High Physiological Levels of LMP1 Result in Phosphorylation of eIF2α in Epstein-Barr Virus-Infected Cells

Abstract

LMP1 is an Epstein-Barr virus (EBV)-encoded membrane protein essential for the proliferation of EBV-infected lymphoblasts (E. Kilger, A. Kieser, M. Baumann, and W. Hammerschmidt, EMBO J. 17:1700-1709, 1998). LMP1 also inhibits gene expression and induces cytostasis in transfected cells when it is expressed at levels as little as twofold higher than the average for EBV-positive lymphoblasts (M. Sandberg, A. Kaykas, and B. Sugden, J. Virol. 74:9755-9761, 2000; A. Kaykas and B. Sugden, Oncogene 19:1400-1410, 2000). We have found that in three different clones of EBV-infected lymphoblasts the levels of expression of LMP1 in individual cells in each clone ranged over 100-fold. This difference is due to a difference in levels of the LMP1 transcript. In these clones, cells expressing high levels of LMP1 incorporated less BrdU. We also found that induction of expression of LMP1 or of a derivative of LMP1 with its transmembrane domain fused to green fluorescent protein instead of its carboxy-terminal signaling domain resulted in phosphorylation of eIF2α in EBV-negative Burkitt's lymphoma cells. This induction of phosphorylation of eIF2α was also detected in EBV-infected lymphoblasts, in which high levels of LMP1 correlated with high levels of phosphorylation of eIF2α. Our results indicate that inhibition of gene expression and of cell proliferation by LMP1 occurs normally in EBV-infected cells.

Latent membrane protein 1 (LMP1), an Epstein-Barr virus (EBV)-encoded oncoprotein, mimics an active cellular receptor, CD40, but in a ligand-independent manner (15, 36). LMP1 stimulates multiple signaling pathways, including NF-κB-, JNK-, and STAT-mediated transcription (10, 14, 29). As an oncoprotein, LMP1 is essential for maintaining the proliferation of EBV-infected lymphoblasts (9, 26), it transforms rodent fibroblasts to proliferate anchorage independently (1, 38), and it also induces B-cell lymphomas in transgenic mice late in their lives (28). LMP1 is an integral membrane protein consisting of a short amino cytoplasmic domain, a six-membrane-spanning domain, and a 200-amino-acid carboxy cytoplasmic domain. The carboxy cytoplasmic domain binds TRAFs, TRADD, and JAK3 and is essential for LMP1's ability to maintain proliferation of EBV-positive lymphoblasts and to activate multiple signaling pathways (9, 10, 14, 20, 29, 30, 33).

In contrast to its ability to stimulate signaling and cell proliferation positively, LMP1 also inhibits RNA accumulation and general protein synthesis and induces cytostasis when it is expressed at levels as little as twofold higher than the average for EBV-infected cells (12, 24, 34). This inhibition requires the first two membrane-spanning domains of LMP1, but the mechanism of inhibition is unknown (5). Because all of the studies of this inhibitory activity have been carried out in transfected cells, it was unclear whether the inhibition occurs in EBV-infected cells and thus whether it could affect the EBV life cycle.

We have sought to understand this inhibitory function of LMP1 by first determining if it occurs in EBV-infected cells. We have found that in three different clones of EBV-infected lymphoblasts, the levels of expression of LMP1 in individual cells in each clone range over 100-fold. This variation in LMP1 protein levels correlated with LMP1 RNA levels. Cell proliferation was diminished in individual cells expressing high levels of LMP1, which is consistent with the notion that high natural levels of LMP1 inhibit cell proliferation and indicates that the inhibitory effect of LMP1 occurs under physiological conditions. We next investigated the mechanism by which LMP1 inhibits gene expression. We found that higher levels of LMP1 result in the phosphorylation of eukaryotic translation initiation factor 2 (eIF2α) that is dependent upon LMP1's transmembrane domains but not its carboxy-terminal signaling domain. Importantly, this induction of phosphorylation of eIF2α occurs normally in EBV-infected cells. In EBV-infected clones, cells expressing high physiological levels of LMP1 had high levels of phosphorylation of eIF2α. Our results indicate that LMP1 inhibits gene expression and cell proliferation when expressed at high physiological levels. This inhibition occurs as one extreme of the normal spectrum of activities mediated by LMP1.

MATERIALS AND METHODS

Cell culture and plasmids.

721 is an EBV-positive lymphoblastoid cell line (23). 11/17-2 and 11/17-3 are EBV-infected lymphoblastoid B-cell clones that have not passed through crisis (35). Cell lines BJAB/HA-LMP1, BJAB/HA6MLMP1-GFP, and BJAB/streptavidin were made by two steps of cloning. BJAB, an EBV-negative Burkitt's lymphoma cell line, was first transfected with plasmid p2547. p2547 encodes a fusion of the tetracycline repressor to KRAB (8), driven by the cytomegalovirus (CMV) immediate-early (IE) promoter, and a bicistronic transcript, EBNA1-IRES-G418R, driven by a phosphoglucokinase promoter (32). A BJAB/p2547 cell line expressing both Tet-KRAB and EBNA1 was used as the parental cell line for further constructing LMP1- and streptavidin-inducible cell lines. p2426 was used to construct a tetracycline-inducible expression plasmid for HA-LMP1, HA6MLMP1-GFP, and streptavidin. p2426 was identical to pcDNA4/TO (Invitrogen) except that the open reading frame for the zeocin-resistant gene was replaced with a puromycin-resistant gene. The expression plasmid for HA-LMP1 (p2587) was generated by cloning a beta-globin intron and hemagglutinin (HA)-tagged LMP1 cDNA downstream of the CMV IE promoter of p2426. The expression plasmid for HA6MLMP1-GFP (p2568) was generated by inserting the coding sequence of this derivative downstream of the CMV IE promoter of p2426. The expression plasmid for streptavidin (p2772) was generated by cloning a beta-globin intron and a streptavidin coding sequence downstream of the CMV promoter of p2426. The BJAB/HA-LMP1, BJAB/HA6MLMP1-GFP, and BJAB/streptavidin cell lines were constructed by transfecting p2587, p2568, and p2772, respectively, into BJAB/p2547 cells and selecting for puromycin-resistant clones. The expression of HA-LMP1, HA6MLMP1-GFP, and streptavidin was confirmed by Western blotting. 721, 11/17-2, and 11/17-3 cells were grown in RPMI 1640 supplemented with 10% fetal bovine serum. The BJAB/HA-LMP1, BJAB/HA6MLMP1-GFP, and BJAB/streptavidin cells were grown in RPMI 1640-10% fetal bovine serum-1 mg of G418/ml-1 μg of puromycin/ml. All cell culture media were supplemented with 200 U of penicillin/ml and 200 μg of streptomycin/ml, and all cells were grown at 37°C in a humidified 7% CO₂ atmosphere.

Immunofluorescent staining of cells.

For the staining of LMP1, 721, 11/17-2, and 11/17-3, cells were either fixed with a 1:1 mixture of acetone-methanol on ice for 5 min or fixed with 4% paraformaldehyde and then permeabilized with 0.1% Triton X-100. Acetone-methanol-fixed cells were used later for Western blotting and extraction of total DNA. Paraformaldehyde-fixed cells were used later for extraction of total RNA. For the staining of CD40, live cells were used. The fixed or live cells were washed two times with 1× phosphate-buffered saline (PBS) and blocked in 1× PBS plus 5% calf serum (CS); stained with either the mouse monoclonal anti-LMP1 antibody CS1-4 (Dako) at 1:1,000, affinity-purified rabbit polyclonal anti-LMP1 antibodies at 1:500, or rabbit polyclonal antibodies recognizing the amino terminus of CD40 at 1:1,000 (Santa Cruz); washed three times with 1× PBS plus 5% CS; and finally stained with appropriate secondary antibodies labeled with AlexaFluor 488 (Molecular Probes). For double staining of bromodeoxyuridine (BrdU) and LMP1, BrdU-labeled and acid-treated cells (see below) were stained with both a mouse monoclonal anti-BrdU antibody (Roche) at 1:200 and rabbit polyclonal anti-LMP1 antibodies at 1:500, washed, and then stained with both AlexaFluor 488-labeled goat anti-mouse (Molecular Probe) and Cy5-labeled goat anti-rabbit antibodies. The stained cells were analyzed on a FACSCalibur machine (Becton Dickinson). Data were analyzed with either CellQuest or FlowJo software. The coefficient of variation (CV) was calculated by the software using the following formula: standard deviation/mean.

Real-time PCR for measuring EBV DNA.

TaqMan probe (ABI)-based real-time PCR was used to measure the copies of EBV-DNA in 721 cells. Total cellular DNA was extracted using DNeasy kits (Qiagen). For each sample, 10 ng of total cellular DNA was used to amplify a fragment of the actin gene using the primers TCACCCACACTGTGCCCATCTA and TGAGGTAGTCAGTCAGGTCCCG and to amplify a fragment of OriP using the primers GGCGCAAGTGTGTGTAATTTGT and GGGCGGGCCAAGATAGG. The TaqMan probe for actin was ATGCCCTCCCCCATGCCATCCTGCGT. The TaqMan probe for OriP was CTCCAGATCGCAGCAATCGCGCT. Serial dilutions of linearized plasmid containing the actin fragment or OriP were used as standards. The real-time PCR was run on an ABI PRISM 7700 sequence detection system. In all cases, the signals measured for the OriP probe were normalized to those measured for actin to give the relative number of OriP molecules detected per cell.

Real-time PCR for measuring LMP1 cDNA.

721 or 11/17-3 cells were fixed with paraformaldehyde, stained for LMP1, and either sorted for high and low levels of LMP1 or unsorted. One million fixed cells were resuspended in 200 μl of proteinase K solution (20 mM Tris-HCl, pH 8.0, 1 mM CaCl₂, 200 mM NaCl, 2% sodium dodecyl sulfate [SDS], 0.5 mg of proteinase K/ml) and incubated at 50°C for 30 min. Total cellular RNA was extracted from the proteinase K-treated cells using Trizol reagent (Invitrogen) following the manufacturer's instructions. Total RNA (from 2.5 × 10⁵ cells) was reverse transcribed using random hexamers and Omniscript Reverse Transcriptase (Qiagen) following the manufacturer's instructions. cDNA from 10⁴ cells was used for each real-time PCR.

LMP1 and actin cDNAs were quantified by TaqMan probe-based real-time PCR. A fragment of LMP1 cDNA was amplified by the primers CTCATCGCTCTCTGGAATTTG and A(T/G)ACCTAAGACAAGTAAGCACC. A fragment of beta actin cDNA was amplified by the primers AGAAAATCTGGCACCACACCTT and CTCAAACATGATCTGGGTCATCTTC. The underlined primers for LMP1 and actin encompassed introns of LMP1 and actin, respectively, and therefore allowed specific detection of the cDNAs of LMP1 and actin. Using these primer pairs, no PCR product was detected when reverse transcriptase was omitted in reverse transcription (data not shown). The TaqMan probe used for detecting LMP1 cDNA was CAGGCATTGTTCCTTGGAATTGTGCTG. The TaqMan probe used for detecting actin cDNA was AATGAGCTGCGTGTGGCTCCCGAG. Tenfold serial dilutions of cDNA from 10⁶ cells were used as standards to determine the relative numbers of copies of LMP1 and actin cDNAs in the samples. The real-time PCR was run on an ABI PRISM 7700 sequence detection system.

Cell sorting.

Exponentially growing cells (<5 × 10⁵ cells/ml) were fixed and stained with CS1-4- and AlexaFluor-488-labeled antibodies as described above. The stained single cells were sorted by a FACSVantage sorter (Becton Dickinson) based on the intensity of the staining. The sorted cells were then solubilized with 1× SDS-polyacrylamide gel electrophoresis (PAGE) sample buffer made from a dilution of a 2× stock solution with 1× RIPA buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholate, 0.5% SDS).

Labeling cells with BrdU.

Exponentially growing cells were incubated in culture medium containing 100 μM BrdU for 30 min. The labeled cells were washed twice with 1× PBS and then fixed with 95% ethanol. The fixed cells were treated with 1 M HCl for 15 min; neutralized by three washings with 50 mM NaCl-100 mM Tris, pH 7.4; and then used for immunofluorescent staining as described above.

Cell cycle analysis.

Cells were stained with either propidium iodide (PI) or Hoechst 33342 for cell cycle analysis. To stain cells with PI, the fixed cells were incubated with DNase-free RNase (1 μg/ml) at 37°C for 15 min and then resuspended in 1× PBS containing 50 μg of PI/ml. To stain cells with Hoechst 33342, the cells were incubated with 1× PBS containing 2.5 μg of Hoechst 33342/ml at room temperature for 15 min.

Western blotting.

Exponentially growing cells were harvested and lysed in 1× RIPA. For analysis of the phosphorylation of eIF2α, it is important to harvest cells at a density of <5 × 10⁵/ml in order to obtain low backgrounds of phosphorylation of eIF2α. Cell lysates (4 × 10⁴ cells per sample) were separated by SDS-10% PAGE for eIF2α or SDS-9% PAGE for LMP1 and EBNA2 and transferred to nitrocellulose membranes. The blots were blocked with 5% nonfat milk and probed with primary antibodies followed by fluorescein isothiocyanate-labeled, alkaline phosphatase-labeled, or ³⁵S-labeled secondary antibodies (Amersham). The signals were quantified by using ImageQuant software. The following primary antibodies were used in the experiments: affinity-purified rabbit anti-LMP1 antibodies at 1:500, a monoclonal anti-EBNA2 antibody (R3) (27) at 1:500, polyclonal antibodies specifically recognizing phosphorylated eIF2α containing a phosphate at serine 51 (Resgen) at 1:1,000, a monoclonal antibody recognizing total eIF2α (Biosource) at 1:2,000, the anti-HA antibody 12CA5 at 1:3,000, and a fluorescein isothiocyanate-labeled monoclonal anti-actin antibody (Sigma) at 1:1,000.

RESULTS

Levels of expression of LMP1 differ among individual EBV-infected cells.

LMP1 inhibits cell proliferation and gene expression when it is expressed in EBV-negative cells at levels as little as twofold higher than the average found in EBV-infected lymphoblasts (12, 24, 34). To assess the possible generality of this inhibition, we measured the levels of expression of LMP1 in individual cells among clones of EBV-positive cells. We examined the levels of expression of LMP1 in three different clones of EBV-infected B cells, 721, 11/17-2, and 11/17-3, by fluorescence-activated cell sorter (FACS) analysis. 721 is an immortalized lymphoblastoid cell line. 11/17-2 and 11/17-3 are EBV-infected B-cell clones that have not gone through crisis. Staining all three clones after their fixation with anti-LMP1 antibodies generated a FACS profile with a broad distribution, indicating that levels of expression of LMP1 differed dramatically among cells of each clone (Fig. 1A). Such a wide distribution of levels of expression is specific to LMP1 and is not detected for a cellular receptor, CD40, that LMP1 mimics functionally (Fig. 1A). The CV of the LMP1 FACS profile was ∼2-fold greater than that of the CD40 profile. Western blotting was used to confirm the results of FACS analysis and to quantify the difference in the levels of LMP1. 721 and 11/17-3 cells were fixed and stained with mouse anti-LMP1 antibodies (CS1-4) and sorted based on levels of fluorescence intensity (Fig. 1B and C). Cells having the highest 5% and the lowest 5% of staining were collected and examined by Western blotting using rabbit anti-LMP1 antibodies (Fig. 1B and C). We noticed that cells having the lowest 5% of staining were consistently smaller than cells having high and intermediate levels of staining (data not shown). The difference in levels of LMP1 between the highest and the lowest 5% of cells was at least 30-fold in 721 cells (Fig. 1B) and was >100-fold in 11/17-3 cells (Fig. 1C; also see Fig. 4). The levels of actin in the cells expressing moderate and high levels of LMP1 were ∼2-fold higher than those in cells expressing low levels of LMP1, a finding consistent with the observation that the lowest 5% of cells were smaller. We also examined the expression of EBNA2, a viral protein important in stimulating LMP1 transcription (21). We did not detect any difference between the levels of EBNA2 in cells expressing the lowest and the highest levels of LMP1 (Fig. 1B and C), indicating that the wide range of protein levels did not obtain for all viral proteins but appears to be specific for LMP1.

FIG. 1.

Levels of expression of LMP1 vary widely in EBV-infected lymphoblasts. (A) FACS analysis of 721, 11/17-2, and 11/17-3 cells stained with CS1-4 anti-LMP1 and anti-CD40 antibodies. The distribution of levels of expression of LMP1 is much broader than that of CD40 expression levels. The ratio of CVs was calculated by calculating the CV for LMP1 and dividing by the CV calculated for CD40. (B) 721 cells were fixed and stained with CS1-4 mouse anti-LMP1 antibody and sorted based on the intensity of staining. The FACS profile represents the unsorted cells and reanalysis of those sorted cells in the highest and the lowest 5% of staining. The unsorted and sorted cells (4 × 10⁴) were analyzed by quantitative Western blotting for LMP1 using rabbit anti-LMP1 antibodies and for EBNA2 and beta-actin. The relative levels of each protein are given. The levels of LMP1 differed by >30-fold, while the levels of EBNA2 and actin differed by no more than 2-fold. (C) 11/17-3 cells were fixed, stained, sorted, and analyzed by Western blotting as described for panel B. The data in panels B and C are representative of three experiments.

FIG. 4.

Expression of LMP1 results in phosphorylation of eIF2α in EBV-negative cells. Two BJAB cell lines in which HA-tagged wild-type LMP1 (A) or a derivative of LMP1 with the transmembrane domain of LMP1 fused to GFP (B) can be inducibly expressed by tetracycline were constructed. BJAB-streptavidin control cells (see the text) or BJAB/HA-LMP1 or BJAB/HA6MLMP1-GFP cells were untreated or treated with 1 or 10 ng of tetracycline/ml for 40 h. The cells were harvested and analyzed by Western blotting using anti-HA (C), anti-phosphorylated-eIF2α (E), and anti-total eIF2α (D) antibodies. The relative levels of phosphorylated eIF2α are indicated. Representative example of three independent experiments are shown. (F) Quantification of induction of phosphorylation of eIF2α in BJAB/HA-LMP1 and BJAB/HA6MLMP1-GFP cells. The levels of phosphorylation of eIF2α in cells not treated with tetracycline were set as 1. The error bars indicate standard deviations.

We asked whether the levels of LMP1 RNA correlate with LMP1 protein levels. To quantify LMP1 RNA, we performed real-time RT-PCR on fixed and sorted 721 and 11/17-3 cells using primers specific to LMP1 cDNA (see Materials and Methods). Cells with the highest 5% of LMP1 expression consistently had ∼16 to 30-fold more LMP1 cDNA and ∼2 to 3-fold more actin cDNA than cells with the lowest 5% of LMP1 expression (Table 1). The levels of the cDNAs of both LMP1 and actin correlated with the levels of their proteins. It is surprising that there was on the order of 20- to 30-fold difference between the levels of LMP1 RNA in cells expressing high and low levels of LMP1, because the levels of EBNA2 were similar in those cells. EBNA2 is the major viral regulator of the LMP1 promoter (21). It remains to be determined whether the difference in LMP1 RNA levels is a result of different levels of transcription or of different half-lives of the RNAs in these cells.

TABLE 1.

Levels of LMP1 RNA differ in cells expressing high and low levels of LMP1 protein^a

Cell line	Expt no.	RNA	Relative no. of copies
			Lowest 5%	Highest 5%	Unsorted
721	1	LMP1	1	36.8	10.7
		Actin	1	2.6	2.0
	2	LMP1	1	16.2	2.2
		Actin	1	2.6	2.5
11/17-3	1	LMP1	1	31.0	6.9
		Actin	1	3.6	2.5

^a721 and 11/17-3 cells were fixed and sorted based on their levels of LMP1 protein. Real-time RT-PCR was performed as described in Materials and Methods. For each sample, cDNA from 10⁴ cells was used for PCR amplification. The relative levels of cDNAs of LMP1 and actin were calculated using a standard generated from serial dilutions of cDNA from 10⁶ cells. In all cases, the levels of LMP1 and actin from cells expressing the lowest levels of LMP1 were set as 1.

We examined whether a difference in the number of viral genomes contributes to the difference in levels of LMP1. Viral-genome numbers were measured in fixed and sorted 721 cells by real-time PCR. The results from the average of three independent experiments showed that cells with the highest 5% of LMP1 expression had threefold-higher numbers of viral genomes than cells with the lowest 5% of LMP1 expression. This small difference in the number of viral genomes is unlikely to account for the 30- to 100-fold differences in levels of LMP1 protein.

High levels of LMP1 correlate with diminished cell proliferation.

LMP1 induces cytostasis in transfected cells when it is expressed at high levels (12, 24). Given that the levels of expression of LMP1 varied dramatically in EBV-infected lymphoblasts, we assessed whether the proliferation of those cells expressing high levels of LMP1 is inhibited as it is in transfected cells expressing high levels of LMP1. We used incorporation of BrdU as an indicator of cell proliferation. In 721 cells, the percentage of BrdU-positive cells was ∼60% after 30 min of labeling. All of the cells in S phase and a fraction of the cells in G₁ and G₂/M had incorporated BrdU (Fig. 2A). This profile is similar to that of other EBV-positive lymphoblastoid cells (37). 721 cells expressing high and low levels of LMP1 had dramatically reduced numbers of cells incorporating BrdU relative to those expressing intermediate levels of LMP1 (Fig. 2C to E). In this experiment, the average percentage of BrdU-positive cells for the population was 56% (Fig. 2B). Of those cells with the highest 1% of LMP1 expression (Fig. 2E), only 23% had incorporated BrdU, while of the cells expressing intermediate levels of LMP1 (Fig. 2D), 60% had incorporated BrdU. Thus, higher levels of LMP1 correlated with fewer BrdU-positive cells. The incorporation of BrdU was also diminished in cells expressing low levels of LMP1. The percentage of BrdU-positive cells was 16% among those cells with the lowest 1% of LMP1 expression (Fig. 2C). This measurement was predicted, because LMP1 is essential for maintaining the proliferation of EBV-infected lymphoblasts.

FIG. 2.

EBV-infected lymphoblasts expressing high levels of LMP1 incorporated less BrdU than those expressing low levels of LMP1. 721 cells were pulsed with BrdU for 30 min, fixed, and stained with anti-BrdU and with either PI or anti-LMP1 antibodies. (A) Representative example of staining for BrdU and PI. (B) Representative example of four independent experiments staining for BrdU and LMP1. (C to E) The percentage of BrdU-positive cells, reflecting the population of cells undergoing DNA synthesis, was analyzed by FACS in cells expressing low (C), intermediate (D), or high (E) levels of LMP1.

To confirm the results of the incorporation of BrdU, we analyzed the relationship between the levels of LMP1 and the cell cycle. A greater proportion of the cells expressing high and low levels of LMP1 than of cells expressing intermediate levels of LMP1 were in G₁ (Fig. 3). The fractions of cells in G₁ were 50% for cells expressing intermediate levels of LMP1, 67% for cells with the lowest 3% of LMP1 expression, and 74% for cells with the highest 3% of LMP1 expression. In summary, these results, when combined with those from transfected cells (12, 24), indicate that high levels of LMP1 inhibit cell proliferation in EBV-infected cells.

FIG. 3.

Cells expressing high and low levels of LMP1 have higher proportions of cells in the G₁ phase of the cell cycle than those expressing intermediate levels. 721 cells were fixed and stained with Hoechst 33342 (A) or Hoechst 33342 and anti-LMP1 antibodies (B) and analyzed by FACS. The fraction of cells in G₁ was analyzed for the cells expressing intermediate levels of LMP1 (mid) and those in the highest 3% (high) and the lowest 3% (low) of LMP1 expression. (C) Averages (plus standard deviations) of the results from analyzing 11 independent subclones of 721 cells. P values were generated by the Wilcoxon rank sum test by comparing results for the whole population with that for cells expressing either high or low levels of LMP1.

LMP1 can result in phosphorylation of eIF2α in EBV-negative cells.

General protein synthesis can be inhibited by inducing phosphorylation of the alpha subunit of eIF2α at serine 51 during cellular responses to multiple forms of stress, including nutritional starvation, the unfolded protein response in the endoplasmic reticulum, viral infection, and heme deficiency in erythrocytes (7, 22). The fact that LMP1 inhibits general protein synthesis led us to ask whether LMP1 might do so through the induction of the phosphorylation of eIF2α. We constructed two BJAB cell lines in which HA-tagged wild-type LMP1 (HA-LMP1) or a derivative of LMP1 with an HA-tagged amino cytoplasmic domain (its transmembrane domain) but with its cytoplasmic signaling domain replaced by green fluorescent protein (GFP) (HA6MLMP1-GFP) are expressed inducibly by treating cells with tetracycline (Fig. 4). A control cell line in which streptavidin is inducibly expressed by tetracycline was also made. The levels of LMP1 in the HA-LMP1-inducible cells were ∼5-fold higher than the average in 721 cells when they were treated with 1 ng of tetracycline/ml and ∼50-fold higher when treated with 10 ng of tetracycline/ml (data not shown). The levels of induction of HA6MLMP1-GFP by tetracycline were similar to that of HA-LMP1 (Fig. 4). Treating the BJAB/HA-LMP1 cell line with tetracycline induced phosphorylation of eIF2α in a dose-dependent manner. The levels of phosphorylated eIF2α were increased ∼4-fold when cells were treated with 1 ng of tetracycline/ml and ∼12-fold when cells were treated with 10 ng of tetracycline/ml (Fig. 4). The inhibitory function of LMP1 is thought to be mediated through its transmembrane domains. Treating the BJAB/HA6MLMP1-GFP cells with tetracycline induced phosphorylation of eIF2α as efficiently as did treating the BJAB/HA-LMP1 cells, confirming this expectation. We did not detect induction of phosphorylation of eIF2α when the BJAB-streptavidin control cells were treated with 1 or 10 ng of tetracycline/ml. These results indicate that efficient expression of LMP1 results in the phosphorylation of eIF2α through the LMP1 transmembrane domain.

High levels of LMP1 correlate with high levels of phosphorylation of eIF2α in EBV-infected cells.

If LMP1 normally induced phosphorylation of eIF2α, it should be possible to detect this modification in EBV-infected cells. Fixed 11/17-3 cells were therefore sorted based on levels of expression of LMP1 and assayed for phosphorylation of eIF2α. Those cells in the highest 5% of LMP1 expression had 3-fold-higher levels of phosphorylated eIF2α than unsorted cells and ∼8-fold-higher levels than the 5% of the cells expressing the lowest levels of LMP1 (Fig. 5). These results indicate that induction of phosphorylation of eIF2α by LMP1 occurs in EBV-infected cells.

FIG. 5.

High levels of LMP1 correlate with high levels of phosphorylation of eIF2α in EBV-infected cells. 11/17-3 cells were fixed and stained with anti-LMP1 antibodies and sorted based on their intensities of staining. The cells in the lowest and the highest 5% of staining were analyzed for LMP1, phosphorylated eIF2α, and beta-actin by quantitative Western blotting. The relative levels of LMP and phosphorylated eIF2α are indicated. A representative example of three independent experiments is shown.

DISCUSSION

LMP1 inhibits gene expression and cell proliferation when expressed at high levels. We have found that efficient expression of LMP1 results in the phosphorylation of eIF2α. This activity of LMP1 requires its membrane-spanning domains but not its carboxy terminus, which is known to associate with cellular signaling molecules. This observation provides a mechanistic explanation for LMP1's inhibitory function. We have investigated whether the inhibition of cell proliferation by LMP1 can be detected in EBV-infected cells as a first step toward understanding its biological consequences. We found that differences among levels of expression of LMP1 in cells ranged up to 100-fold for three clones of EBV-infected lymphoblasts. In these clones, high levels of LMP1 correlated with diminished incorporation of BrdU and high levels of phosphorylation of eIF2α. In transfected cells, high levels of LMP1 induce cytostasis (12, 24) and result in phosphorylation of eIF2α (Fig. 4). Our observations of EBV-positive cells combined with those of transfected cells demonstrate that LMP1 normally results in the inhibition of gene expression and cell proliferation in EBV-infected cells in which LMP1 is expressed at high physiological levels.

Brennan et al. have shown that coexpression of a derivative LMP1 lacking its TRAF- and TRADD-binding domain (LMP1^AAAG) with wild-type LMP1 inhibits LMP1 signaling (3). This result is consistent with our finding that efficient expression of the transmembrane domain of LMP1 inhibits protein synthesis. Our results also indicate that the inhibition by LMP1^AAAG may not result from a dominant-negative action of the derivative but rather from its efficient expression, resulting in an inhibition of general protein synthesis.

It is striking that LMP1, which is essential for the proliferation of normal EBV-positive B lymphoblasts, is expressed at such different levels in cells of a single clone. We have demonstrated that LMP1 protein levels correlate with its RNA levels but not with those of the DNA template. Surprisingly, the viral transcriptional activator EBNA2 varied no more than twofold between the EBV-positive cells that express the lowest and those that express the highest levels of LMP1 RNA and protein. Many questions remain regarding the mechanism by which this wide distribution of expression of LMP1 is established. Is the difference in the LMP1 RNA level a result of different levels of transcription or a result of different half-lives of the RNAs in these cells? Do cells expressing low levels of LMP1 generate progeny expressing high levels of LMP1 and vice versa? What is the fate of cells in which LMP1 inhibition occurs? LMP1 can be measured only in fixed cells, so only a recombinant EBV expressing LMP1 fused to GFP might be able to answer these questions. Nevertheless, our studies indicate that EBV-positive cells proliferate best when LMP1 is expressed at intermediate levels.

There are four kinases in mammalian cells known to phosphorylate eIF2α at serine 51 which are candidates for mediating LMP1 inhibition of protein synthesis. They are PKR, PERK, GCN2, and heme-regulated inhibitor kinase (2, 4, 17, 39). In preliminary experiments in cells null for PKR, LMP1 inhibited gene expression as efficiently as it did in PKR-positive cell lines (unpublished data), indicating that PKR is unlikely to be involved in the LMP1-mediated inhibition of protein synthesis. We are investigating whether any of the other kinases known to act on eIF2α are activated by LMP1. LMP1 inhibition of gene expression and of cell proliferation is clearly not due simply to an accumulation of large amounts of a membrane protein in cells. Several derivatives of LMP1 which have truncated or mutated membrane-spanning domains do not inhibit gene expression and cell proliferation when expressed at levels higher than that required for wild-type LMP1 activity (24, 25).

One benefit LMP1 may provide EBV by affecting phosphorylation of eIF2α is to concomitantly affect the expression of certain cellular genes. Upon the induction of phosphorylation of eIF2α, a subset of genes is activated at both the transcriptional and translational levels despite global protein synthesis being diminished (18, 22). These genes include those for transcription factors, such as ATF4 and GCN4, and the chaperon Bip (16, 19) and genes containing internal ribosomal entry sites (IRES), such as those for c-Myc, PDGF2, and cat-1 (11, 13). Some of these genes, such as those for ATF4, GCN4, and Bip, are clearly involved in the adaptation of cells to stress (18, 31). It has been proposed that initiation of gene expression from cellular IRES elements may be required for cellular differentiation (6, 7, 13). LMP1, when expressed efficiently, may induce a subset of genes through the induction of phosphorylation of eIF2α that are advantageous to EBV in vivo. Comparing the differences between cells expressing high and low levels of LMP1 by using cDNA microarrays or proteomic analyses may be helpful in elucidating the contribution of LMP1 to the EBV life cycle.