In this study we analyzed the in vitro expression of eleven 4.1R cDNAs, previously cloned by our group 3233, by coupled in vitro transcription and translation reactions using a T7 reticulocyte lysate system. Of the eleven 4.1R cDNAs, seven contained exon 2' (4.1R¹³⁵ cDNAs) and, therefore, the upstream translation initiation site ATG1, which is used for the synthesis of long isoforms of protein 4.1R (Figure 1A and Figure 1B). The other four cDNAs lacked exon 2' (4.1R⁸⁰ cDNAs) and hence give rise to short isoforms of protein 4.1R by using the downstream ATG2 present in exon 4 as a translation initiation site (Figure 1A and Figure 1C). For simplicity, we will name 1' to 7' the set of exon 2'-containing cDNAs and 1 to 4 the set lacking exon 2'. It should be mentioned that the long isoforms of protein 4.1R are generally referred to as 4.1R¹³⁵ or ~135 kDa, although their sizes range from ~97 kDa to 135 kDa. Similarly, the short isoforms of protein 4.1R are generally referred to as 4.1R⁸⁰ or ~80 kDa, although their sizes range from ~62 kDa to 80 kDa.

Focusing on the results of the in vitro expression of cDNAs 1' to 7', we observed that all of them gave rise to the expected ATG1-translated proteins (Figure 1D) and, in addition, four of them produced a second product of faster electrophoretic mobility (Figure 1D, lanes 1', 2', 3' and 4'). By analyzing the exon composition of cDNAs 1' to 7', we noticed that those giving rise to two products all included exon 4 (Figure 1B). Notably, cDNAs 2' and 5' generated two and one 4.1R product, respectively, and differed only by the presence (cDNA 2') or absence (cDNA 5') of exon 4 (Figure 1B). cDNAs 1 to 4 use the ATG2 present in exon 4 as a translation initiation site and their exon composition is similar to that of cDNAs 1' to 4', apart from lacking exon 2', so we compared the size of their in vitro-expressed products. Figure 1D shows that the size of the ATG2-translated products originating from cDNAs 1 to 4 was identical to that of the faster-migrating product synthesized from cDNAs 1' to 4'. These results suggest that the ATG2 site is probably used in 4.1R¹³⁵ cDNAs to synthesize the short 4.1R products.

We next performed a directed mutagenesis experiment to replace the ATG2 codon by GTG in two of the 4.1R¹³⁵ cDNAs, 1' and 4' (Figure 2A, 1'ATG2mut and 4'ATG2mut). The mutant cDNAs were in vitro-transcribed and translated, and the synthesized products analyzed by autoradiography (Figure 2B). The mutation completely abolished the synthesis of the short 4.1R product from both 4.1R¹³⁵ cDNAs, whereas the synthesis of the ATG1-translated product remained unchanged (compare lanes 1' and 1'ATG2mut and lanes 4'and 4'ATG2mut). These results are consistent with the hypothesis that the short protein originates from 4.1R¹³⁵ cDNAs translated from the downstream ATG2 site.

The results described above also suggest that the short 4.1R product is unlikely to be produced from proteolytic cleavage of the long 4.1R protein. However, to rule out this possibility definitively we introduced a stop codon in cDNA 4' upstream of the ATG2 by site-directed mutagenesis (Figure 2A, 4'STOP). Figure 2B shows that, indeed, in the absence of the full-length ATG1-translated product, the short protein product was still generated (4'STOP). Additionally, when the 4.1R¹³⁵ cDNAs were expressed in vitro in the absence of the T7 promoter, no protein products were observed (data not shown).

These results indicate that the 4.1R sequence does not contain a cryptic promoter directing the synthesis of the short protein product, and that the short product does not result from proteolytic cleavage of the long one.

All previous experiments were developed in a coupled in vitro transcription-translation reticulocyte lysate system. We next considered whether the ATG2 site of 4.1R¹³⁵ cDNAs was also used in living cells to give rise to the production of the short 4.1R protein. To investigate this, cDNAs 1', 2', 3' and 4' were independently transfected in COS-7 cells and the expressed proteins were analyzed by western blot. To detect the exogenous ATG1-translated 4.1R protein specifically we used the anti-FLAG antibody, since this epitope was added to the N-terminus of the long 4.1R molecule. To detect both the exogenous ATG1- and ATG2-translated 4.1R proteins we used the anti-myc antibody as this epitope was added to the C-terminus of the 4.1R molecule. Figure 3A and 3B shows representative results obtained for cDNAs 1' and 4' that were similar to those observed for cDNAs 2' and 3' (data not shown). As expected, the long protein was synthesized and recognized by both antibodies (Figure 3A lanes 1' and Figure 3B lanes 4'). The short protein was also synthesized and detected by the anti-myc antibody but not by the anti-FLAG antibody (Figure 3A lanes 1' and Figure 3B lanes 4'). The size of the short protein generated from cDNAs 1' and 4' was similar to that of the ATG2-translated 4.1R isoform synthesized from the corresponding 4.1R⁸⁰ cDNA (compare Figure 3A lanes 1' and 1 and Figure 3B lanes 4' and 4). These results indicate that the short protein most probably corresponds to the 4.1R product synthesized from the ATG2 site.

Consistent with the results observed in vitro, all these data suggest that the ATG2 site present in the set of 4.1R¹³⁵ cDNAs containing exon 4 is indeed used in living cells.

A cryptic promoter could be responsible for the production of mRNA species giving rise to the short 4.1R protein detected in the immunofluorescence and western blot analyses. To test this possibility we eliminated the cytomegalovirus (CMV) promoter from the construct containing cDNA 1' (Figure 4A, 1'ΔCMV) and performed transfection experiments in COS-7 cells to analyze the protein expression pattern by western blot (Figure 4B) using the anti-myc and the anti-FLAG antibodies. The long and short 4.1R protein bands were observed for the control cDNA 1' construct, whereas removal of the CMV promoter resulted in a lack of synthesis of both 4.1R proteins (Figure 2B, lanes 1'ΔCMV). Thus, the short 4.1R protein detected in COS-7 cells transfected with the set of 4.1R¹³⁵ cDNAs containing exon 4 is not due to the existence of an internal promoter.

Having ruled out the existence of a cryptic promoter and proteolytic cleavage as the mechanisms for generating the short 4.1R isoforms, we considered internal translation to be the most plausible mechanism responsible for their synthesis. To test this hypothesis, a sequence comprised of 583 nucleotides upstream of the ATG2 (from nucleotide 39 to 621 in the reference sequence AF156225) of 4.1R was inserted into a bicistronic plasmid between the Renilla and firefly luciferase reporter genes (RLuc-4.1s-FLuc). Bicistronic plasmids containing the 583 nucleotides in an antisense orientation (RLuc-4.1a-FLuc) or the foot-and-mouth disease virus (FMDV) IRES (RLuc-FMDV-FLuc) inserted between the two cistrons were used as negative and positive controls, respectively (Figure 5A). These plasmids were independently transfected into COS-7 cells and the expressed proteins were detected by enzymatic activity assays. As shown in Figure 5B, cell lysates transfected with RLuc-4.1s-FLuc were positive for firefly luciferase suggesting that the cloned 583 nucleotides of 4.1R contain a functional IRES. Expression of the second cistron directed by the 4.1R sequence was lower than that directed by the FMDV IRES. A second bicistronic system consisting of DsRed and enhanced green fluorescence protein (EGFP) as first and second reporter cistrons, respectively, corroborated the results obtained with the Renilla/firefly luciferase system (Figure 5B). As the fluorescence intensity of DsRed and EGFP is very easily measured by flow cytometry, we adopted this system for the rest of the experiments in this study.

To investigate the possibility that RNA cleavage or RNA splicing contributed to the expression of the second reporter gene, we examined the integrity of the bicistronic transcript by northern blot analysis (Figure 5C) following transfection of COS-7 cells with the bicistronic vector DsRed-4.1s-EGFP or with DsRed-4.1a-EGFP or with a plasmid coding only for EGFP (pEGFP vector), which was used as monocistronic transcript control (Figure 5A). A band corresponding to the expected size of 2100 bp for the bicistronic RNAs was detected with an EGFP-specific cDNA probe from total RNA extracts isolated from cells transfected with the bicistronic vectors (Figure 5C). Monocistronic EGFP mRNA (720 bp) species were not detected in any of the bicistronic vectors used (compare lanes DsRed-4.1s-EGFP and DsRed-4.1a-EGFP with pEGFP).

A positive bicistronic test indicates that an alternate initiation mechanism is involved in the production of the short 4.1R isoform. A number of internal translational initiation mechanisms have been described: reinitiation, leaky scanning, shunting and internal ribosome entry 27. Reinitiation cannot explain the generation of the short 4.1R isoforms. To distinguish between the other three possibilities, we generated an additional bicistronic construct in which a synthetic stem loop with a ΔG° = -57 kcal/mol, which is known to decrease the scanning of the ribosome 34, was inserted either 5' to the first cistron, HDsRed-4.1s-EGFP, or between the stop codon of the first cistron and the 4.1R sequence, DsRedH-4.1s-EGFP (Figure 6A). These constructs and the original DsRed-4.1s-EGFP construct were independently transfected in COS-7 cells and expression of DsRed and EGFP was determined by flow cytometry. Figure 6B illustrates that the hairpin abolished the expression of the first cistron when added to its 5' end but not the expression of the second cistron (lanes HDsRed-4.1s-EGFP). The second cistron was also expressed when the hairpin was added downstream of the DsRed reporter gene (lanes DsRedH-4.1s-EGFP). These results indicate that the 4.1R sequence comprised between the ATG1 and the ATG2 contains an internal ribosome entry site.

To determine the influence of the 5' region upstream of the ATG2 on the usage of the second cistron, we prepared three different 5' deletion constructs (Figure 7A, DsRed-413s-EGFP, DsRed-215s-EGFP and DsRed-149s-EGFP). As negative controls, we used the same 4.1R sequences in antisense orientations. All these constructs were transfected in COS-7 cells and fluorescence intensity was analyzed by flow cytometry (Figure 7B). The construct lacking 170 nucleotides from the 5' region of exon 4 (made up of 413 nucleotides) had slightly lower efficiency than the construct with the entire sequence (583 nucleotides). The other two deletion constructs had very low efficiencies, almost comparable with that of their corresponding antisense controls. Thus, removal of 5' sequences negatively affects the translation of the second cistron, being nucleotides 170 to 368 essential for IRES activity.

Although the precise mechanism by which cellular IRESs promote translation is not fully understood, the participation of IRES-binding proteins, referred to as ITAFs, is well documented 31. To analyze proteins interacting with the 583 nucleotide segment containing the 4.1R IRES, we performed UV-crosslinking experiments. Reticulocyte lysates were used as a source of ITAFs. As a control we used a 4.1R riboprobe lacking IRES activity, the 149-ribonucleotide fragment corresponding to the 3' terminal end of the 583 ribonucleotide sequence (see Figure 7). Figure 8 shows that the ³²P-labelled 583 riboprobe crosslinked proteins ranging in size from 22 to 80 kDa which, by contrast, the 149 ribonucleotide fragment did not crosslink. The crosslinking of these proteins was globally competed by incubation with a 150-fold molar excess of the unlabelled 583 riboprobe. A protein of approximately 57 kDa was one of the most prominent bands bound to the 583 riboprobe. As PTB is a ~57 kDa protein that interacts with pyrimidine-rich sequences of many viral and cellular IRESs, we analyzed the possible interaction of PTB with the 583 riboprobe. Figure 8B shows that the 583 but not the 149 riboprobe crosslinked purified, recombinant PTB. This result is consistent with the presence of a polypyrimidine-rich tract in the large fragment, which is absent from the short one.

COS-7 cells were grown as described 41. Transfection experiments were performed by electroporation using the Electro Cell Manipulator 600 (BTX, San Diego, CA). Cells were processed 48 hours after transfection.

Anti-FLAG antibody is a rabbit polyclonal antibody (Sigma, Saint Louis, MO). Anti-c-myc monoclonal antibody 9E10 was obtained from the American Type Culture Collection. Anti-horseradish peroxidase-labelled secondary antibodies were obtained from Southern Biotechnology Associates, Inc. Goat anti-rabbit immunoglobulin G (IgG) (H+L) secondary antibody conjugated to Alexa Fluor 488 was obtained from Molecular Probes.

4.1R cDNAs used for in vitro and in vivo expression were generated by RT-PCR using total RNA from MOLT-4 T cells as described previously 3233. These cDNAs encoded 4.1R proteins tagged at the C-terminus with the c-myc epitope. Some of the cDNAs encoded 4.1R proteins also tagged at the N-terminus with the FLAG epitope. The cDNAs were cloned into pCR3.1 (Invitrogen) and verified by sequencing 41.

In two 4.1R cDNAs, previously designated 4.1R¹³⁵Δ16 and 4.1R¹³⁵Δ16,18,19 (for simplicity named in this study cDNAs 2' and 4', respectively) the ATG2 present in exon 4 was mutated and the adenine was substituted by guanosine, thus generating 1' ATG2mut and 4' ATG2mut, respectively. The mutant 4' STOP carried a nucleotide substitution at position 182 of exon 2 that changes the triplet AAG to the stop triplet TAG. These three mutant constructs were obtained using the QuikChange site-directed mutagenesis following the manufacturer's instructions (Stratagene). In vitro protein expression was achieved by coupled in vitro transcription and translation reactions, as previously described 41.

The DsRed coding sequence was amplified by PCR from the pDsRed1-N1 vector (Invitrogen) with the sense primer NheI-red (5'-CCGGTCGCTAGCATGGTGCGCTCC-3') and the antisense primer BglII-stop-red (5'-CTAGAGTAGATCTCGCTACAGGAA-3'). The PCR product was cloned between the NheI and BglII sites of pEGFP-N1 (Invitrogen).

Different 4.1R sequences, amplified by PCR, were inserted into the SacI site of the bicistronic vector. The sense primers used for the DsRed-4.1s-EGFP, DsRed-413s-EGFP, DsRed-215s-EGFP, DsRed-149s-EGFP construct were 5'-GGCCGGAGCTCCACAGCACCAAC-3', 5'-GACTTGACCGAGCTCAAGGAGCGGACA-3', 5'-CCCAATTGCAGAGCTCGAACCGGAAC-3' and 5'-GCAGAAACAGAGCTCGCTCAGGAAGAAC-3', respectively, and the antisense primer was 5'-GCAGTGCATGAGCTCGTGTTTTCTGATTGG-3' in all cases.

The hairpin structure (CCGGATCGG)₃ described by Koromilas et al 34 was cloned between the XhoI and NheI sites of DsRed-4.1s-EGFP to generate the DsRedH-4.1s-EGFP and the HDsRed-4.1s-EGFP constructs, respectively.

The 4.1R sequence amplified with the 5'-GGCCGGAGCTCCACAGCACCAAC-3' and 5'-GCAGTGCATGAGCTCGTGTTTTCTGATTGG-3' primers was cloned into the SacI site of the RLuc-FLuc bicistronic vector in sense (RLuc-4.1s-FLuc) and antisense (RLuc-4.1a-FLuc) orientations.

Relative IRES activity was quantified in vivo as the ratio of the expression of firefly luciferase to that of Renilla luciferase in COS-7 cells transfected with the bicistronic constructs. Luciferase expression was determined from 1–2 × 10⁵ cells using the dual luciferase reporter assay system following the manufacturer's instructions (Promega) in a luminometer (Berthold). Five independent experiments were performed in triplicate.

Total RNA from COS-7 cells transfected with different constructs was extracted using TRIZOL Reagent (Invitrogen Life Technologies). For northern blot analysis, 20 μg of RNA were denatured in 50% formamide and 2.2 M formaldehyde at 65°C, subjected to electrophoresis in a 1% agarose/formaldehyde gel, and transferred to nylon membranes. RNA samples were hybridized under standard conditions to labelled EGFP cDNA. Final blot washing conditions were 0.5 × SSC/0.1% SDS (1 × SSC = 0.15 M NaCl, 0.015 M sodium citrate, pH 7.0) at 65°C.

To generate RNA riboprobes, PCR was performed with specific primers for the indicated 4.1R fragments to which additional sequences were added for incorporating the T7 RNA polymerase promoter at the 5' end. Radiolabeled RNA probes were prepared by transcription with T7 RNA polymerase in the presence of 0.08 mM unlabelled rUTP plus 25 μCi of (α-³²P)UTP (400 Ci/mmol)(Amersham).

12.5 μl of rabbit reticulocyte lysates were incubated with radiolabeled probes at 30°C for 30 minutes. The reaction mixtures were exposed to UV (254 nm) (Stratalinker 1800; Stratagene) for 10 minutes on ice. Then 20 units of RNase A was added to the reaction and incubated during 10 minutes at 37°C. For competition experiments, a 150-molar excess of unlabelled RNA was added 10 minutes before the addition of the radiolabeled probe. For PTB-4.1R interaction experiments, 100 ng of recombinant His-PTB (a gift from Dr. J.M. Izquierdo, Centro de Biología Molecular Severo Ochoa, Madrid) was incubated with the appropriate radiolabeled probes. The RNA-protein complexes were resolved by SDS-PAGE.

COS-7 cells were fixed with 4% formalin (37% formaldehyde solution; Sigma), permeabilized, blocked, incubated with the appropriate antibodies, and processed as described 4. Controls with primary antibodies omitted were included in each experiment. Preparations were examined under a Zeiss epifluorescence microscope.

Protein samples were separated by SDS-polyacrylamide gel electrophoresis and transferred to Immobilon polyvinylidine difluoride (Millipore) in Tris (tris(hydroxyl-methyl)aminomethane)-borate buffer, pH 8.2. Membranes were processed and developed as described 4.

Transfected cells were detached from the dish and suspended at 0.5–1 × 10⁶ cells/ml in phosphate-buffered saline, 2 mM EDTA. Samples were analyzed by flow cytometry using an argon laser at 488 and 558 nm to detect EGFP and DsRed expression, respectively, in a Calibur cytometer (Becton-Dickinson). Four to five independent experiments were performed in triplicate.