Low-Level Genomic Instability Is a Feature of Papillary Thyroid Carcinoma: An Array Comparative Genomic Hybridization Study of Laser Capture Microdissected Papillary Thyroid Carcinoma Tumors and Clonal Cell Lines

Papillary thyroid carcinoma (PTC) is thought to be a relatively genome-stable neoplasm. In contrast to follicular thyroid carcinoma, almost all PTCs are diploid.1 They have been shown to be microsatellite stable2 and generally show low levels of loss of heterozygosity.3,4 However, there is at least some evidence that allelic loss may occur and may be related to poor prognosis.3 

The literature has shown that PTC is associated with several specific molecular genetic events, the best documented of which are ret/PTC rearrangements and the more recently described and, more commonly in the context of sporadic PTC, BRAF^V600E mutation. Although activating point mutations of the RAS oncogene are also considered to account for the development of a small subset of adult sporadic PTCs,5 the true relevance of RAS mutations to the pathogenesis of PTC is undermined when one considers that similar mutations of the RAS gene have been documented at an even higher prevalence in nodules from benign multinodular goiters (21%) and benign microfollicular adenomas (25%).5 

BRAF^V600E now appears to be the leading genetic event in adult sporadic (non–radiation exposed) PTCs6,7 and shows strong concordance with the classic papillary phenotype. ret/PTC and BRAF mutations appear to largely occur in a mutually exclusive manner,8 although overlap, the significance of which is currently unknown, occurs in a minority of cases.9,10 

Thus, BRAF mutations are the most frequent genetic aberration found in sporadic PTCs, and the mutation of one of the intermediates of the RET/PTC-RAS-BRAF pathway accounts for a genetic event, potentially involved in tumor initiation, in two thirds of PTCs.7,8 However, the genetic trigger(s) for most of the remaining PTCs remains unknown. Furthermore, the multistep theory of tumorigenesis predicts that more than one genetic or epigenetic event is associated with the development of carcinoma. Further work on the genetic events associated with initiation and progression of PTC is needed.

Comparative genomic hybridization (CGH) was first described in 1992.11 It is a technique that allows comprehensive analysis of multiple DNA copy number gains and losses across the entire genome, in one single experiment. In metaphase CGH the identification of genes involved in chromosomal aberrations has been difficult because of the limited mapping resolution of this technique, which does not allow precise localization of the genetic aberrations.12,13 The sensitivity of CGH can be adversely affected by contamination of tumor material with normal cells, and a tumor component representing a minimum of 70% of the DNA extracted from the tissue is highly desirable. The use of laser capture microdissection or another microdissection technique is indispensable by ensuring near 100% tumor content for CGH experiments. The sensitivity of CGH depends on the number and size of the copy number changes. Simulation experiments indicate that a copy number increase of 50% should be detectable if the region is 2 megabase pairs (Mbp) or larger, and an amplified region (amplicon) of 250 kilobase pairs (kbp) would need a 400% copy number increase to be detectable with metaphase CGH.14 The lower limit of detection is determined by the product of the excess copy number and the size of the amplified region. When a deletion is 100% (no copies present), a resolution of 1 to 2 Mbp can be achieved.15 The use of arrayed DNA fragments, such as large-insert genomic clones or cDNA clones, as hybridization targets has dramatically increased the resolution of CGH,16 potentially down to ∼100 kbp, and even higher resolution is possible with tiled single nucleotide polymorphism arrays.

In this study the Vysis Genosensor platform (Abbott Molecular Inc, De Plaines, Ill) was used for array CGH. The Array 300 microarray (Abbott Molecular Inc) contains triplicates of 287-target clone DNAs (P1 and BAC clones with 861 discrete spots), representing the most common oncogenes and tumor suppressor genes encountered in the scientific literature. In addition, the array includes clones, which mark known areas of loss of heterozygosity in cancer and loci of common microdeletions and unique subtelomeric sequences relevant to other genetic diseases. The details of clone names, cytogenetic location, and Genbank accession number of these clones are available (www.vysis.com). Although the array provides an average of 40 Mbp coverage of each chromosome, it must be emphasized that this coverage is not homogenous along the chromosome and effectively amounts to 1% coverage of the entire genome. For this reason, comparison between the Vysis Array 300 and the results of conventional CGH are not straightforward. Nevertheless, the Vysis array remains a useful tool for investigating the pathobiology of cancer.

The aim of this study was to assess DNA copy number gain and loss in PTC tumor samples and clonal cell lines using the Genosensor array CGH platform in conjunction with clonal cell lines and laser capture microdissected tissue. Cells from prospectively accessioned PTC were laser capture microdissected to ensure near 100% tumor content. The BRAF status and ret/PTC status of each of the cases were assessed by TaqMan allelic discrimination assay and TaqMan reverse transcriptase polymerase chain reaction (Applied Biosystems, Foster City, Calif), respectively.

All protocols were subject to and approved by the local ethics committee. Samples of PTC were prospectively collected and snap frozen at −80° C. Prior to laser capture microdissection and DNA extraction, frozen section was performed to confirm morphology of frozen tumor samples. The PTC tumor cell lines were cultured and DNA harvested for CGH. Growth conditions and cell culture protocols are available from the authors. Table 1 shows the details of the specimens and cell lines included in the study.

Laser capture microdissection was performed using the Pixcell II platform (Arcturus Bioscience Inc, Mountain View, Calif) on frozen sections stained with hematoxylin-eosin as previously described.17 Both DNA and RNA were extracted from the pure tumor cell populations captured. Nucleic acid extraction was performed using Gentra protocols (Gentra Systems Inc, Minneapolis, Minn) as previously described.18 

Chimeric transcripts of ret/PTC1 and ret/PTC3 were detected using breakpoint-spanning probes with minor groove binding modifications, as previously described.19 Amplification of PCR products was detected in real time. Positive control material included the use of the TPC1 cell line, known to express ret/PTC1, which was used for array CGH and also as a positive control for ret/PTC1 analysis of tissue samples. RNA derived from a previously cloned ret/PTC3–containing plasmid was used as a positive control for ret/PTC3 TaqMan analysis.

An allelic discrimination–type TaqMan assay (Figure 1) was used to type the presence and zygosity of BRAF^V600E. Laser capture microdissection ensured pure populations of malignant cells to prevent contamination by normal cells with wild-type BRAF sequences. The primer sequences and cycling conditions for this assay have been described previously.9 Positive control material was provided by Paula Soares, PhD (IPATIMUP, Portugal) in the form of cell lines of known BRAF mutation status, K2-homozygous for BRAF^V600E, BCPAP-heterozygous for BRAF^V600E, and TPC1 wild-type BRAF expression and known ret/PTC1 expression.

Array CGH was performed as described previously.18 The TPC1, BCPAP, and K2 PTC cell lines, which had been used as sources of control material, were also analyzed by CGH. Laser capture microdissection was used to ensure homogeneous populations of tumor cells.

COSH is a control and validation DNA provided by Vysis. The DNA consists of a mixture of DNAs extracted from 3 tumor cell lines (COLO 320 HSR, SJSA-1, and BT-474). The mixture of cell lines has known amplifications of MYC, GLI, and MDM2, among others. Array CGH of COSH was used as an external validation of operator technique for each batch of experiments. Array CGH performed on COSH revealed amplifications of these genes in accordance with the data from Vysis.

There is no internal control clone or group of clones with the Vysis Genosensor system that can be used for internal experimental validation. One potentially useful way to incorporate an internal control is to use reference DNA of the sex opposite to that of the sample being arrayed (ie, male sample, XY; female reference, XX). In cases where the sex chromosomes are unaffected by tumor-associated aberrations, the results will indicate loss of the X chromosome loci with gain of the Y chromosome loci. This strategy is straightforward for pre- and postnatal clinical samples. With oncology samples, more discrepancies are seen within the sex markers, and fewer markers may be detected than with normal blood. For the sex markers, 5 of the 11 X chromosome loci and both the Y chromosome loci can be reliably used to determine the confidence in the results. The reliable loci on the X chromosome are STS 3′ (Xp22.3), STS 5′ (Xp22.3), KAL (Xp22.3), AR 3′ (Xq11-q12), and XIST (Xq13.2). The two Y loci on the array are SRY (Yp11.3) and AZFa (Yq11). Of course, true tumor-associated aberrations may affect these loci, and anomalous results may therefore occur. The other markers are not always detected because of some homology between X and Y for those specific clones (Teresa Ruffalo, Vysis, written communication, October 2003). For these reasons and because the sex of the cell lines was not known in 2 cases (cases 1 and 2), amplifications and deletions involving the X and Y chromosome were not included in the results.

To determine the variations in the ratios of the spots in normal control DNA, comparative hybridizations using test and reference DNA from 3 normal samples were performed. This approach has been used previously by many other groups.20–22 The mean ratio of the normal array CGH was 1.01 for each of the 3 hybridizations. The mean standard deviation was 0.09. A value of the mean ratio ± 2 standard deviations was set as the cutoff level for normal gene copy number. Ratios of higher than 1.19 and lower than 0.81 were considered to have copy gain and copy loss, respectively. This correlates closely with the manufacturer-recommended figures of 0.8 and 1.2 for a loss and a gain, respectively. As discussed above, internal validation was performed in 2 of the cases (cases 7 and 11), Table 2. In case 7 (female test and male reference), 4 of 5 expected gains on the X chromosome and 1 of 2 expected deletions on the Y chromosome were detected. In case 11 (male test and female reference), 5 of 5 expected losses and 2 of 2 expected gains were detected. In cases where the sexes were matched, occasional aberrations of the X and Y loci were identified; this may rationalize the anomalous ratios for KAL and SRY that occurred in case 7.

The cohort analyzed consisted of 3 cell lines and 8 prospectively accessioned PTCs comprising 4 classic morphology PTCs, 3 follicular variant PTCs, and a single case of PTC (case 8) with mixed/multidivergent differentiation including areas of classical morphology, follicular variant morphology, diffuse sclerosis, and mucoepidermoid differentiation. With regard to this case, all of these areas were separately microdissected, and no evidence of expression of either of the 2 common ret rearrangements or of mutant BRAF was detected in any of the distinct morphologic areas. To ensure adequate DNA quantity, CGH was performed on DNA from a single sample acquired from cells from all these variable areas.

Table 3 shows ret/PTC and BRAF mutation status. Chimeric transcripts of ret/PTC1 were detected in the TPC1 cell line, which is known to express ret/PTC1. No surgical tumor case showed evidence of ret/PTC1. ret/PTC3 was detected in none of the 11 samples. Heterozygous mutations of BRAF (V600E) were detected in the heterozygous control cell line (BCPAP) and in 3 surgical cases (cases 4, 5, and 6). The occurrence of BRAF^V600E was seen in association with 3 of 4 classic variant PTCs and not in any of the 3 cases of follicular variant PTC. Homozygous mutation of BRAF was detected in the homozygous control K2 cell line (case 2).

Array CGH revealed DNA sequence copy number changes in all 3 cell lines and all 8 surgical cases of PTC examined. The numbers of amplifications and deletions were expressed as a percentage of the number of somatic loci on the array (N = 273). With array CGH, this approach more accurately reflects the degree of genetic instability than the total number of aberrations per case commonly used in presenting metaphase CGH data results. Results from the 13 sex chromosome loci were not included because sex chromosomes were used for validation as discussed above. A single somatic cell locus was excluded because although present on 3 of the arrays it was not reliably spotted on different array batches.

Aberrations (including both gains and losses) were detected in all cases, and aberrations were detected at a higher rate in cell lines (mean, 43.8%) than in surgical tissue samples (mean, 6%), Figure 2. The percentage of aberrations detected per locus examined in the tissue samples showed no significant association with either the morphologic subtype (follicular variant or classic PTC) or the mutation status (BRAF wild type vs BRAF^V600E). However, the unusual case of PTC showing mixed/multidivergent differentiation (case 8) predictably showed the highest percentage of total aberrations of the surgical cases.

Figure 3 demonstrates the percentage of gains and losses per case analyzed and shows that both gains and losses occurred in each tumor and cell line with no significant predominance of one versus the other.

Figure 4 shows the fluorescence ratios for all the cases examined across all the somatic loci. Large amplifications only occurred in cell lines (high peaks in cases 1 and 2). For all other cases, the maximum fluorescence ratio for a gain never exceeded 2.5. This indicates that no high-level amplifications of any of the loci occurred in surgical tissue samples.

Tables 4 and 5 show recurrent gains and losses, respectively, occurring in all the cell lines and tissues examined. An arbitrary cutoff threshold of 20% was used to exclude common aberrations occurring at low frequency.

Because of the great number of aberrations seen in the 3 cell lines in comparison with prospectively collected tissue samples and also because of the known inherent genomic instability of cell lines, the data for the tissue samples were independently analyzed for recurrent aberrations occurring exclusively in the 8 tissue samples. Recurrent gains and losses present in the 8 surgical cases are presented in Tables 6 and 7, respectively.

A stated aim of this study was to look for potential molecular triggers in PTC lacking either mutant BRAF or ret/PTC activation. Table 8 demonstrates that gains of PDGFB (22q13), TP 73 (1p36.33), and SNRPN (15q12) were exclusively detected in tumors without these known molecular triggers. Furthermore, all of these gains were found in cases showing variant morphology (mixed or follicular variant), with the exception of gain of SNRPN, which was seen in a single case of classical morphology PTC (Table 8).

The aim of these experiments was to perform an assessment of DNA copy gain and loss occurring in both PTC cell lines and prospectively collected surgical samples of PTC. The rearranged ret and BRAF mutation status of each case was analyzed so as to allow any candidate genes to be assessed with reference to the already-known classic molecular triggers of PTC. No case showed expression of ret/PTC3. Only the TPC1 cell line showed the expected expression of ret/PTC1. All of the cases were collected during the last 2 years and the absence of ret activation reflects a temporal trend that our group has noted in PTC accessioned during the last 20 years.9 In this cohort of PTC, mutant BRAF was exclusively detected in tumors of classic papillary morphology. This finding is consistent with the recent literature.

Previous studies in the literature have used varying technologic approaches to identify candidate oncogenes and tumor suppressor genes in PTC. Classic cytogenetic studies identified abnormalities in 27% of cases, with the most common recurrent changes involving inv (10) (q22.2q21.2) in 7%, t(10;17)(q11.2;q23) in 3%, and chromosome 1 aberrations in 3% of cases.23,24 Loss of heterozygosity studies have also yielded a low rate of detection of aberrations (23%), without any identifiable predilection to specific chromosomal loci.24,25 However, Califano et al26 showed higher levels of loss of heterozygosity than other groups did.

Comparative genomic hybridization is a powerful molecular cytogenetic method that provides enhanced ability to screen solid tumors for genetic aberrations without prior knowledge of potential aberrations, as is required for loss of heterozygosity studies. Previous CGH studies of PTC have yielded varying results. All of the previous studies have used conventional metaphase CGH. Chen et al27 identified several novel copy number changes at a low prevalence in 16 cases of PTC, including gains at chromosome arms 1p34 (4 cases), 1p36 (5 cases), 1q42 (4 cases), 2p21 (3 cases), 2p13 (3 cases), 5q31 (3 cases), 5q33→34 (4 cases), 9q32→34 (4 cases), 14q32 (5 cases), 16q22→24 (4 cases), and 19q13.1 (8 cases) and a loss at 16q12→q13 (3;cl16 cases). In the present study it was possible to corroborate the findings of Chen and coworkers at several loci including the gains of 1p34, 1p36, 5q31, 5q33→34, and 14q32 and the loss at 16q12→13 (Tables 4 and 5). In some cases the Vysis array does not contain clones mapping to loci detected by Chen et al27 (eg, 1q42, 2p21, and 9q32→34). The most common gain in PTC described in the data of Chen et al is 19q13.1; however, although a candidate gene (AKT2) from this locus is represented on the Vysis array, amplification was not detected in any of the 11 cases examined in the present study. These points again emphasize some of the limitations of making direct comparisons of array CGH and classic CGH data.

In a study by Singh et al28 of 21 cases of PTC, genetic abnormalities were identified by CGH in less than half of the cases. A recurrent pattern of aberrations was seen in cases where genetic changes were detected, involving losses at chromosome arms 1p and 9q and chromosomes 17, 19, and 22 and gains of chromosome 4 and chromosome arms 5q, 6q, 9q, and 13q. Loss of chromosome 22 was unique to younger patients (P = .05) and was associated with a higher rate of regional lymphatic metastasis (19% vs 80%, P = .02). Again, the data in the present study partly corroborate the findings of Singh et al, showing the losses of distal 1p and gains of 5q and 13q, but the full-length chromosomal losses and gains have not been reproduced in the present data. Furthermore, the loss of the entire length of chromosome 22, which was associated with younger patients and regional node metastasis, is not seen in any of the 11 cases in the present study.

Hemmer et al29 reported that CGH abnormalities were rare in PTC (3/26 cases, 12%). All of the aberrations in this study were gains, including 5q14→23.3, 5q23.3qter, gain of the entire lengths of chromosomes 7 and 17, and gains of 21q22.1qter and 1q23qter. The presence of these aberrations was correlated with older age and the presence of cervical nodal metastasis. The data of Hemmer et al show no obvious overlap with the current data presented here; however, the patient (case 8) with the greatest number of aberrations detected in our cohort (11.3% of all loci) was aged 73 years at diagnosis and showed unusual mixed morphology. All of the other patients were young (<40 years) at diagnosis and had small numbers of genomic aberrations. Hemmer et al also analyzed follicular carcinoma and found that loss of the entire length of chromosome 22 was commonly associated with this type of thyroid carcinoma but not with PTC. With regard to chromosome 22, the data from the study by Hemmer and colleagues clearly conflict with those of Singh et al.28 

In a study of 25 cases with varying histopathologic and clinical features by Kjellman et al30 from the Karolinska Institute, chromosomal imbalances were detected at a higher rate than in other studies. Our data show a similar high rate of aberrations, 100% of 11 cases versus 84% of the 25 cases reported by Kjellman et al. All of the other studies indicate that aberrations occur in a small minority of cases. Gains and losses were approximately equally frequent (54% vs 46%), a feature also notable in our data (Figure 3).

In the series reported by Kjellman et al,30 the most frequent alteration was gain of 9q33→qter (7/25 cases [28%]). Other recurring alterations were gain of the entire X chromosome (20%), of 1q (16%), of 17q (16%), and of 22q (12%). High-level amplification was detected at distal 1q in 1 case. Losses were most frequently detected on 22q (12%) and on 9q21.3→q32 (12%).

By comparison, the data presented in this paper show no gains of 9q33-qter or 1q, but gains on 17 q and 22q were seen. The gain of chromosome X occurring in male patients was not seen in the 2 male patients (Cases 6 and 8) who had identically sex-matched reference DNA in our cohort. None of the losses described by Kjellman and coworkers are reproduced in our cohort.

Bauer et al31 analyzed 15 cases by CGH. Only 4 (27%) showed aberrations, with an average of 3.25 aberrations per case (range, 2–4). All PTC samples with abnormal CGH had vascular or capsular invasion. There was significant correlation between the presence of chromosomal aberrations and invasion. All 4 cases with CGH abnormalities had a partial or complete gain of chromosome 20, and 3 of the 4 also had a partial or complete loss of chromosome 13. Possible loss at chromosome 4q was observed in cases with younger age. A gain of chromosome 12q24 was observed in a patient with a history of external beam radiation. A single tall cell variant had a gain of chromosome 15. Microsatellite instability was not detected in any case.

The primary finding of the current study is that DNA copy gain and loss appear to occur in all cases of PTC when tumors are examined using a combination of laser capture microdissection and sensitive array-based CGH. Copy gain and loss are more frequent in cell lines, which no doubt reflects the ease with which aggressive-behaving tumors can be maintained in immortal culture, leading to a selection bias for highly genomically unstable tumors, and the artificial effects of prolonged culture and passage. In the surgical cases, which more accurately reflect true tumor biology, aberrations occur at a low percentage of all the loci examined, with a mean percentage of aberrations of 6% of all loci examined (range, 1.45%–11.35%).

In the current study, taking amplified genes occurring in more than 36% of the cases as a threshold (Table 4), the most common amplifications involve loci of well-described oncogenes related to the mitogen activated protein kinase pathway (MAPK), including FGR/SRC2 on 1p36.2→p36.1 (FGR—Gardner Rasheed feline sarcoma viral oncogene homolog, a protein tyrosine kinase), EGFR on 7p12.3→p12.1 (epidermal growth factor receptor, a protein tyrosine kinase), FGF4 on 11q13 (fibroblast growth factor 4, a classical activator of the MAPK pathway), and PDGFB on 22q13.1 (platelet-derived growth factor B [SIS], another classical activator of the MAPK pathway). This finding is significant when one considers that activating mutations of ret/PTC and BRAF all modulate the MAPK pathway and emphasizes the central role played by the MAPK pathway in the pathogenesis of PTC.

It can be seen from reviewing the literature that CGH data on PTC have yielded disparate results and conclusions. This neoplasm is clearly morphologically heterogeneous and shows molecular heterogeneity. It may well be that DNA copy gain and loss occur as a secondary event to established molecular triggers such as ret/PTC activation and BRAF mutation. For this reason, in the current study an attempt has been made to assess DNA copy gain and loss in a small cohort of very well-characterized tumors. Also, an attempt has been made to overcome the effect of normal cellular contamination by using laser capture microdissection and increasing the resolution of the technique by targeting specific known oncogenes and tumor suppressor genes using array CGH. In all of the previous conventional CGH studies cited above, no specific microdissection measures were utilized, which may go some way toward explaining the absence of aberrations detected in many studies.

The most potentially important findings of the current study are the recurrent amplifications of PDGFB, TP73, and SNRPN occurring in those tumors without either of the common molecular triggers. It must be acknowledged that the numbers of cases are too small to infer statistical significance. However, other evidence points to a role in particular for PDGFB in the pathogenesis of PTC. A recent expression microarray study identified PDGF expression as a possible diagnostic marker of PTC.32 Overexpression of the protein was confirmed by immunohistochemistry. It is possible that copy gain of this gene will lie behind the increased expression of the protein. Yano et al also demonstrated over expression of FGF4 in their study. FGF4, which activates MAPK, was the most commonly amplified oncogene in the 11 cases of PTC described in the current study. As discussed above, activation of MAPK pathways is a feature of all the known PTC oncogenes discovered thus far.

The current study demonstrates that copy gain and loss occur in more cases of PTC than heretofore seen in the literature. The combination of laser capture microdissection and array based CGH may be an optimal combination for detecting copy gain and loss in tumors. This study emphasizes the central role played by the MAPK pathway in the pathogenesis of PTC. Amplification of PDGFB occurs exclusively in tumors without mutation of BRAF or expression of ret/PTC chimeras. The mechanism of over expression of both FGF4 and PDGF, recently described in the literature, may well be copy gain at the genomic level.