Purpose:

Interpatient clinical variability in soft-tissue sarcomas (STS) highlights the need for novel prognostic markers supporting patient risk stratification. As sarcomas might exhibit a more mesenchymal or a more epithelial state, we focused on epithelial–mesenchymal and mesenchymal–epithelial transitions (EMT/MET) for prognostic clues, and selected three histotypes with variable aggressiveness.

Experimental Design:

The expression of EMT/MET-related factors was measured by qRT-PCR in 55 tumor samples from patients with leiomyosarcoma, myxofibrosarcoma, or undifferentiated pleomorphic sarcoma. The identified marker was further evaluated by IHC in 31 leiomyosarcomas and by measuring its circulating levels in 67 patients. The prognostic value of a sarcoma-tailored EMT score was analyzed. Epirubicin chemosensitivity and migration were studied in primary STS cultures. Associations with overall survival (OS) were assessed using Kaplan–Meier and Cox regression methods.

Results:

High expression of periostin, a mesenchymal matricellular protein, in sarcoma tissues (P = 0.0024), its high stromal accumulation in leiomyosarcomas (P = 0.0075), and increased circulation (>20 ng/mL, P = 0.0008) were associated with reduced OS. High periostin expression [HR 2.9; 95% confidence interval (CI), 1.3–6.9; P = 0.0134] and circulation (HR 2.6; 95% CI, 1.3–5.1; P = 0.0086), and a mesenchymal EMT score (mesenchymal vs. transitioning; HR, 5.2; 95% CI, 2.1–13.0, P = 0.0005) were associated with increased risk in multivariable models. An intrinsic or induced mesenchymal state enhanced chemoresistance and migration in sarcoma cell lines.

Conclusions:

Although limited to a pilot study, these findings suggest that periostin might contribute prognostic information in the three studied STS histotypes. Moreover, a transitioning EMT score measured in the tumor might predict a less active and a more chemosensitive disease.

Translational Relevance

Soft-tissue sarcomas (STS) are rare mesenchymal tumors of relevant molecular and clinical heterogeneity. Sarcomas may exhibit a more mesenchymal or a more epithelial state, which governs clinical behavior. Thus, we focused on epithelial–mesenchymal transition (EMT) for prognostic clues and performed a pilot study to search for biomarkers supporting patient risk stratification in three potentially aggressive histotypes. Multivariable models showed that periostin expression and circulation levels were significantly associated with increased risk in patients with STS. Moreover, stromal accumulation of periostin was associated with shorter survival in patients with leiomyosarcoma. Periostin was combined with other EMT markers in the computation of a sarcoma-tailored EMT score. A transitioning EMT score, previously shown to denote aggressive carcinomas, predicted a better outcome in STS, suggesting that the clinical behavior of hybrid states might be related to tumor type. Moreover, a transitioning state might attenuate chemoresistance and migration in primary sarcoma cell lines.

Soft-tissue sarcomas (STS) are rare tumors characterized by variable aggressiveness and high heterogeneity, as histopathology differentiates more than 50 subtypes (1, 2). While well-defined molecular alterations drive oncogenesis in few histotypes, such as synovial sarcoma and solitary fibrous tumors, a multifaceted genetic, transcriptomic, and regulomic landscape characterizes most sarcomas (3), including leiomyosarcoma, myxofibrosarcoma, and undifferentiated pleomorphic sarcoma, which share an intrinsic molecular complexity that governs a relevant interpatient clinical variability. Biomarkers with accurate prognostic value supporting clinicopathologic parameters in patient risk stratification are urgently needed.

Sarcomas are mesenchymal tumors that may express epithelial markers, such as E-cadherin or ZO-1, indicating the occurrence of mesenchymal–epithelial transition (MET) during sarcomagenesis (4–6). The clinical value of epithelial and mesenchymal markers in STS needs to be fully understood. In fact, detection of E-cadherin or SNAIL, a mesenchymal transcriptional repressor, in tumor cells was associated with favorable or poor clinical outcome in patients with leiomyosarcoma and other sarcomas, respectively (7, 8). Similarly, intense staining for vimentin, a critical organizer of shape and motility in mesenchymal cells, showed negative prognostic value in patients with leiomyosarcoma (7). Therefore, some histotypes exhibit a more mesenchymal or a more epithelial state, which in turn governs clinical course, highlighting the possible clinical implication of epithelial and mesenchymal factors in sarcoma evolution.

In this monocentric, pilot study, we investigated whether the expression of epithelial–mesenchymal transition (EMT)/MET-related factors, as single or combined parameters, are correlated with patient outcome. Three subtypes with variable aggressiveness were selected. Three primary STS cell cultures were isolated and analyzed for chemoresistance and migration, properties linked to epithelial–mesenchymal plasticity (9, 10).

Patients

A screening phase was conducted to molecularly characterize 55 tumor tissue samples retrospectively selected from a biobank of the Department of Surgery, Oncology and Gastroenterology of Padua University (DISCOG), and prospectively obtained during the initial part of the study. A REMARK flow diagram of samples and analyses is shown in Supplementary Fig. S1 (11). Criteria for patient inclusion were: primary histologically confirmed leiomyosarcoma, myxofibrosarcoma, or undifferentiated pleomorphic sarcoma; availability of fresh or frozen tumor tissue from untreated patients at the time of diagnosis; histologic confirmation of representativeness of the sample; and follow-up of at least 3 years for the screening set (follow-up period end: November 2018). Exclusion criteria were neoadjuvant treatment, ambiguous histopathologic diagnosis, and history of other cancers. Clinical data of all retrospective cases were reviewed.

Results obtained in the discovery stage were further evaluated using two different approaches. First, the selected marker was evaluated by IHC in an independent set of formalin-fixed and paraffin-embedded (FFPE) tissue samples retrospectively selected from the archives of the Anatomy and Pathological Histology Unit (IOV). This set included 31 leiomyosarcomas, 6 leiomyomas, and 15 normal muscular tissues (NMT). Second, the circulating levels of the marker were quantified in serum/plasma samples of 67 patients with STS (47 of the screening set and 20 prospectively enrolled with the same inclusion/exclusion criteria). The time period was April 1998–November 2015 for all cases included in the initial discovery set, and September 1995–November 2017 for those included in the subsequent IHC evaluation and those additional patients included in the measurement of the circulating levels of the marker.

Histologic diagnosis of all retrospectively selected cases was reviewed and confirmed by two expert pathologists (M.C. Montesco and R. Cappellesso), according to the current WHO classification (2). Biobank samples were stored at −80°C until RNA extraction. Fresh surgical tissue samples were snap frozen in liquid nitrogen and stored at −80°C until RNA extraction. Histopathologic diagnosis of all prospectively collected tumor samples was performed by the same pathologists (M.C. Montesco and R. Cappellesso).

Clinical, pathologic, and follow-up data of all patients were collected by a data manager (P. Del Fiore). This study was examined and approved by the IOV Ethics Committee (protocol no. 2014/91) and was conducted in compliance with the guidelines of the Declaration of Helsinki. Written informed consent for the collection and analysis of biological samples was obtained from all biobank participants and prospectively enrolled patients.

Isolation and characterization of primary STS cell cultures

After surgical resection, tumor samples were immediately processed under sterile conditions by mechanical dissociation. The obtained cell suspension was cultured in DMEM-F12 medium (Sigma-Aldrich) supplemented with 20% FCS (Gibco, Thermo Fisher Scientific), 2 mmol/L l-glutamine (Gibco), 50 μg/mL gentamicin (Sigma-Aldrich), 100 U/mL penicillin, 100 μg/mL streptomycin, and 0.25 μg/mL amphotericin (Gibco). This complete medium was replaced twice a week, and when cells reached confluence, they were maintained through serial passages. Short tandem repeat (STR) profiling of the primary STS cultures was performed by BMR Genomics (https://www.bmr-genomics.it/). The isolated primary cell lines were Mycoplasma-free, as confirmed by periodical PCR check (every 10 passages for the first 30 passages, then every 30 passages), and were characterized by IHC, EMT/MET transcriptional profile, chemoresistance, and migration ability.

Quantitative mRNA analyses

Total RNA, enriched with low molecular weight molecules, was extracted using the NucleoSpin miRNA isolation Kit (Macherey-Nagel GmbH & Co. KG) according to manufacturer's protocol from 15–30 mg of tumor tissue that was ground to a fine powder under liquid nitrogen and from cell pellets of primary STS cultures. RNA concentration and purity were assessed as reported previously (12).

The expression of genes linked to an epithelial state (CDH1/E-cadherin and ZO-1/ZO-1) and to a mesenchymal state (SNAI1/SNAIL, SNAI2/SLUG, ZEB1/ZEB1, SIP1/SIP1, ACTA2/alpha-smooth muscle actin or ASMA, VIM/vimentin, CDH2/N-cadherin and POSTN/periostin) was assessed by quantitative real-time PCR (qRT-PCR). First-strand cDNA synthesis was performed with 500 ng of total RNA using RevertAid H Minus Reverse Transcriptase (Thermo Fisher Scientific) in a total volume of 20 μL containing the reaction buffer, 20 U of RiboLock RNase inhibitor, 1 mmol/L dNTPs, 200 U of reverse transcriptase, and 0.2 μg of random primers (Invitrogen, Thermo Fisher Scientific); reactions were carried out according to the manufacturer's protocol. qRT-PCR was carried out using Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen) in a 7900HT Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). Melting curves were analyzed for each reaction to ensure the specificity of the amplicons. The same positive control for each analyzed gene was run in each plate to monitor interplate variability and measurement reproducibility. Primer sets specific for intron-spanning amplicons for CDH1, SNAI1, SNAI2, ZEB1, SIP1 and ACTA2 used in qRT-PCR were published previously (13). Primer pairs were designed for transcripts specific for ZO-1 (forward: 5′-AGGTCTGCCGGGACAACA-3′; reverse: 5′-AGGATCACAGTGTGGTAAGCG-3′), VIM (forward: 5′-CGAAAACACCCTGCAATCTT-3′; reverse: 5′-CTGGA TTTCCTCTTCGTGGA-3′), CDH2 (forward: 5′-GGACAGTTCCTGAGGGATCA-3′; reverse: 5′-CTTGGAGCCTGAGACACGAT-3′), and POSTN (forward: 5′-TGTGGACAGAAAACGACTGTG-3′; reverse: 5′-ATGCCCAGAGTGCCATAAACA-3′). Expression levels of the analyzed genes are reported as relative quantification (RQ) calculated using human porphobilinogen deaminase gene as the endogenous control with the 2−ΔΔCt method (14). Normalized RQ (nRQ) were calculated on the global mean (15).

miRNA quantification

To better evaluate the profile of the epithelial component, rarely found in STS, we increased the number of possible epithelial markers through the selection of miRNAs known to affect EMT and to be aberrantly expressed in STS (16, 17). miRNAs linked to EMT (miR-100–5p; ref. 18), to MET (miR-200b-3p, miR-30b-5p, and miR-30c-5p; refs. 19, 20), and myo-miRs (miR-1, miR-133a-3p, and miR-133b) were quantified using miRNA-specific locked nucleic acids (LNATM) PCR primer sets (Exiqon A/S). The three myogenic miRNAs were included as further independent control of diagnosis and tumor representativeness, as they were previously shown to be highly expressed in leiomyosarcoma samples (16, 17). Ten nanograms of RNA was reverse transcribed to cDNA using the miRCURY LNA Universal Reverse Transcription miRNA PCR cDNA Synthesis Kit II (Exiqon) according to the manufacturer's protocol. The assays were performed in duplicate and each miRNA was quantified in two independent runs using the ExiLENT SYBR Green mastermix (Exiqon) in a 7900HT Real-Time PCR System. To monitor the quality and efficiency of RNA extraction, cDNA synthesis, and PCR amplification, synthetic small RNA molecules (spike-ins, Exiqon) were added into each sample prior to RNA isolation (UniSp2, UniSp4, UniSp5) and prior to cDNA synthesis (UniSp6) according to the manufacturer's protocol. Spike-in amplification was carried out for each sample, and samples not ensuring the highest sensitivity were not included in the analysis. miRNA expression levels were calculated using U6 as the endogenous control with the 2−ΔΔCt method (14). nRQ were normalized on the global mean (15). Criteria used for identifying outliers during quantification of gene-specific mRNAs and miRNAs and for excluding unreliable discoveries due to technical variability have been applied and previously reported (12).

IHC

IHC was performed automatically using the Bond Polymer Refine Detection kit (Leica Biosystems) in the BOND-MAX system (Leica Biosystems) on a 4-μm–thick section from each tissue sample with the primary antibody anti-periostin (rabbit polyclonal, Abcam; dilution 1:800). Sections were then counterstained with hematoxylin. Appropriate positive and negative controls were run concomitantly. IHC for periostin was jointly analyzed and scored for positivity in the cells and in the stroma by two pathologists (R. Cappellesso and M.C. Montesco) blinded to clinical data. A four-tier system was applied to score the maximum intensity of the staining in both the cells and the stroma of each tissue sample: 0 = total absence, 1 = weak, 2 = moderate, and 3 = strong.

Cell pellets of primary STS cultures were formalin-fixed and paraffin-embedded using the Shandon Cytoblock Cell Block Preparation System (Thermo Fisher Scientific), and IHC was performed on 4-μm–thick sections from each cell block as described above using the primary mouse mAbs anti-ASMA (clone 1A4, Merck KGaA), anti-desmin (D33, Abcam), anti-CD34 (QBEND-10, Abcam), anti-MDM2 (IF2, Thermo Fisher Scientific), and the rabbit polyclonal anti-S100 antibody (Dako).

Quantification of circulating levels of periostin

Circulating amounts of periostin were measured in serum or plasma samples from 67 patients diagnosed with leiomyosarcoma, myxofibrosarcoma, or undifferentiated pleomorphic sarcoma. Matched tissue-blood samples were available for 43 patients. Samples were stored at −80°C before testing. Quantification was carried out using a sandwich ELISA kit (DuoSet ELISA Development System Human Periostin/OSF-2, R&D Systems) according to the manufacturer's instructions. All samples were analyzed using an initial 1:25 dilution.

EMT score analysis

The EMT score was calculated as reported previously (21), and tailored on the markers selected as follows. A descriptive network analysis based on Pearson correlation between the logarithm levels of all analyzed markers was performed to identify the strength and direction of existing structures and patterns. The expressed epithelial variable (ZO-1) was scaled and centered and the identified mesenchymal variables were linearly combined in the first principal component using the larger eigenvector extracted from the correlation matrix as weights (21). To compute the EMT score, the difference between the mesenchymal principal component and the scaled epithelial variable was calculated and its prognostic role was assessed both in univariate and multivariable models. The EMT score was calculated using the expression profile of 47 tumor tissues of the screening set that had measurable values for the markers selected; specifically, 4 leiomyosarcomas, 3 myxofibrosarcomas, and 1 undifferentiated pleomorphic sarcoma were excluded from this analysis.

Chemosensitivity and cell migration of primary STS cell lines

Chemoresistance to epirubicin was tested by MTT assay (22). Cells were seeded in 96-well plates (3 × 104/well) and treated with epirubicin (2-fold dilutions from 40 μmol/L to 0.62 μmol/L for S78 and S79 cells, and from 40 μmol/L to 0.31 μmol/L for S57) for 48 hours. IC50 was determined by a nonlinear regression analysis using the GraphPad Prism software (6.07 for Windows).

Cell migration ability was analyzed by cell scratch assay (23). Cells were seeded in 6-well plates (3 × 105/well) and allowed to reach confluence. Scratching was performed with a sterile pipette tip, and the cells were incubated with complete medium with or without recombinant TGFβ1 (TGFβ1, 1 ng/mL; R&D Systems). Phase-contrast photographs of each scratch were taken at 24 and 48 hours after scratching. The percentage of wound closure was calculated as described previously (23). At 48 hours, cells were detached by trypsin treatment (Thermo Fisher Scientific), washed with cold PBS (Oxoid) and processed for total RNA extraction.

Statistical analyses

Quantitative variables were described as median and interquartile range; categorical variables were summarized as counts and percentages. The median follow-up time was based on the reverse Kaplan–Meier estimator. Clinical outcome was analyzed in terms of overall survival (OS), which was the time from the date of diagnosis to death from any cause. Patients who did not develop a survival event during the study period were censored at the date of last observation. The survival probabilities were estimated using the Kaplan–Meier method and compared among strata using the log-rank test. The HRs were obtained from univariate Cox proportional hazards regression models, after checking any deviation from the proportional hazards assumption. Quantitative markers entered the models as dichotomous variables, categorized according to their median value. Statistically significant variables were entered into a multiple Cox proportional hazards regression model to backward select predictors independently associated with the outcome.

The association of IHC score with tissue samples was verified using the Likelihood Ratio χ2 test. The distributions of circulating levels of periostin were compared across clinical characteristics using the Kruskal–Wallis test. Differences in migratory capacity were assessed using the Student t test. P values were adjusted for multiple comparisons using the Benjamini–Hochberg method.

All statistical tests used a two-sided 5% significance level and association measures were provided with their 95% confidence interval (CI). Statistical analyses were performed using the SAS statistical package (SAS, rel. 9.4; SAS Institute Inc.) and RStudio (RStudio: Integrated Development for R. RStudio Inc.).

Prognostic value of periostin expression in leiomyosarcoma, myxofibrosarcoma, and undifferentiated pleomorphic sarcoma

Characteristics of the patients included in the screening are shown in Table 1. The expression level of EMT/MET-related factors was analyzed in tumor tissue from 55 untreated patients. The estimated median follow-up time was 8.2 years, (95% CI, 4.4–14.1), with a 5-year OS of 51.2% (36.9–63.8) and a 10-year OS of 33.7% (19.0–49.1). At the time of this analysis, 61.8% (34/55) of the patients had died.

Table 1.

Clinical and pathologic characteristics of patients with STS of the initial discovery and subsequent evaluation stages.

PatientsHistotypeLMSa
VariableN (%)LMSMFSUPSN (%)
Age (years) ≤70 33 (60.0%) 20 (71.4%) 8 (61.5%) 5 (35.7%) 21 (67.7%) 
 >70 22 (40.0%) 8 (28.6%) 5 (38.5%) 9 (64.3%) 10 (32.3%) 
Gender 28 (50.9%) 13 (46.4%) 8 (61.5%) 7 (50.0%) 12 (38.7%) 
 27 (49.1%) 15 (53.6%) 5 (38.5%) 7 (50.0%) 19 (61.3%) 
Grade G1 3 (5.5%) 1 (3.6%) 2 (15.4%) 7 (22.6%) 
 G2 14 (25.4%) 10 (35.6%) 4 (30.8%) 7 (22.6%) 
 G3 38 (69.1%) 17 (60.7%) 7 (53.8%) 14 (100%) 17 (54.8%) 
Tumor size ≤5 cm 13 (23.6%) 10 (35.7%) 1 (7.7%) 2 (14.3%) 16 (51.6%) 
 >5 cm 42 (76.4%) 18 (64.3%) 12 (92.3%) 12 (85.7%) 15 (48.4%) 
Metastasis at diagnosis Yes 8 (14.6%) 8 (28.6%) 
 No 47 (85.4%) 20 (71.4%) 13 14 31 
Survival status Alive 21 (38.2%) 8 (28.6%) 9 (69.2%) 4 (28.6%) 19 (61.3%) 
 Dead 34 (61.8%) 20 (71.4%) 4 (30.8%) 10 (71.4%) 12 (38.7%) 
Histotype LMS 28 (50.9%)     
 MFS 13 (23.6%)     
 UPS 14 (25.5%)     
PatientsHistotypeLMSa
VariableN (%)LMSMFSUPSN (%)
Age (years) ≤70 33 (60.0%) 20 (71.4%) 8 (61.5%) 5 (35.7%) 21 (67.7%) 
 >70 22 (40.0%) 8 (28.6%) 5 (38.5%) 9 (64.3%) 10 (32.3%) 
Gender 28 (50.9%) 13 (46.4%) 8 (61.5%) 7 (50.0%) 12 (38.7%) 
 27 (49.1%) 15 (53.6%) 5 (38.5%) 7 (50.0%) 19 (61.3%) 
Grade G1 3 (5.5%) 1 (3.6%) 2 (15.4%) 7 (22.6%) 
 G2 14 (25.4%) 10 (35.6%) 4 (30.8%) 7 (22.6%) 
 G3 38 (69.1%) 17 (60.7%) 7 (53.8%) 14 (100%) 17 (54.8%) 
Tumor size ≤5 cm 13 (23.6%) 10 (35.7%) 1 (7.7%) 2 (14.3%) 16 (51.6%) 
 >5 cm 42 (76.4%) 18 (64.3%) 12 (92.3%) 12 (85.7%) 15 (48.4%) 
Metastasis at diagnosis Yes 8 (14.6%) 8 (28.6%) 
 No 47 (85.4%) 20 (71.4%) 13 14 31 
Survival status Alive 21 (38.2%) 8 (28.6%) 9 (69.2%) 4 (28.6%) 19 (61.3%) 
 Dead 34 (61.8%) 20 (71.4%) 4 (30.8%) 10 (71.4%) 12 (38.7%) 
Histotype LMS 28 (50.9%)     
 MFS 13 (23.6%)     
 UPS 14 (25.5%)     

Abbreviations: LMS, leiomyosarcoma; MFS, myxofibrosarcoma; UPS, undifferentiated pleomorphic sarcoma.

aSamples for subsequent IHC evaluation.

E-cadherin was not found to be expressed in any of the analyzed samples. Among the expressed markers, high periostin mRNA levels were found to be significantly associated with decreased OS and increased HR (Fig. 1A), whereas none of the other biomarkers was significantly associated with OS in univariate analysis (Supplementary Table S1).

Figure 1.

Detection and prognostic significance of periostin in leiomyosarcoma, myxofibrosarcoma, and undifferentiated pleomorphic sarcoma. A, Kaplan–Meier curves for OS according to the level of expression of periostin (dichotomized according to the median nRQ value of expression) measured in tumor tissue samples obtained from 55 patients with STS of the screening set. Each sample was run in duplicate in three independent plates. The log-rank test was used to calculate the P value. HR, 95% CI, and the associated P value are also indicated. B, Representative photomicrographs of the immunoreactions for periostin performed on FFPE samples of the normal muscular layer (a), leiomyoma (b), and leiomyosarcoma (c and d). Scale bars, 200 μm. Original magnification 100×. C and D, Kaplan–Meier OS curves according to periostin detection in the stromal compartment or tumor cells, respectively. P values, HR, and 95% CI were calculated as reported above.

Figure 1.

Detection and prognostic significance of periostin in leiomyosarcoma, myxofibrosarcoma, and undifferentiated pleomorphic sarcoma. A, Kaplan–Meier curves for OS according to the level of expression of periostin (dichotomized according to the median nRQ value of expression) measured in tumor tissue samples obtained from 55 patients with STS of the screening set. Each sample was run in duplicate in three independent plates. The log-rank test was used to calculate the P value. HR, 95% CI, and the associated P value are also indicated. B, Representative photomicrographs of the immunoreactions for periostin performed on FFPE samples of the normal muscular layer (a), leiomyoma (b), and leiomyosarcoma (c and d). Scale bars, 200 μm. Original magnification 100×. C and D, Kaplan–Meier OS curves according to periostin detection in the stromal compartment or tumor cells, respectively. P values, HR, and 95% CI were calculated as reported above.

Close modal

Among the clinicopathologic variables significantly associated with decreased OS and increased risk in univariate analysis (Supplementary Fig. S2; Supplementary Table S2), advanced age, high grade, large tumor, and leiomyosarcoma histotype remained independently associated with an unfavorable outcome in a multivariable analysis (Supplementary Table S2). Evaluation of the effect of periostin adjusted for these clinical variables confirmed that high periostin expression in tumor tissue was an independent, unfavorable prognostic factor, in addition to advanced age and high grade (Table 2).

Table 2.

Multivariable analysis for association of periostin expression in tumor tissues and clinical variables with OS.

Multivariable OS
VariableaHR (95% CI)P
Age (years) ≤70  
 >70 2.8 (1.4–5.8) 0.0047 
Grade G1+G2  
 G3 3.3 (1.2–9.0) 0.0189 
Periostin (nRQ) ≤1.6  
 >1.6 2.9 (1.3–6.9) 0.0134 
Multivariable OS
VariableaHR (95% CI)P
Age (years) ≤70  
 >70 2.8 (1.4–5.8) 0.0047 
Grade G1+G2  
 G3 3.3 (1.2–9.0) 0.0189 
Periostin (nRQ) ≤1.6  
 >1.6 2.9 (1.3–6.9) 0.0134 

Abbreviation: nRQ, normalized relative quantification (median value).

aOnly statistically significant variables were reported.

IHC for periostin in leiomyosarcoma

To further investigate the association between periostin and outcome, IHC for this protein was performed on FFPE tissue samples of 31 leiomyosarcomas, 6 leiomyomas, and 15 NMT. Characteristics of this independent set of patients with leiomyosarcoma are shown in Table 1. The estimated median follow-up time was 3.4 years (1.9–9.5), with a 5-year OS of 46.2% (21.3–68.1) and a 10-year OS of 37.0% (13.6–60.9). At the time of this analysis, 38.7% (12/31) of the patients had died.

Tissue samples were classified as either low (scores 0–1) or high (scores 2–3) for periostin detection in stroma and cells. Fig. 1B shows representative IHC staining for periostin in FFPE sections of NMT (a), leiomyoma (b), and leiomyosarcoma (c and d). Stromal periostin detection was significantly different among the three tissue samples (P = 0.0484; Supplementary Table S3), and significantly higher in the stroma of leiomyomas (P = 0.0490) and leiomyosarcomas (P = 0.0447) compared to NMT. The difference between leiomyomas and leiomyosarcomas was not significant. Nevertheless, patients with leiomyosarcoma with high stromal periostin detection showed significantly reduced OS and increased risk (Fig. 1C) compared to patients with low stromal staining. Periostin detection in cells was lower (Supplementary Table S3) and the association with OS and HR in leiomyosarcomas did not reach significance (Fig. 1D).

Circulating levels of periostin

Periostin was quantified in serum/plasma samples from 67 patients with STS, whose characteristics are reported in Supplementary Table S4. The estimated median follow-up time was 7.3 years (4.4–14.1), with a 5-year OS of 55.7% (42.2–67.3) and a 10-year OS of 34.8% (20.6–49.4). At the time of this analysis, 55.2% (37/67) of the patients had died.

As for some patients only serum samples were available, we preliminarily measured protein concentration in matched serum and plasma samples available for 12 patients. Mean levels were analyzed and found to be comparable (meanserum = 22.70, meanplasma = 21.67; paired t test, P = 0.7), suggesting that, in our hands, the type of sample does not influence the assessment of periostin concentration in peripheral blood. Figure 2A shows the distribution of periostin levels according to patient characteristics. High amounts, that is, above the median concentration, of circulating periostin were significantly associated with G3 tumors and leiomyosarcoma and undifferentiated pleomorphic sarcoma histotypes. Elevated levels of circulating periostin were significantly associated with a shorter OS (Fig. 2B), and were confirmed to be an independent risk factor in a multivariable model, in addition to high grade (Table 3A).

Figure 2.

Circulating levels of periostin in 67 STS patients. A, Dot plot graphs showing the distribution of the levels of periostin, expressed as ng/mL, measured in serum/plasma samples according to patient characteristics. Statistical significance of the different distribution is also reported and was calculated by the Kruskal–Wallis test. P values were adjusted for multiple comparisons using the Benjamini–Hochberg method. Median and interquartile range are also indicated. Each serum/plasma sample was tested in duplicate in two independent ELISA assays. LMS, leiomyosarcoma; MFS, myxofibrosarcoma; UPS, undifferentiated pleomorphic sarcoma. B, Kaplan–Meier curves for OS according to the circulating levels of periostin. The P value was calculated by the log-rank test. HR, 95% CI, and the associated P value are also reported.

Figure 2.

Circulating levels of periostin in 67 STS patients. A, Dot plot graphs showing the distribution of the levels of periostin, expressed as ng/mL, measured in serum/plasma samples according to patient characteristics. Statistical significance of the different distribution is also reported and was calculated by the Kruskal–Wallis test. P values were adjusted for multiple comparisons using the Benjamini–Hochberg method. Median and interquartile range are also indicated. Each serum/plasma sample was tested in duplicate in two independent ELISA assays. LMS, leiomyosarcoma; MFS, myxofibrosarcoma; UPS, undifferentiated pleomorphic sarcoma. B, Kaplan–Meier curves for OS according to the circulating levels of periostin. The P value was calculated by the log-rank test. HR, 95% CI, and the associated P value are also reported.

Close modal
Table 3A.

Univariate and multivariable analyses for association of all clinical variables and circulating periostin with OS.

UnivariateMultivariable
VariableEvents/NHR (95% CI)PHR (95% CI)P
Age (years) ≤70 17/39    
 >70 20/28 1.9 (1.0–3.6) 0.0566   
Gender 19/36    
 18/31 1.8 (0.9–3.4) 0.0853   
Grade G1+G2 9/23   
 G3 28/44 3.2 (1.4–6.9) 0.0040 2.4 (1.1–5.6) 0.0352 
Tumor size ≤5 cm 5/12    
 >5 cm 32/55 2.2 (0.8–5.7) 0.1068   
Metastasis at diagnosis No 31/60   
 Yes 6/7 3.2 (1.3–.9) 0.0106 2.0 (0.8–5.1) 0.1280 
Histotype LMS 20/30    
 MFS 7/23 0.3 (0.1–0.7) 0.0073   
 UPS 10/14 1.2 (0.6–2.7) 0.5773   
Circulating periostin ≤20 ng/mL 13/35   
 >20 ng/mL 24/32 3.1 (1.6–6.1) 0.0012 2.6 (1.3–5.1) 0.0086 
UnivariateMultivariable
VariableEvents/NHR (95% CI)PHR (95% CI)P
Age (years) ≤70 17/39    
 >70 20/28 1.9 (1.0–3.6) 0.0566   
Gender 19/36    
 18/31 1.8 (0.9–3.4) 0.0853   
Grade G1+G2 9/23   
 G3 28/44 3.2 (1.4–6.9) 0.0040 2.4 (1.1–5.6) 0.0352 
Tumor size ≤5 cm 5/12    
 >5 cm 32/55 2.2 (0.8–5.7) 0.1068   
Metastasis at diagnosis No 31/60   
 Yes 6/7 3.2 (1.3–.9) 0.0106 2.0 (0.8–5.1) 0.1280 
Histotype LMS 20/30    
 MFS 7/23 0.3 (0.1–0.7) 0.0073   
 UPS 10/14 1.2 (0.6–2.7) 0.5773   
Circulating periostin ≤20 ng/mL 13/35   
 >20 ng/mL 24/32 3.1 (1.6–6.1) 0.0012 2.6 (1.3–5.1) 0.0086 

Abbreviations: LMS, leiomyosarcoma; MFS, myxofibrosarcoma; UPS, undifferentiated pleomorphic sarcoma.

A positive and significant correlation was found between circulating levels of the soluble protein and transcripts in 43 matched tumor tissues (Spearman correlation, ρs = 0.435, P = 0.0036), suggesting that tumor tissue may be the main source of circulating periostin.

EMT score

We combined periostin expression with that of other EMT/MET factors in a global EMT score to study the prognostic value of the overall contribution of epithelial and mesenchymal factors in STS. Network analysis was performed to study the strength and direction of correlations among the EMT/MET-related factors measured in our samples (Fig. 3A). This analysis showed that the most positively inter-correlated variables were ZO-1, ASMA, ZEB1, SIP1, vimentin, N-cadherin, and periostin (ρp range: 0.40–0.87, P ≤ 0.009), in descending order. ZO-1 expression was found to mainly increase concurrently with the expression of ASMA, ZEB1, and SIP1. Similarly, miR-100-5p was positively correlated to 3 epithelial miRNAs, whereas the three myo-miRNAs were highly intercorrelated and their expression was mainly linked to that of ZEB1 and SIP1 (Fig. 3A). Conversely, the expression of two mesenchymal markers, specifically SNAIL and SLUG, was negatively correlated to that of all other mesenchymal factors (Fig. 3A, ρp range: −0.35 to −0.49, P ≤ 0.0241). Therefore, the mesenchymal component was defined by clustering the gene expression data by principal component analysis of all mesenchymal markers except SNAIL and SLUG. The epithelial component was defined using scaled ZO-1 expression levels. Moreover, miRNAs were excluded from this analysis as their relevance in calculating the global EMT status of tumors needs to be further evaluated, as reported previously (21).

Figure 3.

EMT score in STS tissues, and chemoresistance and migration of primary STS cell lines. A, Pearson correlation network analysis of the analyzed biomarkers. Each node in the network represents an epithelial (E), a mesenchymal (M), or a myo-miR (My) biomarker. The thickness of the edge between two nodes represents the strength of correlation. Positive correlations are green, negative correlations are red. Correlations ranged from −0.5 to 1. Correlations with absolute value less than 0.4 are not drawn. B, Kaplan–Meier curves for overall survival according to the EMT score. P values were calculated as reported above. C, Vertical bar charts of EMT scores of 47 samples of STS aligned from the most transitioning to the most mesenchymal. Leiomyosarcoma (L) are reported in red, myxofibrosarcoma (M) in green, and undifferentiated pleomorphic sarcoma (U) in blue, and the histotype is also indicated under each bar. D, Dose–response curves of epirubicin sensitivity of primary STS cell lines. The results, obtained in S57 cells, derived from a myxofibrosarcoma (MFS), in S78 and S79 cells, derived from two leiomyosarcomas (LMS), are the mean ± SD of two independent experiments performed in quadruplicate. Dotted line indicates 50% cell viability relative to untreated cells. The experiments were performed at passage (p) 90 and p92 of S57, p14–15 of S78, and p10–12 of S79. EMT score values of cell lines are reported in the legend. E, Representative images of cell scratch assays performed in S57 (p87–90) and S79 (p35–37) cells after 24 and 48 hours of mock- or TGFβ1 treatment. The bar charts on the right side of the images show the mean ± SD of two independent experiments performed in duplicate at 24 and 48 hours. Differences in migratory capacity were assessed using the Student t test and P values were adjusted for multiple comparisons (*, P < 0.05).

Figure 3.

EMT score in STS tissues, and chemoresistance and migration of primary STS cell lines. A, Pearson correlation network analysis of the analyzed biomarkers. Each node in the network represents an epithelial (E), a mesenchymal (M), or a myo-miR (My) biomarker. The thickness of the edge between two nodes represents the strength of correlation. Positive correlations are green, negative correlations are red. Correlations ranged from −0.5 to 1. Correlations with absolute value less than 0.4 are not drawn. B, Kaplan–Meier curves for overall survival according to the EMT score. P values were calculated as reported above. C, Vertical bar charts of EMT scores of 47 samples of STS aligned from the most transitioning to the most mesenchymal. Leiomyosarcoma (L) are reported in red, myxofibrosarcoma (M) in green, and undifferentiated pleomorphic sarcoma (U) in blue, and the histotype is also indicated under each bar. D, Dose–response curves of epirubicin sensitivity of primary STS cell lines. The results, obtained in S57 cells, derived from a myxofibrosarcoma (MFS), in S78 and S79 cells, derived from two leiomyosarcomas (LMS), are the mean ± SD of two independent experiments performed in quadruplicate. Dotted line indicates 50% cell viability relative to untreated cells. The experiments were performed at passage (p) 90 and p92 of S57, p14–15 of S78, and p10–12 of S79. EMT score values of cell lines are reported in the legend. E, Representative images of cell scratch assays performed in S57 (p87–90) and S79 (p35–37) cells after 24 and 48 hours of mock- or TGFβ1 treatment. The bar charts on the right side of the images show the mean ± SD of two independent experiments performed in duplicate at 24 and 48 hours. Differences in migratory capacity were assessed using the Student t test and P values were adjusted for multiple comparisons (*, P < 0.05).

Close modal

Each tumor sample was then quantitatively scored by subtracting the epithelial component from the mesenchymal component. The obtained EMT scores, ranging from −2.68 to 5.36, were initially subdivided into tertiles, to adjust for the three “canonical” EMT states, that is, epithelial (from −2.68 to -0.55), transitioning/hybrid (from −0.55 to −0.15), and mesenchymal (higher than −0.15). However, a canonical, “pure” epithelial profile does not apply to the sarcoma subtypes under study, as evidenced by the network analysis that showed that ZO-1 expression was strongly positively correlated to that of mesenchymal factors. Therefore a more appropriate, sarcoma-tailored classification should only include a transitioning and a mesenchymal state. Consequently, the first two tertiles were joined into a single transitioning state and tumor samples were subdivided into transitioning and mesenchymal. In univariate analysis, patients with a mesenchymal score showed an increased risk and reduced OS (Fig. 3B). In a multivariable model, a mesenchymal EMT score was confirmed to be an independent, unfavorable prognostic factor, in addition to advanced age and G3 grade (Table 3B). The distribution of samples according to the calculated EMT score is shown in Fig. 3C. Interestingly, a cluster of leiomyosarcoma samples was frankly mesenchymal.

Table 3B.

Multivariable model for association of statistically significant clinical variables and EMT score with OS.

Multivariable OS
VariableaHR (95% CI)P
Age (years) ≤70  
 >70 6.0 (2.5–14.6) <0.0001 
Grade G1+G2  
 G3 4.6 (1.7–12.7) 0.0034 
EMT Score Tran  
 Mes 5.2 (2.1–13.0) 0.0005 
Multivariable OS
VariableaHR (95% CI)P
Age (years) ≤70  
 >70 6.0 (2.5–14.6) <0.0001 
Grade G1+G2  
 G3 4.6 (1.7–12.7) 0.0034 
EMT Score Tran  
 Mes 5.2 (2.1–13.0) 0.0005 

Abbreviations: Mes, mesenchymal; Tran, transitioning.

aOnly statistically significant variables were shown.

In vitro chemoresistance and migration of primary STS cells

Three primary sarcoma cell lines were isolated: S57, derived from a myxofibrosarcoma, and S78 and S79, derived from two leiomyosarcomas. S57 and S79 formed a uniform monolayer at confluence, whereas the S78 culture was composed of long spindle-shaped cells with a pattern of multilayered growth at confluence (Supplementary Fig. S3A). S57 became a stable, self-immortalized cell line after the first 30 passages (p), and it maintained a reliable proliferative rate. S79, after a period of slow growth, reached p37, whereas S78 showed late senescence at p21. STR profiles of these cell lines (Supplementary Table S5) and IHC analyses (Supplementary Fig. S3B) confirmed their unique identity and their derivation.

As sarcoma aggressiveness may be linked to increased therapeutic resistance, we investigated the in vitro sensitivity of these cell lines to epirubicin, an anthracycline used as first-line chemotherapy in aggressive STS subtypes. As shown in Fig. 3D, S57 cells were found to be more sensitive to the cytotoxic activity of epirubicin (IC50 = 2.5 ± 0.5 μmol/L) compared to the two leiomyosarcoma cell lines, as S79 showed a 5.6-fold higher IC50 (13.9 ± 0.61 μmol/L) and S78 was resistant to all analyzed doses. qRT-PCR analyses and EMT score computation showed that S57 was characterized by a hybrid EMT profile, whereas S78 and S79 exhibited a mesenchymal score, as reported in Fig. 3D.

To examine the migration ability of S57 and S79 cells, cell scratch assays were performed in complete medium (mock) and after EMT induction by TGFβ1 treatment (Fig. 3E). Mock-treated S79 cells showed a significantly faster migration into the scratch compared with S57 cells at 24 hours. EMT induction significantly enhanced migration in both cell lines at 24 hours, and promoted a marked morphologic change in S57 cells, characterized by a criss-cross pattern of growth (Fig. 3E). qRT-PCR analyses confirmed a transcriptional reprogramming in EMT-induced cells, resulting in a shift to a more mesenchymal profile, as indicated by the EMT score, even though S57 maintained a hybrid state (Fig. 3E).

Periostin is a matricellular protein implicated in EMT and tumor progression (24). Under physiologic conditions, it localizes at collagen-rich regions to contribute to the mechanical strength of connective tissues (24), that is, the site of sarcoma origin. Diverse normal fetal and adult tissues express periostin at variable levels, mainly depending on stromal abundance (25), but its marked upregulation and secretion have been evidenced in sites of injury and inflammation as well as in several tumor types, highlighting its relevance in the neoplastic process (26). Its protumorigenic activity is linked to integrin binding with subsequent activation of Akt/PKB- and FAK-mediated signaling pathways as well as to its ability to promote remodeling of extracellular matrix (ECM), processes implicated in cell survival, invasiveness, metastasis, and EMT (24, 27–29). Indeed, periostin overexpression and increased detection, mainly in tumor stroma, were associated with poor prognosis in many neoplasms, including colorectal carcinoma, epithelial ovarian cancer, esophageal adenocarcinoma, non–small cell lung cancer, breast cancer, hepatocellular carcinoma, and prostate cancer (refs. 24, 26, and references therein).

Concerning sarcoma, to date periostin expression was shown to be associated with poor survival in patients with osteosarcoma (30). Interestingly, periostin was found to be mainly localized in tumor cells in osteosarcoma tissues, and correlated with vessel density (31). Scanty data are instead available on its detection in STS (32). Our results clearly show for the first time that this multifunctional protein may be overexpressed in tissue samples of the three studied subtypes. Moreover, it may accumulate at high levels in the stromal compartment of leiomyosarcomas. High expression in ex vivo STS tumor samples and high accumulation in leiomyosarcoma stroma were significantly associated with poor survival, indicating that periostin may be a reliable indicator of an active tumor also in patients affected by leiomyosarcoma, myxofibrosarcoma, or undifferentiated pleomorphic sarcoma. As the sample size of this pilot study is small, further studies are needed to validate our findings in an independent and larger STS cohort, possibly also including patients affected by other STS histotypes, and to assess whether this matricellular component might represent a new therapeutic target in aggressive STS.

As other matricellular proteins, periostin is measurable in body fluids, and its increased circulating amounts were associated with aggressiveness in different tumor types (24). Here, data obtained in 67 STS patients supported the prognostic role of this protein, likely secreted by the tumor tissues, as circulating marker. Given the large amount of data indicating that periostin, measured in solid and/or liquid biopsies, might be a prognostic factor in several tumor types, its usefulness as a biomarker for risk assessment on a routine basis should be considered. It has to be pointed out that analyses of association with OS in our exploratory study are data driven, and further analyses are needed to establish whether universal cut-off values for periostin expression and plasmatic levels can be determined or, alternatively, adjusted according to the clinical setting.

No relevant differences were found between detection of periostin in leiomyomas and leiomyosarcomas by IHC. Despite the low number of cases analyzed, this is not a surprising result since leiomyoma is a fibrotic, benign tumor mainly characterized by massive ECM accumulation and remodeling (33). Attention has been focused on several ECM components, but not periostin, responsible for tissue stiffness, which could represent new therapeutic targets for treatment of this common pathology (33). Therefore, periostin too might be a relevant player in this pathologic condition, and further studies are needed to prove this incidental, interesting finding.

The lack of expression of E-cadherin in all our tissue samples may be linked to the number of analyzed samples as well as to the high expression levels of mesenchymal transcriptional repressors of this structural epithelial protein. In parallel, we measured the expression of another epithelial marker, ZO-1, which was found to be positively intercorrelated with mesenchymal markers in the three histotypes, as previously shown for E-cadherin (6). This finding suggests that some STS might express hybrid, metastable profiles between the epithelial and mesenchymal states (4, 5) not necessarily linked to the presence of E-cadherin. Hybrid states of EMT were initially demonstrated in circulating carcinoma cells and were shown to characterize more invasive carcinomas (34, 35). We applied an EMT scoring method (21) based on the expression of a small set of genes including periostin, and subdivided tumor samples into transitioning and mesenchymal. Our data showed that a transitioning EMT score identified STS patients with a better survival. These findings, although limited to a small STS cohort, are in agreement with in silico RNA-seq data analyses of 250 STS samples that correlated a hybrid EMT signature with a favorable outcome (36).

Recent studies conducted in carcinomas highlighted that EMT is frequently not complete in tumor cells, which were shown to exploit multiple transitioning states (9, 10). It is conceivable that different hybrid forms might be linked to opposite outcomes according to the epithelial or mesenchymal state ab initio. In fact, in epithelial tumors, hybrid forms with a higher content of epithelial proteins and limited mesenchymal factors might exhibit an aggressive, metastatic behavior (37). The term “epithelial–mesenchymal plasticity” resumes the multiple bidirectional EMT/MET pathways controlling the flexible attitude of a tumor cell in terms of metastatic potential, stemness, and therapeutic resistance (9, 10). In light of these recent findings, plasticity programs in sarcoma cells need to be fully investigated. Partial EMT states in sarcoma might derive from the acquisition of few epithelial features in a massive mesenchymal landscape, thus possibly attenuating the aggressive nature of these tumors. To explore this hypothesis, we isolated three primary sarcoma cell lines that showed different, intrinsic EMT states, in line with our data on tumor tissues. A self-immortalized cell line derived from a myxofibrosarcoma exhibited a transitioning EMT profile and showed a sensitivity to epirubicin higher than that measured in two leiomyosarcoma cell lines, characterized by a more mesenchymal state. Moreover, an intrinsic or induced mesenchymal state was found to be associated with increased migration. Although preliminary, these findings suggest that the attenuated aggressiveness of sarcomas residing in a hybrid state might be linked to reduced chemoresistance and migration.

In conclusion, results of this exploratory study suggest that expression and circulation of periostin might support clinic-pathologic parameters in risk stratification of patients with the three analyzed STS histotypes. An EMT score including periostin might also add valuable information. Interestingly, our data highlight that, while a transitioning EMT state in carcinomas is associated with tumor aggressiveness, a hybrid profile in STS might predict a more favorable outcome, suggesting that the clinical relevance of the hybrid EMT phenotypes might depend on tumor type. In vitro studies suggested that a hybrid EMT state might attenuate aggressiveness of sarcoma cells by reducing therapeutic resistance and migration. Further clinical and biological studies are needed.

No potential conflicts of interest were disclosed.

Conception and design: A. Brunello, M.L. Calabrò

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Brunello, R. Cappellesso, P. Del Fiore, M. Rastrelli, A. Sommariva, M.C. Montesco, C.R. Rossi, V. Zagonel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.A. Piano, A. Brunello, R. Cappellesso, P. Del Bianco, A. Mattiolo, C. Fritegotto, B. Montini, C. Zamuner, G.L. De Salvo, M.C. Montesco, M.L. Calabrò

Writing, review, and/or revision of the manuscript: M.A. Piano, A. Brunello, R. Cappellesso, P. Del Bianco, A. Mattiolo, B. Montini, A. Sommariva, M.C. Montesco, M.L. Calabrò

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Brunello, C. Zamuner, P. Del Fiore

Study supervision: G.L. De Salvo, C.R. Rossi, V. Zagonel

Other (biological assays, molecular and biological data analyses): M.A. Piano

Other (molecular and biological analyses): A. Mattiolo, C. Fritegotto, B. Montini

The authors thank Giuseppe Opocher for constant support, Maura Digito and Clara Benna of the biobank of tumor tissues of the Department of Surgery, Oncology and Gastroenterology of the University of Padua (Padua, Italy), all the persons who contributed to the establishment and governance of the biobank, Sandro Pasquali and Mark Boyd for helpful comments, Claudia Sedda for contributing to the assessment of circulating periostin, Christina Drace for help in preparing the manuscript and Pierantonio Gallo for artwork. This work was supported by intramural grants from 5 × 1000-IOV2010 (grant CUP no. J94G13000140001, to M.L. Calabrò and A. Brunello) and from 5 × 1000-IOV2012 (grant CUP no. J96D16000040005, to M.L. Calabrò, A. Brunello, M.C. Montesco, and C.R. Rossi). M.A. Piano, A. Mattiolo, C. Fritegotto, B. Montini, and C. Zamuner were recipients of a Ricerca Corrente fellowship, Italian Ministry of Health (IMH). IMH funding was also used for payment of the publication fee.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Clark
MA
,
Fisher
C
,
Judson
I
,
Thomas
JM
. 
Soft-tissue sarcomas in adults
.
N Engl J Med
2005
;
353
:
701
11
.
2.
Fletcher
CDM
,
Bridge
JA
,
Hogendoorn
PCW
,
Mertens
F
.
WHO classification of tumours of soft tissue and bone
.
Lyon, France
:
IARC
; 
2013
. p.
468
.
3.
Cancer Genome Atlas Research Network
. 
Comprehensive and integrated genomic characterization of adult soft tissue sarcomas
.
Cell
2017
;
171
:
950
65
.
4.
Kahlert
UD
,
Joseph
JV
,
Kruyt
FAE
. 
EMT- and MET-related processes in nonepithelial tumors: importance for disease progression, prognosis, and therapeutic opportunities
.
Mol Oncol
2017
;
11
:
860
77
.
5.
Sannino
G
,
Marchetto
A
,
Kirchner
T
,
Grunewald
TGP
. 
Epithelial-to-mesenchymal and mesenchymal-to-epithelial transition in mesenchymal tumors: a paradox in sarcomas?
Cancer Res
2017
;
77
:
4556
61
.
6.
Yang
J
,
Du
X
,
Wang
G
,
Sun
Y
,
Chen
K
,
Zhu
X
, et al
Mesenchymal to epithelial transition in sarcomas
.
Eur J Cancer
2014
;
50
:
593
601
.
7.
Tian
W
,
Wang
G
,
Yang
J
,
Pan
Y
,
Ma
Y
. 
Prognostic role of E-cadherin and Vimentin expression in various subtypes of soft tissue leiomyosarcomas
.
Med Oncol
2013
;
30
:
401
.
8.
Alba-Castellon
L
,
Batlle
R
,
Franci
C
,
Fernandez-Acenero
MJ
,
Mazzolini
R
,
Pena
R
, et al
Snail1 expression is required for sarcomagenesis
.
Neoplasia
2014
;
16
:
413
21
.
9.
Gupta
PB
,
Pastushenko
I
,
Skibinski
A
,
Blanpain
C
,
Kuperwasser
C
. 
Phenotypic plasticity: driver of cancer initiation, progression, and therapy resistance
.
Cell Stem Cell
2019
;
24
:
65
78
.
10.
Lu
W
,
Kang
Y
. 
Epithelial-mesenchymal plasticity in cancer progression and metastasis
.
Dev Cell
2019
;
49
:
361
74
.
11.
McShane
LM
,
Altman
DG
,
Sauerbrei
W
,
Taube
SE
,
Gion
M
,
Clark
GM
, et al
REporting recommendations for tumour MARKer prognostic studies (REMARK)
.
Br J Cancer
2005
;
93
:
387
91
.
12.
Piano
MA
,
Gianesello
L
,
Grassi
A
,
Del Bianco
P
,
Mattiolo
A
,
Cattelan
AM
, et al
Circulating miRNA-375 as a potential novel biomarker for active Kaposi's sarcoma in AIDS patients
.
J Cell Mol Med
2019
;
23
:
1486
94
.
13.
Lignitto
L
,
Mattiolo
A
,
Negri
E
,
Persano
L
,
Gianesello
L
,
Chieco-Bianchi
L
, et al
Crosstalk between the mesothelium and lymphomatous cells: insight into the mechanisms involved in the progression of body cavity lymphomas
.
Cancer Med
2014
;
3
:
1
13
.
14.
Livak
KJ
,
Schmittgen
TD
. 
Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) method
.
Methods
2001
;
25
:
402
8
.
15.
Mestdagh
P
,
Van Vlierberghe
P
,
De Weer
A
,
Muth
D
,
Westermann
F
,
Speleman
F
, et al
A novel and universal method for microRNA RT-qPCR data normalization
.
Genome Biol
2009
;
10
:
R64
.
16.
Renner
M
,
Czwan
E
,
Hartmann
W
,
Penzel
R
,
Brors
B
,
Eils
R
, et al
MicroRNA profiling of primary high-grade soft tissue sarcomas
.
Genes Chromosomes Cancer
2012
;
51
:
982
96
.
17.
Subramanian
S
,
Lui
WO
,
Lee
CH
,
Espinosa
I
,
Nielsen
TO
,
Heinrich
MC
, et al
MicroRNA expression signature of human sarcomas
.
Oncogene
2008
;
27
:
2015
26
.
18.
Chen
D
,
Sun
Y
,
Yuan
Y
,
Han
Z
,
Zhang
P
,
Zhang
J
, et al
miR-100 induces epithelial-mesenchymal transition but suppresses tumorigenesis, migration and invasion
.
PLoS Genet
2014
;
10
:
e1004177
.
19.
Brabletz
S
,
Brabletz
T
. 
The ZEB/miR-200 feedback loop–a motor of cellular plasticity in development and cancer?
EMBO Rep
2010
;
11
:
670
7
.
20.
Kao
CJ
,
Martiniez
A
,
Shi
XB
,
Yang
J
,
Evans
CP
,
Dobi
A
, et al
miR-30 as a tumor suppressor connects EGF/Src signal to ERG and EMT
.
Oncogene
2014
;
33
:
2495
503
.
21.
Tan
TZ
,
Miow
QH
,
Miki
Y
,
Noda
T
,
Mori
S
,
Huang
RY
, et al
Epithelial-mesenchymal transition spectrum quantification and its efficacy in deciphering survival and drug responses of cancer patients
.
EMBO Mol Med
2014
;
6
:
1279
93
.
22.
Burton
JD
. 
The MTT assay to evaluate chemosensitivity
.
Methods Mol Med
2005
;
110
:
69
78
.
23.
Grada
A
,
Otero-Vinas
M
,
Prieto-Castrillo
F
,
Obagi
Z
,
Falanga
V
. 
Research techniques made simple: analysis of collective cell migration using the wound healing assay
.
J Invest Dermatol
2017
;
137
:
e11
e6
.
24.
Gonzalez-Gonzalez
L
,
Alonso
J
. 
Periostin: a matricellular protein with multiple functions in cancer development and progression
.
Front Oncol
2018
;
8
:
225
.
25.
Tilman
G
,
Mattiussi
M
,
Brasseur
F
,
van Baren
N
,
Decottignies
A
. 
Human periostin gene expression in normal tissues, tumors and melanoma: evidences for periostin production by both stromal and melanoma cells
.
Mol Cancer
2007
;
6
:
80
.
26.
Liu
AY
,
Zheng
H
,
Ouyang
G
. 
Periostin, a multifunctional matricellular protein in inflammatory and tumor microenvironments
.
Matrix Biol
2014
;
37
:
150
6
.
27.
Gillan
L
,
Matei
D
,
Fishman
DA
,
Gerbin
CS
,
Karlan
BY
,
Chang
DD
. 
Periostin secreted by epithelial ovarian carcinoma is a ligand for alpha(V)beta(3) and alpha(V)beta(5) integrins and promotes cell motility
.
Cancer Res
2002
;
62
:
5358
64
.
28.
Baril
P
,
Gangeswaran
R
,
Mahon
PC
,
Caulee
K
,
Kocher
HM
,
Harada
T
, et al
Periostin promotes invasiveness and resistance of pancreatic cancer cells to hypoxia-induced cell death: role of the beta4 integrin and the PI3k pathway
.
Oncogene
2007
;
26
:
2082
94
.
29.
Erkan
M
,
Kleeff
J
,
Gorbachevski
A
,
Reiser
C
,
Mitkus
T
,
Esposito
I
, et al
Periostin creates a tumor-supportive microenvironment in the pancreas by sustaining fibrogenic stellate cell activity
.
Gastroenterology
2007
;
132
:
1447
64
.
30.
Hu
F
,
Wang
W
,
Zhou
HC
,
Shang
XF
. 
High expression of periostin is dramatically associated with metastatic potential and poor prognosis of patients with osteosarcoma
.
World J Surg Oncol
2014
;
12
:
287
.
31.
Hu
F
,
Shang
XF
,
Wang
W
,
Jiang
W
,
Fang
C
,
Tan
D
, et al
High-level expression of periostin is significantly correlated with tumour angiogenesis and poor prognosis in osteosarcoma
.
Int J Exp Pathol
2016
;
97
:
86
92
.
32.
Brown
JM
,
Mantoku
A
,
Sabokbar
A
,
Oppermann
U
,
Hassan
AB
,
Kudo
A
, et al
Periostin expression in neoplastic and non-neoplastic diseases of bone and joint
.
Clin Sarcoma Res
2018
;
8
:
18
.
33.
Leppert
PC
,
Jayes
FL
,
Segars
JH
. 
The extracellular matrix contributes to mechanotransduction in uterine fibroids
.
Obstet Gynecol Int
2014
;
2014
:
783289
.
34.
Jolly
MK
,
Boareto
M
,
Huang
B
,
Jia
D
,
Lu
M
,
Ben-Jacob
E
, et al
Implications of the hybrid epithelial/mesenchymal phenotype in metastasis
.
Front Oncol
2015
;
5
:
155
.
35.
Nieto
MA
,
Huang
RY
,
Jackson
RA
,
Thiery
JP
.
Emt: 2016. Cell
2016
;
166
:
21
45
.
36.
Somarelli
JA
,
Shetler
S
,
Jolly
MK
,
Wang
X
,
Bartholf Dewitt
S
,
Hish
AJ
, et al
Mesenchymal-epithelial transition in sarcomas is controlled by the combinatorial expression of microRNA 200s and GRHL2
.
Mol Cell Biol
2016
;
36
:
2503
13
.
37.
Pastushenko
I
,
Brisebarre
A
,
Sifrim
A
,
Fioramonti
M
,
Revenco
T
,
Boumahdi
S
, et al
Identification of the tumour transition states occurring during EMT
.
Nature
2018
;
556
:
463
8
.