Prognostic Value of Myogenic Differentiation in Dedifferentiated Liposarcoma

Abstract

Aim:

Myogenic differentiation (MD) has been claimed to be a poor prognostic factor in dedifferentiated liposarcoma (DDLPS). To validate this, the prognostic significance of MD in a uniformly treated cohort of DDLPS was assessed.

Methods:

A cohort of patients that have been uniformly treated at one institution for DDLPS of the retroperitoneum and pelvis were stained with smooth muscle actin and desmin and semiquantitatively scored for staining focality and strength. Clinical and survival data was collected, and the prognostic significance of MD was evaluated.

Results:

50 patients with uniformly treated DDLPS were evaluated. SMA (p=.052) and combined score of MD (SMA+desmin) showed statistically significant decrease in 5-year DFS (p=.002) in univariate analysis and in multivariate testing combined MD trended toward significance (p=.052). Combined MD was associated with a decreased OS in multivariate analysis (p=.004).

Conclusions:

In a uniformly treated cohort of DDLPS stained for myogenic markers, a combined myogenic score was associated with poor overall survival in multivariate analysis. However, the difference in groups was slight and the clinical application is limited.

Introduction

Liposarcomas are a diverse group of mesenchymal tumors with adipocytic differentiation that have variable biological behavior, classified into 3 main clinicopathologic categories: 1) atypical lipomatous tumor/well-differentiated liposarcoma (ALT-WDLPS) and dedifferentiated liposarcoma (DDLPS), 2) myxoid liposarcoma, and 3) pleomorphic liposarcoma.^1,2 DDLPS most commonly shows undifferentiated spindle to pleomorphic sarcomas but may have homologous transformation, in the form of pleomorphic liposarcoma-like morphology, or heterologous transformation, that may have osteosarcomatous, chondrosarcomatous, or myogenic (malignant smooth or skeletal muscle elements) differentiation. DDLPS accounts for up to 10% of liposarcomas most commonly arising in the retroperitoneum; however, it may involve the mediastinum, spermatic cord/scrotum, head and neck, and extremities.

Prior studies have evaluated the prognostic significance in DDLPS with myogenic differentiation, with one study of retroperitoneal DDLPS suggesting patients with liposarcoma showing myogenic differentiation have a worse survival.³ However, these studies often include a non-uniform population of patients, where histology, clinical and treatment factors have not been controlled. In this study, we sought to evaluate the prognostic significance of myogenic differentiation in a uniformly treated cohort of patients with DDLPS.

Methods

The laboratory information system was searched for dedifferentiated liposarcoma between 1995 and 2016. To maintain uniformity of treatment, only patients who had their primary treatment and resection at this institution were included in this study. Any patients that had prior surgery or chemoradiation at another institution were excluded. Although not specifically excluded, none of the DDLPS included in this study had definitive leiomyosarcomatous or rhabdomyosarcomatous differentiation. Clinical, radiologic and pathologic data was collected in all patients. For survival analysis, patient vital status (alive vs deceased), disease free status (no evidence of disease (NED) vs alive with disease (AWD)/dead of disease (DOD)), overall survival (OS, in days) and disease free survival (DFS, in days) was recorded.

Immunohistochemical staining for smooth muscle actin (SMA), desmin, caldesmon, calponin and myogenin (table 1) was performed on representative 5um sections of formalin-fixed paraffin-embedded tissue from DDLPS whole slides (39 cases) and tissue microarray (11 cases, 3mm cores). Staining was performed on a Leica Bond-Max automated stainer (Leica, Bannockburn, IL) with appropriate positive and negative controls. As previously reported³, only rare cases were positive for caldesmon, calponin or myogenin, and staining was discontinued; the study proceeded with only staining for SMA and desmin. Staining for these myogenic markers was evaluated by semi-quantitative scoring for intensity (0 - no staining; 1 - weak staining; 2 - strong staining) and focality (0 - no staining; 1 - <10% of cells; 2 – 10-50% of cells; 3 - >50% of cells staining) (Figure 1). A combined score (intensity + focality) was determined for each case.

TABLE 1.

Myogenic Immunohistochemical Markers Antibody Source Clone Dilution Pretreatment

Antibody	Source	Clone	Dilution	Pretreatment
SMA	Leica Biosystems	ASM-1	RTU	None
Desmin	Leica Biosystems	DE-R-11	RTU	ER2 for 20 min
Caldesmon	Bio SB Cancer Diagnostics	CALP	RTU	ER2 for 10 min
Calponin	Agilent Dako	26a11	1:1600	ER2 for 10 min
Myogenin	Leica Biosystems	LO26	RTU	ER2 for 40 min

ER2 indicates Leica Bond Epitope Retrieval Solution 2 heat-induced epitope retrieval; RTU, ready to use, no dilution.

Figure 1.

A) Example photomicrograph of dedifferentiated liposarcoma (hematoxylin and eosin; 400X) with B) desmin intensity score of 0; C) desmin intensity score 1; D) desmin intensity score 2.

R program for statistical computing was used for statistical analysis.⁴ Since there is no established staining cutoff to define “myogenic” vs “non-myogenic” expression in DDLPS, OS and DFS was initially evaluated by univariate log-rank test (R package: survival) at a previously published cutoff of 10% cells staining to be considered positive³. To interrogate alternative cutoffs for each marker, receiver operator characteristic curves (R package: pROC) and hazard ratios were calculated (R package: survcomp, rmeta for forest plots). Staining cutoffs that maximized ROC Youden J statistic or hazard ratios were considered “positive” for myogenic differentiation. Finally, multivariable survival analysis was performed with Cox Proportional-Hazard model (package: survival). Continuous variables (age, tumor size) were modeled using cubic splines (R package: splines). P-value <.05 was considered statistically significant.

Results

There were 50 patients with DDLPS of the RP or pelvis that fulfilled the inclusion criteria for this study: 20 females (average age 67 years, range: 35–82, SD: 12) and 30 males (average age 65 years, range: 44–88, SD: 13) (table 2). Seventeen patients were deceased, 40 patients were alive at last follow-up (range 3.8 – 109 months, average 46 months, SD: 31 months). Twenty-five patients had NED, 17 patients had recurrence (REC), 5 patients developed metastatic disease (MET) and 3 patients had gross residual disease (R2) after primary resection (Table 2).

TABLE 2.

Summary of Clinicopathologic Data

Clinicopathologic Factors	n (%)
Sex
Male	30 (60)
Female	20 (40)
Age (y), mean (range, SD)
Male	65 (44–88, 13)
Female	67 (35–82, 12)
Tumor size (cm)
Mean (range, SD)	20.2 (3.0–75, 12.6)
≤10	8 (17)
>10	39 (83)
Grade (FNCLCC)
1	4 (8)
2	18 (36)
3	28 (56)
Radiotherapy
None	12 (24)
Preoperative RT	26 (52)
IORT	10 (20)
Postoperative RT	16 (32)
Chemotherapy	6 (12)

FNCLCC indicates Fédération Nationale des Centres de Lutte Contre le Cancer; IORT, intraoperative radiotherapy; RT, radiotherapy.

There were 39 whole slides (WS) and 11 DDLPS on TMA stained with myogenic markers. There was no statistically significant difference in staining for desmin (average score of 3.5 for WS, average score of 3 for TMA, p=.708). However, there was a significant difference in SMA staining of WS (average 3.3) versus TMA (average 1.4, p=.001).

Using an initial cutoff of >10% cells staining cutoff (focality score >1), as proposed by Gronchi et al³, 31/50 (62%) were positive for desmin, 26/50 (52%) were positive for SMA, and 46/50 (92%) of DDLPS were positive for at least one or both markers.

Disease Free Survival

Univariate analysis of clinical and pathologic data in relation to DFS is summarized in Table 3.

TABLE 3:

P-Values of Univariate Analysis of Multiple Cutoffs of Myogenic Expression of DDLPS at 2-, 5-Year, and Total Survival

Marker	DFS				OS
	Cutoff	2y	5y	Total	Cutoff	2y	5y	Total
Desmin	>1	0.699	0.979	0.648		0.895	0.798	0.606
ROC	combined score >0	0.561	0.165	0.32	combined score >2	0.343	0.648	0.635
HR	combined score >0	0.561	0.165	0.32	combined score >0	0.662	0.411	0.628
SMA	>1	0.069	0.052	0.225		0.382	0.604	0.613
ROC	combined score>2	0.034	0.081	0.32	combined score >3	0.312	0.482	0.813
HR	combined score>3	0.046	0.021	0.034	combined score >3	0.312	0.482	0.813
Any myogenic expression	>1	0.97	0.795	0.596		0.904	0.786	0.219
ROC	combined score >5	0.166	0.086	0.408	combined score >4	0.428	0.563	0.531
HR	combined score >9	0.012	0.002	0.002	combined score >8	0.226	0.14	0.165
Age		0.52	0.32	0.55		0.59	0.55	0.57
Sex	Male/female	0.282	NA	0.236	Male/female	0.52	0.6	0.91
Tumor size		0.13	0.17	0.23		0.419	0.333	0.91
Grade		0.093	0.093	0.154		0.88	0.44	0.24

In the evaluation of DFS in patients with NED vs REC/MET, a preliminary >10% cells staining cutoff (focality score >1), as proposed by Gronchi et al³, was utilized to define positive myogenic differentiation. At this cutoff, each myogenic marker was tested for disease free survival (DFS) by univariate logrank test at 2-year, 5-year and total DFS, with no markers showing statistical significance, although SMA approached significance at 2-year and 5-year DFS (Table 3) (Figure 2). This cutoff is arbitrary, however, and since there is no defined cutoff for “myogenic differentiation”, two further methodologies were employed to identify potentially significant cutoffs: 1) receiver operating characteristic (ROC) curve (marker score vs disease free status) and 2) hazard ratios (marker score vs DFS time). For ROC, thresholds that maximized Youden J statistic were desmin score > 0 (area under curve (AUC) 0.473), SMA score > 2 (AUC 0.554) and combined myogenic score > 5 (AUC 0.566). Although the AUC were low in all tested markers, in univariate analysis at these cutoffs, SMA showed statistically significant DFS at 2 years (p=.034) (Figure 3).

Figure 2.

Kaplan-Meier Survival Curve of DDLPS Disease Free Survival with SMA Positivity at 10% Cutoff.

Figure 3.

Kaplan-Meier Survival Curve of DDLPS Disease Free Survival with SMA Positivity at Combined Score > 2.

Cutoffs that maximized hazard ratios were also evaluated in univariate analysis: desmin score > 0, HR=2.08), SMA (score >3, HR=2.60) and combined score (score >9, HR=5.32). SMA combined score > 2 and myogenic expression combined score >9 showed statistical significance with disease free survival (Figure 4).

Figure 4.

Kaplan-Meier Survival Curve of DDLPS Disease Free Survival with combined myogenic score > 9.

These cutoffs were further evaluated by multivariate Cox Proportional-Hazard model, comparing age, gender, size, grade of dedifferentiated component, myogenic expression combined score > 9. The overall model did not show statistical significance (p=.200), although tumor size (p=.077) and myogenic expression combined score > 9 (p=.056) did approach significance (Table 3), with myogenic expression combined score > 9 nearing significance at 2-year (p=.096) and 5-year (p=.073) DFS.

Overall Survival

The same procedures were performed for overall survival (summarized in Table 2). None of the clinical or pathologic factors showed statistical significance with OS. There was no statistically significant marker for OS at the previously published cutoff of 10%³. Utilizing ROC maximizing Youden J cutoffs - desmin combined score >2 (AUC =.587); SMA combined score >3 (AUC=.606), any myogenic expression combined score >4 (AUC=.585), there was no statistically significant marker at these evaluated cutoffs. Finally, using maximum hazard ratio, desmin combined score >0 (HR 1.45); SMA combined score >3 (HR 1.14) and any myogenic expression combined score >8 (HR 2.31) again failed to show any statistical significance with univariate overall survival analysis. The multivariate Cox Proportional-Hazard model (age, gender, size, grade, myogenic expression combined score >8) overall was significant (p=.03) with myogenic expression combined score > 8 (p=.004) the only covariate showing significance within the overall model (Figure 5). This did not hold true at 2-year (p=.114) or at 5-year (p=.137) OS.

Figure 5.

Kaplan-Meier Survival Curve of DDLPS Overall Survival with Combined Myogenic Expression Score > 8.

Discussion

In this study, we evaluated the prognostic significance of immunohistochemical expression of SMA and desmin in dedifferentiated liposarcomas. We were unable to replicate prior results³, with this uniform cohort of DDLPS at previously utilized cutoffs, finding only SMA and a combined score of myogenic differentiation showing statistical significance at 2-year, 5-year and total disease free survival in univariate analysis and combined myogenic expression (strong SMA and desmin staining) associated with poor OS in multivariate analysis.

Many soft tissue sarcomas are defined by expression of myogenic markers, such as smooth muscle and skeletal muscle tumors¹. Apart from these, many other non-myogenic tumors may show expression of myogenic markers (e.g. adipocytic, fibroblastic/myofibroblastic, pericytic/perivascular, nerve sheath tumors and undifferentiated/unclassified tumors).^5–8 Prognostic and predictive immunohistochemical markers have become an important part of the evaluations of several malignancies and several studies have evaluated the prognostic significance of myogenic marker expression in non-myogenic soft tissue sarcomas with varying results.^3,6–10 One significant issue that many of these studies have in common is their inclusion of multiple sarcoma subtypes when evaluating myogenic differentiation. Because of this, there may be skewing of results by one or more aggressive sarcoma subtypes with myogenic differentiation (e.g. leiomyosarcoma, pleomorphic rhabdomyosarcoma) that falsely increase the “aggressiveness” of unrelated tumors lumped into this “myogenic differentiation” category, e.g. myxofibrosarcoma with SMA positivity.

Fletcher et al (2001)⁶ in re-evaluating malignant fibrous histiocytoma (MFH) reclassified tumors showing myogenic differentiation by immunohistochemistry (not specifically defined) and showed that tumors with myogenic differentiation had poorer 5-year metastasis-free survival (leiomyosarcoma, pleomorphic rhabdomyosarcoma, and undifferentiated pleomorphic sarcoma (UPS) with myogenic expression, etc. grouped together). These studies collectively show that histologic classification of a tumor with myogenic differentiation (e.g. Leiomyosarcoma and pleomorphic rhabdomyosarcoma) is important as they have a more aggressive clinical course. However, the mixture of subtypes masks the importance of muscle marker expression in non-myogenic sarcomas, such as UPS or DDLPS. These prior studies also fail to identify what level of expression should be considered for “myogenic differentiation”. Deyrup et al (2003)¹⁰ showed a 5-year metastasis-free survival in patients without myoid differentiation in pleomorphic sarcomas arising in the extremities of adult patients. This group defined myogenic differentiation as tumors with at least 1 positive myogenic marker. However, this publication also included evaluation of multiple histologic subtypes, even after excluding leiomyosarcoma. In a similar reclassification study, Massi et al (2004)⁷ revised the diagnosis on soft tissue sarcomas of the adult extremity and included in their myogenic sarcoma category multiple different histologic subtypes: leiomyosarcoma, rhabdomyosarcoma and myofibrosarcoma (myofibroblastic sarcoma). In multivariate analysis, myogenic differentiation was shown to correlate with risk of disease progression.

In one tightly controlled study to avoid many clinical and pathologic confounders, Cipriani et al (2014)⁹ evaluated only large (>8cm) UPS uniformly treated at one institution. They considered tumors with >1% of cells staining with any myogenic marker as myogenic differentiation and showed no difference in overall survival or disease specific survival in UPS with or without myogenic differentiation. This study avoids the skewing of results by limiting evaluation to a single sarcoma subtype. However, in doing so the study suffers from a smaller sample size due to more strict inclusion criteria leading to an underpowered study that may be too small to recognize a small but significant survival difference.

To better understand whether myogenic expression is of prognostic significance in liposarcoma, Gronchi et al (2015)³ evaluated muscle marker immunohistochemistry in retroperitoneal WDLPS and DDPLS. Similar to our study, this group included only patients treated at their institution to ensure uniformity of treatment-associated confounders. They evaluated muscle marker expression semiquantitatively and classified myogenic expression positivity as at least 1 myogenic marker was expression by at least 10% of neoplastic cells. However, their study included well-differentiated liposarcomas as well and found tumor grade and myogenic differentiation were significantly associated with a poor OS on multivariate analysis. However, the inclusion of both well-differentiated and DDLPS tumors that expressed myogenic markers in the “myogenic differentiation” subgroup is peculiar. In their study, DDLPS more commonly expressed myogenic markers than WDLPS and, not surprisingly, myogenic differentiation correlated with tumor grade. This may then enrich the “myogenic differentiation” group to higher-grade DDLPS and higher-grade tumors have a poorer OS than G1 WDLPS. Multicollinearity was not addressed in this paper. Several years earlier, Binh et al (2007)⁸ evaluated patients with DDLPS with and without myogenic differentiation and found no difference in 2- or 5-year local recurrence free survival, metastatic risk, or overall survival, even with rhabdomyosarcomatous differentiation. The number of patients they evaluated was much smaller and again might have suffered from the same issues outlined above on underpowered studies.

The strength of our study was in the uniformity of patient population; to avoid contamination by clinical variables, we narrowed our evaluation to only patients with retroperitoneal DDLPS treated primarily (primary resection and follow-up) at this institution. However, the correlate of this strength is the lower number of tumors we were able to evaluate. Therefore, if there is a small but significant difference in myogenic expression and survival, the smaller sample size and underpowered study may fail to recognize the difference (type II/beta error). This may underlie the fact this study did not detect a statistically significant prognostic effect of myogenic expression in the previously published results at a 10% expression level. We did find a statistically significant difference in DFS for SMA and DFS and OS at a combined myogenic score. The clinical applications of these mathematically derived cutoffs in stratifying survival groups was not significant enough to be clinically reliable, until perhaps digital image analysis and quantification becomes more widespread in clinical practice and can be utilized to more accurately and reliably classify patients into prognostically significant subgroups. One issue not addressed in this study is interobserver variability in the assessment of quality and quantity of immunohistochemical stains.

Next generation sequencing efforts and growing repertoire of large-scale genome, exome, transcriptome, etc. sequencing may prove to better classify prognostically significant tumor classes. Recently, favorable and unfavorable methylation patterns have found to have prognostic significance in DDLPS in preliminary studies.¹¹

Overall, this study evaluating the prognostic significance of myogenic expression in DDLPS did find a statistically significant effect of myogenic immunohistochemical expression on disease free and overall survival. However, the difference is slight and likely of little clinical usefulness. Further studies in large-scale transcriptome or proteome-based assays are likely to better identify focused and prognostically significant targets.