Sequencing-Based Genotyping of Mixed Human Papillomavirus Infections by Use of RipSeq Software

Abstract

Sequencing-based pathogen identification directly from clinical specimens requires time-consuming interpretation, especially with mixed chromatograms when multiple microorganisms are detected. We assessed RipSeq Mixed software for human papillomavirus (HPV) genotyping by comparison to the linear array HPV genotyping assay. RipSeq Mixed provided rapid, sequencing-based HPV typing for single-type infections and coinfections with 2 types.

TEXT

Gene sequencing is often used as a reference method for pathogen identification. Broad-range PCR followed by Sanger sequencing is widely used for bacterial identification, and group-specific primers are also used for genotyping of viral pathogens (1–6). However, interpretation is often time-consuming, and the presence of more than one targeted microorganism results in mixed chromatograms that cannot easily be interpreted. The RipSeq Mixed Web application uses a software algorithm for rapid interpretation of sequences from clinical samples containing one or more targeted microorganisms. This includes base-calling of ambiguous positions in mixed chromatograms followed by a probability search against a customizable reference database (7). The software has been used for direct 16S rRNA gene (16S) sequencing of clinical samples from polybacterial infections (8–10). In contrast, interpretation software for mixed chromatograms as a result of viral coinfections has not been described.

Numerous human papillomavirus (HPV) genotyping assays have been described, each with its own advantages and disadvantages that often include labor- or resource-intensive workflow and/or limited resolution (11–13). Most HPV genotyping tests target the L1 open reading frame (ORF) (11, 14, 15). L1 is the basis for phylogenetic classification (16), and broad-range L1 PCR followed by DNA sequencing is a reference method for HPV genotyping. However, approximately 50% of cervical cytology samples contain multiple HPV types, thus limiting the utility of sequencing-based typing and variant detection (17–19). While bacterial 16S sequences are highly conserved between related species (20, 21), HPV genotypes can differ by 10 to 20%, making their identification in mixed infections more difficult (22). Sequences of the same genotype are also significantly more variable. Highly divergent viral targets have also made the development of broad-range primers more challenging, and HPV consensus primers have shown bias for amplification of different HPV types (23). In this study, we tested commonly used HPV consensus primers GP5+/6+ (1), MY09/11 (4), and PGMY09/11 (3) for sequencing-based HPV genotyping and compared the RipSeq Mixed software for identification of one or more HPV types in cervical cytology samples to the linear array genotyping assay (LA; Roche). Of the 3 primer sets, single-round amplification with PGMY09/11 generated amplification products in the largest number of test samples (see Methods in the supplemental material; data not shown) (13). Second-round amplification of PGMY09/11 PCR products with GP5+/6+ or PGMY09/11 primers did not further increase detection rates. Therefore, single-round PCR with PGMY09/11 primers was used for subsequent experiments. The PGMY09/11 set consists of 5 forward and 13 reverse primers and has been shown to improve the number of mixed infections detected over other primer systems (3). PGMY09/11 broad-range primers are also used by LA (11).

We next compared detection of less abundant types in contrived samples containing 2 or 3 HPV genotypes. DNA extracted from CaSki (HPV16) and HeLa (HPV18) cells was mixed at ratios ranging from 1:100 to 100:1 to simulate coinfections at different ratios (see Table S1 in the supplemental material). HPV copies were quantified by type-specific real-time PCR (13). When present in equal concentrations, both HPV genotypes were consistently detected by the RipSeq Mixed algorithm. However, primers PGMY09/11 resulted in a preferential detection of HPV18 when HPV16 was less abundant. To simulate mixed infections with 3 HPV genotypes, the L1 ORF from HPV52, another high-risk (HR) HPV type, was cloned and plasmid pHPV52 was combined with DNA from CaSki and HeLa cells. HPV18 and HPV52 were preferentially detected when present in combination with HPV16 (see Table S1). Thus, despite a combination of 18 individual primers in the PGMY09/11 mix, considerable bias remains (23). Primer bias poses an inherent limitation for all HPV genotyping assays requiring PCR amplification, including hybridization-based methods (23).

Unselected HR HPV-positive liquid-based cervical cytology samples (n = 105) were then tested by sequencing and LA. Reference sequences from the Los Alamos HPV sequence database and GenBank used for RipSeq analysis are shown in Table S2 in the supplemental material. Overall, 34 different genotypes were detected by LA and 31 genotypes were identified by sequencing (see Table S3 in the supplemental material). LA detected a single HPV genotype in 37 (35.2%), 2 and 3 genotypes in 31 and 16 (29.5% and 15.2%), and ≥4 genotypes in 21 (20%) samples. RipSeq Mixed is designed to resolve sequencing chromatograms from a maximum of 3 different sequences (7) and demonstrated 59 (56.2%) samples with one, 36 samples (34.3%) with two, and 9 samples (8.6%) with three genotypes and one uninterpretable sequence.

The same single genotype identified by RipSeq was also detected by LA in 57 (96.6%) samples. In 2 samples (3.4%), the single genotype detected by sequencing could not be identified by LA. However, BLAST analysis revealed 99% and 98% matches with reference sequences for HPV59 and HPV83 confirming RipSeq Mixed results. Both genotypes are targeted by LA. In samples with 2 genotypes by RipSeq, LA detected only 1 of the 2 genotypes in 7 samples (19%), the same 2 genotypes in 10 samples (28%), and the same 2 genotypes plus additional types in 18 samples (50%). Of 9 samples with 3 genotypes by RipSeq, LA detected only 2 of the 3 genotypes in 1 sample (11%), the same 3 genotypes in none, and the same 3 genotypes plus additional types in 8 samples (89%). Partial and complete concordance were therefore 97% and 49% for samples with 1 genotype, 75% and 28% for samples with 2 genotypes, and 89% and 0% for samples with 3 genotypes by sequencing. When limiting the analysis to HR HPV types, LA was positive in 91 samples and RipSeq in 77 (84.6%). In the remaining 14 samples, LA band intensities were low for 9 (64.3%) HR HPV types, indicating that they constituted the less abundant genotypes (see the examples in Fig. 1).

Fig 1.

More abundant HPV types are preferentially detected by sequencing. LA and sequencing-based genotyping results for samples with HR HPV genotypes were compared. HPV types that were more abundant by LA genotyping (darker line) were more frequently detected by sequencing, and less abundant HPV types were more frequently missed. Positive LA controls are β-globin low (B) and β-globin high (C), as well as a genotype decoding reference line (A).

In addition to significant workflow advantages, RipSeq Mixed provided fast and accurate genotype information for cervical cytology samples containing one HPV type. By customizing the sequence database, the user has the ability to select the most accurate and informative reference sequences, enabling more rapid and reproducible genotyping. As expected, concordance of RipSeq Mixed and LA analysis depended largely on the number and relative abundance of HPV genotypes. In samples containing 2 HPV genotypes, RipSeq-based interpretation of partial L1 sequences allows for rapid interpretation of mixed chromatograms that may otherwise be tedious or impossible to resolve. While hybridization-based methods are often labor-intensive and inherently limited to those types that can be reliably differentiated, sequencing-based genotyping enables detection and identification of any genotype, including new (sub)types that may be of epidemiologic or pathogenic relevance (24).

Detection, especially of less abundant viruses in samples containing more than 2 HPV genotypes by methods requiring PCR amplification, is inherently limited by primer bias (23), and their detection with LA also shows variability (25). The reduced sensitivity of Sanger sequencing for minor populations (9) further complicates their identification with the RipSeq Mixed algorithm. For 16S sequencing, primer bias has been reduced by using several group-specific, broad-range primers in separate reactions (9). Similarly, broad-range primer sets targeting different clades of genital-mucosa HPV types may allow for more homogenous amplification. However, the lack of an unbiased reference method therefore complicates comparison of different genotyping methods. In the future, detection of a virtually unlimited number of HPV types by next-generation sequencing may overcome many of the problems of current methods for high-resolution HPV typing (10, 26).