Epstein-Barr virus nuclear antigen EBNA-LP is essential for transforming naive B cells, and facilitates recruitment of transcription factors to the viral genome

The Epstein-Barr virus (EBV) nuclear antigen leader protein (EBNA-LP) is the first viral latency-associated protein produced after EBV infection of resting B cells. Its role in B cell transformation is poorly defined, but it is reported to enhance gene activation by the EBV protein EBNA2 in vitro. We generated two sets of EBNA-LP knockout (LPKO) EBVs containing a STOP codon within each repeat unit of IR1. Intronic mutations in the first of these knockouts suggested a role for the EBV sisRNAs in transformation. LPKOs with intact introns established lymphoblastoid cell lines (LCLs) from adult B cells at reduced efficiency, but umbilical cord B cells, and naive (IgD+, CD27-) adult B cells consistently died approximately two weeks after infection with LPKO, failing to establish LCLs. Quantitative PCR analysis of virus gene expression after infection identified both an altered ratio of the EBNA genes, and a dramatic reduction in transcript levels of both EBNA2-regulated virus genes (LMP1 and LMP2) and the EBNA2-independent EBER genes, particularly in the first 1-2 weeks. By 30 days post infection, these levels had equalised. In contrast, EBNA2-regulated host genes were induced efficiently by LPKO viruses. Chromatin immunoprecipitation revealed that recruitment of EBNA2 and the host factors EBF1 and RBPJ to all latency promoters tested was severely delayed, whereas these same factors were recruited efficiently to several host genes, some of which exhibited increased EBNA2 recruitment. We conclude that EBNA-LP does not simply co-operate with EBNA2 in activating gene transcription, but rather facilitates the recruitment of several transcription factors to the viral genome, to enable transcription of virus latency genes. Additionally, our findings suggest that different properties of EBV may have differing importance in transforming different B cell subsets. Author summary Epstein-Barr virus (EBV) infects almost everyone. Once infected, people harbor the virus for life, shedding it in saliva. Infection of children is asymptomatic, but a first infection during adolescence or adulthood can cause glandular fever (mono). EBV is also implicated in several different cancers. EBV infection of B cells (the immune cell that produces antibodies) can drive them to replicate almost indefinitely (‘transformation’), generating cell lines. We have investigated the role of a virus protein – EBNA-LP – which is thought to support gene activation by the essential virus protein EBNA2. We have made an EBV in which the EBNA-LP gene has been disrupted. This virus (LPKO) shows several properties. 1. It is reduced in its ability to transform adult cells, while immature B cells (more frequent in the young) die two weeks after LPKO infection. 2. Some virus genes fail to turn on immediately after LPKO infection. 3. Binding of EBNA2 to these genes is delayed, as is binding of some cellular factors. 4. EBNA-LP does not affect EBNA2-targeted cellular genes in the same way. This shows that EBNA-LP is more important in immature cells, and that it regulates virus genes – but not host genes – more widely than simply through EBNA2.

numbers of IR1 repeat units below five progressively reduced transformation efficiency 126 [29], but as well as changes to maximum EBNA-LP size, this decrease could be due to 127 the reduced Wp number producing less of the EBNA proteins (particularly EBNA2) or of 128 the recently identified stable intronic RNAs (sisRNA1 and sisRNA2) that are produced 129 from the introns between W exons [30]. 130 These prior studies of EBNA-LP function have been conducted in the context of 131 transfecting isolated genes, and/or in the presence of the truncated EBNA-LP protein 132 produced by the P3HR1 virus, and not in the context of virus infection. Therefore, the 133 aim of this project was to produce a complete EBNA-LP knockout virus, and use it to 134 Genetic analysis of EBNA-LP function 7 establish the importance of (and a role for) EBNA-LP in the transformation of B cells. 135 While our first EBNA-LP knockout was additionally defective due to mutations in the 136 introns between the EBNA-LP exons, a second, cleaner knockout showed that EBNA-137 LP is important but dispensable for the transformation of adult memory B cells, but is

145
Generation and validation of EBNA2-and EBNA-LP-deficient BACs. 146 Because of the multiple copies of Wp, and the alternative splicing of the EBNA-147 LP transcript, the only valid approach to completely knockout EBNA-LP was to 148 introduce a nonsense mutation into the EBNA-LP coding region -with a PvuI restriction 149 site to help screening - (Fig 1A) into each of the IR1 repeat units of EBV. This was done 150 initially using class IIS restriction enzymes to generate an array of 6.6 mutated IR1 151 repeat units (Fig S1), the same size as in the parental EBV BAC (designated wild-type  In order to facilitate comparison with the previous genetic studies of  function in a P3HR1 strain backbone [27,28], we also generated a pair of recombinant 165 viruses (designated YKO) that lacked the protein domains of the Y exons but retained 166 exon Y1 splice acceptor and exon Y2 splice donor (Fig 1B). A revertant (Yrev) was 167 Genetic analysis of EBNA-LP function 9 generated for one of these knockouts. In order to separate the role of EBNA2 from that 168 of EBNA-LP, an EBNA2 knockout (E2KO) EBV -and its revertant, E2rev -were also 169 generated. E2KO retains the entire Y3 exon and its 3' splice site. This allows qPCR 170 detection of Y2-YH EBNA2 transcripts in the E2KO, despite being deleted for the rest of 171 the EBNA2 ORF ( Fig 1C). All of the BACs were screened by restriction digestion and 172 pulsed field gel electrophoresis to ensure they were identical to WT-HB9 except for the 173 intended modifications (Fig S2). the mutations did not alter the levels of any latency proteins other than those that had 182 been mutated (Fig 1C and Fig S4). However, the YKO genomes only produced a very 183 low level of C-terminally truncated EBNA-LP, and neither proteasome inhibition (MG132 184 treatment), nor analyzing whole cell lysates changed this observation (data not shown). 185 We also noted a propensity for LPrev i to produce larger sized and more abundant 186 EBNA-LP isoforms, and that our EBNA2 knockouts exhibit higher EBNA-LP protein 187 levels ( Fig S4). However, overall it appears that in established BL31 cell lines, knockout 188 of EBNA-LP does not alter the protein levels of other EBV latency genes.  (Fig 2A). However, these aggregated cell clumps expanded over the next few 197 days in the wild-type-infected cells whereas LPKO i and YKO cell lines lagged behind in 198 outgrowth ( Fig 2B). We also noted that the expansion of both of the LPrev i -infected cells   LPrev i viruses induced more proliferation than LPKO i but less than the wild-type 224 controls. These observations were consistent for both LPKO i /LPrev i pairs.

225
This suggests that while EBNA-LP contributes to transformation, it is not required 226 for many of the activation functions fulfilled by EBNA2, as the E2KO-infected cells were 227 apparently as inert as uninfected cells. However, there is also some defect in the LPrev i 228 viruses that may also compromise the function of LPKO i . This might have been due to 229 the intronic mutation of the BsmBI restriction site, but resequencing the repeat unit used

245
In order to correct for the intronic defect of LPrev i , and assess whether it also 246 altered the behaviour of LPKO i , we isolated an IR1 repeat unit that matched the B95-8 247 consensus sequence, and used it to generate two new repeat arrays -one wild-type 248 and a second consisting the LPKO mutation described in Fig 1A - the small intron ( Fig S8A): all of the IR1 sequence (other than the defined EBNA-LP 251 mutations) matched the published B95-8 sequence. Both of these repeat arrays were 252 recombined into the IR1-knock-out that had been used to generate LPKO i .2 to make 253 two independent LPKO w BACs (where 'w' indicates wild-type IR1 backbone) and one 254 with a wild-type repeat with no heterogeneity (WT w ) ( Fig S1C). These were validated by 255 pulsed field gel electrophoresis ( Fig S8B) and used to generate virus-producing cell WT w matches (and perhaps exceeds) the transforming capability of the parental wild-260 type BAC and revertants, and is considerably superior to LPrev i . More interestingly, 261 LPKO w is superior to LPKO i in driving infected B cells to undergo proliferation, 262 approaching the level seen for YKO, suggesting that many of the important functions 263 lost in LPKO w are also missing in the YKO infection.

264
Infection of 10 6 B cells with LPKO w at an MOI of 1 rgu/cell consistently induced 265 expansion for approximately 5-7 days, but then appeared to stagnate for the next 1-2 266 weeks, after which cells usually proliferated again, and subsequently established LCLs.

267
The other viruses with reduced transformation efficiency -YKO and LPrev i -did not   285 We also infected mononuclear cells from umbilical cord blood to try to establish 286 LCLs. However, we were repeatedly unable to establish cord blood LCLs with either  In order to quantify this effect, we conducted a dilution cloning experiment 298 comparing transformation of blood from umbilical cord with blood taken at the same 299 time from the baby's mother. This was performed for three donor pairs, and on each 300 occasion, both LPKO w and YKO viruses consistently failed to transform cord blood, 301 despite successfully transforming the maternal cells into LCLs (Fig 4B). In contrast, both 302 the wild type viruses (WT w and Yrev/EBV-BAC) and LPrev i showed no difference in 303 transformation efficiency between cord and maternal lymphocytes. We also observed 304 that WT w -infected cells expanded and acidified the media faster than WT-HB9 and Yrev 305 transformations, but produced the same number of LCL-initiating events (not shown).

306
This demonstrated that the defect in transformation of cord blood is due to the EBNA-LP 307 mutation, and is not a consequence of a generically reduced transformation 308 competence.

309
In order to assess whether the cord cell phenotype was linked to the naïve 310 phenotype of these cells, we used CD27 and IgD status to sort CD19+ve adult cells into   to wild-type levels ( Fig 5 and Fig S10).

348
For EBNA transcripts, EBNA-LP makes no difference to Wp activity (Fig S10A), 349 whereas Cp activity is generally lower in EBNA-LP mutants ( Fig S10B). Transcripts for infection with EBNA-LP mutants ( Fig 5G, Fig 5H, Fig S10I). This is the opposite effect to  Cp promoters between wild-type and EBNA-LP knockout viruses in LCLs or during 387 primary infections (not shown). Since EBNA2 has been repeatedly shown to regulate 388 and bind at these genes, the binding of EBNA2 to both known binding sites and 389 negative control sites was assessed across three 30 day infection time courses.

390
Differences in EBNA2 binding were sometimes detectable on day 2 (not shown) but the 391 ChIP showed a much better signal to noise ratio on day 5 post infection. There is a 392 profound delay in the recruitment of EBNA2 to known transcription factor binding sites 393 at the LMP2A and LMP1/2B promoters on the LPKO w genome compared to WT w (Fig 6   394 and Fig S11). EBNA2 recruitment to its binding site at Cp was modestly reduced in 395 LPKO w infections, but still showed a considerable binding signal. In contrast, EBNA2 396 was efficiently recruited to host genes IL7 and HES1. Indeed, the LPKO w infection 397 consistently showed elevated binding on days 5 and 9, but not at other time points (Fig   398   6B).

399
EBNA2 does not bind directly to DNA, but rather it is directed to many of its 400 binding sites by host proteins, in particular the transcription factor RBPJ (also called 401 RBP-Jκ and CBF1). We therefore also assessed binding to these locations. In our time 402 course, RBPJ binding peaked later than EBNA2 binding, and was slightly (but     Biologically, a number of differences have been reported that separate the 520 behaviour of memory and naïve cells. We noted a slower outgrowth of LCLs from naïve 521 than from memory B cells, although this was not seen in a previous study [37].    newly assembled IR1 repeats containing EBNA-LP mutations. 706 We have used two distinct methods to generate the synthetic IR1. Both 707 approaches generate an IR1 with 6.6 copies, which is a typical size for circulating EBV 708 strains [31] and is the size of IR1 in the parental EBV-BAC clone, WT-HB9. In both 709 cases, the IR1 was assembled in a pBR322-based plasmid in DH5alpha bacteria grown 710 at 30 o C to reduce unwanted recombination.

711
The first approach used to assemble a modified IR3 adapted a strategy that used  wild-type sequence into the knockouts by the same method (Fig S1).

746
BACs were screened for integrity using EcoRI, AgeI, HindIII, NotI and BamHI 747 restrictions digests and run on a CHEF DRII chiller pulsed filed gel electrophoresis 748 system (Bio-Rad). We noted that the family of repeats (FR) region of oriP is smaller in 749 WT-HB9 than predicted by sequence. This reflects a previous observation that the   (Table ST2).