DisV-HPV16, versatile and powerful software to detect HPV in RNA sequencing data

The number of cancer patients has reportedly increased in recent years. It is estimated that there were 14.9 million incident cancer cases and 8.2 million cancer deaths worldwide in 2013 [1]. Approximately 10–15% of human cancers are known to be caused by viruses [2]. Human papillomavirus (HPV) is a sexually transmitted virus causing various benign and malignant diseases including condyloma acuminatum [3], cervical carcinoma(CC) and head and neck squamous carcinoma (HNSC) [4]. Since the first detection by Gissmann in 1982 [5], the presence of human papillomavirus (HPV) in tumor tissue samples from the head and neck, especially oropharyngeal carcinoma, has been increasing worldwide [6]. The HPV family consists of at least 170 different virus types that preferentially infect the mucosa of the genitals [7]. The high-risk HPV type, a sub-group of mucosal HPVs, causes approximately 5% of all human cancers, corresponding to one-third of all virus-induced tumors [8]. Within the high-risk HPV group, HPV16 is the most oncogenic type found in HNSC patients [9].

The detection of viruses in human cancer tissues has significant clinical implications in oncology. Widespread clinical application of next generation sequencing (NGS) and rapid advances in NGS technology in recent years have enhanced the capabilities for virus detection in human cells and enabled large-scale investigations of virus-host interactions [10‐12]. Several software tools have been developed including VirusSeq [13], VirusFinder [14] and ViromeScan [15] that apply a computational subtraction algorithm to distinguish viruses within NGS data. VirusSeq and VirusFinder, however, are disadvantaged by extensive time and multiple alignment tools, respectively [16]. Although ViromeScan is less time consuming than either VirusSeq or VirusFinder, it can only be used to determine the taxonomic composition of virome by aligning sequence reads to completely determined viral genomes [15]. Furthermore, nearly all of the computational tools focus on detecting the presence and integration of virus in sequencing data [17], which are considered the key factors in carcinogenesis [18]. From a disease etiology perspective, however, the more important factor may be the expression of viral oncogenes.

HPV16 oncogenes including E5, E6 and E7 are known to contribute to carcinogenesis. E5 negatively regulates the TGF-β signaling pathway [19]. E6 degrades p53 after binding to both p53 and E6-associated protein ligase [20]. E7 binds to pRb and triggers expression of proteins necessary for DNA replication by activating the E2F transcription factor [21]. Neither E6 nor E7 possess intrinsic enzymatic activity, but functions through direct and indirect interactions with host cellular proteins including several well-known tumor suppressors [22].

In light of the increasing incidence of HPV-associated HNSC, the severity of this disease and the roles of HPV E6 and E7 in carcinogenesis, we undertook the development of a new software tool to enable the detection and analysis of HPV16 oncogenes. The resulting DisV-HPV16 software detects HPV16 in RNA sequencing data and determines the expression levels of HPV16 oncogenes. DisV-HPV16 is faster, more sensitive and more accurate than other software tools. Moreover, its reliability was experimentally validated using RT-PCR.

We compared DisV-HPV16 with VirusSeq and ViromeScan, two previously available tools for the detection of viruses using data from RNA sequencing. VirusSeq required more time (2 h) than ViromeScan or DisV-HPV16 (1 h) to detect virus in RNA sequencing data from cell lines. In analyzing RNA sequencing data from the human transcriptome, the time differential was considerably greater, with VirusSeq requiring more than one day while DisV-HPV16 requiring the least time (1 h). These differences are summarized in Table 1. We ran the three software tools using the same configuration of CPU. DisV-HPV16 was overall faster in detecting virus in either cell line or human transcriptome RNA sequencing data, which we suggest is attributable to the pipeline design of DisV-HPV16 (Fig. 2).

DisV-HPV16 detected HPV16 in SIHA cell line RNA sequencing data. Changing the DisV-HPV16 reference file allowed HPV18 detection in HELA cell line data. Furthermore, DisV-HPV16 evaluated the ratio of E2/E6 in HELA (0.007 or 0.002) and SIHA (0.24) cell line data (Table 2). Previous studies have shown that an E2/E6 ratio between 0 and 1 indicates a combined episomal plus integrated HPV16 status, while an E2/E6 ratio of 0 indicates an integrated viral status [29‐31].

RNA sequencing data from 18 HNSC patients were used to confirm the accuracy and sensitivity of DisV-HPV16. HPV was detected in each of 14 HPV16-positive patient samples by DisV-HPV16, as well as two samples, SRR2932838 and SRR2932841, from HPV35- and HPV33-positive patients, respectively. After changing the DisV-HPV16 reference file, we detected HPV35 in SRR2932838 and HPV33 in SRR2932841 (Table 3). DisV-HPV16 exhibited greater sensitivity than either VirusSeq or ViromeScan in detecting HPV16. We suggest that the enhanced sensitivity may be due to the new reference file we created (Fig. 1). Other genotypes of HPV were detected after changing reference files. Detection of HPV16 in HPV35-positive SRR2932838 and HPV33-positive SRR2932841 indicates the possible occurrence of co-infection in HPV-related cancers, and illustrates the sensitivity and versatility of DisV-HPV16. Previous studies have reported human co-infection by different viruses (e.g. HIV plus HPV) [32] as well as different HPV genotypes (the most frequent were HPV6, 16, 42 and 51) [33]. HIV infection is known to have a significant impact on HPV genital infection [32]. No preferential distribution of specific HPV type(s) with co-infection was identified [33]. In the future, such information may be highly useful in designing vaccination campaigns.

DisV-HPV16 accuracy was experimentally verified by RT-PCR and RNA sequencing of two HNSC cell lines, SIHA and SCC47. For SIHA cells, similar E7/E6 ratio values of 2 (average CT value: E6 = 21.27, E7 = 20.27) and 2.68 (E7 = 1,049,631.5, E6 = 391,666.94) were obtained using RT-PCR and DisV-HPV16, respectively. In SCC47 cells, RT-PCR and DisV-HPV16 analyses yielded E7/E6 values of 1.74 (average CT value: E6 = 21.73, E7 = 20.93) and 2.26 (E7 = 1,378,347.625, E6 = 610,878.63), respectively (Fig. 3). These results confirm the reliability and effectiveness of DisV-HPV16 in detecting and evaluating HPV16 oncogenes.

DisV-HPV16 was used to evaluate HPV16 oncogenes expression in RNA sequencing data from 14 HPV16-positive patients. In a given patient sample the expression levels of different oncogenes were found to vary, as did expression of a given oncogenes among different samples. These observations are summarized in Table 4. Expression levels of HPV16 oncogenes could be clearly depicted in a heatmap based on the FPKM value for each oncogenes (Fig. 4). The 14 samples segment into two groups (four on the left and ten on the right in Fig. 4). Three early genes E6, E7 and E6* are in one cluster while E2, E5 and E1^E4 are in the other, indicating opposite oncogenes expression trends in the two clusters. This result might reflect E2-induced increase in viral replication via splicing-related activities [34], which results in high-level expression of E6 and E7. This may result in clinically significant differences among individual patients suffering from HPV16-positive head and neck cancers.

The new DisV-HPV16 software not only detected the existence of distinct HPV genotypes (HPV16, HPV18, HPV33 and HPV35) in RNA sequencing data (simply by changing reference files) but also revealed the expression levels of HPV16 oncogenes. DisV-HPV16 provides enhanced virus detection and analysis capabilities based on RNA sequencing data and also enlarges the potential for understanding the effects of viral genes on the host genome and elucidating key features of the virus-host relationship. In the present study we tested DisV-HPV16 on four HPV genotypes. Whether the software can be of value for detection and evaluation of additional viruses will be determined in future studies.