Virally Induced Cellular MicroRNA miR-155 Plays a Key Role in B-Cell Immortalization by Epstein-Barr Virus

The indicator vector pNL-SIN-CMV-FLuc, which was used as an internal control for detection of miRNA function, has been previously described, as has pNL-SIN-CMV-RLuc (16). A derivative of pNL-SIN-CMV-RLuc bearing two fully complementary artificial target sites for cellular miR-155 (1552T) has also been described previously (18). For indicator assays for lymphoblastoid cell lines (LCLs), the segment of DNA encoding firefly luciferase (FLuc) or Renilla luciferase (RLuc)-1552T was removed from the pNL-SIN-CMV constructs and inserted into pLCE—the same backbone used for the lentiviral sponge constructs—using restriction sites NheI and XbaI. The resultant plasmids were named pLC-FLuc and pLC-RLuc-1552T. A lentiviral sponge specific for miR-155 was constructed as described previously (12, 17), and the control green fluorescent protein (GFP) and sCXCR4 vectors were published previously (17). For this study, a nine-copy sponge sequence specific for miR-155 (s155) was cloned into lentiviral vector pLCE as described previously (17) by using oligonucleotides 5′-CTAGGACCCCTATCACACCTAGCATTAAGTTTGACCCCTATCACACCTAGCATTAAGTTTGACCCCTATCACACCTAGCATTAATCTAGATTTGAATTC-3′ and 5′-AATTGAATTCAAATCTAGATTAATGCTAGGTGTGATAGGGGTCAAACTTAATGCTAGGTGTGATAGGGGTCAAACTTAATGCTAGGTGTGATAGGGGTC-3′.

The pMSCV/GFP and pMSV/s155 retroviral vectors were generated from pMSCV-puro (catalog no. 634401; Clontech) by excision of the GFP and GFP-s155 expression cassettes from the relevant pLCE-based vector by cleavage with NheI and EcoRI, followed by ligation of the resultant fragments into pMSCV-puro that had been cleaved with HpaI and EcoRI. Similarly, the pTRIPZ/GFP and pTRIPZ/s155 lentiviral vectors were generated from the tetracycline-regulatable vector pTRIPZ-puro (catalog no. RHS4696; Open Biosystems) by excision of the GFP and GFP-s155 expression cassettes from the pLCE-based vectors by cleavage with AgeI and EcoRI, followed by insertion of the resultant fragments into pTRIPZ-puro that had also been cleaved with AgeI and EcoRI.

cDNA libraries for Solexa/Illumina sequencing were prepared as previously described (46). Briefly, small RNAs were isolated by preparative gel electrophoresis and sequentially ligated to 3′ and then 5′ linkers. Primers complementary to the linker sequences were used for reverse transcription (RT) and PCR in order to generate cDNA libraries for deep sequencing. Raw sequencing data were filtered to remove reads lacking identifiable 3′ linker sequences and/or reads falling outside the predicted miRNA size range. A list of final usable reads was then collapsed down into a list of unique sequences which was aligned to the cellular and viral pre-miRNA database from miRBase (release 14.0) by using NCBI BLAST and further parsed using the blastoutparse and filter_alignment scripts of the miRDeep software package (14).

The EBV-positive BL cell lines Mutu I, Mutu III, Namalwa, and Raji, the EBV-positive DLBCL cell line IBL-1, and the EBV-negative BJAB cell line were all maintained in RPMI 1640 supplemented with 10% fetal bovine serum (FBS). EBV-infected human peripheral blood mononuclear cells (PBMCs) were cultured in RPMI 1640 with 15% FBS, and 293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% FBS. All cells were maintained in the presence of antibiotics.

The LCLs SDLCL, EF3D, and LCLd1 were generated by infection of PBMCs with EBV strain B95-8. PBMCs were isolated from buffy coats of normal donors (Carolina Red Cross) by using a Histopaque-1077 column (Sigma). A total of 10⁷ PBMCs were infected with 500 μl of filtered B95-8 virus stock in the presence of 0.5 μg/ml cyclosporine in R15 medium (RPMI 1640 with 15% FBS and 50 μg/ml gentamicin) for 1 h at 37°C. For outgrowth, infected cells were brought to 14.4 ml in R15 medium plus cyclosporine (0.5 μg/ml) and plated in either a 24-well plate for time course analyses (one well was harvested for each time point) and the formation of SDLCL and LCLd1 or a 96-well plate for clonal expansion to generate EF3D.

Lentiviral particles were produced by cotransfection of the pLCE vectors with pMDLgpRRE, pRSV-Rev, and pVSV-G (11) into 293T cells by using Fugene6. pNL-SIN constructs were also packaged in 293T cells but by cotransfection with pcRev, pcTat, and pHIT/G (30). pMSCV was packaged using mouse stem cell virus (MSCV) Gag/Pol and pHIT/G; pTRIPZ was packaged using pCMVΔR8.75 and pMD2G. Supernatants were collected, filtered, and concentrated with Amicon Ultra centrifugal filters (Millipore) 48 h after transfection. The lentivirus- or retrovirus-containing supernatants were then used to transduce the cell lines for which growth curves are shown in Fig. 2. For LCLs, transduction efficiency was approximately 30 to 50%; for all other cell lines except for IBL-1, it was 80 to 90%. MSCV transduction of IBL-1 was only ∼15% efficient; therefore, IBL-1 cells were grown in the presence of puromycin for 1 week to select for transduced cells. Without sorting, all transduced cell populations were a mixture of GFP-positive and GFP-negative cells.

Indicator assays for miRNA activity were performed using previously described lentiviral RLuc- and FLuc-based indicator vectors (16). Briefly, pNL-SIN-CMV-FLuc and either pNL-SIN-CMV-RLuc or pNL-SIN-CMV-RLuc-1552T were packaged into viral particles and then cotransduced (FLuc plus RLuc) into control 293T cells or target B cells. Two days posttransduction, cells were collected and lysed and both FLuc and RLuc levels were quantified using a dual luciferase reporter assay (Promega). FLuc/RLuc ratios were calculated and normalized to ratios obtained from 293T cells not expressing miR-155. For the inducible system, miRNA indicator vectors were transduced into LCLd1/GFP and LCLd1/s155 cells after these had been expanded in the presence of puromycin and the absence of doxycycline. After transduction, the cells were washed and resuspended in media containing doxycycline. GFP-positive cells were sorted 2 days later and immediately lysed for quantitation of luciferase (both RLuc and FLuc) as described above.

LCLs (EF3D and SDLCL) growing as mixed populations of GFP-positive cells (lacking miR-155 function) and GFP-negative cells (retaining miR-155 function) were split to 2 × 10⁵ cells/ml after lentiviral transduction as described above. Twenty-four hours later, while at equal concentrations and while still subconfluent, cells were labeled for 45 min with 25 μM 5-bromodeoxyuridine (BrdU; Calbiochem) and then collected. Cells were fixed with 4% paraformaldehyde on ice for 1 h, pelleted, and then washed twice with phosphate-buffered saline (PBS). Next, cells were permeabilized at room temperature in 0.5% Triton X-100 in PBS (PBS-T). After the cells were washed twice in PBS, DNase I was added and the samples were digested for 45 min at 37°C. Again, cells were washed, but this time they were washed in PBS-T. DNA was labeled with Alexa Fluor-647 anti-BrdU (catalog no. 560209; BD Pharmingen) and resuspended in propidium iodide (PI)-RNase staining buffer (BD Biosciences). A minimum of 10⁵ cells were analyzed using FACSCanto and FlowJo software.

For the detection of apoptotic cells (see Fig. 5C), SDLCL cells were again grown as mixed populations of GFP-positive and GFP-negative cells after transduction with lentiviral vectors as described above. While in exponential growth phase, cells were pelleted and stained with allophycocyanin (APC)-annexin V (BD Biosciences) in binding buffer (0.01 M HEPES, pH 7.4, 0.14 M NaCl, 2.5 mM CaCl₂). GFP-positive cells were analyzed for annexin V positivity using a FACSCanto flow cytometer and FlowJo software. In the inducible system, sorted GFP-positive cells were stained with both annexin V and 7-aminoactinomycin D (7-AAD; BD Biosciences) by costaining in annexin V binding buffer per the manufacturer's protocol 20 h and 44 h after the addition of doxycycline.

For analysis of miRNA expression by stem-loop quantitative RT-PCR (qRT-PCR), total RNA was prepared with TRIzol reagent according to the manufacturer's instructions and 10 ng of total RNA per reverse transcription reaction was used. Primers for RT and probes for quantitative PCR (TaqMan) were ordered from Applied Biosystems by Life Technologies and used as directed. Samples were normalized to miR-16 and RNU48 (assay identification no. 391 and 1006, respectively; Applied Biosystems). The absolute quantity of miR-155 was calculated using a standard curve of known quantities of an miR-155 synthetic RNA oligonucleotide (IDT) with the sequence 5′-rUrUrArArUrGrCrUrArArUrCrGrUrGrArUrArGrGrGrG-3′ (“r” refers to ribonucleotides).

For relative quantitation of LMP1 and EBNA2 mRNA expression levels in LCLd1/GFP and LCLd1/s155 cells, the SYBR green method was used. RNA was isolated from freshly sorted GFP-positive LCLd1/GFP and LCLd1/s155 cells by using TRIzol reagent. Equal amounts of total RNA from each sample were digested with DNase I (NEB) and reverse transcribed using the random primers supplied with the High Capacity reverse transcriptase kit (Applied Biosystems). Primers for quantitative PCR were as previously reported (3). Threshold cycle (C_T) values were normalized to RNU48 mRNA levels and then expressed as values relative to the background levels of LMP1 and EBNA2 mRNA present in 293T cells, which are negative for EBV.

Cells transduced with lentiviral (pLCE) sponge-containing particles—GFP, s155, and sCXCR4—were grown in R10 medium (RPMI 1640 medium with 10% FBS and 50 μg/ml of gentamicin) at 37°C for 3 days before being sorted for the same mean fluorescence intensities (MFI) of GFP on a FACSDiva flow cytometer. Cells transduced with either pMSCV- or pTRIPZ-based vectors were selected with puromycin for 1 week before being sorted for similarly high levels of GFP intensity. For pTRIPZ-transduced cells, doxycycline was added 2 days prior to sorting. Due to the approximate 5% error rate in the sorting process, cells after sorting were expected to be a mixture of ∼95% GFP-positive cells and ∼5% GFP-negative cells. After sorting, cells were resuspended in media at 2 × 10⁵ cells/ml (except for IBL-1 cells, which were resuspended at 6 × 10⁴ cells/ml). At the specified time points postsorting, 100 μl of cells was combined with a known quantity of AccuCount blank particles (Spherotech), 5.1 μm, and analyzed on a FACSCanto instrument (Becton Dickinson). The number of cells was calculated by normalizing to a fixed number of beads added to each sample.

Previously, it has been reported that EBV infection of resting primary human B cells results in the induction of several cellular miRNAs, including miR-155. It has also been reported that EBV-infected B cells exhibiting the viral latency III gene expression program, including LCLs and DLBCLs, express high levels of miR-155, while B cells undergoing EBV latency I, including primary EBV-positive BL tumors, express very low levels of miR-155 (6, 15, 21, 23, 26, 32, 37). To confirm the upregulation of miR-155 in the context of type III latency, we infected primary B cells with EBV and then analyzed the level of expression of miR-155 over time by using qRT-PCR (Fig. 1A). As expected, resting primary B cells, as well as the EBV-negative B-cell line BJAB and the non-B-cell line 293T, expressed low to undetectable levels of miR-155. However, infection of resting B cells with EBV caused miR-155 expression to increase throughout the 13-day time course of this experiment; at the end of the time course, miR-155 expression had been induced by >20-fold (Fig. 1A). Analysis of the two newly established LCLs EF3D and SDLCL, DLBCL cell line IBL-1 (28), EBV latency III BL cell lines Mutu III, Namalwa, and Raji, and EBV latency I BL cell line Mutu I revealed high-level expression of miR-155 in EF3D and SDLCL, undetectable levels of miR-155 in Mutu I, intermediate miR-155 expression in IBL-1, and low-level miR-155 expression in all the other B-cell lines analyzed (Fig. 1B). These data are consistent with previous reports (6, 15, 21, 23, 26, 32, 37) demonstrating that miR-155 is expressed at readily detectable levels in B cells undergoing EBV type III latency but only at very low levels in B cells undergoing EBV latency I.

We next used small-RNA deep sequencing, performed as described previously (46, 48), to analyze the total pattern of miRNA expression in the LCL EF3D. This analysis revealed that miR-155 is, in fact, the single most highly expressed miRNA in EF3D, contributing 799,949 reads (∼21%) to the total annotatable miRNA pool of 3,784,571 reads (Table 1). This analysis also identified miR-21, miR-24, and miR-146a, previously reported to also be induced upon EBV infection of B cells (6), as highly expressed cellular miRNAs, comprising ∼17%, ∼2%, and ∼2% of the total miRNA pool, respectively. In addition, we also detected the expression of eight EBV-encoded miRNAs, of which miR-BHRF1-1 was the most prominent (Table 1). Our inability to detect the other 17 miRNAs encoded by EBV results from the fact that the B95-8 EBV isolate used to generate the EF3D LCL bears a deletion that removes these miRNAs from the viral genome (5). In the EF3D LCL, all of the viral miRNAs together comprised ∼10% of the total cellular miRNA pool, i.e., less than half the level contributed by miR-155 alone.

To address whether miR-155 plays a significant role in the maintenance of EBV-infected B cells, we suppressed miR-155 activity in two recently generated LCLs (EF3D and SDLCL), in the EBV-positive BL cell lines Raji and Namalwa, and in the EBV-negative BL cell line BJAB. miR-155 activity was inhibited by transduction of these cell lines with a GFP-based lentiviral miRNA sponge specific for miR-155 (s155). We also separately transduced a previously described (17) control lentiviral miRNA sponge vector (sCXCR4), which expresses a sponge specific for an artificial small interfering RNA that can inhibit CXCR4 mRNA expression. The second control vector represents the parental lentiviral vector expressing only GFP. miRNA sponges represent a now well-established technique (4, 12, 17, 49) for specifically and stably blocking the activity of an miRNA through overexpression of an mRNA containing the gfp indicator gene linked to multiple (in this case, nine) copies of an incompletely complementary artificial target site for the miRNA of interest. Sponge-expressing cells can be sorted using flow cytometry for GFP-positive cells (17). Following transduction with each sponge and control construct, we sorted the brightest ∼20% GFP-positive cells and then monitored their subsequent growth by FACS for GFP-positive cells.

The growth of the sorted s155-expressing SDLCL and EF3D LCLs was severely attenuated compared to growth of GFP-positive cells transduced with the vectors expressing only GFP and the control sponge sCXCR4 (Fig. 2A and B, respectively). In contrast, the miR-155 sponge had no detectable effect on the growth of the EBV-negative BL cell line BJAB and the EBV-positive BL cell lines Raji and Namalwa (Fig. 2D, E, and F, respectively), even though all three of these cell lines do express detectable levels of miR-155 (Fig. 1B).

We wished to extend this analysis to the DLBCL cell line IBL-1, which is naturally infected with wild-type EBV (28), but we found that the lentiviral pLCE vector was unable to effectively transduce this cell line (data not shown). To circumvent this problem, we recloned the gfp gene, with and without the flanking miR-155 sponge, into a retroviral vector based on MSCV. Transduction of IBL-1 cells with the MSCV/s155 vector resulted in potent inhibition of the growth of GFP-positive cells, while GFP-positive IBL-1 cells that had been transduced with the control GFP expression vector replicated normally (Fig. 2C). We therefore conclude that the DLBCL cell line IBL-1 is similar to LCLs in requiring miR-155 for cell growth.

The most direct way to demonstrate miRNA function is to challenge cells with a vector in which an indicator gene has been linked to artificial target sites for that miRNA inserted into the 3′ untranslated region (16, 54). We have previously described a lentiviral indicator vector specific for miR-155 (17), and we were anxious to confirm the ability of the miR-155 sponges used in the experiments corresponding to Fig. 2 to inhibit miR-155 function. Unfortunately, the strong inhibitory effect of the miR-155 sponge on the growth and viability of the SDLCL, EF3D, and IBL-1 cells (Fig. 2A, B, and C, respectively) made this technically very difficult in these cells (this problem is addressed below). However, we were able to address whether the miR-155 sponge was active in the BJAB, Raji, and Namalwa cell lines, which remained able to replicate after transduction with the miR-155 sponge (Fig. 2D, E, and F, respectively). As shown in Fig. 3, Namalwa cells, which naturally express both miR-155 and the EBV miRNA miR-BHRF1-1, showed significant inhibition of a lentiviral indicator construct carrying the Renilla luciferase (RLuc) gene linked to artificial target sites for either miR-155 or miR-BHRF1-1 compared to 293T cells, which express neither miR-155 nor miR-BHRF1-1. As previously described (16, 17), this miRNA indicator assay is normalized by transducing cells simultaneously with both an RLuc-based lentiviral vector, containing the artificial miRNA target sites, and a firefly luciferase (FLuc)-based lentiviral vector lacking these target sites. Importantly, Namalwa cells transduced with the lentiviral miR-155 sponge expression vector failed to repress expression of the RLuc gene linked to miR-155 target sites but fully retained the ability to repress indicator vectors containing miR-BHRF1-1 target sites (Fig. 3) or target sites for miR-21 or miR-BHRF1-3 (data not shown). These data therefore argue that the miR-155 sponge is indeed able to effectively inhibit miR-155 function after transduction into the B-cell lines analyzed in Fig. 2.

To further analyze the effect of loss of miR-155 function on LCL growth in culture, we next quantified the numbers of GFP-positive and GFP-negative cells over time in cultures of the LCL EF3D that had been transduced with either s155 or GFP. As shown in Fig. 4, the s155 and GFP cells showed similar patterns of GFP expression immediately after sorting. For both samples, a small number of GFP-negative cells was present in the pool of collected cells (<5%). Six days after sorting, the number of GFP-positive cells in the s155 sample had markedly decreased, while the control GFP sample was essentially unchanged (Fig. 4, central panel). Finally, by day 25, nearly all cells present in the s155-transduced culture were GFP negative, while the control GFP-transduced cells maintained their GFP-positive phenotype. These data lend further support to the concept that LCLs lacking miR-155 function are at a severe disadvantage relative to miR-155-expressing LCLs in terms of their growth in culture. Interestingly, preliminary experiments performed in parallel using an miR-21-specific sponge failed to show any significant inhibitory effect on LCL outgrowth (data not shown), even though miR-21 is expressed at high levels in LCLs (Table 1) and has previously been associated with tumorigenesis (1, 29, 34).

To further define the basis for the observed growth deficiency of LCLs lacking miR-155, we next analyzed the cell cycle distributions of s155- and GFP-expressing LCLs. As shown in Fig. 5A and B, we observed a substantial diminution of the ability of s155-transduced cells to enter S phase relative to the abilities of the GFP and sCXCR4 controls. Consistently, the steady-state percentage of G₁ cells among s155-transduced LCLs was increased relative to the percentage among controls. We further observed that the percentage of apoptotic cells, as assessed by annexin V staining, was markedly increased in miR155-deficient LCLs relative to the value for controls (Fig. 5C). A reduced ability to exit the G₁ phase of the cell cycle, as well as an increased propensity to undergo apoptosis, likely explains the poor growth potential of LCLs lacking miR-155 function (Fig. 2 and 4).

A difficulty that we consistently encountered using the pLCE/s155 vector is that this vector expresses the miR-155 sponge constitutively. As loss of miR-155 expression is clearly very deleterious for LCL growth and viability (Fig. 2, 4, and 5), it is difficult to collect sufficient numbers of cells that express the miR-155 sponge for subsequent analysis. To overcome this problem, we generated tetracycline-regulatable lentiviral control and miR-155 sponge expression vectors and used these to transduce a third newly generated LCL called LCLd1. This allowed us to expand the cells after lentiviral transduction and to then induce expression of the miR-155 sponge—and the linked gfp gene—by the addition of doxycycline to the culture media prior to sorting by FACS.

As shown in Fig. 6A, this second approach again resulted in a population of GFP-positive, s155-positive cells that had lost the ability to replicate in culture, while the control GFP-positive LCLs grew normally. Analysis of miR-155 function, using an FLuc-based miR-155 indicator construct analogous to that used in the experiments corresponding to Fig. 3, revealed that the s155 sponge is able to block most, but apparently not all, miR-155 function in the transduced LCLd1 cells after doxycycline addition (Fig. 6B). The partial inhibition of miR-155 function seen in Fig. 6B may reflect the very high level of miR-155 expression in LCLs, compared to the Namalwa cells analyzed in Fig. 3 (Fig. 1B). Nevertheless, it is clear that any residual level of miR-155 function is not sufficient to support LCL growth in culture (Fig. 6A).

Because of published results suggesting that loss of miR-155 function can induce a drop in the EBV genome copy number in both LCLs and Raji cells (27), we next analyzed the expression levels of two key EBV genes in transduced LCLs by performing qRT-PCR for the viral mRNAs encoding LMP1 and EBNA2. As shown in Fig. 6C, we did not observe a significant drop in LMP1 or EBNA2 mRNA expression in the presence of the miR-155 sponge, thus demonstrating that the loss of cell growth seen upon inhibition of miR-155 function in the LCLd1 cells (Fig. 6A) is not due to a loss of LMP1 or EBNA2 expression. Finally, we also repeated our analysis of the effect of loss of miR-155 function in LCLs on apoptosis induction (Fig. 5C) by staining the transduced LCLd1 cells with annexin V, which binds to early apoptotic cells, as well as with 7-AAD, which stains cells that are in the late stages of apoptotic death. This experiment revealed levels of both annexin V and 7-AAD staining after culture of the TRIPZ/s155-transduced cells in doxycycline for either 20 h or 44 h that were substantially higher than the levels for control TRIPZ/GFP-transduced cells (Fig. 6D). These data further confirm our observation (Fig. 5C) that loss of miR-155 function induces apoptotic cell death in EBV-induced LCLs.