Liposarcomas are the most common subtype of soft tissue sarcomas, comprising approximately 20% of adult-onset sarcomas 1. Well-differentiated liposarcoma (WDLPS), also known as atypical lipomatous tumour (ALT), is a slowly growing tumour that exhibits adipocytic differentiation and most often arises in deep soft tissue of the limbs or retroperitoneum. WDLPS at any site can undergo a type of malignant progression termed dedifferentiation. Dedifferentiated liposarcoma (DDLPS) is made up of spindled or pleomorphic cells that proliferate rapidly and fail to undergo adipocytic differentiation. Most DDLPSs have a residual well-differentiated (WD) component upon resection of the primary tumour. WDLPS is incapable of metastasis, but dedifferentiation leads to more aggressive local growth and enables the tumour to metastasize to distant sites.
The mechanisms responsible for progression from WDLPS to DDLPS are incompletely understood. WDLPSs virtually always contain ring or long marker chromosomes which include substantially amplified material from chromosome 12q 2, including the proto-oncogenes MDM2, HMGA2, and CDK4. Amplification of these genes is thought to be an early and essential part of the oncogenic programme of WDLPS. Dedifferentiation of WDLPS likely occurs via multiple alternative genetic alterations. In some DDLPSs, the c-Jun pathway is up-regulated by amplification of the c-Jun proto-oncogene or possibly by amplification of a locus containing ASK1, a MAP3K upstream of c-Jun 3, 4. ASK1 can trigger a phosphorylation cascade that activates the kinase JNK, which then phosphorylates and stabilizes c-Jun. In an immortalized murine cell line, c-Jun overexpression was also shown to block adipogenesis. c-Jun can modulate the activity of transcription factors that regulate adipogenesis, such as C/EBPβ and PPARγ 4, 5. It has been proposed that high c-Jun levels may directly block adipocytic differentiation in liposarcomas, leading to a dedifferentiated phenotype. A c-Jun deoxyribozyme was also recently shown to induce apoptosis in SW872 liposarcoma cells in vitro and inhibit their ability to form tumours in mice 6. We analysed the role of c-Jun in liposarcomas by evaluating a series of liposarcomas for c-Jun amplification and for c-Jun and ASK1 protein levels. We also derived cell lines from DDLPSs with c-Jun amplification and tested the effects of c-Jun knockdown by RNA interference.
Materials and methods
Primary tumour samples
Formalin-fixed, paraffin-embedded (FFPE) tumours were obtained from the Department of Pathology at Brigham and Women's Hospital. Liposarcomas arose in multiple sites, including the retroperitoneum and limbs, and were diagnosed by standard accepted pathological criteria 1, 7. Fresh tumour samples were obtained immediately after surgical resection for the generation of cell lines. Studies were performed in accordance with protocols approved by the Partners Health Care Institutional Review Board.
Antigen retrieval was performed on 4 µm thick FFPE sections. Primary antibodies were incubated for 1 h at room temperature or overnight at 4 °C. Detailed methods and antibodies used may be found in the Supporting information, Supplementary material.
Scoring of immunostains
Tumours were manually scored as positive for c-Jun and other proteins if at least 10% of sarcoma nuclei were positive by visual examination. To obtain a quantitative score, whole sections were scanned and analysed with the Ariol instrument SL-50 (Applied Imaging, Grand Rapids, MI, USA). Detailed scoring methods are given in the Supporting information, Supplementary material.
Fluorescence in situ hybridization (FISH)
FISH was performed on nuclei isolated from 50 µm sections of FFPE tumour containing dedifferentiated or well-differentiated areas. Slides were examined on a Zeiss Axioskop 2+ microscope and images were analysed with Cytovision software from Applied Imaging (San Jose, CA). The reader is referred to the Supporting information, Supplementary material for details of the probes used.
The LPS141, LPS695, and LP6 cell lines were derived from human DDLPS (see the Supporting information, Supplementary material for details of culture conditions).
Lentivirus encoding shRNA to GFP or c-Jun [RNAi Consortium (TRC)] was produced in 293T cells. Details of viral production may be found in the Supporting information, Supplementary material.
Protein lysates were separated by SDS-PAGE and transferred to Immobilon-P membranes (Millipore, Billerica, MA, USA). Immunoblots were probed with antibodies against c-Jun (Cell Signaling, Danvers, MA, USA) or actin (Millipore). Details are given in the Supporting information, Supplementary material.
LP6 cells expressing shRNA to c-Jun or GFP were injected subcutaneously into nude female mice (Nu/Nu, Charles River) and allowed to grow for 2 weeks before analysis. All procedures were performed according to protocols approved by the Institutional Animal Care and Use Committee of the Dana Farber Cancer Institute. Details are given in the Supporting information, Supplementary material.
c-Jun expression in liposarcomas varies by subtype
We evaluated the expression of c-Jun by immunohistochemistry in whole sections of 81 liposarcomas, including 35 dedifferentiated liposarcomas (DDLPSs), 22 well-differentiated liposarcomas/atypical lipomatous tumours (WDLPSs/ALTs), 11 cases of the inflammatory variant of WDLPS, and 13 myxoid liposarcomas, a genetically distinct type of LPS in which the c-Jun expression pattern has not previously been reported. Twenty-seven of the DDLPS cases also contained a well-differentiated (WD) component in which c-Jun expression was evaluated.
Ninety-one per cent of DDLPSs and 59% of the WD components of DDLPS were c-Jun-positive (Figures 1A, 1B, and Table 1). The WD and DD components of a given tumour often exhibited a similar staining intensity (Figure 2B). In contrast, only 27% of non-inflammatory WDLPSs/ALTs were positive for c-Jun by manual scoring (Figures 1A, 1B, and Table 1). Inflammatory WDLPS also exhibited strong c-Jun expression in some cases and overall 64% were scored positive. Only 8% (1/13) of myxoid liposarcomas (including both low-grade and high-grade/‘round cell’ types) were positive for c-Jun (Figures 1A, 1B, and Table 1), suggesting that this genetically distinct type of liposarcoma may not rely on the c-Jun pathway. Benign lipomas (n = 6) were negative for c-Jun expression, as was a high-grade sarcoma with a deletion in the region of chromosome 1 that contains c-Jun (data not shown).
Table 1. Manual scoring of immunostains for total c-Jun, ASK1, phospho-JNK, and phospho-c-Jun in subtypes of liposarcoma. Cases were scored as positive if at least 10% of tumour cells exhibited staining with the indicated antibody. Per cent positivity and total number of cases assessed are reported. A small number of well-differentiated tumours could not be evaluated with all four antibodies for technical reasons
ND = not determined.
WD component of DDLPS
The proportion of c-Jun-positive cells and the intensity of staining varied within tumours. To obtain a more quantitative assessment of c-Jun expression, we used the Ariol slide scanner, which can measure both the intensity and the percentage of positive nuclei and generate a composite score reflecting both parameters. As expected, the composite c-Jun score was significantly lower in pure WD/ALT and myxoid LP tumours than in DDLPS tumours (p = 0.04 and 2 × 10−4, respectively, Figure 1B). WD components of DDLPS had a composite c-Jun score that was lower than DDLPS, but higher than pure WD/ALT (Figure 1B), although these differences did not reach significance. Importantly, the intensity scores of nuclei from DDLPSs and their WD components were virtually identical (data not shown). Inflammatory WDLPS tumours could not be scored accurately by the Ariol software because of the dense lymphoplasmacytic infiltrate, which often outnumbered the neoplastic cell population.
These results are consistent with the previously reported expression of c-Jun in DDLPS 4. However, the relatively high levels of c-Jun in WD components of DDLPS raise the possibility that c-Jun expression may not always be sufficient to inhibit the adipocytic differentiation programme of tumour cells.
c-Jun amplification in liposarcomas
We performed FISH for the c-Jun locus on 33 primary DDLPS tumours and two DDLPS cell lines. In most cases, haematoxylin and eosin (H&E) sections were used to identify dedifferentiated areas of the tumour, and nuclei were extracted from the corresponding portion of the tissue block to ensure that FISH was performed on the DD component, and not the WD component. We detected high-level c-Jun amplification in 6/35 cases (17%), including five primary tumours and the LPS695 cell line (Figure 2A). We also identified one DDLPS that exhibited extensive aneuploidy and copy number gains in c-Jun relative to the centromere 1 probe. Normal copy number of c-Jun was observed in the remaining DDLPSs (28/35). We also performed FISH on a subset of pure WDLPSs (including inflammatory variants) that were positive for c-Jun by immunohistochemistry (n = 6). However, c-Jun copy number was normal in these cases.
In two of the DDLPSs with c-Jun amplification, enough material was present to perform FISH on the WD component of the tumour. In one of the WD components, we found high-level c-Jun amplification that was indistinguishable from the amplification found in the DD component (Figure 2B). Both the WD and the DD components of this tumour (which were not co-mingled) were positive for c-Jun by immunohistochemistry (Figure 2B). Importantly, a previous report found gain of the 1p32 region (containing c-Jun) in the WD component of another DDLPS, although the DD component exhibited an even greater increase in copy number 8.
These data show that c-Jun amplification and expression can be found in well-differentiated liposarcoma cells, suggesting that c-Jun amplification may occur before dedifferentiation. The earliest characterized event in WDLPS development is the amplification of MDM2 in ring and marker chromosomes. We therefore performed dual FISH for MDM2 and c-Jun in metaphase spreads and interphase nuclei from five DDLPSs with known c-Jun amplification, including four primary tumours and the LPS695 cell line. Strikingly, we detected co-amplification of these two genes in all five cases (Figures 2C and 2D). Evaluation of metaphase spreads revealed that multiple copies of the two amplified genes were interspersed along a substantial portion of the marker chromosomes (Figure 2C). This alternating pattern of gene amplification suggests that amplification of the two genes occurred at the same time and would not be expected if c-Jun had been amplified substantially later than MDM2 in the tumour's evolution. The same alternating pattern of amplification has been described with probes to MDM2 and CDK4 in WDLPS 9.
Other mechanisms of c-Jun activation in liposarcomas
Amplification of the MAP3K5 gene (encoding the kinase ASK1) may be an alternative mechanism for activating c-Jun in non-amplified tumours 3. We therefore performed immunohistochemistry for ASK1 on our series of DDLPSs and WDLPSs and correlated the results with c-Jun protein levels.
ASK1 expression was identified in 54% of DDLPSs, 19% of WD components of DDLPSs, and 9.5% of pure WD/ALT evaluated (Table 1). DDLPS cases could be further subclassified as ASK1-diffuse (nearly every tumour cell positive) or -focal (one or more areas of positive cells in a background of negative tumour cells) (Figure 3A). Seventeen per cent of DDLPS (6/35) cases exhibited diffuse ASK1 staining, whereas the WD tumours exhibited at most focal positivity in a minority of cases (19% of WD components and 9.5% of pure WDLPSs, Table 1). ASK1 expression was either focal or negative in DDLPSs with c-Jun amplification.
If ASK1 activates a kinase cascade that up-regulates c-Jun protein levels, c-Jun should be up-regulated in ASK1-diffuse tumours. We classified DDLPS samples into four groups: c-Jun-amplified, ASK1-diffuse, ASK1-focal, and ASK1-negative. We then compared levels of total c-Jun protein between the groups using the composite scores derived from the Ariol. We found that c-Jun-amplified and ASK1-diffuse tumours had equally high composite c-Jun scores (Figure 3B). This finding is consistent with the hypothesis that ASK1 kinase activity might substitute for amplification of c-Jun. Composite c-Jun scores were lower in ASK1-focal DDLPSs, although the difference did not reach significance. Composite c-Jun scores were significantly lower in ASK1-negative DDLPSs than in c-Jun-amplified tumours (p = 0.02, Figure 3B).
If kinase cascades driven by ASK1 and other MAP3Ks are active in DDLPS, we predicted that phosphorylated forms of JNK and c-Jun would be detectable. Indeed, the majority of DDLPSs were positive for p-JNK and p-c-Jun (51% and 69%, respectively), whereas these phosphoproteins were detected in only a minority of WD components and pure WD/ALT (Figure 3A and Table 1).
c-Jun knockdown inhibits the tumourigenicity of DDLPS cells without affecting differentiation
To study the function of c-Jun amplification in DDLPS, we derived a primary cell line (designated LP6) from a retroperitoneal DDLPS with c-Jun amplification. We confirmed by FISH that LP6 cells retained marker chromosomes with c-Jun amplification (data not shown). We found that c-Jun protein levels were much higher in these cells than in the DDLPS cell line LPS141, which lacks c-Jun amplification (Figure 4A, left). To modulate c-Jun levels, we used lentivirus to deliver c-Jun shRNAs, two of which significantly reduced c-Jun levels in both cell lines (Figure 4A, left). We found that viable LP6 cell number was significantly reduced by c-Jun shRNA compared with GFP shRNA (p < 0.01, Figure 4A, right). We also found that proliferation of LP6 cells stably expressing c-Jun shRNA was significantly diminished compared with cells expressing shGFP (p < 1 × 10−4; Supporting information, Supplementary Figure 1). Long-term observation (>3 weeks) of stable c-Jun knockdown cells demonstrated that many of them became flatter than controls and survived for extended periods of time without proliferating. Evaluation of lipid droplets with Nile Red dye at multiple time points revealed no increase in intracellular lipid content in c-Jun knockdown cells compared with controls (data not shown).
To determine whether c-Jun was required for in vivo tumour formation, LP6 cells expressing shRNA to GFP or c-Jun were injected subcutaneously into nude mice. LP6-shGFP cells formed rapidly growing tumours that were histologically similar to the high-grade DDLPSs from which they were derived (Figure 4B, left). LP6-shJun cells formed slowly growing tumours that were significantly smaller than LP6-shGFP tumours (168 mm3 versus 32 and 28 mm3, p < 0.01, Figure 4B, right). LP6-shJun tumours consisted of variable numbers of pleomorphic, high-grade sarcoma cells in a background of murine fibroblasts, leukocytes, and blood vessels in residual matrigel. Despite their lower density, the individual shJun tumour cells resembled their shGFP counterparts cytologically. There was no histological evidence of adipocytic differentiation in the shJun tumours.
c-Jun protein was detectable in a variable but substantial number of cells in LP6-shJun tumours by immunohistochemistry (data not shown). It is likely that the c-Jun-positive cells in these tumours either escaped infection and selection in vitro or down-regulated shRNA expression in vivo. Nevertheless, the overall Ki67 rate was significantly higher in LP6-shGFP than in LP6-shJun tumours (51% versus 27% and 31%, p = 0.002 and 0.03, Figure 4C). These results show that c-Jun knockdown inhibits proliferation in vivo as well as in vitro, and suggest that there is a strong selection for retention of c-Jun expression in these liposarcoma cells in vivo.
We have confirmed previous observations that the c-Jun proto-oncogene is amplified and overexpressed in dedifferentiated liposarcomas (DDLPSs), but not pure well-differentiated liposarcomas (WDLPSs) 4. We have also found that c-Jun is rarely expressed in myxoid liposarcomas (regardless of grade), suggesting that c-Jun may not play a role in the oncogenesis of this genetically distinct type of liposarcoma. In DDLPS lacking c-Jun amplification, the c-Jun pathway may be activated by other mechanisms 3, because ASK1, phospho-JNK, and phospho-c-Jun are readily detected in a subset of these tumours. The derivation of new DDLPS cell lines from primary tumours with genomic amplification of ASK1 will be required to test this hypothesis directly.
Our study provides several lines of evidence that adipocytic differentiation in liposarcomas is not always inhibited by c-Jun. First, c-Jun protein levels are often similarly high in both the WD and the DD components of DDLPS (Figures 1A, 1B, and 2B). Second, we identified a liposarcoma in which both the DD and the WD components exhibited high-level c-Jun amplification (Figure 2B). Copy number gain of the 1p32 locus, which contains c-Jun, has also been reported in one WD component of DDLPS, although the DD component contained an even greater copy number gain of 1p32 8. Third, the co-localization of c-Jun and MDM2 on ring and long marker chromosomes (Figures 2C and 2D) suggests that c-Jun is often incorporated into the ring chromosomes and amplified with MDM2 prior to genetic progression from WDLPS to DDLPS. Finally, the most prominent effect of c-Jun knockdown on liposarcoma cells is an inhibition of proliferation and tumour formation rather than an induction of adipogenesis (Figure 4). Taken together, these results show that c-Jun overexpression in human liposarcomas is not always sufficient to block adipocytic differentiation and suggest that other genetic or epigenetic events may be responsible for the transition from WDLPS to DDLPS.
In contrast to the WD components of DDLPS, c-Jun amplification and overerexpression have never been observed in pure WDLPS. In addition to the six WDLPSs with normal c-Jun copy number in our study, at least 92 pure WDLPSs have been previously studied by comparative genomic hybridization (traditional or array-based) 4, 8–18. In these studies, high-level amplification of the c-Jun locus at 1p32 was never observed; only one case of low-level copy number gain involving 1p31 to 1p36 was reported 13.
We propose the following model to explain these observations (Figure 4D). We hypothesize that in some WDLPSs, only the minimum number of oncogenes required for tumour initiation is amplified or otherwise activated during ring chromosome formation. These tumours may grow and evolve slowly, often presenting as pure WDLPS but still progressing to DDLPS at some fixed rate. In contrast, other WDLPSs may stochastically incorporate additional oncogenes, such as c-Jun, into ring chromosomes during tumour initiation. c-Jun would provide an additional oncogenic stimulus, causing the tumour to proliferate and progress to a dedifferentiated state so rapidly that DDLPS elements would usually be present at the time of diagnosis. Such a model would be in accordance with the observation that all of the c-Jun-amplified tumours in our study and a prior study 4 presented with dedifferentiated components. This model is a specific example of the hypothesis that the malignant potential of a tumour can be determined by its earliest genetic alterations 19.
Additional studies, such as global mRNA profiling of c-Jun knockdown cells, will be needed to identify the key effectors of c-Jun in liposarcoma. Several targets of c-Jun regulate cell proliferation and survival, including cyclin D1, p16, p21, and p53 20. Our data do not preclude the possibility that c-Jun amplification can cooperate with other genetic changes to promote dedifferentiation. Indeed, c-Jun knockdown may lead to up-regulation of a subset of genes that promote adipogenesis but are insufficient to induce lipid accumulation under our experimental conditions.
Our data suggest that inhibiting the c-Jun pathway is likely to be a potent therapeutic strategy in DDLPS. Further investigations are warranted to determine whether c-Jun pathway inhibitors can synergize with inhibitors of other oncogenes (such as MDM2 and HMGA2) that are universally amplified in DDLPS.
We thank Jon Aster (Brigham and Women's Hospital) for the c-Jun probe and Andre Oliveira (Mayo Clinic) for the MDM2 probe. We thank Samuel Moss for assistance with mouse experiments. This work was supported by the Center for Molecular Oncologic Pathology of the Dana Farber Cancer Institute and Brigham and Women's Hospital.
Supporting information may be found in the online version of this article.