The first 4 authors contributed equally to this article.
We thank the Lobo, Margolis, and Jackson Families for their continued support of our liposarcoma research and Dr. Torsten O. Nielsen for his helpful comments.
Myxoid liposarcoma (MLPS), a disease especially of young adults with potential for local recurrence and metastasis, currently lacks solid prognostic factors and therapeutic targets. The authors of this report evaluated the natural history and outcome of patients with MLPS and commonly deregulated protein biomarkers.
Medical records were retrospectively reviewed for patients who presented to the authors' institution with localized (n = 207) or metastatic (n = 61) MLPS (1990 to 2010). A tissue microarray of MLPS patient specimens (n = 169) was constructed for immunohistochemical analysis of molecular markers.
The 5-year and 10-year disease-specific survival rates among patients with localized disease were 93% and 87%, respectively; male gender, age >45 years, and recurrent tumor predicted poor outcome. The local recurrence rate was 7.4%, and the risk of local recurrence was associated with recurrent tumors and nonextremity disease location. Male gender was the main risk factor for metastatic disease, which occurred in 13% of patients. Forty percent of patients who had localized disease received chemotherapy, mostly in the neoadjuvant setting. Immunohistochemical analysis revealed significantly higher expression of C-X-C chemokine receptor type 4 (CXCR4) and platelet-derived growth factor beta (PDGFR-β) in metastatic lesions versus localized lesions. Tumors with a round cell phenotype expressed increased levels of CXCR4, p53, adipophilin, PDGFR-α, PDGFR-β, and vascular endothelial growth factor relative to myxoid phenotype. Only the receptor tyrosine kinase encoded by the AXL gene (AXL) was identified as a prognosticator of disease-specific survival in univariate analysis.
Liposarcoma (LPS), a mesenchymal malignancy cohort that demonstrates lipogenic differentiation, is the most common adult soft tissue sarcoma subtype.1 LPS consists of 3 categories: 1) well differentiated LPS and dedifferentiated LPS (WDLPS/DDLPS), 2) myxoid and round cell liposarcoma (MLS and RCL [MLPS]), and 3) pleomorphic liposarcoma (PLS). The RCL variant of MLPS represents histologic progression of pure MLS to hypercellular round cell morphology, defined as >5% RCL phenotype in a given tumor, and is associated with a poor prognosis.2 Both variants bear characteristic reciprocal chromosomal translocations, most frequently the t(12;16) (translocation involving bands 12 and 16) in “fused in sarcoma/DNA-damage–inducible transcript 3” (FUS/DDIT3)3 or, rarely (<5%), Ewing sarcoma breakpoint region 1 (EWSR1)/DDIT3 t(12;22)(q13;q12) (translocation involving 12q13r and 22q12).4 The resultant fusion proteins may play multiple roles in tumorigenesis, including cancer initiation,5 specifically blocking complete MLPS adipocytic differentiation6 however, the exact mechanisms of action remain unknown. MLPS incidence peaks in the fourth and fifth decades of life, predominantly arising in the lower extremities and buttock. Metastasis, which occurs in up to 33% of patients, is characterized by an unusual predilection for fat-bearing areas, including the abdomen, axilla, and bone.7 Surgical resection with or without radiotherapy is the standard of care for patients with localized MLPS. Chemotherapy is usually reserved for patients who have high-risk/advanced disease. Accurately determining prognosis remains problematic. Neither clinical factors (ie, age, resection margin status, tumor location) nor histologic factors (eg, necrosis) consistently correlate with outcome.2,8–13 Several studies have suggested that an RCL subtype either predicted prognosis2,8–11 or was not associated with prognosis,12,13 differences perhaps caused by inconsistent assessment of round cell changes per se.
Prognostic MLPS biomarkers have been suggested and have included the overexpression of p53, insulin-like growth factor 2, insulin-like growth factor receptor 1, the ret proto-oncogene, etc,14,15 without definitive confirmation. Such knowledge deficits, coupled with our limited understanding of MLPS pathogenesis cellular pathways, compromise prognostic assessments while limiting our abilities to develop specific treatment. Consequently, we used human MLPS specimens assembled in a tissue microarray (TMA) to investigate MLPS natural history and clinical outcomes, seeking to identify disease-specific survival (DSS) prognosticators and commonly deregulated molecular processes/biomarkers.
MATERIALS AND METHODS
After we received Institutional Review Board approval from The University of Texas MD Anderson Cancer Center (UTMDACC), records from 268 patients who had histologically proven MLPS (from April 1990 to March 2010) were used to construct a clinical database containing patient, tumor, treatment, and follow-up information.
The TMA included 169 MLPS biopsy and surgical resection samples from 110 patients and 14 control samples of normal fat, WDLPS, PLS, and myxoma; and it was constructed as described previously.16
Immunohistochemical staining was performed on 4-mm-thick TMA sections. Commercially available antibodies against Ki67, cyclin D1, B-cell lymphoma 2 (BCl2), matrix metalloproteinase 9, p53, chemokine (C-C motif) receptor 7, C-X-C chemokine receptor type 4 (CXCR4), human epidermal growth factor receptor 2 (HER2), receptor tyrosine kinase encoded by the AXL gene (AXL), platelet-derived growth factor alpha (PDGF-α), PDGF-β, PDGF receptor alpha (PDGFR-α), PDGFR-β, epidermal growth factor receptor (EGFR), vascular endothelial growth factor (VEGF), peroxisome proliferator-activated receptor gamma (PPAR-γ), adipophilin, cMET (a proto-oncogene that encodes the hepatocyte growth factor receptor), cKIT (a proto-oncogene also known as mast/stem cell growth factor receptor or tyrosine-protein kinase Kit), estrogen receptor, and progesterone receptor were used. Specific antigen retrieval, blocking agents, antibody specifications, and concentrations are available upon request. Horseradish peroxidase-labeled secondary antibodies or biotinylated systems (4-plus system; Biocare Medical, Concord, Calif) were used. Labeling intensity was scored as 0 (none), 1 (weak), 2 (moderate), or 3 (strong); and the percentage of positive tumor cells was estimated. Samples that exhibited an intensity of 0 or 1 were considered “low expression, ” whereas intensities of 2 or 3 were considered “high expression.” Because of poor reproducibility of the distinction between staining intensity of 1 and 2 for PDGFR-α, PDGFR-β, and CXCR4, only cases with a score of 3 were considered to represent high scoring.
Correlations between TMA biomarker expression and tumor or disease status were calculated using the Fisher exact test. Spearman correlation coefficient analysis was used to determine associations between biomarkers. Correlations of clinicopathologic data and immunohistochemical biomarker expression with patient outcomes were evaluated using univariate and multivariable Cox proportional hazards models for DSS, recurrence-free survival, and metastasis-free survival. All variables that were significant in univariate analysis were used in the multivariate model, and only those variables are illustrated.
Characteristics of Localized Myxoid/Round Cell Liposarcoma
Localized MLPS clinicopathologic variables are summarized in Table 1. The mean presenting age was 41 years (range, 13.7–79 years), with male predominance (57.5%). Most patients presented with a primary lesion (89.3%), especially of the extremities (74%). Tumors usually were located in deep soft tissue (76%); the median tumor size was 10 cm (range, 1.7–48 cm), and tumors typically presented as a painless mass (81%). Nearly all patients (96%) underwent surgery. Only 8 patients were deemed unresectable because of anatomic constraints or medical comorbidities. Microscopically clear resection margins (R0) were achieved in 83% of patients, microscopically positive (R1) margins were achieved in 13% of patients, and gross tumor (R2 margins) remained in 4% of patients. Forty percent of patients who were identified as high-risk (tumors >10 cm; RCL histology) received chemotherapy (usually neoadjuvant, 77%), especially doxorubicin and ifosfamide. Seventy-four percent of patients received radiotherapy (usually neoadjuvant).
Outcomes of Patients With Localized Myxoid/Round Cell Liposarcoma
The median follow-up for patients with localized disease was 68 months (range, 1–249 months). The local recurrence rate was 7.4%, and the median time to recurrence was 31 months (range, 6–83 months). The risk of local recurrence was associated with disease status (local recurrence vs primary tumor) at the time of diagnosis and nonextremity disease location (Table 2). Metastasis occurred in 13% of patients at a median of 34 months (range, 5–141 months) after diagnosis. Higher rates of metastasis were observed in patients who had RCL patients compared with patients who had MLS (21% vs 10%; P = .042). Although multiple univariate analysis factors were associated with increased risk of metastasis, only male gender remained statistically significant in multivariate analysis (Table 3). The DSS rate was 99% at 1 year, 93% at 5 years, and 87% at 10 years (Fig. 1A). Age >45 years, male gender, and locally recurrent disease were statistically significant prognosticators in multivariate analysis (P = .0252, P = .0159, and P = .049, respectively) (Table 4). Overall survival analysis of patients with localized versus metastatic disease and those with primary tumors versus recurrent disease revealed statistically significant differences (P > .0001 and P = .0127, respectively), mirroring the DSS analysis results, whereas a comparison of overall survival between the MLS and RCL subtypes revealed only a trend toward a better outcome for patients with MLS (P = .0723; data not shown).
Characteristics and Outcomes of Patients With Metastatic Myxoid/Round Cell Liposarcoma
The median follow-up for the 61 patients who presented to UTMDACC with metastatic disease was 24 months (range, 1.7–95 months); their clinicopathologic variables are summarized in Table 5. The median patient age at presentation was 50 years (range, 32–78 years) with a predilection among men (70%). The median time to metastasis from diagnosis of primary disease was 35 months (range, 1–195 months). Fifty-seven percent of patients presented with a single metastatic site, most commonly the abdomen (49%; including the pelvis, groin, and retroperitoneum) followed by bone (23%). Nineteen of 26 patients who presented with multiple-site metastases had abdominal involvement; 14 patients had lung and pleura metastasis; 14 patients had bone lesions; 11 patients had axillary, chest wall, and/or mediastinal involvement. Synchronous primary and metastatic lesions appeared in 23% of patients with disseminated disease. Twenty-five patients underwent metastasectomy, and 6 of those patients achieved R0 resection. The majority of patients (90%) received chemotherapy, most commonly doxorubicin and ifosfamide. In 34% of treated patients, doxorubicin and ifosfamide were combined with additional drugs, such as dacarbazine, gemcitabine, and docetaxel. Twenty-five percent of patients with metastatic disease received trabectedin. Twenty-four patients (40%) received radiotherapy before or after surgical interventions (n = 13), and 2 patients received radiotherapy for palliation only. The DSS rate was 78% at 1 year and 8.2% at 5 years, and no patients who presented with metastasis survived beyond 9 years (P < .0001) (Fig. 1A). In univariate analysis, the presence of lung metastasis or multiple metastases predicted poor outcome, whereas surgical metastasectomy was associated with improved DSS. However, only surgical metastasectomy remained significant as a predictor of improved outcomes in multivariate analysis (Table 6). All but 6 patients with metastatic disease received chemotherapy, including 2 patients who were terminal and chose no treatment, 1 patient who refused treatment, 2 patients with stable disease, and 1 patient for unspecified reasons.
Table 5. Metastatic Myxoid/Round Cell Liposarcoma: Patient, Tumor, Treatment, and Outcome Variables, N = 61
Candidate markers, which were selected based on their expression in other malignancies, included cell proliferation, survival, angiogenesis, migration, invasion, and metastasis factors. Biomarker expression levels are illustrated in Figure 2 and are summarized in Table 7; all samples expressed cytoplasmic and nuclear CXCR4 and adipophilin. AXL or PDGFR-β was expressed in most samples. cKIT and cMET staining was negative in all evaluable samples (n = 117 and n = 131, respectively). To evaluate the role of biomarkers in disease progression, expression levels in primary, recurrent, and metastatic tumors were compared. Significantly lower levels of BCL2 and VEGF were observed in primary versus recurrent and metastatic tumors (P = .004 and P = .0042, respectively). CXCR4 and PDGFR-β expression levels were significantly higher in metastatic versus localized lesions (P = .0036 and P ≤ .0001, respectively). Tumors with an RCL phenotype expressed significantly increased levels of nuclear CXCR4, p53, adipophilin, PDGFR-α, PDGFR-β, and VEGF relative to tumors with an MLS phenotype (Table 8). Prognostically, only AXL emerged as a significant predictor of DSS on univariate analysis (hazard ratio, 19; P = .032); on multivariate analysis, AXL failed to predict outcome.
A positive correlation means statistically significant higher expression in RCL vs. MLS.
P values are 1-sided, and values < .025 are considered statistically significant.
This report comprises a comprehensive clinical, pathologic, and molecular study of patients with MLPS based on a large, single-institution cohort in which therapeutic decisions and strategies were managed by an expert multidisciplinary team. The results demonstrated excellent 1-year, 5-year, and 10-year DSS compared with previous studies.2,8–10 Univariate analysis suggested several variables as DSS prognostic factors (Table 4), supporting findings from previous studies.8–10,12 However, multivariate analysis revealed only age >45 years, male gender, and locally recurrent disease as significant prognosticators of a poor outcome. The survival outcome results from of our series are supported further by the similar trend in both DSS and overall survival.
Our 7.4% local recurrence rate is similar to that reported by a recently published large MLPS cohort11 and is better than the rates reported in other published series,8–10,12,13 perhaps because of more frequent positive margins reported by Haniball et al (34%), Kilpatrick et al (49%), and ten Heuvel et al (24%) compared with our current series (17%); less frequent use of radiotherapy; and/or the inclusion of patients with primary retroperitoneal myxoid lesions. Our preferred therapeutic approach since the 1980 s has been radiotherapy (50 grays) followed by surgery, which facilitates excellent local control with smaller radiation fields17 than postoperative radiotherapy, which we reserve for patients who are referred to us after undergoing primary tumor resection elsewhere. Our results indicate that tumor location (nonextremity vs extremity) and tumor status (recurrent vs primary) are predictors of poor recurrence-free survival, whereas resection margins and histology type (MLS vs RCL) do not predict recurrence-free survival, in contrast to other studies.8,9,11 Our 13.1% metastasis rate is at the lower end of published metastasis ranges2,8–11 (up to 38%), perhaps reflecting the aggressive therapeutic algorithms used at UTMDACC. Our univariate analysis indicated that radiotherapy is associated with poorer metastasis-free survival, perhaps because of selection bias, in which patients who have tumors that are deemed more aggressive are more likely to receive radiotherapy.
Male gender was the only significant multivariate analytic prognostic factor for metastasis-free survival. A male predilection in MLPS has been described previously2,8–13,17. however, to the best of our knowledge, this is the first time it has been identified as prognostic of a poor outcome. It is noteworthy that most other MLPS series did not evaluate the association between sex and outcome; therefore, we believe that ours is the largest series to date addressing this question. Seeking an explanation for this finding, we investigated whether hormone receptors (estrogen, progesterone, and androgen receptors) played a role in MLPS. However, negligible expression of these receptors was observed. Other factors may portend an inferior prognosis for men with sarcoma; eg, mutation in RASSF1A (Ras association domain family member 1).18 Although we have not yet identified sex-specific drivers in MLPS, future studies may more clearly accomplish this objective. Surgical resection remains the standard of care for patients with localized MLPS. Recently, the role of radiotherapy in local disease control has been demonstrated,17 and the rates of radiotherapy receipt have increased, as we and others have demonstrated.9,11 The role of chemotherapy for patients with localized MLPS is uncertain, and its rate of administration at some treatment centers is as low as 6%11 compared with 40% at UTMDACC, where any patient with localized disease who is considered high-risk (tumor size >10 cm, RCL histology, and positive resection margins) receives chemotherapy. Chemotherapy did not prolong survival in our patients. However, the absence of a robust “no-treatment” comparison group may be associated with selection bias in favor of treating all patients who are deemed “high-risk.” Chemotherapy in selected patients with MLPS can achieve positive response rates19 and may account in part for the improved local and distant control rates we observed. Nonetheless, the need persists for enhanced identification of patients who may benefit from chemotherapy. The MLPS metastatic pattern is unique. In contrast to other soft tissue sarcomas, MLPS primarily metastasizes to extrapulmonary sites, most commonly the abdomen.20 The analysis of our subgroup of patients who presented with metastatic disease is consistent with this reality. Not surprisingly, lung metastases predicted worse outcomes on univariate analysis, because these generally appear at later disease stages. We speculate that the ability to colonize the lungs may represent an adaptation of the tumor to a nonfatty environment, perhaps reflecting increased tumor aggressiveness. Chemotherapy is the mainstay treatment of metastasis; reminiscent of local disease, our patients who received chemotherapy did not have outcomes superior to those of nontreated patients; whereas surgery conferred a positive outcome effect. These findings may reflect biases in nonrandomized selection for treatment with either modality. Metastasectomy, especially of isolated lesions, may favorably affect outcome. Patient selection for metastasectomy is individualized and is based on general condition, prognosis etc; indications include low tumor grade, age <50 years, a greatest tumor dimension <15 mm, and slow disease progression (>12 months between pulmonary recurrences or >24 months between extrapulmonary recurrences).
Myxoid/Round Cell Liposarcoma-Related Biomarkers
We investigated adipocytic differentiation biomarkers to determine what role (if any) these may play in MLPS tumorigenesis. All samples expressed high levels of adipophilin, a known marker of adipogenesis that appears early in the differentiation process,21 perhaps suggesting that both MLS tumors and RCL tumors differentiate beyond the initial stage(s) before the disruption of complete adipocytic maturation. Moreover, we observed significantly higher levels of adipophilin in RCL tumors versus MLS tumors, perhaps suggesting that adipophilin participates in disease progression. The adipogenesis regulator PPAR-γ is overexpressed in MLPS,15 and our data support this observation. However, it is also known that PPAR-γ has protumorigenic properties22; in this setting, PPAR-γ may function more as a tumorigenic factor than an adipogenic factor.
CXCR4 is a chemokine receptor in many cancers and may facilitate metastases through effects on cell migration, invasion, and angiogenesis.23 All of our MLPS samples demonstrated CXCR4 protein expression, which was higher in the clinically more aggressive RCL variant than in MLS. Cytoplasmic CXCR4 expression also was higher in metastatic MLS compared with localized MLS, a pattern that has been observed in breast cancer, lung cancer, colon cancer, melanoma and other malignancies.24,25 CXCR4 may be involved in aggressive MLPS tumor behavior (RCL and metastases) and may serve as a future viable therapeutic target.
The overexpression of p53 may reflect dysfunctional antiapoptotic signaling and has been linked to a poor response to chemotherapy.26 Expression of p53 in MLPS varies from 7% to 100% positivity in published series.2,14,27 In our study, p53 overexpression in RCL versus MLS tumor variants was comparable to that reported in other studies; however, it did not correlate with outcome, in contrast to others.
Receptor tyrosine kinases (RTKs) and their ligands are expressed in many mesenchymal malignancies and have been targeted by an ever-increasing array of directed therapies. Tumors often express multiple RTKs because of cross-talk with other pathways, such as the RAS/MAPK (rat sarcoma/mitogen-activated protein kinase) and angiogenic signaling pathways. Therefore, we assessed the expression of several RTKs to determine their theragnostic significance. AXL is an RTK associated with tumor invasion, angiogenesis, and metastasis28; its expression correlates with a poor outcome in several malignancies, including MLPS, as reported here, and may be an attractive candidate for targeting given the availability of AXL inhibitors. AXL was expressed in almost all of our MLPS samples (97%); and, for the first time, we demonstrated that high AXL expression predicts a poor outcome in patients with MLPS (univariate analysis). However, the importance of this finding is unclear, because it was not confirmed on multivariate analysis.
Previous reports indicated that PDGFR-β was expressed in MLPS.29 Our finding that PDGFR-β expression was higher in metastatic lesions compared with localized lesions is intriguing given its role in bone metastasis.30 Considering the relatively high prevalence of MLPS bone metastasis and the availability of PDGFR-β inhibitors, this may be an attractive locus for further investigation.
Our study has several limitations. Because it is a retrospective study, it requires prospective validation and cannot be used to justify therapeutic decisions. The major caveat of our histologic analysis is that, of 268 patients who were included in our study, only 169 tissue samples were analyzed, because patients whose tumors were resected elsewhere were not included in our TMA. Also, those patients who received preoperative therapies before surgery, resulting in no residual tumor identified in the resection specimen, were not represented on the TMA. This resulted in a TMA inclusion selection bias that favored more clinically resistant tumors and/or larger tumors that were inadequately resected elsewhere. Finally, although we identified several biomarkers that had prognostic utility, this does not mean that these proteins are integral to MLPS tumor biology; consequently, they may be irrelevant in therapeutic targeting.
In conclusion, we present a large MLPS patient study incorporating comprehensive follow-up and biomarker analysis. Our cohort was notable for the aggressive treatment strategy for localized tumors, which may correlate with lower metastatic rates than previously described. In addition, we identified age, male gender, and abnormal expression of several proteins as potential prognostic factors. Therefore, future prospective studies confirming our results appear to be warranted.
This work was supported in part by a Deutsche Forschungsgemeinschaft training grant (supporting Dr. Ghadimi) and an Advanced Imaging Research Center fellowship grant (supporting Dr. Creighton). Dr. Hoffman is the recipient of a fellowship grant from the American Physicians Fellowship for Medicine in Israel.