INTRODUCTION

Top Abstract Introduction Materials and methods Results Discussion References

Epstein-Barr virus (EBV), a B-lymphotropic gamma herpesvirus, is widespread in all human populations, where it is carried by the great majority of individuals as a lifelong asymptomatic infection. This same virus has potent B-cell growth transforming ability and is strongly implicated in the pathogenesis of several human malignancies, notably in the endemic and in some sporadic cases of Burkitt's lymphoma (BL), in some cases of classical Hodgkin's disease (HD), in a specific type of nasal T-cell lymphoma, and in undifferentiated nasopharyngeal carcinoma (NPC) (23, 26). How the virus might contribute to the pathogenesis of such a diverse set of malignancies remains to be determined. One hypothesis, however, is that particular EBV strains could be associated with particular tumor types, possibly through a change in cell tropism or through the acquisition of mutations in growth-transforming latent-cycle genes that endow the virus with increased oncogenic potential (1, 7, 8, 16, 17, 19, 20, 22). Interest in this possibility has grown as the extent of EBV's genetic diversity has become more apparent. In this context there are two major types of EBV, now called types 1 and 2, that are distinguished by linked polymorphisms in the latent-cycle genes encoding the nuclear antigens EBNAs 2, 3A, 3B, and 3C (9, 31). This remains the only genetic classification for which there is a clear biological correlate, in that type 1 strains have stronger in vitro transforming ability for resting B cells than do type 2 strains (27). Within each virus type, however, there are now many polymorphic markers that allow individual strains to be distinguished from one another; some markers are informative among viral strains from the same geographic area (11, 32, 33), whereas others represent differences that largely correlate with the geographical origin of the virus (1, 2, 10, 17, 18, 21, 24).

The present work focuses on the latent-cycle gene encoding the virus genome maintenance protein EBNA1. Though EBNA1 does not display any obvious type-specific polymorphism, a degree of interstrain sequence variation has been noted from the analysis of EBV-associated tumors and of viral strains detectable in the blood and/or throat washings of asymptomatic carriers (4, 14, 35, 38). The EBNA1 protein is composed of unique N-terminal (residues 1 to 89) and C-terminal (residues 328 to 641) domains flanking a large Gly-Ala repeat, and most sequencing studies have focused on a region (residues 466 to 527) which is within the molecule's DNA binding-dimerization domain (3) and which, from X-ray crystallographic data, contains at least some of the important DNA contact residues (5, 6). Recently, Bhatia and colleagues (4, 14) have used signature changes at residue 487 to classify five distinct EBNA1 subtypes; these were the prototype B95.8 strain sequence P-ala, a closely related subtype P-thr, and three more distant variants V-pro, V-leu, and V-val. These authors reported that in a heterogeneous panel of EBV-positive BLs (24 from endemic and nonendemic areas of Africa and 12 from North and South America), the distribution of EBNA1 subtypes was markedly different from that detectable in the blood and/or throat washings of a similarly heterogeneous panel of healthy controls. In particular, almost 50% of the tumors carried a V-leu subtype sequence, whereas this was never found in controls (4).

The implication, that certain EBNA1 subtypes carried a greater lymphomagenic risk, was particularly interesting because EBNA1 is the only viral protein detectably expressed in BL tumor cells (30). Furthermore, there is circumstantial evidence from transgenic mouse studies (37), from in vitro work with a BL cell line (34), and from transactivation assays with reporter gene constructs (36) to suggest that EBNA1 has other activities besides virus genome maintenance and that these could underpin a more direct role for the protein in BL pathogenesis. We were therefore interested in addressing two outstanding questions raised by these studies. First, might the process of EBV-associated lymphomagenesis involve the de novo generation of a rare pathogenic viral variant by mutation of the patient's dominant preexisting strain? For that purpose, we study here 12 cases of endemic BL and compare the viral strain within the tumor with that rescued from the patient's normal B cells. Second, does the difference between BL-associated and control donor-derived EBNA1 subtypes hold true when comparisons are focused within more circumscribed geographic areas? Here we report appropriately controlled studies on BL from two endemic areas, East Africa (Uganda and Kenya) and New Guinea, and on sporadic cases of BL from Europe.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and methods Results Discussion References

BL and control samples from East Africa. A total of 55 endemic BLs from East Africa (44 carrying type 1 EBV and 11 carrying type 2 EBV) were obtained as freshly excised biopsies transported on ice in culture medium (RPMI 1640 supplemented with 2 mM glutamine, 40 µg of gentamicin per ml, and 10% [vol/vol] fetal calf serum) and were received within 48 to 72 h. Of these, 12 (8 type 1 and 4 type 2) came from the Lake Victoria-Machakos regions of Kenya; 9 of these Kenyan tumors were studied as derived cell lines with confirmed BL-associated chromosomal translocations that have been fully described elsewhere (13, 28), and 3 were studied as biopsy cell suspensions. A further 43 BLs (36 type 1 and 7 type 2) came from the West Nile region of Uganda; 27 of these were studied as derived cell lines with confirmed BL-associated chromosomal translocations (13a), and 16 were studied as biopsy cell suspensions. Heparinized blood samples from the same BL patients and from appropriate control donors were transported at ambient temperature in an equal volume of RPMI 1640 medium and were received within 48 to 72 h. In 12 cases of BL (2 from Kenya and 10 from Uganda) the resident EBV strain in the patient's normal peripheral B cells was rescued by spontaneous outgrowth of blood lymphocyte cultures to EBV-positive lymphoblastoid cell lines (LCLs) as described elsewhere (43); between 1 and 10 independently-derived LCLs were obtained per patient. In addition, a total of 32 control EBV isolates (25 type 1 and 7 type 2) were likewise rescued in vitro from normal individuals resident in the same BL endemic areas. Of these 20 (14 type 1 and 6 type 2) came from Kenyan donors as described elsewhere (43) and 12 (11 type 1 and 1 type 2) came from Ugandan donors.

BL and control samples from Europe. A total of three EBV-positive sporadic BLs (all carrying type 1 EBV) were available as derived cell lines with confirmed BL-associated chromosomal translocations. These lines, kindly provided by G. Lenoir (IARC, Lyon, France), were derived from French Caucasian individuals and have been described elsewhere (29). A total of 32 control EBV isolates (23 type 1 and 9 type 2) were generated by spontaneous in vitro transformation of peripheral blood B cells from British Caucasian donors. Of these 22 were healthy adults, 2 were infectious mononucleosis patients (prefix IM), 4 were HIV-positive hemophilia patients (prefix QEH, BCH, or OX), and 4 were HIV-positive homosexuals (prefix EBH). Note that many of these control isolates were from earlier studies (39-42), and the panel was deliberately selected so that the numbers of type 2 Caucasian isolates matched those in the East African panels.

BL and control samples from New Guinea. A total of four cell lines derived from endemic BL (two type 1 EBV and two type 2) and four spontaneously derived LCLs from normal individuals (two type 1 and two type 2) were available for study. These, kindly provided by D. Moss (QIMR, Brisbane, Australia), have been described elsewhere (2, 25, 29) and come from lowland regions of New Guinea where BL is endemic.

EBNA1 gene sequencing. DNA was prepared from cell pellets by standard methods, and the relevant regions of the EBNA1 gene were amplified by PCR as follows. The N-terminal coding region was first amplified with the 5' primer 5'-GTCTGCACTCCCTGTATTCA-3' (B95.8 coordinates 107881 to 107900) and the 3' primer 5'-CTTTGCAGCCAATGCAA-3' (B95.8 coordinates 108199 to 108183) under the following PCR conditions (94°C for 30 s, 46°C for 60 s, 72°C for 120 s, 35 cycles). The C-terminal coding region was first amplified with the 5' primer 5'-GAAAAGAGGCCCAGGAGTCCCAGTAGTCAG-3' (B95.8 coordinates 109081 to 109110) and the 3' primer, 5'-AACAGCACGCATGATGTCTACTGGGGATTT-3' (B95.8 coordinates 109969 to 109940) under the following PCR conditions (94°C for 60 s, 62°C for 90 s, 72°C for 240 s, 35 cycles). The PCR products were then gel purified with a Qiaex agarose gel extraction kit (Qiagen, West Sussex, United Kingdom) according to the manufacturer's instructions and then sequenced by PCR with a Perkin-Elmer Amplicycle kit (Perkin-Elmer/Applied Biosystems, Warrington, United Kingdom) and either the above primers or, for the C-terminal coding region, the following internal primers: E1 sequence 1 5'-GGTTCCAACCCGAAATTTGA-3' (B95.8 coordinates 109366 to 109385), E1 sequence 2 5'-AAGGGAGGTCTTACTACCTC-3' (B95.8 coordinates 109500 to 109481), and E1 sequence 3 5'-AGAAGGCCCAAGCACTGGAC-3' (B95.8 coordinates 109278 to 109297). All primers were ³²P end labeled prior to PCR sequencing under the following conditions (94°C for 30 s, 42°C for 30 s, 70°C for 60 s, 30 cycles).

EBV type and strain analysis. All BL-associated and normal donor-derived EBV isolates were classified as type 1 or type 2 by using standard PCR assays across type-specific regions of the EBNA2 and EBNA3C genes as described earlier (31). Selected isolates were additionally screened for three polymorphisms of the latent membrane protein (LMP) 1 gene, namely, by the presence or absence of a XhoI restriction site, the presence or absence of the 30-bp deletion, and the number of 33-bp repeats (24); all were assayed by standard PCR methods as described previously (17).

	RESULTS

Top Abstract Introduction Materials and methods Results Discussion References

Comparison of BL- and normal B-cell-derived virus strains from endemic BL patients. The initial set of experiments focused on 12 East African BL patients (10 Ugandan and 2 Kenyan) from whom a tumor cell line or tumor biopsy cells were available and also on one or more LCLs derived by spontaneous outgrowth from cultures of the patient's normal peripheral blood B cells. For each of these pairs of isolates, the resident EBNA1 gene was sequenced across codons 1 to 70 and 395 to 641; this covers most of the unique N-terminal and C-terminal regions of the EBNA1 protein, including all positions where amino acid changes have been previously reported (4, 14, 35, 38). The results, summarized in Table 1, show amino acid changes relative to the B95.8 prototype and also show codons with silent nucleotide changes. Most importantly, within each individual pair of BL-derived and normal B-cell-derived virus isolates the EBNA1 nucleotide sequence was identical. However, there were significant differences between pairs, and the individual results are arranged in Table 1 to illustrate the different EBNA1 sequence patterns observed.

View larger version (53K): [in this window] [in a new window]   FIG. 1.   Representative sequencing gels for the Jada-BL, Angu-BL, and B95.8 virus isolates across codons 1 to 40 (N-ter) (A) and codons 464 to 504 (C-ter) (B) of the EBNA1 gene. Nucleotide changes in Jada and Angu relative to the B95.8 prototype are shown in bold capital letters, and codon changes are shown in bold numbers; the corresponding nucleotides and codons in B95.8 are shown in lowercase and lighter print. The N-terminal and C-terminal sequence patterns are identified below the relevant gels by using the signature amino acids at positions 16 and 487, respectively.

EBNA1 subtype distribution among BL- and control donor-derived viral strains from particular geographic areas. To allow a more extensive analysis of EBNA1 subtype distribution among virus strains from BL patients and control donors, we elected to sequence all isolates across codons 1 to 60 and 460 to 510 of the EBNA1 gene. As is clear from Table 1, these regions of EBNA1 contain many of the informative N- and C-terminal polymorphisms, including the signature codons 16 and 487, respectively.

(i) East African BL patients and controls. In all, we studied 55 cases of BL, either as recently established cell lines or as fresh tumor cell suspensions, from classical tumor-endemic regions of Uganda and Kenya. The distribution of BL-associated EBNA1 subtypes was very similar in the two countries, and so the data were combined. Table 3 summarizes the changes observed for all 55 BL-derived virus isolates over the designated N- and C-terminal regions vis-à-vis the B95.8 prototype sequence. In all, four distinct EBNA1 subtypes could be identified. Thus, 6 of 55 BL isolates displayed the Q/T pattern, 12 of 55 displayed the Q'/T pattern, and 29 of 55 displayed the E/L pattern; these three subtypes have already been introduced in the sequencing study shown in Table 1. In addition, the remaining 8 of 55 BL isolates showed the Q' pattern at the N terminus with a C-terminal sequence that was identical to the L sequence except for the absence of a change at codon 471 and the presence of an alternative codon (GAC as opposed to GAT) at residue 500; in view of these differences plus the consistent association of this L-related sequence with Q' (as opposed to E) at the N terminus, it was classified as L'. We noted that type 1 BL isolates, which constituted 80% of the total, were distributed among all four EBNA1 subtypes; by contrast 10 of 11 type 2 isolates were clustered in the E/L subtype group.

(ii) European BL patients and controls. EBV genome-positive tumors form a small minority of the sporadic BL cases arising in Europe (23) and only three cases, all carrying a type 1 virus strain, were available for analysis (29). The data from these cases are shown in Table 5. Two tumors showed a typical Q/T EBNA1 subtype pattern with a series of coding and silent changes identical to those seen in Q/T isolates from East Africa. The third tumor followed the B95.8 prototype sequence at both the N and the C termini (except for a single codon change at residue 499) and was therefore classified as EBNA1 subtype E/A.

(iii) New Guinea BL patients and controls. Although there were limited numbers of BL cell lines and normal control isolates available from New Guinea, we felt it important to include these in the analysis since New Guinea is the only region outside equatorial Africa with classical endemic BL (23). The summary data are shown in Table 7. All four BL cases (two type 1 and two type 2) displayed an E/T EBNA1 subtype pattern, hitherto never seen in East Africa and seen only as a rare subtype among the European control donors. Importantly, this same pattern was also shown by two of the four control isolates from New Guinea (one type 1 and one type 2). The other two New Guinea control donors carried quite different EBNA1 sequence combinations. Both were characterized by N termini that were variants of the Q sequence and were, for the present purpose, classified as Q" and Q. The Q" sequence was found in association with a C terminus with coding changes at positions 487, 499, and 502 that was designated as V from the signature residue at position 487, hence giving the subtype Q"/V. The other virus had a hitherto unreported C-terminal sequence with multiple changes relative to the B95.8 prototype but retaining the A residue at position 487, giving the subtype Q/A^ng.

	DISCUSSION

Top Abstract Introduction Materials and methods Results Discussion References

The present study was prompted by reports from Bhatia and colleagues (4, 14) suggesting that EBV strains with particular EBNA1 sequences in the C-terminal 466 to 527 region, especially a sequence designated V-leu, were preferentially associated with both the endemic and nonendemic forms of BL. These authors could not detect the V-leu sequence by direct amplification from the blood and/or throat washings of their control donors, but they did detect up to four other C-terminal sequences in these control samples, often with more than one sequence present in the same individual. The sequences were designated as "prototype" if they were identical or closely related to B95.8 (i.e., P-ala or P-thr, with residue 487 as the signature residue) or "variant" if they were more distant from the B95.8 sequence (i.e., V-pro or V-val). Because these variants were only detected in the presence of a coresident prototype sequence, these authors suggested that the variants were being generated de novo by ongoing mutations in the EBNA1 gene (14); more recent observations of EBNA1 sequences in EBV-positive nasal T-cell lymphomas have been interpreted in a similar way (15). This led to a scenario for BL pathogenesis in which the tumor selectively involves a rare EBNA1 variant that has arisen in the patient against the background of a preexisting infection with a prototype virus. Our immediate objective, therefore, was to look for evidence of such events by comparing the tumor-derived virus strain from that independently isolated from the patient's normal B-cell pool. In all 12 cases of endemic BL studied, the pairs of isolates had identical EBNA1 sequences (Table 1) and indeed were also concordant at several other polymorphic loci (Table 2). The only way in which such data can be reconciled with the above hypothesis would be if the postulated lymphomagenic variant were to become the dominant virus in the patient's general B-cell pool or were to be selectively rescued from that pool in vitro. Though such explanations remain formally possible, we consider them to be unlikely in view of the results obtained in the remainder of this study.

Our subsequent objective was to reexamine the question of EBNA1 subtype distribution among BL-derived versus control donor-derived viral isolates but now focusing within more-circumscribed geographic areas than were used in the original reports (4, 14). In this context we would point out that our control virus sequences were all amplified from in vitro isolates rescued by spontaneous LCL outgrowth rather than by being amplified directly from ex vivo lymphocyte preparations. We think that this is unlikely to be a major source of error since, at least for healthy European virus carriers, we find that the EBNA1 subtype of the derived LCL accurately reflects the dominant virus strain detectable in the blood by direct amplification (15a); however, we acknowledge that this might not necessarily be the case for virus carriers in BL-endemic areas. With this caveat, our study went ahead with a unique collection of BL and control donor isolates built up over several years from two geographically adjacent areas of East Africa in which BL occurs in its classical endemic form. The work was then extended to rare cases of EBV-positive sporadic BL occurring in Europe, again compared to a large group of controls, and finally to a limited number of BL and control donor samples from a second quite different BL-endemic area, namely New Guinea. Our results (Tables 3 to 8) lead to the following conclusions.

First, as earlier work has implied (14, 35), it is necessary to identify both N- and C-terminal sequences to gain a true picture of EBNA1 subtype polymorphism. Within any one geographic area, certain N- and C-terminal signature sequences tend to be present in particular combinations; however, one cannot use the sequence at the N terminus as a predictor of the C-terminal sequence pattern (or vice versa) since "recombinant" EBNA1 subtypes also exist. For example, among our European control group the most frequent N-terminal/C-terminal combinations were Q/T and E/A; however, a small number of isolates displayed a different combination, E/T (Table 6). Such a subtype could have been produced by a recombination event within the intervening repeat sequence in the EBNA1 gene. Interestingly, one of the original EBNA1 sequences reported by Wrightham et al. (38), designated AM, likewise appears to be a recombinant, this time within the C-terminal region itself.

Second, different geographic areas have different spectra of EBNA1 subtypes present within their host populations, though these spectra can overlap. Thus, as summarized in Table 8, of the five EBNA1 subtypes found in East African control donors, one of the most common (Q/T) was also prevalent among the European controls. On the other hand, the most common East African subtype (E/L) was found rarely and other subtypes (Q'/T, Q'/L', and Q/L) were not found at all among the European group. Although only a few isolates were available from New Guinea, of the three EBNA1 subtypes observed, two (Q"/V and Q/A^ng) were unique to this area, while a third (E/T) was identical to that found in a minority of Europeans. Interestingly, the B95.8 prototype sequence (E/A) was common in Europe but was not seen in either of the other two areas. EBNA1 sequence polymorphisms therefore reinforce the picture of geographic differences between EBV strains that has been built up from earlier studies of other polymorphic loci (1, 2, 10, 17, 18, 21, 24).

Third, the East African and European control groups contained sufficient numbers of type 1 and type 2 isolates to allow tentative conclusions to be drawn regarding EBNA1 subtype polymorphism in relation to EBNA2/EBNA3C virus type. In both geographic areas, type 1 viruses were heterogeneous in terms of EBNA1 subtype, whereas type 2 viruses tended to be more uniform. Thus, East African type 2 isolates either fell into the E/L subtype group or carried a unique subtype Q/L (Table 4), whereas all European type 2 isolates followed the B95.8 EBNA1 prototype E/A (Table 6). These observations agree with earlier results from immunoblotting experiments which suggested that the EBNA1 proteins encoded by type 2 viruses show significantly less size variation than do their type 1 counterparts (43); it may be that within any one host population, type 2 viruses show relatively limited heterogeneity. However, it is important to note that the type 2 isolates from East Africa and those from Europe are quite distinct from one another at the EBNA1 locus. These findings are consistent with the view that both type 1 and type 2 viruses were carried out of Africa in the early days of human migration into Europe and that both have coevolved with their host population since that time. Thus, type 2 virus strains that are found occasionally in the general European population (12, 41) and at increased incidence among HIV-positive homosexuals in that same population (42) (see EBH 43, 46, and 48 [Table 6]) carry a characteristically European EBNA1 sequence and therefore do not represent recent imports from Africa.

Finally, and most importantly, the EBNA1 subtypes associated with BL in any one geographic area are also prevalent among viral strains within the general host population in that same area. The strongest evidence in this context comes from the 55 cases of endemic BL seen in East Africa. Here, the most common BL-associated EBNA1 subtype, E/L, was also the most common subtype seen among the geographically matched control group (Table 8). This makes it clear that virus isolates with a Leu residue at EBNA1 position 487 (V-leu in the terminology of Bhatia and colleagues) are not preferentially associated with the Burkitt tumor but are widespread in the East African population in which the tumor arises. Of course, it might be argued that the high prevalence of such an EBNA1 subtype in the general population could explain why BL occurs with such high frequency in East Africa. However, the EBNA1 subtypes associated with BL in East Africa were not found in tumors in another area of classical BL endemicity, New Guinea; all four New Guinea BLs studied carried a different subtype, E/T, which was also found in two of the four normal donor isolates studied from this region. Likewise, the EBNA1 subtypes seen in the three cases of sporadic BL from Europe fell within the major EBNA1 subtype groups, Q/T and E/A, found in the normal European population. We therefore conclude that there is no obvious relationship, of the kind proposed by Bhatia and colleagues (4, 14) between particular EBNA1 subtypes and the Burkitt tumor; our study is far too small, however, to eliminate the more subtle possibility that certain EBNA1 mutations confer a marginally higher lymphomagenic risk. Whether there is any relationship between particular EBNA1 subtypes and other EBV-positive tumors remains to be seen. On the one hand, the evidence from EBV-positive nasal T-cell lymphomas might suggest this (15). On the other hand, we noted that the particular EBNA1 sequence variant (with a Val residue at position 487) reportedly frequent in Chinese NPC (14, 35) was actually very similar to that present in one of our New Guinea control isolates (L19; Table 7B). Reasoning that this may represent another example of a polymorphism common among Southeast Asian virus strains (10, 18), we recently analyzed several peripheral blood-derived isolates from Chinese control donors and found this EBNA1 variant present in every case (23a). Caution must be exercised, therefore, when data are available from tumor-derived virus strains in the absence of geographically matched control isolates.