Well-differentiated and dedifferentiated liposarcomas (WDLPS/DDLPS) contain amplified sequences from the 12q13-15 region, including the MDM2 and CDK4 genes (Fig. 1).1, 2 The detection of MDM2 and CDK4 expression or amplification is now a reliable tool for distinguishing benign adipose tissue tumors from WDLPS, and DDLPS from poorly differentiated tumors.3 However, many questions concerning the genomic structure and the role of the 12q13-15 amplified region in the pathogenesis of WDLPS/DDLPS remain unresolved. Most data concerning the characterization of the amplified regions in WDLPS/DDLPS were obtained before the establishment of complete human genome maps and were also based on small series of cases. It is now established that MDM2 (12q15) and CDK4 (12q14.1) are separated by a distance of 11.06 Mb (http://genome.ucsc.edu) on the chromosome 12 long arm (Fig. 1) and that the 12q13.3-12q15 region contains 164 genes (http://www.ncbi.nlm.nih.gov/projects/genome/genemap99/). These data raise the question of the role of 12q13-15 genes other than MDM2 and CDK4 in the amplification process of WDLPS/DDLPS. In particular, the amplification and expression status of HMGA2 (12q14.3) and DDIT3 (12q13.3), the roles of which are respectively well known in lipomas2 and myxoid/round cell liposarcomas (MLS),2 need to be determined in WDLPS/DDLPS. Another gene of interest is YEATS4 (GAS41) (Fig. 1), which is telomeric to MDM2 and is amplified in gliomas.4, 5 In addition, YEATS4 encodes a protein that interferes, like MDM2, with the p53 pathway.6, 7
We investigated the genomic structure of the 12q13-15 amplified region in a series of 38 WDLPS/DDLPS samples using fluorescence in situ hybridization (FISH) with a panel of 12q13-21 bacterial artificial chromosome (BAC) probes. We aimed first to determine whether MDM2 and CDK4 belonged to a unique amplicon, and second to study the amplification status of HMGA2, DDIT3, DYRK2, TSPAN31 and YEATS4 in a large series of WDLPS/DDLPS. In addition, using quantitative RT-PCR (QRT-PCR), we studied the expression levels of MDM2, CDK4, HMGA2, DDIT3, DYRK2, TSPAN31 and YEATS4 in 11 of the 38 cases in comparison to normal subcutaneous adipose tissue (NSAT), normal visceral adipose tissue (NVAT) and lipomas.
Material and methods
WDLPS and DDLPS specimens
Twenty-eight WDLPS and 10 DDLPS cases were studied. The clinico-pathological features of the 38 samples from 35 patients are summarized in Table I. In all cases, the diagnosis of WDLPS or DDLPS was established according to the World Health Organization Classification of Tumors.2
Table I. Clinicopathologic and Cytogenetic Characteristics of 38 Cases of WDLPS/DDLPS and 5 Cases of Lipomas
Control samples: Normal subcutaneous and visceral adipose tissues and lipoma specimens
NSAT from 5 nonobese patients and NVAT from 5 other nonobese patients who underwent surgery for nonmalignant disease were studied as controls. Five cases of lipomas were selected as controls for QRT-PCR analysis. Clinico-pathological and cytogenetic features of the 5 lipomas are summarized in Table I.
Cytogenetic and FISH analysis
After surgical removal, tumor tissue was disaggregated and cells from short-term cultures (4–10 days) were used for metaphase cell preparations, according to standard procedures. Long-term cultures were also performed in cases 2, 16 and 17. The cytogenetic features of the 38 samples are summarized in Table I. Cytogenetic and preliminary FISH results for cases 1–7, and 17 have previously been described in Pedeutour et al.1 and Sirvent et al.8 The FISH probes used in our study are listed in Table II. BAC clones from the Roswell Park Cancer Institute library were selected according to their location on the University of California Santa Cruz database (http://genome.ucsc.edu/; March 2006 release). They were obtained from the Children's Hospital Oakland Research Institute (CHORI) (http://bacpac.chori.org/), the Resources for Molecular Cytogenetics from the University of Bari (http://www.biologia.uniba.it/rmc/) or from Invitrogen (Carlsbad, CA), and prepared as probes for FISH analysis according to standard procedures. For each probe, the expected chromosomal location at 12q was verified by FISH on normal metaphase cells from peripheral blood lymphocytes. In cases 19 and 29–32, FISH analysis was performed on tumor interphase nuclei because no metaphase cells could be obtained after short-term culture. A rearrangement of HMGA2 was detected with RP11-30I11 and RP11-118B13 in lipoma cases L2 and L3 while no rearrangement was detected in lipoma cases L4 and L5. Microscopic analysis was performed using a DM6000B microscope (Leica Microsystems, Wetzlar, Germany). FISH images were processed using the ISIS (MetaSystems, Altlussheim, Germany) or the Leica CW4000 CytoFISH (Leica Microsystems) softwares. At least 100 nuclei per slide were analyzed. For each probe, genomic amplification was defined as the presence of at least 10 fluorescent signals per cell in ≥1% of cells.
Table II. Description and Origin of the Fish Probes used in the Characterization of the Tumor Samples
QRT-PCR was used to determine the expression levels of MDM2, CDK4, HMGA2, DDIT3, DYRK2, TSPAN31 and YEATS4 in the 11 cases of WDLPS/DDLPS for which frozen material was available (cases 8, 12, 20, 22, 25, 27, 31–35). Total RNAs were extracted from WDLPS/DDLPS, lipomas samples, NSAT and NVAT samples using the RNeasy lipid tissue minikit (Qiagen, Hilden, Germany) and were treated by DNA-free™ (Applied Biosystems, Foster City, CA). One micrograms of total RNA was reverse-transcribed into cDNA using the High capacity cDNA Reverse Transcription kit (Applied Biosystems) and Q-PCR was performed in duplicate with the ABI PRISM 7000 Detection System and FAM dyes (Applied Biosystems) according to the manufacturer's protocol. RPLP0 (the large P0 subunit of the acidic ribosomal phosphoprotein) was used as endogenous control for normalization.9 QRT-PCR was performed using the following TaqMan gene expression assays (Applied Biosystems): Hs01066938_m1 (MDM2), Hs00364847_m1 (CDK4), Hs171569_m1 (HMGA2 exons1-2), Hs00971724_m1 (HMGA2 exons3-4), Hs00971725_m1 (HMGA2 exons 4-5), Hs01090850_m1 (DDIT3), Hs00389256_m1 (DYRK2), Hs00195585_m1 (TSPAN31), Hs00232423_m1 (YEATS4) and Hs99999902_m1 (RPLP0). The reaction mix consisted of 10 μl of TaqMan master mix 2X, 1 μl of TaqMan gene expression mix and 5 μl of 1/10 cDNA in a final volume of 20 μl. The PCR conditions were: 2 min at 50°C and 10 min at 95°C, followed by 40 cycles at 95°C for 15 sec and 60°C for 1 min. The comparative Ct (threshold cycle) method was used to achieve relative quantification of gene expression. The mRNA levels of genes of interest (R) were normalized to the mRNA levels of RPLP0: ΔCt = CtR−CtRPLP0. The relative amount of mRNA between controls (NSAT or NVAT) and WDLPS/DDLPS or lipomas was given by 2−ΔΔCt where ΔΔCt = ΔCtR of WDLPS/DDLPS or lipoma—mean of ΔCtR of controls (NSAT or NVAT). The ΔCt values of the controls were homogenous. No statistically significant difference between NSAT and NVAT was detected for the levels of expression of MDM2, CDK4, HMGA2, DYRK2, TSPAN31 and YEATS4. In contrast, DDIT3 expression was less abundant in NVAT than in NSAT (mean ΔCt subcutaneous tissue = 4.1 vs. mean ΔCt visceral tissue = 5.3). This difference was statistically significant (p = 0.009).
Statistical analysis of quantitative RT-PCR results
Results of QRT-PCR results are expressed as means ± Standard Error of the Mean. Statistical significance of differential gene expression between 2 study groups was determined using the nonparametric Kruskal–Wallis test with the ΔCt of each group. Correlation analyses were made using the Pearson's correlation test (perfect correlation r = 1.0). Results were considered to be statistically significant when the p value was 0.05 or less.
MDM2 (12q15) is consistently amplified and overexpressed in WDLPS/DDLPS
We detected an amplification of MDM2 in all 38 cases with the FISH probe RP11-797C20 containing MDM2 (Table III; Fig. 2). QRT-PCR analysis showed an overexpression of MDM2 in the 11 cases analyzed [mean level compared to NSAT: 40.5 ± 7.2, range: 14.8–88.0; mean level compared to NVAT: 27.8 ± 4.9, range: 10.2–60.5 (Fig. 1)]. MDM2 was significantly less expressed in the 5 lipomas in comparison with WDLPS/DDLPS (mean level compared to NSAT: 2.6 ± 0.4, p < 0.0001; mean level compared to NVAT: 1.8 ± 0.3, p < 0.0001). Altogether, our results show a consistent amplification and overexpression of MDM2 in WDLPS/DDLPS.
Table III. Amplification of the 12q13-21 region in 38 cases of well-differEntiated and dedifferentiated liposarcomas
Composition of the 12q15 amplicon in WDLPS/DDLPS: Frequent coamplification and co-overexpression of YEATS4 with MDM2
To characterize the amplification status of the region surrounding MDM2, we performed FISH analysis with three 12q15 probes. We detected amplification in 25 out of 35 cases (71%) with RP11-512G13 (centromeric to MDM2, containing CAND1), in 22 out of 35 cases (63%) with RP11-112N10 (centromeric to MDM2, containing DYRK2) and in 32 out of 37 cases (86%) with RP11-159A18 (telomeric to MDM2, containing YEATS4) (Table III; Fig. 2). The telomeric border of the MDM2 amplicon variably extended up to the 12q21 region: we detected amplification in 8 out of 35 cases (23%) with RP11-89P15 (12q21.1) and in 9 out of 35 cases (26%) with RP11-25J3 (12q21.2) (Table III). QRT-PCR analysis showed an overexpression of DYRK2 in 10 cases out of 11 WDLPS/DDLPS cases [mean level compared to NSAT: 23.6 ± 7.4, range: 4.6–78.5; mean level compared to NVAT: 30 ± 10.7, range: 4.8–107.6 (Fig. 1)]. We detected no significant overexpression of DYRK2 in the 5 lipomas (mean level compared to NSAT: 1.4 ± 0.4; mean level compared to NVAT: 2 ± 0.5). Moreover, while we found no YEATS4 overexpression in lipomas (mean level compared to NSAT: 0.9 ± 0.1; mean level compared to NVAT: 1.1 ± 0.1), we detected an overexpression in 10 out of 11 WDLPS/DDLPS cases as compared to NSAT (mean level: 19.9 ± 5.3, range: 3.0–57.3) and NVAT (mean level: 22.5 ± 6.0, range: 3.4–64.9) (Fig. 2). In all cases involving both FISH and QRT-PCR analyses (n = 10), we found a perfect correlation (Pearson correlation coefficient r = 1) between the amplification and the overexpression of YEATS4 (Fig. 2).
HMGA2 (12q14.3) is consistently amplified, rearranged and overexpressed in WDLPS/DDLPS
We investigated the HMGA2 status with 5 FISH probes located at 12q14.3 (Table II): RP11-30I11 (proximal to the 5′ end of HMGA2), RP11-299L9 (containing HMGA2 exons 1 and 2), RP11-23C9 (containing HMGA2 exons 1-3), CTC-782I9 (containing HMGA2 exons 4 and 5) and RP11-118B13 (distal to the 3′ end of HMGA2). As summarized in Table IV, we observed an amplification of HMGA2 in all cases. The pattern of amplification was often complex, with amplification of exons 1 and 2 in all cases and no amplification of exons 4 and 5 in 8 cases Table III and Fig. 2. Moreover, the 5′ (RP11-30I11) and 3′ (RP11-118B13) regions of HMGA2 were amplified only in 8 (23%) and 18 cases (51%), respectively. These results indicate that structural rearrangements within or in close vicinity of HMGA2 are associated with its amplification. We next investigated the expression status of HMGA2 exons 1-2, exons 3-4 and exons 4-5, using QRT-PCR with 3 TaqMan probes spanning exon junctions (HMGA2 1-2, HMGA2 3-4 and HMGA2 4-5, respectively). HMGA2 transcripts for the whole gene were almost undetectable in NSAT and NVAT. Results are summarized in Table IV. We detected an overexpression of the 5 exons of HMGA2 in lipomas L2 and L3 with a 12q rearrangement but no overexpression for cases L1, L4 and L5 lacking 12q aberration. In all the 11 WDLPS/DDLPS cases analyzed, we found an overexpression of HMGA2 exons 1 and 2. We detected no overexpression of HMGA2 exons 3-4 and exons 4-5 in cases 25, 27 and 33. Our results show that amplification of HMGA2 occurs at a similar frequency to that of MDM2 in WDLPS/DDLPS and is always associated with intragenic or close extragenic rearrangements.
Table IV. Expression of HMGA2 Exons 1-2, HMGA2 Exons 3-4, HMGA2 Exons 4-5 in 11 Cases of Well-Differentiated and Dedifferentiated Liposarcomas and Five Cases of Lipoma
HMGA2 (1-2) Ct
HMGA2 (3-4) Ct
HMGA2 (4-5) Ct
RPLP0 and HMGA2 mRNA levels were determined by quantitative RT-PCR and expressed as Ct (fluorescence cycle threshold). Values corresponding to an overexpression compared to normal adipose tissue are indicated in bold characters.
Normal subcutaneous adipose tissue
Normal visceral adipose tissue
CDK4 and TSPAN31 (SAS) are frequently but not always amplified and overexpressed in WDLPS/DDLPS and are the core of a distinct, recurrent but inconsistent amplicon at 12q14.1
We detected amplification of CDK4 and TSPAN31 in 33 out of 38 cases (87%) with the FISH probe RP11-571M6 (Table III). We found an overexpression of CDK4 in 9 out of 11 cases (mean level compared to NSAT: 23.0 ± 8.1, range: 2.4–88.6; mean level compared to NVAT: 22.6 ± 8.0, range: 2.4–86.8), whereas no overexpression was detected in the 5 lipomas (mean level compared to NSAT: 0.9 ± 0.2; mean level compared to NVAT: 0.9 ± 0.2). We found an overexpression of TSPAN31 in 7 out of 11 cases (mean level compared to NSAT: 20 ± 6.9, range: 2.8–68.6; mean level compared to NVAT: 26 ± 9.9, range: 3.2–95.7), whereas no overexpression was detected in the 5 lipomas (mean level compared to NSAT: 0.9 ± 0.1; mean level compared to NVAT: 0.8 ± 0.1).
Composition of the 12q14.1 amplicon in WDLPS/DDLPS: DDIT3 and GLI are rarely coamplified with CDK4 and TSPAN31
We performed FISH analysis with seven probes located in a 420 kb region spanning CDK4 and TSPAN31 (Tables II and III; Fig. 1). Three probes were centromeric to CDK4: RP11-1077C21 (containing GLI), RP11-258J5 (containing GLI and DDIT3) and CTD-2337E14. Four probes were telomeric to CDK4: CTD-2320N9, CTB-54M23, CTD-2575M15, RP11-424B7. Results are summarized in Table III. We detected no amplification with RP11-1077C21 (GLI) except for DDLPS case 28. We found amplification in 3 out of 38 cases (8%) with RP11-258J5 (GLI, DDIT3), 19 out of 38 cases (50%) with CTD-2337E14, 7 out of 35 cases (20%) with CTD-2320N9, 9 out of 35 cases (26%) with CTB-54M23 and 8 out of 35 cases (23%) with CTD-2575M15. We detected no amplification with the most telomeric probe RP11-424B7. Our results indicate that the CDK4-TSPAN31 amplicon is smaller than the MDM2 amplicon. The centromeric border of the CDK4-TSPAN31 amplicon was most often located in a narrow region distant from 18 kb (16 cases) up to 86 kb (14 cases) of the 5′ telomeric end of DDTI3 (Fig. 1). In less than 10% of cases, the CDK4-TSPAN31 amplicon extended up to DDIT3 and GLI. In contrast, the telomeric border of the CDK4 amplicon was located in a variably broad region, distant from 65 kb (21 cases) up to 2.6 Mb (8 cases) downstream of the 5′end of CDK4-TSPAN31 (Table III).
Different levels of expression of DDIT3 between NSAT and NVAT: Implication for the DDIT3 expression pattern in WDLPS/DDLPS
Using NSAT as control, we found no overexpression of DDIT3 (fold change < 2) in 8 WDLPS/DDLPS cases whereas we detected a moderate 2.5-fold increase in 1 case and a very strong level of overexpression (91.7-fold and 29.0-fold, respectively) in 2 cases with amplification of DDIT3 (Fig. 2). We found a lower abundance of DDIT3 expression in NVAT than in NSAT. Therefore, using NVAT as control, we showed an overexpression of DDIT3 (fold change > 2) in 9 out of 11 cases (Fig. 2). The overexpression was missing in the 2 cases showing no amplification of CDK4. DDTI3 was not overexpressed in the 5 lipomas. We found a perfect correlation between the presence of overexpression of DDIT3 (classified as < or ≥ 2-fold compared to NVAT) and CDK4 amplification (Pearson correlation coefficient: r = 1.0).
In our study, we have characterized the genomic structure of the 12q13-15 region amplified in WDLPS/DDLPS.
Previous studies have suggested a lower frequency of amplification of CDK4 compared to MDM2.1, 10, 11 We demonstrated here that MDM2 and CDK4 belong to 2 distinct amplicons. First, none of our 38 cases showed continuity of amplified sequences between MDM2 and CDK4. We consistently found a nonamplified region of at least 5 Mb separating the 2 genes. Second, although MDM2 was consistently amplified, CDK4 and neighbor genes were not amplified in 13% of cases. Therefore, we conclude that the amplification of CDK4 is not as indispensable as the amplification of MDM2 in WDLPS/DDLPS. However, since overexpression of CDK4 may account for an increase in cellular proliferation by enhancing the G1-S transition,12 further investigations will be necessary to determine whether the absence of CDK4 amplification is correlated to a specific clinico-histopathological profile or compensated by another genomic anomaly. When amplified, CDK4 was always coamplified but not always co-overexpressed with TSPAN31, located at 3 kb from CDK4. The CDK4-TSPAN31 amplicon is smaller in size and has less variable borders than the MDM2 amplicon. In particular, its centromeric limit is located directly downstream of the 5′end of DDIT3. This indicates the presence of sequences prone to breakage in WDLPS/DDLPS in the vicinity of 5′DDIT3. Several studies have demonstrated the initiation of genomic amplification by a DNA double-strand break across chromosomal fragile sites.13 Our results suggest that the 5′DDIT3 region might act as a “fragile region” the breakage of which might be at the origin of the CDK4-TSPAN31 amplicon formation and DDIT3 dysregulation. Belonging to the CCAAT/enhancer binding protein family of transcription factors,14 DDIT3 (C/EBP-zeta) regulates adipocyte differentiation and is known for its role in MLS t(12;16) or t(12;22) translocations.2, 15DDIT3 is expressed at very low levels under normal conditions. We have shown here that DDIT3 has higher expression levels in NSAT than in NVAT. Such differences have also been reported for other C/EBP family members.16 We found an amplification of DDIT3 in 8% of WDLPS/DDLPS cases. Interestingly, we found an overexpression of DDIT3, independently of its own amplification status, only in tumors with CDK4 amplification. This suggests that DDIT3 overexpression in CDK4-amplified WDLPS/DDLPS might be related to the breakages in the DDIT3 5′ region. Our results indicate that the role of DDIT3 in adipose tumors is not restricted to MLS and could also be important in WDLPS/DDLPS. This may have clinical implications since an association of DDIT3 overexpression in several tumor types with sensitivity of these tumors to cytotoxic agents has been suggested.17–19 MLS shows significantly higher response rate to chemotherapy compared to other liposarcomas.20 The overexpression of DDIT3 might therefore be considered in future studies as a potential marker for selecting WDLPS/DDLPS patients who could benefit from chemotherapy.
Rearrangement and overexpression of HMGA2 are observed in lipomas and in other benign tumors such as uterine leiomyomas and salivary gland pleomorphic adenomas.21–23 In lipomas, the preferential clustering of HMGA2 breakpoints in the third intron23 results in the replacement of the acidic C-terminal domain by a variety of ectopic sequences. In addition, extragenic breakpoints located 5′ or 3′ to HMGA2 similar to the rearrangements observed in uterine leiomyomas also occur in lipomas.24, 25 Some reports based on small series have shown the amplification and overexpression of HMGA2 in a subset of WDLPS/DDLPS.1, 26–28 Here we have demonstrated the amplification of exons 1-2 of HMGA2 in all the 38 WDLPS/DDLPS cases studied, against the nonconsistent amplification of exons 3-5. Some authors have proposed that the truncation of HMGA2 followed or not by the addition of ectopic sequences might be the critical steps in HMGA2 oncogenicity.29, 30 However, the mechanism of HMGA2 activation in tumors might be more complex than originally expected. Indeed Zaidi et al.31 reported no difference in the phenotype of benign tumors developed in transgenic mice expressing either a full-length or a truncated HMGA2 transcript under the control of a differentiated adipocyte-specific promoter. Moreover, some tumors with 12q rearrangements retain the full coding region of HMGA2, suggesting a nondependence of the HMGA2 oncogenic potential on the nature of the transcript.24, 25, 32, 33 In addition, HMGA2 overexpression could also result from the disruption of the 3′UTR causing the loss of function of let-7, a microRNA specifically involved in the posttranscriptional repression of HMGA2.34, 35 Our results identifying variability in the dysregulation pattern of HMGA2 in WDLPS/DDLPS, strongly suggest that a misexpression of HMGA2, whatever the mechanism, might be sufficient for tumorigenesis. Notably, the observation that the tumors developed in HMGA2 transgenic mice were always benign31, 36, 37 raises the question of the role of HMGA2 in malignancy. We can hypothesize that in adipose tissue tumors, the dysregulation of HMGA2 may be sufficient to lead to only benign lesions. In contrast, the dysregulation of HMGA2 might cause a malignant phenotype when coupled with MDM2 amplification, as suggested by a study showing that the expression of MDM2 or CDK4 allowed human diploid fibroblast to by-pass HMGA2-induced proliferation arrest.38 Although CDK4 certainly plays an important role in a majority of cases, our results favor MDM2-HMGA2 instead of MDM2-CDK4 as the crucial couple of genes in WDLPS/DDLPS pathogenesis. However, cooperation between HMGA2 and CDK4, when amplified, may also play an important role.
We have also shown that DYRK2 was frequently amplified and overexpressed in WDLPS/DDLPS. This gene is amplified and/or overexpressed in gliomas, gastrointestinal stromal tumor, esophageal cancer and nonsmall-cell lung cancer.39, 40 Interestingly, DYRK2 encodes a protein which plays an important role in the p53 function although its potential interaction with MDM2 remains to be elucidated.41, 42 In addition, we identified YEATS4 (GAS41) as a novel oncogene important in WDLPS/DDLPS pathogenesis. YEATS4 is a transcription factor amplified in human gliomas.4, 5 Shown to inhibit p53 transactivation activity and decrease the p53 proapoptotic response,6 the inhibition of YEATS4 results in p53 stabilization and activation.7 The functional impairment of the p53 pathway plays a crucial role in WDLPS/DDLPS oncogenesis. Indeed several studies have demonstrated that the MDM2 antagonist, Nutlin 3A, induced cell cycle arrest and apoptosis in WDLPS or DDLPS cell lines.43, 44 Moreover, it caused tumor shrinkage in MDM2 amplified sarcoma xenografts in nude mice.45 Our findings that showed the frequent amplification and overexpression of YEATS4 in WDLPS/DDLPS make this gene a potential new target for the development of therapeutic strategies activating the p53 pathway.
The authors thank Dr. M. Rocchi (University of Bari, Italy) for giving some of the BAC clones used in this study.