Points of Recombination in Epstein-Barr Virus (EBV) Strain P3HR-1-Derived Heterogeneous DNA as Indexes to EBV DNA Recombinogenic Events In Vivo

Abstract

Deletions and rearrangements in the genome of Epstein-Barr virus (EBV) strain P3HR-1 generate subgenomic infectious particles that, unlike defective interfering particles in other viral systems, enhance rather than restrict EBV replication in vitro. Reports of comparable heterogeneous (het) DNA in EBV-linked human diseases, based on detection of an abnormal juxtaposition of EBV DNA fragments BamHI W and BamHI Z that disrupts viral latency, prompted us to determine at the nucleotide level all remaining recombination joints formed by the four constituent segments of P3HR-1-derived het DNA. Guided by endonuclease restriction maps, we chose PCR primer pairs that approximated and framed junctions creating the unique BamHI M/B1 and E/S fusion fragments. Sequencing of PCR products revealed points of recombination that lacked regions of extensive homology between constituent fragments. Identical recombination junctions were detected by PCR in EBV-positive salivary samples from human immunodeficiency virus-infected donors, although the W/Z rearrangement that induces EBV reactivation was frequently found in the absence of the other two. In vitro infection of lymphoid cells similarly indicated that not all three het DNA rearrangements need to reside on a composite molecule. These results connote a precision in the recombination process that dictates both composition and regulation of gene segments altered by genomic rearrangement. Moreover, the apparent frequency of het DNA at sites of EBV replication in vivo is consistent with a likely contribution to the pathogenesis of EBV reactivation.

As a ubiquitous human tumor virus linked to diverse pathologies of epithelial, hematopoietic, and mesenchymal origins, Epstein-Barr virus (EBV) offers the paradox of being typically acquired, maintained, and shed asymptomatically. The natural history of EBV infection is one of rapport with its host, from the establishment of latency upon infection of naïve B cells to persistence in memory B cells to release of infectious virions upon B cell maturation into plasma cells (26). Distinct from its natural life cycle is the apparent burst of lytic replication reflected in the unusual antibody profile preceding the onsets of EBV-associated cancers. Extreme elevations of titers of antibodies to viral proteins of the lytic cycle precede by months to years the onsets of Burkitt's lymphoma (BL), Hodgkin's lymphoma, gastric carcinoma, and nasopharyngeal carcinoma, bearing sufficient prognostic value in the last case as to provide a useful cancer screening tool in south China, where nasopharyngeal carcinoma is endemic (12, 27, 30, 31, 41, 47).

As a portent of disease, this reactivation process is likely to differ substantially from intermittent EBV shedding in lifelong carriers, which accompanies physiological cell differentiation (26) but fails to engender a like humoral immune response. The uniquely permissive EBV-positive BL cell line P3HR-1 offers a case in point, generating a subgenomic infectious particle that reactivates the standard viral genome from a state of latency (28, 29, 38, 40). Its components have been termed heterogeneous (het) DNA, which is characterized by endonuclease digestion fragments varying in both size and homology relative to the standard viral genome (3, 8, 14, 15, 29).

The deleted, rearranged DNA is composed of four noncontiguous regions of EBV DNA joined by three intramolecular recombination events (Fig. 1A), which is the equivalent of approximately one-third of the standard genome (5, 6, 19, 29). Although a general consensus on the abnormally juxtaposed DNA segments making up the novel restriction fragments in het DNA exists, their detection in a variety of configurations, including palindromes, has suggested that there may in fact be a family of defective particles (5, 14, 19, 20). Defective genomes contain origins of replication, terminal sequences required for packaging, and open reading frames encoding transactivators of lytic cycle genes (6, 19) yet still depend on helper functions from standard virus. In contrast to the standard viral genome, which is vertically partitioned as a plasmid, het DNA is not stably maintained in cells (29). Defective virions are generally thought to be independent replicons that are maintained by cell-to-cell spread (28).

FIG. 1.

Points of recombination in P3HR1-derived het DNA. (A) BamHI restriction map of standard P3HR-1 genome adapted from Cho et al. (5) and Rabson et al. (37). Boxes 1 to 4 approximate regions comprising defective het DNA, with arrows indicating orientation in standard genome versus that in het DNA (see panel D). (B) Ethidium bromide gel showing PCR amplicons spanning each junction (W/Z, E/S, and M/B1) of rearranged het DNA. BHRF1 is an EBV gene present in the standard genome only. P3HR-1 clone 5 contains het DNA and standard virus; P3HR-1 clone 13 and BL line Namalwa (NMW) contain standard EBV genomes; BL-2 is an EBV-negative BL line. (C) Sequences at points of recombination in het DNA fragments. Bold, underlined type indicates sequences in W/Z, E/S, and M/B1 after recombination; boxed residues indicate points of recombination at base pairs which are shared by both BamHI fragments and which may be contributed by either one; nonbolded type indicates residues excluded from the het DNA recombination product. Numerical superscripts identify residues specific for EBV type 1 (2, 34) (NCBI GenBank accession number AJ507799) or EBV type 2 (9) (NCBI GenBank accession number DQ279927), P3HR-1 being an EBV type 2 strain: 1, C in type 1 and A in type 2; 2, C in type 1 and T in type 2; 3, A in type 1 and C in type 2; 4, C in type 1 and T in type 2; 5, G in both type 1 and type 2; 6, A in type 1 and G in type 2. The sequence of fusion junction in WZhet DNA determined here by PCR methodology is identical to that previously reported by Jenson and Miller (18). (D) Orientation of het DNA components (adapted from Cho et al. [5]) in relation to the standard genome shown in panel A, with the map coordinates (GenBank accession number AJ507799) at each recombination point. Coordinate 32786 in the seventh BamHI W repeat was arbitrarily chosen, an identical sequence being located at coordinate 29714, 26642, 23570, 20498, 17426, or 14254. The BamHI B1 restriction fragment of P3HR1 strain is encompassed by the BamHI I segment of the AJ507799 sequence. (E) Schematic diagram showing distribution and alteration of EBV open reading frames in het DNA. Standard EBV ORF nomenclature is shown despite reorientation of DNA segments 2 and 3 (e.g., the leftward reading frame BZLF1 is now rightward reading).

The unusual biological property of disrupting viral latency has been attributed to positive regulatory elements in the BamHI W fragment of EBV DNA newly positioned upstream of BZLF1 (BamHI Z leftward reading frame 1), an immediate-early gene which in the standard genome is located more than 55 kb away and in an inverse orientation (18, 40). Now constitutively expressed, the BZLF1-encoded product (termed ZEBRA, Zta, or EB1) transactivates the lytic origin of EBV DNA replication as well as EBV lytic cycle gene promoter elements. On the basis of sequence data defining the point of recombination in WZhet DNA from a single-cell clone of P3HR-1 (18), similar recombinants have been identified by PCR analysis in oral hairy leukoplakia of AIDS, the single known pathological manifestation of EBV replication (35). Subsequent studies linked a WZhet fragment to individual cases of BL (39), Hodgkin's lymphoma (10), thymic carcinoma (36), and the productive infection accompanying idiopathic pulmonary fibrosis in immunocompetent patients (21).

The presence of WZhet DNA in human lesions, together with the unique biology manifest by that rearrangement in vitro, led us to consider the pathogenic potential implicit in the remaining two recombination sites. Here, we defined at the nucleotide level the points of recombination between BamHI M and BamHI B1 (M/B1) and between BamHI E and BamHI S (E/S) in het DNA found in cell clone 5 of the P3HR-1 cell line and verified their presence in clinical materials.

MATERIALS AND METHODS

Cell lines and virus preparations.

EBV-infected, BL-derived P3HR-1 cell clone 5 (het DNA positive) and clone 13 (het DNA negative) were gifts from G. Miller (Yale University). BL2 is an EBV-negative BL cell line. Raji, Namalwa, and B95-8 are EBV-positive B-cell lines that maintain virus in latent, integrated, and partially replicative states, respectively. P3HR-1 clone 5 cells were maintained in RPMI 1640 medium (CellGro) supplemented with 100 μg/ml penicillin-streptomycin (CellGro) and 7% fetal bovine serum (HyClone) and incubated at 34°C in 5% CO₂. Other cell lines were similarly maintained, but in 10% fetal bovine serum at 37°C. Virus stocks were obtained by treatment of cells with 20 ng/ml tetradecanoylphorbol-13-acetate. Culture supernatants were concentrated 20× using a Labscale TFF system (Millipore) and viral stocks stored at −80°C. A common virus stock was used for all experiments.

Salivary samples.

Oral specimens were collected with an absorbent cotton salivette (Sardstat) placed sublingually after informed donor consent was gained under an Institutional Review Board-approved protocol of the AIDS Clinical Trials Unit of Harborview Medical Center, Seattle, WA. The cotton pledget was vortexed in 5 to 10 ml of phosphate-buffered saline and a cell pellet obtained by centrifugation. DNA was harvested from the cell pellet with a QIAamp DNA blood mini kit (Qiagen).

DNA PCR and quantitative reverse transcription-PCR (qRT-PCR).

The PCR primer sets used successfully to amplify points of recombination in het DNA are listed in Table 1. DNA PCR was performed for 35 cycles (94°C for 30 s, 55°C [BHRF1, W/Z, and M/B1] or 52°C [E/S] for 1 min, and 72°C for 1 min) in a 50-μl reaction mixture containing 0.6 μM primers, 0.2 mM deoxynucleoside triphosphates (Promega), and 0.75 U Taq DNA polymerase in the manufacturer's buffer (Promega) with a PTC-200 thermal cycler (MJ Research). PCR products were electrophoresed on a 2% agarose gel with 1× Tris-acetate-EDTA. Gels were denatured with 0.5 M NaOH-1.5 M NaCl and then neutralized with 0.5 M Tris-HCl (pH 7)-1.5 M NaCl. DNA was transferred by Southern blotting to nylon membranes (GE Osmonics), cross-linked by UV (Bio-Rad Laboratories), and hybridized with oligonucleotide probes end labeled with a [γ-³²P]ATP by T4 polynucleotide kinase (USB). Limiting-dilution DNA PCR was performed using 10-fold dilutions of total cellular DNA, virion DNA, or plasmids containing cloned junctions of het DNA fragments. For PCR of virion DNA derived from P3HR-1 clone 5, 150 μl of concentrated cell-free virus was incubated at 37°C for 30 min with 5 U DNase (Promega). DNase was inactivated with 20 μl of stop solution at 65°C for 15 min. DNA was harvested with a QIAamp DNA blood mini kit (Qiagen) and 2 μl used in the initial dilution of the template for the PCRs.

TABLE 1.

Sequences of primers and probes for PCR analysis of P3HR-1 het DNA^a

Target	Sequence (5′→3′)	GenBank accession no. (coordinates)
Defective EBV genome
W/Z fragmentb (product = 269 bp)
W (5′ primer)	GTCCAGCGCGTTTACGTAAG	M20820 (1649-1630)
Z (3′ primer)	GCACATTAGCAATGCCTGTG	M20820 (1381-1400)
Probe	CATGCAGCAGACATTCATCATTTAGAAATG	M20820 (1498-1527)
E/S fragment (product = 447 bp)
e1c (5′ primer, type 1)	ACTTAAATCCAGAGGAGCGT	AJ507799 (88352-88333)
e1 (5′ primer, type 2)	ACTCAAATCCAGAGGAGTGT	DQ279927 (89004-88985)
S (3′ primer)	ATGGCAGAGGAAGACCGGTA	AJ507799 (73313-73332)
E probe (type 1)	TCACACTTGAGTTCCATGTCGCCCACACAG	AJ507799 (88263-88292)
E probe (type 2)	TCACACTTGAGCTCCATGTCGCCCACGCAG	DQ279927 (88915-88944)
M/B1 fragment (product = 260 bp)
M (5′ primer)	ACGGTGATTAGGCACTGGAC	AJ507799 (70206-70187)
B1d (3′ primer)	TCGGGGCATTGGTGTAACAA	AJ507799 (140729-140710)
Probe	GCGGTTATAACTGCCAGAGC	AJ507799 (70096-70077)
Standard EBV genome
BHRF1 (product = 240 bp)
5′ primer	GTGCATGGAAATGGTA	AJ507799 (42142-42157)
3′ primer	AAGGCTTGGGTCTCC	AJ507799 (42381-42367)
Probe	AAACACCTCTCCGCCTTTCG	AJ507999 (42191-42210)

^aThe P3HR-1 strain is a type 2 EBV. BamHI E and BamHI e1 sequences vary between type 1 (GenBank accession number AJ507799) and type 2 (GenBank accession number DQ279927) EBVs.

^bThe point of recombination in WZhet was previously reported (18).

^cThe BamHI e1 fragment (GenBank accession number AJ50799) is termed BamHI c′ in the restriction map of P3HR-1 (Fig. 1A), a deletion mutant of Jijoye (37).

^dBamHI B1 in deletion variant P3HR-1 (Fig. 1A) is equivalent to BamHI I in the type 1 (GenBank accession number AJ507799) and type 2 (GenBank accession number DQ279927) sequences.

For donor salivary cells, DNA amplification reactions employed 1 μg DNA. Controls included BL-2 (EBV-negative) cells, P3HR-1 cell clone 13 (het DNA negative/standard EBV positive), buffer only (template negative), and P3HR-1 cell clone 5 (het DNA positive/standard EBV positive). All samples yielding PCR products were confirmed as positive in a separate amplification reaction; negative samples were verified by increasing the PCR cycle number (from 35 to 45), varying the quantity of template DNA (from 1 μg to 2 μg), or using nested PCR.

The transcript levels of EBV genes BZLF1 and BSMLF1 (SM) in acutely infected BL2 cells were determined by qRT-PCR with a Bio-Rad iCycler, using Sybr green chemistry. In brief, RNA was extracted according to the RNA STAT60 protocol (Tel-test). For cDNA synthesis, 5 μg of RNA was treated with 5 U RQ1 RNase-free DNase (Promega) and converted to cDNA by first-strand synthesis using random hexamer primers and Moloney murine leukemia virus reverse transcriptase (Invitrogen). Fifty nanograms of cDNA was amplified using the following exon-specific primer pairs at the given concentrations: for SM, 5′AACGAGGATCCCGCAGAGA3′ (5′ coordinate 71955; GenBank accession number AJ507799; 200 nM) and 5′ATCGCAGTCTGTGTTGGTGTCT (5′ coordinate 71781; 200 nM); for BZLF1, 5′CCGGCACGACGCACA3′ (5′ coordinate 90405; 600 nM) and 5′TTATTTCTAGTTCAGAATCGCATTCC3′ (5′ coordinate 90212; 600 nM); and for GAPDH (glyceraldehyde-3-phosphate dehydrogenase), 5′GAAGGTGAAGGTCGGAGT3′ (5′ coordinate 108; GenBank accession number NM_002046; 400 nM) and 5′GAAGATGGTGATGGGATTTC3′ (5′ coordinate 333; 400 nM). Standard two-step cycling conditions were used with a 60°C annealing temperature. The comparative cycle threshold method was used to assay differences in transcript levels, with cycle threshold values determined by automated threshold analysis. All samples were run in triplicate, together with reactions to quantify expression of an internal control gene, the human GAPDH gene. Uninfected BL2 cells and reverse transcriptase-negative samples served as negative controls.

Cloning and sequencing PCR products.

PCR products were purified from gels with a QIAquick gel extraction kit (Qiagen). For efficient ligation, a poly(A) tail was added by incubating 7 μl of a purified PCR product with 1 μl of 2 mM dATP and 2 U Taq DNA polymerase (Promega) for 30 min at 70°C. The product was ligated into pCR2.1 (TA-cloning kit; Invitrogen) and transformed into one-shot INVαF′ competent cells per the manufacturer's protocol (Invitrogen). Bacterial colonies were transferred to membranes that were sodium dodecyl sulfate treated, denatured, and hybridized to oligonucleotide probes (Table 1). Plasmid DNA in three positive colonies from each sample was harvested using a QIAprep spin mini prep kit (Qiagen), and the sequence was determined by fluorescent dye-terminator sequencing (SeqWright DNA Technology). The coordinates for the M/B1 and E/S points of recombination were determined from the wild-type EBV genomic sequence data (GenBank accession number AJ507799) (2, 34).

Gardella gel system.

To separate circular, linear, concatemeric, and other replicative intermediate forms of EBV DNA occurring over an acute time course postinfection, BL-2 cells infected in vitro were embedded in agarose plugs and run on Gardella gels as previously described (11), but with modifications. Briefly, 10³ to 10⁵ cells were resuspended in PBS at fixed time points postinfection, mixed with equal volumes of 1% low-melting-temperature SeaPlaque agarose (FMC), and transferred to a disposable mold (Bio-Rad Laboratories). Solidified plugs were incubated overnight at 50°C in 500 μl lysis buffer (0.5 M EDTA [pH 8.5], 1% Sarcosyl, 0.5 mg/ml proteinase K) and stored at 4°C. Plugs were loaded into wells of a horizontal 0.8% SeaPlaque agarose gel and electrophoresed in 0.5× Tris-borate-EDTA at 200 V for 15 h at 4°C. For gel lanes analyzed by Southern blotting, DNA was transferred to nylon membranes, cross-linked by UV, and hybridized with a randomly [α-³²P]dCTP-labeled BamHI M fragment of EBV DNA as a probe. For gel segments used in conjunction with DNA PCR (7), 13 equivalent gel slices were obtained from the length of each lane, perpendicular to the direction of electrophoresis. Each gel segment was melted at 65°C, and 1 μl of the molten agarose was used for PCR. Three separate reactions were performed on each slice for the W/Z, M/B1, and E/S junctions. The PCR products were run on a 2% agarose gel and Southern blotted onto a nylon membrane that was hybridized with oligonucleotide probes 5′ end labeled with [γ-³²P]ATP. As a control for EBV genome configuration, similar analyses were performed on the established cell lines Raji (episomes) and B95-8 (episomes and linear forms) as well as on purified virion DNA (linear forms) from B95-8.

Nucleotide sequence accession numbers.

The sequence data defining points of recombination in het DNA fragments E/S and M/B1 have been assigned GenBank accession numbers EU669823 and EU669824, respectively.

RESULTS

Minimal homology at points of recombination in P3HR-1 het DNA.

Guided by endonuclease restriction maps of P3HR-1 het DNA (5, 6, 19, 20), we empirically chose sets of PCR primers to span the presumed locations of the junctions creating the unique BamHI M/B1 and E/S recombination fragments of het DNA. BamHI B1 of the P3HR-1 laboratory strain is equivalent to BamHI I in the wild-type genome and contains the region deleted in B95-8 (2, 34). The primer sets yielding products from het DNA-positive P3HR-1 clone 5 but not het DNA-negative clone 13 (Fig. 1B) are listed in Table 1. PCR products were cloned into a TA vector and sequenced to identify points of recombination at the nucleotide level. Junctions between truncated BamHI E and S fragments and BamHI M and B1 fragments contained two or three shared residues, respectively, that could have been derived from either fragment, but with no areas of extensive homology (Fig. 1C). The nucleotide sequence of WZhet, defined previously by Jenson and Miller from a cloned WZhet fragment (18), matches exactly our junctional sequence obtained from P3HR-1 cellular clone 5 by PCR.

EBV sequence coordinates at points of recombination were determined in reference to the entire wild-type EBV genome (GenBank accession number AJ507799) (2, 34) (Fig. 1D). The E/S recombination truncated the 5′ portions of two open reading frames, BERF4 and BSLF1, contained in respective BamHI fragments and transcribed in opposing directions, their 3′ ends now fused back to back (Fig. 1E). The M/B1 fusion joined BamHI M of the standard P3HR-1 genome to BamHI B1 at a point between open reading frames BMLF1 and BMRF2, leaving intact in het DNA the region (BSLF2 and BMLF1) that encodes the SM lytic-cycle early protein. Likewise, BamHI B1 was fused just 3′ to the polyadenylation signal of LF3 (leftward reading frame 3), an EBV early gene with a putative role in lytic DNA replication (45). LF3 is positioned adjacent to a lytic origin of replication (termed D_R) also retained in het DNA (Fig. 1E).

Detection of comparable het DNA fusion fragments in clinical samples.

As a first step toward assessing a potential role for het DNA in the biology of EBV reactivation in vivo, we determined if junctions newly defined in P3HR-1 het DNA could also be detected at sites of replication in human infections. We applied the PCR analysis outlined above to salivary specimens obtained on a single occasion from 20 human immunodeficiency virus (HIV)-positive individuals known to be susceptible to mucosal lesions characterized by exuberant EBV replication and high levels of recombinogenic activity (43). Of the 19 donors positive for standard EBV, two contained recombinants manifesting not only the WZhet rearrangement previously found in clinical materials but also the two newly defined points of recombination in E/S and M/B1. An additional two donors were positive only for the WZhet rearrangement and two others for just the M/B1 junction (Table 2). Cloning and sequencing of PCR products confirmed that the points of recombination in the clinical samples were identical to those found in het DNA fragments from the P3HR-1 virus strain, each junctional sequence being ascertained from three to six clones of amplified product per patient and in two or more patients (Tables 2 and 3).

TABLE 2.

Presence of P3HR-1 het DNA-defined points of recombination in salivary samples from HIV-positive patients

Patient	Presence of:
	Standard EBVa	W/Z	E/S	M/B1
A	+	−	−	−
B	+	+	−	−
C	+	−	−	+
D	+	−	−	−
E	+	−	−	+
F	+	+	+	+
G	+	−	−	−
H	+	+	−	−
I	+	+	+	+
J	+	−	−	−
K	+	−	−	−
L	+	−	−	−
M	+	−	−	−
N	+	−	−	−
O	+	−	−	−
P	+	−	−	−
Q	+	−	−	−
R	+	−	−	−
S	−	−	−	−
T	+	−	−	−

^aStandard EBV genome detected by PCR for the BHRF1 sequence that is absent in het DNA.

TABLE 3.

Persistence of het DNA in oral samples

Patient no.	No. of days from start of study	Presence of:
		Standard EBV	W/Z	E/S	M/B1
1	0	+	+	−	−
	138	+	+	−	−
	222	+	+	+	−
	309	+	+	−	−
2	0	+	+	−	−
	83	+	+	−	−
	148	+	+	+	−
3	0	+	+	−	−
	35	+	+	−	−
	70	+	+	−	−
4	0	+	+	−	−
	35	+	+	−	−
5	0	+	+	−	−
	35	+	+	−	−
	78	+	+	−	−
	113	+	+	−	−
	148	+	−	−	−
	175	+	+	−	−
	218	+	+	−	−
	253	+	−	−	−
6	0	+	−	−	−
	29	+	−	−	−
	35	+	−	−	−
	64	+	+	−	−
	94	+	+	−	−
	129	+	+	−	−
	213	+	+	−	−
	319	+	+	−	−
7	0	+	+	−	−
	84	+	+	−	−
	185	+	+	−	−
	266	+	−	−	−
	352	+	+	−	−

Several lines of evidence argue against the possibility of contamination of clinical material with PCR products and/or cloned material derived from P3HR-1 as an explanation for the sequence identity at points of recombination in het DNA. First, PCRs have been performed at two institutional sites by separate investigators (K. Ikuta and S. K. Srinivas), the PCR results were reproducible for every clinical sample tested, and the reactions were run concurrently with the controls (buffer only [DNA template negative], clone 13 [het DNA negative/standard genome positive], clone 5 [het DNA positive/standard genome positive], and EBV-negative BL2 cells) that gave the appropriate outcomes. Second, single-nucleotide polymorphisms verified by separate PCR amplifications distinguished clinical from P3HR-1-derived het DNA in several donors (data not shown). Third, in select cases PCR results were confirmed with a second set of primers, targeting sequences outside the region originally amplified in P3HR-1 het DNA (not shown).

Persistence of WZhet recombinants at sites of EBV replication in vivo.

To determine if het rearrangements could be consistently detected in successive oral samples over periods of up to 1 year, 33 specimens were evaluated from an additional seven donors selected on the basis of WZhet-positive samples (Table 3). In the analysis of multiple samplings from single donors, the WZhet rearrangement overwhelmingly predominated, with 27 of the 33 samples found positive. One sample from each of two separate donors subsequently contained a second rearrangement (E/S), detected only by a nested PCR approach. Because the primer sets also recognize sequences in the standard genome but exponentially amplify only rearranged or deleted het DNA, limiting dilutions of total cellular DNA from P3HR-1 clone 5 were used as templates to assess whether the failure to detect all junctions in clinical samples might reflect various degrees of competition for primer sets by the standard genome. P3HR-1 cells contain from 30 to several hundred EBV genome equivalents per cell. In contrast, only 1 out of approximately 200 cells is het DNA positive (10, 38). Despite the high level of standard-viral-genome background, each het primer set the amplified targeted sequences with similar efficiency (Fig. 2A).

FIG. 2.

Packaging of het fusion junctions into viral particles. (A) The efficiencies of the PCR primer sets targeting the het junctions are comparable, despite the abundance of the standard genome, as shown by a Southern blot of PCR products, using total cellular DNA from P3HR-1 clone 5 (het DNA positive) as a template. The starting dilution contained 100 ng of total cellular DNA. BHRF1 is an EBV gene present in standard virus but absent from het DNA. (B) PCR products from 10-fold dilutions of DNA obtained from DNase-treated virions in concentrated culture supernatant from the P3HR-1 clone 5 cells shown in panel A. The starting dilution contained 150 ng virion DNA.

Proportionate representation of het DNA fragments in virus preparations from P3HR-1 supernatants.

The exact organization of het DNA fragments in defective particles is likely to be variable, as evidenced by in vitro studies (5, 19, 25). The prevalence in clinical material of WZhet rearrangements alone, for example, suggests that individual recombinations may exist in isolation. Alternatively, additional defective particles that contain rearrangements other than those defined by laboratory strain P3HR-1-derived het DNA may exist in vivo. Because some single-cell subclones of P3HR-1 clone 5 have also been reported to contain less than the full measure of het DNA fragments (25), we questioned whether het DNA packaged into cell-free infectious particles contained equal representations of all junctional fragments. Virus stock prepared from P3HR-1 clone 5 cell culture supernatant was treated with DNase to remove any intracellular EBV DNA released from damaged cells. Virion DNA protected within enveloped particles was then amplified from limiting dilutions by using primers that targeted the three defined recombination points. With respect to the endpoint, the results were comparable to the outcome observed with P3HR-1 total cellular DNA, where amplifications of het fragments were roughly equivalent (Fig. 2B). The greater intensity of the hybridization signal from WZhet at each dilution of the template could represent arrangement into a palindrome, as reported previously (19, 20). What the approach does not adequately address is whether all het fragments were packaged uniformly within a single defective particle or whether some were present individually in distinct replicons.

Varied molecular configurations of individual het fusion fragments in acutely infected BL cells.

Infection of EBV-negative cells with virus inoculums containing het genomes has previously been reported to increase the molar ratio of het DNA to standard genomic DNA (5). Using that observation to investigate the possibility of diversity among het replicons, we first infected the EBV-negative BL-derived BL2 cell line with a virus preparation from P3HR1 clone 5 cells containing het particles and standard virus. Internalization of virus from mixed inoculums (clone 5) or from preparations of standard (clone 13) P3HR-1 virus alone was ascertained at 48 h postinfection by a Southern blot hybridized with a probe that recognized both the standard 4.6-kb BamHI M DNA fragment and the 7.5-kb M/B1 fragment of het DNA (Fig. 3A). Despite equivalent uptakes of standard genome from both virus preparations, viral activity was altered in BL2 cells by the presence of het DNA. As shown by qRT-PCR, transcripts from the immediate-early and early genes BZLF1 and BSMLF1 (SM) were markedly elevated in het DNA-containing cells, consistent with the ability of het DNA to induce the viral lytic cycle via constitutive expression of BZLF1 (Fig. 3B).

FIG. 3.

Relative expression levels of EBV lytic genes in BL2 cells infected with P3HR-1 het DNA-positive and het DNA-negative virus stocks. (A) Radiograph of Southern blot of BamHI-restricted BL-2 cellular DNA 48 h after infection with virus preparation from P3HR-1 clone 5 (het and standard genomes) or P3HR-1 clone 13 (standard genome). The standard BamHI M fragment is 4.6 kb; the 7.5-kb band represents the M/B1 fusion fragment of het DNA. The figure represents a composite of lanes run on the same gel. (B) qRT-PCR amplification profiles of transcripts from EBV immediate-early gene BZLF1 and early gene SM, present in both standard and het genomes. Expression was normalized to GAPDH levels. Values for cells infected with standard virus alone (clone 13) were arbitrarily set at 1. Each cycle number difference at the threshold (where product plot crosses the horizontal red line) represents a twofold product difference. Samples were run in triplicate. Red represents het DNA-positive P3HR-1 clone 5-derived virus; blue represents het DNA-negative P3HR-1 clone 13-derived virus.

To ascertain whether any one of the three points of recombination could be localized to a separate replicative intermediate based on variable electrophoretic mobility, we differentiated circular, linear, and higher-order genomic structures of EBV DNA in acutely infected BL2 cells via an in situ cell lysis gel as described by Gardella et al. (11). We employed the modification of the method of Decker and colleagues effective at localizing low-copy-number DNA forms by DNA PCR of sequential gel slices taken from the length of a lane (7). Figure 4A shows the results of the PCR/Gardella approach directed toward the standard EBV genome in established controls: the B95-8 cell line, which contains a mixture of latently and lytically infected cells (episomal and linear configurations); the tightly latent Raji cell line (episomal forms only); and B95-8 virions (linear forms only).

FIG. 4.

PCR/Gardella gel analysis of het DNA configuration in acute infection. (A) Relative positions of episomal and linear configurations of EBV DNA. (Left) Radiograph of Southern transfer from Gardella gel, with three lanes each of 10⁵ B95-8 cells (latent and lytic infection) and Raji cells (latent infection). A single lane contained concentrated culture supernatant from B95-8 cells. The probe was derived from the BamHI M segment of EBV DNA. Gel segments numbered 1 to 14 approximate gel slices taken from replicate lanes (not shown) bearing 10³ cells that were analyzed by PCR (right), with the EBV BHRF1 gene segment targeted for amplification. The locations of the standard linear and episomal configurations of EBV DNA determined by this approach served as a control for analysis of het DNA. (B) Sequential appearance of het DNA rearrangements on distinct molecules within acutely infected BL2 cells. (Left) Southern blot with individual Gardella gel lanes containing 10⁵ BL2 cells at sequential time points after infection with the P3HR-1 clone 5-derived virus stock (standard and het genome positive) previously characterized in Fig. 2B. Note the appearance of the hybridization signal in wells at 48 h with the concurrent incremental increase in the linear signal. (Right) PCR analysis of replicate lanes containing 10³ acutely infected BL2 cells at comparable time points. Each of 13 gel slices at every time point was analyzed for all three points of recombination (W/Z, E/S, and M/B1). Hybridized blots of PCR products are grouped to show changes in relative position within the Gardella gel of DNA molecules bearing a single recombination joint over time. The signal at 0.5 h postinfection (hpi) represents input defective particles (linear molecules) prior to DNA replication.

In a variation of that approach, we applied PCR primers targeting the W/Z, E/S, and M/B1 sites of recombination to allow determination of the relative mobilities of het DNAs against the backdrop of the standard viral genome. Figure 4B shows the autoradiogram from the Southern blot of a Gardella gel in which each lane contained 10⁵ infected BL2 cells at successive time points postinfection. A progressively increasing hybridization signal at the position of linear DNA was detected over the 96-hour time course, suggesting a replicative infection. At 48 h, an additional hybridization signal appeared in the wells (the site of the agarose plug), a location previously interpreted to signify the presence of replicative intermediates and/or recombination products whose structures inhibit entry into the gel (17). Consistent with that interpretation is the enhanced signal representing linear processed progeny genomes at 48 h onward. Noteworthy was the absence of a detectable episomal band of lower mobility, since circularization of input linear DNA is known to occur within hours of viral entry (1).

Comparable lanes, run simultaneously with the blotted portion of the gel but removed for PCR analysis, were transected at points indicated numerically (Fig. 4B), the original autoradiogram serving as a reference for excision of gel slices and assessment of het DNA mobility. Three PCRs employing primer pairs targeting separate points of recombination were performed on each of the 13 sequential gel slices excised from the lanes representing sequential time points (Fig. 4B). Two findings deserve mention. First, the three fusion joints of het DNA localized to distinct DNA molecules. At 30 min postinfection, prior to the commencement of replication, a PCR product was obtained with primer sets for W/Z and M/B1 only, each located at the position of linear DNA and representing input het genomes but located in separate gel slices (segments 10 and 11, respectively). Because Gardella gels are quite effective at distinguishing linear molecules from their circular and higher-order genomic structures but lack sensitivity in resolving linear molecules of similar size, the separation clearly places them within discrete DNA molecules. Disparate mobilities were similarly observed at postreplication time points (e.g., 72 h, when W/Z only is in gel segment 6, M/B1 only in segment 7, and E/S only in segment 4) (Fig. 4B).

The second finding of note is that E/S was not detected until 48 h postinfection and then only in the well (Fig. 4B). This finding appears at odds with the roughly equivalent representation of junctional fragments in the original inoculum (Fig. 2B). E/S may have been below detectable limits until amplified by an initial round of replication, hence its initial detection in the well. However, the detection of M/B1 at 30 min postinfection (Fig. 4B) and the sensitivity of the DNA PCR-modified Gardella approach puts this explanation in doubt (7). Alternatively, its absence may reflect poor infectivity of a separate class of replicons bearing this rearrangement, with the appearance at 48 h representing generation de novo by recombination events accompanying WZhet-driven EBV replication. In a similar vein, the reappearance of M/B1 in the well at 24 h may also signify a new recombination event or, alternatively, differences in efficiency of replication among distinct het DNA fragments. The same temporal order of het fragment appearance postinfection delineated by our Gardella gel analysis (W/Z followed by M/B1 and E/S) was apparent in earlier data from Miller and colleagues (28), where less-sensitive restriction fragment analysis postinfection led to detection of WZhet at 3 days, followed sequentially by detection of M/B1 and E/S over the ensuing 4-day interval.

In short, contrary to what would be expected from intracellular propagation of a single het replicon, where every primer pair would yield a product from the same gel slice, recombinations appeared to be situated on multiple DNA molecules. The findings suggest the existence of a population of defective molecules, some of which may predominate in terms of infectivity. Not all may represent functional replicons. However, a hybridization signal both in the well and in the linear position for all fragments at 96 h supports the likelihood that all three rearrangements are present in at least a subpopulation of het DNA molecules.

DISCUSSION

By defining at the nucleotide level the points of recombination between four noncontiguous segments of EBV DNA comprising the het genome in a subset of P3HR-1 cells, we have been able to document comparable EBV DNA rearrangements at sites of viral replication in human infection. All three points of recombination defined in laboratory-derived het DNA were ascertained in salivary samples from 2 of 19 EBV-positive donors with HIV infection. We emphasize that the detection of junctional sequences per se by no means establishes the existence of defective genomes functionally equivalent to that produced by P3HR-1. However, in the context of the atypical burst of EBV replication retrospectively profiled in serological responses of cancer patients, our findings warrant reevaluation of a variant, once considered an oddity of cell culture, that has the property of activating an otherwise inefficient EBV replicative function.

The fidelity of sites of recombination in three novel het EBV DNA restriction fragments derived from both the laboratory strain P3HR-1 and the wild-type virus may indicate a selection for sequences required for survival or having potential biologic significance consequent to altered gene expression. Our data by no means exclude the possibility of additional points of recombination, such as those previously reported for WZhet DNA (35), given the selection bias inherent in our excision, from ethidium bromide-stained gels, of PCR products for cloning that fell within the expected size range. Unlike the case with WZhet, neither of the two newly defined points of recombination appear likely to confer new biological properties upon het DNA as far as can be determined by sequence alone, either through generating novel open reading frames or by altering the regulation of the expression of a standard EBV protein.

The E/S recombination occurred via the 3′ portions of the “rightwardly” and “leftwardly” transcribed genes BERF4 and BSLF1, truncating the 5′ regions of both (Fig. 1E). Of more interest is what was left intact after the M/B1 rearrangement. Recombination in BamHI B1 occurred 51 nucleotides beyond the polyadenylation site for the LF3 gene, preserving this early gene and the adjacent viral oriLyt region (45). Given that LF3 function remains poorly defined, any advantage that additional transcripts from LF3 might provide in lytic infection is at this point purely speculative. Most standard viral strains contain two nearly identical repetitive sequences with oriLyt activity, designated D_L and D_R (i.e., left and right duplications), adjacent to internal reiterations IR2 and IR4, respectively. Other than lytic origins, the duplications contain multiple promoters regulating the adjacent genes BHLF1 and LF3, respectively, whose products are the most abundant poly(A)⁺ viral transcripts synthesized during the EBV lytic cycle (32, 33, 46). LF3 is less tightly regulated than BHLF1, being identified in tumors of both epithelial and lymphoid origin, often in conjunction with Zta (44, 46).

The detection in clinical material of an incomplete complement of het-specific fragments over multiple samplings parallels findings for P3HR-1 cell subclones, where some het DNA restriction fragments, but not others, were identified (25). The consistent detection of WZhet DNA alone over sampling periods extending for periods up to a year indicates that the defective genome can be sustained at least for the short term in infected hosts. WZhet persistence has previously been reported in the clinical setting of an EBV-positive thymic carcinoma, where detection at diagnosis was confirmed 1 year later at sites of tumor reoccurrence (36). Two donors positive for the WZhet fragment alone on initial testing became E/S positive on subsequent sampling occasions (Table 3), raising at least the possibility that WZhet-induced replication of latent virus may generate rearrangements anew in the accompanying standard genome.

Analysis of viral DNA configuration in BL2 cells immediately postinfection, whether in the intact Gardella gel, where standard and het genomes were indistinguishable, or in lanes transected for het DNA analysis, revealed a temporal pattern of hybridization that progressed from linear genomes to DNA in the wells to an increased abundance of linear genomes without detection of unit length circular intermediates. The absence of the latter may reflect a purely lytic infection, the establishment of episomal forms by standard virus being delayed or altogether abrogated by het DNA. Although it has been suggested that DNA configurations other than circles could serve as the initial templates for viral replication (17), the issue was not specifically addressed here. Noteworthy for its similarity, however, is the absence of a predominant, detectable population of episomal EBV DNA in the permissive infection characteristic of oral hairy leukoplakia in HIV infection/AIDS (13), where het DNA is also found (35).

It remains unclear whether a viral strain similar to that of the nontransforming P3HR-1 type 2 virus is required for generation of het DNA in vivo. Deemed a laboratory variant, P3HR-1 was isolated in a cell clone of the BL cell line Jijoye (16) and contains a 6,800-bp deletion in the EBNA2-encoding region of its transformation-competent parent Jijoye (4, 24, 37). The type-specific single-nucleotide polymorphisms contained in the BamHI E sequences retained in het DNA allow some characterization of the parental virus (Fig. 1C). In donor saliva in which both type 1 and type 2 EBVs were detected, het DNA was type 2 (data not shown). Should het derivation be limited to nontransforming parental strains, it need not diminish the relevance of the current findings. EBNA2 rearrangements are regularly detected in lesions of oral hairy leukoplakia as well as in oral washings of healthy donors (42, 43). The recent evidence that generation of EBNA2 deletion variants within EBV-positive endemic BL confers superior survival advantage upon tumor cells moves these so-called rare products of recombination closer to pathogenic center stage (22, 23).

When viewed in the context of the natural history of EBV infection, interplay between EBV het DNA and its parental helper strain may elucidate circumstances surrounding the dramatic antibody response to EBV lytic cycle antigens that precedes EBV-linked cancers and distinguishes EBV reactivation occurring as a harbinger of disease from intermittent shedding characteristic of the healthy virus carrier state. As yet unexplained and largely unexplored, these early precipitating events in the immunocompetent host may mimic the exuberant replication and recombinogenic activity occurring in the abbreviated time frame of HIV infection/AIDS, where the incidence of EBV-associated malignancies is high. In each scenario, primary infection by EBV antedates the onset of disease by years, but the time to neoplasia from EBV reactivation is short by comparison. Such a temporally shortened sequence of events implies cause and effect and raises important questions regarding the part played by viral DNA rearrangement and recombination in EBV pathogenesis.