Comparison of benign and malignant follicular thyroid tumours by comparative genomic hybridization.

DNA copy number changes were compared in 29 histologically benign follicular adenomas, of which five were atypical, and 13 follicular carcinomas of the thyroid by comparative genomic hybridization. DNA copy number changes were frequent in adenomas (14 out of 29, 48%). Most changes were gains, and they always involved a gain of the entire chromosome 7 (10 out of 29, 34%); other common gains involved chromosomes 5 (28%), 9 (10%), 12 (24%), 14 (21%), 17 (17%), 18 (14%) and X (17%). Losses were found only in four (14%) adenomas. Two of the five atypical adenomas had DNA copy number losses, and none had gains. Unlike adenomas, gains were rare and losses were frequent in carcinomas. A loss of chromosome 22 or 22q was particularly common in carcinomas (6 out of 13, 46%), whereas a loss of chromosome 22 was found in only two (7%) adenomas, one of which was atypical (P = 0.002). A loss of 1p was also frequent in carcinomas (31%), but gains of chromosomes 5, 7, 12, 14 or X that were common in adenomas were not found. Loss of chromosome 22 or 22q was present in six of the eight widely invasive follicular carcinomas, but in only one of the five minimally invasive carcinomas. We conclude that large DNA copy number changes are common in thyroid adenomas. These changes are strikingly different from those found in follicular carcinomas consisting of few losses and frequent gains, especially those of chromosome 7. A loss of chromosome 22 is common in widely invasive follicular carcinoma.

defined as an encapsulated benign tumour showing, follicular cell differentiation (Hedmg,er et al. 1988). Despite the fact that follicular adenomas are benign neoplasms and do not gix-e rise to metastases. one quarter of them are DNA aneuploid by flow cytometry (Joensuu et al. 1988). and clonal chromosome abnormalities have been found in karvotype analvses (Bondeson et al. 1989: Tevssier et al. 1990: v-an den Berg et al. 1990: Sozzi et al. 1992: Antonini et al. 1993: Roque et al. 1993a: Belge et al. 1994: Criado et al. 1995. One-third of follicular adenomas have been reported to harbour either numerical chromosomal abnormalities or rearrangyements that are mainlx balanced (Sozzi et al. 1992). Trisomies of several chromosomes. including 5. 7. 12. 14. 18. 20 and 22 have been described in karyotype analyses (Antonini et al. 1993: Heim et al. 1995. Most frequent are trisomies of chromosomes 5. 7 and 12. detected in about 20%7 of adenomas w-ith an abnormal karvotvpe (Heim et al. 1995). Van den Berg et al (1990) suggested that a combination of numerical abnormalities. including a gain of chromosomes 4. 5. 7. 9. 12 or 16. is characteristic of follicular adenoma. In addition to follicular adenoma. nodular goitre has also show n trisomies in the same chromosomes as follicular adenoma (Roque et al. 1993b). indicating a close relationship between some types of nodular hy-perplasia and adenoma.
Follicular carcinoma is the second most common type of thyroid carcinoma after papillary carcinoma It is much less common than follicular adenoma. which is diatnosed about ten times as often as follicular carcinoma Cvtogenetic information about follicular carcinoma is limited, and only a few tumours hax-e been examined (Bondeson et al. 1989: Jenkins et al. 1990: Tevssier et al. 1990: Herrman et al. 1991: van den Berg et al. 1991: Roque et al. 1993c: Grebe et al. 1997. The short arm of chromosome 3 has been reported to contain rearrangements. and a minimal common deleted region of 3p25-pter has been described (Herrman et al. 1991: Roque et al. 1993c. Van den Berg et al ( 1991 ) reported idic(22:22 p l:pll ) and additional structural abnormalities in chromosome 22 in a case of follicular carcinoma. and Jenkins et al ( 1990) reported aberrations that were mainly deletions in three cases.
Some follicular adenomas may has-e a close morphological resemblance to follicular carcinoma: the main difference is the presence of invasion of tumour into the capsular blood vessels in carcinoma. Because of this resemblance. a question arises whether follicular carcinoma originates from a pre-existinadenoma (Franssila 1997). Follicular adenoma might represent a precancerous lesion that could transform into carcinoma through copy number chanaes in critical aenes controlling mx asion and affecting, metastasis formation. In the present study. we used comparative gyenomic hybridization (CGH) to study DNA copy number changes in follicular thyroid tumours. To our knowledgye. these tumours have not been studied by this method earlier. The results showthat.
although DNA copy number changes can frequently be detected in both types of thyroid neoplasms. the chancges differ greatly. suggesting that. in spite of similar morphology. different cenetic mechanisms may give rise to these neoplasms.

MATERIALS AND METHODS Tumour specimens and DNA isolation
The series consists of 29 follicular adenomas and 13 follicular carcinomas of the thyroid. stored in the frozen tissue bank of Department of Pathology. Helsinki University Central Hospital.
Finland (Table 1). Twenty-four (83%c) of the patients with an adenoma and ten (77%7 ) of those ith a carcinoma were women. In follicular adenoma. the median age at diagnosis was 46 years (range 25-87) and in follicular carcinoma 71 years (range 34-85). All origainal histological diagnoses were re-examined (KY) without know ledge of the CGH results. The classification used w-as that of WHO (Hedinger et al. 1989).
The tissue samples had been frozen in liquid nitrogen upon arrixal at the Department of Pathology. and stored at -80'C until analyvsis. Frozen sections were cut. stained with toluidine blue and examined to verifxthat the tissue examined contained mainly tumour tissue. In all cases. at least 70%7 of the cells analysed were tumour cells. Twenty to thirty 5-im sections were cut from each tumour specimen. and genomic DNA was isolated usinc a standard phenol-based method (Sambrook et al. 1989).
Comparative genomic hybridization (CGH) CGH w-as performed according to the method of Kallioniemi et al (1994) with some modifications. and according to the protocol descrnbed by El-Rifai et al (1997). Tumour DNA was labelled with fluorescein-dUTP and fluorescein-dC(TP (Dupont. Boston. MA. USA). and the normal reference DNA (extracted from the blood of a healthy man or woman) w-as labelled w-ith Texas red-dUTP and Texas red-dCTP (Dupont) in a standard nick translation reaction. Equal amounts (1 ji) of the labelled test and reference probes were used for hybridization. ith 10 jig of unlabelled human Cot-1 DNA to block the binding of repetitive sequences in 10 jil of the hybridization buffer [50% formamide. 10%c dextran sulphate. 2 x SSC (1 x SSC is 0.15 -x sodium chlonrde. 0.015 N sodium citrate. pH 7)]. The DNA was then denatured for 5 min at 75°C before applying it to normal lymphocyte preparations. Before hybridization. the metaphase preparations were dehydrated in a series of 70. 80 and 1001% ethanol concentrations and denatured at 65CC for 2 min in a formamide solution (70% fornamide/2 x SSC). The slides were then dehydrated on ice as described above. Then they were treated with proteinase-K at 37°C for 7.5 min (0.2 jg ml-' in 20 mnLi Tris-HCl. 2 m-i calcium chlonrde. pH 7). and once again dehydrated in a senres of nising ethanol concentrations as indicated above. Hvbridization was performed in a moist chamber at 37C for 48 h. Post-hybridization washes were as follow's: three times in 50% formamide/2 x SSC/pH 7. twice in 2 x SSC. and once in 0.1 x SSC at 45CC followed by 2 x SSC and 0.1 xI sodium dihydrogen phos-phatelO. 1 mI disodium hydrogen phosphatelO. 1%c Nonidet P40/pH 8 and distilled water at room temperature for 10 min each. The slides were counterstained with 4'.6-diamidino-2-phenylindole (DAPI) at a concentration of 0.1 jIg ml-' in an anti-fade solution.  (8) of chroxosomes with no aberrations obtained from varnus negative control expenments from the analysis (e.g. chromosomes that wvere heavilv bent or overlapping or those that had oxerly ing artefacts). Chromosomal regions were interpreted as amplified (a gain) when the red-green ratio exceeded 1.17: as highly amplified. when the ratio exceeded 1.5: and underrepresented (a loss) when the ratio was less than 0.85 (Figure 1). All findings were confirmed using a confidence interval of 99%7. A positive control with known chromosomal aberrations and a negatix e control were included in each hvbridization to x erifx the reliabilitv of the method. Chromosomal regions in the centromenrc areas of chromosomes 1. 9. 16 and Y and the p-arms of acrocentric chromosomes were discarded from the analysis because of their large heterochromatic areas. As CGH recognizes only proportional changes in DNA copy number. the ratio profiles do not indicate the absolute copy number changes. In diploid and near-diploid cells. a ratio of 1.5 indicates a 100%7 increase in the copy number in a chromosome arm or in an area of the size of a chromosome band (Knuutila et al. 1998). If this threshold is not reached the increase is only 50%7. suggesting chromosomal trisomy.

Statistical analysis
Fisher's exact test w-as used in statistical analysis. All P-x-alues are tw o-tailed.
The most frequently involxved chromosomes in follicular adenoma Axere 7 (ten cases: 34%-> and 5 (eight cases: 28%/). Particularly. a gain of the entire chromosome 7 was present in all ten adenomas that displayed one or more gains. The other chromosomes commonly gained were 12 (sexen cases: 24%). 14 (six cases: 21%9 ). X (fixe cases: 17%/-c). 17 (fixe cases: 17%/,r). 18 (four cases: 14%7) and 9 (three cases: 10%). In four cases. the onlx changes were a loss of an entire chromosome or chromosomes. These losses wxere in chromosomes 1. 2. 6. 11. 13 (case 26). 3 (case 13) and 22 (cases 22 and 25). Txxo of these four follicular adenomas x-ith a loss (nos. 25 and 26) xxere histologicallN classified as atypical adenomas. A DNA copy number gain was not found in any of the fixve atypical follicular adenomas. Follicular carcinomas displayed copy number changes in 9 (69%c out of the 13 cases analysed. On axerage. there xxere 1.6 changes per case (median 1: range 0-6). and the number of changres detected among the 13 cases was 21. Unlike adenomas. losses xxere more common than gains (16 X s. 5). and the frequency of tumours with a loss (8 out of 13. 62%7c) w-as greater than in adenomas (4 out of 29. 14%e. P = 0.003). Aberrations xxere detected in eirht different chromosomes: 1. 9. 13. 17. 18. 19. 21 and 22. Chromosome 22 was most frequentlx involved, and it was deleted in as manx as six (46%-) carcinomas. Fixe of these six tumours showed a loss of the wxhole chromosome 22. and in one case there x-as a loss of the chromosome arm 22q 12.3-qter. Loss of chromosome 22 or 22q was present in six of the eight widelx ix asix-e folhicular carcinomas. but only in one of the fi e minimalix invasixe carcinomas. In adenomas. a loss of chromosome 22 xxas found only in txo (7%c) tumours (P=0.002). one of xxhich was an atypical adenoma A loss of lp. lp21-22 and lpl3-23 xas detected in four carcinomas (cases 35. 38. 41 and 42). In addition. loss of the wxhole chromosome 9. 18 or 19 A-ere each detected in one carcinoma. and chromosome 13 in txwo carcinomas. A gain xx-as found only in four chromosomes. lq22-qter (cases 35. 38). lq24-qter (case 42) and the long arm of chromosome 17 (cases 32. 33 . No highly amplified chromosomal regions were detected either in follicular adenoma or carcinoma. None of the carcinomas had aberrations in chromosome 7. w-hich wvas gained in 10 out of the 29 adenomas (P = 0.02). and no gains were found in chromosomes 5. 12. 14 or X. which were also frequently gained in adenomas. There were only four losses that were found both in adenomas and carcinomas. They were found in chromosomes 1 (one adenoma and four carcinomas). 9 (one adenoma and one carcinoma). 13 (one adenoma and two carcinomas) and 22 (two adenomas and seven carcinomas). The only gain that was found in both types of tumours w-as that of chromosome 17q (five adenomas and two carcinomas).

DISCUSSION
We found DNA copy number changes to be frequent in follicular thyroid adenomas. Typical aberrations were gains of the entire chromosomes 5. X. 7. 12. 14. 17 and 18. Our results are in line with results of karvotvpe studies in which numerical chromosomal abnormalities. mainly trisomies. and especially trisomies of chromosomes 5. 7 and 12. have been described (Antonini et al. 1993: Belge et al. 1994: Heim et al. 1995. In a fluorescence in situ hybridization study. not only one extra copy but also several extra copies of chromosomes 7 and 12 were reported (Criado et al. 1995). Our CGH data did not. however. suggest several extra copies in these chromosomes. as high DNA copy number changes were not detected (see Materials and methods. section Digital image analysis).
In follicular carcinoma. DNA copy number losses were commonly found. A typical aberration was deletion of a part of or the entire chromosome 22. which was found in about one-half of all carcinomas. It appeared to be more frequent in widely invasive than in minimally invasive carcinomas. which suggests that this deletion mav be associated with malignant progression of follicular carcinoma. Further studies are needed to find out if deletion of chromosome 22 is correlated to survival.
Also in karyotvpe studies. deletions have been found in follicular carcinomas (Jenkins et al. 1990). A monosomy or DNA copy number loss of chromosome 22 is not unique to follicular thyroid carcinomas as it has been described in other types of human neoplasms such as meningioma. glioma. mesothelioma and gastrointestinal stromal tumour (Tonk et al. 1992: Mohapatra et al. 1995: El-Rifai et al. 1996: Bjorqkvist et al. 1997 (Schofield et al. 1996). The significance of these and other suppressor genes located in 22q in the genesis of follicular thyroid carcinoma is unsettled. Although the CGH profiles of most folhcular adenomas and carcinomas differed greatly from each other. in four follicular adenomas the only aberration detected was a loss of an entire chromosome or chromosomes. and in two of them the lost chromosome w-as 22. Of the five atypical adenomas. losses w-ere found in two. and in one of them chromosome 22 had been lost. These findings might suggest that some follicular adenomas. including the atypical adenoma. may have a common genetic origin with foflicular carcinoma. Atypical adenoma is not a well-defined entity. but rather a tumour that shows architectural and cytological features resembling those seen in follicular carcinoma. It. however. lacks invasion, the most important criterion of follicular carcinoma. In earlier studies. a subgroup of follicular carcinoma called follicular carcinoma without invasion was described (Woolner et al. 1961). These tumours were associated with excellent prognosis and would probably be called atypical adenomas at present.
In conclusion. the results indicate that in follicular thyroid adenomas extensive chromosomal changes are often present. and that these changes are mainly gains of entire chromosomes. Unlike adenomas. gains are not frequent in follicular carcinoma. whereas losses. especially those of chromosome 22 or 22q. are found often. Loss of chromosome 22 maxv be associated with the widely invasive type of follicular carcinoma.