The statistical significance of genomic aberrations for each subtype is shown. RAE q-values [left axis; for visualization, q-values ≤ 0.05 are considered significant, corresponding false discovery rate (FDR) ≤ 5%] and scores (right axis) for gains and amplifications (red) and losses and deletions (blue) are plotted across the genome (chromosomes indicated at bottom). Genes harboring somatic nucleotide alterations in this study are indicated in each subtype in which they were discovered (Table 2).

A. Somatic mutations in the NF1 protein in myxofibrosarcoma and pleomorphic liposarcoma (black triangles) and the position of the RasGAP and Cral domains (dark and light green respectively) are juxtaposed to known mutations in malignant peripheral nerve sheath tumors (MPNSTs; open triangles). B. Transcript expression according to copy number and sequence status in myxofibrosarcoma and pleomorphic liposarcoma compared to normal adipose tissue samples (black/red and green respectively, log₂ expression from Affymetrix array profiling data; p-value=1.94×10⁻⁵, ANOVA; mutated tumors are indicated). One of the two R304* mutant tumors lacked expression data.

A. Survival for patients with tumors that harbor helical-domain mutations (red) versus kinase-domain mutations (grey), and wildtype PIK3CA (blue). The analysis includes the 65 patients for whom outcome information was available. Patients with mutations in either the helical or the kinase domain had a shorter disease–specific survival compared to those with wildtype PIK3CA (p-value = 0.0363, log-rank test). The difference in disease-specific survival between patients with helical-domain mutant tumors and those with wildtype PIK3CA tumors was significant (p-value=0.013, log-rank test). B. Western blots of myxoid/round-cell liposarcoma tumor lysates comparing the phosphorylation levels of Akt, PRAS40, and S6 kinase, as well as their protein levels, in patients with wild-type PIK3CA or with mutations in PIK3CA helical or kinase domains.

(A) Integrated profile of statistically significant genomic gains/amplifications as assessed by both RAE and GISTIC (combined as described in Methods; FDR, false-discovery rate) is followed by a heatmap of copy number segmentation on 12q13.2-q32.1 in 50 patient samples of dedifferentiated liposarcomas (red is amplification, blue is deletion, each row indicates one tumor sample). Below is the position of genes from our screen encoded by this region of 12q whose knockdown is anti-proliferative in dedifferentiated liposarcoma. Bold gene symbols indicate those whose amplification produced over-expression of its transcript or those over-expressed in tumor relative to normal adipose tissue. Genes in green are highlighted in panels B–C and E. Alternative genomic regions encoding genes not on 12q whose knockdown is anti-proliferative are also included. (B) Effect of three validated shRNAs targeting CDK4 on the proliferation of two cell lines, LPS141 and DDLS8817, at various time points (x-axis) with negative controls (pLKO empty vector and GFP473). Below are western blots showing the effect of shRNAs on levels of CDK4 protein (as indicated). (C) G1 arrest induced in LPS141 and DDLS8817 cell lines by treatment with the CDK4/CDK6 inhibitor PD0332991. MDA-MB-435 (Rb-positive) and H2009 (Rb-negative) were included as sensitive and insensitive controls. Error bars are s.d. of replicate measurements. (D–E) As in panel (B), effect on proliferation of three shRNAs targeting MDM2 (panel D) and YEATS4 (panel E) (negative controls: pLKO empty vector and scrambled shRNA) where each targeting shRNA resulted in reduced protein levels (at bottom). Error bars are propagated error from the ratio of mean and s.d. of measurements/replicates to time 0.