Heterogeneous Epstein-Barr virus latent gene expression in AIDS-associated lymphomas and in type I Burkitt's lymphoma cell lines

EpsteinBarr virus (EBV) is associated with different tumours, such as endemic Burkitt's lymphomas (BL) and nasopharyngeal carcinoma. EBV is rarely associated with non-Hodgkin's lymphoma (NHL) in non-immunocompromised patients but is involved in approximately 60 % of lymphoma cases in human immunodeficiency virus (HIV)-infected patients, although the frequency may vary with the localization and histology of the tumour (Hamilton-Dutoit et al., 1993; Gaidano et al., 2000).

B lymphocytes are immortalized by EBV infection in vitro, generating permanent lymphoblastoid cell lines (LCLs), in which an array of virus-encoded proteins, including six EBV nuclear antigens (EBNA-1, -2, -3A, -3B, -3C and leader protein) and three latent membrane proteins (LMP-1, -2A and -2B), are expressed. The small EBV-encoded non-polyadenylated nuclear RNAs (EBER-1 and -2) are also expressed. In LCLs, all six EBNAs are generated by differential splicing of a primary transcript that originates from one of the two promoters located in the BamHI C or W fragments (Cp or Wp). This form of latency is termed latency III (Lat III). In human tissue, EBV expression is often more restricted. EBNA-1, the LMPs and the EBERs are expressed in latency II (Lat II), whereas only EBNA-1 and the EBERs are expressed in latency I (Lat I). In Lat I and Lat II, EBNA-1 transcription originates from a different promoter, located in the BamHI Q fragment of the EBV genome (Qp). During the lytic cycle, EBNA-1 mRNA is transcribed from the F promoter (Fp), which lies upstream of Qp (reviewed by Kieff & Rickinson, 2001).

Recently, Kelly et al. (2002) identified a subset of BL tumours in which the Lat III-associated EBNA promoter Wp drove expression of the EBNA-3 genes. EBNA-2 production was abrogated by a gene deletion.

Here, we analysed extensively EBV gene expression in NHL arising in HIV-infected patients using immunohistochemistry (IHC) and/or RT-PCR to monitor the expression of EBNA-1, -2, -3A, -3B, -3C and LMP-1 and -2, as well as BZLF1 (the EBV immediate-early antigen), in 14 biopsies of NHL of HIV-infected patients. Moreover, our results show that expression of EBNA-3 genes can be directed from Fp in BL from HIV-infected patients as well as in Akata and Mutu I BL cell lines.

Tissues.
EBV-positive lymphomas were taken from a series of HIV-related NHL tumours collected by the French Study Group (coordinator, M. Raphaël, Avicenne Hospital, Bobigny, France). Lymphomas were classified according to the Revised EuropeanAmerican classification system (Harris et al., 1994) as well as the recent WHO classification system (Raphaël et al., 2001): BL, Burkitt's lymphoma; BLL, Burkitt's-like lymphoma; DLCL, diffuse large B cell lymphoma; IBP, immunoblastic lymphoma with plasmatoid differentiation. Frozen, formalin-fixed and paraffin-embedded biopsy specimens from 14 cases were used for RT-PCR, IHC and in situ hybridization (ISH). Clinical and pathological data are indicated in Table 1.

Table 1. Clinical and pathological data

IHC of EBV latent and replicative proteins.
Immunophenotypic studies were performed on frozen tissue sections using an immunoperoxidase reaction. Phenotypic antigen CD20 was detected using L26 monoclonal antibody (mAb) (Dako). EBV latent and replicative gene expression was assessed using mAbs against LMP-1 (CS1-4, Dako), EBNA-2 (PE2, Dako) and BZLF1 (BZ, Dako). Tumour cells were detected by conventional microscopy; the tumour was considered as positive when at least 510 % of malignant tumour cells were present. The proportion of malignant cells was greater than 90 % in 9 of 14 patients. In the five other patients, tumour infiltration ranged between 10 and 60 % of the analysed tissue.

Preparation of RNA and RT-PCR.
Frozen specimens were pulverized and RNA extracted using RNasol B (Bioprobe Systems). RNA was then treated with RQ1 DNase (Promega). RNA (1 µg) was reverse-transcribed with MoMLV reverse transcriptase (Gibco) after priming with oligo(dT). PCR was carried out with the cDNA samples obtained from 33 ng of total RNA. PCR was performed as described (Martel-Renoir et al., 1995). Second-round PCR was carried out with 1/50 of the first-round PCR mixture. Primers used are listed in Table 2.

Table 2. Oligonucleotide sequences To compare the efficiency of different PCR amplifications, 10 fg of plasmid DNA containing the target sequences were amplified using specific sets of primers. Amplified DNA was analysed on agarose gels after 21, 24, 27 and 30 cycles of amplification. Efficiency of amplification was similar in all cases (data not shown).

PCR products were analysed by electrophoresis on agarose gels and transferred onto Hybond filters (Amersham) by Southern blotting. Filters were hybridized with specific ³²P-labelled probes to confirm the specificity of the PCR product generated, as described (Martel-Renoir et al., 1995). Comparison of the efficiency of the different PCR amplifications was performed using a semi-quantitative RT-PCR method. A sample of 10 fg of plasmid DNA containing the target sequence was amplified using the specific set of primers. Amplified DNA was analysed on agarose gels after 21, 24, 27 and 30 cycles of amplification. Bands of the expected sizes were visible after 27 cycles of amplification for EBNA-2, -3A, -3B and -3C, indicating that in all cases the efficiency of PCR was similar (data not shown). DNA sequences were determined using the Prism Ready Reaction Dideoxy Terminator Cycle Sequencing kit (Applied Biosystems) on a Model 373A automatic sequencer.

ISH.
EBER ISH (Barletta et al., 1993) was carried out on routinely processed paraffin sections with FITC-labelled EBER-1- and -2-specific oligonucleotides, according to the manufacturer's instructions (Dako).

Western blots.
Western blots were performed using anti-EBNA-3B (Exalpha Biologicals) and the A10 anti-EBNA-3C (Radkov et al., 1997), as described previously (Fahmi et al., 2000).

Characterization of EBV present in each tumour
ISH using EBER-1- and -2-specific probes indicated the presence of EBV in all tumour samples (Table 1). Tumours #8 and #16 were not checked due to an insufficient amount of available material.

PCR was then performed with primers specific for type A and type B EBV (Rowe et al., 1989) on each of the tumour samples, with the exceptions of tumours #1, #7 and #14 (due to a lack of material). Most of the tumour biopsies analysed contained the type A variant, while two samples (#3 and #10) contained both type A and type B strains (Table 1).

EBV gene expression
RT-PCR and IHC were used to determine the specific pattern of EBV gene expression in each tumour cell. RT-PCR was performed for EBNA-2, -3A, -3B and -3C and LMP-1, -2A and -2B and ZEBRA. IHC was performed on EBNA-2, LMP-1 and ZEBRA.

Of the 14 tumours studied, four followed a typical pattern of gene expression: tumour #14 (BLL) showed the classical Lat I pattern, tumour #1 (BLL) showed the classical Lat II pattern and tumours #3 (BL) and #17 (DLCL) showed the Lat III pattern.

Non-canonical patterns that did not follow one of these expressions were also observed. Indeed, the remaining tumours showed various levels of heterogeneity in their patterns of viral gene expression. The majority of these tumours express at least one of the EBNA-3 genes in the absence of EBNA-2, either with or without LMP-1.

Markedly, in one DLCL (#13), we observed expression of EBNA-2 and LMP-1 and -2 genes without the detection of any transcripts of the EBNA-3 gene family. Moreover, in one IBP (#7), three DLCL (#5, #6 and #16), two BL (#10 and #11) and three BLL (#4, #8 and #9), expression of transcripts of the EBNA-3 gene family was observed without detection of the EBNA-2 gene product. An illustration is detailed in Fig. 1: in tumour #11, in the absence of EBNA-2 expression (Fig. 1b, ii), amplification of EBNA-3A, -3B and -3C cDNAs was observed (Fig. 1b, iiiv). Furthermore, in tumour #4, only expression of two members of the EBNA-3 gene family (EBNA-3B and -3C) was detected (Fig. 1b, iv and v), again without EBNA-2. The same results were obtained in three independent experiments.

(25K): Fig. 1. Detection of EBNA transcripts by RT-PCR in RNA preparations from BL and BLL lymphomas of HIV-infected patients. Primer combinations used for RT-PCR are indicated with arrowheads (a) and products obtained after RT-PCR were visualized on agarose gels stained with ethidium bromide (b). For mRNA detection, the following primers were used: (i) EIAS and U172S were used for EBNA-1 detection; (ii) Y2E2S and BYRF1AS for EBNA-2; (iii) RA-AE3S and RA-A/BE3AS for EBNA-3A; (iv) ABE4S and E4AS for EBNA-3B; and (v) RA-AE6S and RA-A/BE6AS for EBNA-3C. The quality of the cDNA products was checked by amplification of hypoxanthine phosphoribosyl transferase (HPRT) (vi). PCR products were hybridized as described by Martel-Renoir et al. (1995) with specific 32P-labelled probes to confirm the specificity of the generated PCR products (data not shown). Tunours #3, #4 and #11 were used in the assay. DG75 and BL41 were used as negative controls, while Raji and BL41/B95-8 served as positive controls. BL41/P3HR1 represents a negative control for EBNA-2.

In the last cases, detection of the three EBNA-3 RNAs was only seen in two DLCL, while in all other cases, either EBNA-3B alone or EBNA-3B and -3C RNAs were detected. Precise characterization of the EBNA-3 transcripts in tumour #4 has been performed.

Characterization of promoter usage
In three lymphomas, #3, #4 and #11, RT-PCR was performed to determine promoter activity. Cp and Wp activities were assessed as described previously (Tierney et al., 1994). F/Qp activity was assayed using a sense primer in the F/Qp region and an anti-sense primer within the BKRF1 open reading frame (ORF) encoding EBNA-1 (Fig. 2). The activity of all promoters, Cp, Wp and F/Qp, was detected in tumour #3, while only the F/Qp promoter seems to be active in tumours #4 and #11, even though EBNA-3 expression was detected in these samples.

(33K): Fig. 2. Detection of Cp-, Wp- and F/Qp-initiated transcripts by RT-PCR analysis in RNA preparations from BL and BLL lymphomas of HIV-infected patients. Primer combinations used for RT-PCR are indicated with arrowheads (a) and products obtained after RT-PCR were visualized on agarose gels stained with ethidium bromide (b). To detect Wp activity, W0W1/S and CpWp/AS were used for first-round PCR and W1SN and W2ASN for second-round PCR (i). C1C2/S and CpWp/AS were used to detect Cp activity (ii) and QPS and E1AS for F/Qp activity (iii). Lymphomas #3, #4 and #11 were used in the assay. BL41 was used as a negative control, BL41/P3HR1 and BL41/B95-8 were used as positive controls.

Characterization of EBNA-3B transcription
As the F/Qp promoter was the only detectably active promoter in tumours #4 and #11, we checked whether EBNA-3B transcription could be initiated from F/Qp. Tumour #4, as well as different cell lines, were assayed by RT-PCR using a forward primer located either 42 bp downstream (FPS) or 43 bp upstream (UpFPS) of the Fp TATA box and a reverse primer overlapping the first (BERF2a) and second (BERF2b) exon of the EBNA-3B gene. No amplification was observed when the UpFPS primer was used, while the positive control PCR with UpFPS and FAS gave a band of the expected size (61 bp) (data not shown). Conversely, when the FPS primer together with primer E4RTAS was used, a band of 750 bp was detected after amplification of cDNAs obtained from tumour #4 and from the BL cell line Akata (Fig. 3a). Tumour #11 was not assayed due to the lack of sufficient material. This result suggests that Fp might direct EBNA-3B transcription in tumour #4 and Akata cells. The sequences of the 750 bp PCR products obtained from Akata and tumour #4 were determined; the sequences obtained from both Akata cells and tumour #4 were identical to the B95-8 strain of EBV (Fig. 3b, c). Sequence analysis showed a region spanning the Q fragment (nt 62 39762 458) that linked to the U172 exon (nt 67 47767 649) and BERF2a by splicing events. The splicing event that joined the Q fragment to U172 was identical to that seen in the F/Qp-driven EBNA-1 mRNA. Exon U172 is the same as that in EBNA-1 and EBNA-3C cDNAs (Speck & Strominger, 1985; Bodescot & Perricaudet, 1986). Exon U172 is linked to the first exon of EBNA-3B (BERF2a) at the splicing acceptor site, as was predicted from sequencing and RNase mapping analyses (Kerdiles et al., 1990). The 5' sequence of BERF2a has not been characterized definitively yet. Since the EBNA-3B cDNA contained sequences upstream of the start site of Qp-driven transcripts, our results suggest that Fp was used to initiate EBNA-3B mRNA expression. This represents the first evidence for EBNA-3B transcription from Fp.

(49K): Fig. 3. Detection of EBNA-3B expression directed through Fp by RT-PCR analysis in RNA preparations from tumour #4 and from the Akata cell line (a). The primer combination used was FPES and E4RTAS. (b) Schematic representation of the 750 bp PCR product obtained from Akata and tumour #4. Exons are indicated by rectangles. The sequence was established for nt 62 39795 498 of the EBV genome. The coordinates of exon boundaries are indicated. The expected sequence of FQUBERF2aBERF2b is represented in (c). The sequence determined experimentally is given in italics. Primers used in this study are underlined and the positions of Fp, Qp and transcription-initiation sites are indicated.

In order to investigate the possibility that FpU172EBNA-3B belongs to a putative EBNA-1 pre-mRNA, RT-PCR was performed with RNA extracted from Akata cells using a sense primer (ABE4S), located within the BERF2b exon, and an anti-sense primer (EIAS), located in the 5' region of the EBNA-1 ORF. No amplification was observed, suggesting that the FpU172EBNA-3B PCR product is not derived from an EBNA-1 pre-mRNA (data not shown).

Fp-driven EBNA-3 gene expression
In Akata and Mutu I cells, the entire EBNA-3 gene family can be expressed through Fp. This expression occurs predominantly after induction of the EBV lytic cycle. EBNA-3A, -3B and -3C expression through Fp was assayed by RT-PCR of RNA from Akata cells. In each case, two rounds of PCR were necessary to visualize a band of the expected size (data not shown). However, if RNA was prepared from Akata cells, in which the EBV lytic cycle had been induced by treatment with 1 % anti-human IgG (Dako), only one round of PCR was needed to amplify EBNA-3A (Fig. 4a), -3B (Fig. 4b) and -3C (Fig. 4c) cDNAs. The three amplified bands hybridized with the ³²P-labelled oligonucleotide U172AS, suggesting that the cDNAs harbour the U172 exon. This result shows that in Akata cells, EBNA-3 expression from Fp occurs predominantly after the induction of the EBV lytic cycle. Similar results were obtained for Mutu I cells. Fig. 4(df) shows expression of EBNA-3A, -3B and -3C, respectively, in Mutu I cells in which the lytic cycle has been induced by treatment with TGF-β1 (transforming growth factor-β1). As with Akata cells, no expression of mRNA for any of the EBNA-3 genes was detectable in the absence of lytic cycle induction, though, as expected, EBNA-3B and -3C proteins were detected by Western blot in Mutu III cells (Fig. 5). Fig. 4(df) shows that no expression of EBNA-3A, -3B or -3C occurs through Fp in Mutu III cells, irrespective of whether the virus is induced to lytic cycle or not (virus induction was verified by Western blot of the ZEBRA protein, data not shown).

(23K): Fig. 4. Detection of EBNA-3A, -3B and -3C expression directed through Fp by RT-PCR analysis in RNA preparations from BL cell lines uninduced or induced to the lytic cycle. For Akata cells, the primer combination used was FPS and RA-A/BE3AS to detect EBNA-3A expression (a), FPES and E4RTAS for EBNA-3B (b) and FPES and RA-A/BE6AS for EBNA-3C (c). For Mutu I, Mutu III, Kem I, Kem III and Sav cell lines, the primer combination used was FPES and RA-A/BE3AS to detect EBNA-3A expression (d), FPES and E4RTAS for EBNA-3B (e) and FPES and EBNA6AS for EBNA-3C (f). Products were analysed by Southern blot and hybridized with the 32P-labelled U172AS oligonucleotide.

(31K): Fig. 5. Immunoblot analysis for the detection of EBNA-3B and EBNA-3C in group I BL cell lines upon induction of the lytic cycle. The equivalent of 100 µg of protein was loaded per lane. Blots were probed using anti-EBNA-3B (a) and anti-EBNA-3C (b) mAbs. B95-8 and Mutu III served as positive controls.

Fp-driven EBNA-3 expression was also investigated in Kem I and Sav I type I BL cell lines. In these cells lines, Fp-driven EBNA-3 expression was not detectable, even when the lytic cycle was induced. In light of these results, it seems that Fp-driven EBNA-3 expression may occur in some, but not all, type I BL cell lines. Furthermore, this expression was never detected in type III BL cell lines, as observed for Mutu III (Fig. 4df) or with other group III cell lines such as B95-8 or Kem III (data not shown).

EBNA-3B and -3C proteins were detected in Akata and Mutu I BL cell lines
To verify that the EBNA-3B and -3C transcripts detected in our RT-PCR experiments could in fact give rise to their corresponding proteins, Western blots were performed using anti-EBNA-3B and the A10 anti-EBNA-3C mAbs (Fig. 5). EBNA-3B protein was detected in Akata and Mutu I cells, as well as in Mutu III and B95-8 cells, which were used as controls. In both Akata and Mutu I cells, the levels of detected protein were similar, irrespective of whether or not the lytic cycle was induced (Fig. 5). However, in Akata cells, the signal for EBNA-3B expression was only seen after overexposure of the blot. Hence, the EBNA-3B protein is expressed in certain group I BL cell lines and it is possible that at least some of the protein detected in our Western blots was the product of the mRNA transcripts observed by RT-PCR. Results obtained for EBNA-3C were slightly different. Unlike EBNA-3B, EBNA-3C protein was not detected in Akata cells, whether induced or not. However, EBNA-3C was detected at similar levels in both induced and uninduced Mutu I cells.

Lytic gene expression is detected in NHL of HIV-infected patients
In 7 of 14 lymphomas, EBV expression was not wholly latent, since the immediate-early protein ZEBRA was detected. Fig. 6 shows IHC detection of ZEBRA in frozen sections of tumour #4. The large nuclei of numerous tumour cells stain positive, demonstrating that virus reactivation had occurred in this tumour. This correlates with results described in Fig. 3 showing that the lytic promoter Fp is active in tumour #4.

(164K): Fig. 6. Expression of ZEBRA protein. ZEBRA was detected by IHC on frozen sections using the BZ mAb. Antigenantibody complexes were visualized using an immunoperoxidase reaction and observed by conventional microscopy. Magnification, x500.

In the categories of classical BL and BLL, as characterized by heterogeneous morphology showing a morphological spectrum from BL to IBP, we have not detected any significant differences in EBV expression. On the other hand, our results showed a heterogeneous pattern of latency expression in EBV-positive AIDS-related lymphomas. Except in two cases where classical Lat III patterns of expression were observed, the most striking feature was the presence of EBNA-3 transcripts (EBNA-3A and/or -3B and/or -3C) associated or not with LMP-1 but without detection of EBNA-2. This non-canonical pattern of latent gene expression was also observed in Akata and Mutu I cell lines; this transcription occurs predominantly after induction of the productive cycle.

The BL cell lines Oku and Sal have been shown to produce Wp/Cp-driven EBNA-3 transcripts from a virus in which the EBNA-2 gene has been deleted (Kelly et al., 2002). In Akata and Mutu I cell lines, the EBNA-2 gene is not deleted, since we could obtain amplification of the corresponding ORF using primers E2S and E2AS (data not shown). In this case, the EBNA-2 gene is silent, as the promoter that drives EBNA-3 gene expression is located downstream of the EBNA-2 ORF. Our results show that EBV can use a mechanism other than EBNA-2 deletion to produce EBNA-3 without EBNA-2.

The Q promoter is used in Lat I and II to generate EBNA-1 transcripts. During the lytic cycle, transcription from Fp generates EBNA-1 mRNA (Lear et al., 1992; Sample et al., 1991). We show that expression of the EBNA-3 gene family occurred through Fp. This promoter is used for EBNA-1 and -3 expression during the lytic cycle and we did not observe any pre-mRNA containing both EBNA-3B- and -1-encoding sequences. The U172 exon is present in EBNA-1 transcripts and in the latent mRNA harbouring the EBNA-3C gene (Speck & Strominger, 1985; Bodescot & Perricaudet, 1986). Here, results show that the U172 exon is also present in the Fp-driven EBNA-3 transcripts and sequence analysis shows that U172 is spliced to the BERF2a exon. We then confirmed by sequencing the splicing acceptor site of BERF2a, which has been predicted from early RNase mapping analyses (Kerdiles et al., 1990) but has never been identified definitively.

It has been reported that EBNA-1 (Lear et al., 1992; Nonkwelo et al., 1996) as well as the truncated form of the LMP-1 protein (Hudson et al., 1985) are expressed during the lytic cycle. Results presented here show that the EBNA-3 gene family can also be expressed during the lytic cycle. This represents a new type of transcription pattern observed in some type I BL cell lines as well as in lymphomas of immunocompromised patients. Expression of EBNA-3 proteins in lymphomas of immunodeficient patients is important as these proteins are immunodominant targets for CD8⁺ cells (Rickinson & Moss, 1997).

We are grateful to Alan Rickinson for providing Sav I, Sav III, Kem I and Kem III cell lines, Jean Feuillard for providing RNAs from HIV lymphoma, to Martin Rowe for the A10 mAb and to Audrey Alberga and Nathan Laborde for careful reading of the manuscript. This work was supported by ANRS.