The Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/pmc/) is an open database, which provides high-throughput data for biological research. The GSE21122 dataset is based on the Affymetrix Human Genome U133A Array (HG-U133A) GPL96 platform and includes a variety of soft-tissue sarcoma specimens (11). The dataset contained data on 89 liposarcoma specimens (46 dedifferentiated liposarcoma specimens, 20 myxoid liposarcoma specimens and 23 pleomorphic liposarcoma specimens) and 9 NF specimens, which were downloaded for further analysis.

The original file was parsed to obtain the signal intensity of each probe. The dataset and the signal intensity values of each probe set were acquired using the robust multi-array average algorithm (12). Due to the design principle of the chip itself, and the artificial or other unavoidable factors in the experimental process, there were a large number of unqualified or invalid detection points in the chip raw data. The probe sets in the lowest 20% of the signal intensity order of all the probe sets in the two sample groups were filtered and considered as background noise. Subsequently, the variable coefficient of the same probe group in the same sample group was calculated using the coefficient of variation method (a method for comparing the discreteness of two sets of data), and the probe groups with a coefficient of variation >25% in both groups were filtered out. The limma package in R v3.4.3 (13) was used to identify DEGs and a linear model based on the empirical Bayesian distribution was used to calculate the significant difference level (P-value). The Benjamini-Hochberg method (14) was used to correct the significant difference level and to obtain the false discovery rate (FDR). An FDR<0.05 and |log FC|>1 were used as the cut-off criteria for DEGs, wherein a |log FC|<0 was considered as a downregulated gene and a |log FC|>0 was considered as an upregulated gene. The volcano map and heatmap were developed using the ggplot2 package and pheatmap package in R version 3.4.3, respectively (15).

Total RNA was extracted from cell lines and frozen tissues using TRIzol^® reagent (cat. no. 15596026; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. RT was performed using 5X All-In-One RT MasterMix kit (cat. no. G490; Applied Biological Materials, Inc.) according to the manufacturer's instruction. The procedure for RT was: 25°C for 10 min, 42°C for 15 min and 85°C for 5 min. qPCR was performed using EvaGreen 2X qPCR MasterMix (cat. no. Master Mix-LR; Applied Biological Materials, Inc.) on an ABI 7500 fast real-time PCR Detection system (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. The thermocycling conditions for qPCR were: Pre-denaturation at 95°C for 10 min; followed by 40 cycles of denaturation at 95°C for 15 sec, annealing at 60°C for 1 min and extension at 72°C for 30 sec. Primers were designed with Primer v6.0 (Thermo Fisher Scientific, Inc.). GAPDH was used as an internal reference gene. All reactions were performed in triplicate. The specificity of qPCR was confirmed by melting-curve analysis. The relative expression was calculated and analyzed using the 2^−ΔΔCq method (16). The primer sequences are shown in Table SI.

Short hairpin RNA (shRNA) was used to construct the gene knockdown vector (cat. no. GV115; Shanghai GeneChem Co., Ltd.). Three RNA interference targets were designed for each gene (Table SII), and three plasmids carrying different targets were mixed in equal proportions for lentivirus packaging. The 94T778 cells were infected with lentivirus containing the shRNAs using X-tremeGENE HP DNA Transf. Reag. 1.0 ml (cat. no. 6366236001; Roche Pharmaceutical, Ltd.). The cells were imaged using a Celigo^® Image Cytometer (Nexcelom Bioscience) with a fluorescence microscope. For each experimental well of a 96-well plate, Celigo^® Image Cytometer scanned four fields of view at each time point (magnification, ×40), and the cells were counted in the images with the corresponding analysis software in the Celigo^® Image Cytometer to obtain the number of cells in the experimental well.

All specimens of RLPS and NF tissues were obtained from patients who underwent surgical resection between March 2015 and January 2018 at the Peking University Cancer Hospital Sarcoma Center (Beijing, China). None of the patients received chemotherapy or radiotherapy prior to surgery. A total of 47 RLPS tissues (from 27 males and 20 females, with age range between 20 and 80 years, and the mean age was 55.3±10.7 years) were formalin-fixed immediately after resection, and embedded in paraffin prior to immunohistochemical analysis. The detailed clinicopathological characteristics of the 47 patients with RLPS are presented in Table I. The histopathological analysis of tumor tissues was performed by two pathologists, independently. For qPCR, 21 RLPS tissues (12 males and 9 females, the mean age was 55.0±8.5 years old) and 10 NF tissues (6 males and 4 females, the mean age was 57.4±8.4 years old) were snap-frozen in liquid nitrogen immediately after resection and transferred to −80°C freezer for long-term storage. The present study was approved by the Ethics Committee of Peking University Cancer Hospital (approval no. 2019KT19) and written informed consent was obtained from each participant.

The 4-µm-thick slices were heated for 1 h at 72°C and then dewaxed in xylene (cat. no. 10023418; Sinopharm Chemical Reagent Co., Ltd.) for 30 min at room temperature, and hydrated in 100, 95 and 85% ethanol solutions (cat. no. 10009259; Sinopharm Chemical Reagent Co., Ltd.) for 5 min at room temperature, respectively. The slices were incubated in 3% hydrogen peroxide for 15 min at room temperature to block endogenous peroxidase activity. Heat-mediated antigen retrieval was performed using EDTA buffer (cat. no. ZLI-9069; pH=9.0; OriGene Technologies, Inc.), after the pressure valve of the pressure cooker is exhausted for the first time, the firepower is reduce and the high pressure is continue for 2 min and 30 sec. Goat serum (cat. no. ZLI-9056; OriGene Technologies, Inc.) was used to block slices at 37°C for 1 h, after cooling to room temperature. Subsequently, the slices were incubated with the thymidylate synthase (TYMS) antibody (1:100; cat. no. ab108995; Abcam) overnight at 4°C. The following day, the slices were incubated with the goat anti-mouse/rabbit IgG-HRP conjugated antibody secondary antibody (at working solution; cat. no. PV-6000; OriGene Technologies, Inc.). The slices were visualized at room temperature using DAB (Dako; Agilent Technologies, Inc.) containing hydrogen peroxide for staining specific protein, when the specific staining is strong and the background color is shallow, the slice can be washed, and then using hematoxylin (cat. no. HK100-9K; Biogenex) for staining nucleus for 1 min at room temperature. Following TYMS staining, semi-quantitative classification of TYMS staining was performed according to the percentage of positive cells (PP) and staining intensity (SI) by two pathologists, who were blinded to the clinical information. The PP was scored as: 0, negative; 1, <25; 2, 25–75%; and 3, >75%. SI was scored as: 0, negative; 1, weak; 2, moderate; and 3, strong. The immunoreactivity score (IRS) was defined as PP multiplied by SI, where IRS=0 was considered as ‘negative’, and IRS>0 was considered as ‘positive’.

Gene Expression Profiling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku.cn/?from= timeline&isappinstalled=0) was used to analyze the association between TYMS expression and the overall survival (OS) and disease-free survival (DFS) of patients with sarcoma.

The human RLPS cell lines 93T449, 94T778 and SW872 were purchased from American Type Culture Collection and cultured in RPMI-1640 medium (Gibco; Thermo Fisher Scientific, Inc.) containing 10% FBS, 100 µg/ml penicillin, and 100 µg/ml streptomycin. Cells were maintained in a humidified incubator at 37°C with 5% CO₂.

Total protein was extracted using RIPA lysis buffer (cat. no. CW2333S; CWBIO) and supplemented with a protease and phosphatase inhibitor cocktail (cat. nos. CW2200S and CW2383S; CWBIO). A BCA Protein assay kit (cat. no. CW0014S; CWBIO) was used to quantify extracted proteins. Equivalent quantities of protein (5 µg) were separated by 8 or 10% SDS-PAGE and then transferred onto PVDF membranes. The membranes were blocked with 5% skimmed milk for 1 h at room temperature and then incubated with a series of primary antibodies at 4°C overnight. The following day, the membranes were incubated with secondary anti-rabbit or anti-mouse antibody for 1 h at room temperature. For detection of protein bands, Immobilon Western HRP Substrate Luminol Reagent (cat. no. WBKLS0500; EMD Millipore) and an enhanced chemiluminescence detection system (Amersham Imager 600; GE Healthcare) were used. The loading control used was β-actin. Image J software version 1.8.0 (National Institutes of Health) was used to analyze the intensities of western blot bands for the phosphorylated and non-phosphorylated proteins. All primary and secondary antibodies are listed in Table SIII.

All three cell lines were seeded (3.5×10⁵ for 93T449 cells; 1.5×10⁵ for 94T778 cells; 3.0×10⁵ for SW872 cells) on tissue culture (TC)-treated glass coverslips (cat. no. YA0351; Beijing Solarbio Science & Technology Co., Ltd.) for 24 h, then fixed with 4% paraformaldehyde for 20 min at room temperature, and incubated with a TYMS antibody (1:50; cat. no. 9045; Cell Signaling Technology, Inc.) at 4°C overnight. After washing, the cells were incubated with the goat anti-rabbit IgG-FITC conjugated antibody (1:100; cat. no. ZF-0311; OriGene Technologies, Inc.) for 30 min at room temperature, and the nucleus was stained with DAPI solution in the dark for 10 min at room temperature (0.5 µg/ml; cat. no. D523; Dojindo Molecular Technologies, Inc.). Images were acquired using a Multiphoton Laser Scanning Microscope (magnification, ×630; LSM780; Carl Zeiss Inc.).

shRNA, targeting TYMS, was constructed, packaged and purified by Shanghai GeneChem Co., Ltd. The lentiviral vector (cat. no. GV493; Shanghai GeneChem Co., Ltd.) containing non-silencing shRNA (Lenti-shCtrl) was used as the negative control. The stably transfected cells were selected using 1 µg/ml puromycin for 93T449 and 94T778, and 2 µg/ml puromycin for SW872. After lentivirus infection for 72–96 h, and screening by puromycin for 48 h, the cells were used for subsequent experiments. The efficiency of knockdown was confirmed using qPCR and western blot analysis.

A Cell Counting Kit-8 (CCK-8; Dojindo Molecular Technologies, Inc.) assay was performed to evaluate cell proliferation. Cells were plated in 96-well plates (3×10³ per well for 93T449; 1.5×10³ per well for 94T778 and SW872 cells), 10 µl of CCK-8 solution was added to each well, followed by the measurement of absorbance at 450 nm using a microplate reader (iMark; Bio-Rad Laboratories, Inc.) after 24, 48, 72, 96 or 120 h.

Cells were plated in 6-well plates at a density of 300 cells per well. The cells were cultured at 37°C for three weeks. Subsequently, colonies were fixed with 4% paraformaldehyde (cat. no. P1110; Beijing Solarbio Science & Technology Co., Ltd.) for 20 min at room temperature and stained with 0.1% crystal violet for 15 min (cat. no. G1063; Beijing Solarbio Science & Technology Co., Ltd.) at room temperature. The number of colonies were imaged and counted using the Promega Colony Counter software version 10 (Promega Corporation).

All three cell lines were cultured on 96-well plates (4.0×10⁴ cells per well for 93T449; 2.0×10⁴ cells per well for 94T778; 3.0×10⁴ cells per well for SW872) for 24 h, then fixed with 4% paraformaldehyde (as aforementioned), and subjected to cell apoptosis analysis using a TUNEL cell apoptosis assay kit (cat. no. T2190; Beijing Solarbio Science & Technology Co., Ltd.) according to the manufacturer's protocol. The nucleus was stained using DAPI solution (0.5 µg/ml; cat. no. D523; Dojindo Molecular Technologies, Inc.) in the dark for 10 min at room temperature, and images were acquired using a Multiphoton Laser Scanning Microscope (magnification, ×630; LSM780; Carl Zeiss Inc.). The number of TUNEL-positive cells were counted using the Promega Colony Counter software version 10 (Promega Corporation).

Cells were collected for cell cycle analysis. First, cells were digested with trypsin and washed with PBS, followed by overnight fixation at 4°C with 75% ice-cold ethanol. On the second day, the fixed cells were centrifuged at 250 × g, and washed with PBS at room temperature, followed by incubation for 15 min at room temperature in the dark with 300–500 µl PI/RNase Staining Buffer (cat. no. 550825; BD Pharmingen; BD Biosciences). BD Accuri C6 flow cytometry (Becton, Dickinson and Company) was used for cell collection and the ModFit LT software version 3.2 (Verity Software House, Inc.) was used for cell cycle analysis of the samples.

In the migration and invasion assays, chambers (migration assay, cat. no. 3422; invasion assay, cat. no. 354480; Corning, Inc.) were hydrated in serum-free RPMI-1640 medium (Gibco; Thermo Fisher Scientific, Inc.) for at least 2 h at 37°C. A total of 1×10⁵ 93T449 cells, 3×10⁴ 94T778 cells or 3×10⁴ SW872 cells were mixed in serum-free medium with a total volume of 200 µl for migration or 500 µl for invasion. Cells were added to the upper chamber, and 500 µl RPMI-1640 medium, containing 20% FBS, was added to the lower compartment of the chamber. After 24 h of incubation for migration and 48 h for invasion, the cells in the upper chamber were removed. The cells that had invaded/migrated were fixed with 4% paraformaldehyde for 20 min at room temperature and stained with 0.1% crystal violet for 15 min at room temperature. The stained cells were imaged in five randomly selected fields using an optical microscope (magnification, ×400), and the number of cells was calculated using ImageJ version 1.52t (National Institutes of Health).

Protein microarray analysis was performed using a Proteome Profiler™ Array-Human Phospho-Kinase Array kit (cat. no. ARY003B; R&D Systems, Inc.) according to the manufacturer's protocol.

Statistical analysis was performed using SPSS version 20.0 (IBM Corp.) and GraphPad Prism version 7.0 (GraphPad Software, Inc.). A two-tailed χ² test and Fisher's exact test were used to evaluate the association between TYMS expression and clinicopathological characteristics of patients with RLPS. Survival analysis was performed using Kaplan-Meier survival analysis and a log-rank test. An independent samples t-test was performed to compare two groups of quantitative data, and one-way ANOVAs with post-hoc Tukey's tests was performed to compare multiple groups of quantitative data. Results are presented as the mean ± standard deviation. P<0.05 was considered to indicate a statistically significant difference.

The microarray data of GSE21122 contained 89 LPS samples and 9 NF samples. The cut-off criteria for a gene to be considered as significantly differentially expressed were an FDR<0.05 and |logFC|>1. A total of 855 DEGs were extracted from GSE21122, of which 334 genes were upregulated and 521 genes were downregulated (Fig. S1; Table SIV). Fig. 1A shows the expression of DEGs in LPS and NF samples, and Fig. 1B shows the percentage of upregulated and downregulated genes.

To screen for functional genes which promoted the occurrence of RLPS, RT-qPCR was used to determine the expression of 40 genes that were considered to be upregulated in the RLPS cell line, 94T778, none of which have been previously studied in RLPS, to the best of our knowledge (Table SV). In 94T778 cells, 30 of the 40 examined genes exhibited higher expression compared with the other 10 genes (Table SVI), suggesting that expression of these 30 genes was relatively high in 94T778 cells and were thus more likely to serve a role in the development of RLPS.

Lentivirus-mediated shRNA was used to infect 94T778 cells and knockdown the 30 genes with high expression. Subsequently, the effect of these genes on proliferation of 94T778 cells was evaluated using the HCS method. The results showed that following knockdown, 16 of the 30 genes, including TYMS, KIF20A, BUB1B, TMSB15A, KIAA0101, SNAI2, LMNB1, TIA1, TPX2, RRM2, PKM, PTTG1, CKS2, ZWINT, MELK and RACGAP1, significantly slowed down the proliferation of 94T778 cells (Table SVII; Fig. 1C). Based on the results of HCS, knockdown of the TYMS gene had the most notable effect on the proliferative capacity of RLPS cells, and thus was selected to further explore its effect on the biological behaviors and the potential mechanisms involved in RLPS cells.

mRNA expression of TYMS in 21 RLPS tissues and 10 NF tissues was assessed using qPCR. Relative mRNA expression of TYMS was notably higher in RLPS tissues compared with the NF tissues (P<0.001; Fig. S2A). Furthermore, in the immunohistochemical assay, TYMS was expressed in 35 of the 47 RLPS tissues (74.5%). IHC showed that TYMS was predominantly localized in the cytoplasm (Fig. S2B).

As shown in Table II, TYMS expression was 42.9, 91.3, 80.0 and 80.0% positive in well-differentiated, dedifferentiated, myxoid/round cell and pleomorphic subtypes, respectively (P=0.009). TYMS expression was significantly higher in patients with high-grade RLPS compared with patients with low-grade RLPS (low-grade vs. high-grade: 42.9% vs. 87.9%; P=0.003).

In the survival analysis, five patients were lost to follow-up, and in the remaining 42 cases, the median OS time was 29.21±24.68 months in the TYMS-positive patients and 38.72±10.16 months in TYMS-negative patients. The median DFS was 26.20±23.53 months in TYMS-positive patients and 35.58±9.83 months in TYMS-negative patients. Kaplan-Meier survival analysis and a log-rank test showed that TYMS expression was negatively associated with both OS and DFS time of patients with RLPS (Fig. S2C and D; OS, P=0.024; DFS, P=0.030). Furthermore, survival analysis of TYMS and sarcoma using GEPIA database showed that patients with sarcoma and TYMS-high expression had lower OS and DFS time compared with those with TYMS-low expression (Fig. S3A and B; OS, P=0.0065; DFS, P=0.019). The aforementioned results showed that high TYMS expression was associated with progression of RLPS or sarcoma.

Protein expression of TYMS was evaluated by western blotting in three human RLPS cell lines (93T449, 94T778 and SW872). The TYMS protein levels were considerably high in all three RLPS cell lines, particularly in the 93T449 and SW872 (Fig. S4A). TYMS was found to be localized in the cytoplasm in all three human RLPS cell lines (Fig. S4B). Therefore, all three cell lines were used for subsequent experiments.

To examine the role of TYMS in RLPS cells, 93T449, 94T778 and SW872 cells were transfected with lentiviral vectors containing non-specific shRNA (Lenti-shCtrl) and shRNAs targeting TYMS (Lenti-shTYMS-1 and Lenti-shTYMS-2). The transfection efficiency was confirmed using RT-qPCR and western blotting (Fig. S4C and D). The results of a CCK-8 assay showed that TYMS downregulation significantly reduced cell viability (Fig. 2A; P<0.01) and decreased colony-forming capacity (Fig. 2B and C; P<0.05) in all three RLPS cell lines. To evaluate the apoptotic status of cell lines following knockdown of TYMS, TUNEL staining was performed on the RLPS cell lines. Lenti-shTYMS cells showed a significant increase in apoptosis compared with the Lenti-shCtrl transfected cells (Fig. S5).

TYMS downregulation decreased cell proliferation, therefore, cycle cell distribution was assessed to determine whether this attenuation was associated with cell cycle arrest. Flow cytometry analysis demonstrated that TYMS knockdown promoted progression of the cell cycle from G₁ to S phase. As shown in Fig. 3A and B, in 93T449 cells, the proportion of cells in the G₁ phase decreased significantly, from 88.05±1.01% in Lenti-shCtrl cells to 84.64±0.42% and 84.79±0.40% in Lenti-shTYMS-1 and Lenti-shTYMS-2 cells, respectively (both P<0.05). The proportion of S phase cells increased significantly. Similar results were observed in the 94T778 and SW872 cells. These results suggest that TYMS downregulation arrests the cell cycle at the S phase.

In order to determine the effect of TYMS expression on cell migration and invasion, Transwell migration and invasion assays were performed. As shown in Fig. 4, in 93T449 cells, the number of migrated cells per field was 95±35 in shTYMS-1 cells (P<0.001) and 102±10 in shTYMS-2 cells (P<0.001) vs. 1,234±114 in the control cells; the number of invaded cells per field was 118±11 in shTYMS-1 cells (P<0.001) and 124±21 in shTYMS-2 cells (P<0.001) vs. 849±71 in control cells. Similar results were obtained in 94T778 and SW872 cells. The results showed that TYMS knockdown significantly decreased both the migration and invasive capacities of RLPS cells, thereby indicating that TYMS downregulation may inhibit RLPS progression.

The aforementioned results suggest that TYMS serves an important role in the development and progression of RLPS; however, the mechanism is unclear. Thus the possible mechanisms of TYMS in RLPS cells (94T778 and SW872) were assessed using protein microarray analysis. The results showed that phospho-STAT3 (Ser727), phospho-STAT5 (Y694) and phospho-STAT6 (Y641) were downregulated in Lenti-shTYMS transfected 94T778 and SW872 cells (Fig. S6). Therefore, it was hypothesized that TYMS knockdown may have resulted in decreased activity of the JAK/STAT signaling pathway. Western blotting was used to assess the expression levels of major proteins associated with the JAK/STAT signaling pathway in Lenti-shCtrl and Lenti-shTYMS cells, including total Jak1-3 and Tyk2 of the JAK family and the corresponding phosphorylated proteins, as well as the expression levels of total STAT1, STAT3, STAT5 and STAT6, and the corresponding phosphorylated proteins of the STAT family. The results showed that TYMS knockdown resulted in downregulation of the expression of phosphorylated Tyk2 in Lenti-shTYMS cells compared with the control cells, and the expression of phosphorylated Jak1, total Jak2, phosphorylated Jak2, total Jak3 and phosphorylated Jak3 were not detected. Meanwhile, phosphorylated STAT1 (Tyr701), phosphorylated STAT3 (Ser727), phosphorylated STAT5 (Y694) and phosphorylated STAT6 (Y641) in the STAT family were significantly downregulated in Lenti-shTYMS cells, consistent with the results of protein microarray analysis (Fig. 5; Fig. S7). These results suggest that TYMS knockdown may downregulate the activity of the JAK/STAT signaling pathway in RLPS cells, thereby inhibiting the progression of RLPS.