Human Herpesvirus 6B Induces Hypomethylation on Chromosome 17p13.3, Correlating with Increased Gene Expression and Virus Integration

ABSTRACT

Human herpesvirus 6B (HHV-6B) is a neurotropic betaherpesvirus that achieves latency by integrating its genome into host cell chromosomes. Several viruses can induce epigenetic modifications in their host cells, but no study has investigated the epigenetic modifications induced by HHV-6B. This study analyzed methylation with an Illumina 450K array, comparing HHV-6B-infected and uninfected Molt-3 T cells 3 days postinfection. Bisulfite pyrosequencing was used to validate the Illumina results and to investigate methylation over time in vitro. Expression of genes was investigated using quantitative PCR (qPCR), and virus integration was investigated with PCR. A total of 406 CpG sites showed a significant HHV-6B-induced change in methylation in vitro. Remarkably, 86% (351/406) of these CpGs were located <1 Mb from chromosomal ends and were all hypomethylated in virus-infected cells. This was most evident at chromosome 17p13.3, where HHV-6B had induced CpG hypomethylation after 2 days of infection, possibly through TET2, which was found to be upregulated by the virus. In addition, virus-induced cytosine hydroxymethylation was observed. Genes located in the hypomethylated region at 17p13.3 showed significantly upregulated expression in HHV-6B-infected cells. A temporal experiment revealed HHV-6B integration in Molt-3 cell DNA 3 days after infection. The telomere at 17p has repeatedly been described as an integration site for HHV-6B, and we show for the first time that HHV-6B induces hypomethylation in this region during acute infection, which may play a role in the integration process, possibly by making the DNA more accessible.

IMPORTANCE The ability to establish latency in the host is a hallmark of herpesviruses, but the mechanisms differ. Human herpesvirus 6B (HHV-6B) is known to establish latency through integration of its genome into the telomeric regions of host cells, with the ability to reactivate. Our study is the first to show that HHV-6B specifically induces hypomethylated regions close to the telomeres and that integrating viruses may use the host methylation machinery to facilitate their integration process. The results from this study contribute to knowledge of HHV-6B biology and virus-host interaction. This in turn will lead to further progress in our understanding of the underlying mechanisms by which HHV-6B contributes to pathological processes and may have important implications in both disease prevention and treatment.

INTRODUCTION

Human herpesvirus 6B (HHV-6B) is a betaherpesvirus that shares 90% of its nucleotide sequence with the closely related HHV-6A (1, 2). HHV-6B is almost universally acquired by humans before the age of 2 years (3–5), and its primary infection causes febrile illnesses, including roseola (5, 6). As is the case with all herpesviruses, HHV-6B can achieve latency after primary infection and can be reactivated later in life, resulting in potentially serious secondary complications, particularly in immunocompromised patients. One example is that of allogeneic hematopoietic cell transplantation patients, in whom HHV-6B reactivation is common and can lead to encephalitis (reviewed in reference 7). HHV-6B has also been identified in brain tissue surgically removed from patients treated for temporal lobe epilepsy (8–10). However, the pathological implications of this for epileptogenesis are unclear.

The ability to establish latency in the host is a hallmark of all herpesviruses, but the mechanisms differ. Both HHV-6A and HHV-6B can achieve latency by integrating their genomes into the telomeric regions of host cell chromosomes (11, 12). Integration can occur in somatic cells or germ cells; in germ cells, there is a 50% probability that the offspring would have this viral genome integrated in every cell of the body. This is called inherited chromosomally integrated HHV-6A/6B (ici-HHV-6A/6B) and has been detected in approximately 1% of the human population (reviewed in reference 13). It has been reported that reactivation of these integrated viruses can occur (11, 14) and that ici-HHV-6A and ici-HHV-6B are associated with different disease groups (14).

The mechanism by which HHV-6A and HHV-6B (HHV-6A/B) integrate their genomes into the host cell genome is not fully understood but has been suggested to involve the protein U94 (15). This protein is unique to HHV-6A/6B and has no orthologue in the other human herpesviruses, but it has an overall amino acid identity of 24% with the adenoassociated virus type-2 (AAV-2) Rep68 integrase (16, 17). Interestingly, U94 possesses all of the functions required for HHV-6A/6B chromosomal integration: it can bind telomeric DNA sequences, hydrolyze ATP, and function as both helicase and exonuclease (17). However, recent findings revealed that HHV-6A integration take place even in the absence of U94 (18) but not without the arrays of telomeric repeats (TMRs) present in the virus genome (19). This suggests that the process of HHV-6A/B integration is complex and that more research is required to understand this process.

Epigenetic control of gene expression primarily involves DNA methylation and histone modifications, where there are strong correlations among DNA methylation, histone deacetylation, the presence of tightly packed chromatin, and transcriptional repression (reviewed in references 20 and 21). DNA methylation occurs when a methyl group is covalently added to the carbon-5 position of a cytosine base in the DNA. The methyl transfer is executed by DNA methyltransferases (DNMTs). Even though methylation is regarded as a stable modification, methylated cytosines can be demethylated. This can be achieved by passive demethylation or by active oxidation executed by ten-eleven translocation (TET) enzymes. The term “passive demethylation” refers to a lack of methyl groups or methyl donors or an absence of DNMTs and thus to a process that can occur only when there is also DNA replication. In active demethylation, oxidization of the methyl group to hydroxymethyl, formyl, or carboxyl groups occurs and the oxidized products can then be converted to unmethylated cytosine via thymine DNA glycosylase (TDG)-mediated base excision repair (reviewed in references 22, 23, and 24).

The host epigenetic machinery is used in the defense against viruses, but viruses have also evolved strategies to use this machinery for their own survival (reviewed in references 20, 25, and 26). For example, viruses can use methylation as a way to regulate their life cycle, with latent virus DNA being heavily methylated and thereby protected from immune detection but with the ability to become hypomethylated, resulting in reactivation (27). Another example is that some viruses encode proteins that can interact with epigenetic modulators, which may repress expression of immune response genes. Modulation of the epigenetic landscape in this way can have serious implications for the host cells, as many genes are affected. One example is cancer caused by Epstein-Barr virus (EBV) infection, where EBV induces expression of DNA methyltransferase 3b, resulting in increased DNA methylation and decreased expression of many important host cell genes (28, 29). How HHV-6B affects host cell DNA methylation has not yet been investigated, and the objective of this study was therefore to analyze DNA methylation patterns and effects on gene expression. This is of great importance for understanding HHV-6B biology and the interplay between host and pathogen and if HHV-6B-induced molecular and cellular changes might have any pathological consequences.

(A previous version of this paper was included in the doctoral thesis “Human herpesvirus 6A and 6B: assay validation, virus-host interaction and clinical relevance” [E. Engdahl, Karolinska Institutet] [30].)

RESULTS

HHV-6B infection in vitro induces CpG hypomethylation at the end of chromosomes.

As the effect of HHV-6A/6B on DNA methylation had not yet been investigated, a hypothesis-free approach was chosen for investigating the effect of HHV-6B on >450,000 CpG sites in the human genome. The Illumina 450K array revealed that HHV-6B infection of Molt-3 cells is associated with hypomethylation of many CpG sites in the host cell genome at 3 days postinfection (dpi) (Fig. 1). This hypomethylation was nonrandom and was specifically taking place close to the ends of chromosomes, especially at chromosome 17p13.3 (Fig. 2). In total, 406 CpG sites showed a significant (adjusted P value, <0.05) change in methylation between HHV-6B-infected and uninfected Molt-3 cells. (For the chromosomal location and the relationship to the gene, see Table S1 in the supplemental material.) Remarkably, 351 (86%) of these significant CpGs were located <1 M bases from a chromosomal end and 100% of these 351 CpGs were hypomethylated in virus-infected cells.

FIG 1.

Volcano plot of the Illumina data in vitro. Each dot represents the difference in methylation between uninfected and HHV-6B-infected Molt-3 cells in one CpG site. A negative M value indicates HHV-6B-induced hypomethylation, while positive M values indicate HHV-6B-induced hypermethylation. The red line indicates the threshold for significance (FDR < 0.05).

FIG 2.

Manhattan plot of the Illumina data in vitro. Each dot represents a CpG site, and dots reaching values above 1.3 (FDR < 0.05) are regarded as representing statistical significance. Increased densities of dots indicate that that region is differently methylated, here mostly observed close to the beginning and the end of the chromosomes.

A 0.6-Mb region at the end of 17p13.3 was by far the most hypomethylated region, and this region was further investigated. Looking into the Illumina results from this 0.6-Mb region at greater depth, 38% (127/333) of the investigated CpGs were significantly hypomethylated by HHV-6B. Among the nonsignificant CpGs, 89% (184/206) also appeared to be less methylated in virus-infected cells, indicating that more CpG sites may be affected in the same direction, although the results did not reach a false-discovery rate (FDR) of <0.05. The average beta values of the 333 CpGs in this region were 0.78 in uninfected cells and 0.68 in HHV-6B-infected cells, indicating an average 10% loss of methylation in this region at 3 dpi. This extremely hypomethylated region of chromosome 17p13.3 contains four genes: DOC2B, RPH3AL, RFLNB, and VPS53.

In order to technically validate some of the findings obtained with the Illumina array, new triplicate 3-dpi samples were analyzed with bisulfite pyrosequencing for CpGs located at chromosome 17p13.3. Five CpG sites with results that were significant in the Illumina array were chosen for validation, and seven extra CpGs adjacent to these CpGs were analyzed at the same time. Confirming the Illumina results, all 12 analyzed CpGs were significantly less methylated in HHV-6B-infected Molt-3 cells than in uninfected cells (Fig. 3).

FIG 3.

Methylation frequency at 3 dpi, obtained with pyrosequencing. Investigated CpG sites were annotated to the (A) VPS53 gene, (B) RPH3AL gene, and (C) RFLNB gene. CpG sites with a cg number are CpG sites present on the Illumina array. VPS53 CpG 1 and 2 were obtained with primer pair 1 and CpG 3 and 5 with primer pair 2. Graphs show means and standard deviations of results from triplicate samples. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001, ***** = P < 0.00001.

To investigate the temporal pattern of the hypomethylation observed at 3 dpi, one CpG in the VPS53 gene (cg20843650) was further investigated over time in vitro. As shown in Fig. 4, the hypomethylation induced by HHV-6B was already observable at 2 dpi.

FIG 4.

Methylation of cg20843650 (VPS53) over time. Data represent means and standard deviations of results from triplicate mock-infected and HHV-6B-infected Molt-3 cell samples. hpi, hours postinfection. ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001.

HHV-6B induces TET2 expression.

For demethylation to occur in nondividing cells (HHV-6B induces cell cycle arrest and does not proliferate; data not shown) within 2 days after infection, the process has to be active. Active demethylation most often involves the TET enzymes. Therefore, expression of the three known TET enzymes, TET1, TET2, and TET3, was investigated. Most strikingly, TET2 expression was significantly upregulated by HHV-6B (Fig. 5B). Expression of TET1 and TET2 was also affected by the virus, as evident by HHV-6B having induced decreased TET1 expression at 6 dpi and increased TET3 expression at 3 dpi in virus-infected cells (Fig. 5A and C).

FIG 5.

Relative gene expression levels of (A) TET1, (B) TET2, and (C) TET3 over time. Data represent means and standard deviations of results from at least triplicate samples obtained from mock-infected or HHV-6B-infected Molt-3 cells. * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

Hydroxymethylation.

As the TET enzymes can oxidize methyl to hydroxymethyl but cytosines with any of these epigenetic marks are measured as 5-methylcytosine (5mC) using the methods described above, the amount of 5-hydroxymethylcytocine (5hmC) was measured in a separate analysis using triplicate HHV-6B-infected and mock-treated (here referred to as uninfected) samples at 3 dpi. HHV-6B-infected Molt-3 cells displayed significantly (P = 0.029 and 0.015, respectively) more hydroxymethylation at the two investigated CpG sites in the VPS53 gene than uninfected cells (results for cg20843650 are shown in Fig. 6). The CpG sites investigated in the RPH3AL and RFLNB genes did not exhibit significantly different levels of hydroxymethylation in infected and noninfected cells.

FIG 6.

Methylation and hydroxymethylation of cg20843650 (VPS53) at 3 dpi. Data represent means of results from triplicate samples obtained from mock-infected and HHV-6B-infected Molt-3 cells. 5mC = 5-methylcytosine, 5hmC = 5-hydroxymethylcytosine.

Hypomethylated genes show increased expression.

In order to determine the functional effect of the HHV-6B-induced methylation changes, we analyzed expression of the four genes located in the region with HHV-6B-induced hypomethylation at 17p13.3. Significantly higher gene expression of RPH3AL, RFLNB, and VPS53 was observed in HHV-6B-infected cells than in uninfected cells at 3 dpi (Fig. 7A). The expression of DOC2B can be regarded as extremely upregulated by HHV-6B as its expression was substantial in the virus-infected cells (threshold cycle [C_T] value, 33.9 ± 0.16) but could not be detected at all in uninfected cells (C_T value, >40). Unfortunately, the absence of signal in uninfected cells limited the calculation of its relative expression level or inclusion in statistics. If the C_T value is set to 40 in wells with cDNA levels outside the detection limit, the level of expression of DOC2B can be conservatively computed to be over 100 times greater in HHV-6B-infected cells than in uninfected cells.

FIG 7.

Relative gene expression of (A) RPH3AL, RFLNB, and VPS53 at 3 dpi (n = 5 + 5), with mock-infected cells used for calibration, or (B) VPS53 over time (n = 3 + 3), with untreated cells used for calibration. Means and standard deviations are shown. ** = P ≤ 0.01, **** = P ≤ 0.0001.

To further investigate the expressional increase over time, we chose to follow VPS53 expression in the same material used to investigate the DNA methylation of this gene over 6 days. As seen in Fig. 7B, the increased expression of VPS53 was already observable at 2 dpi with a pattern very similar to but opposite that of the decreased methylation of cg20843650 (Fig. 4). This indicates a strong correlation between the DNA methylation and expression of this gene.

The HHV-6B-induced hypomethylation may play a role in the integration process.

As 17p is a known integration site for HHV-6A and HHV-6B, the link between HHV-6B-induced hypomethylation and virus integration was investigated. Using PCR primers specific for HHV-6 DR_R and the 17p subtelomere, we could see evidence of integration. Comparing HHV-6B-infected cells and mock-infected cells at 6 dpi, amplification of a DNA fragment of anticipated size was visible only in the HHV-6B-infected cells (data not shown). To further investigate the timeline of integration, triplicate samples taken before infection and at 1, 2, 3, and 6 dpi were investigated. An amplified gene product of approximately 1.5 kb was found in HHV-6B-infected cells at 3 and 6 dpi in all of the triplicate samples and at 2 dpi in one of the samples (a representative photo of one infected cell culture analyzed over time is shown in Fig. 8B).

FIG 8.

PCR amplification of integration site. PCR was performed with primers binding to HHV-6B and the 17p telomere DNA; an ∼1.5-kb DNA fragment indicates virus integration. (A) Approximate location of primers used in integration PCR. Black = telomere, gray = subtelomere. DR_L—U—DR_R, virus genome (12). The figure is not in the correct scale. (B) Agarose gel electrophoresis investigating virus integration. Lanes 1 and 8, DNA ladders ranging from 75 to 20,000 bp (1.5-kb fragments are indicated with arrows); lane 2, uninfected cells (0 dpi); lanes 3 to 6, HHV-6B-infected Molt-3 cells at 1 to 6 dpi; lane 7, mock-infected cells at 6 dpi.

To further investigate the amplified PCR product, the DNA fragments were sequenced (for the results, see Table S2). The fragments sequenced from the HHV-6B direction (obtained by the HHV-6B DR_R primer) showed the highest identity to HHV-6B (strain Z29) DNA. Sequencing from the other direction (using the Chr 17p subtelomere primer) yielded a sequence ≥98% identical to that of DNA of chromosomes 7, 9, 11, 16, and 17 in all three virus samples. No sequence was identified in the mock-infected samples. These results strongly suggest chromosome 17p telomere as an integration site, but the possibility of integration into other telomeres cannot be ruled out.

DISCUSSION

We have for the first time shown that HHV-6B infection results in hypomethylated regions close to the telomeres, particularly 17p13.3, shortly after infection. The virus-induced DNA demethylation was most likely caused by the TET2 enzyme, which was shown to be upregulated by HHV-6B. However, further studies are needed to elucidate whether the expressional upregulation results in an increased protein level and causes the observed demethylation. In contrast to the HHV-6B effect, TET2 expression is repressed by EBV infection and, as expected, loss of TET2 increases EBV-induced host cell DNA methylation (31). However, TET2-induced hydroxymethylation of EBV promoters can also affect EBV reactivation from latency (32) and can affect the EBV latency type (33).

In order to investigate whether the increase in TET2 expression also affects the level of hydroxymethylation, we preformed oxidative bisulfite pyrosequencing to measure the portion of hydroxymethylation in three loci located in the VPS53, RPH3AL, and RFLNB genes at 3 dpi. In one of these loci (representing both CpG sites analyzed in the VPS53 gene), significant differences in hydroxymethylation were indeed detected. 5hmC has previously been regarded as a transient intermediate state between 5mC and unmethylated C, but increasing evidence points to 5hmC being a stable epigenetic modification participating in, e.g., activation of genes (32, 34) (reviewed in reference 24). The difference between the investigated loci is in itself interesting and may indicate a functional difference between the loci. If demethylation of 5mC occurs through modifying the 5mC first to become 5hmC and then to become further oxidized and replaced by unmethylated cytosines, all three hypomethylated loci must have had the 5hmC mark at one time point. This implies that the 5hmC modification had been further oxidized at 3 dpi in two of the three loci but that the investigated region in the VPS53 gene might have recruited proteins that stabilize this mark, possibly having a functional effect. The nature of HHV-6B-induced 5hmC needs further investigation, primarily at both earlier and later time points, to get the full picture.

The telomere region at 17p13.3 has repeatedly been described as a chromosomal integration site for both HHV-6A and HHV-6B (11, 14, 35–37), and the observed virus-induced hypomethylation in this region might be one mechanism by which the virus makes the host cell DNA more accessible to virus integration by manipulating chromatin structure. This needs further investigation, as this study did not investigate chromatin structure or histone modifications. However, hypomethylation of subtelomere DNA has been associated with increased telomere recombination frequencies (38), indicating that the hypomethylation induced by HHV-6B might affect the telomere integrity. In addition, telomeres containing chromosomally integrated HHV-6A/6B have been shown to be unstable (39). The data from our investigation of the possible connection between hypomethylation and virus integration in a T cell line infected in vitro suggest that the HHV-6B-induced hypomethylation at 17p13.3 may play a role in the process of integration of the virus into the telomere of 17p. However, the PCR assay could not specifically determine an exclusive site of integration into chromosome 17, and as integration into all separate telomeres was not investigated in this study, the possibility cannot be excluded that integration had occurred into a telomere without subtelomeric hypomethylation. Additional studies are needed to explore the connection between DNA methylation and virus integration in detail.

Similar region-specific hypomethylation during active infection observed in this study has, to our knowledge, not been reported for any virus before. Human cytomegalovirus (HCMV), a closely related betaherpesvirus, has been described to alter the host cell DNA methylation capacity by relocalization of DNMT from the nucleus to the cytosol, but no difference in the level of host cell DNA methylation could be observed at 3 dpi (40), indicating that the HHV-6B-induced hypomethylation is not a general antiviral response by the host cell. There are other indications that epigenetics plays a role in HHV-6A integration. The histone deacetylase inhibitor trichostatin A (TSA) has been shown to reactivate HHV-6A from latent integration (11). Reactivation of HHV-6A upon histone acetylation indicates that histone deacetylation is important in silencing of integrated viruses. We hypothesize that the subtelomeric hypomethylation during active infection observed in this study and the telomeric histone deacetylation during latency are connected, but that these epigenetic events play roles in different time windows: DNA hypomethylation during the integration process and histone deacetylation during latency when the host cell machinery represses the transcription of viral genes. Further investigations are needed to resolve these molecular mechanisms in detail.

HHV-6A was found to be able to integrate already into the 17p telomeric region of the JJhan cell line during productive infection (11), which is in line with the timeline observed in this study. However, the time period investigated here was short and the observed changes could have been transient, warranting performance of a long-term experiment to investigate the host DNA methylation also after the lytic phase of infection. In an attempt to investigate the effect of HHV-6B on DNA methylation in vivo, DNA extracted from 12 HHV-6B DNA-positive and 11 HHV-6B DNA-negative hippocampus samples from temporal lobe epilepsy (TLE) patients (selected from a larger cohort on which data were already published [41]) was investigated with the Illumina array described in Materials and Methods. No difference in methylation could be observed between the HHV-6B DNA-positive and -negative patient groups (data not shown). These negative results can be interpreted in different ways. They might indicate that hypomethylation is present only during the initial acute phase of the infection, and, as we suspect that the infection seen in brain of people with epilepsy is of a more chronic type, such hypomethylation might not be present at the time of biopsy. However, these negative findings could also be due to the low percentage of infected cells, to cell heterogeneity in brain biopsy specimens, or to the small number of included patients, representing a lack of adequate power to detect any differences. A larger cohort, in combination with methods such as single-cell analysis, would be needed to investigate the presence of HHV-6B-induced hypomethylation in epilepsy.

The region-specific epigenetic modification of the telomere at 17p identified in this study is interesting, and an obvious issue is that of what makes this site more prone to HHV-6A/6B integration than other sites in the genome. We hypothesize that the genomic sequence in this subtelomere/telomere is better suited for recombination with the telomeric repeats (TMRs) at the end of the viral direct repeat (DR) regions required for HHV-6A integration (19). This hypothesis is in line with findings by Tweedy et al. (42), as they were able to show that all in vivo HHV-6A integration sites at 17p contained TMRs distinct from integration sites at other chromosomes, possibly facilitating homologous recombination with virus TMRs. How this is connected to the HHV-6B-induced hypomethylation is still an open issue, but the data open up a new dimension in this research field.

It is also plausible that the increased expression of the studied hypomethylated genes is due to an ability of HHV-6B to specifically target these genes and that they might be needed for virus survival and pathogenesis in the host cell. Vps53 is one of four proteins in the Golgi-associated retrograde protein (GARP) complex. GARP is important for successful transfer of biomolecules between organelles, more specifically, for the reception of endosome-derived retrograde transport carriers at the trans-Golgi network (reviewed in reference 43). Both Noc2 (encoded by RPH3AL) (44) and Doc2b (reviewed in reference 45) are involved in calcium-triggered exocytosis. RefilinB is an actin regulatory protein that, together with actin-binding filamin A, organizes the perinuclear actin network (46). All four of these proteins could potentially be needed by the virus for its transport within the cell and exocytosis of new virus particles. However, since this study focused on DNA methylation and its effect on gene expression but did not investigate protein expression or phenotypic changes, we cannot draw conclusions regarding protein effects.

The reported findings observed in this study were obtained using a cell line, and even though the HHV-6B pathogenesis in this T cell line resembles that in primary T cells, the epigenetic landscape is not the same in cell lines as in primary cells. Continued investigation of both HHV-6A and HHV-6B in different primary cell types might reveal different patterns that would be more generalizable. The epigenetic modifications investigated here are not exhaustive, and additional investigations of the effect of HHV-6B/6A on, for example, histone modifications might reveal additional strategies of virus-induced regulation of the host cell. Although there are questions left to be answered, our study was the first to report that HHV-6B specifically induces hypomethylated regions close to the telomeres and that integrating viruses might use the host methylation machinery in their integration process. This, in turn, is key to understanding the latency of these types of viruses, which may be modifiable to prevent reactivation and health complications.

MATERIALS AND METHODS

Cell and virus culture.

HHV-6B strain Z29 was propagated in the Molt-3 T-cell line. Cell-free supernatant was obtained by centrifugation for 15 min at 280 × g when the cytopathic effect in the cell culture was >50%, and aliquots were stored at −80°C. Cell-free supernatant from uninfected Molt-3 cells was harvested and used for the mock infections. The 50% tissue culture infective dose (TCID₅₀) was determined by quantitative PCR (qPCR) readout as previously described (47) using HHV-6B-specific primers (48).

Molt-3 cells were incubated under mock infection conditions or with 1.4 × 10⁴ TCID₅₀ of HHV-6B/10⁶ cells for 3 h before the medium was changed to fresh RPMI medium supplemented with GlutaMAX (Gibco), 1% PEST (Gibco), and 10% fetal bovine serum (FBS) (Gibco) and were further cultured in the presence of 5% CO₂ at 37°C. Three separate experiments were started on different days, each with triplicate cultures for both conditions.

All cell cultures were stained for viral protein p41 (Santa Cruz Biotechnology) at 3 days postinfection (dpi), yielding >95% positive cells in HHV-6B cultures but 0% positive cells in mock-treated (here referred to as uninfected) cultures.

Genome-wide methylation analysis.

DNA was extracted with a QIAamp DNA minikit (Qiagen) from three uninfected and three HHV-6B-infected cell cultures obtained at 3 dpi according to the manufacturer's protocol. Methylation was investigated for >485,000 CpG sites with an Infinium HumanMethylation450 BeadChip array according to the guidelines of the manufacturer (Illumina). The BeadChips were scanned in an Illumina iScan system and then preprocessed by the use of Illumina GenomeStudio software (no background subtraction or normalization in Genome Studio). The quality control (QC) and color balance check was passed. The raw data were processed using the Bioconductor lumi package (49), where color channel bias was corrected using a within-array smooth quantile normalization. A between-array quantile normalization was then performed, after which beta and M values were calculated. Annotations were made against the genome (version GRCh37/hg19), and this version was used for obtaining chromosome lengths and for investigating the data.

Bisulfite pyrosequencing.

To technically validate the findings observed with the Illumina 450K array and to further investigate the methylation pattern over time, bisulfite pyrosequencing was performed over selected loci (Table 1). Primers were designed in PyroMark Assay Design 2.0 (Qiagen) and were optimized for the best annealing temperature.

TABLE 1.

Primers for bisulfite pyrosequencing^a

Annotated gene	Annealing temp (°C)	Primer type	Primer sequence
RPH3AL	57	Forward	GGGTGAGATTATTTTTATGGTGAGAA
		Reverse	CACCCCTAATATAAAAAATACACTTAAACC*
		Sequencing	GAGGGGTTGGGGGAA
RFLNB (FAM101B)	57	Forward	GGGATTGGTAGGAGGTTTTAAGAA
		Reverse	ACACCAAACAAACTAAAATACATTCA*
		Sequencing	GAGGTAGGTTTGAAGT
VPS53 (1)	56	Forward	GGAGTAAAGGGTTTTTTGTTGTTGTTAT*
		Reverse	ACCTCTACTAACCCCATATCCAAATAC
		Sequencing	CCATTCTAACCCAAACTT
VPS53 (2)	56	Forward	TTGTGAGGATGTTGGTGGGAAGTA
		Reverse	CCCCAACATACACAATCTTCCCATTA*
		Sequencing	TGTTTAAATAGAGGGAATGTTTAG

^aThe concentration used for the forward and reverse primers was 0.2 μM, and the concentration used for the sequencing primer in the wells was 0.4 μM. *, biotinylated primer. (1) and (2) are two different primer pairs for the same gene.

DNA was extracted with an All Prep DNA/RNA minikit (Qiagen) from five untreated Molt-3 cells (0 dpi) and triplicate mock-infected or HHV-6B-infected Molt-3 cells at 0.5 dpi, 1 dpi, 2 dpi, 3 dpi, and 6 dpi. DNA was subjected to bisulfite conversion using an EZ DNA Methylation-Gold kit (Zymo Research) according to the manufacturer's protocol, and 20 ng DNA/reaction was used for PCR amplification of selected regions (Table 1) using a PyroMark PCR kit (Qiagen) according to the manufacturer's protocol. Pyrosequencing was performed on a PyroMark Q96 ID machine (Qiagen) using PyroMark Gold Q96 reagents (Qiagen) according to the manufacturer's protocol. All assays were validated for biases with amplified commercial DNA with 0 to 100% methylation.

Analysis of locus-specific DNA hydroxymethylation.

Both the Illumina array and the bisulfite pyrosequencing described above cannot distinguish between methylated and hydroxymethylated cytosines, and CpG sites with either of these two modifications are reported as methylated. In order to separate 5-hydroxymethylcytosine (5hmC) from 5-methylcytosine (5mC), DNAs from the triplicate mock-infected and virus-infected 3-dpi samples used in the Illumina array were both oxidized and subjected to bisulfite conversion using a TrueMethyl Seq kit (CEGX, Cambridge, United Kingdom) according to the protocols of the manufacturers. After conversion, these samples were subjected to pyrosequencing as described above, except that VPS53 primer pair 2 was excluded (Table 1). The amount of 5hmC present was calculated by subtracting the level of CpG “methylation” in the oxidized fraction from the level in the bisulfite-converted fraction in each sample.

Gene expression.

Total RNA was extracted with an RNeasy minikit or an All Prep DNA/RNA minikit (Qiagen) with the addition of DNase treatment (Qiagen), after which the RNA was transcribed into cDNA using a High-Capacity cDNA reverse transcription kit (Applied Biosystems). Gene expression was investigated according to the manufacturer's protocol in triplicate wells, where each well contained 20 ng cDNA and TaqMan Universal PCR master mix, and TaqMan gene expression assays (Applied Biosystems) were performed for peptidylprolyl isomerase A (PPIA; Hs99999904_m1); TET1 (Hs04189344_g1); TET2 (Hs00325999_m1); TET3 (Hs00896441_m1); rabphilin 3A-like (without C2 domains) (RPH3AL; Hs00383179_m1); VPS53, GARP complex subunit (VPS53; Hs00217606_m1); double C2 domain beta (DOC2B; Hs00186064_m1); or refilin B (RFLNB [former name, FAM101B]; Hs00823804_m1). A “no-reverse-transcription” control and a “no-template” control were included in triplicate wells and gave no signal.

Expression of TET enzymes was analyzed in at least triplicate samples obtained on days 0, 1, 2, 3, and 6. Five mock-infected and five HHV-6B-infected (3-dpi) samples were analyzed for VPS53, RPH3AL, and RFLNB expression, whereas expression of VPS53 was further investigated over time in a separate experiment using new triplicate samples obtained on days 0, 0.5, 1, 2, 3, and 6. PPIA was used as the reference gene, as this gene can be considered stably expressed during HHV-6B infection (50). Relative expression was calculated with the 2^−ΔΔCT method (51) using the average ΔC_T value for untreated or mock-treated (3 dpi) samples as the calibrator.

Integration.

To investigate HHV-6B integration into the telomeric region of chromosome 17p, PCR primers were used to amplify a 1.5-kb DNA fragment consisting of the virus/chromosome fusion (12). One primer bound to HHV-6B DR_R (5′-CATAGATCGGGACTGCTTGAAAGCGC-3′) and the other to the subtelomere of chromosome 17p (5′-AACATCGAATCCACGGATTGCTTTGTGTAC-3′). The purified DNA used for methylation studies was also used in this experiment: 100 ng DNA was amplified in 20-μl reaction mixtures containing 0.5 μM primers and Platinum Green Hot Start PCR master mix (Invitrogen). The PCR program was initiated by a 94°C denaturation step for 2 min followed by 35 cycles of 94°C denaturation for 30 s, 60°C annealing for 30 s, and 72°C extension for 90 s. Agarose gel electrophoresis was used for visualization of DNA fragments as follows: 7 μl amplified product and 4 μl GeneRuler 1-kb Plus DNA Ladder (Thermo Scientific) were loaded onto a 1.3% agarose-GelRed gel, and bands were detected with UV light.

To further investigate the amplified product, the PCR products from the 6-dpi samples were sequenced. For this analysis, 500 ng DNA was used as start material in the PCR and the annealing temperature was increased to 62°C. The same primers as those used for PCR amplification were used for sequencing; i.e., the amplified products were sequenced from both ends. To remove the primers left in the PCR tube, the samples were treated with ExoSAP-IT PCR product cleanup reagent (ThermoFisher Scientific) before the start of sequencing. Sequencing was performed using a BigDye Terminator v3.1 cycle sequencing kit and an ABI 3730 Prism DNA analyzer, according to manufacturer's guidelines (Applied Biosystems). Chromas 2.6.2 (Technelysium) was used to analyze the sequencing data, and the obtained sequences were matched to published DNA sequence information using the nucleotide BLAST function on the NCBI webpage.

Statistics.

Empirical Bayes-moderated t statistics, as implemented in the Bioconductor package limma (52), was used to find differently methylated probes in the Illumina arrays comparing M values between samples. The Benjamini-Hochberg adjustment was applied to correct for multiple testing. A false-discovery rate (FDR, adjusted P value) of <0.05 was regarded as statistically significant. For comparisons of only two groups, the t test was used in Excel and the threshold for significance was set at a P value of <0.05.

A Manhattan plot and a volcano plot were made in R software using the qqman and ggplot package, while all other graphs were made in GraphPad Prism 5.0.