Virology 400 (2010) 53–67

Contents lists available at ScienceDirect

Virology

j ourna l homepage: www.e lsev ie r.com/ locate /yv i ro

Cellular factors associated with latency and spontaneous Epstein–Barr virusreactivation in B-lymphoblastoid cell lines

Michael L. Davies a, Shushen Xu a, James Lyons-Weiler b, Adam Rosendorff c, Steven A. Webber d,Laura R. Wasil a, Diana Metes e, David T. Rowe a,⁎a Department of Infectious Diseases and Microbiology, Graduate School of Public Health, University of Pittsburgh, 435 Parran Hall, 130 DeSoto Street, Pittsburgh, Pennsylvania, USAb GPCL Bioinformatics Analysis Core, University of Pittsburgh, 3343 Forbes Avenue, Pittsburgh, Pennsylvania, USAc Department of Pathology, Children's Hospital of Pittsburgh of UPMC, Rangos Research Center, 530 45th Street, Pittsburgh, Pennsylvania, USAd Department of Pediatrics, University of Pittsburgh School of Medicine and Children's Hospital of Pittsburgh of UPMC, Pennsylvania, USAe Thomas E. Starzl Transplantation Institute and Department of Surgery, School of Medicine, University of Pittsburgh, Pittsburgh 15213, Pennsylvania, USA

⁎ Corresponding author. Fax: +1 412 383 7490.E-mail address: rowe1@pitt.edu (D.T. Rowe).

0042-6822/$ – see front matter © 2010 Elsevier Inc. Adoi:10.1016/j.virol.2010.01.002

a b s t r a c t

a r t i c l e i n f o

Article history:Received 5 October 2009Returned to author for revision10 November 2009Accepted 4 January 2010Available online 11 February 2010

Keywords:Epstein–Barr virusHerpesvirusLymphoblastoid cell linesLatency IIILytic reactivationB cellsB cell differentiationUnfolded protein response

EBV-immortalized B-lymphoblastoid cell lines are used as models for cellular transformation and as antigen-presenting cells in immunological assays. LCLs vary in surface markers and other phenotypic properties, butit is not known how this heterogeneity relates to the EBV life cycle. To explore correlations, we examined 62LCLs for cellular and viral phenotypes. LCLs generated from pediatric and adult donors could similarly becategorized as either low in EBV copy number or fluctuating within a high range. High-copy statusaccompanied higher lytic viral gene expression and lower latent gene expression. Inhibiting lytic EBVreplication did not affect cellular phenotype or lytic switch protein expression, indicating that an LCL's lyticpermissivity was a stable property. Among the cellular genes overexpressed in permissive LCLs wereunfolded protein response genes and plasma cell markers. Among genes overexpressed in non-permissiveLCLs were transcription factors involved in maintaining B cell lineage, in particular EBF1. This study suggestspreviously undetected mechanisms by which cellular pathways influence the lytic reactivation of EBV.

ll rights reserved.

© 2010 Elsevier Inc. All rights reserved.

Introduction

Epstein–Barr virus is a human gammaherpesvirus that infectsepithelial cells and B cells, and is distinguished from otherherpesviruses by its ability to induce proliferation and expansionof B cells. In vivo EBV can cause mononucleosis during primaryinfection, characterized by massive proliferation of lymphocyteswhich then leads to viral persistence in the B lymphocytesthroughout one's lifetime. EBV persistence has been associatedwith tumors including Burkitt's lymphoma and Hodgkin's disease,and less conclusively with autoimmune diseases, particularly SLE.Like most human herpesviruses EBV can induce complications inimmunosuppressed people, after primary infection or duringreactivation from latency; unlike most other herpesviruses (theexception: HHV-8), EBV has the unusual ability to cause tumors inthese patients (Jenkins et al., 2003). Upon infection of a B cell, EBVexpresses six Epstein–Barr nuclear antigens (EBNAs) and two latentmembrane proteins (LMPs); this is known as the Latency III profile.

B cell proliferation requires functions encoded in most of the EBNAsand contributions from LMP1 and LMP2A proteins that mimic CD40and B cell receptor signaling respectively (Rickinson and Kieff,2007). In vivo Latency III-transformed cells are quickly recognizedand killed off by cytotoxic T cells, although a minority of infectedcells survive, escaping detection by switching from Latency III tomore limited Latency I or II profiles of viral gene expression.Without effective immune surveillance, Latency III-transformed Bcells can proliferate, and in immunosuppressed transplant recipientsthis can lead to post-transplant lymphoproliferative disorder (PTLD)(Dolcetti, 2007).

In vitro EBV infection is a convenient way to produce B-lymphoblastoid cell lines (LCLs), which can be used as targets forassays of cytotoxic T cells, as models for immortalization andsenescence, or as sources of patient-specific cells for genotyping.LCLs are often used as models for PTLD (Markasz et al., 2006), as theyhave similar morphology and phenotype involving high expression ofactivation and adhesion markers, and grow readily in SCID/hu mice.LCLs have been characterized for properties such as time to clonality,rate of episome multiplication early in proliferation, frequency of Igisotypes, and the mutations and selection that eventually lead to trulytumorigenic cell lines (Ryan et al., 2006; Sugden et al., 1979; Sugimoto

54 M.L. Davies et al. / Virology 400 (2010) 53–67

et al., 2004). In general, LCLs are considered to be more similar toCD38− CD23+ PTLDs derived from germinal center-experienced“bystander” B cells than the CD38+ CD23− PTLDs that originatefrom centroblasts (Rochford et al., 1993). LCLs are phenotypicallymore heterogeneous than most studies presume, though they arefrequently treated as interchangeable. One study disagreed with theconsensus (CD10− CD19+ CD20+ CD21+ CD23+ CD30+ CD40+

CD40L− CD77− Fas+ OX40− OX40L−, IL-6 secreting) profile of LCLs(Durandy et al., 1997; Gregory et al., 1991; Shinozaki et al., 2001) bydescribing a CD19−CD20− subgroup, which also produced lesssoluble Ig and negligible amounts of the paracrine growth factorsIL-6 and IL-10 (Wroblewski et al., 2002). LCLs vary significantly insurface expression of CD38 and other markers (Dolcetti et al., 1999;Khanolkar et al., 2003; Subklewe et al., 2005), and the nature of theirmalignant mutations (Takahashi et al., 2003). These earlier findingsmay suggest that LCLs should not be used interchangeably in clinicalstudies.

Lytic reactivation is generally induced in vitro by the addition ofhistone deacetylase (HDAC) inhibitors, which enhance acetylation ofthe histones associated with the EBV episome, decondensing itschromatin structure and remodeling the promoters for the Z and Rgenes that encode the lytic switch proteins BZLF1/ZEBRA and BRLF1/Rta (Countryman et al., 2008). These proteins are uniquely necessaryto induce the cascade of lytic EBV replication, which involves theproduction of early viral proteins, followed by replication of viral DNAby viral DNA polymerase, and production of late viral proteinsincluding the structural proteins that make up the viral particle.Several human transcription factors are known to influence activationof the Z promoter; two recent studies (Bhende et al., 2007; Sun andThorley-Lawson, 2007) added to this list XBP-1, a protein that inducesB cells to exit either a proliferating or quiescent phenotype andterminally differentiate into antibody-secreting plasma cells (PCs).XBP-1-mediated induction of ZEBRA expression confirms a series ofassociations between B cell terminal differentiation and lyticreactivation which began with the findings of Wendel-Hansen et al.(1987) that in a given LCL, a non-proliferating minority of cellscontain high levels of cytoplasmic Ig and low levels of latent EBVantigens. This finding also fits into a model for in vivo viruspropagation over decades. In this model, latent virus is maintainedin quiescent memory cells in the blood that, when they becomeactivated, traffic to the secondary lymphoid organs and differentiateinto PCs. Concomitantly, they activate the lytic program and makevirus particles, leading to the infection of bystander B cells in thoseorgans (Thorley-Lawson and Gross, 2004).

Gene expression profiling has been used in vitro to investigatethe effects of EBV infection on both epithelial cells and Blymphocytes. Studies have looked at genes influenced by LatencyIII infection, Latency I infection, expression of isolated EBV proteins(LMP1, LMP2A), infections using knockout strains of virus lackingEBNA2 or EBNA3B/3C, and the effects of inducing EBV lyticreplication in Latency I-type cells (Cahir-McFarland et al., 2004;Carter et al., 2002; Chen et al., 2006; Morris et al., 2008; Portis andLongnecker, 2003; Spender et al., 2006; Yuan et al., 2006; Zhaoet al., 2006). Microarrays have been used clinically for manydiseases including multiple myeloma, Hodgkin's disease, PTLD, NPC,and SLE to search for gene expression profiles that correlate withacute and/or EBV-positive cases. In this study we characterized howLCLs' properties correlated in their permissivity for lytic reactiva-tion. We tested the hypotheses that LCLs which differentiate towarda PC phenotype become permissive for EBV reactivation, and thatcell lines with fewer PC-like properties will experience less EBVreactivation. Representative permissive and non-permissive LCLswere examined by microarray gene expression profiling, andconfirmed with quantitative real-time RT-PCR, to identify severalgenes whose expression levels were associated with higher or lowerlevels of lytic permissivity.

Results

Levels of EBV DNA/cell correlate with spontaneous lytic reactivationwithin a characteristic range in both adult- and juvenile-derived LCLs

We hypothesized that LCLs vary in permissivity for EBV reactiva-tion in a manner influenced by the ontology of the cells that gave riseto the LCL. We measured the EBV DNA content in 62 independentlyestablished LCLs which could be divided into two groups (adult orpediatric) depending on the age of the B cell donor. In both groups themajority of cell lines had b200 copies of EBV DNA/cell with someranging above 500 copies/cell, when measured at a single time point(Fig. 1) LCLs derived from healthy adult donors had a mean value of254.4 (95% CI 151.7–357.2) and a median value of 124.0. LCLs derivedfrom the peripheral B cells of pediatric transplant candidates had amean of 193.4 copies of EBV DNA/cell (95% CI 132.6–254.2) and amedian value of 123.8. LCLs from adult and pediatric donors were notstatistically different in EBV DNA content, either in this cross-sectional measurement or in longitudinal observations (not shown).Indeed, we found no significant differences regardless of how the LCLswere stratified (age, gender or, for adults, HIV status). Of the 62 celllines, 24 (38%) had N200 copies of EBV DNA/cell and, upon furtheranalysis, were eventually classified as permissive for spontaneousEBV lytic reactivation.

Permissive and non-permissive LCLs were also subcloned atlimiting dilutions and measured for EBV DNA/cell to confirm thattheir lytic permissivity phenotype was shared by the daughter celllines (Fig 1C). In addition to resembling their parent lines inmorphology and growth rate, all sublines from the low-copy lineswere low-copy like the parent, and only three of 22 sublines from oneof the high-copy lines were measured as low-copy. That subcloningdid not affect the phenotype was expected since all our LCLs had beengrowing in culture for at least 3 months prior to use in theseexperiments. This is in agreement with the study of Ryan et al. (2006)showing that LCLs reach clonality or stable biclonality within 2months of establishment. However, as our aim was to characterize aset of LCLs that were representative of those widely used in clinicaland scientific assays, our cell lines were not all subclones. Typically,LCLs are not normally subcloned or investigated for clonality beforebeing used in the laboratory.

It has been previously shown that within 2 months post-infectionthe number of latent EBV episomes per cell in LCLs commonlyincreases from 1 or 2 per cell to dozens through mechanisms notrelated to classical episome maintenance (replication once per cellcycle in S phase from oriP) (Kieff and Rickinson, 2007). Since mostLCLs undergo some level of spontaneous lytic reactivation, and thislytic permissivity varies among LCLs (Adhikary et al., 2007; Keatinget al., 2002), when LCLs are used in therapeutic protocols they areroutinely cultured in the presence of acyclovir (a nucleoside analoguethat inhibits herpesviral DNA polymerase and prevents lytic, but notlatent, viral replication) to suppress the production of infectious virus(Bollard et al., 2004). To investigate whether the high levels of EBVDNA in some LCLs were the consequence of lytic reactivation, we grewLCLs in the presence of acyclovir (ACV). When grown with ACV, allLCLs had their levels of EBV DNA/cell reduced below 100 (data fromfour representative LCLs shown in Fig. 2A). Thus, for LCLs maintainedwithout lytic inducers or lytic-inhibiting drugs, the amount of EBVDNA/cell is a marker for lytic permissivity of the population. As LCLsgrow in culture and undergo spontaneous lytic replication, a typicalLCL might be expected to gradually increase its number of episomes/cell as a result of re-infection by virus in the supernatant.Alternatively, as the cells in each line that are disposed to enter alytic replication cycle reactivate and die off, the cell line mightdecrease over time in its lytic permissivity. Neither trend wasobserved (Fig. 2). When grown without ACV, each LCL with greaterthan a baseline level of lytic permissivity showed levels of EBV DNA/

Fig. 1. Sets of proliferating LCLs from pediatric and adult donors have comparable levelsof EBV DNA content. LCLs generated from pediatric transplant candidates (n=32) andfromhealthy adults (n=30)weremaintained forN2months andmeasured for EBVDNAcontent 2 days after being split to equal densities. (A) In each group, the majority of celllines had b200 copies of EBV DNA/cell, while some ranged above 500 copies/cell. (B)Boxes show the 25th and 75th percentiles; diamonds represent values outside 1.5× theinterquartile range. (C) Four LCLswere subcloned, with subclones allowed to diverge for6 weeks before beingmeasured for EBV DNA/cell. White diamonds represent subclones(mean of 2 time points), while black diamonds represent parent lines (mean of 7 timepoints). Bars indicate geometric means of subclones (n=18–22) for each LCL.

55M.L. Davies et al. / Virology 400 (2010) 53–67

cell that fluctuated significantly but stayed within a certain range.These transient increases and decreases did not correlate with thedays on which cultures were passaged, or with the density of cells inthe media. Although the level of EBV DNA/cell in each LCL fluctuatedfrom day to day, there was no steady trend upward or downward overtime.

Cell lines with consistently high levels of EBV reactivation (so-called permissive LCLs) might be engaged in a self-reinforcingreactivation cycle that could be broken by the addition of acyclovir.To test this, we cultured permissive and non-permissive LCLs for14 days in ACV and then removed it and continued culturepassage. Non-permissive LCLs, represented by LCL21 (Fig. 2B),maintained the low-copy phenotype before, during and after theapplication of ACV. Permissive LCLs, represented by LCL23 (Fig. 2C),had their EBV DNA/cell reduced to baseline level when grown inACV, but immediately after the removal of ACV returned to theirhigh-copy phenotype. This suggests that the rate at which spon-taneous EBV lytic reactivation takes place in an LCL is influencednot by the number of viral episomes or by ongoing virus pro-duction, but by cellular factors whose expression is a characteristicof that cell line.

All LCLs show a Latency III pattern of gene expression with a detectablelevel of lytic reactivation

To characterize LCLs for their lytic permissivity, and to confirmthat they had comparable Latency III profiles of viral geneexpression, we quantified 15 viral lytic and latent mRNAtranscripts using real-time RT-PCR. For a subset of LCLs (n≥21),we measured the level of EBV DNA/cell for 5 days, with RNAextracted on the third day. Every LCL expressed transcripts for thefull complement of EBV Latency III genes (Fig. 3). Every LCL alsoexpressed detectable levels of the transcripts for three early lyticgenes (BZLF1, which produces the lytic switch protein ZEBRA;BMLF1, which produces the abundant nuclear EB2 protein; and thelytic F promoter, which represented the majority of transcriptsdetected using primers specific for sequences in the downstream Qregion). The expression of these lytic mRNAs correlated positivelywith the number of EBV genomes in an LCL (Fig. 3B–D). Significantlinear correlations were also found between the number of EBVgenomes/cell and the percentage of cells expressing early (ZEBRA)or late (gB/gp110) lytic viral proteins as detected by flowcytometry (Supplementary Fig. 1). These data show that in lyticallypermissive LCLs the entire process of lytic reactivation is occurring,rather than an abortive expression of early lytic proteins and EBVDNA replication.

The expression of latent EBV genes in an LCL correlates negatively(but weakly) with its permissivity for lytic reactivation

We also measured the relative expression levels of all six EBNAgenes and two LMP genes in the same subset of LCLs (n≥21). All eightwere expressed in every LCL; themost abundant transcript was LMP1,and the least abundant was EBNA3A (Fig. 3A). Fig. 4 shows the level ofmRNA for each latent gene plotted against the level of EBV DNA/cell.For EBNA1, -2, -3A, -3B, -3C, -LP, LMP1, and LMP2B, the relationship issimilar and slightly negative, non-linear, and attributable to aminority of non-permissive LCLs that contain high levels of thetranscript in question. Only EBNA3A (Fig. 4D) and EBNA3C (Fig. 4F)generated a significant negative linear regression (pb0.05) betweengene expression levels and EBV DNA/cell.

The only latent gene product whose expression positivelycorrelated with the level of lytic EBV DNA/cell was LMP2A (Fig. 4H).This relationship is not as strong as the correlation between EBV DNAcontent and BZLF1 and BMLF1 gene expression, and it resembles therelationship between EBV DNA content and the other latent genes, inthat non-permissive LCLs show a wider range of gene expressionlevels than permissive LCLs. No inverse relationship was detectedbetween LMP1 and LMP2A expression, even though LMP1 is known torepress lytic reactivation in its role as a CD40 mimic (Adler et al.,2002), while LMP2A signaling can downregulate the expression ofLMP1 (Stewart et al., 2004).

Fig. 2. In some LCLs, EBV DNA/cell fluctuates within a certain range and is reduced to a baseline level by the antiviral drug acyclovir. (A) Four LCLs were grown in complete media andpassed regularly to equal densities. The number of EBV genomes/cell in each LCL did not remain stable or steadily increase over time. In the presence of acyclovir EBV DNA/cellreached a stably low level, with no effect on cellular morphology. Non-permissive LCL21 (B) and highly permissive LCL23 (C) were grown inmedia without ACV from days 0 to 20. Onday 4, each LCL was split into two flasks, with or without ACV. On day 14, each LCL growing under ACV was split into two flasks, with or without ACV. Upon removal of ACV, LCL23returned to its high-copy phenotype.

56 M.L. Davies et al. / Virology 400 (2010) 53–67

The C and W promoters are both active in LCLs

After months in culture LCLs are expected to have completed theshift from using the W promoter (Wp) to using the C promoter (Cp)for expressing EBNA genes. RT-PCR primers were designed to detectthe 5′ ends of polycistronic EBNA transcripts and amplify RNAs withC2 toW1 splices (representing transcription from the Cp), orW0 toW1

Fig. 3. Viral gene expression in LCLs. Cell lines (n≥21) were measured for EBV DNA and RNreal-time RT-PCR. Each of nine latent transcripts and three lytic transcripts was detectable inDNA content. Levels of EBV DNA/cell (mean of 5 time points) were plotted against levelsrepresenting the lytic transcripts BZLF1/ZEBRA (B); BMLF1/EB2 (C); and F promoter (D).variation adjusted for sample size.

splices (representing transcription from the Wp). Importantly,alternative splices between these exons determine whether themRNA contains an EBNA-LP start codon (LP+) or no EBNA-LP startcodon (LP−) Therefore, two sets of 5′ primers were used todistinguish LP+ from LP− transcripts (Supplementary Table 1).EBNA-LP expression level was estimated by adding together thedetected levels of C2:W1 and W0:W1 LP+ splices. LP− splices were

A content after growing for N3 months. (A) Levels of mRNA/cell were measured usingevery LCL measured. (B–D) Permissivity for EBV lytic reactivation correlates with EBV

of EBV mRNA (1 time point, taken 2 days after cells were split to a common density)Statistics shown are the p-value derived from Student's t test, and the coefficient of

Fig. 4. No strong correlation exists between EBV latent gene expression and EBV DNA content. LCLs (n≥17) were measured for EBV DNA and RNA content after growing for N3months, as described in Fig. 3. Gene expression wasmeasured for EBV nuclear antigens 1 (A), 2 (B), leader protein (C), 3A (D), 3B (E), and 3C (F), and EBV latent membrane proteins 1(G), 2A (H) and 2B (I). Statistics shown are as in Fig. 3. Linear regressions are shown when p≤0.05.

Fig. 5. Both W and C promoters are actively making EBNA transcripts in all LCLs. LCLs(n=22) were measured for EBV DNA and RNA content after growing for N3 months, asdescribed in Fig. 3. We detected polycistronic EBNA transcripts representing four splicevariants—either C2:W1 or W0:W1, and either containing or not containing an EBNA-LPstart codon. Each LCL contained all four splice variants. Non-permissive LCLs (b200 EBVgenomes/cell) contained more transcripts of each splice variant, as with all other latenttranscripts except LMP2A (Figs. 3–4).

57M.L. Davies et al. / Virology 400 (2010) 53–67

presumed to encode at least one downstream EBNA protein, but notEBNA-LP. As predicted, the RNA levels for the 5′ ends of EBNAtranscripts (sum of the Cp and Wp cDNAs) were similar to the RNAlevels for the 3′ ends of EBNA transcripts (sum of the EBNA ORFcDNAs). This concordance suggested that the efficiency of the reversetranscription did not grossly favor some transcripts over others.Although C transcripts were more abundant than W transcripts, allfour targets were detected in every LCL (Fig. 5). There was nocorrelation between time in culture and having a greater ratio of Cpromoter to W promoter activity (data not shown). W and Cpromoters appear to share the property of being more highlyexpressed in non-permissive LCLs, and this relationship reachesstatistical significance for two of four transcripts (Fig. 5). In agreementwith recent findings (Elliott et al., 2004), our data indicate that theWpis never totally silenced even after more than 6 months in culture.

Inhibition of lytic replication does not affect the rate of spontaneousBZLF1 induction

Varying levels of lytic replication in LCLs might not be directlyrelated to the levels of ZEBRA induction, but instead correlate moreclosely with the level of success in translating that initial lytic switchinto full-fledged viral replication accompanied by production of latelytic proteins. To characterize spontaneous reactivation at the level ofindividual cells rather than in pooled populations, we used flowcytometry to detect the number of cells in each cell line expressing

lytic virus proteins, and the level of expression of these proteins, whengrown with or without acyclovir. Just as with lytic mRNA expression,there was a linear relationship between the number of EBV lytic

58 M.L. Davies et al. / Virology 400 (2010) 53–67

genomes in a cell line and the number of cells producing both early(ZEBRA) and late (gp110) lytic proteins (Supplementary Fig. 1). Thereactivating cells in non-permissive LCL38, though less numerousthan those in permissive LCL01, stain just as brightly for ZEBRA andgp110, indicating that although LCLs vary in the fraction of cells thatswitch into lytic reactivation, the cells that do so produce similarlevels of lytic proteins (Fig. 6). The ratio of double-positive (ZEBRA+

gp110+) cells to single-positive (ZEBRA+) cells was consistentlybetween 10% and 30% in both permissive and non-permissive LCLs,indicating that although LCLs vary in permissivity for ZEBRAinduction, they are similar in their ability to undergo the full cycleof lytic replication once ZEBRA is induced.

Permissive LCLs, maintained in the presence of ACV for 4 weeks,were compared with the same LCLs grown with no antiviral drugs.ACV greatly reduced the production of gp110, but had no effect oneither the number of cells producing ZEBRA or the level of ZEBRAproduced in those cells (Fig. 6A). Flow cytometry analysis for twopermissive LCLs grown with or without ACV at four time points foundthat ACV did not affect the number of cells producing ZEBRA (Fig. 6C),but did reduce the number of ZEBRA+ gp110+ double-positive cells,expressed both as a percent of the total cell population (Fig. 6D) andas a percent of the ZEBRA+ cells (Fig. 6E). This further suggests thatthe “high-copy” status of an LCL is a stable phenotype characterized bygreater permissivity for BZLF1 activation, and is independent ofwhether lytic replication of the genome actually occurs.

Profiling LCLs by their display of surface markers

We hypothesized that high rates of lytic permissivity might beassociated with LCLs of a non-memory cell lineage, or with those thatexperience more spontaneous differentiation toward PC status. Also,LCLs with reduced lytic reactivation might have lost certain aspects of

Fig. 6. Permissive LCLs contain more cells expressing lytic EBV proteins. (A, B) Permissiveprotein ZEBRA(Z) and the late viral protein gB/gp110. In cells grown without acyclovir,representative of 9 LCLs at 4 time points. (C, D, E) Quantification of flow scatterplots. Two perexpression. (C) The ratio of ZEBRA-positive cells to total cells at each time point. (D) The rapositive cells to ZEBRA-positive cells. Even long-term treatment with ACV only reduces the n

the B cell phenotype, such as surface immunoglobulin expression.Panels of fluorescently labeled antibodies were used to characterizethe LCLs for numerous cell surface markers, including CD138/Syndecan1 and CD38 (plasma cells); CD27 (memory and plasmacells); B220 (naive cells); CD23 and CD69 (activated B cells); CD30(activated B cells and some tumors); HLA class I and class II; pan-B cellmarkers CD20, BAFFR/BR3, and CD45; CD40 (a pan-B cell marker andsignaling molecule mimicked by LMP1); CD19 (pan-B cell marker andpart of the B cell receptor complex); isotypes IgM, IgD, IgG and IgA;and a reagent that stains all human antibodies (goat pan-Ig). NineteenLCLs were maintained without ACV and measured for all of thesemarkers.

Although there were examples of LCLs having low levels of onesurfacemarker or another (CD19, CD40, CD45, HLA II), and they variedin intensity of CD38, CD27, and CD138, there was no coherentphenotypic pattern (e.g. enhanced PC differentiation) that could beassociated with EBV reactivation. Linear regression of each surfacemarker suggested a negative correlation between EBV DNA contentand the percent of cells positive for CD45 (p=0.006, R2adj=33.3%),CD38 (p=0.020, R2adj=23.5%), and CD40 (p=0.031, R2adj=20.0%), and a positive correlation between EBV DNA content andthe percent of cells positive for B220 (p=0.017, R2adj=25.0%).When we used mean fluorescence intensity of the cells, rather thanpercent positive, as the metric, the only correlation suggested bylinear regression was a positive one between B220 and EBV DNAcontent (p=0.035, R2adj=19.0%). Table 1 shows that there was noassociation between EBV DNA content and class-switched or IgM+

IgD+ status, and no consistent correlations with other phenotypicmarkers. For eight cell lines examinedwith andwithout ACV, the drughad no effect on surfacemarker expression, confirming that long-terminhibition of EBV lytic replication did not impact a cell line's surfacemarker phenotype (data not shown).

LCL01 and non-permissive LCL38 were stained for expression of the immediate earlythe proportion of cells expressing Z and/or gB is greater in 01 than in 38. Data aremissive LCLs, grownwith or without ACV, were stained on 4 days each for ZEBRA and gBtio of ZEBRA/gB-double-positive cells to total cells. (E) The ratio of ZEBRA/gB-double-umber of cells expressing late lytic genes—not the % of cells initiating lytic reactivation.

Table1

Cellsu

rfaceph

enotyp

eis

notsign

ificantly

correlated

withpe

rmissivity

forEB

Vreactiva

tion

.

LCL

DNA/

cell

IgA

IgM

IgD

IgG

pan-

IgHLA

IIHLA

ICD

19CD

20BR

3CD

40CD

45CD

30CD

27B2

20CD

23CD

69CD

138

CD38

FITC

APC

Pacific

Orang

ePE

FITC

PE-Cy7

Pacific

Blue

PE-Cy5

.5APC

-Cy7

PEPa

cific

Orang

ePa

cific

Blue

A64

7APC

-A75

0PE

-Cy7

A64

7Pa

cific

Blue

PEPa

cific

Orang

e

QW

171.4

/+++++

+++

+++++

+++++

+++++

+++++

+++++

+++++

++++

+++++

+++

+++++

++++++

++

++++

0713

84.4

−+++++

/+++

++++

+++++

+++++

/++++

+++++

++++

+++

++

/+++++

+++

/14

1081

.0−

++++

−++

++++++

+++++

+++++

++++

+++++

++++

++

++

+++

+++

+++++

++

−33

1570

.9−

++++

−++

++++++

+++++

+++++

+++++

+++++

+++

+++

++

+++

−++

+++

+20

8.7

/++++

++

++++

+++++

+++++

+++++

+++++

+++++

+++++

+++++

+++++

+++

+++

/+++++

++++

++++

1740

.5−

++++

/++++

+++++

+++++

+++++

+++++

+++

+++++

++++

++++

++

/+++++

/+

/37

240.1

−+++

−/

−++++

+++++

+++++

+++++

+++++

+++++

+++++

+++

+++

/+++++

+/

−26

606.1

−++

+++++

+++

+++++

+++++

+++++

++++

+++++

+++++

++++

+++

+++

++++++

/++

+21

21.4

−−

−+++++

+++++

+++++

+++++

+++++

+++++

+++++

++

+++++

++++

++++

++++++

++

++

3876

.0−

−−

++++

++

+++++

+++++

+++++

+++++

+++++

+++++

+++++

+++

+++

−++++

/++

/72

274.9

−−

−+++++

+++

+++++

+++++

+++++

+++++

+++++

+++++

+++++

++++

++++

++

+++++

++

++

+10

301.4

−−

−++++

+++

+++++

+++++

+++++

++++

+++++

++++

+++

+++

++++

++

+++++

//

/23

342.5

−−

−+++

++++

+++++

+++++

+++++

+++++

+++++

+++++

++

++++

+++++

+++++

//

/06

963.5

−−

−+++++

++++

+++++

+++++

+++++

+++++

+++++

++++

+++

+++

+++++

++++

+++++

/+

−03

1014

.6−

−−

++++

++++

+++++

+++++

+++++

+++++

+++++

++++

++++

+++

++++

++++

+++++

/+

/TR

1177

.1−

−−

++++

+++++

+++++

+++++

+++++

+++++

+++++

++++

++

+++

+++++

+++

+++++

//

−32

1259

.6−

−−

++++

++

+++++

+++++

+++++

+++++

+++++

+++

++

++++

/+++++

+/

/76

1357

.6−

−−

+++

+++

+++++

+++++

+++++

+++++

+++++

++++

+++++

++++

+++++

+++

+++++

−+

+77

207.9

+++++

−−

/++++

++

+++++

+++++

++++

++++

++++

+++++

++++

++

/+++++

/+

++

85-100

%po

sitive

:+++++.6

5-85

%po

sitive

:++++.4

5-65

%po

sitive

:+++.3

0-45

%po

sitive

:++.1

5-30

%po

sitive

:+.5

-15%

positive

:/.

0-5%

positive

:-.

59M.L. Davies et al. / Virology 400 (2010) 53–67

Comparison of gene expression between permissive and non-permissiveLCLs by microarray

From the HapMap project it is clear that LCLs differ in theirexpression of a wide range of cellular proteins, as a result ofpolymorphisms (Zhang et al., 2008), so we used LCLs as a system todetect connections between cellular gene expression and EBVactivity in latently infected cells. We selected two permissive LCLs(LCL33 and LCL77) and two non-permissive LCLs (LCL21 and LCL38),maintained them in separate ACV+ and ACV− conditions for 2months, and isolated DNA and RNA 2 days after passage at the samedensity. cDNA was hybridized to Illumina HumanRef-8 v2 BeadChipscontaining probes for 20589 human mRNA sequences. For everyprobe, we first compared the average intensity of the 4 LCLs grownwith ACV with the average intensity of the 4 LCLs grown without ACV(Fig. 7A), and found no significant effects on any cellular genes. Wethen compared the average intensity of the permissive samples(LCL33 ACV+ and ACV−, LCL77 ACV+ and ACV−) with the averageintensity of the non-permissive samples (LCL21 ACV+ and ACV−,LCL38 ACV+ and ACV−), and found a much greater degree ofvariation (Fig. 7B), although the R2 value of 0.993 indicates howsimilar the phenotypes of the cells in this study are.

We used efficiency analysis of several transformations of the datato identify the process that would generate the most repeatable list ofgenes with significant absolute differences in expression betweenpermissive and non-permissive LCLs. Online caGEDA softwareindicated that the most consistent determination of significancecame from data that were untransformed and had been defined assignificant using the J5 formula. By this measure, the most repeatablethreshold for significance was J5=32.5, and 34 transcripts werefound to be above this threshold—20 overexpressed in permissiveLCLs, and 14 overexpressed in non-permissive LCLs (Fig. 7C).Supplementary Tables 2 and 3 list all the transcripts whose over-expression in either non-permissive (Table S2) or permissive (TableS3) LCLs led to a J5 score above 10.0. We also quantified the differencebetween permissive and non-permissive LCLs by measuring the folddifference, or ratio between the averages of the groups. Supplemen-tary Tables 4 and 5 describe the transcripts that were overexpressedby a fold difference above 1.75 in either non-permissive (Table S4) orpermissive (Table S5) LCLs.

Differentially expressed sets of genes in permissive andnon-permissive LCLs

In light of the small sample size usedwith this microarray, we usedstatistical analysis software not just to identify individual genes, butalso to seek out patterns of overexpression in entire groups of genesassociated with pathways, biological processes, or the curated resultsof other microarray studies. GSEA (Gene Set Enrichment Analysis)was used to analyze the results as a whole, ranking all genes bysignificance of their difference between the two groups of LCLs, andthen identifying gene sets that were overrepresented at eitherextreme of the ranking (Subramanian et al., 2005). Curated genesets associated with non-permissive LCLs included genes under-expressed in AIDS-associated primary effusion lymphoma (Klein et al.,2003); downregulated by p21 and p53 (Wu et al., 2002); over-expressed in CD10+ hematopoietic progenitor cells with B cellpotential (Haddad et al., 2004); and upregulated in multiple myelomacell lines after activation by N-ras or IL-6 (Croonquist et al., 2003).Curated gene sets associated with permissive LCLs included genesupregulated by p53, IFN-α or IFN-γ (Browne et al., 2001; Inga et al.,2002; Sana et al., 2005); overexpressed in all types of plasma cells invivo (Tarte et al., 2003); overexpressed in peripheral blood B cells oflupus patients (Bennett et al., 2003); and associated with the ARTvesicular trafficking pathway.

Motif-based gene sets associated with non-permissive LCLsincluded those with promoter sequences predicted to be activated

60 M.L. Davies et al. / Virology 400 (2010) 53–67

or repressed by transcription factors PAX8, PAX6, NF-IL3, FOXO1,EGR1/2/3, E2F1, CUTL1/CDP, and several microRNAs including theoncogenic miR-155, -221, and -222, and the anti-inflammatory miR-346 (Alsaleh et al., 2009; Lambeth et al., 2009; Teng andPapavasiliou, 2009). Motif-based gene sets associated with permis-sive LCLs were those with motifs for ATF-6, XBP-1, HTF-1, andSTAT5A, and microRNAs miR-338 and miR-518A/E/F. Althoughdozens of both motif-based and curated gene sets were enriched ata false discovery rate (FDR) b0.25 in non-permissive LCLs, only sevenmotif-based gene sets were enriched at FDR b0.50 in permissiveLCLs. The most enriched gene sets detected by GSEA are described inSupplementary Table 6.

DAVID, employing a canonical database of gene ontologies andother groups of genes such as KEGG pathways (Huang et al., 2009),

Fig. 7. Genes differentially expressed between permissive and non-permissive LCLs. (A) Fmaintained without acyclovir. No genes show significant differential expression between thwere compared with 2 non-permissive LCLs grown with and without ACV. Red points requantified for each gene using the J5 formula, and efficiency analysis predicted that themost33 genes had significance above 32.50.

was used to analyze lists of genes significantly overexpressed inpermissive or non-permissive LCLs and identify gene ontologiesthat contained significant genes. We used four different definitionsof significance (J5N10.0; J5N7.0; fold difference N1.75; folddifference N1.5) to create these lists, to see whether the genesidentified by one method would also be identified by others. Allgene ontologies that contained at least one significant gene, andwere associated with immunology or transcription factor activity,were collected. Similar ontologies (e.g. “transcription”, “transcriptionfactor activity”, and “regulation of transcription”) were combined,and all such ontologies are listed in Supplementary Table 7. Thesmall number of differentially expressed genes identified from thesecategories is expected from a comparison of cells with such similarphenotypes.

or all genes, 4 LCLs maintained with acyclovir were compared with the same 4 LCLse two populations. (B) For all genes, 2 permissive LCLs, grown with and without ACV,present genes with J5N32.50. (C) The absolute difference between populations wasrepeatable threshold for significancewas J5=32.50. Thirty four transcripts representing

Table 2Genes found by RT-PCR to be most differentially expressed between permissive andnon-permissive LCLs.

pa Fold change

pb .05 (non-permissive vs. permissive)CD40 b0.001 1.76EVL 0.001 1.84LRMP (Jaw-1) 0.005 1.98ETS1 0.006 1.38CD79B 0.009 1.54TNFSF13B (BAFF) 0.011 1.68CD99 0.021 1.38PAX5 (BSAP) 0.026 1.42ITGB (Integrin β2) 0.035 1.47RAC2 0.035 1.23TCF3 (E2A) 0.039 1.31NFATC1 0.048 1.67

Fold differenceN2.0 (non-permissive/permissive)IFI27 0.191 15.46EGR1 (Zif-268) 0.075 3.98CD38 0.114 3.38KLF2 0.077 2.11CCR7 (EBI1) 0.057 2.10GNA15 (G α15) 0.076 2.09

BothEBF1 0.002 3.25PTPRK 0.005 3.42ID2 0.005 4.23TSC22D3 (GILZ) 0.014 2.20CD69 0.031 2.63EBI2 0.038 2.03

pb0.05 (permissive vs. non-permissive)DNAJB11 0.012 1.59CCR10 0.015 1.49SDF2L1 0.016 1.77CALR 0.018 1.52ARMET 0.045 1.49

Fold differenceN2.0 (permissive/non-permissive)LEPREL1 0.067 4.53SDC1 (Syndecan) 0.101 2.85RAB31 (Rab22b) 0.155 2.11CD27 0.086 2.02

BothHSP90B1 (Grp94) 0.009 2.16

n=7 in each population.a From a 2-sample t test.

61M.L. Davies et al. / Virology 400 (2010) 53–67

Genes overexpressed in permissive or non-permissive LCLs, detected byquantitative RT-PCR

To further explore hypotheses suggested by themicroarray results,we subjected a larger group of LCLs (seven permissive and seven non-permissive) to an RT-PCR based assay for cellular gene expression.Weinvestigated 92 genes, which were identified by strongly differentialexpression in the microarray, by relevance to B cell or EBV biology, orboth. Seven of the selected genes (BCL11A, BCL6, CD40LG, CT45-4, HLA-DRB3, HLA-DRB5, U2B7) were not detected in the majority of samples.For the remaining 85 genes, we used the 2−ΔΔCt method to quantifyrelative gene expression, normalizing the results to an endogenouscontrol and then comparing that value to the value for a referencesample, LCL17. Each measurement was normalized twice, with eitherGAPDH or B2M as the endogenous control gene, and these twonormalized values were averaged together. The mean for sevenpermissive LCLs (mean EBV DNA/cell 42.9, SD 23.6) was comparedwith the mean for seven non-permissive LCLs (mean EBV DNA/cell1029.1, SD 520.6), and all 85 genes (plus two more endogenouscontrols, HPRT1 and ACTB) were characterized for significance of thedifference between permissive and non-permissive LCLs. All cellulargenesmeasured by RT-PCR are listed in Supplementary Tables 8 and 9,ranked both by p-value of the difference derived from a two-sample ttest (Table S8) and by relative fold difference between the twopopulations of LCLs (Table S9).

Table 2 shows all the genes whose differential expression wassignificant at a level of pb0.05, or a fold difference greater than twofold.Although the three PC-associated transcription factors investigated(XBP1; IRF4/MUM1; PRDM1/Blimp-1) were not overexpressed inpermissive LCLs, we did detect a difference in expression of SDC1,encoding the plasma cell surface marker Syndecan-1/CD138; andCCR10, encodinga chemokine receptor (theCCL27/CCL28 receptor) thatis upregulated on PCs and EBV-immortalized B cell lines, but not EBV+

Burkitt's lymphoma lines (Nakayama et al., 2002, 2003). The differencein CCR10 expression was small but consistent, while the high folddifference in SDC1 expressionwas the result of twodata points (Fig. 8B).Interestingly, although permissive LCLs did not show transcriptionfactor expression indicative of increased plasma cell differentiation,non-permissive LCLs overexpressed EBF1, PAX5, TCF3/E2A, and ETS1,which are all components of a transcription factor network thatmaintains the proliferating B cell phenotype and inhibits the XBP-1/Blimp-1 network (Fig. 8C). The three genesmost overexpressed in non-permissive LCLs were EBF1, ID2, and PTPRK. ID2 encodes a non-DNA-binding transcription factor thought to inhibit Pax-5 activity and in turnbe downregulated by EBF-1 and Pax-5 (Thal et al., 2009).PTPRK encodesa receptor tyrosine phosphatase that regulates activity of proteinsincluding Src and EGFR, and mediates anti-proliferative and pro-migratory effects of TGF-β (Wang et al., 2005; Xu et al., 2005). Table 2shows that these were the only genes that were both overexpressed asdetermined by significance of pb0.01, and overexpressed by folddifference greater than threefold, in non-permissive LCLs.

Five of the ten genes significantly overexpressed in permissiveLCLs (Table 2, right column) are associated with ER stress or theunfolded protein response (UPR). The UPR is a sequence of eventsinduced by accumulation of unfolded proteins in the ER, leading torapid breakdown of these proteins followed by enhanced new proteinsynthesis and sometimes autophagy. UPR genes suggested by themicroarray results and then measured using RT-PCR included ARMET,ATF6, CALR, DDIT4, DNAJB11, FKBP11, HSP90B1, HSPA8, PPIB, andSDF2L1, encoding ER proteins, and the downstream transcriptionfactors ATF4 and XBP1. Eight of the ten ER genes were overexpressedin permissive LCLs (Fig. 8A); ATF4 and XBP1 expression did not differbetween the two groups of LCLs, but correlated positively and linearlywith LMP1 expression (p=0.005 for ATF4, p=0.034 for XBP1).

One branch of the UPR leads to upregulation of XBP-1 mRNA, andanother leads to splicing of XBP-1 mRNA to produce its active XBP-1s

isoform. Although lytic permissivity did not correlate with XBP-1expression, we considered that it might correlate with its splicing. Tolook for such a relationship, we determined the relative quantities ofspliced and unspliced transcripts by SYBR Green PCR followed bydissociation curve analysis to distinguish the two isoforms. Norelationship was detected (Fig. 8D), suggesting that EBV reactivationdoes not induce the UPR despite being found at higher levels in celllines with more propensity for UPR activity.

Discussion

Lymphoblastoid cell lines are useful models for the processes ofEBV Latency III, for EBV-infected B cells in disease (e.g. PTLD), forliving genetic repositories and as HLA-matched immunologicaltargets. By profiling a large panel of LCLs we have investigated theextent to which latently infected B cells vary in permissivity for EBVreactivation, a property which influences long-term viral persistenceas well as transmission by the production of free virus in the tonsils.Spontaneous reactivation in the cultured LCLs was characterized byviral DNA replication and the production of early and late lyticproteins. Depending on the cell line, between 0.1% and 8% of cells in

62 M.L. Davies et al. / Virology 400 (2010) 53–67

an LCL were producing the lytic switch protein ZEBRA and, of theseproducers, about 20% made late structural proteins. In general, LCLscould be classified into two groups. One group was composed of“permissive” LCLs with high and fluctuating amounts of EBV DNA/cell (N200 copies/cell), lower levels of latent gene expression, andhigh frequency of lytic gene positive cells. The other group consistedof “non-permissive” LCLs that had low amounts of EBV DNA/cell(10–200 copies/cell), variable levels of latent gene expression, and alow frequency of lytic gene positivity. Statistical analysis suggested aweak negative relationship between the abundance of latent viralgenes in an LCL and its permissivity for lytic reactivation.

The only latent gene for which there was a positive correlation wasLMP2A. Although LMP2A may maintain latency by blocking cross-linked BCR from inducing lytic reactivation (Fukuda and Longnecker,2005), there is evidence that LMP2A downregulates Pax-5 and EBF-1expression, enhances differentiation into plasma cells, and induces Zpromoter activity, all factors associated with increased lytic reactiva-tion (Portis and Longnecker, 2003; Schaadt et al., 2005; Swanson-Mungerson et al., 2006). For every latent transcript, except LMP2A,the mean expression level in permissive LCLs was lower than that innon-permissive LCLs, although this difference was only statisticallysignificant for the EBNA3 genes. These findings do not support modelsin which EBV latency proteins restrict the induction of the lytic cyclein proliferating cells. Indeed, of all the latent genes measured, LMP1had the highest level of expression and by far the least difference inexpression between permissive and non-permissive LCLs, making itimplausible that variations in LMP1 expression account for permis-sivity towards EBV reactivation. All LCLs were immortalized with thesame strain of EBV (B95-8), which minimizes the possibility thatpermissivity related to virus strain variation. Lytic permissivity did notcorrelate with the amount of time spent in culture, and culturing LCLswith an inhibitor of lytic replication did not alter their permissivity. Allof the above analyses suggest that permissivity for spontaneous lyticreactivation was not influenced by viral factors, with the possibleexception of LMP2A.

Fig. 8. Differential expression between permissive and non-permissive LCLs determined byexpression of 90 cellular mRNAs, using RT-PCR and normalizing to both B2M and GAPDH. (genes. (C) Relative expression of transcription factors that maintain B cell lineage. (D) Permby SYBR Green RT-PCR followed by amelt curve used to detect the relative amounts of XBP-1permissive (open) groups, and asterisks indicate p≤0.05 from Student's t test.

It was clear from longitudinal analyses of LCLs that the differencein lytic permissivity was a stable property of an LCL. It wasreasonable to suspect that lytic permissivity could be due toproperties that were intrinsic to the founder B cell. Gross phenotypicproperties such as doubling time, clumping, or the morphology andsize of cells were not associated with lytic permissivity. Cellularfactors, whose expression is expected to be a stable characteristic ofa cell line, were investigated. Initially we analyzed expression of arange of B cell surface markers that define ontological status andobserved that only in permissive LCLs did more than 30% of the cellsstain positive for B220, a marker for naive primary B cells. Althoughthis was intriguing, there was not an accompanying low level ofCD27 and IgD on permissive LCLs, which would be expected for anaive cell lineage. We also investigated whether permissive LCLscontained a higher percentage of cells with the CD27hi CD38hi

phenotype characteristic of in vitro plasmacytoid differentiation, butthis was not the case.

If intrinsic cellular factor(s) were responsible, the variation wouldneed to be subtle. In addition to unpredictable epigenetic effects ongene expression in transformed cell lines, genetic polymorphisms canlead to varying expression levels of genes in otherwise comparablecells from different individuals. Genetic analyses have connectedsingle nucleotide polymorphisms (SNPs) in the IFNG, TGFB1, IL10 andIL1A genes with EBV-associated PTLD-like disease (Babel et al., 2007;Dierksheide et al., 2005; Hatta et al., 2007). EBV reactivation aftertransplantation has been correlated with an IFNG genotype thatproduces low basal levels of IFN-γ (Bogunia-Kubik et al., 2005). SNPsthat lead to a high basal level of IL-10 expression (da Silva et al., 2007),and SNPs in HLA-A (Niens et al., 2006), have been associated withsusceptibility to EBV+ HD. To learnmore about the patterns of cellulargene expression that may influence LCLs' permissivity for lyticreactivation, we first conducted an RNA microarray.

The results from gene set analysis of our microarray suggested thatgenes upregulated by IFN-γ and IFN-αwere overexpressed as a wholein non-permissive LCLs. IFNG itself is listed in Supplementary Table 7

RT-PCR. Permissive LCLs (n=7) and non-permissive LCLs (n=7) were measured forA) Relative expression of UPR/chaperone genes. (B) Relative expression of plasma cellissive LCLs (n=11) and non-permissive LCLs (n=9) were measured for XBP-1 splicings and XBP-1umRNA. Horizontal bars show themeans of the permissive (filled) and non-

63M.L. Davies et al. / Virology 400 (2010) 53–67

among the genes overexpressed in non-permissive LCLs, along withother cytokine genes including IL15, GDF15 (MIC-1), TNFSF4 (OX40L),TNFSF13B (BAFF), two MIP-1α chemokines (CCL3, CCL3L3) and oneMIP-1β chemokine (CCL4L2). In permissive LCLs, IL4 (an anti-inflammatory cytokine known to counteract IFN-γ) was over-expressed along with pro-inflammatory cytokines IL1A, MIF, PRL(prolactin), and three other cytokines, CKLF, CMTM3, and CMTM7.Most of these genes were low in absolute amount of transcriptsproduced and the results were less robust than for other genesdiscussed below. The overexpression of pro-inflammatory cytokinesin permissive LCLs may suggest a role in creating a milieu that lowerssignaling thresholds for activation of the lytic Z or R promoters.

Also overexpressed in non-permissive LCLs were potential targetgenes of oncogenic microRNAs 155 and 221/222. miR-155′s targetsinclude PU.1, CEBP/β, BACH1, SOCS1, AID, and SHIP1, and it isupregulated by TGF-β signaling (Kong et al., 2008; Lu et al., 2009;O'Connell et al., 2009; Yin et al., 2008). This microRNA has been linkedto Latency III, Hodgkin's disease, PTLD, and DLBCL and AML leukemias,and is not associated with Latency I lymphomas (Kluiver et al., 2006;Teng and Papavasiliou, 2009). miR-155 orthologs are encoded byoncogenic herpesviruses HHV-8 and MDV-1, and miR-155 is inducedby EBV LMP1 (Gatto et al., 2008; Zhao et al., 2009). miR-221/222enhances the progression of several malignancies by inhibiting thecell cycle regulator p27Kip1 and is upregulated byMDV-1 in lymphomacells (Lambeth et al., 2009; le Sage et al., 2007).

Plasma cell-associated genes were expressed differentially in themicroarray. Significantly, of the curated gene sets gleaned from earlierpublications, the one most overexpressed in permissive LCLs was agroup of 80 genes overexpressed in plasma cell and plasmablastsubsets and underexpressed in peripheral blood and tonsillar B cells(Tarte et al., 2003). A few of these genes were individually over-expressed in permissive LCLs, including HYOU1, TNFRSF17, PRG1,DDOST, PPIB, HSPA5, ARMET, and HSP90B1, the last five of which havebeen associated with the unfolded protein response. Also, whenexamining sets of genes which shared a known promoter motif, wefound that among the few overexpressed in permissive LCLs were thesets of genes whose promoters contained recognition sites for XBP-1,ATF-6, and HTF-1. HTF-1 is a rat homologue of XBP-1, and ATF-6induces the expression of XBP-1 (Yoshida et al., 2001).

These findings suggest that non-permissive LCLs have geneexpression profiles that support stable proliferation rather thanterminal differentiation into PC. In this regard, non-permissive LCLsoverexpressed a set of genes downregulated by the tumor suppressorgene p21 in a p53-dependent fashion, and permissive LCLs over-expressed a set of genes upregulated by p53. The gene for p21(CDKN1A) was seen in later experiments to be overexpressed in non-permissive LCLs, though this was not statistically significant (1.3-foldenrichment; p=0.108). With regard to PC differentiation, non-permissive LCLs overexpressed a gene set associated with a“proliferation” subgroup of multiple myeloma (MM) that predicts abad prognosis compared with other MM cases (Zhan et al., 2006), andgene sets upregulated in MM cells activated by the oncogene N-ras orexposed to the pro-proliferative cytokine IL-6. Also overexpressed innon-permissive LCLs was a set of genes downregulated in primaryeffusion lymphoma, identified in a study that concluded that PEL maybe derived from plasmablasts (Klein et al., 2003).

When we compared LCLs grown with and without acyclovirthere were negligible differences in gene expression, suggesting thatneither the drug treatment nor late lytic viral events influenced theresults of the microarray. Genes identified by this microarray studyas potentially involved in lytic permissivity, and other genes ofinterest, were selected for further study using RT-PCR. Many geneexpression studies have been done on the effects of EBV infection ofB cells, but this is the first to characterize a variety of geneticallydistinct LCLs, all immortalized in the same way with the same strainof virus.

Thus far, the list of genes known to induce or repress lytic EBVreplication is short. In physiological situations, the R promoter mainlyresponds to BZLF1 (Kieff and Rickinson, 2007). The Z promoter hasbeen extensively characterized for binding sites, including motifsactivated by XBP-1 and the BZLF1 protein it encodes (Sun andThorley-Lawson, 2007), and repressed by signaling from CD40 as wellas EBV LMP1 (Adler et al., 2002). We saw that CD40 expression didcorrelate negatively with lytic permissivity, although LMP1 expres-sion did not. CD40 was the most differentially expressed gene in theRT-PCR comparison of seven permissive and seven non-permissiveLCLs (Table 2). Because CD40 emerged as significant in this objective,large-scale comparison, we are encouraged that functional signifi-cance might attach to other genes detected in the same study.

The unfolded protein response (UPR) is activated when B cellsbecome antibody-secreting cells and during some productive virusinfections, although there is no evidence that UPR activation isinduced by herpesvirus reactivation. Some herpesviruses inhibit theUPR during productive infection to create an environment thatsustains lytic replication, or induce it as part of an immune evasionstrategy (Lee et al., 2009). EBV LMP1 induces the PERK/ATF-4pathway of the UPR, which leads initially to further upregulation ofLMP1, and then at higher levels to LMP1 degradation by autophagy(Lee and Sugden, 2008). This biphasic pattern probably titrates LMP1expression to ensure an appropriate level of constitutive signaling tosustain the proliferation of EBV-transformed cells. Expression of twotranscription factors downstream of UPR activation, ATF4 and XBP1,correlated positively with LMP1 expression in our study, suggestingthat LCLs vary in the degree to which they undergo the LMP1/PERK/ATF-4 feedback cycle. Importantly, this variation was not associatedwith the level of lytic permissivity. We also investigated ten UPRproteins that localize in the ER, and eight were overexpressed inpermissive LCLs, including five of the ten most significantly over-expressed (HSP90B1, DNAJB11, SDF2L1, CALR, ARMET). These ERproteins did not correlate with LMP1, and did not include any of thestress response proteins recently found to be induced by EBNA3A(Young et al., 2008). We cannot rule out the possibility that thisassociation of UPR gene overexpression with lytic permissivityactually reflects the induction of the UPR by EBV reactivation.However, the same association was found in LCLs cultured withACV, and it is unclear whether these genes are indicative of enhancedUPR activity, or are inherently present at higher levels enabling thecells to have a more acute response if and when ER stress sensors areactivated. Experiments now underway suggest that UPR proteins areexpressed at a consistently high level in permissive LCLs, and are notconcentrated in the minority of cells undergoing lytic reactivation.

In general, as shown in Table 2 (a subset of the larger RT-PCR resultsshown inSupplementary Tables 8 and9),more geneswere significantlyoverexpressed in non-permissive LCLs than in permissive LCLs. In ourmicroarray results the lists of genes overexpressed in non-permissiveLCLs also contained more transcription factors, a trend that exists foractivators, repressors, coactivators and corepressors (SupplementaryTable 7). This suggests that some EBV-transformed B cell lines represslytic reactivation more tightly because of a milieu of factors thatcounteract the cells' tendency to experience a positive feedback loop ofBZLF1 expression and thus trip over into lytic reactivation.

Signaling from CD40 or LMP1 leads to the activation of transcrip-tion factors that repress the Z promoter. We detected a number ofgenes whose expression was greater in the non-permissive LCLs, butnone are known to be upregulated by CD40 or LMP1. One study foundthat Pax-5 and EBF-1 activity were enhanced by CD40 signaling, butCD40 did not affect the levels of these factors (Merluzzi et al., 2004).Several genes that have been seen as upregulated by lytic replicationof EBV or other herpesviruses were nonetheless associated with thenon-permissive phenotype in our LCLs, further suggesting that ourfindings did not reflect downstream effects of EBV reactivation. Forexample, EGR1/Zif-268 is directly induced by BZLF1 and directly

64 M.L. Davies et al. / Virology 400 (2010) 53–67

induces BRLF1 (Chang et al., 2006; Zalani et al., 1995), while ID2 isupregulated by lytic replication of EBV, HHV-8, CMV, and HSV(Khodarev et al., 1999; Naranatt et al., 2004; Yuan et al., 2006; Zhuet al., 1998). However, we found both these factors among the genesoverexpressed in LCLs with less lytic permissivity. Many genes weinvestigated by RT-PCR had been detected in earlier studies to beupregulated or downregulated by EBV Latency III proteins, but werenot regulated in either direction in our study (e.g. JUND, EBI3, TIMP1,TNFRSF14). This might be expected, since every LCL was producing thefull complement of Latency III proteins, and EBNA2 and LMP1expression did not correlate with permissivity.

Most significant was the group of transcription factors whichmaintain B cell lineage from the pre-B cell phase to the point ofterminal differentiation into PCs. During the transition to a plasma-blastic phenotype this network of factors, including EBF-1, Pax-5, andE2A, gives way to a network of factors that had been repressed, led byBlimp-1, which suppresses Pax-5 and thus removes restraints on XBP-1 production (Nutt and Kee, 2007). Although lytically permissive LCLsdid not containmore PRDM1/Blimp-1, IRF4, unspliced XBP1, or splicedXBP1 mRNA, the major members of the opposite network were alloverexpressed in non-permissive LCLs. EBF1, PAX5, and TCF3/E2Awere among the 18 genes with significantly enhanced (pb0.05)expression in non-permissive LCLs, and EBF1 had the most significantp-value of any gene with a N3-fold difference between the means ofthe two populations. In addition, ETS1, which encodes a transcriptionfactor that upregulates EBF-1 and Pax-5 by repressing Blimp-1 activity(John et al., 2008), was one of eight overexpressed genes with pb0.01in this comparison.

Taking our results as a whole, we suggest that each line of LatencyIII-transformed B cells has a characteristic frequency of EBV lyticreactivation which is attributable to the intrinsic levels of a few keycellular proteins that influence gene expression. We further suggestthat in addition to the known effects of XBP-1 on BZLF1 expression,the XBP-1-containing pathway may be relevant in a different way.When LCLs produce high amounts of factor(s) that repress the activityof XBP-1 and Blimp-1, they also repress the level of spontaneous lyticreactivation. From this study the best single candidate for such acontrolling factor is EBF-1.

Materials and methods

Establishment of LCLs and cell culture

Lymphoblastoid cell lines were established by suspending PBMCsin RPMI-1640 including 20% FBS and tacrolimus, to which was addedsupernatant from the B95-8 virus-producing cell line. After 10–15days of incubation the viral supernatant was removed and immortal-ized cells weremaintained for at least 4 weeks before being used; theywere then grown in RPMI-1640 including 10% FBS, 2 mM L-glutamine,100 units/ml penicillin and 100 μg/ml streptomycin. For inhibitinglytic viral replication, acyclovir was added to the media at aconcentration of 22.5 μg/ml. A total of 62 LCLs from 62 differentdonors were studied. These lines had been established for twoseparate immunological studies. The collection of pediatric LCLs wasestablished during a PTLD study, and the adult LCLs were establishedduring an HIV-related study. After determining the average genomecopy number/cell, cell lines with N200 copies/cell were classified aspermissive for spontaneous lytic virus induction.

Cell lysis and PCR

For DNA measurement, cell pellets were lysed in 10 mM Tris (pH7.6) containing 50 mM KCl, 2.5 mM MgCl2, 1% Tween 20, and0.1 mg/ml proteinase K. For measuring the number of EBV genomes,we detected a sequence in the BLLF1 gene encoding viral gp350(forward primer 5′-GTATCCACCGCGGATGTCA-3′; reverse primer 5′-

GGCCTTACTTTCTGTGCCGTT-3′; probe 5′FAM-TGGACTTGGTGT-CACCGGTGATGC-TAMRA-3′). For normalizing to the amount ofcellular DNA, we detected a sequence in human GAPDH (Xu et al.,2006). Lysates of the EBV-negative DG75 B cell line were used asstandards for GAPDH DNA, while lysates of the Namalwa cell line,each of which contains two integrated copies of the EBV genome,were used as standards for EBV DNA.

For RNA measurement, RNA was extracted from cell pellets using aVersaGene cell kit (Gentra); RNAwas thenDNAse-treatedusing a ZymoResearch kit, and transcribed into cDNA using MultiScribe reversetranscriptase (ABI). For each target transcript, the standards used werepurifiedPCR product containing a knownnumber of copies of the targetsequence, and normalized to cell number by measuring β2-micro-globulin cDNA, using a predesigned primer set (Applied Biosystems).

PCR targets

EBV mRNA targets included the latent genes EBNA1, EBNA2,EBNA3A, EBNA3B, EBNA3C, LMP1, LMP2A, and LMP2B, and the lyticgenes ZEBRA and BMLF1. Also quantified were the number oftranscripts from the C, W, and Q promoters. Qp is a latent promoter,but a measurement of its transcripts also detects transcripts from thelytic F promoter. For both Cp and Wp, two alternative splices weremeasured—a nonproductive splice that encodes EBNA1, -2, -3A, -3B, or-3C; and a productive splice that encodes EBNA-LP as well as one of theother EBNA proteins. The amount of EBNA-LP mRNA can be measuredas the sum of the Cp and Wp productive splices.

All EBV transcripts except BMLF1 were quantified with primerscrossing an mRNA splice site, detecting cDNA but not genomic DNA.Primer sequences are supplied in Supplementary Table 1. All PCR wasquantitative real-time PCR performed on an ABI 7500 instrument.

Flow cytometry

Proliferating cells were washed in cold PBS, and chilled and stainedin FACS buffer (HBSS including 2.0% BSA and 0.1% sodium azide). Thefollowing mouse α-human monoclonal antibodies were used: PE-α-IgG (clone G18-145, BD); PE-α-CD138 (clone M115, BD); PE-α-BR3(clone 11C1, BioLegend); PE/Cy5.5-α-CD19 (clone SJ25-C1, Caltag);PE/Cy7-α-HLA-DR (clone L243, BioLegend); APC-α-IgM (clone MHM-88, BioLegend); Alexa647-α-CD23 (clone D3.6, BioLegend); Alexa647-α-CD30 (clone MEM-268, BioLegend); APC/Cy7-α-CD20 (clone 2H7,BioLegend); APC/Alexa750-α-CD27 (clone CLB-27/1, Caltag); PacificBlue-α-HLA-A,B,C (clone W6/32, BioLegend); Pacific Blue-α-CD45(clone HI30, BioLegend); Pacific Blue-α-CD69 (clone FN50, Bio-Legend); biotin-α-IgD (clone IA6-2, BD); biotin-α-CD38 (clone AT13/5, AbD Serotec); and biotin-α-CD40 (clone LOB7/6, AbD Serotec). Wealso used PE/Cy7-conjugated rat α-mouse B220 (BD), and FITC-conjugated polyclonal goat antibodies (Biosource) specific for humanIgA and for all human Igs (pan-Ig). Antibodies against viral proteinsincludedmouseα-ZEBRA (clone BZ.1, Santa Cruz), andmouseα-gp110(clone 5B2, Abcam), which were labeled with Alexa Fluor-conjugatedZenon goat F(ab)2 fragments (Molecular Probes). Cell surface stainingwas done with live cells, with streptavidin-Pacific Orange (MolecularProbes) used to label biotinylated antibodies; in intracellular stainingfor viral genes, 1% formaldehyde and 0.1% Triton X-100 was used to fixand permeabilize cells. The samples were measured in a BD FACSAria,and data were analyzed using FACSDiva and FlowJo software.

Gene expression microarrays

RNA was isolated using the TRIzol process (Invitrogen), andtechnicians at the University of Pittsburgh Genomics and ProteomicsCore Laboratories (GPCL) transcribed it into cDNA, which was hybri-dized to Illumina HumanRef-8 v2 BeadChip arrays. Software used toanalyze the data included caGEDA (http://bioinformatics2.pitt.edu/

65M.L. Davies et al. / Virology 400 (2010) 53–67

GE2/GEDA.html) to do efficiency analysis and get J5 scores; DAVID(http://david.abcc.ncifcrf.gov/) and Ingenuity Pathways Analysis(Ingenuity) to get information about gene ontologies, pathways andbiological roles of each gene of interest; and GSEA (http://www.broad.mit.edu/gsea/) to get information about entire sets of genes, ratherthan individual genes, that were upregulated in each sample popula-tion. GSEA analyses were done with the following settings: 1000permutations; genes identified by HUGO gene symbol with collapsedataset set to false; weighted enrichment score normalized by gene setsize; gene sets from GSEA databases c2.all.v2.5.symbols.gmt (curated)and c3.all.v2.5.symbols.gmt (motif); andgene sets ofb15orN500genesfiltered out. The J5 score is a measure of the scale of difference of thesample group means for a given gene relative to the average differenceof all the genes on the array (Patel and Lyons-Weiler, 2004).

RT-PCR arrays

RNAwas isolated using the TRIzol process and reverse-transcribedusing the High-Capacity kit (ABI); cDNA was mixed with TaqManGene Expression Master Mix and applied to custom-made 384-wellplates (Low-Density Arrays, ABI) containing primers for 92 genes ofinterest and 4 endogenous controls. Seven permissive and seven non-permissive LCLs were measured. For four LCLs, RNA was taken on twodates and measurements were averaged together; the rest weremeasured on one date. Five were assessed both with and withoutacyclovir, to reaffirm that viral replication has a negligible influenceon the cellular gene expression across the whole cell line; the restwere only measured without acyclovir. PCR was performed on an ABI7900HT at the GPCL, and data were analyzed using SDS2.1 software(ABI), Microsoft Excel, and Minitab 15. Results from RT-PCR arrayswere quantified using the 2−ΔΔCt method (for each sample, taking thedifference between each experimental gene's cycle of detection (Ct)and that of an endogenous control gene, and then comparing thisdifference to the difference in a reference sample).

Relative quantitation of XBP-1u and XBP-1s mRNAs

XBP-1 cDNAs were generated by random hexamer RT, andamplified using the ABI 7300 Real-time PCR System, SYBR greenqPCR master mix reagent (Fermentas), and the following primers:XBP-1 forward: 5′-GTTGAGAACCAGGAGTTAAG-3′; XBP-1 reverse 5′-GAGAAAGGGAGGCTGGTAAG-3′ to generate 357nt (XBP-1u) and331nt (XBP-1s) amplicons. Amplifaction conditions were 95 °C for20 s, followed by 21 cycles of 95 °C for 5 s and 60 °C for 30 s. Toquantitate relative XBP-1 splicing, dissociation curve analysis wasperformed post-run, according to the manufacturer's instructions.Relative XBP-1 splicing was calculated by comparing peak heightsfrom the unspliced (Tm 85.7 degree) and spliced (Tm 82.7 degree)peaks, and measured as the ratio of XBP1s / (XBP1s + XBP1u).

Acknowledgments

We thank Holly Bilben, LuAnn Borowski, and Jose Luis Garrido forlaboratory assistance; Deborah Hollingshead for microarray sampleprocessing; and Monica Tomaszewski-Flick for critical reading anddiscussions.

This work was partially supported by NIH (NHLBI) grant P50HL74732, and made possible in part by grant 5 UL1 RR024153 fromthe National Center for Research Resources (NCRR), a component ofthe National Institutes of Health (NIH), and NIH Roadmap for MedicalResearch. Its contents are solely the responsibility of the authors anddo not necessarily represent the official view of NCRR or NIH.Information on NCRR is available at http://www.ncrr.nih.gov/.Information on Re-engineering the Clinical Research Enterprise canbe obtained from http://nihroadmap.nih.gov/clinicalresearch/overview-transcriptional.asp.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.virol.2010.01.002.

References

Adhikary, D., Behrends, U., Boerschamm, H., Pfünder, A., Burdach, S., Moosmann, A.,Witter, K., Bornkamm, G.W., Mautner, J., 2007. Immunodominance of lytic cycleantigens in Epstein–Barr virus-specific CD4+ T cell preparations for therapy. PLoSONE 2 (7), e583.

Adler, B., Schaadt, E., Kempkes, B., Zimber-Strobl, U., Baier, B., Bornkamm, G.W., 2002.Control of Epstein–Barr virus reactivation by activated CD40 and viral latentmembrane protein 1. Proc. Natl. Acad. Sci. U.S.A. 99, 437–442.

Alsaleh, G., Suffert, G., Semaan, N., Juncker, T., Frenzel, L., Gottenberg, J.-E., Sibilla, J.,Pfeffer, S., Wachsmann, D., 2009. Bruton's tyrosine kinase is involved in miR-346-related regulation of IL-18 release by lipopolysaccharide-activated rheumatoidfibroblast-like synoviocytes. J. Immunol. 182 (8), 5088–5097.

Babel, N., Vergopoulos, A., Trappe, R.U., Oertel, S., Hammer, M.H., Kairavanov, S.,Schneider, N., Reiss, H., Papp-Vary, M., Neuhaus, R., Gondek, L.P., Volk, H.D., Reinke,P., 2007. Evidence for genetic susceptibility towards development of posttransplantlymphoproliferative disorder in solid organ recipients. Transplantation 84 (3),387–391.

Bennett, L., Palucka, A.K., Arce, E., Cantrell, V., Borvak, J., Banchereau, J., Pascual, V., 2003.Interferon and granulopoiesis signatures in systemic lupus erythematosus blood.J. Exp. Med. 197 (6), 711–723.

Bhende, P.M., Dickerson, S.J., Sun, X., Feng, W.-H., Kenney, S.C., 2007. X-box-bindingprotein 1 activates lytic Epstein–Barr virus gene expression in combination withprotein kinase D. J. Virol. 81 (14), 7363–7370.

Bogunia-Kubik, K., Mlynarczewska, A., Jaskula, E., Lange, A., 2005. The presence of IFNG3/3 genotype in the recipient associates with increased risk for Epstein–Barr virusreactivation after allogeneic haematopoietic stem cell transplantation. Br. J.Haematol. 132 (3), 326–332.

Bollard, C.M., Aguilar, L., Straathof, K.C., Gahn, B., Huls, M.H., Rousseau, A., Sixbey, J.,Gresik, M.V., Carrum, G., Hudson, M., Dilloo, D., Gee, A., Brenner, M.K., Rooney, C.M.,Hislop, H.E., 2004. Cytotoxic T lymphocyte therapy for Epstein–Barr virus+Hodgkin's disease. J. Exp. Med. 200 (12), 1623–1633.

Browne, E.P., Wing, B., Coleman, D., Shenk, T., 2001. Altered cellular mRNA levels inhuman cytomegalovirus-infected fibroblasts: viral block to the accumulation ofantiviral mRNAs. J. Virol. 75 (24), 12319–12330.

Cahir-McFarland, E.D., Carter, K., Rosenwald, A., Giltnane, J.M., Henrickson, S.E., Staudt,L.M., Kieff, E., 2004. Role of NF-{kappa}B in cell survival and transcription of latentmembrane protein 1-expressing or Epstein–Barr virus Latency III-infected cells.J. Virol. 78 (8), 4108–4119.

Carter, K.L., Cahir-McFarland, E., Kieff, E., 2002. Epstein Barr virus-induced changes inB-lymphocyte gene expression. J. Virol. 76 (20), 10427–10436.

Chang, Y., Lee, H.-H., Chen, Y.-Y., Lu, J., Wu, S.-Y., Chen, C.-W., Takada, K., Tsai, C.-H.,2006. Induction of the early growth response 1 gene by Epstein–Barr virus lytictransactivator Zta. J. Virol. 80 (15), 7748–7755.

Chen, A., Zhao, B., Kieff, E., Aster, J.C., Wang, F., 2006. EBNA-3B- and EBNA-3C-regulatedcellular genes in Epstein–Barr virus-immortalized lymphoblastoid cell lines. J. Virol.80 (20), 10139–10150.

Countryman, J.K., Gradoville, L., Miller, G., 2008. Histone hyperacetylation occurs onpromoters of lytic cycle regulatory genes in Epstein–Barr virus-infected cell lineswhich are refractory to disruption of latency by histone deacetylase inhibitors.J. Virol. 82 (10), 4706–4719.

Croonquist, P.A., Linden, M.A., Zhao, F., Van Ness, B.G., 2003. Gene profiling of amyeloma cell line reveals similarities and unique signatures among IL-6 response,N-ras-activating mutations, and coculture with bone marrow stromal cells. Blood102 (7), 2581–2592.

da Silva, G.N., Bacchi, M.M., Rainho, C.A., de Oliveira, D.E., 2007. Epstein–Barr virusinfection and single nucleotide polymorphisms in the promoter region ofinterleukin 10 gene in patients with Hodgkin lymphoma. Arch. Pathol. Lab. Med.131 (11), 1691–1696.

Dierksheide, J.E., Baiocchi, R.A., Ferketich, A.K., Roychowdhury, S., Pelletier, R.P.,Eisenbeis, C.F., Caligiuri, M.A., VanBuskirk, A.M., 2005. IFN-{gamma} genepolymorphisms associate with development of EBV+ lymphoproliferative diseasein hu PBL-SCID mice. Blood 105 (4), 1558–1565.

Dolcetti, R., 2007. B lymphocytes and Epstein–Barr virus: the lesson of post-transplantlymphoproliferative disorders. Autoimmun. Rev. 7 (2), 96–101.

Dolcetti, R., Quaia, M., Gloghini, A., De Re, V., Zancai, P., Cariati, R., Babuin, L., Cilia, A.M.,Rizzo, S., Carbone, A., Boiocchi, M., 1999. Biologically relevant phenotypic changesand enhanced growth properties induced in B lymphocytes by an EBV strainderived from a histologically aggressive Hodgkin's disease. Int. J. Cancer 80 (2),240–249.

Durandy, A., Le Deist, F., Emile, J.-F., Debatin, K., Fischer, A., 1997. Sensitivity of Epstein–Barr virus-induced B cell tumor to apoptosis mediated by anti-CD95/Apo-1/fasantibody. Eur. J. Immunol. 27 (2), 538–543.

Elliott, J., Goodhew, E.B., Krug, L.T., Shakhnovsky, N., Yoo, L., Speck, S.H., 2004. Variablemethylation of the Epstein–Barr virusWp EBNA gene promoter in B-lymphoblastoidcell lines. J. Virol. 78 (24), 14062–14065.

Fukuda, M., Longnecker, R., 2005. Epstein–Barr virus (EBV) latent membrane protein 2Aregulates B-cell receptor-induced apoptosis and EBV reactivation through tyrosinephosphorylation. J. Virol. 79 (13), 8655–8660.

66 M.L. Davies et al. / Virology 400 (2010) 53–67

Gatto, G., Rossi, A., Rossi, D., Kroening, S., Bonatti, S., Mallardo, M., 2008. Epstein–Barrvirus latent membrane protein 1 trans-activates miR-155 transcription through theNF-{kappa}B pathway. Nucleic Acids Res. 36 (20), 6608–6619.

Gregory, C.D., Dive, C., Henderson, S., Smith, C.A., Williams, G.T., Gordon, J., Rickinson,A.B., 1991. Activation of Epstein–Barr virus latent genes protects human B cellsfrom death by apoptosis. Nature 349 (6310), 612–614.

Haddad, R., Guardiola, P., Izac, B., Thibault, C., Radich, J., Delezoide, A.L., Baillou, C.,Lemoine, F.M., Gluckman, J.C., Pflumio, F., Canque, B., 2004. Molecular character-ization of early human T/NK and B-lymphoid progenitor cells in umbilical cordblood. Blood 104 (13), 3918–3926.

Hatta, K., Morimoto, A., Ishii, E., Kimura, H., Ueda, I., Hibi, S., Todo, S., Sugimoto, T.,Imashuku, S., 2007. Association of transforming growth factor-{beta}1 genepolymorphism in the development of Epstein–Barr virus-related hematologicdiseases. Haematologica 92 (11), 1470–1474.

Huang, D.W., Sherman, B.T., Lempicki, R.A., 2009. Systematic and integrative analysis oflarge gene lists using DAVID bioinformatics resources. Nat. Protoc. 4 (1), 44–57.

Inga, A., Storici, F., Darden, T.A., Resnick, M.A., 2002. Differential transactivation by thep53 transcription factor is highly dependent on p53 level and promoter targetsequence. Mol. Cell. Biol. 22 (24), 8612–8625.

Jenkins, F.J., Rowe, D.T., Rinaldo, C.R., 2003. Herpesvirus infections in organ transplantrecipients. Clin. Diagn. Lab. Immunol. 10 (1), 1–7.

John, S.A., Clements, J.L., Russell, L.M., Garrett-Sinha, L.A., 2008. Ets-1 regulates plasmacell differentiation by interfering with the activity of the transcription factorBlimp-1. J. Biol. Chem. 283 (2), 951–962.

Keating, S., Prince, S., Jones, M., Rowe, M., 2002. The lytic cycle of Epstein–Barr virus isassociated with decreased expression of cell surface major histocompatibilitycomplex class I and class II molecules. J. Virol. 76 (16), 8179–8188.

Khanolkar, A., Fu, Z., Underwood, L.J., Bondurant, K.L., Rochford, R., Cannon, M.J., 2003.CD4+ T cell-induced differentiation of EBV-transformed lymphoblastoid cells isassociated with diminished recognition by EBV-specific CD8+ cytotoxic T cells.J. Immunol. 170 (6), 3187–3194.

Khodarev, N.N., Advani, S.J., Gupta, N., Roizman, B., Weichselbaum, R.R., 1999.Accumulation of specific RNAs encoding transcriptional factors and stress responseproteins against a background of severe depletion of cellular RNAs in cells infectedwith herpes simplex virus 1. Proc. Natl. Acad. Sci. U. S. A. 96 (21), 12062–12067.

Kieff, E., Rickinson, A.B., 2007. Epstein–Barr virus and its replication, In: Howley, P.M.,Griffin, D.E., Lamb, R.A., Martin, M.A., Roizman, B., Straus, S.E. (Eds.), 5th ed. FieldsVirology, Vol. 2. Lippincott,Williams &Wilkins, Philadelphia, pp. 2603–2654. 2 vols.

Klein, U., Gloghini, A., Gaidano, G., Chadburn, A., Cesarman, E., Dalla-Favera, R., Carbone,A., 2003. Gene expression profile analysis of AIDS-related primary effusionlymphoma (PEL) suggests a plasmablastic derivation and identifies PEL-specifictranscripts. Blood 101 (10), 4115–4121.

Kluiver, J., Haralambieva, E., de Jong, D., Blokzijl, T., Jacobs, S., Kroesen, B.J., Poppema, S.,van den Berg, A., 2006. Lack of BIC and microRNA miR-155 expression in primarycases of Burkitt lymphoma. Genes Chromosomes Cancer 45 (2), 147–153.

Kong, W., Yang, H., He, L., Zhao, J.-J., Coppola, D., Dalton, W.S., Cheng, J.Q., 2008.MicroRNA-155 is regulated by the transforming growth factor {beta}/Smadpathway and contributes to epithelial cell plasticity by targeting RhoA. Mol. Cell.Biol. 28 (22), 6773–6784.

Lambeth, L.S., Yao, Y., Smith, L.P., Zhao, Y., Nair, V.K., 2009. MicroRNAs 221 and 222target p27Kip1 in Marek's disease virus-transformed tumour cell line MSB-1. J. Gen.Virol. 90 (5), 1164–1171.

le Sage, C., Nagel, R., Egan, D.A., Schrier, M., Mesman, E., Mangiola, A., Anile, C., Maira, G.,Mercatelli, N., Ciafrè, S.A., Farace, M.G., Agami, R., 2007. Regulation of the p27Kip1

tumor suppressor by miR-221 and miR-222 promotes cancer cell proliferation.EMBO J. 26 (15), 3699–3708.

Lee, D.Y., Sugden, B., 2008. The LMP1 oncogene of EBV activates PERK and the unfoldedprotein response to drive its own synthesis. Blood 111 (4), 2280–2289.

Lee, D.Y., Lee, J., Sugden, B., 2009. The unfolded protein response and autophagy:herpesviruses rule! J. Virol. 83 (3), 1168–1172.

Lu, L.-F., Thai, T.-H., Calado, D.P., Chaudhry, A., Kubo, M., Tanaka, K., Loeb, G.B., Lee, H.,Yoshimura, A., Rajewsky, K., Rudensky, A.Y., 2009. Foxp3-dependent microRNA155confers competitive fitness to regulatory T cells by targeting SOCS1 protein.Immunity 30 (1), 80–91.

Markasz, L., Stuber, G., Flaberg, E., Jernberg, Å., Eksborg, S., Olah, E., Skribek, H., Szekely,L., 2006. Cytotoxic drug sensitivity of Epstein–Barr virus transformed lympho-blastoid B-cells. BMC Cancer 6, 265.

Merluzzi, S., Moretti, M., Altamura, S., Zwollo, P., Sigvardsson, M., Vitale, G., Pucillo, C.,2004. CD40 stimulation induces Pax5/BSAP and EBF activity through an APE/Ref-1-dependent redox mechanism. J. Biol. Chem. 279 (3), 1777–1786.

Morris, M.A., Dawson, C.W., Wei, W., O'Neil, J.D., Stewart, S.E., Jia, J., Bell, A.I., Young, L.S.,Arrand, J.R., 2008. Epstein–Barr virus-encoded LMP1 induces a hyperproliferativeand inflammatory gene expression programme in cultured keratinocytes. J. Gen.Virol. 89 (11), 2806–2820.

Nakayama, T., Fujisawa, R., Izawa, D., Hieshima, K., Takada, K., Yoshie, O., 2002. Human Bcells immortalized with Epstein–Barr virus upregulate CCR6 and CCR10 anddownregulate CXCR4 and CXCR5. J. Virol. 76 (6), 3072–3077.

Nakayama, T., Hieshima, K., Izawa, D., Tatsumi, Y., Kanamaru, A., Yoshie, O., 2003.Cutting edge: Profile of chemokine receptor expression on human plasma cellsaccounts for their efficient recruitment to target tissues. J. Immunol. 170 (3),1136–1140.

Naranatt, P.P., Krishnan, H.H., Svojanovsky, S.R., Bloomer, C., Mathur, S., Chandran, B.,2004. Host gene induction and transcriptional reprogramming in Kaposi'ssarcoma-associated herpesvirus (KSHV/HHV-8)-infected endothelial, fibroblast,and B cells: Insights into modulation events early during infection. Cancer Res. 64(1), 72–84.

Niens, M., van den Ber, A., Diepstra, A., Nolte, I.M., van der Steege, G., Gallagher, A.,Taylor, G.M., Jarrett, R.F., Poppema, S., te Meerman, G.J., 2006. The human leukocyteantigen class I region is associated with EBV-positive Hodgkin's lymphoma: HLA-Aand HLA complex group 9 are putative candidate genes. Cancer Epidemiol.Biomark. Prev. 15 (11), 2280–2284.

Nutt, S.L., Kee, B.L., 2007. The transcription regulation of B cell lineage commitment.Immunity 26 (6), 715–725.

O'Connell, R.M., Chaudhuri, A.A., Rao, D.S., Baltimore, D., 2009. Inositol phosphataseSHIP1 is a primary target of miR-155. Proc. Natl. Acad. Sci. U. S. A. 106 (17),7113–7118.

Patel, S., Lyons-Weiler, J., 2004. caGEDA: a web application for the integrated analysis ofglobal gene expression patterns in cancer. Appl. Bioinformatics 3 (1), 49–62.

Portis, T., Longnecker, R., 2003. Epstein–Barr virus (EBV) LMP2A interferes with globaltranscription factor regulation when expressed during B lymphocyte development.J. Virol. 77 (1), 105–114.

Rickinson, A.B., Kieff, E., 2007. Epstein–Barr virus, In: Knipe, D.M., Howley, P.M., Griffin,D.E., Lamb, R.A., Martin, M.A., Roizman, B., Straus, S.E. (Eds.), 5th ed. Fields Virology,Vol. 2. Lippincott, Williams & Wilkins, Philadelphia, pp. 2655–2700. 2 vols.

Rochford, R., Hobbs, M.V., Garnier, J.-L., Cooper, N.R., Cannon, M.J., 1993.Plasmacytoid differentiation of Epstein–Barr virus-transformed B cells in vivois associated with reduced expression of viral latent genes. Proc. Natl. Acad. Sci.U. S. A. 90, 352–356.

Ryan, J.L., Kaufmann, W.K., Raab-Traub, N., Oglesbee, S.E., Carey, L.A., Gulley, M.L., 2006.Clonal evolution of lymphoblastoid cell lines. Lab. Invest. 86 (11), 1193–1200.

Sana, T.R., Janatpour, M.J., Sathe, M., McEvoy, L.M., McClanahan, T.K., 2005. Microarrayanalysis of primary endothelial cells challenged with different inflammatory andimmune cytokines. Cytokine 29 (6), 256–269.

Schaadt, E., Baier, B., Mautner, J., Bornkamm, G.W., Adler, B., 2005. Epstein–Barr viruslatent membrane protein 2A mimics B-cell receptor-dependent virus reactivation.J. Gen. Virol. 86 (3), 551–559.

Shinozaki, K., Yasui, K., Agematsu, K., 2001. Direct B/B–cell interactions in immuno-globulin synthesis. Clin. Exp. Immunol. 124 (3), 386–391.

Spender, L.C., Lucchesi, W., Bodelon, G., Bilancio, A., Karstegl, C.E., Asano, T., Dittrich-Breiholz, O., Kracht, M., Vanhaesebroeck, B., Farrell, P.J., 2006. Cell target genes ofEpstein–Barr virus transcription factor EBNA-2: Induction of the p55{alpha}regulatory subunit of PI3-kinase and its role in survival of EREB2.5 cells. J. Gen.Virol. 87 (10), 2859–2867.

Stewart, S.E., Dawson, C.W., Takada, K., Curnow, J., Moody, C.A., Sixbey, J.W., Young, L.S.,2004. Epstein–Barr virus-encoded LMP2A regulates viral and cellular geneexpression by modulation of the NF-{kappa}B transcription factor pathway. Proc.Natl. Acad. Sci. U. S. A. 101 (44), 15730–15735.

Subklewe, M., Sebelin, K., Block, A., Meier, A., Roukens, A., Paludan, C., Fonteneau, J.-F.,Steinman, R.M., Münz, C., 2005. Dendritic cells expand Epstein–Barr virus-specificCD8+ T cell responses more efficiently than EBV-transformed B cells. Hum.Immunol. 66 (9), 938–949.

Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Gillette, M.A.,Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., Mesirov, J.P., 2005. Gene setenrichment analysis: a knowledge-based approach for interpreting genome-wideexpression profiles. Proc. Natl. Acad. Sci. U. S. A. 102 (43), 15545–15550.

Sugden, B., Phelps, M., Domoradzki, J., 1979. Epstein–Barr virus DNA is amplified intransformed lymphocytes. J. Virol. 31 (3), 590–595.

Sugimoto, M., Tahara, H., Ide, T., Furuichi, Y., 2004. Steps involved in immortalizationand tumorigenesis in human B-lymphoblastoid cell lines transformed by Epstein–Barr virus. Cancer Res. 64, 3361–3364.

Sun, C.C., Thorley-Lawson, D.A., 2007. Plasma cell-specific transcription factor XBP-1sbinds to and transactivates the Epstein–Barr virus BZLF1 promoter. J. Virol. 81 (24),13566–13577.

Swanson-Mungerson, M., Bultema, R., Longnecker, R., 2006. Epstein–Barr virus LMP2Aenhances B-cell responses in vivo and in vitro. J. Virol. 80 (14), 6764–6770.

Takahashi, T., Kawabe, T., Okazaki, Y., Itoh, C., Noda, K., Tajima, M., Satoh, M., Goto, M.,Mitsui, Y., Tahara, H., Ide, T., Furuichi, Y., Sugimoto, M., 2003. In vitro establishmentof tumorigenic human B-lymphoblastoid cell lines transformed by Epstein–Barrvirus. DNA Cell Biol. 22 (11), 727–735.

Tarte, K., De Vos, J., Zhan, F., Klein, B., Shaughnessy, J.D., 2003. Gene expression profilingof plasma cells and plasmablasts: toward a better understanding of the late stagesof B-cell differentiation. Blood 102, 592–600.

Teng, G., Papavasiliou, F.N., 2009. Shhh! Silencing by microRNA-155. Philos. Trans. R.Soc. Lond., B, Biol. Sci. 364 (1517), 631–637.

Thal, M.A., Carvalho, T.L., He, T., Kim, H.-G., Gao, H., Hagman, J., Klug, C.A., 2009. Ebf1-mediated down-regulation of Id2 and Id3 is essential for specification of the B celllineage. Proc. Natl. Acad. Sci. U. S. A. 106 (2), 552–557.

Thorley-Lawson, D.A., Gross, A., 2004. Persistence of the Epstein–Barr virus and theorigins of associated lymphomas. N. Engl. J. Med. 350 (13), 1328–1337.

Wang, S.E., Wu, F.Y., Shin, I., Qu, S., Arteaga, C.L., 2005. Transforming growth factor{beta} (TGF-{beta})-Smad target gene protein tyrosine phosphatase receptor type{kappa} is required for TGF-{beta} function. Mol. Cell. Biol. 25 (11), 4703–4715.

Wendel-Hansen, V., Rosén, A., Klein, G., 1987. EBV-transformed lymphoblastoid cellsdown-regulate EBNA in parallel with secretory differentiation. Int. J. Cancer 39 (3),404–408.

Wroblewski, J.M., Copple, A., Batson, L.P., Landers, C.D., Yannelli, J.R., 2002. Cell surfacephenotyping and cytokine production of Epstein–Barr virus (EBV)-transformedlymphoblastoid cell lines (LCLs). J. Immunol. Methods 264 (1), 19–28.

Wu, Q., Kirschmeier, P., Hockenberry, T., Yang, T.-Y., Brassard, D.L., Wang, L.,McClanahan, T., Black, S., Rizzi, G., Musco, M.L., Mirza, A., Liu, S., 2002.Transcriptional regulation during p21WAF1/CIP1-induced apoptosis in humanovarian cancer cells. J. Biol. Chem. 277 (39), 36329–363377.

67M.L. Davies et al. / Virology 400 (2010) 53–67

Xu, Y., Tan, L.-J., Grachtchouk, V., Voorhees, J.J., Fisher, G.J., 2005. Receptor-type protein-tyrosine phosphatase-{kappa} regulates epidermal growth factor receptor. J. Biol.Chem. 280 (52), 42694–42700.

Xu, S., Green, M., Kingsley, L., Webber, S., Rowe, D., 2006. A comparison of quantitative-competitive and realtime PCR assays using an identical target sequence to detectEpstein–Barr virus viral load in the peripheral blood. J. Immunol. Methods 137,205–212.

Yin, Q., McBride, J., Fewell, C., Lacey, M., Wang, X., Lin, Z., Cameron, J., Flemington, E.K.,2008. microRNA-155 is an Epstein–Barr virus-induced gene that modulatedEpstein–Barr virus-regulated gene expression pathways. J. Virol. 82 (11),5295–5306.

Yoshida, H., Matsui, T., Yamamoto, A., Okada, T., Mori, K., 2001. XBP1 mRNA is inducedby ATF6 and spliced by IRE1 in response to ER stress to produce a highly activetranscription factor. Cell 107 (7), 881–891.

Young, P., Anderton, E., Paschos, K., White, R., Allday, M.J., 2008. Epstein–Barr virusnuclear antigen (EBNA) 3A induces the expression of and interacts with a subset ofchaperones and co-chaperones. J. Gen. Virol. 89 (4), 866–877.

Yuan, J., Cahir-McFarland, E., Zhao, B., Kieff, E., 2006. Virus and cell RNAs expressedduring Epstein–Barr virus replication. J. Virol. 80 (5), 2548–2565.

Zalani, S., Holley-Guthrie, E., Kenney, S., 1995. The Zif268 cellular transcription factoractivates expression of the Epstein–Barr virus immediate-early BRLF1 promoter.J. Virol. 69 (6), 3816–3823.

Zhan, F., Huang, Y., Colla, S., Stewart, J.P., Hanamura, I., Gupta, S., Epstein, J., Yaccoby, S.,Sawyer, J., Burington, B., Anaissie, E., Hollmig, K., Pineda-Roman, M., Tricot, G., vanRhee, F., Walker, R., Zangari, M., Crowley, J., Barlogie, B., Shaughnessy, J.D., 2006.The molecular classification of multiple myeloma. Blood 108 (6), 2020–2028.

Zhang, W., Ratain, M.J., Dolan, M.E., 2008. The HapMap resource is providing newinsights into ourselves and its application to pharmacogenomics. Bioinform. Biol.Insights 2, 15–23.

Zhao, B., Maruo, S., Cooper, A., Chase, M.R., Johannsen, E., Kieff, E., Cahir-McFarland, E.,2006. RNAs induced by Epstein–Barr virus nuclear antigen 2 in lymphoblastoid celllines. Proc. Natl. Acad. Sci. U. S. A. 103 (6), 1900–1905.

Zhao, Y., Yao, Y., Xu, H., Lambeth, L., Smith, L.P., Kgosana, L., Wang, X., Nair, V., 2009. Afunctional microRNA-155 ortholog encoded by the oncogenic Marek's diseasevirus. J. Virol. 83 (1), 489–492.

Zhu, H., Cong, J.-P., Mamtora, G., Gingeras, T., Shenk, T., 1998. Cellular gene expressionaltered by human cytomegalovirus: global monitoring with oligonucleotide arrays.Proc. Natl. Acad. Sci. U. S. A. 95 (24), 14470–14475.