Stress-Inducible Alternative Translation Initiation of Human Cytomegalovirus Latency Protein pUL138^▿

Cells.

Human embryonic lung fibroblasts (MRC5) and human embryonic kidney 293 (HEK-293) (ATCC, Manassas, VA) cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 10 mM HEPES, 1 mM sodium pyruvate, 2 mM l-glutamine, 0.1 mM nonessential amino acids, 100 U/ml penicillin, and 100 μg/ml streptomycin. For experiments requiring serum starvation, MRC5 cells were cultured in normal growth medium without serum for 24 h prior to nucleofection and through the duration of the experiment. Mononuclear cells were isolated from human cord blood or bone marrow from donors at University Medical Center at the University of Arizona by using Ficoll/Paque Plus (Amersham Pharmacia Biosciences) density gradient centrifugation with a protocol approved by the Institutional Review Board. Magnetic cell separation (Miltenyi Biotec) was used to enrich for CD34⁺ hematopoietic progenitor cells. CD34⁺ cells were maintained in Iscove's modified Dulbecco's medium (IMDM) supplemented with 10% BIT serum substitute (StemCell Technologies), 2 mM l-glutamine, 20 ng/ml human low-density lipoprotein (Sigma), and 50 μM 2-mercaptoethanol.

Viruses.

A fusion-inducing factor X (FIX) virus strain bacterial artificial chromosome (BAC) that expresses the green fluorescent protein (GFP) was used as the parental strain (FIXwt) for all recombinations described. Viruses were propagated by electroporating infectious BAC DNA into MRC5 cells and purified by density gradient centrifugation as described previously (52). Virus titers were determined by 50% tissue culture infective dose (TCID₅₀) assays on MRC5 cells.

Recombinant viruses (FIX-UL133_myc, FIX-UL135_myc, FIX-UL136_myc, FIX-LUC-ΔUL138) were constructed in Escherichia coli by linear recombination in a two-step positive/negative selection method that leaves no trace of the engineering process (52, 73, 76). Briefly, in the first step, PCR products containing galK flanked by homologous viral sequences target the galK insertion into the viral BAC by homologous recombination. GalK⁺ recombinants were selected on minimal medium plates containing galactose as a sole carbon source. In the second step, the galK insertion was replaced with PCR products containing the myc epitope tag flanked by viral sequences homologous to the desired insertion site. Similarly, the FIX-LUC-ΔUL138 recombinant virus was generated by replacing galK with the PCR product encoding firefly luciferase (FL) amplified from the pGL3-Basic plasmid (Promega). Recombinants were selected on M63 minimal plates containing 0.2% 2-deoxygalactose and 15 μg/ml chloramphenicol and further screened by BAC digestion, PCR, and sequencing. The generation of the FIX-UL138_myc recombinant virus was described previously (52). All primers used for virus recombination are listed in Table 1.

Plasmids.

All primers used for plasmid cloning are listed in Table 2. The bicistronic, dual luciferase reporter construct (pCDNA3-rLuc-polIRES-fLuc) containing the poliovirus IRES (pv-IRES) was kindly provided by N. Sonenberg (McGill University, Canada) (55). The poliovirus IRES was excised using KpnI and BamHI restriction sites, and portions of UL138 cDNA upstream of UL138 or the Kaposi sarcoma-associated herpesvirus (KSHV) ORF72 IRES (nucleotide position 123206 to 122973) (23) were cloned into the plasmid using primers flanked with KpnI and BamHI restriction sites and the FIXwt BAC DNA or KSHV genomic DNA as the template. The KSHV genome DNA was derived from the KSHV-positive primary effusion lymphoma (PEL) cell line BCBL-1, a kind gift from P. Schaffer (University of Arizona). Sequences 5′ of UL138 or the KSHV-IRES were cloned into the promoterless pGL3-Basic plasmid (Promega) using primers with KpnI and HindIII restriction sites. Generation of pGL3-major immediate-early promoter (pGL3-MIEP₁₄₀₀) was described previously (52). cDNAs (pFL) encoding each of the UL138 transcripts in which FL had been substituted for UL138 were cloned into pEF1/myc-His-B (Invitrogen) using XbaI and PmeI restriction sites and FIX-LUC-ΔUL138 as the template.

The pCIG lentivirus expression vector (32) obtained from L. Lybarger (University of Arizona) was further modified to (i) provide a more versatile multiple-cloning site (MCS)-IRES-GFP cassette resulting in pCIG2-IRES-EGFP (where EGFP is enhanced GFP) and (ii) to remove all posttranscriptional regulatory elements, including the IRES GFP and woodchuck posttranscriptional regulatory element (WPRE), resulting in pCIG2-IGW. Briefly, pCIG2-IRES-EGFP was generated with an expanded MCS by annealing and elongating two primers with partial homology (bold) with EcoRI and BamHI restriction enzyme sites (underlined) at each end (forward, 5′-GGGG GAATTC GG TGA TCA GGG TAT ACG GCT CGA GGG GAT ATC GGG TTT AAA CGG GGC GC-3′; reverse, 5′-GGGG GGATCC CCT TAA TTA ACC TTC GAA CCG CGA TCG CCC GGC GCG CCC CGT TTA AAC C-3′). Next, the encephalomyocarditis virus IRES (EMCV-IRES) was PCR amplified from pIRESpuro (Clontech) using primers containing BamHI and BglII enzyme sites (underlined) (forward, 5′-GGGG GGATCC GCC CCT CTC CCT CCC CCC CCC C-3′; reverse, 5′-GGGG AGATCT GGT TGT GGC AAG CTT ATC ATC G-3′). EGFP was amplified from pCMS-EGFP (Clontech) using primers containing BglII and SalI enzyme sites (underlined) while maintaining the contextual Kozak sequence (bold) immediately preceding the ATG (forward, 5′-GGGG AGATCT CGC CAC CAT GGT GAG CAA GGG CGA GGA GC-3′; reverse, 5′-GGGG GTCGAC CTG CCC CAG CTG GTT CTT TCC GCC-3′). The MCS, ECMV-IRES and EGFP PCR amplicons were digested, ligated, and inserted into the EcoRI and SalI sites of pCIG lacking the parental IRES and GFP to form pCIG2-IRES-EGFP. The modified MCS contains the following enzyme sites in 5′-to-3′ order: XbaI, NheI, EcoRI, BclI, BstZ17I, XhoI, EcoRV, PmeI, AscI, AsiSI, BstBI, PacI, and BamHI. The pCIG2-IRES-EGFP vector was used to clone individual myc-tagged ORFs amplified from pEF1/myc-His-B expression plasmids containing each myc-tagged ORF using the EcoRV restriction site. This vector contains an IRES sequence downstream of the multiple cloning site that allows for the coexpression of enhanced GFP as a marker for infection. For vectors containing UL138 cDNAs, the pCIG2-IRES-EGFP expression vector was further modified to pCIG-IGW. Briefly, the IRES, EGFP, and WPRE posttranscriptional regulatory element were excised using PacI and KpnI followed by religation. UL138 cDNAs were cloned into pCIG-IGW using XbaI and PmeI or EcoRV and AsiSI restriction sites with FIXwt BAC as the template.

Transient transfection and luciferase reporter assays.

For transient expression in HEK-293 cells, molar equivalents (1 × 10¹⁰ to 7 × 10¹⁰ copies) of the indicated constructs were transfected into confluent HEK-293 cells using Lipofectamine 2000 (Invitrogen) and Opti-MEM I (Invitrogen) according to the manufacturer's recommendations. At 4 h posttransfection, the medium was replaced with normal growth medium, and protein expression was assayed at 48 h posttransfection. Luciferase activity was measured using either the Dual-Luciferase reporter assay system (Promega) or the luciferase assay system (Promega) as described previously (52). Nucleofection of subconfluent MRC5 cells was performed with molar equivalents (3 × 10¹¹ copies) of constructs using the basic Nucleofector kit for primary mammalian fibroblasts (Lonza) program U-023 according to the manufacturer's recommendations. Protein expression was analyzed at 6 h postnucleofection. For RNA experiments, molar equivalents (3 × 10¹¹ copies) of in vitro-transcribed RNA (iv-RNA) were transfected into HEK-293 cells using Lipofectamine 2000 (Invitrogen) and Opti-MEM I (Invitrogen), and FL activity was assayed 6 h posttransfection.

Immunoblotting.

Protein lysates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to a 0.45-μm polyvinylidene difluoride (Immobilon-FL [Millipore]) membrane, and imaged as described previously (52). Primary antibody conditions for rabbit anti-myc, mouse anti-myc, mouse anti-α-tubulin, and mouse anti-HCMV IE1 and IE2 proteins (generous gift from T. Shenk; clone 3H4) were described previously (52). Custom rabbit polyclonal, primary antibodies (Open Biosystems) were generated for pUL133, pUL135, pUL136, and UL138. Each antibody was affinity purified using the immunizing peptide and used at the following concentrations: anti-UL133 (2 μg/ml), anti-UL135 (2 μg/ml), and anti-UL136 (5 μg/ml).

Northern blotting.

Total RNA was isolated from HEK-293 cells transfected with the bicistronic luciferase constructs and prepared as described previously (52). A pEF1/myc-His-B (Invitrogen) plasmid containing the reverse complement of firefly luciferase was linearized with PmeI and used to generate an [α-³²P]CTP-radiolabeled antisense riboprobe using T7 in the SP6/T7 Riboprobe combination system (Promega) according to the manufacturer's instructions. Probe hybridization and visualization were in accordance with previous methods (52).

RACE.

RNA expressed during HCMV (FIXwt) infection was analyzed by RNA ligase-mediated rapid amplification of cDNA ends (RLM-RACE) using the FirstChoice RLM-RACE kit (Ambion). Briefly, primary human CD34⁺ cells or MRC5 fibroblasts were infected at a multiplicity of infection (MOI) of 5. At 18 h postinfection (hpi), total RNA was isolated and DNase was treated using the RNA II kit (Macherey-Nagel), processed as directed by the FirstChoice kit recommendations, and reverse transcribed with 90 U of Moloney murine leukemia virus reverse transcriptase (MMLV-RT) (Ambion) and 20 U of SuperScript III (Invitrogen) and 10% dimethyl sulfoxide (DMSO). The reverse transcription proceeded for 45 min at 45°C and then for 20 min at 55°C. The RNA template was degraded by the addition of 10 U RNase H for 30 min at 37°C. As a negative control (NC), RNA was processed as above; however, the tobacco acid pyrophosphatase enzyme (TAP-null) was omitted during the RNA preparation and ligation. The cDNA generated from capped mRNA species was amplified using the 5′ FirstChoice outer primer and gene-specific primers UL138 gsp2 (5′-TAGCTGCACTGGGAAGACACTT-3′), UL135 gsp (5′-ACCGCCTTTTCCAAGAGTT-3′), and UL133 gsp (5′-ATGCTGCTGCTGCTGTTGTTGTA-3′) using Phusion polymerase (New England Biolabs), 500 nM primers, 200 nM deoxynucleoside triphosphates (DNTPs), 3% DMSO, and 1× Phusion GC buffer. Reactions were denatured at 98°C for 1 min and then at 35 cycles of 98°C for 12 s, 65°C for 30 s, and 72°C for 40 s, followed by a final extension time of 5 min at 72°C. The primary PCR amplicon was then subjected to 35 cycles of nested PCR using the 5′ FirstChoice inner primer, which contains a BamHI restriction site, and primers UL138 REV-411nt (5′-GTATCTAGAGTCGTGCCAATGGTAAGCTAGA-3′), UL135 REV-62nt (5′-GTATCTAGAGCAGGCACCACGCTTATCAGAA-3′), and UL133 REV-187nt (5′-GTATCTAGACACGAAGGTTTGTTCTTCAGTTTA-3′) (with the XbaI insertion site indicated by underlining) using the same thermocycling parameters that were used during the primary PCR. Resulting products were digested with BamHI-HF and XbaI (New England Biolabs), gel purified, cloned, and sequenced. For the 3′ RACE, total RNA was reverse transcribed as described above and primed by a poly(A) adapter primer supplied with the FirstChoice RLM-RACE kit (Ambion). cDNA was then amplified with the 3′ outer primer and a primer for UL138 FWD (5′-ATGGACGATCTGCCGCTGAACGT-3′) using Phusion polymerase with GC buffer, 500 nM primers, 200 nM DNTPs, and 3% DMSO. Cycling parameters were 98°C for 30 s and then 35 cycles of 98°C for 10 s, 65°C for 30 s, and 72°C for 30 s, followed by a final extension at 72°C for 5 min. The primary PCR amplicon was then subjected to 35 cycles of nested PCR using the 3′ FirstChoice inner primer and UL138 gsp1 (5′-TCTAGCTTACCATTGGCACGAC-3′), with the same cycling parameters. The resulting product was digested with BamHI-HF (New England Biolabs), gel purified, cloned, and sequenced.

In vitro transcription and translation.

pEF1/myc-His-B expression plasmids containing FIX-LUC-Δ138 cDNAs were linearized using PmeI, phenol-chloroform extracted, and transcribed using the mMessage mMachine T7 kit (Ambion) and subsequently poly(A) tailed using the poly(A) tailing kit (Ambion) according to the manufacturer's instructions. Retic lysate IVT (Ambion) was used for translation of in vitro-transcribed RNA using a 1:1 mix of the 20× translation mix lacking Met and Leu (potassium acetate concentration, 80 mM) according to the manufacturer's instructions.

RNA structure prediction.

RNA secondary structure prediction was generated using the RNA Mfold server, version 3.2 (77) (http://mfold.bioinfo.rpi.edu/cgi-bin/rna-form1.cgi). Default parameters were used, except for the following: structure draw mode was set at “untangle with loop fix,” exterior loop type was set to “flat,” and structure annotation was set to “p-num.” Free energy of stable stem loops was calculated by submitting the sequence for each stem loop individually to the Mfold server.

qRT-PCR.

RNA was isolated and DNase treated using the NucleoSpin RNA II kit (Machery-Nagel) according to the manufacturer's instructions. cDNA was generated from 1 μg input RNA or in vitro-transcribed and poly(A)-tailed RNA using an anchored oligo[dT]₁₈ primer and the transcriptor first-strand cDNA synthesis kit (Roche) according to the manufacturer's instructions. Quantitative reverse transcription PCR (qRT-PCR) was performed with the LightCycler 480 Probes Master (Roche) according to the manufacturer's instructions along with Universal Probe Library (Roche) probes and primers specific for UL133 (forward, 5′-TTTCTGGACCTGGCATCG-3′; reverse, 5′-TACGCGCAGAGGAAAATCAT-3′; probe number 114, 5′-CTGTGGTC-3′), UL135 (forward, 5′-GCGGTGTACGTCGCTCTAC-3′; reverse, 5′-GGAAACTCGGGTTTATCTATCG-3′; probe number 55, 5′-GGCAGACCATCCTCTCCTATG-3′), UL136 (forward, 5′-GCGGCTGTCATTATCCTGAG-3′; reverse, 5′-TAGTACATGGCCCCGTTCA-3′; probe number 152, 5′-TCGCCGTC-3′), UL138 (forward, 5′-GGTTCATCGTCTTCGTCGTC-3′; reverse, 5′-CACGGGTTTCAACAGATCG-3′; probe number 126, 5′-TCTCTGGT-3′), and firefly luciferase (FL) (forward, 5′-TGAGTACTTCGAAATGTCCGTTC-3′; reverse, 5′-GTATTCAGCCCATATCGTTTCAT-3′; probe number 29, 5′-GAAGACGG-3′) and the prevalidated UPL human beta-actin gene assay (Roche). Linear fragments for the generation of UL133, UL135, UL136, UL138, and firefly luciferase standard curves were synthesized using the following primer sets, using the FIX BAC or pGL3-Basic plasmid (Promega) as a template: UL133 (forward, 5′-ATGGGTTGCGACGTGCACG-3′; reverse, 5′-TTACGTTCCGGTCTGATGCTGC-3′), UL135 (forward, 5′-ATGTCCGTACACCGGCC-3′; reverse, 5′-TCAGGTCATCTGCATTGACTCG-3′), UL136 (forward, 5′-AGAGAGAAAGAGAGTATGTCAGTC-3′; reverse, 5′-TTACGTAGCGGGAGATACGGC-3′), UL138 (forward, 5′-ATGGACGATCGTCCGCTGAA-3′; reverse, 5′-TCACGTGTATTCTTGATGATAA-3′), firefly luciferase (forward, 5′-ATGGAAGACGCCAAAAACATAAAGAAAG-3′; reverse, 5′-TTGGCCTTTATGAGGATCTCTCTGATTT-3′). qRT-PCR conditions were as follows: 95°C for 10 min and then 60 cycles of 95°C for 10 s and 60°C for 30 s. Reactions were run on a LightCycler 480 (Roche), and data analysis was performed using the LightCycler 480 software (version 1.5) using the second derivative max method.

A fold change of 5′ and 3′ probe amplification for each transcript was used to quantify transcript ends compared to cDNA ends derived from in vitro-transcribed RNA (iv-RNA) using the following equation (53):

where Eff is efficiency, Ref is reference sample (iv-RNA), Exp is the experimental sample, and Ct is the crossing point. The efficiency of each probe set was determined by standard curve amplification: UL133 (1.646), UL135 (1.979), UL136 (1.921), UL138 (1.953), and FL (1.897). In our analysis, we considered a 2-fold change to be within the confidence interval for full-length transcripts.

pUL138 is encoded on multiple transcripts in fibroblasts and hematopoietic progenitor cells.

We have previously identified two transcripts of 3.6 and 2.7 kb encoding pUL138 by using Northern blotting and rapid amplification of cDNA ends (RACE) (52). Further studies using alternative RACE methods to amplify the cDNA ends generated from polyadenylated RNA isolated from FIX-infected fibroblasts and CD34⁺ hematopoietic progenitor cells (HPCs) identified a third transcript corresponding to 1.4 kb (Fig. 1). The 1.4-kb transcript 5′ start maps to nucleotide position 188206, 292 nucleotides downstream of the predicted start of UL136. All three transcripts share a 3′ end at nucleotide position 186833, which corresponds to a polyadenylation site just downstream of the putative UL138b ORF (previously ORF11). It is not clear why the 1.4-kb transcript was not identified in our original studies, as all RACE methods should favor the amplification of the smaller transcript. One possibility is that the 1.4-kb transcript is present in lower abundance during infection or that the RNA ligase-mediated rapid amplification of cDNA ends (RLM-RACE) technology used in this study provides enhanced sensitivity. We also failed to detect this transcript by Northern blotting. The presence of this band on Northern blots may have been obscured, as it runs at the same molecular mass as 18S rRNA. Alignment of the UL138 cDNAs with the HCMV genome suggests that splicing events are not involved in generation of the UL138 transcripts (data not shown). All three 5′ RACE products were detected in fibroblasts and CD34⁺ HPCs, indicating that these transcripts are expressed in multiple contexts of infection.

Putative ORFs upstream of UL138 encode proteins during HCMV infection.

Our transcript-mapping studies demonstrate extensive sequence (∼0.5 to 2.5 kb) upstream of UL138 with the potential to encode multiple ORFs. These putative ORFs were originally annotated as start-to-stop codons encoding at least 80 amino acids (7, 13, 47). The presence of putative ORFs upstream of UL138 raises the possibility that these transcripts encode more than one protein or that they contain a substantial 5′ untranslated region (UTR). To examine the coding potential of each putative ORF, we cloned each individual ORF into the pCIG2-IRES-EGFP expression vector with an in-frame C-terminal myc epitope tag. HEK-293 cells were transiently transfected with each construct, and expression of the UL133, UL134, UL135, UL136, UL137, and UL138 ORFs was analyzed by immunoblotting using a mouse monoclonal antibody to the myc epitope. Only UL133, UL135, and UL136 exhibited evidence of protein coding capacity (Fig. 2A). Excluding the myc epitope tag, UL133, UL135, UL136, and UL138 coded for proteins of approximately 28, 36, 27, and 21 kDa, respectively. The molecular masses of pUL133, pUL135, and pUL136 myc-tagged proteins are approximately 10 kDa greater than their predicted molecular mass based on sequence annotation, suggesting that these proteins are posttranslationally modified. While we detected multiple bands for some proteins, the significance of these bands is not known. In the case of pUL138, the additional lower-molecular-mass protein bands appear to be artifacts of transient overexpression of the myc-tagged construct and are not observed in infection with the wild-type virus (52) or a recombinant virus expressing a myc-tagged version of pUL138 (Fig. 2B). Protein products from ORFs UL134 and UL137 were not detected in this assay, suggesting that these ORFs may not encode stable proteins in a transient expression assay. It should be noted that both UL134 and UL137 are transcribed from the opposite DNA strand relative to the other putative ORFs analyzed.

To determine if UL133, UL135, and UL136 encode proteins during HCMV infection, we constructed recombinant viruses with in-frame myc epitope tags inserted at the 5′ end of UL133, UL135, or UL136 of the FIX strain BAC. Infection of MRC5 fibroblasts with the recombinant viruses, termed FIX-UL133_myc, FIX-UL135_myc, and FIX-UL136_myc, resulted in protein production from each ORF, as detected by immunoblotting at 48 hpi with a myc monoclonal antibody (Fig. 2B). Each virus produced a single protein product, except for FIX-UL136_myc, which produced two major protein species during infection. The larger-molecular-mass pUL136 band corresponds to the protein species produced following transfection of the pUL136 expression plasmid encoded by the full-length UL136 ORF (Fig. 2A). The smaller-molecular-mass species may represent a truncated form of pUL136. HCMV IE proteins, IE1 and IE2, show similar levels of infection between the recombinant viruses such that the observed levels of pUL133, pUL135, pUL136, and pUL138 reflect their relative levels of expression during infection. These results are the first to demonstrate the coding potential of the ULb′ ORFs UL133, UL135, and UL136.

UL138 is expressed from polycistronic mRNAs.

The existence of bona fide ORFs upstream of UL138 suggests that the UL138 transcripts are polycistronic. To explore this possibility, cDNAs generated for each transcript were cloned into pCIG-IGW, and molar equivalents of each were transfected in HEK-293 cells. Protein lysates were harvested 48 h posttransfection and analyzed by immunoblotting with rabbit polyclonal antisera generated against peptides specific to pUL133, pUL135, pUL136, or pUL138 coding sequences (Fig. 3A). Lysates from cells transfected with an expression construct encoding the myc-tagged version of each individual ORF (as described for the experiments shown in Fig. 2) were used as positive controls.

The 3.6-kb cDNA expressed a 36-kDa protein corresponding to pUL133 in addition to low-intensity bands corresponding to pUL138. The 2.7-kb cDNA resulted in reactive bands corresponding to pUL135 and pUL138. The expression of pUL135 from the 2.7-kb cDNA was surprising, as the 5′ end of the transcript does not include the predicted start codon for pUL135 (Fig. 1). The expression of this protein from the 2.7-kb cDNA suggests that the translation start for pUL135 may correspond with an in-frame AUG 28 nt downstream of the 5′ end of the 2.7-kb transcript (Fig. 1). Several bands are observed for UL135 expressed during infection (Fig. 2B) or transfection (Fig. 3), indicative of possible alternative starts or posttranslational modifications. The 1.4-kb cDNA expressed two proteins, one corresponding to pUL138 and the other corresponding to a smaller-molecular-mass pUL136. This result was unexpected, since the 5′ start of the 1.4-kb transcript maps downstream of the predicted AUG of UL136 (Fig. 1). This band most likely corresponds to the smaller-molecular-mass species of pUL136 observed during virus infection (Fig. 2B). Full-length pUL136 was not expressed from either the 3.6- or the 2.7-kb cDNAs during transfection, although it was detected during infection (Fig. 2B). These results suggest that expression of full-length pUL136 from the polycistronic transcripts may be specific to the context of viral infection. Importantly, pUL138 was expressed from each of the three cDNAs, with the greatest expression resulting from the transfection of the 1.4-kb cDNA (Fig. 3B).

Our results suggest that UL138 transcripts are polycistronic. To rule out the possibility that internal promoter activity or splicing could result in monocistronic transcripts encoding pUL138, we analyzed the 5′ and 3′ ends of the RNAs by qRT-PCR following transfection of the cDNA constructs. Equivalent amplifications of 5′ and 3′ ends argue against the presence of smaller distinct RNAs that may arise from internal promoter activity or splicing events, since it is unlikely that distinct RNAs would be present at equivalent levels. These results were evaluated against comparable probe amplification of cDNA generated from full-length, in vitro-transcribed RNA (iv-RNA), thus eliminating the possibility of mammalian promoter or splicing activity. Our results indicate equivalent amplifications (within a 2-fold confidence level) of the 5′ and 3′ transcript ends derived from the transfected cells relative to iv-RNAs (Fig. 3C). We interpret this result to indicate that the transcripts supporting pUL138 expression are full length and that monocistronic UL138 transcripts cannot be responsible for pUL138 protein production, as shown in Fig. 3A. In addition, transcript numbers from each cDNA transfection were comparable, indicating that levels of pUL138 expression cannot be explained by differences in transfection efficiency.

A sequence overlapping the 5′ end of the UL136 ORF promotes translation of a downstream cistron.

The expression of multiple proteins from a single transcript necessitates alternative mechanisms of translation initiation. We hypothesized that pUL138 translation is mediated by upstream sequences that facilitate translation initiation. Given the extensive sequence (∼0.5 to 2.5 kb) 5′ of UL138, we favored the possibility that an internal ribosome entry site (IRES) was responsible for pUL138 expression. To explore this possibility, we cloned sequences 5′ of UL138 into a bicistronic reporter construct in which the 5′ cistron encodes Renilla luciferase (RL) and the 3′ cistron encodes firefly luciferase (FL) (Fig. 4A). The cloned sequences were designated IRES candidate sequences (ICS). A construct containing the poliovirus IRES (PV-IRES) or the Kaposi's sarcoma-associated herpesvirus IRES (KSHV-IRES) served as a positive control. A construct lacking an ICS or IRES was used as a negative control (empty). The RL and FL activities of each construct were measured 48 h posttransfection of HEK-293 cells, and IRES activity was expressed as a ratio of FL to RL. Negligible FL expression was measured in the empty control, whereas the PV- or KSHV-IRES resulted in substantial expression of the downstream FL cistron. Of the ICS analyzed, only ICS-7 FL expression was comparable to that of the PV-IRES, and to a greater extent than that of the KSHV-IRES (Fig. 4B). ICS-7 is a 663-nt sequence that starts 65 nt upstream of the start for UL136. ICS-5, -6, and -8 overlapped with ICS-7; however, these sequences supported moderate to low expression of FL. These results suggest that a sequence element upstream of UL138 has the potential to mediate translation of pUL138.

To further define the sequences required for maximal IRES activity, we cloned fragments extending from the 5′ and 3′ ends of ICS-7 (Fig. 4C). Interestingly, none of these sequences showed FL activity beyond that seen with ICS-7. ICS-10 and ICS-12 exhibited 52.7% and 47.2% of the activity of ICS-7, respectively. However, ICS-9, ICS-11, ICS-13, and ICS-14 exhibited less than 10% of the activity of ICS-7. These data indicate that an additional sequence around ICS-7 is not required for activity and that the context of the sequence is likely critical for proper folding of the RNA sequence into a functional element.

Next, we sought to determine the minimal ICS-7 sequence required for IRES function. Shorter fragments of ICS-7 as well as the reverse complement of ICS-7 (ICS-7RC) were cloned into the bicistronic dual luciferase reporter construct and analyzed for FL expression (Fig. 4D). None of the sequences supported comparable FL expression to ICS-7. Importantly, the reverse complement of ICS-7 (ICS-7RC) exhibited diminished activity relative to ICS-7. The ICS-7RC exhibited only 26% of the activity of ICS-7, whereas the reverse complement of the KSHV-IRES (KSHV-RC) exhibited 52% of the activity of the KSHV-IRES (Fig. 4D). Our results indicate that ICS-7 is the optimal sequence element required for expression of the downstream cistron. While our results do not exclude other mechanisms of alternative translation initiation, they are consistent with this element functioning as an IRES.

Transcripts originating from the dual luciferase vector are full length, and ICS-7 does not have inherent promoter activity.

Possible alternatives to our interpretation that downstream cistron expression results from internal ribosome entry include the presence of monocistronic UL138 transcripts resulting from (i) alternative splicing or RNA fragmentation or (ii) cryptic promoter activity. To rule out these possibilities, we examined the length of transcripts generated in HEK-293 cells transfected with the bicistronic luciferase vector by Northern blotting. A radiolabeled anti-sense RNA probe specific to FL was hybridized to total RNA isolated from HEK-293 cells transfected with the empty bicistronic luciferase construct or the constructs containing PV-IRES, ICS-7, or ICS-7RC. RNA from untransfected cells was used as a negative control (NC). We detected a predominant reactive transcript corresponding to the predicted size of each full-length bicistronic luciferase transcript (Fig. 5A). Smaller transcripts unique to the ICS-7 construct that could independently result in FL activity were not detected. These results were confirmed by using qRT-PCR to demonstrate equivalent levels of RL and FL transcripts for each of the bicistronic constructs compared to cDNA derived from in vitro-transcribed RNA (data not shown), suggesting that monocistronic transcripts are not responsible for expression of the FL cistron.

Previously, in vitro transcription and translation assays have been used to rule out the possibility of promoter activity in testing the existence of an IRES. We analyzed the ability of full-length bicistronic luciferase iv-RNA containing ICS-7 as a template for expression of the downstream FL cistron. The empty vector or two ICS sequences that were either determined to be negative (ICS-9) or positive (ICS-10) for IRES activity during our initial IRES screen were used as controls. Transcripts were translated in vitro using rabbit reticulocyte lysates (RRL). Similar to the IRES activity measured by DNA transfection (Fig. 4), ICS-7 and ICS-10 supported expression of the downstream cistron, whereas only modest levels of FL were supported by ICS-9 (Fig. 5B). These data suggest that cryptic promoter activity or splicing cannot explain FL expression, as shown in Fig. 4.

We also tested ICS-7 directly for promoter activity by cloning it upstream of FL in a promoterless plasmid relative to the HCMV major immediate-early promoter (MIEP₁₄₀₀). The empty plasmid and a plasmid containing the KSHV-IRES were used as negative controls. In addition to ICS-7, we included ICS-4 and ICS-7RC in our analysis, since these sequences did not stimulate expression of the downstream cistron, as shown in Fig. 4. FL activity was measured 48 h after transfection of HEK-293 cells. Relative to the MIEP₁₄₀₀, ICS-7 did not promote significant FL activity (Fig. 5C). Instead, ICS-7 FL activity was similar to those of the KSHV-IRES and HCMV sequences that did not exhibit IRES activity in bicistronic assays. The FL activity from each of these constructs was 0.02 to 0.05% of MIEP₁₄₀₀ activity. It is therefore unlikely that ICS-7 facilitates gene expression through promoter function.

ICS-7 Mfold prediction suggests extensive RNA secondary structure.

IRES sequences have extensive RNA secondary structures that allow for the recruitment of the translation machinery. We analyzed the RNA secondary structure of ICS-7 using Mfold, a computer-based RNA modeling program (Fig. 6) (77). We used the p-num function to identify highly favorable structures. The shaded stem loops represent predicted highly stable structures (−17 Kcal/mol) with p-num scores that correspond to 95.8 to 99.8% probability that these bases will pair only with each other. The likelihood of these structures forming was considerably higher than that of other structures within the same molecule, independent of whether only ICS-7 or the entire 3.6- or 2.7-kb transcripts were folded (data not shown). With the exception of a single base pair substitution at the 5′ end of ICS within structure ii (Fig. 6), the ICS-7 sequence is highly conserved among all HCMV strains for which a sequence is known (data not shown). These results support the idea that the hairpins are structurally conserved in the mRNA molecule and may be involved in IRES function (78). Our primary sequence analysis also revealed a region containing a 15-base polypyrimidine tract (bold line) with a polypyrimidine tract binding protein (PTB) consensus sequence (bold) (43, 50). The polypyrimidine tract is conserved with 100% identity in all HCMV strain sequences available through NCBI. Polypyrimidine tracts have been found in many IRES elements and may facilitate PTB binding to stabilize the mRNA and then recruitment of ribosomes (36). This region is relatively unstable (−7.6 Kcal/mol) compared to the conserved hairpins. The presence of highly stable, conserved hairpins, combined with a polypyrimidine tract, supports our hypothesis that this region functions as an IRES.

ICS-7 is stress-inducible.

Alternative mechanisms of translation initiation are often active during cellular stress when canonical translation is repressed. This allows for a select subset of mRNAs to be expressed during stress. Therefore, we investigated the ability of ICS-7 to maintain translation in the presence of stress in a cell type that is permissive for HCMV infection. Bicistronic constructs containing either no IRES (empty), the PV-IRES, or ICS-7 were nucleofected into MRC5 fibroblasts that were fed under normal growth conditions or serum stressed for 18 h prior to nucleofection, and luciferase levels were quantitated 6 h later. ICS-7 IRES activity was increased approximately 2-fold under serum-stressed conditions relative to fed conditions, whereas the PV-IRES was unaffected by serum stress (Fig. 7). These results suggest that ICS-7 may act as an inducible IRES in a context-specific manner.

UL138 cistron expression from the 3.6- and 2.7-kb transcripts, but not the 1.4-kb transcript, is stress inducible.

Next, we analyzed the impact of serum stress on UL138 cistron expression from the 3.6-, 2.7-, or 1.4-kb cDNAs in which we substituted UL138 with FL as a reporter gene (Fig. 8A). MRC5 fibroblasts were cultured under fed or serum-stressed conditions for 18 h and then nucleofected with each pFL-cDNA construct. Lysates were analyzed for FL expression 6 h later, and FL expression under starved conditions was normalized to FL expression under fed conditions. Consistent with the ability of ICS-7 to induce expression of a downstream cistron (Fig. 7), we demonstrated enhanced FL expression from the 3.6-kb (pFL-3.6) and 2.7-kb (pFL-2.7) cDNAs. FL expression was induced 2.14- and 9.43-fold from the 3.6- and 2.7-kb cDNAs, respectively, whereas FL expression from the pFL-1.4 cDNA decreased ∼2-fold in serum-stressed cells (Fig. 8B). Similar to FL expression from the pFL 1.4-kb cDNA, FL expression was decreased (∼2-fold) from a monocistronic construct encoding only FL under serum-starved conditions, consistent with the inhibition of cap-dependent translation. From these data, we conclude that FL is expressed by a canonical mechanism from the pFL-1.4 transcript, whereas FL is expressed by alternative IRES-like mechanisms from the pFL-3.6 and pFL-2.7 cDNAs. Taken together, these results suggest that pUL138 expression is ensured under a variety of cellular contexts and that UL138 cistron expression from the pFL-3.6 and pFL-2.7 cDNAs is inducible under stress.

In parallel, we isolated RNA from cells nucleofected under serum-stressed or fed conditions and used qRT-PCR to confirm that the increase in FL expression measured from the pFL-3.6 and pFL-2.7 cDNAs in Fig. 8B was not due to (i) increased transcript levels or (ii) shorter, monocistronic transcripts specific to serum stress. We quantitated the 5′ and 3′ ends of transcripts generated after nucleofection by using qRT-PCR, as we did for the experiment shown in Fig. 3C. Our results demonstrate equivalent levels of 5′ and 3′ ends under fed or stressed conditions for each pFL-cDNA (e.g., stressed/fed) (Fig. 8C), suggesting that increased FL expression from pFL-3.6 and pFL-2.7 is not due to an increase in FL transcripts in the stressed condition. Additionally, we measured a ≤2-fold change in amplification of 5′ and 3′ transcript ends relative to cDNA derived from iv-RNA for all samples (see Materials and Methods), with the exception of pFL-3.6 (stressed) and pFL-1.4 (fed) (Fig. 8C). These data suggest that most transcripts are full length and independent of cellular context. The two points outside the 2-fold confidence interval indicate that there are a greater number of 5′ ends relative to 3′ ends and therefore cannot account for the increased FL expression. Importantly, equivalent 5′ and 3′ transcript ends in the pFL-2.7 fed and stressed samples indicate that the 9-fold induction of FL expression (Fig. 8B) is not due to changes in transcription.

To confirm that our cDNA nucleofection results are reflective of bona fide IRES activity, we analyzed full-length RNAs transcribed in vitro for their ability to support expression from the downstream UL138 cistron. RNA was transcribed in vitro using linearized pFL-cDNA (Fig. 8A) as a template, and then equal molar equivalents were either translated in RRL (Fig. 8D) or transfected into HEK-293 cells (Fig. 8E). Levels of FL expression from the pFL-2.7 and pFL-1.4 cDNAs in the RRL in vitro translation system were comparable (Fig. 8D). In contrast, FL expression from the pFL-3.6 cDNA was minimal, consistent with previous experiments. Similarly, when the pFL-2.7 and pFL-1.4 iv-RNAs were transfected into cells, similar levels of FL expression were measured (Fig. 8E). Taken together, these results suggest that the pFL-2.7 and pFL-1.4 cDNAs function efficiently as templates for translation. These data further support our interpretation that translation of the UL138 cistron from the pFL-3.6 or pFL-2.7 cDNAs requires alternative mechanisms of translation initiation and cannot be explained by the presence of monocistronic RNAs encoding the UL138 cistron.