Underexpression of peroxisome proliferator-activated receptor (PPAR)γ in PAX8/PPARγ-negative thyroid tumours

The expression of peroxisome proliferator-activated receptor (PPAR)γ in thyroid neoplasias and in normal thyroid (NT) tissues has not been fully investigated. The objectives of the present work were: to study and compare the relative expression of PPARγ in normal, benign and malignant thyroid tissues and to correlate PPARγ immunostaining with clinical/pathological features of patients with thyroid cancer. We analysed the expression of PPARγ in several types of thyroid tissues by reverse transcription–polymerase chain reaction (RT–PCR), interphase fluorescent in situ hybridisation, real-time RT–PCR and immunohistochemistry. We have demonstrated that NT tissues express PPARγ both at mRNA and at protein level. PAX8-PPARγ fusion gene expression was found in 25% (six of 24) of follicular thyroid carcinomas (FTCs) and in 17% (six of 36) of follicular thyroid adenomas, but in none of the 10 normal tissues, 28 nodular hyperplasias, 38 papillary thyroid carcinomas (PTCs) and 11 poorly differentiated thyroid carcinomas (PDTCs). By real-time RT–PCR, we observed that tumours negative for the PAX8-PPARγ rearrangement expressed lower levels of PPARγ mRNA than the NT. Overexpression of PPARγ transcripts was detected in 80% (four of five) of translocation-positive tumours. Diffuse nuclear staining was significantly (P<0.05) less prevalent in FTCs (53%; 18 of 34), PTCs (49%; 19 of 39) and PDTCs (0%; zero of 13) than in normal tissue (77%; 36 of 47). Peroxisome proliferator-activated receptorγ-negative FTCs were more likely to be locally invasive, to persist after surgery, to metastasise and to have poorly differentiated areas. Papillary thyroid carcinomas with a predominantly follicular pattern were more often PPARγ negative than classic PTCs (80% vs 28%; P=0.01). Our results demonstrated that PPARγ is underexpressed in translocation-negative thyroid tumours of follicular origin and that a further reduction of PPARγ expression is associated with dedifferentiation at later stages of tumour development and progression.

Peroxisome proliferator-activated receptors (PPARs) are members of the nuclear receptor superfamily, which form heterodimers with retinoid X receptor. The heterodimers activate the transcription of specific genes in response to binding of a ligand. Three PPAR isoforms have been described: a, b (also called d, NUC-1 or FAAR) and g (Desvergne and Wahli, 1999). PPARg is the most intensively studied isoform. It has been shown that this nuclear receptor is important in several biological pathways involving cell differentiation, insulin sensitivity, atherosclerosis and cancer (Rosen and Spiegelman, 2001). There are two protein isoforms (PPARg 1 and PPARg 2 ) generated by alternative splicing and alternative promoter usage. Peroxisome proliferator-activated receptorg 1 isoform is encoded by three transcripts, which differ in 5 0 -untranslated region (variants g 1 , g 3 and g 4 ). Variant g 2 encodes isoform PPARg 2 Sundvold and Lien, 2001). Peroxisome proliferator-activated receptorg 2 contains 30 additional amino acids in the N-terminus (Tontonoz et al, 1994). Peroxisome proliferator-activated receptorg is activated by natural ligands (fatty acids and eicosanoids) (Chawla and Lazar, 1994;Tontonoz et al, 1994;Forman et al, 1995;Kliewer et al, 1995) and by synthetic ligands (thiazolidinediones) (Lehmann et al, 1995). Peroxisome proliferator-activated receptorg activation was reported to inhibit the growth and, in some cases, to induce apoptosis or differentiation of tumour cells from different lineages: liposarcoma (Tontonoz et al, 1997;Demetri et al, 1999), breast cancer (Elstner et al, 1998;Mueller et al, 1998), prostate cancer (Kubota et al, 1998), colorectal cancer (Brockman et al, 1998;Sarraf et al, 1998;Kitamura et al, 1999), bladder cancer (Guan et al, 1999), nonsmall-cell lung carcinoma (Chang and Szabo, 2000), pancreatic cancer (Motomura et al, 2000), gastric cancer (Sato et al, 2000), renal carcinoma (Inoue et al, 2001), testicular cancer (Hase et al, 2002) and liver cancer (Toyoda et al, 2002). Kroll et al (2000) reported that t(2;3)(q13;p25), a chromosomal translocation detected in a subset of follicular thyroid carcinomas (FTCs), originates a fusion gene composed by DNA-binding domain of the thyroid transcription factor PAX8 and domains A to F of PPARg. Recently, our group and others (Marques et al, 2002;Nikiforova et al, 2002;Cheung et al, 2003) have detected the expression of PAX8-PPARg gene not only in FTCs but also in follicular thyroid adenomas (FTAs). Ohta et al (2001) studied the expression of PPARg in papillary thyroid carcinoma (PTC) cell lines and in thyroid tumours. They showed that most cell lines and half of PTCs expressed PPARg, while normal adjacent tissue and two FTAs were negative. This group as well as Martelli et al (2002) also demonstrated that PPARg agonists induce apoptosis and inhibit the growth of thyroid tumour cells.
Several studies have demonstrated that, compared to their normal counterparts, the expression of PPARg in tumour cells is either overexpressed, such as in renal cell carcinoma (Inoue et al, 2001) and testicular cancer (Hase et al, 2002), underexpressed, such as in oesophageal carcinomas (Terashita et al, 2002) or is equal to the normal tissue, such as in colonic adenocarcinomas . This last group has also identified somatic mutations of PPARg in four of 55 sporadic primary colorectal carcinomas . The expression of PPARg in thyroid neoplasias and in the normal thyroid (NT) tissue has not been fully investigated. We have expanded our previous study (Marques et al, 2002) and analysed the expression of PPARg in a series of thyroid tumours and correspondent normal tissue by reverse transcription -polymerase chain reaction (RT -PCR), interphase fluorescent in situ hybridisation (FISH), real-time RT -PCR and immunohistochemistry. We observed that PPARg expression is usually underexpressed in multiple types of thyroid tumours, and that this may be an important event in the development of thyroid neoplasias.

Materials
The number of cases analysed by each technique for the different histological groups is represented in Table 1. Paraffin-embedded tissues and frozen tissues were available in 247 samples and in 131 samples, respectively. Haematoxylin-and eosin-stained sections from each sample were evaluated histologically by two pathologists to classify tumours according to the 1988 World Health Organisation histological classification of thyroid tumours. The extent of papillary carcinomas was classified according to the system of DeGroot et al (1990) and the metastasis-age-completeness-ofresection-invasion-size-score (MACIS) (Hay et al, 1993). The system of DeGroot et al (1990) categorises the patients with PTC by clinical class: I, with intrathyroidal disease; II, with cervical adenopathies; III, with extrathyroidal invasion and IV, with distant metastasis. The prognostic score defined as MACIS was calculated according to Hay et al (1993): MACIS ¼ 3.1 (if aged p39 years) or 0.08 Â age (if aged X40 years), þ 0.3 Â tumour size (in centimetres), þ 1 (if incompletely resected), þ 1 (if locally invasive) and þ 3 (if distant metastasis present).

RNA extraction, RT -PCR and sequencing
Total RNA was extracted from frozen tumours using TRIzol reagent (Life Technologies, Inc., Gaithersburg, MD, USA), accord-ing to the manufacturer's protocol. RNA was quantified by UV spectrophotometry (optical density measured at 260 nm). Complementary DNA (cDNA) was synthesised from 1 mg of RNA at 371C for 90 min, using oligo-(dT) primers (Life Technologies, Inc.) and reverse transcriptase (Life Technologies, Inc.). PAX8-PPARg fusion gene expression was analysed by RT -PCR as described previously (Marques et al, 2002).
To analyse the expression of the various PPARg mRNA isoforms, segments from the 5 0 -terminal region of the PPARg gene were amplified by PCR using forward primers, located in exons A 1 , A 2 and B and reverse primers located in exon 1. Primer sequences are presented in Table 2. First round amplifications were performed using 1 ml of cDNA, forward primers P 1 , P 3 and P 5 and reverse primer P 7 . A measure of 1 ml of each amplification reaction was then used as template for second amplification reactions with nested primers P 2 , P 4 , P 6 (forward) and P 8 (reverse). A total of 25 ml reactions were carried out on over 35 cycles using the following conditions: 951C for 1 min, 55 -571C for 1 min and 721C for 1 min. Amplification reactions contained final concentrations of 20 mM Tris-HCl (pH 8.4), 50 mM KCl, 200 mM dNTPs (Amersham Pharmacia Biotech, Uppsala, Sweden), 1.0 -2.5 mM MgCl 2 , 10 pmol of each primer (forward and reverse) and 1.5 U of Taq DNA Polymerase (Life Technologies, Inc.). Negative controls for cDNA synthesis and PCRs, in which the template was replaced by sterile water, were included in each experiment. RNA integrity and efficiency of cDNA synthesis were tested in each sample by performing RT -PCR amplification for the housekeeping gene phosphoglycerate kinase-1 (Sugg et al, 1998). Normal colon tissue was used as positive control for the analysis of PPARg expression .
PCR products were analysed and purified by electrophoresis in a 2% agarose gel stained with ethidium bromide. Polymerase chain reaction products were also subjected to automatic sequencing (ABI Prism 310 Genetic Analyser using the ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction Kit Version 2; Applied Biosystems, PE Corporation, Foster City, CA, USA).

Interphase FISH analysis
Fluorescent in situ hybridisation was performed on isolated nuclei extracted from 50 mm paraffin-embedded sections of thyroid tumours with BAC probes for PPARg (RPCI 1130 G23, BAC PAC Resources) and PAX8 (RPCI 1165 I12, BAC PAC Resources). Briefly, PPARg clone DNA was labelled with digoxigenin and PAX8 DNA with biotin by random priming, using the Bioprime DNA labelling system (Invitrogen S.A., Barcelona, Spain). Nuclear suspensions were spotted on SuperFrost slides (Menzel-Glaser, GMbH, Memmert, Germany) and pretreated with 0.1% pepsin (Sigma-Aldrich, St Louis, MO, USA) in 0.2% HCl at 371C. Probe mixture in 50% formamide in 2 Â SSC was codenatured with nuclear DNA at 801C for 2 min. Detection of the digoxigeninlabelled PPARg probe was accomplished using an anti-digoxigenin   Kroll et al (2000).  (Lazar et al, 1999).

Immunohistochemistry
Formalin-fixed paraffin-embedded sections (3 mm) were attached to glass slides pretreated with gelatin. The sections were then dried at 371C overnight and dewaxed with xylol. Endogenous peroxidase was inhibited with 0.6% H 2 O 2 in methanol for 10 min. Antigen retrieval was performed using a stainless-steel 6-l capacity pressure cooker, with 0.01 M sodium citrate buffer (pH 6.0), for 6 min at full pressure. Slides were incubated with normal goat serum 1 : 10 (DAKO X907, DAKO Corp., Golstrup, Denmark) for 10 min before blocking the endogenous avidin and biotin (Vector SP-2001, Vector Laboratories, Inc., Burlingame, CA, USA). Peroxisome proliferator-activated receptorg primary antibody 1 : 30 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) was incubated for 30 min. Specificity of PPARg immunostaining was demonstrated by preincubating the samples with PPARg blocking peptide 1 : 10 (Santa Cruz Biotechnology, Inc.). Bound primary antibody was detected using biotinylated goat anti-mouse and anti-rabbit immunoglobulin G, being subsequently amplified with streptavidin conjugated to horseradish peroxidase (DAKO K5001; DAKO Corp.). All incubations were performed at room temperature. The peroxidase staining reaction was revealed with a solution containing 3,3 0 -diaminobenzidine tetrachloride. Sections were counterstained with Mayer's haematoxylin, dehydrated and mounted.

Statistical analysis
The frequencies of PPARg transcript variants and the level of PPARg mRNA in each tumour histotype were analysed by w 2 test and unpaired t-test, respectively. Peroxisome proliferator-activated receptorg mRNA levels in PTCs and in corresponding NT tissues were compared using a paired t-test. Peroxisome proliferator-activated receptorg immunostaining for nodular hyperplasias (NH) and thyroid tumours was compared with the staining in NT tissues by two-tailed Fisher's exact test. We also correlated the PPARg immunostaining in FTCs and PTCs with clinical/ pathological features of the patients by unpaired t-test, two-tailed Fisher's exact test or w 2 test as appropriate. P-values less than 0.05 were considered significant. Statistical analysis was performed using Graph Pad Prism version 2.0 (San Diego, CA, USA).

Analysis of PPARc transcript variants
RNA from 72 frozen samples (6 normal tissues, 29 FTAs, 9 FTCs, 24 PTCs and four poorly differentiated thyroid carcinomas (PDTCs)) was analysed by RT -PCR. Peroxisome proliferatoractivated receptorg transcript variants were determined by combining the RT -PCR results obtained for each primer pair. Peroxisome proliferator-activated receptorg 3 could be detected only in the five cases that did not present PPARg 1 , because RT -PCR with primer pairs P 3 P 7 or P 4 P 8 originated products with exactly the same size in both variants. Most thyroid tissues expressed PPARg 1 , PPARg 2 and PPARg 4 , and the proportion of specific variants expressed was similar in NT tissues and in the various types of thyroid tumours (data not shown).

PAX8-PPARc fusion gene expression
The fusion gene was detected by RT -PCR and/or interphase FISH analysis. Six out of 24 (25%) FTCs and six out of 36 (17%) FTAs were positive for PAX8-PPARg fusion gene expression. The rearrangement was not detected in 10 NT tissues, 28 NHs, 38 PTCs and 11 PDTCs.

Quantitative analysis of PPARc gene expression
The mRNA level of PPARg in thyroid tissues is represented in Figure 1.   Figure 2 Quantitative analysis of PPARg mRNA by real-time RT -PCR in thyroid tumours and in the corresponding normal adjacent tissue. The expression level in each PTC was lower than in normal adjacent tissue. The PPARg mRNA level in two FTAs was also lower than in corresponding normal tissues. One FTC translocation-positive exhibited a ratio similar to its surrounding normal tissue. PTC -papillary thyroid carcinoma; FTAfollicular thyroid adenoma; FTC -follicular thyroid carcinoma; NT -normal thyroid.  observed that 86% (six of seven) of FTCs with distant metastasis were PPARg negative (P ¼ 0.03). There was also a trend for negative tumours to be more locally invasive (75%), to have poorly differentiated areas (71%) and to have persistent disease after surgery (80%). Most (75%) of the widely invasive FTCs were PPARg negative, whereas only 38% of minimally invasive tumours did not shown PPARg staining. The two widely invasive tumours positive for PPARg had two different components: a follicular area, which stained positive, and an insular area, which was negative. Positive PPARg staining was not correlated with age, gender, tumour size or vascular invasion. In PTC cases, we observed that 72% (13 of 18) of the tumours with a classic pattern were positive, whereas 80% (12 of 15) of the follicular variants were negative (P ¼ 0.01). Peroxisome proliferator-activated receptorg staining did not correlate with any other prognostic variable but, interestingly, all class IV tumours (n ¼ 3) were negative.

DISCUSSION
We (Marques et al, 2002) and others (Kroll et al, 2000;Nikiforova et al, 2002;Aldred et al, 2003;Cheung et al, 2003) have detected cases of thyroid tumours, such as FTC, FTA or PTC, that exhibit mild or moderate diffuse PPARg nuclear staining, even though they are RT -PCR negative for the PAX8-PPARg fusion gene. The question was then whether such cases present or not overexpression of PPARg. It is important to discriminate between these two possibilities, because underlying different pathogenic mechanisms may be present. For instance, if PPARg expression is found to be upregulated in PAX8-PPARg-negative tumours, it could reflect either a breakpoint between PAX8 and PPARg in a location outside the primers used in the RT -PCR reaction, or a rearrangement between PPARg and a non-PAX8 partner, or overexpression of wild-type PPARg or point mutations in the PPARg gene. The objectives of the present work were two-fold: 1 -to study, and compare, the relative expression of PPARg in the normal gland and in benign and malignant diseases of the thyroid; and 2 -to correlate PPARg immunostaining with clinical and pathological characteristics of patients with thyroid carcinomas of follicular origin. We chose to examine PPARg expression in thyroid tissues by RT -PCR, interphase FISH, real-time RT -PCR and immunohistochemistry. We first demonstrated that NT tissues express PPARg both at mRNA and at the protein level. This is in contrast with the findings by Ohta et al (2001), who detected PPARg mRNA in four of six PTC cell lines and in three of six PTCs, but not in NT tissues or in FTAs. However, our results are in concordance with the recent data of Aldred et al (2003), who have also demonstrated PPARg expression in seven of seven NT specimens. Interestingly, the mean ratio of PPARg/GAPDH mRNA obtained by the semiquantitative method of Aldred et al (2003) of 0.7970.30 is not far from the ratio of 0.6870.40 obtained by our quantitative method. As the human PPARg gene gives rise to four mRNAs, PPARg 1 -4 , that differ at their 5 0 -end as a consequence of alternate promoter usage and splicing, and these mRNAs code two protein isoforms, PPARg 1 and PPARg 2 , which may exert distinct biological effects, we investigated the expression of the different PPARg transcripts in the thyroid tissues. We were able to show that thyroid cells express all mRNA isoforms, but the proportion of specific variants was similar in normal tissues and in the various types of thyroid tumours studied. However, because we did not perform quantitative RT -PCR, it is possible that some tumour types predominantly express one of the isoforms. To compare PPARg expression between tumours and NT tissues, we performed quantitative analysis by real-time RT -PCR. Our assay did not distinguish wild-type transcripts from PAX8-PPARg fusion mRNAs. We observed that tumours negative for the rearrangement expressed lower levels of PPARg mRNA than NT. This was particularly evident in the PTC cases (n ¼ 7) in which the normal adjacent tissue of the same patient was also available for analysis ( Figure 2). Upregulation of PPARg mRNA levels was found in four of the five (80%) translocation-positive tumours (three FTC and two FTA) analysed. However, we detected one PAX8-PPARgpositive case (FTC) with a PPARg/GAPDH ratio within the mean of the normal group. Interphase FISH analysis revealed that only a small subset of cells in this case harboured the translocation, which is consistent with the normal expression level of PPARg as assessed by real-time RT -PCR. Overall, there was a direct correlation between our real-time analysis of PPARg expression and the immunoreactive protein: strong immunostaining was present only in tumours with upregulated PPARg mRNA levels and mild or moderate staining was revealed in the remaining tumours, as well  as in normal tissues. Notably, the translocation-positive FTC with normal PPARg/GAPDH ratio showed diffuse and faint nuclear staining. Aldred et al (2003) performed semiquantitative RT -PCR analysis of PPARg expression in 14 NT tissues and in 19 FTCs and also showed that nontranslocation tumours had underexpression of PPARg. A larger number of tissues were examined by immunohistochemistry in order to determine, and compare, the prevalence of PPARg staining between normal, hyperplastic and neoplastic tissues, and to correlate staining with known prognostic variables of thyroid carcinomas. Compared to NT tissues, staining was significantly (Po0.05) less prevalent in FTCs, PTCs and PDTCs. This trend was also present in FTAs, although not statistically significant (P ¼ 0.09). Previous studies have shown that FTCs harbouring the fusion gene, hence strongly reactive with a PPARg antibody, are somewhat smaller in size (Cheung et al, 2003;Nikiforova et al, 2003), more overtly invasive, and occur at a younger age than tumours without the rearrangement (Nikiforova et al, 2003). In the study of French et al (2003), FTCs with PPARg rearrangement had vascular invasion and a solid/nested histology more frequently than translocation-negative tumours. We observed that PPARg-negative FTCs were more likely to be locally invasive, to persist after surgery, to metastasise and to have poorly differentiated areas.
We could not correlate PPARg staining with any of the prognostic variables analysed in the group of PTCs, except for tumours presenting a predominantly follicular pattern that were more often negative (80%) than classic PTCs (28%; P ¼ 0.01).
In summary, we have demonstrated underexpression of PPARg in PAX8/PPARg-negative thyroid tumours of follicular origin, and that a further reduction of PPARg expression is associated with dedifferentiation at later stages of tumour development.