Highly conserved intragenic HSV-2 sequences: results from next-generation sequencing of HSV-2 U_L and U_S regions from genital swabs collected from 3 continents

Abstract

Introduction

Understanding the variability in circulating herpes simplex virus type 2 (HSV-2) genomic sequences is critical to the development of HSV-2 vaccines.

Methods

Genital lesion swabs containing ≥10⁷ log₁₀ copies HSV DNA collected from Africa, the USA, and South America underwent next-generation sequencing, followed by K-mer based filtering and de novo genomic assembly. Sites of heterogeneity within coding regions in unique long and unique short (U_L_U_S) regions were identified. Phylogenetic trees were created using maximum likelihood reconstruction.

Results

Among 46 samples from 38 persons, 1468 intragenic base-pair substitutions were identified. The maximum nucleotide distance between strains for concatenated U_{L_}U_S segments was 0.4%. Phylogeny did not reveal geographic clustering. The most variable proteins had non-synonymous mutations in <3% of amino acids.

Conclusions

Unenriched HSV-2 DNA can undergo next-generation sequencing to identify intragenic variability. The use of clinical swabs for sequencing expands the information that can be gathered directly from these specimens.

Introduction

Herpes simplex virus type 2 (HSV-2) is a highly prevalent sexually transmitted human pathogen (1) which is the leading cause of genital ulcer disease worldwide (2–5), and fuels the HIV epidemic by increasing the risk of HIV acquisition 2–3 fold (6). The clinical manifestations of genital HSV-2 infection are highly variable, with most people having unrecognized infection due to mild symptoms (7). While development of a prophylactic HSV-2 vaccine is a global priority to improve sexual health worldwide (8, 9), several candidates tested in clinical trials have failed to show efficacy (10–12). Knowledge of genomic HSV-2 sequence variation may allow us to design vaccines to optimize efficacy globally, and may provide insight into the pathogenesis of HSV-2 infection. Although several groups have sequenced HSV-2 cultured isolates, to date sequencing of HSV-2 DNA directly from genital swabs has not been described. HSV PCR is more sensitive than culture and is a preferred method for diagnosis of genital herpes (13, 14). Furthermore, results are not affected by ambient temperature at time of collection and swabs can be stored for prolonged periods at room temperature (13, 15). Thus, the ability to sequence HSV-2 without the need to culture the virus would expand the number of samples that can be sequenced from diverse locations and further contribute to rational vaccine design and understanding of viral evolution.

Improvements in DNA sequencing, resulting in lower costs from increasingly efficient platforms, have led to the recent publication of multiple HSV genomes. Publication of 20 full length HSV-1 sequences collected from around the world has provided information about genomic diversity, and a phylogenetic analysis of 31 full length HSV-1 sequences revealed a minimum of 6 clades with geographic clustering (16, 17). The first full length HSV-2 isolate sequenced was HG52, a clinical strain collected prior to 1971 in Scotland (18) that has been passaged multiple times in the laboratory (19). SD90e, a clinical isolate collected in South Africa that has undergone minimal laboratory passaging, was the second HSV-2 full-genome sequence, and multiple polymorphisms were identified compared to HG52 (19). Recently, 5 low-passage cultured clinical strains were sequenced by Kolb et al. and 5 additional clinical strains were sequenced by Petro et al.; all samples were collected in the USA (20, 21). Additionally, 34 low-passage clinical and laboratory cultured strains from the USA, Japan, and Uganda were sequenced by Newman et al. and 18 low-passage clinical isolates from multiple African countries were sequenced by Burrel et al. (22, 23).

Full-genome viral DNA has been sequenced without need for culture for other herpesviruses (24–26), and human papillomavirus (HPV) has been sequenced without culture from formalin fixed paraffin embedded tissue as well as skin swabs from sites of condyloma acuminata (27, 28). Our group has enrolled participants into natural history studies and clinical trials of genital HSV-2 infection in Seattle, WA, for more than 30 years, and we have provided laboratory support for studies conducted in Peru and East and southern Africa. We maintain a repository of well-characterized samples from these studies, and we selected a subset of samples with high copy number of HSV DNA to develop a high-throughput sequencing platform. Here we demonstrate that genomic sequence of the coding regions located in unique long (U_L) and unique short (U_S) portions of the HSV-2 genome can be obtained from DNA swabs without culture amplification.

Results

Demographic information

Forty-six samples with ≥7 log₁₀ copies/swab HSV DNA collected from genital mucosa had >10-fold average coverage of the HSV-2 genome (Table 1). These samples were from 38 persons; 8 persons contributed 2 samples. Eleven (29%) persons were from the USA, 12 (32%) were from Peru, and 15 (40%) were from sites in Africa (Zimbabwe, Zambia, South Africa, Kenya, Tanzania, and Uganda) (Table 1). Nineteen (50%) participants were men, and 24 (62%) participants were HIV seronegative (Table 1). For the 8 participants who had 2 samples sequenced from different time points, the median time between specimen collection was 2.1 years (range 3 months-15.7 years).

Table 1.

Demographics and genome coverage of included samples.

Sample Name	Sample number	Year	Country	Sex	HIVstatus	Meancoverage	% of genome with ≥NX coverage			% of genome with “N”	Genbank accession #
							N=1	N=3	N=10
Kenya.1	2009–4556	2009	Kenya	Male	Positive	210.7	98.9	98.3	97.3	1.4	KX574894.2
Kenya.2	2009–4550	2009	Kenya	Male	Positive	43.4	98.1	96.7	91.4	3.8	KX574893.2
Peru.1	2007–38205	2007	Peru	Male	Negative	618.4	98.9	98.2	97.1	1.4	KX574860.2
Peru.2	2007–38606	2007	Peru	Male	Negative	208.7	98.1	96.8	93.8	0.9	KX574873.2
Peru.3	2007–38910	2007	Peru	Male	Negative	16.8	93.7	88.9	73.9	3.3	KX574877.2
Peru.4	2007–38901	2007	Peru	Male	Negative	157.7	98.1	96.8	93.5	2.7	KX574876.2
Peru.5	2007–39303	2007	Peru	Male	Negative	106.4	97.2	95.8	90.8	2.3	KX574878.2
Peru.6	2007–38807	2007	Peru	Male	Negative	251.3	98.3	97.4	95.0	6.7	KX574875.2
Peru.7	2007–38249	2007	Peru	Male	Negative	60.4	97.3	95.3	89.6	3.1	KX574872.2
Peru.8	2007–38627	2007	Peru	Male	Negative	106.8	97.8	96.4	92.2	1.6	KX574874.2
Peru.9	2006–30839	2006	Peru	Female	Positive	50.5	96.7	93.9	87.2	5.1	KX574865.2
Peru.10	2006–45028	2006	Peru	Female	Positive	25.3	96.2	93.0	83.8	2.3	KX574867.2
Peru.11	2006–30937	2006	Peru	Female	Positive	186.7	98.2	97.0	94.7	1.4	KX574866.2
Peru.12	2004–4373	2004	Peru	Male	Positive	38.3	96.7	94.0	86.7	3.1	KX574861.2
South Africa.1	2009–3532	2009	South Africa	Male	Positive	16.8	96.7	92.8	71.1	5.6	KX574892.2
South Africa.2	2007–39521	2007	South Africa	Female	Negative	49.2	97.1	95.0	88.4	4.9	KX574879.2
South Africa.3	2007–39712	2007	South Africa	Female	Negative	166.2	98.3	97.3	94.4	5.4	KX574880.2
South Africa.4	2007–39729	2007	South Africa	Female	Negative	64.6	97.1	95.0	89.0	4.0	KX574881.2
Tanzania.1	2009–409	2009	Tanzania	Female	Positive	34.7	97.7	95.5	88.2	5.9	KX574902.2
Uganda.1	2011–34727	2011	Uganda	Male	Negative	66.2	97.7	95.8	90.1	4.3	KX574899.2
Uganda.2	2012–9333	2012	Uganda	Male	Positive	12.9	93.4	86.3	64.9	4.0	KX574906.2
USA.1a	2003–16029	2003	USA	Male	Positive	11.2	93.1	84.8	57.7	6.5	KX574903.2
USA.1b	2007–22031	2007	USA	Male	Positive	16.1	93.8	87.0	68.4	6.8	KX574870.2
USA.2a	2004–20601	2004	USA	Male	Positive	22.4	94.2	88.8	78.8	4.4	KX574904.2
USA.2b	2008–37496	2008	USA	Male	Positive	11.6	91.9	84.8	61.6	5.2	KX574885.2
USA.3a	2007–40807	2007	USA	Female	Negative	39.6	96.5	93.6	85.6	6.2	KX574882.2
USA.3b	2008–16404	2008	USA	Female	Negative	136.2	98.1	97.1	94.3	4.9	KX574884.2
USA.4a	2008–7539	2008	USA	Female	Negative	147.7	98.4	97.5	95.3	3.8	KX574887.2
USA.4b	2011–33490	2011	USA	Female	Negative	31.3	96.5	93.2	84.9	5.0	KX574898.2
USA.5a	2009–33778	2009	USA	Female	Negative	30.6	96.9	94.4	85.8	2.7	KX574891.2
USA.5b	2010–6416	2010	USA	Female	Negative	97.1	97.8	96.6	92.6	3.5	KX574896.2
USA.6a	2009–13458	2009	USA	Female	Negative	63.7	97.3	95.7	90.5	2.7	KX574889.2
USA.6b	2009–31735	2009	USA	Female	Negative	84.7	97.5	95.9	91.1	3.6	KX574890.2
USA.7a	2009–1245	2009	USA	Female	Negative	14.8	92.4	85.0	66.0	5.5	KX574888.2
USA.7b	2010–21	2010	USA	Female	Negative	46.7	96.5	92.0	85.1	4.3	KX574895.2
USA.8a	2006–49895	2006	USA	Male	Negative	74.4	95.1	91.6	86.3	6.5	KX574868.2
USA.8b	2006–50074	2006	USA	Male	Negative	44.6	94.2	90.0	83.5	5.6	KX574869.2
USA.9	2012–32825	2012	USA	Female	Positive	171.6	97.3	95.6	91.2	5.9	KX574901.2
USA.10	2008–37744	2008	USA	Female	Negative	45.7	95.4	91.0	84.7	4.5	KX574886.2
USA.11	2011–21761	2011	USA	Male	Positive	25.9	94.7	90.2	80.1	4.6	KX574897.2
Zambia.1	2009–2198	2009	Zambia	Male	Positive	30.7	95.8	92.8	78.8	3.6	KX574905.2
Zambia.2	2006–29060	2006	Zambia	Female	Negative	49.3	96.7	93.9	86.3	2.9	KX574864.2
Zambia.3	2006–19390	2006	Zambia	Female	Negative	585.6	99.1	98.2	97.1	1.8	KX574862.2
Zambia.4	2007–8222	2007	Zambia	Female	Negative	121.1	97.9	96.5	92.3	1.1	KX574883.2
Zimbabwe.1	2007–38120	2007	Zimbabwe	Female	Negative	178.3	98.1	97.1	94.5	2.0	KX574871.2
Zimbabwe.2	2006–21832	2006	Zimbabwe	Female	Negative	274.8	98.5	97.7	96.1	1.0	KX574863.2

Coverage of the HSV-2 genome

Most reads from each sample were human (median 97.8%, IQR 95.6%-99%). DNA sequences matching bacterial species were also detected (data not shown). Despite this, we were able to obtain a minimum of 10-fold average coverage of the HSV-2 genome to perform our analyses. The median average depth of coverage of the HSV-2 genome was 55.5X average fold coverage (range: 11.2 – 618.4) (Table 1). The median number of unmapped viral reads per sample was 3.9% (range: 0.9%-6.8%). The median number of nucleotides per sample with ≥10-fold coverage was 88.7% (range: 57.7%-97.3%). However, the genome coverage was not uniform, with poor coverage in the repetitive long (R_L) and repetitive short (R_S) regions and much greater depth of coverage in ORFs in unique long (U_L) and unique short (U_S) regions in samples with overall low- as well as high- fold coverage (data not shown).

Identification of sites of genetic diversity

SD90e (KF781518) was used as the reference strain (19). Overall, we identified 2532 variable loci compared to SD90e; 580 were indels. Of the remaining 1952 loci with base substitutions, 1906 were considered variant loci, with 46 others considered to be unreliable because they had poor coverage at that SNP across specimens (defined as less than 5-fold coverage at that SNP in 26 of 46 genome sequences). Among the 1906 SNPs, 438 were located in non-coding regions. The locations of the remaining base substitutions within coding regions (n=1468 unique SNPs, with 16 overlapping SNPS identified in complementary coding regions of another ORF) are shown in Figure 1. Sites of variability are spread throughout the concatenated genome fragment, rather than clustered in hotspots, although there are some areas with concentrated and prevalent SNPs, such as in U_L39 and U_L30. Analysis of SNP prevalence relative to SD90e demonstrates many highly prevalent SNPs (Figure 1, blue bars). Because the reference sequence represents a minority or rare sequence at some loci, SNP prevalence is at or near 100% at some loci. Per sample, we found that the median number of SNPs that differed from SD90e was 196 (IQR 165,214). The maximum number of SNPs in a single sample compared to SD90e was 266.

Figure 1.

Circos plot of HSV-2 SNPs found in 38 samples, compared to SD90e. Only ORFs located in the U_L_U_S portion of the genome are included in this plot; intervening non-coding regions are removed. ORFs are indicated in gray in the outer circle, with gene name shown. The inner red lines indicate SNP positions. The innermost blue lines indicate the prevalence of each SNP at the specific location, relative to reference genome SD90e, with higher lines showing greater prevalence of each SNP.

We also determined how many samples contained each identified SNP. The median number of samples that differed from SD90e per variant SNP was 1 (IQR=1–4), with 898 (61%) identified SNPs found in only one specimen, and an additional 129 (8%) found in 2 specimens (data not shown).

Phylogeny and divergence

We explored the genetic diversity of our samples, using the concatenated U_L_U_S regions, with the diploid repetitive regions (R_L and R_S), resulting in 122,619 nucleotides remaining (79.6% of the genome, using 154,126 nucleotides as the denominator). We found that all samples were highly related, with maximum nucleotide divergence of 0.4% between the strains in the concatenated U_L_U_S region (data not shown). Phylogenetic trees were created to explore the hypothesis that there HSV-2 strains that differ on the basis of geography. This analysis did not reveal geographic clustering, with strains from USA, Peru, and several African countries intermingled throughout the tree (Figure 2A).

Figure 2. Phylogenetic analysis of HSV-2 sequences.

R_s and R_L regions were excluded from the phylogenetic analysis, leaving the concatenated U_L_U_S region. Maximum likelihood phylogeny reconstruction with bootstrapping (1000 replicates) was used. Support node values are indicated. Figure 2A. Phylogeny of 40 sequenced HSV- 2 samples, including SD90e and HG52. Sequences are indicated with a country name and number, with geographic origin shown in the legend. For the 8 participants who provided 2 samples, only the first sample collected was included. Support values for each branch are indicated. Figure 2B. Eight participants from Seattle provided 2 paired samples collected a median of 2.1 years apart. Support values for each branch are indicated. Median (range) distance between pairs other than USA.1 and USA.2 were on a scale of 0.00002 (0.00000 to 0.00004).

The paired specimens collected from 8 individuals over time were closely related or the same, except in the cases of USA.1 and USA.2. These participants appeared to be infected with 2 different strains (Figure 2B), with the paired samples of USA.1 and USA.2 each differing by 186 base pairs. Both pairs had a Tamura-Nei distance of 0.002. In contrast, 6 pairs of samples collected from the same person over time appeared to be nearly identical, with a median of 0 (range: 0–1) basepair differences within each pair and a Tamura-Nei distance of 0.00002 (range 0.00000–0.00004).

Amino acid variation

To determine the ORFs with the highest variability, we explored proportions of overall and non-synonymous (dN) mutations for each ORF (Figure 3). Similar to the nucleotide level, amino acid variability was spread throughout the genome and did not appear to be clustered for any functional or kinetic class (Figure 3). The most variant protein, encoded by U_L49A (probable virion membrane protein), had ~2.5% protein length adjusted amino acids with non-synonymous mutations, but is a particularly short ORF (88 amino acids). The ORFs with the greatest degree of amino acid variability also included U_S4 (glycoprotein G), U_S5 (glycoprotein J), U_L39 (ribonucleotide reductase large subunit), U_L29 (single-stranded DNA binding protein), U_L56 (membrane protein), U_L49 (tegument protein), U_S9 (tegument protein) and U_S1 (ICP4 immediate early transactivator).

Figure 3.

Amino acid variation in each protein, shown as percent of synonymous and non-synonymous mutations. The percent was calculated by taking the number of dS or dN mutations and dividing by the length of the ORF. Synonymous nucleotide variation is noted in light gray, non-synonymous mutations are noted in dark gray. ORFs are organized functional group when available, or kinetic class.

To determine if our results were analogous to those obtained from cultured HSV-2, we compared the percentage of SNPs detected in each gene in our analysis to the recent report of 34 low-passage isolates by Newman et al. (22). Both methods detect a similar proportion of SNPs in each ORF, r=0.67, p<0.0001) (data not shown).

Variation in HSV-2 vaccine candidate proteins

Next, we examined the sequence variation in selected proteins currently being developed as potential subunit or DNA vaccines for both therapeutic and prophylactic indications (9) (Figure 4). gD2, encoded by gene U_S6, is a documented human CD4, CD8, and neutralizing antibody target (29–31) which has been used in several prophylactic and therapeutic subunit vaccines (10, 32). We identified novel variants, K147R, I201L and A335T, in strains South Africa.3, Zambia.1, and Uganda.2, respectively (Figure 4). We also detected the previously described V169A and V353A variants in our specimens (33, 34). Three of the five SNPs were detected in only one strain, indicating that gD2 structure is highly constrained. In contrast, we observed a larger degree of variation in other HSV-2 proteins previously or currently under investigation as subunit vaccines. For example, the virion tegument protein VP11/12 (gene U_L46) has immune modulatory function, is recognized by human CD4 and CD8 T-cells (29, 35) and is included in a candidate DNA vaccine (36, 37). We detected 9 non-synonymous changes, with several novel variants identified (K110N, D407N, D488G, and N698S). However, 4 of the non-synonymous changes were detected in only 1 or 2 sequences.

Figure 4.

Sites of amino acid changes in ORFs of selected vaccine candidates and proteins of diagnostic and therapeutic significance. Only sites of variability are indicated. Amino acid changes compared to SD90e (top row) are shown in colors, while identical amino acids are indicated by dots and amino acids for which we could not make an amino acid call are indicated with an X. Percentages of sequences matching the reference alleles are shown at the bottom. Examples are U_S6 (gD2), U_L23 (thymidine kinase), and U_L46 (VP11/12). Also included is U_S4 (gG2), the protein that forms the basis for type-specific serologic testing.

Mutations in therapeutic and diagnostic targets

U_L23 (thymidine kinase) which phosphorylates the nucleoside analogue acyclovir and is associated with antiviral resistance mutations (38–41), had amino acid variability at ~1% of amino acids (Figure 3). In addition, of the 11 amino acid variants identified in U_L23, only 3 variants, all of which had been previously identified (42), occurred in more than one sample (G39E, N78D, L140F) (Figure 4). We identified 2 novel amino acid variants, each present in only 1 sample (L230M, in a sample Kenya.1, and L263P in sample Uganda.1). The functional importance of these novel variants is unknown. Antibody to gG2 (encoded by U_S4) is used to diagnose HSV-2 infection with type-specific serologic assays (43). Glycoprotein G2, a large protein with 700 amino acids, was among the most variable HSV-2 proteins, with 39 amino acid substitutions (Figure 3). The most common site of amino acid variability (L423P) was conserved in 63% of strains, while most amino acid substitutions (25 [64%]) were found in only one sample (Figure 4). Interestingly, sample Zambia.4 had 14 unique non-synonymous changes in gG2. Thus, even the most variable protein had non-synonymous mutations found in the minority of samples, indicating a high degree of conservation.

Our group has used a gB probe-primer set to detect HSV infection over the past 20 years (44). We examined the degree of variability of the sequence in the primer set in our 39 samples. We found that none of the samples had any variants detected within the forward and reverse primers or the probe (data not shown). In addition, we use a 16 base pair gB2 sequence for strain typing (45), and we did not find any base pair mutations in this sequence. These data suggest that the selected probe-primer pairs to detect and type HSV-2 are highly conserved and remain suitable for diagnosis of HSV-2 infection.

Discussion

Our study contributes 46 partial HSV-2 sequences from diverse global locations to a growing sequence database. We demonstrate that HSV-2 genomes can be recovered from genital swabs containing unenriched HSV-2 DNA to identify variable sites within coding regions of U_L and U_S. Importantly, we find that the maximum nucleotide diversity, the overall number of SNPs per specimen, and the detection of SNPs in open reading frames are similar to those found using tissue culture (20, 22). We found a high degree of conservation between strains, particularly in coding regions and specifically in vaccine candidate genes and portions of the genome used for diagnostic assays. These data will provide insight into the development of new tools to treat HSV-2 infection, including selection of vaccine candidates.

Recent HSV-2 sequencing efforts have contributed multiple additional full-length HSV-2 sequences. Kolb et al. sequenced 5 low-passage clinical isolates and 1 high-passage laboratory isolate from the United States and Newman et al. sequenced 34 low-passage cultured isolates from the USA, Africa, and Japan (20, 22). These cultured isolates yielded much greater depth of coverage than our analysis, with 140–203 fold coverage in Kolb et al. (20) and 3100–9300 fold coverage in Newman et al. (22). However, similar to our findings, these sequencing efforts with much better depth of coverage also had limited coverage of the R_L and R_S regions. These areas are difficult to sequence and map due to high GC content and repetitive regions. Importantly, our work used genital swabs with HSV DNA that did not undergo culture prior to sequencing, demonstrating that culture is not required to obtain high quality HSV sequence of the non-repetitive regions of the HSV-2 genome without enrichment. Studies of HSV during serial passage reveal instances of acquired mutations, even in the absence of known selective pressures such as antiviral drugs, and in the absence of phenotypic changes such as altered plaque morphologies that usually correlate with glycoprotein coding mutations (46, 47). To understand changes that may occur even during early passage of virus, it will be of interest to compare simultaneously collected viral culture isolates to DNA swabs. Given the large number of DNA swab samples that we and others have collected from around the world and over time in the same person, the ability to utilize these samples for sequencing greatly expands the number of samples available for sequence analysis. For example, further studies of viral evolution within a single participant during distinct shedding episodes, or during a symptomatic and an asymptomatic shedding episode when lower quantities of virus may preclude culture, could be performed.

We found that HSV-2 is highly conserved from diverse geographic locations, with maximum genomic distance of 0.4% and a low number and prevalence of nonsynonymous mutations, including proteins that are being developed for vaccines, such as gD2, or used in HSV diagnosis, such as gG2 and gB2. The genomic distance is the same as that reported by Newman et al. and Kolb et al. (20, 22). Although Petro et al. described a much higher divergence in a small number cultured clinical isolates, additional sequencing will be required to extend this finding and determine the genomic distance between strains (48). The global diversity of HSV glycoproteins has been described, with a unique amino acid signature gG2 involving amino acids 291, 338, 372, and 546 in specimens from East Africa (49). None of the specimens analyzed, including 6 from East Africa, had this amino acid signature. In 6 of 8 persons with paired specimens we found that the virus was virtually unchanged over a median of 2.1 years. This is in contrast to other viruses, such as HIV, which have a median of 25–35% substitutions in some proteins between subtypes from different geographic areas and significant divergence within the host over time (50). The high sequence conservation for HSV-2 suggests that development of an effective vaccine using conserved epitopes may not be hindered by global antigenic diversity, although additional sequencing of genomes from as yet unexplored geographic regions may reveal additional sources of diversity, such as HSV-1xHSV-2 recombinants as recently described (51). For example, Burrel et al. described HSV-2v, a strain with divergent UL30 sequences which to date has been found in persons from Africa (23, 52). With increased knowledge of HSV-2 specific and cross-reactive T cell epitopes to alphaherpesviruses (53), future studies may also explore the issues of T cell mediated immune escape between and within persons.

We found that two men with paired samples appear to have been infected with a second strain of HSV-2; both men were infected with HIV. While superinfection has been described using techniques such as PCR assays based on variation in DNA repeats (54, 55), greater knowledge of HSV-2 diversity will allow for more sophisticated methods to identify and quantify the prevalence of HSV-2 superinfection.

We used specimens collected from around the world to enhance understanding of geographic variability of HSV-2, which will likely have important implications for vaccine development. We did not find evidence of geographic clustering, similar to what has been described by Newman et al. in 34 cultured strains from the USA, Africa, and Japan (22). However, this finding is remarkably different from HSV-1, is which geographic clustering has been described in Asian and African samples (16). Prior analyses of HSV-2 recombination with full length sequences have detected 5 major crossover events, a much lower rate than for HSV-1 (16, 22). Additional analyses of recombination events in our sequences and others have recently been published and reveal recombinants between HSV-1 and HSV-2 in U_L29, U_L30, and U_L39 (23, 56). Our phylogenetic analyses did not account for recombination, and therefore, may have underestimated the importance of geography. However, we did not find evidence of geographical grouping in the U_L39 HSV-1 X HSV-2 recombination site among a much larger sample size (183 specimens collected from around the world) (51). Recent viral divergence analyses using sequences from 12 glycoproteins have suggested that HSV-2 evolved from chimpanzee herpes simplex virus (ChHV) 1.6 million years ago, after a more distant co-divergence of HSV-1 and ChHV 6 million years ago (57). Availability of full length HSV-2 sequences will further clarify the evolutionary origins of HSV-2.

Our study had several strengths and limitations. Although we had limited depth of coverage, particularly in repetitive regions of the genome, our goal was to identify population prevalent SNPs in coding regions, both to identify targets for vaccine development and to develop robust genotyping methods. Our approach did not allow for analysis of the highly repetitive regions R_S and R_L regions of HSV-2, which are difficult to resolve even with much deeper coverage of the genome (16, 19, 20, 22). There may be important regions of variability in these repetitive areas that we were not able to identify with this analysis. In addition, it is possible that SNPs within U_L_U_S coding regions of individual samples were not identified in this study due to low coverage at that site. Finally, the limited coverage depth was not optimal for the study of within-sample minor alleles, which is another potential important source of variation.

We sequenced genital swabs with high quantities of HSV DNA, many of which were collected from genital lesions. It is possible that these samples with high quantities of HSV DNA may be different from samples obtained during asymptomatic shedding episodes, which are often associated with low quantities of HSV. Enrichment strategies may be required to sequence from HSV obtained from swabs of asymptomatic genital HSV-2 shedding episodes, which often have less HSV DNA detected (58). Increased understanding of HSV-2 heterogeneity has great potential to contribute to vaccine development, novel therapeutics, genotyping, and to enhance understanding of viral pathogenesis, evolution, and host immune responses to infection. Additional sequencing of well-characterized clinical HSV-2 strains from diverse geographic locations will provide a richer understanding of viral diversity and pathogenesis.

Materials and methods

Patients and samples

Genital lesion swabs with ≥7 log₁₀ copies HSV DNA/swab specimen were collected from participants in natural history studies and clinical trials in the United States, Peru, and sites in sub-Saharan Africa (Zimbabwe, Zambia, South Africa, Kenya, Tanzania, and Uganda) between 1993 and 2012 (59, 60). Eight participants from Seattle each had 2 swab samples collected at different time points sequenced. Swabs were collected from the genital area, often at the site of a genital lesion, using a Dacron swab, and placed into 1 ml of 1X PCR buffer. Swabs were stored at −20°C in Seattle, WA. All participants provided informed consent for the collection of genital swabs and procedures were approved by the University of Washington Human Subjects Review Committee.

Laboratory Methods

DNA was extracted and HSV PCR was performed, using HSV-1/HSV-2 type-common primers to glycoprotein B (gene U_L27) as previously described (61). Twenty nanograms total DNA extracted from swabs was used to create libraries using the NuGen (San Carlos, CA) low input kit. Samples were multiplexed 6 per lane (62) and 100 bp paired end reads were performed on the Illumina (San Diego, CA) HiSeq2500 platform.

Alignment and Variant Calling

Basecalling and demultiplexing of reads was performed by Illumina HiSeq RTA and CASAVA 1.8.2 software, with resulting Fastq files filtered to exclude read pairs that failed standard Illumina quality checks. We used a k-mer based filtering approach (bbduk v36 with k = 31 and hdist = 2) to identify reads matching either the HG52 or SD90e strains of HSV-2 (Genbank JN561323.2and KF781518.1 respectively) and all other reads were discarded (63). We then performed further filtering to remove adapter contaminants (bbduk with k = 20, mink = 11, hdist = 1) and low quality regions with Q < 28 at the 3′ and 5′ ends (Trimmomatic v0.36) (64). Reads shorter than 20 nt after trimming were discarded and reads passing quality controls were mapped to HG52 and SD90e. De novo assembly was performed using SPAdes with read error correction and automatic selection of k-mer sizes (65). Assembled scaffolds were mapped to SD90e using Mugsy; gaps between scaffolds were filled using mapped reads and a consensus sequence was constructed from merged alignments (66). SAMtools and BCFtools v1.3.1 were used to compute genotype likelihoods and call variants at each position relative to the reference sequence SD90e (67, 68). We did not include an analysis of heterozygotes, defined as two different base calls for the same nucleotide site within the same specimen, in the variant calls because we were not confident that we would be able to detect minority variants with the limited coverage depth.

The mpileup command from Samtools and a python script were used to create U_L_U_S fasta files. These files contained the unique long (U_L) and unique short (U_S) regions concatenated together, including individual ORFs and intervening regions. The repetitive long (R_L) and repetitive short (R_S) regions were excluded from this analysis (KF781518 positions 1-9403, 117737-13341, 147702-154764). This removed coding genes R_L1 (protein, γ34.5), R_L2 (protein, ICP0), R_S1 (protein, ICP4), and DNA specifying the latency associated transcripts from analysis. These files were combined with the U_L_U_S region of SD90e (69) with ClustalW (70).

Phylogeny

phangorn within R was used to construct phylogenetic trees (71). Model selection was based on Akaike Information Criterion (AIC); the generalised time-reversible model using the gamma distribution with testing of invariant sites was selected (72). Further optimization was performed using maximum likelihood (73). Bootstrapping with 1000 replicates was used to compute support values (74).

DNA and protein analysis

Using the U_L_U_S file, Circos software (version 0.64) was used to create a genetic map showing the presence of base substitutions in the open reading frame (coding) regions of the genome (75). Among all identified base substitutions from the 38 sequences from unique participants and SD90e, only base substitutions with a valid base call at that position in at least 20 of the sequences were included. For participants that had 2 samples sequenced, we included only the first submitted sample to avoid having duplicate strains contribute twice to the analysis.

DNA sequences were translated to the amino-acid level, accounting for reverse and complementary sequences, using Bioconductor (version 2.14) in R (version 3.1.1). The frequency of variability for each allele was calculated by taking the column frequency against the reference sequence SD90e, as the proportion of amino acids at each location that differ from the reference. DNA codons with missing alleles for any of the three nucleotides were considered unreadable codons and were removed from variability analysis. In each annotated ORF in the reference genome, the total number of SNPs divided by the length of coding region and the number of base substitutions causing variation on amino acid level divided by the length of coding region was calculated to determine the frequency of synonymous and nonsynonymous mutations. Graphical presentations of the amino acid alignments and variants for proteins with diagnostic and therapeutic significance were created in R (version 3.1.1). Within these proteins, unique non-synonymous amino acid mutations which have not been previously described were confirmed by manual inspection of sequence.

GenBank accession number

All sequences were entered into GenBank with accession numbers shown in Table 1. Raw reads were also deposited in the NCBI Sequence Read Archive under project number PRJNA280778

Highlights.

• HSV-2 genomes were recovered from genital swabs containing unenriched HSV-2 DNA.

• HSV-2 was highly conserved without strains based on geography among 3 continents.

• Among selected HSV-2 vaccine and diagnostic targets, coding variation was rare.