ORIGINAL ARTICLE

Predicting dedifferentiation in liposarcoma: a proteomicapproach

Colt M. McClain & David B. Friedman & Tahar Hajri &Cheryl M. Coffin & Justin M. M. Cates

Received: 12 February 2013 /Revised: 5 April 2013 /Accepted: 17 April 2013 /Published online: 26 May 2013# Springer-Verlag Berlin Heidelberg 2013

Abstract There are no known morphologic characteristics,cytogenetic aberrations, or molecular alterations predictiveof dedifferentiation in liposarcomas. Identification of such aprognostic marker could potentially affect surgical and ad-juvant therapy and/or follow-up surveillance for these pa-tients. Two-dimensional difference gel electrophoresis wasutilized to characterize protein expression patterns in lipo-ma, atypical lipomatous tumor (ALT), and the well-differentiated components of dedifferentiated liposarcoma(DDL). Protein spots were identified by peptidemapping/fingerprinting using matrix-assisted laser desorp-tion ionization time-of-flight mass spectrometry. No signif-icant differences in protein expression were identifiedbetween lipoma and ALT or DDL. Proteins that were sig-nificantly down-regulated in the well-differentiated compo-nent of DDL compared to ALT included mitochondrialaldehyde dehydrogenase 2 (ALDH2, >3-fold reduction)and selenium-binding protein-1 (SELENBP1, >4-fold re-duction). Subsequent validation studies were performed byimmunohistochemistry (IHC) on a separate series of ALT(n=30) and the well-differentiated components of DDL(n=28). IHC stains were evaluated in a semi-quantitative

manner, and the results were analyzed using the Mann–Whit-ney test and receiver–operator curve analysis. Decreased IHCstaining for SELENBP1 in the well-differentiated componentof DDL was confirmed. Cytoplasmic ALDH2 levels deter-mined by IHC were not significantly different in ALT andDDL; no nuclear staining for ALDH2 was observed. Expres-sion of SELENBP1 is decreased in the well-differentiatedcomponent of DDL compared to ALT. However, variabilityin the staining patterns in liposarcoma precludes its use as apredictive marker for dedifferentiation.

Keywords Liposarcoma . Dedifferentiation . Proteomics .

Selenium-binding protein-1 . Two-dimensional differencegel electrophoresis

Introduction

Atypical lipomatous tumor (ALT) is a locally aggressivelipogenic neoplasm with virtually no potential for distant me-tastasis. Nevertheless, contemporaneous (de novo or primary)or metachronous (secondary) dedifferentiation to anonlipogenic sarcoma with significant risk of distant metastasisoccurs in a subset of these patients. While some experts havesuggested that tumor progression is a time-dependent phenom-enon, the most reliable predictor of dedifferentiation remainsthe anatomic site of the tumor [1, 2]. Whereas the reportedincidence of dedifferentiation in ALT is <5% in tumors of thetrunk or extremities, approximately 20% of retroperitoneal oringuinoscrotal tumors dedifferentiate [2–10]. However, it is notpossible to predict which ALTs are capable of tumor progres-sion and dedifferentiation by light microscopic examination orcytogenetic analysis [8, 11, 12]. Furthermore, there are noknown molecular markers that signal an increased risk ofdedifferentiation in well-differentiated tumors [13, 14]. Suchdiagnostic markers would be clinically useful to identify pa-tients at high risk for dedifferentiation, as this cohort might

Electronic supplementary material The online version of this article(doi:10.1007/s00428-013-1416-2) contains supplementary material,which is available to authorized users.

C. M. McClain : C. M. Coffin : J. M. M. Cates (*)Department of Pathology, Microbiology and Immunology,Vanderbilt University Medical Center, 1161 21st Ave. South,Nashville, TN 37232, USAe-mail: justin.m.cates@vanderbilt.edu

D. B. FriedmanDepartment of Biochemistry and Mass Spectrometry ResearchCenter, Vanderbilt University Medical Center, Nashville, TN, USA

T. HajriDepartment of Surgery, Vanderbilt University Medical Center,Nashville, TN, USA

Virchows Arch (2013) 463:85–92DOI 10.1007/s00428-013-1416-2

benefit from more aggressive surgical or adjuvant radiationtherapy and more frequent surveillance [15].

Two-dimensional gel electrophoresis (2D-DIGE) was usedto compare protein expression patterns in lipoma, ALT, andthe well-differentiated components of dedifferentiatedliposarcoma (DDL). Although no proteins differentiallyexpressed in lipoma and ALTor DDL were identified, expres-sion levels of six proteins were found to be significantlydifferent in the well-differentiated components of DDL com-pared to ALT. Mass spectra signal intensities were sufficientfor identification of two of these proteins as mitochondrialaldehyde dehydrogenase 2 (ALDH2) and selenium-bindingprotein-1 (SELENBP1). Since these candidate proteinmarkers may be of diagnostic utility as indicators of increaseddedifferentiation potential and an aggressive clinical course inpatients initially presenting with ALT, validation studies wereperformed using immunohistochemical (IHC) stains on aseparate cohort of ALT and DDL.

Materials and methods

Protein isolation

Protein extracts were prepared from snap-frozen tumor sam-ples as previously described with slight modifications [16].Hematoxylin and eosin-stained slides from all cases includ-ed in this study were reviewed by a soft tissue pathologist(JMMC) and the diagnoses were confirmed usingestablished morphologic and cytogenetic criteria [1, 2];clinicopathologic data for the samples used in protein iso-lations and 2D-DIGE are presented in Table 1. Briefly,tissue samples from eight lipomas, eight ALT, and well-differentiated areas from three DDL were homogenized onice with an Omni Tip™ Homogenizing Kit (Omni Interna-tional, Kennesaw, GA, USA) in 1.0 ml of cold protein lysisbuffer (1% Triton X-100, 50 mM Tris–HCl [pH 7.6],150 mM NaCl) with Complete protease inhibitor cocktail(Roche Applied Science, Indianapolis, IN, USA). Tissuehomogenates were centrifuged at 12,000 rpm for 12 min at4 °C, and the protein content of the supernatants was mea-sured using a Bradford-based protein assay (Bio-Rad Labo-ratories, Hercules, CA, USA).

Two-dimensional difference gel electrophoresis

The mixed internal standard methodology of Friedman et al.was used as previously described [17, 18]. Each sample wasco-resolved with the same pooled sample internal standardon 24-cm pH 4–7 isoelectric focusing (IEF) gradientscontaining 500 μg of total protein. The DeCyder suite ofDIGE software tools (GE Healthcare, Life Sciences,Piscataway, NJ, USA) was used to quantitatively compare

the abundance of each protein from each sample relative tothe unique Cy2-signal for that protein from the internalstandard within each gel and normalize quantitative abun-dance values for each protein between gels. Statistical con-fidence was associated with each change in abundance orcharge-altering posttranslational modification using analysisof variance (ANOVA) and post hoc Student’s t tests. DIGEgels were post-stained with SYPRO Ruby (MolecularProbes Inc., Eugene, OR, USA) to ensure accurate proteinexcision. An automated Spot Handling Workstation(Amersham Biosciences, Sunnyvale, CA, USA) was usedto robotically excise proteins from the gels and process themfor in-gel protease digestion and mass spectrometry. Pro-teins were digested with trypsin protease within the gelplugs. Peptides were extracted from the gel plugs and mixedwith α-cyano-4-hydroxycinnamic acid (5 mg/ml in 60%acetonitrile, supplemented with 1 mg/ml ammonium citrate)and spotted onto a stainless steel target for subsequent massspectral analysis.

Protein identification

Matrix-assisted laser desorption ionization, time-of-flight(MALDI-TOF) mass spectrometry was used to acquire themasses of the intact molecular peptide ions to within 20 ppm(peptidemassmapping), andMALDI-TOF/TOF tandemmassspectrometry was used to fragment selected ions to generateamino acid sequence information, both using a Voyager 4700mass spectrometer (Applied Biosystems, Foster City, CA,USA). Fragmentation spectra were acquired in a data-dependent fashion based on the MALDI-TOF peptide massmap for each protein and internally calibrated using trypsinautolytic fragments present in the samples to provide massaccuracy within 20 ppm. Both types of mass spectral datawere collectively used to interrogate protein databases togenerate statistically significant candidate identificationsusing GPS Explorer software (Applied Biosystems) runningthe MASCOT v1.9 database search algorithm (Matrix Sci-ence, Boston, MA, USA). MOlecular Weight SEarch(MOWSE) scores (−10 ⋅ log(P), where P is the probabilitythat the observed match is a random event) are represented asA, B(x), C, D, where A = combined MS and MS/MS searchscores, B(x) = number of peptides matched (number ofunmatched peptides), C = number of peptides with MS/MSdata, and D = percent of amino acids accounted for by thematching peptides (protein coverage). First-pass searcheswere performed against the SWISS PROT and/or the NCBInrdatabases without constraining protein molecular weight orisoelectric point or species. Carbamidomethylation of cysteine(reduction and alkylation with DTT and iodoacetamide) wasperformed and partial oxidation of methionine residues wasalso allowed in the search parameters. Protein scores >55 arewithin the 95th percentile confidence interval (P<0.05).

86 Virchows Arch (2013) 463:85–92

Immunohistochemistry

A separate set of 58 samples of ALT (n=30) and the well-differentiated components of DDL (n=28) was used for thevalidation study. Formalin-fixed, paraffin-embedded tissue sec-tions were stained immunohistochemically for ALDH2 andSELENBP1. Slides were deparaffinized and placed on theLeica BOND-MAX IHC autostainer (Leica Biosystems,Buffalo Grove, IL, USA). Heat-induced antigen retrievalwas performed using Epitope Retrieval 2 solution (LeicaBiosystems) for 20 min. The sections were separatelyincubated with anti-ALDH2 (Epitomics, Inc., Burlin-game, CA, USA) diluted 1:500 or anti-SELENBP1 (Sig-ma-Aldrich Co., St. Louis, MO, USA) diluted 1:100 for1 h. The Bond Refine Polymer detection system (LeicaBiosystems) was used for visualization. IHC-stainedslides were scored for both nuclear and cytoplasmic inten-sity (0, negative; 1+, weak; 2+, moderate; 3+, strong staining)and distribution (0, negative; 1+, <1 %; 2+, 1–10 %; 3+,11–33 %; 4+, 34–66 %; 5, >66 % of cells staining). Thesetwo scores were summed to generate an IHC index.

Eleven (39%) of the 28 DDL were FNCLCC grade 2 and17 (61%) were grade 3. In 24 of the 28 cases (86 %), thededifferentiated component was present in the initial surgicalresection specimen (primary or de novo DLL). In the fourothers, the median time to dedifferentiation was 2.9 years(range, 0.3–13.7 years). Other clinicopathologic characteris-tics are listed in Table 2. The study protocol was approved bythe Institutional Review Board at Vanderbilt University.

Statistics

Average volume ratios were calculated and statistical anal-ysis of 2D-DIGE data was performed using DeCyder 2DSoftware v6.5 (GE Healthcare, Life Sciences). Categoricaldata were compared using Fisher’s exact test. Ordinal andcontinuous data were compared using the Mann–Whitneytest. The utility of the IHC markers as differential diagnosticmarkers was assessed by receiver–operator curve (ROC)analysis. Kaplan–Meier metastasis-free survival curves werecompared using log-rank tests, and statistical significancewas defined as α=0.05. Other than statistical analysis of the2D-DIGE data, statistical calculations were performed usingStata v12.1 (StataCorp, College Station, TX, USA).

Results

Two-dimensional difference gel electrophoresis

2D-DIGE using different cyanine fluorescent dyes and amixed internal standard allowed direct comparison and relativequantification of specific proteins among lipoma, ALT, andDDL samples resolved together on the same IEF gels(Fig. 1). Twenty-nine of 69 spots of interest were tentativelyidentified as unique proteins by peptide mapping/fingerprintingusing MALDI-TOF/TOF tandem mass spectrometry and data-base interrogation (Supplementary Table 1). Post hoc t testscomparing ALT and DDL with lipoma identified only three

Table 1 Samples used in2D-DIGE analysis

2D-DIGE two-dimensional dif-ference gel electrophoresis, ALTatypical lipomatous tumor, DDLdedifferentiated liposarcoma, nadata not available, M male, Ffemale, MDM2− ratio of MDM2(12q13) and CEP12 signals <2.0by fluorescence in situ hybridi-zation, MDM2+ ratio of MDM2(12q13) and CEP12 signals ≥2.0by fluorescence in situhybridization

Diagnosis Age/sex Anatomic site Cytogenetic analysis

Lipoma na/na Superficial na

Lipoma na/F Intramuscular 46,XX

Lipoma na/M Intramuscular 46,XY, t(1;12)

Lipoma 54/M Intramuscular 46,XY, t(1;12), del13, +8

Lipoma 18/F Superficial 46,X, t(X;8)

Lipoma 66/F Intramuscular 46,XX, t(4;12)

Lipoma 57/F Intramuscular MDM2−

Lipoma 39/F Intramuscular MDM2−

ALT na/M Intramuscular 46,X,−Y, +r

ALT 49/F Intramuscular 46,XX, add(12)(q13)

ALT 65/M Intramuscular MDM2+

ALT 69/F Intramuscular na

ALT 66/M Intramuscular +r, +mar

ALT 59/M Intramuscular MDM2+

ALT 59/M Retroperitoneum +r, +mar

ALT 70/F Intramuscular MDM2+

DDL 54/F Retroperitoneum 57,XX, add(12)(q24.3), +11mar

DDL 85/M Retroperitoneum 67-76,XY, add12(p12), +8 ~ 12mar, 3-6dmin

DDL 61/M Retroperitoneum +r, +mar

Virchows Arch (2013) 463:85–92 87

proteins with P values≤0.005, two of which represented cleav-age fragments of actin; the other was present in very lowabundance and could not be identified. In contrast, six proteinswere identified as differentially expressed in the well-differentiated component of DDL compared to ALT. Massspectra signal intensities were sufficient for tentative identifica-tion of three of these proteins. Levels of ALDH2 (Fig. 1, spot#20; 56.8 kDa; pI 6.6) and SELENBP1 (Fig. 1, spot #21;52.9 kDa; pI 5.93) were significantly lower in the well-differentiated component of DDL compared to ALT (Fig. 2).The third protein, ubiquitin carboxyl-terminal hydrolase iso-zyme L1, was not investigated further as the MOWSE scoresfrom protein identification were marginal. The low signalintensities generated from several of these candidate bio-markers most likely indicate that these proteins are of lowabundance.

Fig. 1 Two-dimensional difference gel electrophoresis map. Compar-ison of the average intensities of protein spots among the three diag-nostic groups by one-way ANOVA identified six proteins of interest(P≤0.002). The amount of protein extracted for identification analysisof three of these proteins was insufficient, suggesting that they are verylow abundance proteins. Two of the proteins identified were mitochon-drial aldehyde dehydrogenase 2 (spot #20; 56.8 kDa; pI 6.6) andselenium-binding protein-1 (spot #21; 52.9 kDa; pI 5.93), the levelsof which were significantly decreased in the well-differentiated com-ponents of dedifferentiated liposarcoma compared to atypical lipoma-tous tumor and lipoma

Fig. 2 Two-dimensional difference gel electrophoresis profiles gener-ated using DeCyder software. Graphs depict normalized log abundanceratios of mitochondrial aldehyde dehydrogenase 2 (ALDH2) and sele-nium-binding protein-1 (SELENBP1) in lipoma, atypical lipomatoustumor (ALT), and the well-differentiated component of dedifferentiatedliposarcoma (DDL) relative to the cognate signals present in the pooledinternal standard specific for each protein. Levels of ALDH2 weresignificantly decreased in DDL compared to ALT (>3-fold; Student’s ttest, P=2 ⋅ 10−3). The abundance profile of SELENBP1 demonstrated4-fold reduction in the well-differentiated component of DDL com-pared to ALT (Student’s t test, P=7 ⋅ 10−4)

Table 2 Clinicopathologic characteristics of the lipomatous tumorsused in validation study

ALT (n=30) DDL (n=28) Pa

Age (years) 64 (39–89) 57 (31–91) 0.28

Gender 0.06

Male 15 (50 %) 21 (75 %)

Female 15 (50 %) 7 (25 %)

Follow-up (years) 4.7 (0.01–10.7) 1.2 (0.1–20) 0.04

Died of disease 1 (3 %) 4 (14 %) 0.19

Local recurrence 6 (20 %) 11 (39 %) 0.15

Metastasis 0 (0 %) 11 (39 %) <0.001

Greatest dimension (cm) 17 (4.1–42.1) 14 (2.5–45.0) 0.45

Tumor volume (cm3) 503 (14–4,948) 471 (3–11,341) 0.94

Anatomic site <0.001

SQ/superficial 0 (0 %) 2 (7 %)

Intramuscular 23 (77 %) 6 (21 %)

Retroperitoneal 3 (10 %) 17 (61 %)

Other 4 (13 %) 3 (11 %)

ALT atypical lipomatous tumor, DDL dedifferentiated liposarcoma, SQsubcutaneousa Categorical data were compared using Fisher’s exact test; continuousdata are presented as median (range) and were compared using theMann–Whitney test

88 Virchows Arch (2013) 463:85–92

Immunohistochemistry

A separate set of archived, formalin-fixed, paraffin-embedded tumor samples of ALT and DDL was stainedfor ALDH2 and SELENBP1 by IHC (Fig. 3). The medianIHC indices for both nuclear and cytoplasmic SELENBP1were significantly decreased in the well-differentiated com-ponents of DDL compared to ALT (Table 3). In contrast, thecytoplasmic IHC index for ALDH2 was not significantly

different in these tumor subtypes. As expected, no nuclearstaining for mitochondrial ALDH2 was detected.

Log-rank tests of metastasis-free survival curves demon-strated that at the optimal cut-off points determined by ROCanalysis, decreased cytoplasmic and nuclear SELENBP1staining was associated with shorter metastasis-free intervals(log-rank test, P=0.0065 and P=0.05, respectively). Cyto-plasmic ALDH2 was not associated with metastasis-freesurvival. Too few deaths occurred in this cohort over the

Fig. 3 Representativeimmunohistochemical stains forALDH2 (a) and SELENBP1(b) in atypical lipomatous tumorand the well-differentiatedcomponents of dedifferentiatedliposarcoma

Virchows Arch (2013) 463:85–92 89

follow-up period to perform overall survival analysis. ROCanalysis disclosed that cytoplasmic ALDH2 IHC index cor-rectly classified a liposarcoma as dedifferentiated in only 60% of cases (sensitivity 68%; specificity 53%). Both nuclearand cytoplasmic SELENBP1 IHC indices failed to performas well as ALDH2 in ROC analysis. Thus, although proteinexpression levels and IHC indices for SELENBP1 are lowerin the well-differentiated components of DDL and are asso-ciated with decreased metastasis-free survival, the variabil-ity of IHC staining patterns in ALT and DDL precludes theuse of this marker as a predictor of dedifferentiation inliposarcoma (Fig. 4).

Interestingly, the median IHC index of SELENBP1 wassignificantly decreased in primary, de novo DDL compared

to metachronous, secondary DDL (cytoplasmic staining,0 vs. 6.5, P=0.026; nuclear staining, 3.0 vs. 7.5, P=0.031,Mann–Whitney test); levels of these proteins in ALT andsecondary DDL were similar. This suggested to us thatsecondary DDL might be less aggressive than primaryDDL. Indeed, metastasis-free survival (defined as the inter-val from the date of dedifferentiation) is significantly re-duced for patients with primary DDL (52 % 5-year survival)compared to those with secondary DDL (100 % 5-year sur-vival, log-rank test, P=0.049); tumor location, tumor size,FNCLCC grade, patient age, and gender were not significant-ly associated with the duration of metastasis-free survival.

Discussion

Although the majority of ALTs do not progress, a subsetundergoes dedifferentiation to an aggressive, intermediate tohigh grade sarcoma with significant risk of distant metasta-sis and disease-related mortality [2–10]. Other than the siteof origin, there are no clinicopathologic characteristics thatidentify ALT with an increased potential for dedifferentia-tion and tumor progression [13, 14]. Several biomarkershave been studied in attempts to identify the molecularmechanisms of dedifferentiation. Such candidate proteinsare of interest as diagnostic/prognostic markers to help predictthe risk of dedifferentiation, as this cohort might benefit frommore aggressive surgical or adjuvant radiation therapy andmore frequent surveillance [15]. In addition, these markersmight represent potential pharmacotherapeutic targets.

Previous studies failed to show differences in the expres-sion patterns of calreticulin, TP53, MDM2, CDKN1A, pRb,loss of heterozygosity at the RB1 locus, or Ki67 indices inALT and the well-differentiated components of DDL[19–21]. However, recent array-based studies have demon-strated that ALT and the well-differentiated components ofDDL show distinct comparative genomic hybridization andgene expression profiles [22–24]. Amplification andoverexpression of the proto-oncogene JUN has been pro-posed as a mechanism of dedifferentiation through inhibi-tion of lipogenesis and adipocytic maturation [25, 26].Indeed, the JUN oncoprotein is detected more often in thewell-differentiated components of DDL than in ALT (59 vs.27%) [27]. These findings suggest that although ALT andthe well-differentiated components of DDL are histopatho-logically indistinguishable, molecular markers might be ableto predict which ALT are capable of dedifferentiation.

In this study, proteomic profiles of lipoma, ALT, and thewell-differentiated components of DDL were comparedusing 2D-DIGE. Interestingly, only one protein was identi-fied as significantly increased in ALT compared to lipomausing this technique. Unfortunately, the levels of this proteinwere too low for accurate identification. Limited differences

Table 3 Mitochondrial aldehyde dehydrogenase (ALDH2) and sele-nium-binding protein-1 (SELENBP1) in atypical lipomatous tumorand the well-differentiated components of dedifferentiated liposarcoma

ALT DDL Pa

Cytoplasmic ALDH2 index 3 (0–8) 4 (3–6) 0.42

Nuclear ALDH2 index 0 0 na

Cytoplasmic SELENBP1 index 5 (0–6) 0 (0–4.5) 0.009

Nuclear SELENBP1 index 6 (4–8) 5 (3.5–5.5) 0.02

ALT atypical lipomatous tumor, DDL dedifferentiated liposarcoma,ALDH2 mitochondrial aldehyde dehydrogenase, SELENBP1 seleni-um-binding protein-1, na not applicablea Data are presented as median (interquartile range) and comparedusing the Mann–Whitney test

Fig. 4 Distributional plots of immunohistochemical indices forALDH2 and SELENBP1. Median immunohistochemical indices forALDH2 were not significantly different between atypical lipomatoustumor and the well-differentiated components of dedifferentiatedliposarcoma. Although the median nuclear and cytoplasmicSELENBP1 immunohistochemical indices were significantly lower inthe well-differentiated components of dedifferentiated liposarcomacompared to atypical lipomatous tumor, there was significant variationin the staining indices for both lipomatous tumors

90 Virchows Arch (2013) 463:85–92

in the RNA expression patterns of lipoma and ALT havealso been reported [23]. These data suggest that thetranscriptome and the proteome of lipoma and ALT arehighly similar, as one might expect given the morphologicsimilarities between these tumors. Molecular differencesbetween lipoma and ALT may be limited to low abundanceregulatory proteins not detectable by conventional microar-ray or DIGE methods. Proteins known to be overexpressedin ALT compared to lipoma, such as MDM2, CDK4, andCDKN2A, were not detected by 2D-DIGE, possibly be-cause these proteins are “masked” by higher abundanceproteins of similar molecular weights and isoelectric pointswithin the 2D gels or because the levels of these proteins arebelow the sensitivity of the methodology.

In contrast, six proteins were differentially expressed inthe well-differentiated components of DDL. Mass spectrasignal intensities were sufficient for confident identificationof two of these proteins as ALDH2 and SELENBP1.ALDH2 is an instrumental enzyme in the oxidation ofalcohols and plays an important role in protecting the cellagainst oxidative stress and DNA damage caused by alde-hyde derivatives [28, 29]. Although the physiological role ofSELENBP1 is not completely understood, low levels of thisselenoprotein in many carcinomas suggest that this proteinhas tumor suppressor effects [30–32]. Noteworthy in thisrespect is the finding that the levels of both these proteinswere significantly lower in the well-differentiated compo-nents of DDL, a tumor with greater malignant potential thaneither ALT or lipoma.

Although the reduced levels of ALDH2 and SELENBP1in well-differentiated components of DDL detected byDIGE were statistically significant, only SELENBP1showed significant differences in cytoplasmic and nuclearexpression as determined by IHC. However, there is sub-stantial overlap in the IHC profiles of SELENBP1 in thesetumor subtypes, the degree of which precludes the use ofSELENBP1 as a diagnostic/prognostic biomarker.

It was also noted that nuclear and cytoplasmicSELENBP1 levels were significantly decreased in thewell-differentiated components of primary, de novo DDLcompared to secondary, metachronous DDL and that theIHC indices between ALT and the well-differentiated com-ponents of secondary DDL were comparable. Thissuggested that ALT and secondary, metachronous DDL aremore similar to each other than to the well-differentiatedareas of primary DDL. About 85–90% of all DDL arise inassociation with a synchronous well-differentiated compo-nent, whereas the remainder present metachronously as alocally recurrent or metastatic tumor [10, 33–35]. Since themetastatic rate of primary DDL appears to be significantlygreater than that of secondary DDL (relative risk=3.7) [35],we assessed whether the timing of dedifferentiationinfluenced the interval to development of distant metastasis.

Although median metastasis-free survival was not reachedfor either cohort during this follow-up interval, the propor-tion of patients with metastatic disease at 5 years follow-upwas markedly higher for those with primary DDL comparedto secondary DDL (48 vs. 0 %). Thus, DDLs in which thededifferentiated component arises in association with theprimary tumor appear to be inherently more aggressive neo-plasms than those with metachronous dedifferentiation.

In summary, 2D-DIGE identified two proteins signifi-cantly down-regulated in the well-differentiated componentof DDL compared to ALT, one of which was confirmed inIHC validation studies. Overlap in the protein expressionpatterns of SELENBP1 in these tumor subtypes prohibitsusing this biomarker as a predictor for tumor progressionand an aggressive clinical course in patients who initiallypresent with ALT without evidence of synchronous dedif-ferentiation. However, this may not necessarily be a clini-cally important distinction, as these patients seem to have amuch better prognosis than those who present with synchro-nous dedifferentiated and well-differentiated components ofliposarcoma in the primary neoplasm. 2D-DIGE identifiedother low abundance proteins differentially expressed inDDL that could not be identified by MALDI-TOF massspectrometry on the samples currently available for analysis.These proteins may represent important regulatory proteins,and additional studies may yield sufficient material to en-able peptide matching and subsequent characterization.

Acknowledgments This work was funded by the Shelby L. RichterMemorial Research Award (JMMC) granted by the Sarcoma Founda-tion of America.

Conflict of interest The authors declare that they have no financialrelationship with the Sarcoma Foundation of America or other conflictsof interest.

References

1. Dei Tos A, Marino-Enriquez A, Pedeutour F, Rossi S (2013)Dedifferentiated liposarcoma. In: Fletcher CDM, Bridge JA,Hogendoorn PCW, Mertens F (eds) WHO classisification of tu-mours of soft tissue and bone. IARC, Lyon, pp 37–38

2. Dei Tos AP, Pedeutour F (2013) Atypical lipomatous tumour. In:Fletcher CDM, Bridge JA, Hogendoorn PCW, Mertens F (eds)WHO classisification of tumours of soft tissue and bone. IARC,Lyon, pp 33–36

3. Azumi N, Curtis J, Kempson RL, Hendrickson MR (1987)Atypical and malignant neoplasms showing lipomatous differenti-ation: a study of 111 cases. Am J Surg Pathol 11:161–183

4. Evans HL, Soule EH, Winkelmann RK (1979) Atypical lipoma,atypical intramuscular lipoma, and well differentiated retroperito-neal liposarcoma: a reappraisal of 30 cases formerly classified aswell differentiated liposarcoma. Cancer 43:574–584

5. Kindblom LG, Angervall L, Fassina AS (1982) Atypical lipoma.Acta Pathol Microbiol Immunol Scand A 90:27–36

Virchows Arch (2013) 463:85–92 91

6. Kooby DA, Antonescu CR, Brennan MF, Singer S (2004) Atypicallipomatous tumor/well-differentiated liposarcoma of the extremityand trunk wall: importance of histological subtype with treatmentrecommendations. Ann Surg Oncol 11:78–84

7. Lucas DR, Nascimento AG, Sanjay BK, Rock MG (1994) Well-differentiated liposarcoma: the Mayo Clinic experience with 58cases. Am J Clin Pathol 102:677–683

8. Rosai J, Akerman M, Dal Cin P et al (1996) Combined morpho-logic and karyotypic study of 59 atypical lipomatous tumors:evaluation of their relationship and differential diagnosis withother adipose tissue tumors (a report of the CHAMP study group).Am J Surg Pathol 20:1182–1189

9. Rozental TD, Khoury LD, Donthineni-Rao R, Lackman RD(2002) Atypical lipomatous masses of the extremities: outcomeof surgical treatment. Clin Orthop Relat Res 398:203–211

10. Weiss SW, Rao VK (1992) Well-differentiated liposarcoma (atypicallipoma) of deep soft tissue of the extremities, retroperitoneum, andmiscellaneous sites: a follow-up study of 92 cases with analysis of theincidence of "dedifferentiation". Am J Surg Pathol 16:1051–1058

11. Fletcher CD, AkermanM,Dal Cin P et al (1996) Correlation betweenclinicopathological features and karyotype in lipomatous tumors: areport of 178 cases from the Chromosomes and Morphology(CHAMP) Collaborative Study Group. Am J Pathol 148:623–630

12. Dal Cin P, Kools P, Sciot R, De Wever I, Van Damme B, Van DeVen W, Van Den Berghe H (1993) Cytogenetic and fluorescence insitu hybridization investigation of ring chromosomes characteriz-ing a specific pathologic subgroup of adipose tissue tumors.Cancer Genet Cytogenet 68:85–89

13. Sandberg AA (2004) Updates on the cytogenetics and moleculargenetics of bone and soft tissue tumors: liposarcoma. Cancer GenetCytogenet 155:1–24

14. Pilotti S, Della Torre G, Lavarino C, Di Palma S, Sozzi G,Minoletti F, Rao S, Pasquini G, Azzarelli A, Rilke F, PierottiMA (1997) Distinct mdm2/p53 expression patterns in liposarcomasubgroups: implications for different pathogenetic mechanisms. JPathol 181:14–24

15. Dalal KM, Antonescu CR, Singer S (2008) Diagnosis and man-agement of lipomatous tumors. J Surg Oncol 97:298–313

16. Elizalde M, Ryden M, Van Harmelen V, Eneroth P, GyllenhammarH, Holm C, Ramel S, Olund A, Arner P, Andersson K (2000)Expression of nitric oxide synthases in subcutaneous adipose tis-sue of nonobese and obese humans. J Lipid Res 41:1244–1251

17. Lilley KS, Friedman DB (2004) All about DIGE: quantificationtechnology for differential-display 2D-gel proteomics. Expert RevProteomics 1:401–409

18. Friedman DB, Hill S, Keller JW, Merchant NB, Levy SE, CoffeyRJ, Caprioli RM (2004) Proteome analysis of human colon cancerby two-dimensional difference gel electrophoresis and mass spec-trometry. Proteomics 4:793–811

19. Adachi T, Oda Y, Sakamoto A, Saito T, Tamiya S, Masuda K,Tsuneyoshi M (2001) Immunoreactivity of p53, mdm2, andp21WAF1 in dedifferentiated liposarcoma: special emphasis onthe distinct immunophenotype of the well-differentiated compo-nent. Int J Surg Pathol 9:99–109

20. Schneider-Stock R, Boltze C, Jaeger V, Stumm M, Seiler C, Rys J,Schutze K, Roessner A (2002) Significance of loss of heterozygosity

of the RB1 gene during tumour progression in well-differentiatedliposarcomas. J Pathol 197:654–660

21. Hisaoka M, Matsuyama A, Nakamoto M (2012) Aberrantcalreticulin expression is involved in the dedifferentiation ofdedifferentiated liposarcoma. Am J Pathol 180:2076–2083

22. Chibon F, Mariani O, Derre J, Malinge S, Coindre JM, Guillou L,Lagace R, Aurias A (2002) A subgroup of malignant fibroushistiocytomas is associated with genetic changes similar to those ofwell-differentiated liposarcomas. Cancer Genet Cytogenet 139:24–29

23. Shimoji T, Kanda H, Kitagawa T et al (2004) Clinico-molecularstudy of dedifferentiation in well-differentiated liposarcoma.Biochem Biophys Res Commun 314:1133–1140

24. Coindre JM, Pedeutour F, Aurias A (2010) Well-differentiated anddedifferentiated liposarcomas. Virchows Arch 456:167–179

25. Mariani O, Brennetot C, Coindre JM, Gruel N, Ganem C, DelattreO, Stern MH, Aurias A (2007) Jun oncogene amplification andoverexpression block adipocytic differentiation in highly aggres-sive sarcomas. Cancer Cell 11:361–374

26. Tap WD, Eilber FC, Ginther C et al (2011) Evaluation of well-differentiated/de-differentiated liposarcomas by high-resolution ol-igonucleotide array-based comparative genomic hybridization.Genes Chromosomes Cancer 50:95–112

27. Snyder EL, Sandstrom DJ, Law K et al (2009) C-Jun amplificationand overexpression are oncogenic in liposarcoma but not alwayssufficient to inhibit the adipocytic differentiation programme. JPathol 218:292–300

28. Ohta S, Ohsawa I, Kamino K, Ando F, Shimokata H (2004)Mitochondrial ALDH2 deficiency as an oxidative stress. Ann NYAcad Sci 1011:36–44

29. Matsuda T, Yabushita H, Kanaly RA, Shibutani S, Yokoyama A(2006) Increased DNA damage in ALDH2-deficient alcoholics.Chem Res Toxicol 19:1374–1378

30. Raucci R, Colonna G, Guerriero E, Capone F, Accardo M, CastelloG, Costantini S (2011) Structural and functional studies of thehuman selenium binding protein-1 and its involvement in hepato-cellular carcinoma. Biochim Biophys Acta 1814:513–522

31. Jerome-Morais A, Me W, Liu R, Yang W, Jackson MI, Combs GFJr, Diamond AM (2012) Inverse association between glutathioneperoxidase activity and both selenium-binding protein 1 levels andGleason score in human prostate tissue. Prostate 72:1006–1012

32. Jeong JY, Wang Y, Sytkowski AJ (2009) Human selenium bindingprotein-1 (hSP56) interacts with VDU1 in a selenium-dependentmanner. Biochem Biophys Res Commun 379:583–588

33. Henricks WH, Chu YC, Goldblum JR, Weiss SW (1997)Dedifferentiated liposarcoma: a clinicopathological analysis of155 cases with a proposal for an expanded definition of dediffer-entiation. Am J Surg Pathol 21:271–281

34. McCormick D, Mentzel T, Beham A, Fletcher CD (1994)Dedifferentiated liposarcoma: clinicopathologic analysis of 32cases suggesting a better prognostic subgroup among pleomorphicsarcomas. Am J Surg Pathol 18:1213–1223

35. Ghadimi MP, Al-Zaid T, Madewell J, Peng T, Colombo C,Hoffman A, Creighton CJ, Zhang Y, Zhang A, Lazar AJ (2011)Diagnosis, management, and outcome of patients withdedifferentiated liposarcoma systemic metastasis. Ann SurgOncol 18:3762–3770

92 Virchows Arch (2013) 463:85–92