B cells undergoing EBV-driven immortalization express high levels of CD23 6 7 . We examined the kinetics of expression of CD23 on CD23⁺ cells at early times following exposure of total primary B cells to EBV. Figure 1 shows that following exposure to EBV, the fraction of CD23⁺ cells that expressed high levels of CD23 (CD23^hi) increased. There was a rapid shift in the pattern of CD23 expression from 2.3% CD23^hi cells at time 0 to 59.6% CD23^hi cells at 90 h resulting in a 25.9-fold increase at 90 h relative to time 0. Therefore, exposure to EBV resulted in sub-populations of B cells with differential levels of expression of CD23. No alterations in expression of CD23 from baseline were observed when un-infected cells were cultured (data not shown).

To distinguish distinct sub-populations within CD23⁺ B cells early after exposure to EBV, we correlated expression of CD23 with expression of other B cell markers including CD58, IL6, CD57, CD86, HLA Class II, PD1, and IL10. Table 1 shows that there was a rapid increase in the fraction of CD23⁺ cells that expressed CD58 in the first 18 h relative to time 0 (9.7-fold in subject 1 and 21.5-fold in subject 2). While nearly a third of CD23^- cells expressed CD58 at time 0, there was less than a two-fold increase over the duration of the experiments. Thus among CD23⁺ cells, there was a rapid increase in expression of CD58.

Intracellular expression of IL6 increased in CD23⁺ cells by 42 to 66 h (15.8-fold at 42 h in subject 1 and 24.1-fold at 66 h in subject 2) as compared to time 0. Very few CD23^- cells expressed IL6. CD57 expression increased substantially on CD23⁺ cells (15.8-fold) at 42 h in only subject 1. No substantial change in expression on CD23⁺ cells was observed for the other molecules. Although levels of expression of CD58 (LFA3), an adhesion molecule, and the cytokine IL6 altered most rapidly following exposure to EBV, expression of CD58 on CD23⁺ cells was the earlier marker of infection with EBV.

Next, we examined whether there was a correlation between the level of expression of CD23 and CD58. Figure 2 shows B cells from a representative healthy EBV-seropositive subject (A) and a representative healthy EBV-seronegative subject (B) that were un-treated or exposed to EBV for four days. When compared with un-infected cells, two distinct sub-populations of CD23⁺ cells emerged after EBV infection. Sub-population R2 was characterized by expression of CD58 and low levels of expression of CD23 (CD23^loCD58⁺) while sub-population R3 expressed CD58 and high levels of CD23 (CD23^hiCD58⁺). Thus, two distinct sub-populations of B cells could be identified based on presence of CD58 and either high or low level of expression of CD23 early after exposure to EBV of cells from both EBV-seropositive and -seronegative individuals.

Since CD58 and IL6 were expressed early on CD23⁺ cells, we asked if any of these markers correlated with proliferation of cells. To examine proliferation, we labeled B cells with CFSE prior to exposure to EBV. CFSE is redistributed equally among daughter cells resulting in approximate halving of fluorescence intensity with each round of proliferation. A representative experiment shows that five days after exposure to EBV, 15.3% (G1+G2+G3) of CD58⁺ cells had proliferated (Figure 3A) and approximately 5% of cells had undergone more than one round of proliferation. In comparison, none of the CD58^- cells had undergone more than one round of proliferation. Among CD58⁺ cells, 38.5% expressed IL6 while only 0.3% of CD58^- cells expressed IL6. Thus, although CD58 expression was associated with expression of IL6, proliferating cells expressed CD58 but not IL6. Results of Figure 3A show that there were at least three sub-groups of CD58⁺ cells: one group that proliferated but did not express IL6, another that expressed IL6 but did not proliferate, and a third that did neither.

To determine how expression of CD58 correlated with proliferation, we temporally followed the evolution of the two CD23⁺CD58⁺ sub-populations and the CD23⁺CD58^- sub-population after exposure to EBV. Figure 3B shows that there was an approximate doubling in the fraction of cells in R3 (CD23^hiCD58⁺ cells) from 2.3% to 4.8% between day 5 and day 6 and from 4.8% to 9.5% between day 6 and day 7. In contrast, the two other sub-populations R1 (CD23^loCD58^- cells) and R2 (CD23^loCD58⁺ cells) did not show a similar doubling. Tracking the three sub-populations from day 4 to day 30 revealed a progressive increase in the fraction of CD23^hiCD58⁺ cells resulting in the exclusive presence of CD23^hiCD58⁺ cells by day 30 (Figure 3C, upper panels). Un-infected cells showed no outgrowth of LCL (Figure 3C, lower panels) and were dead after 10 days by Trypan blue exclusion (data not shown). These findings suggested that CD23^hiCD58⁺ cells were likely to be proliferating. Figure 3C also shows that neither CD23^loCD58⁺ cells nor CD23^hiCD58⁺ cells emerged when B cells from EBV-seropositive individuals were placed in culture in the absence of exogenously added EBV.

Proliferation of cells in each sub-population was examined by exposure of CFSE-labeled B cells to EBV for five days (Figure 3D). Nearly 80% of CD23^hiCD58⁺ cells (R3) had proliferated with 31.7% cells in G1, 38.8% of cells in G2, and 7.8% of cells in G3. In contrast, the vast majority of CD23^loCD58^- cells (89.4% in R1) and CD23^loCD58⁺ cells (79.5% in R2) had not proliferated. The earliest time at which proliferation was observed was four days after exposure to EBV (data not shown). Simultaneous examination for expression of IL6 revealed that 54.4% of CD23^loCD58⁺ cells expressed IL6. Minimal to no IL6 expression was observed in CD23^hiCD58⁺ cells (7%) and CD23^loCD58^- cells (0%). Thus, expression of IL6 and proliferation were mutually exclusive (Figure3A and 3D). The expression pattern CD23^hiCD58⁺IL6^- was characteristic of cells that underwent proliferation.

We examined whether CD23^hiCD58⁺ cells were able to proliferate in the absence of EBV-exposed non-proliferating B cells. Three days after exposure to EBV, FACS-sorted CD23^hiCD58⁺ cells, representing 0.5% of the culture, were re-introduced into culture after mixing with un-infected autologous primary B cells as feeder cells to maintain CD23^hiCD58⁺ cells at 0.5% of the culture. Three day old pre-sort culture and post-sort CD23^hiCD58⁺ cells are shown in Figure 4A and B, respectively. Un-infected, EBV-exposed, EBV-exposed but mock-sorted, and mixed culture of sorted-CD23^hiCD58⁺ cells plus un-infected cells were harvested on day 7. Un-infected B cells gave rise to only 0.04% CD23^hiCD58⁺ cells (C). Un-disturbed EBV-exposed B cells gave rise to 9.7% (D) and mock-sorted EBV-exposed B cells gave rise to 7% (E) CD23^hiCD58⁺ cells demonstrating that manipulation of cells during staining and sorting did not substantially hinder proliferation of CD23^hiCD58⁺ cells. In contrast, the percentage of CD23^hiCD58⁺ cells four days after mixing sorted-CD23^hiCD58⁺ cells with un-infected cells remained unchanged (F) as compared to the fraction of CD23^hiCD58⁺ cells in the pre-sorted population (A) suggesting dependence of CD23^hiCD58⁺ cells on the non-proliferating EBV-exposed sub-populations.

We asked whether susceptibility to proliferation or IL6 expression was pre-determined by the differentiation state of B cells before infection. Peripheral B cells were FACS-sorted into CD27⁺ memory and CD27^- naïve B cells. CD27 is considered a general marker for peripheral memory B cells 17 18 . Sorting strategy and purity of sorted populations of a representative experiment are shown in Figure 5A. Simultaneous staining of B cells with CD27 or matched isotype control antibody allowed detection of CD27⁺ cells prior to sort (data not shown). Sorted cells were examined on day 5 after exposure to EBV for expression of CD23 and IL6 (Figure 5B). Since CD23^hiCD58⁺ cells underwent proliferation (Figure 3D) and nearly all CD23^hi cells also expressed CD58 after exposure to EBV (Figure 2), 5% of memory B cells (CD23^hi) were predicted to proliferate after exposure to EBV (Figure 5B). In comparison, only 0.1% of naïve B cells were CD23^hi. Nearly 30% of memory B cells but only 0.2% of naïve B cells expressed IL6. Among memory B cells, expression of IL6 and expression of high levels of CD23 were mutually exclusive. Both sub-populations of B cells, whether susceptible to proliferation or expression of IL6, were derived from memory B cells. Thus, memory or naïve B cells did not exclusively give rise to either proliferating or IL6-expressing cells.

To determine which sub-populations of CD23⁺ cells expressed viral latency genes, CD23^loCD58^- (R1), CD23^loCD58⁺ (R2), and CD23^hiCD58⁺ (R3) cells were FACS-sorted four days after exposure of B cells to EBV. A representative experiment showing sorting strategy and purity of sorted sub-populations is shown in Figure 6A. Both CD23^hiCD58⁺ and CD23^loCD58⁺ sub-populations expressed EBNA1, LMP1, and EBNA2 transcripts, a pattern consistent with viral type 3 latency which is characteristic of EBV-immortalized B cells (Figure 6B). In comparison, CD23^loCD58^- cells expressed only EBNA2 mRNA. Since expression of mRNA often does not correlate well with expression of protein and because qRT-PCR is not informative about the fraction of cells expressing a gene of interest or the level of expression at the single cell level, we examined EBV-exposed cells for expression of LMP1 protein at the earliest possible time, on day 3, by flow cytometry. Figure 6C shows that 87% of CD23^hiCD58⁺ cells and 82.3% of CD23^loCD58⁺ cells expressed similar high levels of LMP1 when compared with isotype control antibody-staining of corresponding sub-populations. In contrast, only 4.8% of CD23^loCD58^- cells expressed low levels of LMP1. Sorted cells also demonstrated a similar pattern of expression of LMP1 in CD23^hiCD58⁺ and CD23^loCD58⁺ cells by immunofluorescence (data not shown). Thus, both distinct sub-populations of cells namely CD23^hiCD58⁺ IL6^- and CD23^loCD58⁺ IL6⁺ expressed viral latency genes including LMP1, a critical oncoprotein necessary for growth transformation in LCL 19 ; yet only one of these populations proliferated. This data also demonstrated that lack of proliferation of CD23^loCD58⁺ cells was not due to lack of transition from an early EBNA2-expression stage to subsequent LMP1-expression in this sub-population. Both proliferating and IL6-expressing sub-populations also expressed high levels of EBNA2 transcript. Characteristics of the two sub-populations CD23^hiCD58⁺IL6^- and CD23^loCD58⁺IL6⁺ are shown in Table 2.

Figure 6D shows that after exposure of sorted memory or naïve B cells (as in Figure 5) to EBV, 61.6% of memory B cells and 19.2% of naïve B cells were LMP1-positive. Within memory cells, 30.7% (18.9/[18.9 + 42.7]) of LMP1⁺ cells expressed IL6 while 25.8% (9.9/[9.9 + 28.5]) of LMP1^- cells expressed IL6. Although LMP1⁺ cells contributed to two-thirds of the IL6-expressing population, LMP1^- cells contributed to the other third. It is unclear whether this latter population is infected with EBV. Nearly 20% of naïve B cells expressed LMP1; yet no IL6 expression was observed. Among memory B cells, IL6-producing cells were CD23^lo, non-proliferating cells (Figure 5B) and LMP1⁺ cells largely contributed towards IL6 expression (Figure 6D), suggesting that the majority of IL6-producing cells were non-proliferating, CD23^lo, LMP1⁺, memory B cells.

Peripheral Blood Mononuclear Cells (PBMC) were obtained from healthy adults (five males and two females) as described 34 . The use of human subjects was approved by the Human Investigation Committee at Yale University. Informed written consent was obtained from volunteers. EBV-seroreactivity was determined by presence of antibodies to EBNA1 and viral capsid antigen using Western Blot. Experiments were performed with cells from five EBV-seropositive and two EBV-seronegative individuals. Experiments were repeated with cells from up to six subjects. B cells were isolated using negative selection by immunomagnetic-depletion of CD3⁺ cells (Invitrogen) with the exception of experiments in Figure 5 and Figure 6D in which B cells were positively selected and sorted by FACS. Negative selection of B cells was performed to avoid inadvertent activation of B cells during isolation. At least 95% of CD19⁺ B cells expressed CD21, the receptor for EBV (data not shown). Monocytes were depleted by adherence to plastic ( 35 ; data not shown).

EBV was isolated from the supernatant of B95-8 cells as described 36 . Infectivity of virus preparations was assessed by infection in triplicate of EBV-negative BJAB cells with serial dilutions of virus. After 48 h of culture, cells were examined for expression of EBNA by indirect immunofluorescence as described 37 and virus titer was calculated. Infections were performed using titered EBV at multiplicity of infection of 50-100 to maximize the number of infected B cells. After incubation of cells with virus for two hours at 37°C, cells were washed twice and placed in culture in the presence of 5% CO₂ at 2 × 10⁶ ml^-1 in RPMI 1640 containing 10% FBS.

Carboxyfluorescein diacetate, succinimidyl ester (CFSE; Invitrogen) labeling of un-infected B cells was performed as described 38 . We had experimentally determined that 2 μM CFSE allowed detection of proliferation for up to four generations with minimal toxicity to cells (data not shown).

Cells were surface-stained with saturating concentrations of antibodies including anti-CD23-PE (BD Bioscience), anti-CD23-FITC (Dako), anti-CD23-biotin (BD Bioscience), anti-CD58-FITC (ABD Serotec), anti-CD58-PE-Cy5 (Biolegend), anti-CD58-biotin (Gene Tex, Inc.), anti-CD86-APC (BD Bioscience), anti-CD57-FITC (BD Bioscience), anti-MHC Class II- FITC (BD Bioscience), anti-PD1-APC (eBioscience), and anti-CD27-biotin (Biolegend). Murine Ig at 1 mg ml^-1 was included to inhibit nonspecific binding. Isotype-matched control antibodies included murine IgG1-PE, IgG1-FITC, IgM-FITC, IgG2a-FITC, IgG1-APC, IgG1-PE-Cy5, and IgG1-biotin. Biotinylated antibodies were detected using Avidin-PE-Cy7 (BD Bioscience) or Avidin-FITC (Zymed). For intracellular staining, cells were fixed and permeabilized with Cytofix/Cytoperm (BD Pharmingen) or 90% cold methanol (for LMP1 staining). Antibodies included anti-IL6-FITC (eBioscience), anti-IL6-APC (Biolegend), anti-IL10 Pacific blue (eBioscience), and anti-LMP1 (CS1-4, Dako) followed by anti-mouse IgG1-PE or FITC. Isotype control antibodies included murine IgG2b-FITC, rat IgG1κ-APC, rat IgG1-Pacific blue, and murine IgG1. Data was acquired on LSR II (BD Bioscience) or FACS Calibur and analyzed using WinMDI. Gates were set on live lymphocytes based on forward- and side-scatter profiles. Quadrants or gates were drawn after comparing cells stained with an antibody of interest to cells stained with a matched isotype control antibody. For analysis of cell proliferation, gates were manually drawn on live cells for each CFSE peak as described 39 40 . Unlabeled but EBV-exposed cells and labeled but un-infected cells were used as controls. Sorting was performed on FACS Vantage or Aria Cell Sorter.

Primer sequences targeting the BamW fragment of EBV were: Forward: 5'GACTCCGCCATCCAAGCCTAG3'; Reverse: 5'TGGACGAGGACCCTTCTACGG3'.

Relative transcript levels of EBV latency genes were determined by real-time reverse transcription-PCR (qRT-PCR) of cDNA from sorted cells using gene-specific primers. Primer sequences were: EBNA1: Forward: 5'AGGGGAAGCCGATTATTTTG3';

Reverse: 5'CTCCTTGACCACGATGCTTT3'; LMP1: Forward: 5'TGAGTGACTGGACTGGAGGA3'; Reverse: 5'GGCTCCAAGTGGACAGAGAA3'; EBNA2: Forward: 5'CGGTCCCCGACTGTATTTTA3'; Reverse: 5'GGCTCTGGCCTTGAGTCTTA3'. Relative expression levels were calculated using standard curves generated from serial dilutions of EBV⁺ Akata Burkitt Lymphoma cell total RNA and normalized to 18 S rRNA.