Prognostic Impact of P53 Status, TLS-CHOP Fusion Transcript Structure, and Histological Grade in Myxoid Liposarcoma: A Molecular and Clinicopathologic Study of 82 Cases¹

MLS,³ the most common subtype of liposarcoma, occurs predominantly in the extremities of adults and has a tendency either to recur locally or to metastasize to unusual soft tissue locations (1, 2, 3). A proportion of cases show histological progression to RC morphology, a feature significantly associated with a poor prognosis (2). The karyotypic hallmark of MLS and RC liposarcomas is the t(12;16)(q13;p11), which is specific for this tumor type and is present cytogenetically in >95% of cases (4). The translocation leads to the fusion of the CHOP and TLS genes at 12q13 and 16p11, respectively, and the generation of a TLS-CHOP hybrid protein (5, 6, 7). The TLS gene is also known as FUS (6). In rare cases of MLS, a variant chromosomal translocation has been described, t(12;22)(q13;q12), resulting in an EWS-CHOP fusion gene (3, 8, 9).

The CHOP gene is a member of CHOP family of leucine zipper transcription factors, implicated in adipocyte differentiation and growth arrest (10, 11, 12). The CHOP gene is normally expressed at very low levels in most cells, including adipocytes; however, it is markedly activated by perturbations that induce cellular stress. In MLS, the hybrid TLS-CHOP gene encodes a protein that consists of the 5′ portion of TLS fused to the entire coding region of CHOP. The TLS-CHOP protein is thought to function primarily as an aberrant transcriptional regulator that interferes with adipocyte differentiation, favoring proliferation over terminal differentiation.

TLS-CHOP fusion transcripts occur as different recurrent structural variants reflecting the inclusion of different portions of TLS, caused by different genomic breakpoints (13). Of the possible TLS genomic breakpoints, only breaks in TLS introns 5, 7, and 8 give rise to in-frame fusion transcripts joining TLS exons 5, 7, and 8, respectively, to exon 2 of CHOP. Thus, three major recurrent fusion transcript types have been reported in cases of MLS: type 7-2 (also known as type I); type 5-2 (also known as type II); and type 8-2 (also known as type III). Exons 1–5 of TLS are thus invariably present in all TLS-CHOP fusions.

On the basis of experience with other translocation-associated sarcomas, such as Ewing sarcoma, alveolar rhabdomyosarcoma, or synovial sarcoma, in which the genetic variability of the fusion structure impacts on survival (see “Discussion”), we sought to evaluate the potential impact of TLS-CHOP fusion transcript structure on clinical outcome in MLS. We also analyzed P53 status, because P53 alteration has been found to be a strong negative prognostic factor in other sarcomas with chromosomal translocations, for instance Ewing sarcoma and synovial sarcoma (14, 15). The prognostic value of TLS-CHOP fusion transcript structure and P53 status was also compared with established conventional clinical and pathological prognostic factors in MLS.

This study included a total of 82 patients with MLS from three institutions, including MSKCC (52 patients); Cleveland Clinic Foundation (19 patients), and the UNMC (11 patients). The retrospective collection of frozen tumor samples included patients treated at the above institutions over a 17-year period (1983–2000). All tumors had a typical histological appearance of MLS. Among the 82 patients, there were 50 (61%) males and 32 (39%) females. Mean age at diagnosis was 45.8 years, and ages ranged from 13 to 79 years of age. The tumors were frequently located within the lower extremity [67 (82%)], with a strong predilection for the thigh [57 (70%)]. The remaining tumors were located in the upper extremity, 1 (1%) case; trunk, 10 (12%) cases; and retroperitoneum/pelvis, 4 (5%) cases. Most of the tumors were large, measuring ≥10 cm in 60 (73%) cases. Twenty (25%) tumors were ≥5 cm but <10 cm; only 2 (2%) cases were <5 cm. The majority of tumors were localized at presentation (89%), with only 9 (11%) patients presenting with distant metastases at the time of diagnosis (Table 1).

In all cases, microscopic slides from the en-bloc resection specimen of the primary tumor were available for pathological review, and our standard practice was to submit one block/per cm³ of tumor for histological examination in each case. The tumors were scored for the percentage of RC component or transitional areas (increased cellularity; Ref. 16). The percentage of RC component was estimated by scanning all individual sections, using the entire tumor volume as a denominator. The cases were divided into low and high histological grades using, in parallel, two previously described cutoff points, first with ≥5% (17) RC component and the second with ≥25% RC component (2), to assign low and high grades. The transitional areas were defined as areas of increased cellularity where the tumor cells were not closely packed and which retained a small amount of intercellular myxoid stroma and a discernible plexiform vascular pattern. In cases with <5% RC component, the degree of increased cellularity was also assessed, because previous reports suggested a possible prognostic role (16). Therefore, these cases were divided into two groups: (a) pure MLS cases, lacking RC component or increased cellularity; and (b) MLS with areas of increased cellularity (≥5%). The degree of necrosis was also assessed, arbitrarily using ≥ 5% microscopic areas of necrosis as a threshold for positive cases.

All primary tumors were managed by surgical resection with a curative intent. In 46 (56%) cases, the tumors were entirely resected with negative microscopic margins; in 14 (17%) cases, the inked surgical margins were free but close (<1 mm); in the remaining 22 (27%) cases, there were positive microscopic margins. Adjuvant chemotherapy and/or radiotherapy were given based on evaluation of prognostic factors in each individual case and were thus not randomized. Therefore, adjuvant therapy was not included in the analyses performed.

Tumor samples for molecular analysis were snap-frozen in liquid nitrogen and stored at −70°C. The tumor tissues available for molecular analysis were obtained from primary tumor sites in 58 (71%) cases, from a local recurrence in 14 (17%) cases, or from a distant metastatic site in 10 (12%) cases. All cases included for analysis were previously confirmed to harbor CHOP rearrangement by Southern blotting (87% of cases) at MSKCC or by conventional cytogenetic demonstration (at UNMC) of a t(12;16)/t(12;22) (13% of cases), using standard protocols as described previously (18). Forty cases were included in a previous study of CHOP and TLS genomic rearrangements in MLS (19).

Extraction of total RNA was based on the guanidinium isothiocyanate-phenol chloroform method using the RNA Wiz reagent (Ambion, Inc., Austin, TX). Analysis by RT-PCR for TLS-CHOP and if indicated, for EWS-CHOP transcripts, was performed on all 82 cases at one of the centers (MSKCC). The adequacy of the extracted RNA was assessed by RT-PCR, using primers for phosphoglycerate kinase gene transcripts. Negative controls that lacked either tumor RNA or reverse transcriptase were used routinely. Three μg of total RNA were subjected to RT-PCR using Qiagen 1 Step RT-PCR kit (Qiagen, Inc., Valencia, CA), using a forward primer within exon 5 of TLS (5′-CAG CCA GCA GCC TAG CTA TG-3′) or in exon 7 of EWS (5′-CTG GAT CCT ACA GCC AAG CTC CAA G-3′) and a reverse primer in exon 3 of CHOP (5′-TGT CCC GAA GGA GAA AGG CAA TG-3′). The Qiagen 1 Step RT-PCR conditions included: (1) the reverse transcription step: 50°C for 30 min; followed by (2) 95°C for 15 min; (3) the PCR cycles of 95°C for 45 s, 64°C for 45 s, 72°C for 1 min (35 cycles); and (4) 72°C for 7 min. The RT-PCR products were identified by agarose gel electrophoresis. The expected sizes were 526 bp for type 7-2, 250 bp for type 5-2, and 625 bp for type 8-2 TLS-CHOP fusion transcripts. All non-type 5-2 transcripts and randomly selected type 5-2 transcripts were confirmed by direct automated sequencing.

To independently confirm the RT-PCR results, LR PCR was performed in 61 (74%) cases with adequate/available DNA. DNA was isolated from snap-frozen tissue using a standard organic extraction protocol. One μg of DNA was amplified using the Expand System LR-PCR kit (Roche Molecular Biochemicals, Indianapolis, IN), per package instructions. Three sets of primers were used: one set each for the detection of either TLS-CHOP or CHOP-TLS genomic junction fragments in cases with type 5-2 (5-2) transcripts, and one set for the detection of TLS-CHOP genomic junction fragments in both type 7-2 (7-2) and type 8-2 (8-2) cases. The primers for TLS-CHOP in cases with type 5-2 transcripts were: LR TLS-5 forward (5′-GTG GAG GTG GAG GTG GAG-3′), designed close to the 3′ end of exon 5 of TLS (nucleotides 584–601, GenBank accession no. S62138) and LR CHOP-2 reverse (5′-CAT CTG CTT TCA GGT GTG GTG-3′), ∼30 bp from the 5′-end of exon 2 CHOP (nucleotides 908–928; GenBank accession no. S62138). The expected product size range was 60–3250 bp, and the annealing temperature used was 60°C. The primers for reciprocal CHOP-TLS junction fragment used in cases with the type 5-2 transcript were: LR CHOP-1 forward (5′-GCA GCG ACA GAG CCA AAA TCA GAG C-3′) and LR TLS-6 reverse (5′-TCC ACC ACT CTG GTC TTG ATT GC-3′; position 662–684 in TLS). The expected product size range was 850-3220 bp with an annealing temperature of 62°C (20). The primer set used to amplify TLS-CHOP genomic junction fragments in cases with type 7-2 (7-2 fusion) or type 8-2 (8-2 fusion) transcript was: LR TLS-7 forward (5′-CCG TGG TGG CTT CAA TAA A-3′), positioned at the 3′-end of exon 7 (position 852–870) TLS and the same LR CHOP-2 reverse primer, in CHOP exon 2, with an annealing temperature of 58°C. The expected size range for products was 949-4046 bp (for the type 8-2 fusion) and 80–3504 bp (for the type 7-2 fusion).

In 71 (87%) cases, paraffin-embedded representative tumor tissue was available for P53 immunohistochemical analysis. The single tumor tissue paraffin block used for the immunohistochemical analysis was selected to contain the most cellular area or the RC component from each case. A mouse monoclonal antibody, clone DO7 (DAKO, Carpinteria, CA; 1:500; 0.2 μg/ml) was used for detection of a P53 epitope located between amino acids 19 and 26 of wild-type and mutant human P53 proteins. Deparaffinized sections were treated with 3% H₂O₂ to block endogenous peroxidase activity. Sections were subsequently immersed in boiling 0.01% citric acid (pH 6.0) in a microwave oven for 15 min to enhance antigen retrieval, allowed to cool, and incubated with 10% normal horse serum (mouse monoclonal antibodies) to block nonspecific tissue immunoreactivities. Primary antibodies were then incubated overnight at 4°C. Biotinylated horse antimouse IgG antibodies (Vector Laboratories, Burlingame, CA; 1:500 dilution) were applied for 1 h, followed by avidin-biotin-peroxidase complexes that were applied for 30 min (Vector Laboratories; 1:25 dilution). Diaminobenzidine was used as the final chromogen, and hematoxylin was used as the nuclear counterstain. Nuclear immunoreactivities were scored on a continuous scale with values that ranged from undetectable levels or 0% to homogeneous staining or 100%. The staining profile in tumor cells was compared with the negative internal tissues, as well as with both negative and positive control samples. On the basis of other reports (15, 21) a cutoff point of ≥10% nuclear staining was used to designate positive cases. This type of aberrant P53 staining in sarcomas correlates well with the presence of P53 point mutations.

All time-to-event end points were modeled using Kaplan-Meier survival plots and analyzed by the log-rank test. Significant factors by log-rank test were analyzed for independent prognostic importance by the Cox proportional hazards regression using a stepwise procedure. Because of the modest number of cases in this study, only factors found to have P < 0.2 by univariate analysis were entered into the Cox model. Fisher’s exact tests were used to assess associations between factors. The statistical analysis was performed for all patients with available follow-up and for the subgroup of patients presenting with localized disease.

The clinical end points analyzed were time to local recurrence (LRFS), to distant metastasis (MDFS), and to death from disease (DSS). The factors analyzed included: age, size, location, histological grade (≥5% RC; ≥25% RC; and ≥5% increased cellularity in the subgroup of patients with <5% RC), necrosis (≥5% of the tumor mass), surgical margins, and TLS-CHOP fusion transcript type, and status of P53 protein overexpression.

All cases had a typical morphological appearance of MLS. The high-grade areas were restricted to the RC component. None of the 82 cases showed areas of dedifferentiation or nonlipogenic spindle cell proliferation. Furthermore, no histological features overlapping with other types of liposarcoma, such as well-differentiated or pleomorphic types, were noted.

Using ≥5% RC as the cutoff point, there were 47 (57%) low-grade and 35 (43%) high-grade cases. When ≥25% RC cutoff was used, there were 67 (82%) low-grade and 15 (18%) high-grade cases. Of the 47 cases with <5% RC, 27 (57%) showed areas of increased cellularity. Thirty-nine (48%) cases showed microscopic evidence of tumor necrosis in ≥5%.

The most common TLS-CHOP fusion type was type 5-2, identified in 55 cases (67%), followed by type 7-2 in 16 cases (20%) and type 8-2 in 8 cases (10%; Fig. 1). In 2 cases, the EWS-CHOP transcript was identified (2%; Fig. 1). There was no evidence of alternative splicing identified by RT-PCR, i.e., in no case did we detect more than one type of TLS-CHOP fusion transcript.

A total of 47 (57%) RT-PCR products were sequenced as follows: 20 cases (40%) of type 5-2 fusion and all type 7-2 and 8-2 TLS-CHOP transcripts, and other fusion variants. From the 20 type 5-2 cases sequenced, only one case showed a variation in size and sequence, having two amplified bands (Fig. 1). The smaller band (169 bp) showed a fusion of the 5′ portion of exon 5 of TLS with a 26-bp sequence from intron 1 of CHOP (nucleotides 2358–2383; GenBank accession no. Y09999), which functioned as a cryptic exon spliced to exon 2 of CHOP. This fusion transcript was in-frame (Fig. 2). The TLS portion of the fusion transcript lacked 107 nucleotides of exon 5. The break had occurred in exon 5 of TLS (at position 497; GenBank accession no. S62138). The larger band (782 bp) was a transcript consisting mostly of CHOP intron 1 sequences, which contained stop codons in all frames. The possibility of this latter product arising from genomic DNA was excluded by a negative result on RT-PCR lacking reverse transcriptase. Because the shorter in-frame fusion product in this case was most similar to a “5-2,” it was included in the type 5-2 group for further analysis.

Another case showed a novel and unique variant fusion transcript (Fig. 1), joining the 5′ portion of exon 6 of TLS to exon 2 of CHOP. The TLS break was located within exon 6 (position 733). Therefore, a portion of exon 6 (lacking 129 bp) was fused in-frame with exon 2 of CHOP, resulting in a product size of 382 bp (Fig. 2).

Direct sequencing of the two EWS-CHOP cases showed the same 179-bp product with an in-frame junction of exon 7 of EWS to exon 2 CHOP. One of these cases was reported previously (3).

Because the distribution of transcript types in the present study differed from some previous smaller studies (13, 22, 23) in that the proportion of type 5-2 cases was higher, we sought to independently confirm the TLS-CHOP fusion type by LR DNA PCR. Among the 61 cases studied by LR DNA PCR, an amplified product in the expected size range was identified in 9 of 9 (100%) type 7-2 (Fig. 3), 6 of 7 (85%) type 8-2, and 28 of 45 (62%) type 5-2 cases tested (Fig. 4), using appropriate primer pairs, as described in “Patients and Methods.” In the 28 positive type 5-2 cases, the TLS-CHOP genomic junction fragment was amplified in 8 cases, and the reciprocal CHOP-TLS fragment was amplified in 20 cases (Fig. 4). Because of case-to-case variation in both genomic junction fragment size and DNA quality, the occurrence of some negative cases with LR DNA PCR assays is not unexpected. As stated in “Patients and Methods,” all cases were previously confirmed to harbor CHOP rearrangement by Southern blotting or by conventional cytogenetic demonstration of a t(12;16), including those negative by LR DNA PCR but positive by RT-PCR.

To exclude the possibility that some cases with TLS intron 7 or 8 breaks were resulting in type 5-2 transcripts through alternative splicing, we also studied 44 type 5-2 cases with the LR TLS7-CHOP2 primer set (used for types 7-2 and 8-2). These assays were all negative, arguing against the possibility of more downstream breaks with alternative splicing. Thus, taken together, the LR DNA PCR results were entirely consistent with the RT-PCR results, thereby validating the fusion typing results.

From the 71 cases available for immunohistochemical analysis, 12 (17%) showed nuclear immunoreactivity for P53 in ≥10% of tumor cells. Except for one case, the nuclear reactivity was predominantly restricted to the RC component or the hypercellular areas (Fig. 5). Eight of the 29 (27%) high-grade MLS and 3 of 27 (11%) low-grade MLS with increased cellularity showed overexpression of P53. Only 1 of the 15 (7%) pure MLS cases studied showed P53 immunoreactivity.

Clinical follow-up was available in 79 of 82 patients (96%). Median follow-up was 44 months (range, 3–288 months). Of the 79 patients with follow-up data, 22 (28%) developed local recurrence, 30 (38%) had distant metastases, and 20 (25%) patients died of disease. At last follow-up, 44 (56%) patients were alive with no evidence of disease; 15 patients (19%) were either alive or dead with disease; and 20 (25%) were dead of disease. The 5-year DSS in this group was 73%. Among the metastatic sites, an overrepresentation of unusual locations was noted, such as soft tissue (63%) and bone (37%), with only 33% spreading to the lung. Furthermore, in a significant number of patients, soft tissue and bone metastases occurred early in the disease course, before clinical dissemination or pulmonary metastases.

In the subgroup of patients presenting with localized disease at diagnosis and available follow-up data (70 cases), 19 (27%) had local recurrences and 21 (30%) distant metastases. Thirteen (19%) patients died of disease. In this group, 44 (63%) patients were alive with no evidence of disease, 13 (19%) were alive or dead with disease, and 13 (19%) were dead of disease at last follow-up. The 5-year LRFS was 73%, 5-year MDFS 66%, and 5-year DSS 80%. If the analysis was restricted to cases with histologically low-grade tumors, the 5-year rates were 86% MDFS and 83% DSS for the MLS cases with increased cellularity and 79% MDFS and 100% DSS for pure MLS.

By univariate analysis, high histological grade (defined as ≥5% RC) was a strong predictor of MDFS in the localized group (P < 0.01) and of DSS both in the entire (P < 0.01) and localized groups (P = 0.01; Fig. 6). By multivariate analysis, localized tumors with ≥5% RC were independently associated with a poorer DSS (P = 0.02). The second cutoff point used to define high histological grade (≥25% RC) did not reach statistical significance in any of the groups or survival functions tested. Also, the presence of increased cellularity in the cases with <5% RC did not reach statistical significance for DSS or MDFS, although a trend (P = 0.08) was identified in predicting DSS in the localized cases.

The presence of necrosis (≥5% of tumor mass) correlated with DSS in the entire group (P < 0.001 by univariate and P < 0.01 by multivariate analysis) and in the localized group (P = 0.01, by univariate analysis; Fig. 7). In addition, the presence of necrosis predicted MDFS in the patients with localized disease at presentation by univariate analysis (P < 0.05; Table 2). The tumor size was not found to predict clinical outcome in the present cohort, presumably because the majority of cases (73%) were >10 cm.

Of the two biological factors tested, TLS-CHOP fusion type and P53 status, only the latter showed an impact on survival. There was no significant association between TLS-CHOP fusion transcript type and histological grade or survival (MDFS, DSS). Overexpression of P53 (defined as nuclear staining in ≥10% of cells) was associated with an unfavorable outcome, predicting MDFS in localized tumors (P < 0.001 by univariate and P = 0.02 by multivariate analysis; Table 3 and Fig. 8) and DSS in the entire group (P = 0.001 by univariate and P < 0.05 by multivariate analysis; Table 3). Interestingly, there was an association between TLS-CHOP fusion transcript type and P53 status. The majority of P53-positive cases contained the type 5-2 TLS-CHOP fusion, and none of the type 7-2 or type 8-2 fusion cases overexpressed P53 protein (P < 0.01). Of 12 P53 positive cases, 11 had a type 5-2 TLS-CHOP fusion transcript and 1 case had a EWS-CHOP fusion.

The LRFS was significantly related to negative surgical margins (P < 0.01) and extremity location (P = 0.02) in the entire group, by multivariate analysis. The same adverse factors were found to be significant in the localized cases, both by univariate and multivariate analysis. By univariate analysis, the presence of increased cellularity (defined as ≥5%; P < 0.01) or tumor location other than thigh (P = 0.05) also correlated with a higher risk of local recurrence in the localized group of patients.

The most common TLS-CHOP fusion type detected in our study was type 5-2, followed by type 7-2 fusion, a finding in concordance with data from previous smaller series (13, 24, 25). However, in contrast to some studies (22, 26), we and Kuroda et al. (23) found a higher frequency of type 8-2 fusion (10%) and a very small number of fusion variants (1%). Two cases exhibited the EWS-CHOP transcript, bringing the number of reported cases to six (3, 8, 9).

The different combinations of TLS and CHOP genomic breakpoints produce in-frame transcripts because the splice junctions of exons 5, 7, and 8 of TLS and exon 2 of CHOP occur at the same codon nucleotide position. The two main types, fusion of TLS exon 5 to CHOP exon 2 (type 5-2) and fusion of TLS exon 7 to CHOP exon 2 (type 7-2), account for ∼85% of TLS-CHOP fusions. In contrast to a study that found that 3 of 13 cases contained two transcripts (types 7-2 and 5-2), presumably as a result of alternative splicing (13), none of our 82 cases showed more than one in-frame transcript type, indicating that this phenomenon is uncommon in MLS.

This study represents the first large scale molecular analysis of the genetic heterogeneity of TLS-CHOP fusion transcript and its relationship to survival in MLS patients. In contrast to some other sarcomas characterized by specific translocations, such as Ewing sarcoma, synovial sarcoma, or rhabdomyosarcoma (27, 28, 29, 30), our results indicate that the molecular variability of TLS-CHOP fusion transcripts in MLS does not appear to have a clear impact on clinical outcome. This would suggest that within TLS-CHOP, TLS exons 6–8 do not contribute an additional functional domain relevant to the biology of MLS. In the EWS-FLI1 fusion, functional variability seems to be determined primarily by the inclusion of additional portions of FLI1, rather than of portions encoded by EWS exons 8–10 (27, 31). In terms of exon structure and domains encoded, TLS exons 6–8 correspond to exons 8–10 of EWS (32, 33), but there is actually only limited sequence similarity, either at the DNA or protein level (6, 33). This portion of TLS and EWS contains a domain of uncertain function, the RGG domain (defined by multiple copies of the sequence Arg-Gly-Gly) on the NH₂-terminal side of the RNA-binding domain.

In contrast to other liposarcoma variants or myxoid sarcomas of the extremities, MLS tends to metastasize to unusual soft tissue locations, such as the retroperitoneum, opposite extremity, and axilla, among others (2, 3, 34). In the present study, we found a high incidence of soft tissue metastases, 24% for the entire group and almost two-thirds of those with metastases. A significant number of these soft tissue implants were identified early in the disease course, before the manifestation of disease in sites where sarcomas usually metastasize, such as the lung. MLS is the most common type of soft tissue sarcoma presenting with either synchronous or metachronous multifocal disease (3, 34, 35). We have recently demonstrated the monoclonal origin of multifocal MLS by the analysis of TLS-CHOP and EWS-CHOP genomic rearrangements, establishing the metastatic nature of distant soft tissue lesions in these cases (3).

Histological grading in MLS is based on the extent of RC component. The transition from morphologically low-grade to higher grade areas (whether hypercellular or RC) can be either abrupt or, more often, gradual (36). Therefore, accurate assessment of histological grade in ML is not only biased by sampling error but also by difficulty in objective quantitation of the RC component, particularly in those cases with gradual transition to a RC component. The distinction between the “RC” phenotype and a hypercellular focus can be histopathologically challenging. According to some studies, the presence of any hypercellular areas seems to correlate with an increased risk of recurrence (16, 17, 37), whereas the presence of ≥25% RC shows strong statistical correlation with poor outcome (2). Furthermore, Evans (37) stressed that hypercellularity in recurrent lesions, rather than primary tumors, was that which correlated with poor outcome. He also noted that nearly one-half of MLS with <5% hypercellularity eventually metastasized. Azumi et al. (38) noted that no case developed metastasis unless it showed hypercellular areas either in the primary or recurrent tumor. In our study, similar to the study by Smith et al. (17), the presence of ≥5% RC strongly correlated with an unfavorable outcome, by both univariate and multivariate analysis, whereas the presence of increased cellularity in low-grade tumors showed only a tendency to predict DSS in localized cases. Our findings indicate that the 5% cutoff for the presence of a RC component is statistically more powerful than 25%, suggesting that it also may be more biologically meaningful. This is best reflected by the high 5-year disease-specific survival rates, 100% for the pure MLS cases and 83% for the low-grade MLS with increased cellularity. Similar to the study by Killpatrick et al. (2), our findings also suggest that the presence of tumor necrosis correlates with a more aggressive behavior.

We and others have confirmed the strong specificity of TLS-CHOP for MLS in a large number of myxoid sarcomas (19, 39). Because of overlapping morphological features with other myxoid neoplasms, the differential diagnosis of MLS can be sometimes quite challenging. Myxofibrosarcoma in the extremities and WDLS with extensive myxoid change in the retroperitoneum are the most common diagnostic pitfalls (40). Detection of TLS-CHOP (or EWS-CHOP) fusion transcripts can serve as a useful diagnostic adjunct in such cases. The availability of this large series of MLS with confirmatory translocation data allows us to further comment on two clinicopathological issues: putative mixed or dedifferentiated forms of MLS, and the occurrence of MLS in the pediatric age group. There have been occasional reports of apparent mixed-type liposarcoma, with features of both WDLS and MLS, and a recent description of three cases of MLS with areas of dedifferentiation or a high-grade nonlipogenic component, a phenomenon typically associated with WDLS (41, 42, 43). In the present study, no cases with TLS-CHOP rearrangement showed either areas of well-differentiated or dedifferentiated liposarcoma histologically, further supporting the concept that MLS and WDLS are distinct entities and suggesting that the aforementioned “mixed” types of liposarcoma might be attributable to incomplete histopathological and/or molecular analysis. MLS is a disease of younger adults, with the age at presentation (present series, 45.8 years) on average a decade lower than other histological subtypes of liposarcoma. In the pediatric age group (<18 years), MLS is an extremely rare occurrence (44, 45) and can be confused morphologically with the more common lipoblastoma. Few if any cases reported as pediatric MLS have been confirmed to contain the TLS-CHOP fusion. In our study, a single case of MLS occurred in this age group. This was a 13-year-old boy presenting with a histologically low-grade MLS of the thigh, which showed a type 5-2 fusion.

Few prior studies have analyzed the incidence of P53 alterations in the myxoid and round cell forms of MLS (46, 47, 48), and they did not examine associations with survival. In these studies, the P53 overexpression varied from <5% (46) up to 30% of cases (47). In the present study, we found that overexpression of P53, typically associated with underlying missense mutation, although present in only a minority of tumors (17%), was associated with poor outcome. In the study by Smith and Goldblum (48), 7% of MLS cases analyzed immunohistochemically were positive for P53 and were present only in high-grade tumors (≥5% RC) and localized within the RC component. These findings are in concordance with our results, in which the majority of P53-positive tumors were seen in high-grade tumors and the nuclear immunoreactivity present in the RC to hypercellular areas. P53 alterations have a similar strong negative impact on survival in at least two other translocation-associated sarcomas, Ewing’s sarcoma (14) and synovial sarcoma (15). We also observed an association between the type 5-2 fusion transcript and P53 overexpression, suggesting a nonrandom relationship between fusion transcript type and P53 mutation. This needs to be confirmed in future studies.

In summary, we find that high histological grade (≥5% RC), presence of necrosis, and P53 overexpression are independent predictors of unfavorable outcome in localized MLS.

UNMC tissue procurement was performed under IRBH281-96, supported by NCI-P30 CA36727. We would like to thank Drs. H. Rosenthal, G. Vergara and G. Maale for their patient sample contribution, Wen Chen for technical support, and Irina and Nicolae Tinca for charitable contributions.