Access the full text.
Sign up today, get DeepDyve free for 14 days.
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
Hindawi Journal of Oncology Volume 2020, Article ID 8106212, 11 pages https://doi.org/10.1155/2020/8106212 Research Article Prognostic Value of a Three-DNA Methylation Biomarker in Patients with Soft Tissue Sarcoma 1 1,2 2 3 2 1 Yuxiao Chen , Rui Zhu , Min Chen , Wenna Guo, Xin Yang , Xin-Jian Xu , and Liucun Zhu Department of Mathematics, Shanghai University, Shanghai 200444, China School of Life Sciences, Shanghai University, Shanghai 200444, China School of Life Sciences, Zhengzhou University, Zhengzhou, Henan 450001, China Correspondence should be addressed to Xin-Jian Xu; firstname.lastname@example.org and Liucun Zhu; email@example.com Received 5 January 2020; Accepted 25 April 2020; Published 15 May 2020 Academic Editor: Rossana Berardi Copyright © 2020 Yuxiao Chen et al. *is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Soft tissue sarcomas (STS) are a highly aggressive and heterogeneous group of malignant mesenchymal tumors. *e prognosis of patients with advanced or metastatic STS remains poor, and the main therapy of STS patients combines primary surgery, radiotherapy, and chemotherapy. Aberrant DNA methylation shows close association with the pathogenesis and tumor pro- gression. *erefore, DNA methylation biomarkers might have the potential in accurately predicting the survival of STS patients. In order to identify a prognostic biomarker based on DNA methylation sites, a comprehensive analysis of the DNA methylation proﬁle of STS patients in the Cancer Genome Atlas (TCGA) database was performed. All samples were randomly divided into training and testing datasets. Cox proportional hazards regression analysis was performed to identify a prognostic biomarker that contains three DNA methylation sites. *e Kaplan–Meier analysis demonstrated that the 3-DNA methylation biomarker dis- criminated patients into high-risk and low-risk groups, both in the training and in the testing datasets, and the area under the receiver operating characteristic curve values (AUCs) were 0.844 (P< 0.001, 95% CI: 0.740–0.948) and 0.710 (P � 0.002, 95% CI: 0.595–0.823), respectively. Besides, this biomarker presented superior prognostic performance in STS patients with diﬀerent age, sex, tissue of origin, therapy, and histologic subtypes. Compared with other prognostic biomarkers, this biomarker tended to be a more precise prognostic factor in STS patients. Moreover, methylation sites in this biomarker might provide a new way for clinicians to make decisions regarding the intervention and assess the eﬀectiveness of an individual therapeutic strategy. only 15 months, especially in patients with lung metastases . 1. Introduction For most tumors, risk stratiﬁcation and targeted therapy can Soft tissue sarcomas (STS) are a heterogeneous group of signiﬁcantly improve the therapeutic eﬀects, and researches on malignant tumors that mainly arise from mesenchymal stem STS show similar results . Surgery is often considered cu- cells which can present in various parts of the human body, rative for localized STS, radiotherapy can degrade the risk of including the extremities, trunk, and retroperitoneum . local recurrence, and chemotherapy is usually reserved for More than 50 tumor subtypes that are relevant to STS have managing metastatic patients [7–9]. *ese studies have indi- been identiﬁed . According to the recent statistics, the STS cated that it is important to stratify the risks of STS patients cases might reach an estimated number of 12,750 and 5,270 accurately and select appropriate therapeutic modalities for deaths worldwide in 2019, accounting for 0.7% of all cancer patient management in the future. However, the current re- cases and 0.9% of cancer deaths separately . *e survival rate search studies in this ﬁeld, which are referred to as prognostic and prognosis of patients with STS remain very disappointing. researches, are relatively less and limited to focus on a speciﬁc *e 5-year survival rate of STS patients is about 50%, showing STS histologic subtype that is of no help to the risk stratiﬁ- no improvements over the years . *e median survival is cation of most STS patients [10–12]. 2 Journal of Oncology In recent years, DNA methylation has been widely measured based on the Inﬁnium HumanMethylation450 regarded as a prognostic biomarker with great potential. BeadChip. *e β-values of 485577 DNA methylation sites Aberrant DNA methylation plays a crucial role in cancer that represent the ratio between the methylated array in- pathogenesis and progression . Hypermethylation of the tensity and the total array intensity falling between 0 (no promotor regions inhibits certain tumor suppressor genes, methylation) and 1 (full methylation) were listed in each which is considered as a shared trait of many malignant sample. During the process of data cleaning, the samples that tumors . Genetic and nongenetic factors shape DNA satisﬁed the following criteria were removed: (1) diagnosed methylation collectively, which can reﬂect the cumulative as nonsarcoma; (2) nonprimary solid tumors; (3) samples exposure of patients towards some risk factors such as were not obtained from the frozen tissues; and (4) the transgenerational inheritance, in utero environment, obe- survival time of patients was equal to 0 days. Meanwhile, the sity, smoking, age, chronic inﬂammation, etc. . *ere- methylation sites containing missing values were excluded. fore, DNA methylation markers might aid in deducing Accordingly, 257 samples with 374,831 methylation sites tumor pathogenesis and progression. Furthermore, DNA were retained for subsequent analysis. *e histological methylation markers are more stable when compared with subtypes of these samples belong to STS (Supplementary 1). RNA- and protein-based markers . Due to these ad- It is worth noting that patients are censored if they are alive, vantages, more and more studies have illustrated the po- and whose survival is recorded as the number of days from tentiality of DNA methylation as a prognostic marker in the start till the end of follow-up. *e entire steps for data cancers. For example, in renal cell carcinoma, GATA5 cleaning are presented in Figure 1. All samples were ran- hypermethylation is frequently presented and shows a sig- domly grouped, and half of which was used as a training niﬁcant association with shortened progression-free survival dataset (Supplementary 2) for constructing the prognostic . Hypomethylation of long interspersed nucleotide el- model and the remaining half as a testing dataset (Sup- ement-1 (LINE-1) is unfavorable with regard to the prog- plementary 3) for verifying the eﬃcacy of the model. nosis of colorectal cancer in patients . Some scholars have conducted studies on STS as well. Hypermethylation of 2.2. Statistical Analysis. Statistical analysis involved in this RASSF1A promoter is associated with shorter duration of study was based on the open-source software R (version survival in patients with stage II and III STS . DNA 3.4.4) and Perl (version 18). *e survival package in R was (INK4a) hypermethylation in p16 gene promoter can lead to used for Cox proportional hypothesis test and Cox pro- the shorter survival in patients with malignant STS . portional hazards regression analysis. *e ROC analysis was However, most of the STS methylation analyses involve a performed to demonstrate the accuracy of the biomarker small number of samples or merely focus on the methylation model in predicting the overall survival (OS) of patients, and of a single speciﬁc gene. the AUCs were calculated by Perl. *e Kaplan–Meier *e Cancer Genome Atlas (TCGA) is a large-scale and analysis was used to clarify the relationship between the open-access data platform that provides 269 soft tissue prognostic model and the OS of the patient. *e above sarcoma samples and a corresponding DNA methylation analysis involved some statistical hypothesis tests, in which proﬁle. To obtain a more systematic and comprehensive the likelihood ratio test was used to evaluate the signiﬁcance understanding of how DNA methylation takes part in the of one or more independent variables in Cox regression pathogenesis and development of STS, the genome-wide analysis . Whether the AUC was signiﬁcantly equal to 0.5 DNA methylation proﬁle of STS samples in TCGA was was examined by Z-test . *e Log-rank test and Breslow analyzed. As a consequence, a 3-DNA methylation prog- test were used to assess signiﬁcant diﬀerence in the cu- nostic biomarker was identiﬁed with multivariate Cox re- mulative survival rate between patients with high and low gression analysis. *e receiver operating characteristic risks. *e former is sensitive to long-term survival, while the (ROC) curve and Kaplan–Meier analysis were performed to latter is more sensitive to short-term survival . *e show the performance of this biomarker in predicting the diﬀerence in DNA methylation levels between patients with survival of STS patients. *e predictive utility of this bio- short-term survival (≤3 years) and long-term survival (>3 marker was further validated by regrouping patients with years) was tested by the Mann–Whitney U test. diﬀerent clinical characteristics. Meanwhile, the prognostic performance of the 3-DNA methylation biomarker with other known molecular biomarkers was compared. Fur- 2.3. Establishment of the Prognostic Biomarker Model. thermore, we analyzed the biological function and the latent First, the Cox proportional hazards hypothesis test (P> 0.05) impacts of these methylation sites on the pathogenesis and and univariate Cox proportional hazards regression analysis development of STS, and a few recommendations directed at (P< 0.01) were used to screen DNA methylation sites that the therapy of STS patients were proposed. were signiﬁcantly associated with the OS of STS patients, and sites in the training dataset that satisﬁed the requirements were used as candidate markers. *en, all the possible combinations 2. Materials and Methods of two, three, four, and ﬁve candidate markers were con- 2.1. Patients and DNA Methylation Data from TCGA. structed, and multivariate Cox regression analysis was used to Clinical information of 269 STS patients and relevant DNA further screen these combinations that were signiﬁcantly as- sociated with patient survival. Finally, according to the P methylation data involved in this study were downloaded from the TCGA database. *e DNA methylation proﬁle was values of likelihood ratio test, a 3-DNA methylation biomarker Journal of Oncology 3 where i represents the patients and n represents the number STS DNA methylation of DNA methylation sites in the model; a is the Cox re- data in TCGA j (269, 485577) gression coeﬃcient of site j. β is the methylation level of site i,j j in patient i. According to formula (1), the risk scores of all patients were calculated, and the median risk score was set as a threshold to classify patients into the low-risk group and high-risk group. Next, the Kaplan–Meier curve and Log- rank test and Breslow test were used to assess the diﬀerences Tumor samples Normal samples regarding the cumulative probability of survival between the (265, 485577) (4, 485577) two groups, both in the training and testing datasets. Fur- thermore, the whole dataset was regrouped based on dif- ferent clinical characteristics such as sex, age, tissues of origin, histological subtypes, and adjuvant therapeutic modalities to further validate the prognostic ability of the signature by performing ROC and Kaplan–Meier analyses. Primary solid tumor Solid tumor of came from frozen tissue other types (260, 485577) (5, 485577) 3. Results 3.1. Clinical Characteristics of STS Patients. *e median age of the 257 STS patients included in this study was 61 (ranging from 20 to 89) and the mean survival was 1189 days (ranging within 15–5723 days). In this dataset, the number Survival days of samples Survival days of samples of male patients diﬀered from the number of female patients not equal to 0 equal to 0 to a lesser extent, which was 117 (45.53%) and 140 (54.47%), (257, 485577) (3, 485577) respectively. Besides, the locations of STS were various. Most of the samples were originated from the connective soft tissues in head and neck, chest, pelvis, limbs, and trunk (N = 114, 44.36%). *is category was brieﬂy abbreviated as connective soft tissue. *e rest of the samples were origi- nated from the retroperitoneum (N = 99, 38.52%), uterus DNA methylation DNA methylation (N = 31, 12.06%), and other tissues (N = 13, 5.06%) including sites without ‘NA’ values sites with ‘NA’ values (257, 374831) (257, 110746) kidney, stomach, and ovary. According to the histological subtypes, all samples were classiﬁed into 39 (15.17%) cases Figure 1: Data cleaning steps for STS DNA methylation data. In with ﬁbromyxosarcoma, 59 (22.96%) cases with lip- the bracket of each box, the number of samples was on the left and osarcoma, 104 (40.47%) cases with leiomyosarcoma, 10 the number of methylation sites on the right. *e total data of 269 (3.89%) cases with synovial sarcoma, 3 (1.17%) cases with STS samples and 485577 methylation sites were downloaded from giant cell sarcoma, 9 (3.5%) cases with malignant peripheral the TCGA database. First of all, 4 normal samples were removed, as nerve sheath tumor, and 33 (12.84%) cases with undiﬀer- this study aimed to investigate the prognosis of STS patients. entiated sarcoma. All these 257 patients underwent surgery; Among the remaining 265 tumor samples, 5 other types of solid among them, 137 (53.31%) cases received adjuvant phar- tumor samples were dropped, which included 3 recurrent solid maceutical therapy and the remaining 120 (46.69%) received tumor samples, 1 metastatic solid tumor sample, and 1 primary solid tumor sample obtained from formalin-ﬁxed paraﬃn-em- adjuvant radiation therapy. *e detailed information against bedded tissue that was proven to be ineﬀective for sequencing clinical characteristics is summarized in Table 1. analysis. As a result, 260 primary solid tumor samples were retained. To avoid possible misunderstandings on the follow-up survival time of STS patients, 3 samples with a survival time of 0 3.2. Establishment of the 3-DNA Methylation Prognostic days were excluded. Meanwhile, all the DNA methylation sites Biomarker Model. *e univariate Cox proportional hazards containing 'NA' values were removed. Finally, a total of 257 samples regression analysis and Cox proportional hazards hypothesis with clinical information and 374831 DNA methylation sites were test were used to identify the DNA methylation sites that preserved for subsequent analysis. were signiﬁcantly associated with the OS of patients with STS in the training dataset. *e screen rules were followed by was identiﬁed, which was the optimal biomarker when setting the thresholds of the P value of the likelihood ratio compared with other combinations. *e risks of STS patients test (P< 0.01) and Cox proportional hazards hypothesis test were stratiﬁed with the following risk score formula, con- (P> 0.05) in Cox proportional hazards regression analysis. structed by the Cox regression coeﬃcients of the model: As a result, 15551 sites were retained as candidate markers. Multivariate Cox regression analysis was then performed on candidate markers. Finally, a biomarker model containing risk score � a β , (1) i j i,j three CpG sites (cg19804488, cg20542822, and cg07898500) j�1 was selected to analyze the prognosis of patients. According 4 Journal of Oncology Table 1: Clinical characteristics of soft tissue sarcoma patients in the TCGA database. Patients Total Training cohort Testing cohort Characteristics Groups (N � 257) (N � 129) (N � 128) No. % No. % No. % Male 117 45.53 59 45.74 58 45.31 Sex Female 140 54.47 70 54.26 70 54.69 Median 61 60 62 Range 20–89 20–89 24–87 Age at diagnosis <61 128 49.80 66 51.16 62 48.44 ≥61 129 50.20 63 48.84 66 51.56 Alive 158 61.48 77 59.69 81 63.28 Vital status Dead 99 38.52 52 40.31 47 36.72 Uterus 31 12.06 16 12.40 15 11.72 Connective soft tissue 114 44.36 51 39.54 63 49.22 Tissue of origin Retroperitoneum 99 38.52 54 41.86 45 35.16 Other 13 5.06 8 6.20 5 3.90 FMS 39 15.17 13 10.08 26 20.31 LPS 59 22.96 30 23.26 29 22.65 LMS 104 40.47 56 43.41 48 37.50 Histological type SS 10 3.89 6 4.65 4 3.13 GCS 3 1.17 2 1.55 1 0.78 MPNST 9 3.50 5 3.87 4 3.13 US 33 12.84 17 13.18 16 12.50 Pharmaceutical 137 53.31 70 54.26 67 52.34 Adjuvant treatment type Radiation 120 46.69 59 45.74 61 47.66 FMS: ﬁbromyxosarcoma; LPS: liposarcoma; LMS: leiomyosarcoma; SS: synovial sarcoma; GCS: giant cell sarcoma; MPNST: malignant peripheral nerve sheath tumor; and US: undiﬀerentiated sarcoma. to the Cox regression coeﬃcient of each site in the model, the 3.3. Prognostic Performance of the 3-DNA Methylation Bio- risk score formula was presented as follows: marker in the Training and Testing Datasets. A useful prognostic biomarker should be closely related to the OS of risk score � 7.537 × β − 3.301 × β cg19804488 cg20542822 patients. *e Kaplan–Meier survival analysis was performed (2) − 3.825 × β , to identify the relationship between the 3-DNA methylation cg07898500 biomarker and patient survival in both the training and where β represents the β value of cg19804488, and cg19804488 testing datasets. Based on the above risk score formula (1), the same represents the other two sites cg20542822 and the risk score of each patient was calculated. According to cg07898500. Obviously, high DNA methylation level of the median risk score, patients were divided into the low-risk cg19804488 can result in the increasing risk of STS patients, group and high-risk group, and the Kaplan–Meier curve was while high methylation levels of cg20542822 and cg07898500 drawn. *e risk score distribution and median of the two reduce the risk. *e genes corresponding to the three CpG datasets are presented in Supplementary Figure 1. *e Log- sites are GALK1 (galactokinase 1), COL6A5 (collagen type rank test and Breslow test revealed that the cumulative VI alpha 5 chain), and VCX3B (variable charge X-linked 3B). survival rates of the low-risk group in the training and Relevant Cox regression coeﬃcients, P values, gene symbols, testing datasets were much higher than those of the high-risk and chromosomal positions are shown in Table 2. Fur- group (Figure 3). At the same time, Kaplan–Meier survival thermore, DNA methylation levels of the three sites in analysis was performed for individual DNA methylation patients with short- and long-term survival were examined. sites on the testing dataset, and neither of which can dis- Figure 2 explicates that the methylation level of cg19804488 tinguish high- and low-risk patients (Supplementary Fig- in the group of patients with short-term survival was sig- ure 2). *is implied that the 3-DNA methylation biomarker niﬁcantly higher than that in the group of patients with long- exhibited better performance in distinguishing the risks of term survival (P � 1.03E − 03). In contrast, both STS patients than these sites alone as signatures. cg20542822 and cg07898500 had a higher methylation level To verify the prognostic performance of the 3-DNA in patients with long-term survival compared with patients methylation biomarker, ROC curve analysis was performed in with short-term survival; the P values were 9.91E−05 and the training and testing datasets. *e AUCs were 0.844 2.07E−03 separately. *e result was consistent with the Cox (P< 0.001, 95% CI: 0.740–0.948) and 0.710 (P � 0.002, 95% regression analysis above. CI: 0.595–0.823) (Figure 4), respectively. It indicated that the Journal of Oncology 5 Table 2: Information about the 3 DNA methylation sites. Probe ID Coeﬃcient P value Gene symbol Chromosome location cg19804488 7.537 8.20E−07 GALK1 Chr17: 75764282–75764283 cg20542822 −3.301 1.89E−08 COL6A5 Chr3: 130379491–130379492 cg07898500 −3.825 2.59E−04 VCX3B ChrX: 8464947–8464948 1.2 1.0 0.8 0.6 0.4 0.2 0.0 P = 1.03E – 03 P = 9.91E – 05 P = 2.07E – 03 –0.2 cg19804488 cg078998500 cg20542822 Short-term survival Long-term survival Figure 2: Violin plots of methylation level at a single CpG site for STS patients with short- and long-term survival in the training dataset. Here, short-term survival refers to a survival time that was less than or equal to 3 years, and long-term survival refers to a survival time for more than 3 years. Blue violin plots are for STS patients with short-term survival, and green violin plots are for patients with long-term survival. *e dotted line through the black dot represents the median, the range of the white box is from the lower quartile to the upper quartile, the thin white line represents the 95% conﬁdence interval, and the outer shape is the density. *e two on the left are violin plots of methylation distribution of cg19804488 for patients with short-term survival and long-term survival. Similarly, the two in the middle correspond to cg20542822, and the two on the right correspond to cg07898500. *e Mann–Whitney U test showed that there was a signiﬁcant diﬀerence in long-term and short-term patients with respect to the methylation levels of the three methylation sites. 3-DNA methylation biomarker performed well in predicting the prognostic eﬀectiveness of our biomarker. Also, the the OS of STS patient. *erefore, the 3-DNA methylation patients were categorized into two groups by the median age biomarker can be used as a prognostic factor for STS patients. of 61 years as the cut-oﬀ point (Supplementary Figure 3). *e group of patients under 61 years (n � 128, 49.8%) of age and the group of patients older than 61 (n � 129, 50.2%) 3.4. Identifying the Prognostic Performance of the Biomarker showed a signiﬁcant diﬀerence in the survival rate between by Regrouping the Dataset with Diﬀerent Clinical high-risk and low-risk patients (P< 0.001). *e AUCs were Characteristics. Clinical characteristics are important indi- 0.734 (P � 0.002, 95% CI: 0.599–0.870) and 0.815 (P< 0.001, cators for predicting the survival rate in patients with STS 95% CI: 0.721–0.908). In the male and female patient groups, , which include age, histological subtypes, and tumor the survival time of the high-risk patient group was also locations. *erefore, samples were regrouped based on signiﬁcantly lower than that of the low-risk patient group diﬀerent clinical characteristics for further validation of the (Supplementary Figure 4). *e AUCs were 0.785 (P< 0.001, prognostic robustness and independence of the 3-DNA 95% CI: 0.665–0.915) and 0.773 (P< 0.001, 95% CI: methylation biomarker. *e dataset was classiﬁed into a 0.676–0.870). For diﬀerent histological subtypes, taking the group of patients who received adjuvant pharmaceutical limited sample sizes into account, the samples were real- therapy and a group of patients who received adjuvant located into three groups: leiomyosarcoma (N � 104, radiation therapy. In the two groups, the median risk scores 40.47%), liposarcoma (N � 59, 22.96%), and other subtypes were −2.174 and −2.233, respectively. Kaplan–Meier analysis of sarcoma (N � 94, 36.57%). In the three groups, the AUCs showed that the 3-DNA methylation biomarker signiﬁcantly were 0.792 (P< 0.001, 95% CI: 0.665–0.920), 0.790 distinguished the high- and low-risk patients, with the P (P � 0.002, 95% CI: 0.626–0.953), and 0.806 (P< 0.001, 95% values of Log-rank and Breslow test being less than 0.001 CI: 0.692–0.919), respectively (Supplementary Figure 5). (Figure 5). Moreover, the AUCs were 0.769 (P< 0.001, 95% Furthermore, no matter where the tumor located, including CI: 0.659–0.879) and 0.801 (P< 0.001, 95% CI: 0.698–0.905). retroperitoneal, connective soft tissue, and other tissues, the It suggested that diﬀerent adjuvant therapies did not impede Kaplan–Meier and ROC analyses showed that in each group, DNA methylation 6 Journal of Oncology Training dataset Testing dataset 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 Log-rank P value = 1.87E – 02 Log-rank P value = 6.66E – 16 0.0 0.0 Breslow P value = 7.72E – 03 Breslow P value = 7.90E – 14 0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000 6000 Time (days) Time (days) Low risk (9/65) Low risk (19/64) High risk (43/64) High risk (28/64) +/+ Censored data +/+ Censored data (a) (b) Figure 3: Kaplan–Meier curves showing the ability of the 3-DNA methylation biomarker in distinguishing high and low risk of STS patients. Blue line represents the survival curve of low-risk patients, and green line represents the survival curve of high-risk patients. “+” represents the censored samples. *e data on the left side of the brackets represents the number of death samples of the current group, and the right side represents the total number of samples of the current group. (a) Kaplan–Meier curves for high- and low-risk patients in the training dataset (N � 129). *e low risk score in patients was signiﬁcantly correlated with the better prognosis (Log-rank P value � 6.66E−16, Breslow P value � 7.90E−14). (b) Kaplan–Meier curves for high- and low-risk patients in the testing dataset (N � 128). A signiﬁcant diﬀerence was shown in the cumulative survival between patients with high and low risks (Log-rank P value � 1.87E−02, Breslow P value � 7.72E−03). Training dataset Testing dataset 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 P < 0.001 P = 0.002 AUC = 0.844 AUC = 0.710 95% CI: 0.740–0.948 95% CI: 0.595–0.823 0.0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 1 – speciﬁcity 1 – speciﬁcity (a) (b) Figure 4: ROC curves showing the prognostic utility of the 3-DNA methylation biomarker. (a) *e ROC curve of the 3-DNA methylation biomarker in predicting the OS of STS patients in the training dataset. *e AUC is 0.844 (P< 0.001, 95% CI: 0.740–0.948). (b) *e ROC curve of the 3-DNA methylation biomarker prognostic predictor in the testing dataset. *e AUC is 0.710 (P � 0.002, 95% CI: 0.595–0.823). Sensitivity Proportional survival Sensitivity Proportional survival Journal of Oncology 7 Pharamaceutical therapy (N = 137) Pharamaceutical therapy (N = 137) 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 P < 0.001 Log-rank P value = 5.17E – 07 AUC = 0.769 0.0 Breslow P value = 4.24E – 07 95% CI: 0.659–0.879 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0 1000 2000 3000 4000 5000 1 – speciﬁcity Time (days) Low risk (15/69) High risk (39/68) +/+ Censored data (a) (b) Radiotherapy (N = 120) Radiotherapy (N = 120) 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 P < 0.001 Log-rank P value = 2.75E – 06 AUC = 0.801 0.0 Breslow P value = 2.20E – 06 95% CI: 0.698–0.905 0.0 0 1000 2000 3000 4000 5000 0.0 0.2 0.4 0.6 0.8 1.0 Time (days) 1 – speciﬁcity Low risk (14/60) High risk (31/60) +/+ Censored data (c) (d) Figure 5: ROC and Kaplan–Meier analyses in groups of patients receiving diﬀerent adjuvant therapies. (a) and (b) Kaplan–Meier and ROC curves in the group of patients receiving adjuvant pharmaceutical therapy. (c) and (d) Kaplan–Meier and ROC curves in the group of patients receiving adjuvant radiotherapy. It indicates that regardless of the adjuvant treatment therapies the patient received, our biomarker distinguished the high- and low-risk patients signiﬁcantly. the prognosis of low-risk patients was better than that of 3.5. Prognostic Superiority of the 3-DNA Methylation Bio- high-risk patients (Supplementary Figure 6). All the above markerasComparedwithOtherBiomarkers. Previous studies results are summarized in Table 3. have reported several prognostic biomarkers that are correlated Proportional survival Proportional survival Sensitivity Sensitivity 8 Journal of Oncology Table 3: Results of Kaplan–Meier and ROC analysis in diﬀerent groups based on diﬀerent clinical characteristics. Kaplan–Meier Kaplan–Meier AUC 95% CI Log rank P value Breslow P value Regrouping Sample Group factors size MST1- MST1- 3- MST1- MST1- 3-meth 3-meth 3-meth meth meth meth meth meth Female 140 4.02E−07 1.70E−02 2.45E−07 1.03E−02 0.773 0.657 0.676–0.870 0.535–0.779 Sex Male 117 1.69E−05 7.63E−03 1.22E−05 6.34E−03 0.785 0.636 0.665–0.915 0.490–0.782 Age at <61 128 1.10E−04 1.87E−04 5.26E−05 3.13E−04 0.734 0.714 0.599–0.870 0.572–0.856 diagnosis ≥61 129 2.11E−06 1.07E−01 2.54E−07 7.01E−02 0.815 0.599 0.721–0.908 0.478–0.720 Connective soft 114 2.19E−03 3.21E−01 8.44E−04 2.94E−01 0.800 0.483 0.677–0.923 0.299–0.666 Tissue of tissue origin Retroperitoneum 99 9.24E−05 1.69E−04 6.75E−05 3.10E−04 0.787 0.701 0.661–0.914 0.574–0.827 Other tissues 44 8.68E−04 1.44E−01 6.19E−04 1.32E−01 0.798 0.692 0.655–0.941 0.504–0.881 Leiomyosarcoma 104 3.66E−05 1.15E−02 3.49E−05 1.17E−02 0.792 0.657 0.665–0.920 0.502–0.812 Histological Liposarcoma 59 2.42E−04 1.45E−04 2.48E−04 1.92E−04 0.790 0.753 0.626–0.953 0.616–0.890 type Other subtypes 94 7.57E−05 4.36E−02 3.19E−05 3.73E−02 0.806 0.579 0.692–0.919 0.402–0.756 Treatment Pharmaceutical 137 5.17E−07 2.75E−04 4.24E−07 4.51E−04 0.769 0.673 0.659–0.879 0.545–0.800 type Radiation 120 2.75E−06 8.08E−02 2.20E−06 7.25E−02 0.801 0.6100 0.698–0.905 0.471–0.750 To be mentioned, 3-meth is the biomarker identiﬁed in this paper. *e Z-test was used to assess whether the AUC of the 3-DNA methylation biomarker is higher than the AUC of the MST1 methylation biomarker. “ ” represents the P value of Z-test being less than 0.001, which illustrated that the AUC of our 3- DNA methylation biomarker was signiﬁcantly higher than the MST1 methylation biomarker. with soft tissue sarcomas. For example, Jia et al. explained that Testing dataset 1.0 high expression of genes p16 and NM23-H1 in STS was closely associated with favorable prognosis of patients . PARP1 expression was considered as a prognostic factor in patients with STS . Low expression of genes EGFR and HIF-1 in 0.8 STS patients usually contributes to a poor prognosis . *e results of Pollino et al. illustrated that increased gene ex- pression of SDP35 promoted the progression of STS metastasis 0.6 and could be used as an independent marker of poor prognosis in patients . Hypermethylation of MST1 showed signiﬁcant correlation with favorable prognosis in patients with STS . 0.4 For prognostic biomarkers of STS that were established in the above studies, the prognostic eﬃciency was assessed and compared with the biomarker identiﬁed in this study. *e 0.2 AUCs of these biomarkers are shown in Figure 6 and more information is presented in Supplementary Table 1. It revealed that the prognostic performance of the 3-DNA methylation 0.0 biomarker was much better than other biomarkers. Notably, 0.0 0.2 0.4 0.6 0.8 1.0 the performance of the MST1 methylation biomarker in dis- 1 – speciﬁcity tinguishing the high- and low-risk STS patients was also assessed. *e results are summarized in Table 3. Kaplan–Meier Our 3-meth: 0.710 NM23-H1: 0.665 and ROC curves are shown in Supplementary Figures 7–11; P16: 0.629 EGFR + HIF1a: 0.572 they showed that the MST1 methylation biomarker repre- PARP1: 0.592 MST1meth: 0.603 sented poor performance in the group of patients aged over 61 SDP35: 0.623 years, the group of patients receiving radiotherapy, and the Figure 6: ROC curves of the 3-DNA methylation biomarker group of patients with primary lesions of connective soft tissues compared with other biomarkers that were previously reported. in the head, neck, chest, etc. Moreover, in all groups, the AUC *e red bold curve is the ROC curve of the signature established in values of the MST1 methylation biomarker were signiﬁcantly this research and its AUC is signiﬁcantly higher than other lower than the 3-DNA methylation biomarker. *erefore, in prognostic markers. terms of methylation biomarkers, our 3-DNA methylation biomarker has more advantages. TCGA database. *e high rate of metastasis and mortality remains a hang-up in most of the STS patients . *erefore, 4. Discussion it is of great signiﬁcance to ﬁnd prognostic markers for risk stratiﬁcation and guiding clinicians to scheme therapeutic *is study is the ﬁrst to systematically analyze the genome- regimens in STS patients. As a result, a 3-DNA methylation wide DNA methylation proﬁle of the STS patients in the Sensitivity Journal of Oncology 9 since one of the abilities of VCX-A, which has a high se- biomarker containing 3 CpG sites (cg19804488, cg20542822, and cg07898500) was identiﬁed through univariate and quence similarity with VCX3B, is to inhibit mRNA decapping. More gene functions and their association with multivariate Cox regression screening. Aberrant methylation of the 3 sites was closely bound up with the survival of STS STS should be evaluated by conducting further experiments. patients. Kaplan–Meier and ROC analyses showed that this We are also concerned about the phenomenon that biomarker displayed superior performance in predicting the patients with visceral sarcomas and tumors that originate survival of STS patients. To further conﬁrm the practicability, from the viscera often have similar symptoms. For example, the dataset was regrouped according to diﬀerent clinical some gastrointestinal symptoms such as abdominal pain, characteristics such as age, sex, the tissue of origin, etc. In weight loss, melasma, or anemia are usually shared between patients with gastrointestinal stromal tumors and common particular, no matter which adjuvant therapy was used, the Kaplan–Meier analysis showed that the 3-DNA methylation adenocarcinomas. Patients with uterine leiomyosarcoma and common uterine malignancies are both accompanied by biomarker discriminated the survival diﬀerence of patients with high and low risks. Besides, the AUCs in all these groups painless vaginal bleeding . *erefore, some visceral sarcomas are often managed as primary organ tumors rather exceeded 0.7. At the same time, compared with other prognostic biomarkers, the AUC of our biomarker was much than STS . It is excusable if this phenomenon does not higher. All the above results indicate that the biomarker is a interfere with the therapeutic eﬀects of STS patients. useful prognostic factor independent of clinical characteristics However, our research revealed that the therapeutic eﬃ- in STS patients. ciency of STS in diﬀerent organs can be predicted by using More samples and cross-platform datasets could broadly the same prognostic biomarker. *is suggested that visceral conﬁrm the results of this research which require continuous sarcomas have something in common that distinguishes them from their tumor counterparts in the primary organ. attention. It is expected that there will be more datasets that can be used to strengthen our conclusions in the future. At Accordingly, treating visceral sarcomas diﬀerently and carrying out targeted therapy might be an eﬀective new idea. the same time, more investigations are required on how the 3-DNA methylation biomarker is involved in the patho- genesis and development of STS. 5. Conclusion As mentioned above, in terms of STS, there are few Based on the analysis of the genome-wide DNA methylation studies on DNA methylation prognostic biomarkers and proﬁle of 257 STS patients in TCGA, a 3-DNA methylation none of them has been applied in clinical practice so far. biomarker that was signiﬁcantly associated with the survival However, with the continuous development of DNA of patients with STS was identiﬁed, which likewise embodied methylation detection technology, our research may provide the prognostic practicality for STS patients with diﬀerent a new direction for clinicians to make decisions regarding age, sex, histologic subtypes, primary organs, etc. Moreover, interventions and assess the eﬀectiveness of individual the 3-DNA methylation biomarker exhibited more advan- therapeutic strategies. Galactose kinase (GALK1) is the main tages in predicting the survival of STS patients when enzyme for galactose metabolism. Tang et al. have reported compared with other prognostic markers that were reported that GALK1 might be a new target for treating hepatocellular previously. *erefore, the biomarker identiﬁed in this study carcinoma (HCC) and also revealed that galactose metabolic can be used as a superior biomarker for the prognosis of pathway exhibited a new posttranscriptional regulation ef- patients with STS. fect on the protein expression of PI3K/AKT signaling pathway in HCC . However, the PI3K/AKT signaling Data Availability pathway has been proved to be a potential therapeutic target in STS [32, 33]. *erefore, whether GALK1 could regulate *is study used public data accessible in *e Cancer Ge- the PI3K signaling pathway through the galactose pathway nome Atlas (TCGA) database. in patients with STS, thereby promoting the pathogenesis and progression of tumors, should be further studied. Type 6 Conflicts of Interest collagen 5 chain (COL6A5) is also known as COL29A1, and its variation is closely associated with speciﬁc dermatitis *e authors declare that there are no conﬂicts of interest . Besides, COL6A5 acts as a functional ligand of LAIR-1, regarding the publication of this paper. which can directly inhibit the activation of immune cells. *ese cells will in turn lose the antitumor response and Acknowledgments promote tumor progression. . Immunotherapy is one of the major breakthroughs in oncology, but there is still much *is work was supported by the New Medicine Postgraduate controversy going on in STS. Some immune checkpoints Innovation Fund Program of Shanghai University and the including PD-1/PD-L1, CTLA-4, and other blocking ther- National Natural Science Foundation of China (Grant no. apies only have subtle eﬀects in STS . Perhaps distinct 11771277). immune targets should be investigated. *e possible regu- lating mechanism of COL6A5 on LAIR-1 may provide a new Supplementary Materials strategy. *e function of VCX3B, a member of the VCX/Y family, has not been clearly described yet. It can be deduced Supplementary 1: the 257 sarcoma samples used in this that this gene might increase the stability of some mRNAs, study; their corresponding histological subtypes all belong to 10 Journal of Oncology  M. Esteller, “Epigenetics in cancer,” New England Journal of soft tissue sarcoma. Supplementary 2: the 129 samples in the Medicine, vol. 358, no. 11, pp. 1148–1159, 2008. training dataset. Supplementary 3: the 128 samples in the  M. Kulis and M. Esteller, “DNA methylation and cancer,” testing dataset. Supplementary Table 1: the OS and DNA Epigenetics and Cancer, Part A, vol. 70, no. 2, pp. 27–56, 2010. methylation level or gene expression of the 3-DNA meth-  M. Widschwendter, A. Jones, I. Evans et al., “Epigenome- ylation biomarker and other known biomarkers listed above based cancer risk prediction: rationale, opportunities and in the testing dataset. Supplementary Figure 1: distribution challenges,” Nature Reviews Clinical Oncology, vol. 15, no. 5, histograms of the risk score based on the 3-DNA methyl- pp. 292–309, 2018. ation biomarker both in the training and in the testing  S. P. Liu and X. T. Hu, “Chapter 18-epigenetic biomarkers,” in datasets. Supplementary Figure 2: Kaplan–Meier and ROC Handbook of Epigenetics, T. Tollefsbol, Ed., pp. 277–292, analyses of the 3 sites in testing dataset. Supplementary Elsevier, Amsterdam, Netherlands, 2nd edition, 2017. Figure 3: Kaplan–Meier and ROC curves in two groups;  I. Peters, H. Eggers, F. Atschekzei et al., “GATA5 CpG island methylation in renal cell cancer: a potential biomarker for grouping is based on their median age 61 at initial diagnosis. metastasis and disease progression,” BJU International, Supplementary Figure 4: Kaplan–Meier and ROC curves for vol. 110, no. 2, pp. 144–152, 2012. STS patients in diﬀerent sex groups. Supplementary Fig-  K. Hur, P. Cejas, J. Feliu et al., “Hypomethylation of long ure 5: Kaplan–Meier and ROC analyses for STS patients interspersed nuclear element-1 (LINE-1) leads to activation of whose tumors belong to diﬀerent histologic subtypes. proto-oncogenes in human colorectal cancer metastasis,” Gut, Supplementary Figure 6: Kaplan–Meier and ROC analyses vol. 63, no. 4, pp. 635–646, 2014. for STS patients with tumor originating from diﬀerent tis-  C. Seidel, F. Bartel, M. Rastetter et al., “Alterations of cancer- sues. Supplementary Figures 7–11: Kaplan–Meier and ROC related genes in soft tissue sarcomas: hypermethylation of analyses of the MST1 methylation biomarker for STS pa- RASSF1A is frequently detected in leiomyosarcoma and as- tients with diﬀerent clinical characteristics. (Supplementary sociated with poor prognosis in sarcoma,” International Materials) Journal of Cancer, vol. 114, no. 3, pp. 442–447, 2005.  T. Knosel, A. Altendorf-Hofmann, L. Lindner et al., “Loss of p16 (INK4a) is associated with reduced patient survival in soft References tissue tumours, and indicates a senescence barrier,” Journal of Clinical Pathology, vol. 67, no. 7, pp. 592–598, 2014.  S. A. G. D. Demetri and R. S. Benjamin, “Soft tissue sarcoma,”  F. E. Harrell, Regression Modeling Strategies, Springer, New Journal of =e National Comprehensive Cancer Network, York, NY, USA, 2017. vol. 3, no. 2, pp. 158–194, 2005.  N. Pandis, “Comparison of 2 means (independent z test or  L. A. Doyle, “Sarcoma classiﬁcation: an update based on the independent t test),” American Journal of Orthodontics and 2013 world health organization classiﬁcation of tumors of soft Dentofacial Orthopedics, vol. 148, no. 2, pp. 350-351, 2015. tissue and bone,” Cancer, vol. 120, no. 12, pp. 1763–1774, 2014.  A. Hazra and N. Gogtay, “Biostatistics series module 9:  R. L. Siegel, K. D. Miller, and A. Jemal, “Cancer statistics, survival analysis,” Indian Journal of Dermatology, vol. 62, 2019,” CA: A Cancer Journal for Clinicians, vol. 69, no. 1, no. 3, pp. 251–257, 2017. pp. 7–34, 2019.  D. Schottenfeld and J. F. Fraumeni, Cancer Epidemiology and  E. G. Demicco, R. G. Maki, D. C. Lev et al., “New therapeutic Prevention, Oxford University Press, New York, NY, USA, targets in soft tissue sarcoma,” Advances in Anatomic Pa- thology, vol. 19, no. 3, pp. 170–180, 2012.  J. Jia, P. Yin, X. Zhang et al., “Correlation of p16 and nm23-H1  M. F. Brennan, C. R. Antonescu, K. M. Alektiar, and expression levels with incidence and prognosis of soft tissue R. G. Maki, Management of Soft Tissue Sarcoma, Springer, sarcoma,” Oncology Letters, vol. 17, no. No. 6, pp. 4865–4870, Basel, Switzerland, 2nd edition, 2013.  R. A. De Mello, A. Tavares, and G. Mountzios, International  F. Bertucci, P. Finetti, A. Monneur et al., “PARP1 expression Manual of Oncology Practice, Springer, Basel, Switzerland, in soft tissue sarcomas is a poor-prognosis factor and a new 2nd edition, 2015. potential therapeutic target,” Molecular Oncology, vol. 13,  J. N. Cormier and R. E. Pollock, “Soft tissue sarcomas,” CA: A no. 7, pp. 1577–1588, 2019. Cancer Journal for Clinicians, vol. 54, no. 2, pp. 94–109, 2004.  S. Rot, H. Taubert, M. Bache et al., “Low HIF-1alpha and low  M. A. Clark, C. Fisher, I. Judson, and J. M. *omas, “Soft- EGFR mRNA expression signiﬁcantly associate with poor tissue sarcomas in adults,” New England Journal of Medicine, survival in soft tissue sarcoma patients; the proteins react vol. 353, no. 7, pp. 701–711, 2005. diﬀerently,” International Journal of Molecular Sciences,  R. Grimer, I. Judson, D. Peake, and B. Seddon, “Guidelines for vol. 19, no. 12, 2018. the management of soft tissue sarcomas,” Sarcoma, vol. 2010,  S. Pollino, M. S. Benassi, L. Pazzaglia et al., “Prognostic role of Article ID 506182, 15 pages, 2010. XTP1/DEPDC1B and SDP35/DEPDC1A in high grade soft-  G. Lee, S. Y. Lee, S. Seo et al., “Prognostic factors and clinical tissue sarcomas,” Histology and Histopathology, vol. 33, no. 33, outcomes of urological soft tissue sarcomas,” Korean Journal pp. 597–608, 2018. of Urology, vol. 52, no. 10, pp. 669–673, 2011.  C. Seidel, U. Schagdarsurengin, K. Blumke et al., “Frequent  J. Liu, R. Li, X. Liao, and W. Jiang, “Comprehensive bio- hypermethylation of MST1 and MST2 in soft tissue sarcoma,” informatic analysis genes associated to the prognosis of lip- Molecular Carcinogenesis, vol. 46, no. 10, pp. 865–871, 2007. osarcoma,” Medical Science Monitor, vol. 24, Article ID  B. Kasper, “Standards and novel therapeutic options in the 913043, 2018. treatment of patients with soft tissue sarcoma,” Reviews on  N. Aggerholm-Pedersen, K. Maretty-Kongstad, J. Keller, and Recent Clinical Trials, vol. 2, no. 3, pp. 206–211, 2007. A. Safwat, “Serum biomarkers as prognostic factors for  M. Tang, E. Etokidem, and K. Lai, “*e leloir pathway of metastatic sarcoma,” Clinical Oncology, vol. 31, no. 4, galactose metabolism-A novel therapeutic target for pp. 242–249, 2019. Journal of Oncology 11 hepatocellular carcinoma,” Anticancer Research, vol. 36, no. 12, pp. 6265–6271, 2016.  J. Quesada and R. Amato, “*e molecular biology of soft- tissue sarcomas and current trends in therapy,” Sarcoma, vol. 2012, Article ID 849456, 16 pages, 2012.  H. J. Lim, X. Wang, P. Crowe, D. Goldstein, and J. L. Yang, “Targeting the PI3K/PTEN/AKT/mTOR pathway in treat- ment of sarcoma cell lines,” Anticancer Research, vol. 36, no. 11, pp. 5765–5771, 2016.  C. Soderhall, I. Marenholz, T. Kerscher et al., “Variants in a novel epidermal collagen gene (COL29A1) are associated with atopic dermatitis,” PLoS Biology, vol. 5, no. 9, Article ID e242,  R. J. Lebbink, T. de Ruiter, J. Adelmeijer et al., “Collagens are functional, high aﬃnity ligands for the inhibitory immune receptor LAIR-1,” Journal of Experimental Medicine, vol. 203, no. 6, pp. 1419–1425, 2006.  W. Liu, Q. Jiang, and Y. Zhou, “Advances of systemic treatment for adult soft-tissue sarcoma,” Chinese Clinical Oncology, vol. 7, no. 4, p. 42, 2018.  M. P. W. T. Pisters, M. Weiss, R. Maki, and C. P. Raut, Soft- Tissue Sarcomas, Cancer Network, Cranbury Township, NJ, USA, 2016.  J. Y. Hui, “Epidemiology and etiology of sarcomas,” Surgical Clinics of North America, vol. 96, no. 5, pp. 901–914, 2016.
Journal of Oncology – Hindawi Publishing Corporation
Published: May 15, 2020
Access the full text.
Sign up today, get DeepDyve free for 14 days.