Introduction
Osteosarcoma is a rare cancer which affects 0.2–0.35/100 000 across countries with a slightly higher frequency in male than female (Mirabello et al., 2009a; Valery et al., 2015). It is the most common malignant bone tumor (Mirabello et al., 2009b). The highest peaked incidence was observed in women at 10–14 years and in men at age 15–19 years worldwide. There is a lower second peak in elderly (Mirabello et al., 2009a; Hung et al., 2014). The combination of surgery and chemotherapy remains standard of care. Introduction of effective chemotherapy during 1960s−1980s increased the cure rate from < 10% to 60–75% but no more gain later on (Jaffe et al., 2013). Many studies were carried out for understanding more about the mechanism of osteosarcoma and to look for more new effective treatment.
The initial sequencing of the human genome in 2001 showed that repetitive DNA sequences accounted for at least 50% of the genome (Lander et al., 2001). But more recent computational approaches stated a higher proportion as two-thirds of repetitive elements in the human genome (de Koning et al., 2011). They were classified into five classes: transposon-derived repeats, simple sequence repeats; segmental duplications; blocks of tandemly repeated sequences and ribosomal gene clusters (Lander et al., 2001). They can also be classified into two broad classes: tandem repeats and transposable elements (Padeken et al., 2015). Tandem repeats included satellite DNA, minisatellite and microsatellite (Padeken et al., 2015). While transposable elements were classified into retrotransposon (class I) and DNA transposons (class II). About 45% of genome belongs to transposon-derived repeats. Class I retrotransposons includes long terminal repeat (LTR)/human endogenous retroviruses (HERVs) and non-LTR retrotransposons such as LINEs (Long Interspersed Nuclear Elements) and SINEs (Short Interspersed Nuclear Elements) (Lander et al., 2001; Rebollo et al., 2012). Human endogenous retroviruses (HERVs) are a family of viruses integrated in our genome which have similarities with the nowadays exogenous retroviruses (Nelson et al., 2003). HERVs account for about 8% of human DNA (Lander et al., 2001; Cegolon et al., 2013). They are typically composed of gag, pol and env regions sandwiched between the two LTRs (long terminal repeats) (Nelson et al., 2003; Bannert and Kurth, 2004; Mager and Stoye, 2015).
Repetitive elements are found to be associated to some epithelial cancers. Overexpression of CT (centromeric) and PCT (pericentric) sequences was found in cancer tissues compared to normal tissues of the same patient who got testicular, liver, ovarian, and lung cancers (Eymery et al., 2009). Digital gene expression analysis of 15 pancreatic ductal carcinomas showed a median 21-fold increased expression of total amount of all satellite transcripts compared with normal pancreas. And the pericentromeric human satellite II (HSATII) was the biggest differentially expressed satellite subfamily which was not able to be detected in normal human pancreas and was expressed minimally in other normal tissues. Overexpression of HSATII was also observed in other human cancers including lung (2 of 2), kidney (2 of 2), ovarian (2 of 2), and prostate (3 of 3) (Ting et al., 2011; Bersani et al., 2015). The presence of a level of GSATII, TAR1, and/or SST1 satellite transcripts below the reference level indicates that subject has a tumor (Schiavetti et al., 2002; Ting et al., 2012; Bersani et al., 2015). Alpha human satellite DNA was overexpressed (43 times) in pancreatic cancer compared to normal pancreas (Ting et al., 2011). Transcripts from HERV-K HML-2 have been found to be associated with many cancers such as melanoma (Schiavetti et al., 2002) leukemia and lymphoma (Contreras-Galindo et al., 2008) as well as tumors of the breast (Pichon et al., 2006; Wang-Johanning et al., 2007) testis (Pichon et al., 2006) and ovary (Wang-Johanning et al., 2007). The HERV-E family has been found to be correlated with prostate, kidney, ovarian and uterine cancers (Wang-Johanning et al., 2003; Gimenez et al., 2010). HERV-H sequences was found to be overexpressed in colorectal carcinogenesis (Pérot et al., 2015).
For non-LTR retrotransposons, LINE-1 was found to be overexpressed in tumor samples of pancreatic and prostate cancers (Contreras-Galindo et al., 2008; Criscione et al., 2014). De novo L1 insertions were found in colorectal cancer (Solyom et al., 2012). L1-mediated retrotransposition was suggested a potentially crucial source of mutations that can decrease the tumor suppression of somatic cells in hepatocellular carcinoma (Shukla et al., 2013). Few SINE subfamily were expressed differently in prostate cancer (Criscione et al., 2014).
Interesting findings about repetitive elements especially on epithelial cancers led to approval of patent for using some kinds of REs as biomarker in detection, prognosis and follow-up of several carcinomas. It raised the question about REs expression in sarcoma, a different kinds of malignant diseases. These repeated DNA sequences are not yet analyzed in case of osteosarcoma and it would be necessary to analyse their impact on the disease. Hence, we conducted this study with aim to reveal the differential expression of repeated DNA elements in osteosarcoma.
Discussion
To our present knowledge, no similar study was carried out in osteosarcoma. We found herein 82 repetitive elements which were expressed differentially with significance FDR < 0.05 between osteosarcoma and adjacent normal bone in paired samples. Of which, 35 were upregulated and 47 were downregulated with FDR < 0.05.
The top most significant differentially expressed repetitive elements (with lowest FDR) were THE1C-int, LTR5, MER57F, and MER87B. Interestingly, they all belong to LTR elements or HERVs with DNA transposons made up of transposable elements (TEs). THE1C-int is a ERV3, a retrovirus-like MaLR element, consisting of 375 bp. LTR5 is ERV2, clone of HERV-K18 consisting of 969 bp. MER57F is a ERV1 with 435 bp and MER87B is also a ERV1 of 509 bp (Jurka et al., 2005; Bao et al., 2015). To our knowledge, THE1C-int, MER57F, and MER87B were not yet mentioned to be correlated with diseases in previous literature. While LTR5 hypomethylation was found to be correlated to systemic lupus erythematosus mechanism (Nakkuntod et al., 2013). Many studies and observations suggested an important role of transposable elements in genomic instability, transcriptional control, non-coding RNA regulation, oncogenic activation and chromosomal rearrangements, Anwar et al. (2017). HERVs was mentioned a lot recently as a TE that may have a role in carcinogenesis in many different cancers that we got some evidences based on present literature.
In prostate cancer, HERV-E (and/or ERV3) env genes was expressed only in prostate tumor cells that suggested as targets for immunotherapy for the disease (Wang-Johanning et al., 2003). HERV-K was also identified to be involved to prostate cancer. It was found that forward and reverse transcripts of several HERV-K loci were detected in prostate cancer cell lines (Agoni et al., 2013) and overexpression of HERV-K (22q11.23) was associated with hypomethylation of the HERV-K locus in the disease (Goering et al., 2011). Interestingly, in a study of 14 patients with different grades of prostate cancer in paired samples (prostate tumor and normal tissue), 475 retrotransposon subfamilies were detected to be significantly differential expression in tumor tissue with FDR < 0.05. Among thoses elements, LTR was the most prevalent and most of them were endogenous retroviruses with ERV1 being the most represented (Ren et al., 2012; Criscione et al., 2014).
In ovarian cancer, HERV-K env protein was expressed in epithelial ovarian cancer with high frequency (90%) without expression in normal and benign ovarian surface epithelial tissue. Interestingly, ERV3 and HERV-E were found to expressed simultaneously in the same ovarian cancer tissues and antibodies to HERVs were present in the sera of ovarian cancer. Authors suggested that HERVs should be further studied to provide a new ovarian cancer screening tool and potentially serve as a new target for detection, diagnosis and treatment for this cancer (Wang-Johanning et al., 2007). Methylation levels of HERV-K and HERV-E were decreased in advanced ovarian cancer. HERV-K hypomethylation was found to be correlated to ovarian cancer aggressiveness and poor response to treatment (Iramaneerat et al., 2011).
In breast cancer, HERV-K env was suggested to be a tumor marker as it was expressed only in breast cancer tissues and cell lines analyzed but not detected in normal breast tissues (Wang-Johanning et al., 2001). HERV-K Env proteins were expressed in more than 85% cases of breast cancer patients which induced both serologic and cell-mediated immune responses (Wang-Johanning et al., 2008). High titers of of env RNA of HERV-K (HML-2) was found in the plasma of patients with breast cancer (Contreras-Galindo et al., 2008).
Transposable elements were also studied in hepato-gastrointestinal cancers. The HERV-H sequence was found to be overexpressed in up to 30–50% of different gastrointestinal cancers such as colorectal, gastric and pancreas cancers (Wentzensen et al., 2007). Additionally, the gag gene of a HERV-H locus on the chromosome Xp22 was commonly expressed in colon malignant tissues (Alves et al., 2008). LTRs were found to be upregulated and LTR-007 was noted to be the most frequently expressed LTR in hepatocellular carcinoma (Hashimoto et al., 2015).
Additionally, RNA from human endogenous retrovirus K (HERV-K) (HML-2) can be found at a very high titers in the plasma of lymphomas patients and it decreased dramatically with succesful treatment of lymphoma (Contreras-Galindo et al., 2008). A very interesting study on tumorigenesis of Hodgkin's lymphoma provided evidence that endogenous LTR activation can be oncogenic and hypomethylation of the THE1B LTR (a MaLR family LTR retrotransposon, the same family THE1 with THE1C-int found in our study) caused the CSF1R (colony stimulating factor 1 receptor gene) oncogene activation. It was emphasized that CSF1R was strongly expressed in these malignant cells (Lamprecht et al., 2010).
HERVs were also studied in melanoma. HERV-K mRNA and proteins were found in melanoma tissues and cell lines (Muster et al., 2003; Büscher et al., 2005) and HERV-K (HML-2) loci being transcribed in melanoma tissues was identified (Schmitt et al., 2013).
HERVs was found in some germinal cancers. Production of HERV-K virus-like particle was identified in teratocarcinoma cell lines (Boller et al., 1993; Löwer et al., 1993); HERV-K Gag and Env proteins were expressed in germ cell tumors and interestingly antibodies against these proteins were also detected (Sauter et al., 1995, 1996). Hypomethylation of several HERV-W loci seems to be important in the HERV-W activation in testicular cancer (Gimenez et al., 2010).
Though many studies revealed the role of HERVs in oncogenesis, further studies need to be carried out for further understanding the direct role of these elements in cancer innitialization. Our findings here maybe the first in osteosarcoma and they match with the present understanding about transposable elements in cancer generally.
Ranking by logFC, we can generate some of the most upregulated repetitive elements with highest logFC (Table 3). Among those, SAR, HSATII and -CATTC-n were the top 3 most upregulated ones.
SAR is a human satellite I DNA consisting of 84 BP with sequence: acagtatata atatatattt tgggtacttt gatattttat gtacagtata taatatatat tttgggtact ttgatatttt atgt (Jurka et al., 2005; Bao et al., 2015). It has not been mentioned in previous studies on cancer. This finding may suggest for further looking for its role in osteosarcoma and we can discuss the function of SAR in the light of another satellite element HSTAII, that is pericentromeric human satellite II element. HSTAII was the second most upregulated element in osteosarcoma after SAR. HSTAII was found to be upregulated in some kinds of epithelial carcinomas such as pancreatic adenocarcinoma, lung cancer, renal cancer, ovarian, prostate cancer, colon cancer (Ting et al., 2011; Bersani et al., 2015). In another study analyzing 15 human pancreatic adenocarcinoma samples, 21-fold increased expression of all satellites transcripts compared with normal pancreas has been described. The highest differential expression was found for pericentromeric satellite HSATII. It was increased 131-fold in pancreatic cancer and undetectable in normal pancreas or very low expression in other normal tissues (Ting et al., 2011). Overexpression of HSATII was also found in other cancers such as lung cancer, kidney, ovarian and prostate. More interestingly, using RNA-ISH analysis of endoscopic ultrasound-guided fine-needle aspirates (EUS-FNA) of pancreatic masses revealed that HSATII-positive cells were detected in 10/10 cases confirmed adenocarcinoma in which 2 cases were non diagnostic with FNA (Ting et al., 2011). This is very promising for further study in the future on osteosarcoma with the present findings in epithelial carcinomas. HSATII copy gain in colon cancer was associated with a significant reduction in overall survival compared with no gain tumors (Bersani et al., 2015). Interesting experiment with colon cancer cells line treated by inhibiting HSATII rdDNA formation in 3D conditions and mouse tumor xenografts lead to tumor shrinking with reduction of HSATII copy number gain (Bersani et al., 2015).
Simple satellite repeat (CATTC)n was found to be overexpressed in human colon cancer simultaneously with HSATII and ALR/alpha. This finding completely supports our results where we found similar simultaneous upregulation of the REs (Bersani et al., 2015).
Taken together, our findings of changed expression of repetitive elements in osteosarcoma match with previous results in epithelial, germinal, and hematologic cancers. It suggests that REs could have potential biomarkers in detection, diagnosis and follow-up osteosarcoma. It needs further studies to verify this preliminary results.
We admitted that we have some limitations to be improved. Our samples were heterogenous but it was acceptable to make comparision between malignant group and control group due to paired samples. Moreover, in the Figure 2, we can see that patients without chemotherapy seem not to have higher similarity in their subgroup. But it is difficult to make a conclusion because only 3 patients did not receive chemotherapy in 18 patients totally. Anyway, it raised a question to be verified if the repetitive elements expression is changed with chemotherapy.