Materials and methods

Selection of cases

After approval of the Research Ethics Committee, 129 cases of soft tissue tumours were selected from the surgical pathology archives of the Santa Rita’s Hospital of Cancer, Porto Alegre, Brazil. The cases were studied at the Center for Research and Post-graduation Heitor Cirne Lima, Federal University for Health Sciences of Porto Alegre, Porto Alegre, Brazil. All tumours specimens were fixed in buffered formalin. Surgeries were held between 1997 and 2007 and initial diagnoses were obtained from pathology reports.

The specimens included incisional and excisional biopsies from primary, recurrent or metastatic tumours. All cases were reviewed and reclassified by two general surgical pathologists (PBA, AAH) according to the latest WHO classification of bone and soft tissue tumours.2 Selected controversial cases were reviewed by a known soft tissue pathologist (AMO). All cases had their diagnosis based in morphology. IHC and FISH analyses were performed in some cases to corroborate histopathological classification.

The tumours’ final diagnoses were as follows: WDLPS (n = 19), DDLPS (n = 10), spindle cell lipoma (SC-LP) (n = 3), pleomorphic lipoma (PL-LP) (n = 2), non-conventional lipoma (NC-LP) (n = 6), hibernoma (n = 1), timolipoma (n = 1), angiomyolipoma (n = 1), neural fibrolipoma (n = 1), intramuscular lipoma (n = 1), fibrolipoma (n = 1), angiosarcoma (n = 1), gastrointestinal stromal tumour (GIST) (n = 4), synovial sarcoma (SS) (n = 8), clear cell sarcoma (n = 1), pleomorphic liposarcoma (PLPS) (n = 2), myxoid-round cell liposarcoma (MRLPS) (n = 4), malignant peripheral nerve sheath tumour (MPNST) (n = 7), fibrosarcoma (FBS) (n = 3), leiomyosarcoma (LMS) (n = 8), Kaposi sarcoma (n = 2), epithelioid sarcoma (EPITS) (n = 3), dermatofibrosarcoma protuberans (DFSP) (n = 1), high grade undifferentiated pleomorphic sarcoma (UPS) (n = 26), epithelioid hemangioendothelioma (n = 1), osteosarcoma (n = 3), rhabdomyosarcoma (RMS) (n = 2), mixofibrosarcoma (MXFBS) (n = 5), low grade fibromyxoid sarcoma (LGFMS) (n = 1) and solitary fibrous tumour (SFT) (n = 1).

Immunohistochemistry

IHC staining for MDM2/CDK4 was performed on representative paraffin blocks. Sections (4 μm thick) were cut and mounted on previously silanised glass slides. The slides were dried for 60 minutes at 60°C and were left resting overnight at room temperature. Next morning, the sections were deparaffinised in xylene and rehydrated with ethanol. They were then treated with citrate buffer (0.01 M citric acid, pH 6.0), and heat-based antigen retrieval was carried out in a hot bath at 94°C for 40 minutes. Endogenous peroxidase was blocked using 95 ml of methanol plus 5 ml of 3% hydrogen peroxide solution. Preparations were washed in distilled water and phosphate-buffered saline solution. Unspecific protein binding was blocked with 1% bovine serum albumin for 30 minutes. Primary antibodies MDM2 (clone 1B10, dilution 1:50, Novocastra) and CDK4 (clone DCS-31, dilution 1:200, Biosource International) were incubated overnight at 4°C on a level surface and in a dark humid chamber. Next day, the sections were stained by the streptavidin–biotin–peroxidase method using the LSAB kit plus HRP system (DAKO), according to the manufacturer’s directions. To reveal the reaction a diaminobenzidine solution was employed and counterstaining was made with haematoxylin. A well characterised DDLPS case was used as a positive control for both antibodies in each staining run. Negative controls were obtained by omitting the primary antibodies.

Both antibodies react inside the nucleus, so any specific nuclear weak to strong immunostaining was evaluated by microscopy. The cases were stratified into four categories of staining pattern: (1) diffuse positivity, when >30% of cells were strongly immunoreactive; (2) moderate positivity, when 10–29% of cells were moderate to strongly immunoreactive; (3) focal positivity, when <10% of cells were immunoreactive; and (4) negative, when no cell was immunoreactive. A percentage of tumour cell reactivity was determined, approximately, by eyeballing the slides and counting 100 tumour cells, at medium/high power, in at least five different fields. A cut-off point of more than 10% moderate to strong nuclear immunoreactivity on the tumour cells was defined as a positive expression of MDM2/CDK4. Stains were evaluated individually by two independent pathologists (PBA and AAH) and discordant cases were revised together. The staining pattern categorisation was developed by the authors’ own experience.

Fluorescence in situ hybridisation

FISH analysis was performed on paraffin-embedded tissue sections with appropriate controls as previously described26 using MDM2 and CPM amplification probes. Three cases were selected for FISH because they were difficult to diagnose by morphology alone.

Statistical analysis

The cases were divided in four groups of study: WDLPS, DDLPS, benign adipocytic tumours (BAT), and other mesenquimal tumours (OMT). Sensitivity, specificity, predictive values and accuracy of positive expression for MDM2/CDK4 proteins was determined in WDLPS/DDLPS groups. Additionally, results were examined when only diffuse immunoreactivity was considered to define a case as positive. The tested hypothesis expected positive nuclear immunostaining for MDM2/CDK4 on WDLPS/DDLPS and negative staining for these immunomarkers on the BAT/OMT. Sensitivity for MDM2/CDK4 was calculated as the number of WDLPS/DDLPS cases that tested positive, divided by the total number of cases in each group. Specificity for MDM2/CDK4 was calculated by dividing the number of BAT/OMT cases that tested negative, divided by the total number of cases in each group. The diagnostic test efficacy was measured by calculating the κ and the Youden index (J) value.

Results

Tables 1–6 present the clinical and pathological data, histological diagnoses, immunohistochemistry results and statistical analyses of all tumour groups.

Well differentiated liposarcoma (tables 1 and 4)

MDM2 immunoreactivity was seen in all 19 cases (100%) of WDLPS (fig 1D,F). The positivity was strong and diffuse in 17/19 (89.4%) cases. Only two cases showed expression in no more than 30% of the neoplastic cells. Immunoreactivity for CDK4 was observed in 13/19 cases (68.4%); it was strong and diffuse in 12/13 cases (92.3%) (fig 2A,D–F). Both MDM2 and CDK4 protein expression was seen in 13/19 (68.4%) cases, being both strong and diffuse in 11/13 cases (84.6%). The case that showed moderate positivity for CDK4 (case 106) stained strong and diffuse for MDM2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

MDM2 immunostainings. (A) Case 29: diffuse immunostaining in a retroperitoneal low grade dedifferentiated liposarcoma (DDLPS). (B) Case 23: a non-conventional lipoma stained focally in reactive cells. (C) Case 70: well differentiated area of a DDLPS. (D) Case 27: sclerosing retroperitoneal well differentiated liposarcoma (WDLPS). (E) Case 13: multinucleated giant tumoural cells in a high-grade DDLPS. (F) Case 17: 48-year-old woman; sclerosing retroperitoneal WDLPS.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

CDK4 immunostainings. (A) Case 75: lipoblast in a retroperitoneal sclerosing well differentiated liposarcoma (WDLPS). (B) Case 29: diffuse immunostaining in a retroperitoneal dedifferentiated liposarcoma (DDLPS). (C) Case 70: well differentiated area of a DDLPS. (D) Case 27: sclerosing retroperitoneal WDLPS. (E) Case 18: WDLPS adjacent to renal parenchyma. (F) Case 17: 48-year-old woman; sclerosing retroperitoneal WDLPS.

Dedifferentiated liposarcoma (tables 1 and 4)

MDM2 immunoreactivity was seen in 9/10 cases (90%). The staining patterns were strong and diffuse in all 9 cases (100%) (fig 1A,C,E). Immunoreactivity for CDK4 was observed in 7/10 cases (70%) and it was strong and diffuse in all cases (100%) (fig 2B,C). Both MDM2 and CDK4 protein immunoreactivity was seen in seven cases (70%), being strong and diffuse in all cases (100%).

Benign adipocytic neoplasms (tables 2 and 4)

A deep-seated lipoma (case 12) did not show MDM2 gene amplification by FISH (Mayo Clinic, Minnesota, USA). One hibernoma (case 35) and one PL-LP did not show CPM gene amplification by FISH (Mayo Clinic). MDM2 immunoreactivity was seen in 7/17 (41.1%) tumours. In one case the immunoreactivity pattern was strong and diffuse (case 110). The remaining six positive cases (85.7%) showed moderate immunoreactivity (cases 23, 25, 49, 55, 76 and 113) (fig 1B). Immunoexpression for CDK4 was observed in two (11.7%) of the 17 cases; it was not strong and diffuse in any case. One PL-LP (case 3) and one SC-LP (case 49) stained in no more than 30% of the neoplastic cells. Only case 49 expressed both MDM2 and CDK4 of all 17 benign adipose tissue tumours (5.8%); both proteins were expressed in moderate levels.

Other mesenquimal tumours (tables 3 and 4)

MDM2 immunoreactivity was seen in 29/83 (34.9%) cases from the OMT group. The immunoreactivity pattern was strong and diffuse in 19/29 (65.5%) positive cases. Ten positive cases (34.4%) showed moderate immunoexpression in no more than 30% of the neoplastic cells. Immunoreactivity for CDK4 was seen in 3 (3.6%) of the 83 cases (cases 21, 40 and 43); it was strong and diffuse in only one case (case 21). Two cases (cases 40 and 43) exhibited moderate positivity in no more than 30% of the neoplastic cells. Both MDM2 and CDK4 protein immunoexpression was seen in 2/83 (2.4%) tumours (cases 40 and 43); the immunostaining pattern was moderate in both cases.

Statistical analysis (tables 5 and 6)

MDM2 and CDK4 were accurate in detecting WDLPS/DDLPS among other soft tissue tumours. Sensitivity and specificity for MDM2/CDK4 immunoexpression in WDLPS/DDLPS was 96.5% and 64%, and 68.9% and 95%, respectively. The combination of MDM2/CDK4 immunoreactivity increased the specificity of the test (97%) (κ = 0.71; J = 0.66). When considering a case positive with only strong and diffuse immunoexpression, the test gained in specificity (80% and 99%, respectively). In combination, diffuse positivity for MDM2/CDK4 achieved the best specificity (100%) (κ = 0.71; J = 0.62).

In distinguishing WDLPS from BAT, the sensitivity and specificity for MDM2/CDK4 immunoexpression was 100% and 58.8%, and 68.4% and 88.2%, respectively. With the combination of MDM2/CDK4 the specificity increased (94.1%), with no change increase in sensitivity (κ = 0.61; J = 0.62). When considering only diffuse positivity, MDM2 and CDK4 alone gained in specificity (94.1% and 100%, respectively) (κ = 0.83; J = 0.83). When combining MDM2/CDK4, diffuse immunoreactivity gained in specificity (100%) (κ = 0.56, J = 0.57).

In detecting DDLPS among OMT the sensitivity and specificity for MDM2/CDK4 positive expression was 90% and 65%, and 70% and 96.3%, respectively. With the combination of MDM2/CDK4 the specificity of the panel increased (97.5%) (κ = 0.7; J = 0.67). When considering only diffuse positivity for MDM2/CDK4, the test gained in specificity (77.1% and 98.8%, respectively), with no change in the sensitivity. In combination, diffuse positivity for MDM2/CDK4 gained in specificity (100%) (κ = 0.8; J = 0.7).

Discussion

IHC is said to have no role in diagnosing lipomatous tumours.2 4 Rarely, S100 protein immunoreactivity is used to identify adipocytic cells and lipoblasts in liposarcomas.2 Recently, peroxisome proliferator-activated receptor γ, a nuclear receptor that regulates growth and differentiation of adipocytes, was shown to stain lipogenic and non-lipogenic areas of DDLPS.22 In this study, the expression of MDM2 and CDK4 was examined by IHC in a series of soft tissue tumours. It was shown that all 19 (100%) cases of WDLPS were positive for MDM2 immunostaining, and most (13/19) were also positive for CDK4. When trying to differentiate WDLPS from BAT, MDM2 showed good sensitivity (100%) and CDK4 showed good specificity (88.2%). These findings are similar to those reported by Binh et al19 (100% and 96%, respectively), in a study with 44 WDLPS and 48 benign adipose tissue tumours (table 7). In agreement with their findings, the combination of MDM2 and CDK4 immunostaining increased the specificity of the test (94.1%). Moreover, it was shown that, when dealing with WDLPS, it is better to consider only strong and diffuse immunoreactivity in more than 30% of the neoplastic cells, to achieve the best specificity for MDM2 (94.1%) and CDK4 (100%).

Molecular cytogenetic studies using FISH were available on three cases of the BAT group (one non-conventional lipoma, one hibernoma and one PL-LP). Results showed that the lesional cells did not contain MDM2 gene (case 12) or CPM (carboxypeptidase M) gene (cases 65 and 128) amplification (Mayo Clinic). These three cases were both negative for MDM2/CDK4 immunostaining. The CPM gene is telomeric to the MDM2 locus on chromosome 12q15, and is also co-amplified with MDM2 in virtually all cases of WDLPS (unpublished data, Dr Andre Oliveira, 2009).

MDM2/CDK4 proteins when overexpressed act as stimulators of the cell cycle through the G1-S phase.23 24 Overexpression of these two proteins could lead to loss of cell cycle control with consequent development of neoplastic growth.23 24 25 This study found 90% of the DDLPS cases to have positive immunoreactivity for MDM2 and 70% to be immunoreactive for CDK4. When trying to identify DDLPS among other mesenquimal tumours (including its simulators), MDM2 showed good sensitivity (90%) and CDK4 showed good specificity (96.3%). These findings corroborate those reported by Binh et al,19 who found 95% of sensitivity and specificity for MDM2/CDK4. Like them, this study observed that combining MDM2/CDK4 immunostaining increased the specificity of the test (97.5%). In addition, it was shown that the most specific way to separate DDLPS from other sarcomas is to combine MDM2/CDK4 testing and consider a case positive when strong and diffuse immunoreactivity is seen in more than 30% of the neoplastic cells (97.5%).

One study investigated MDM2/CDK4 alterations by IHC in 200 soft tissue tumours and compared the results with FISH and RT-PCR.18 Their results observed that several tumours expressed these two proteins despite the absence of gene amplification, suggesting a mechanism of protein overexpression other than gene amplification. They stated that immunoreactivity alone for MDM2 or CDK4 does not always correlate with the diagnosis of WDLPS/DDLPS.18 This study demonstrated MDM2 immunoexpression in ten UPS, four MXFBS, one FBS, two LMS, three MPNST, two SS, one EPITS, two GIST, one DFSP, one MRLPS, one PLPS and one LGFMS; and CDK4 expression in one UPS, one MXFBS and one PLPS. Although we cannot totally exclude that some of these cases may be in fact DDLPS (eg, case 35), this study reinforces that MDM2 IHC alone has low specificity in diagnosing DDLPS.

In summary, this study gathered good evidence that the detection of MDM2/CDK4 protein overexpression by IHC can be used for clinical purposes to diagnose WDLPS/DDLPS. The best specificity result was achieved with the combination MDM2/CDK4 immunoreactivity. Furthermore, it was shown that specificity could be even higher if one considers a case positive for MDM2/CDK4 expression when strong and diffuse immunoreactivity is seen in more than 30% of the neoplastic cells. We do not recommend the use of MDM2 immunostaining alone to differentiate pleomorphic sarcomas from DDLPS because it has low specificity, even in cases with strong and diffuse immunoreactivity.