly-morphisms were identified at nucleotide 626 (G/A), nucleotide

Fig. 2. A) Non-nested reverse transcription–polymerase chain reaction (RT–PCR) of the fragile histidine triad (FHIT) exons 5–9. A band similar in size tothe expected 411-base-pair (bp) product is seen in all cell lines except C33A andHeLa A (showing a nonspecific band), which have homozygous deletions ofexon 5. MS751 demonstrates an amplified product despite its heterogeneouspopulation partially deleted in exon 5. All primary cervical tumors (prefix X)appear to express wild-type transcripts. Open arrow marks the 380-bp product(HT-3, SiHa, Caski, and NL4) that represents a transcript with an AT-rich insertspliced between exons 5 and 9, with associated loss of exons 6–8 as determinedby sequence analysis. Filled arrow4 the 342-bp exon 8 splice variant seen in

HT-3, SiHa, Caski, ME180, NL1, NL4, and NL5 confirmed by cycle-sequenceanalysis. This likely represents a normal splice variant since it is not tumorspecific. The complementary DNA of sample NL 2 was degraded and not am-plified by glyceraldehyde-3-phosphate dehydrogenase primers.B). Single-strandconformation polymorphism analysis of the FHIT nucleotides 637–856. TumorsX7 and X16 both show abnormally migrating bands. Sequencing of X7 aberrantbands showed an insertion of 35 bp, with no homology found within the Gen-bank database as determined by a BLASTN search. Sequencing of the homo-zygous abnormal migrating band in tumor X16 revealed a small 6-bp insertionbetween exons 9 and 10 which may represent an error in splicing.

436 ARTICLES Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from

656 (A/G), and nucleotide 545 (G/A). A point mutation wasfound in exon 4 at nucleotide 307 (G/A) outside the protein-coding region in tumor X31. SSCP analysis of nucleotides 637–856 revealed an additional cDNA abnormality in tumor X7 thatwas not detected by non-nested or nested RT–PCR analysis (Fig.2, B). A hemizygous 35-bp insertion was found between nucleo-tides 811 and 830, with no matched homology in a BLASTNsearch. This insertion lies outside the FHIT coding exons.

FHIT/FRA3B Allele Loss and HPV 16 Status

Because of the presence of aberrant FHIT transcripts in nineof the fresh tumor specimens but in none of the normal tissuesand because of the known examples of HPV 16 integration inFRA3B, we wondered if all of these observations might reflectthe presence of HPV integration in this region. As a way to startstudying this possibility, we went back and re-examined theassociation of FHIT LOH and oncogenic HPV type in our pre-vious study(1), which reported a 56% LOH at FHIT (nine of 16informative cases) in 20 microdissected early-stage cervical can-cers by use of intragenic FHIT markers (seeTable 3). Oncogenic

HPV types were found in 18 (90%) of the 20 tumors tested(Table 3). Oncogenic HPVs are reported to randomly integratewithin the host genome but may induce clonal selection by in-tegrating within fragile sites or a protooncogene or a tumorsuppressor gene region. However, since HPV 16 has been re-ported to integrate within intron 4 of the FHIT gene, we exam-ined the association between HPV 16 and FHIT/FRA3B LOH.Fifteen of the 20 cases studied were informative for LOH at anyFHIT locus and also demonstrated the presence of HPV DNA(Table 3). Two cases (cases 6 and 8) lacked HPV DNA, and fourcases (cases 1, 6, 16, and 17) were noninformative for any LOHat the FHIT locus. There was a statistically significant (P 4.041) association between HPV 16 sequences and FHIT LOH asdetermined by use of a two-sided Fisher’s exact test. We havenot yet been able to demonstrate actual HPV integration withinFRA3B in these tumor samples, while the cDNAs from the freshtumor samples did not allow for any HPV testing.

We have previously shown that LOH within FHIT/FRA3B isa frequent and early event in invasive cervical tumors(1). Here,we show that homozygous FHIT/FRA3B exon deletions werefound in cervical cancer cell lines in conjunction with aberrant

Fig. 3. A). Nested reverse transcription–polymerase chain reaction (RT–PCR) ofthe fragile histidine triad (FHIT) exons 3–10 showed numerous aberrant tran-scripts in cervical cancer cell lines but not in normal epithelia. The wild-type702-base-pair (bp) transcript is shown by the shaded double arrow with the largerheteroduplexes formed due to an 11-bp exon 10 alternate splice product (opendouble arrow). No FHIT transcripts were noted in HeLa A and Caski. HeLa Alost the exon 3 primer site as determined by the homozygous deletion in thegenomic analysis. Caski, although having apparent intact exons 3 and 10 bygenomic analysis, may have an altered priming site at either end. The one-sidedgray-filled arrows represent aberrant bands that were excised and sequenced forthe cervical cell lines and primary tumors. Examples of these results are asfollows: arrows for HT-3 mark the doublets that have the same heteroduplexpattern for the wild-type FHIT transcripts and represent an exon 4-deleted splicevariant heterogeneous with the 11-bp alternate splice product in exon 10. C4IIhad a single-splice variant deleted for exons 4–8 as marked by the single arrow.

Sequencing of other unlabeled representative aberrant transcripts revealed anexon 5–7- and exon 5–8-deleted splice variants in C33A; exon 5–8 deletion withan Alu sequence2 insert in C4I; and an exon 4-deleted splice variant and exon5–7 and 5–8-deleted splice variants in SiHa.B andC) Nested RT–PCR of theFHIT exons 3–10 in primary cervical tumors. Significant aberrant transcripts areseen in tumors X7, X9, X13, and X16 (black-filled arrows) with associated lowerlevels of normal-sized transcripts. Low-level aberrant bands were seen in tumorsX27, X29, X31, X10, and X11 (open single arrows). Sequencing of these ab-errant transcripts revealed similar splice abnormalities as was seen in the cervicalcancer cell lines. Most of these transcripts within the primary tumors had dele-tions of single exons and/or serial exons (4–8). Similar ALU inserts within areasof deleted exons, as seen some cell lines, were also seen in primary tumor X7.Tumor X26 is a malignant melanoma derived from a pelvic lymph node (primarysite uncertain) and is not further evaluated in this study.

Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998 ARTICLES 437

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from

and occasionally absent wild-type FHIT transcripts. However,we also find that some of these deletions may result fromin vitroselection (Table 1). Several smaller-sized FHIT transcripts (exon8 deleted and AT insertion with deletions) are seen in cell lines,primary tumors, and some normal epithelia, and are thus nottumor specific. We find that tumor cell lines and fresh tumors,but not normal epithelia, frequently exhibit other aberrant FHITtranscripts. However, in nearly every case, we also found awild-type transcript involving the FHIT open-reading frame.Furthermore, no point mutations were found in the FHIT codingregion and only a single somatic point mutation has been de-scribed in a primary tumor(22). It is possible that the abnormaltranscripts seen predominantly in the cell lines and tumors occuras a bystander effect of the inducible FRA3B fragile site. It istempting to speculate that instability of the fragile site in somecervical tumors may be due to HPV 16 integration withinthe FHIT region. It is also possible that random integration ofHPV 16 causes greater genomic instability than that caused byother oncogenic HPVs, which is reflected within this region, butthese notions need to be further studied. Thus, until additionalinformation on FHIT function or protein expression is obtained,our studies are consistent with the idea that FHIT does notfunction as a classical tumor suppressor gene in human cervicaltissue.

Table 2. Abnormal fragile histidine triad (FHIT) gene in cervical cancer cell lines and primary tumors

Reverse transcription–polymerase chain reaction products* Reverse transcription–polymerase chain reaction products*

Nestedwt†

Nestedaberrant‡

Non-nestedwt

Non-nestedexon 8 del

Non-nestedrepetitive

AT/and delNested

wt†Nested

aberrant‡Non-nested

wtNon-nestedexon 8 del

Non-nestedrepetitive

AT/and del

Cell lines TumorsMS751A − 0 + − − X6 + 0 + + +MS751B + 2 + − − X7 + 3 Weak + − +HeLa A − 0 − − − X9 + 5 Weak + − v weak +HeLa B − 3 + Weak + − X10 + 4 + − −HT-3 − 2 + + Weak + X11 + 4 + + −C4I − 4 + − − X12 + 0 + + −SiHa − 3 + + + X13 + 4 + + −Caski − 0 + + + X14 + 0 + + −C33A − 3 − − − X15 + 0 Weak + − −ME180 + 0 + + − X16 − 1 Weak +§ − −C4II − 1 + − − X17 + 0 + + −

Normal epithelia X18 + 0 + + −NL 1 + 0 + + − X23 + 0 + + −NL 2 Degraded Degraded Degraded Degraded Degraded X25 + 0 + − −NL 3 + 0 + + − X27 + 3 + − −NL 4 + 0 + + + X28 + 0 + − −NL 5 + 0 + + − X29 + 4 + + −NL 6 + 0 + − − X31 + 1 + + −NL 7 + 0 + − − X32 + 0 + − −NL 8 + 0 + − − UT105 + 0 + + −NL 9 + 0 + − −NL 10 + 0 + − −

*Reversed transcribed from total messenger RNA and subsequently amplified bymeans of the polymerase chain reaction (PCR). Nested4 nested PCR primers, wt4 wild-type, aberrant4 aberrant RT–PCR species, nonnest4 non-nested PCR(single round, one primer pair), repetitive AT/del4 transcripts that have lost exonsin the open-reading frame with subsequent repetitive AT-rich sequences insertedbetween exons 5 and 8. Results are as follows: +4 present, −4 absent, weak +4 only present on long film exposure (confirmed by sequencing after reamplifica-tion), and v weak4 present on longer exposures on most repetitive assays, likelyinsignificant.

†Nine of 11 cell lines tested had no normal transcripts as determined by nestedPCR (MS751 B demonstrated low-level wild-type transcript). Of interest, HeLa B,which did not demonstrate genomic homozygous deletions within the tested exonsand introns, did show wild-type transcripts in both nested and non-nested PCR.

‡Actual number of four aberrant transcripts are listed—after DNA sequenceconfirmation of representative bands. For example, four bands are seen in HT-3,representing heteroduplex formation of two aberrant transcripts (exon 4-deletedsplice variant and exon 4-deleted splice variant with an additional 11-base-pair (bp)loss at nucleotides 812–823). SiHa shows three sets of doublets (doublets resultingfrom the presence or absence of the above-described 11-bp loss), representing thesame exon 4 loss as seen in HT-3, a loss of exons 5 to 7, and a loss of exons 5 to8. Seven of nine primary tumors that demonstrated aberrant FHIT transcripts hadone to five aberrant transcripts. No aberrant transcripts were detected in normaltissues by nested reverse transcription-PCR; however, aberrant transcripts that werenot tumor specific were seen in the non-nested RT–PCR.

§The presence of wild-type transcripts in the X16 tumor is likely due to stromalcell contamination.

Table 3. Summary of FHIT (fragile histidine triad) gene and FRA3B regionloss of heterozygosity (LOH) and human papillomavirus (HPV) types in

archival, primary invasive cervical cancers

Samples* D3S1300* D3S4103*Any FHIT

locusHPVtype

Case 1 NI NI NI 16Case 2 LOH LOH LOH 16Case 3 LOH LOH LOH 16Case 4 Heterozygous LOH LOH 16Case 5 LOH Heterozygous LOH 16Case 6 NI NI NI No HPVCase 7 Heterozygous NI Heterozygous 18Case 8 NI LOH LOH No HPVCase 9 NI Heterozygous Heterozygous 16Case 10 LOH LOH LOH 18Case 11 NI LOH LOH 16Case 12 LOH Heterozygous LOH 16Case 13 NI LOH LOH 16Case 14 Heterozygous Heterozygous Heterozygous 18Case 15 Heterozygous NI Heterozygous 18Case 16 NI NI NI 16Case 17 NI NI NI 16Case 18 Heterozygous Heterozygous Heterozygous 33Case 19 Heterozygous NI Heterozygous 31Case 20 Heterozygous NI Heterozygous 16

*Data on 3p LOH for cases 1–20 have previously been published(1). NI 4

noninformative. Markers D3S1300 and D3S4103 are within intron 5 of the FHITgene.

438 ARTICLES Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from

References

(1) Wistuba II, Montellano FD, Milchgrub S, Virmani AK, Behrens C, ChenH, et al. Deletions of chromosome 3p are frequent and early events in thepathogenesis of uterine cervical carcinoma. Cancer Res 1997;57:3154–8.

(2) Hung J, Kishimoto Y, Sugio K, Virmani A, McIntire DD, Minna JD, et al.Allele-specific chromosome 3p deletions occur at an early stage in thepathogenesis of lung carcinoma. JAMA 1995;273:1908.

(3) Chen LC, Matsumura K, Deng G, Kurisu W, Ljung BM, Lerman MI, et al.Deletion of two separate regions on chromosome 3p in breast cancers.Cancer Res 1994;54:3021–4.

(4) Kovacs G, Erlandsson R, Boldog F, Ingvarsson S, Muller-Brechlin R,Klein G, et al. Consistent chromosome 3p deletion and loss of heterozy-gosity in renal cell carcinoma. Proc Natl Acad Sci U S A 1988;85:1571–5.

(5) Maestro R, Gasparotto D, Vukosavljevic T, Barzan L, Sulfaro S, BoiocchiM. Three discrete regions of deletion at 3p in head and neck cancers.Cancer Res 1993;53:5775–9.

(6) Jones MH, Nakamura Y. Deletion mapping of chromosome 3p in femalegenital tract malignancies using microsatellite polymorphisms. Oncogene1992;7:1631–4.

(7) Kohno T, Takayama H, Hamaguchi M, Takano H, Yamaguchi N, Tsuda H,et al. Deletion mapping of chromosome 3p in human uterine cervical can-cer. Oncogene 1993;8:1825–32.

(8) Karlsen F, Rabbitts PH, Sundresan V, Hagmar B. PCR–RFLP studies onchromosome 3p in formaldehyde-fixed, paraffin-embedded cervical cancertissues. Int J Cancer 1994;58:787–92.

(9) Wilke CM, Hall BK, Hoge A, Paradee W, Smith DI, Glover TW, et al.FRA3B extends over a broad region and contains a spontaneous HPV16integration site: direct evidence for the coincidence of viral integration sitesand fragile sites. Hum Mol Genet 1996;5:187–95.

(10) Ohta M, Inoue H, Cotticelli MG, Kastury K, Baffa R, Palazzo J, et al. TheFHIT gene, spanning the chromosome 3p14.2 fragile site and renal carci-noma-associated t(3;8) breakpoint, is abnormal in digestive tract cancers.Cell 1996;84:587–97.

(11) Beahrs OH, Henson DE, Hutter RVP, Myers MH, editors. Manual forstaging of cancer. 3rd ed. Philadelphia: Lippincott, 1988:151–3.

(12) Kingston RE. Preparation of RNA from eukaryotic and prokaryotic cells.In: Ausubel FM, Brent R, Kingston RE, Moore DD, Seidman JG, Smith JA,et al. Current protocols in molecular biology. New York: John Wiley &Sons, 1997.

(13) Fong KM, Biesterveld EJ, Virmani A, Wistuba I, Sekido Y, Bader SA, etal. FHIT and FRA3B 3p14.2 allele loss are common in lung cancer andpreneoplastic bronchial lesions and are associated with cancer-related FHITcDNA splicing aberrations. Cancer Res 1997;57:2256–67.

(14) Sozzi G, Alder H, Tornielli S, Corletto V, Baffa R, Veronese ML, et al.Aberrant FHIT transcripts in Merkel cell carcinoma. Cancer Res 1996;56:2472–4.

(15) Sozzi G, Veronese ML, Negrini M, Baffa R, Cotticelli MG, Inoue H, et al.The FHIT gene 3p14.2 is abnormal in lung cancer. Cell 1996;85:17–26.

(16) Virgilio L, Shuster M, Gollin SM, Veronese ML, Ohta M, Huebner K, etal. FHIT gene alterations in head and neck squamous cell carcinomas. ProcNatl Acad Sci U S A 1996;93:9770–5.

(17) Negrini M, Monaco C, Vorechovsky I, Ohta M, Druck T, Baffa R, et al.The FHIT gene at 3p14.2 is abnormal in breast carcinomas. Cancer Res1996;56:3173–9.

(18) Ahmadian M, Wistuba II, Fong KM, Behrens C, Kodagoda D, SaboorianMH, et al. Analysis of the FHIT gene and FRA3B region in sporadic breastcancer, preneoplastic lesions, and familial breast cancer probands. CancerRes 1997;57:3664–8.

(19) Thiagalingam S, Lisitsyn NA, Hamaguchi M, Wigler MH, Willson JD,Markowitz SD, et al. Evaluation of the FHIT gene in colorectal cancers.Cancer Res 1996;56:2936–9.

(20) Boldog F, Gemmill RM, West J, Robinson M, Robinson L, Li E, et al.Chromosome 3p14 homozygous deletions and sequence analysis ofFRA3B. Hum Mol Genet 1997;6:193–203.

(21) Druck T, Hadaczek P, Fu TB, Ohta M, Siprashvilli Z, Baffa R, et al.Structure and expression of the human FHIT gene in normal and tumorcells. Cancer Res 1997;57:504–12.

(22) Gemma A, Hagiwara K, Ke Y, Burke LM, Khan MA, Nagashima M, et al.FHIT mutations in human primary gastric cancer. Cancer Res 1997;57:1435–7.

Notes1LINE 4 long interspersed nuclear elements; MER4 medium reiteration

frequency interspersed elements. These transposon-like DNA elements are in-terspersed through the human genome.

2Alu sequences are transposon-like elements that are interspersed throughoutthe human genome; the diploid genome contains approximately 750 000 Alusequences.

Supported by Public Health Service grant HD00849 from the National Insti-tute of Child Health and Human Development, National Institutes of Health,Department of Health and Human Services. C. Y. Muller is an AAOGF-NIHCDFellow of the Reproductive Scientist Development Program. J. D. Minna issupported in part by Texas ARP grant 003660-080.

Manuscript received May 22, 1997; revised December 19, 1997; acceptedJanuary 14, 1998.

Journal of the National Cancer Institute, Vol. 90, No. 6, March 18, 1998 ARTICLES 439

by guest on Decem

ber 23, 2014http://jnci.oxfordjournals.org/

Dow

nloaded from