The apoptotic protease activating factor (Apaf-1) plays a central role in apoptosis: interaction of this protein with procaspase-9 leads to cleavage and activation of this initiator caspase. In common with other mRNAs whose protein products have a major regulatory function, the 5′ untranslated region (UTR) of Apaf-1 is long, G-C rich and has the potential to form secondary structure. We have shown that the 5′ UTR of Apaf-1 contains an internal ribosome entry segment, located in a 233 nucleotide region towards the 3′ end of the leader, and that the translation initiation of this mRNA occurs only by internal ribosome entry. The Apaf-1 IRES is active in almost all human cell types tested, including Human cervical carcinoma (HeLa), Human liver carcinoma (HepG2), Human breast carcinoma (MCF7), Human embryonic kidney (HK293), African Green Monkey kidney (COS7) and Human lung (MRC5). The Apaf-1 IRES initiates translation as efficiently as the HRV IRES, but is less active than the c-myc IRES. We propose that the Apaf-1 IRES ensures that a constant cellular level of Apaf-1 protein is maintained even under conditions where cap-dependent translation is compromised.
Programmed cell death or apoptosis is a process that is important in numerous biological systems including tissue development and immune system maintenance (Jacobson et al., 1997). The disruption of apoptosis can result in the progression of cancer and the onset of degenerative disorders.
In the nematode Caenorhabditis elegans, morphogenetic apoptosis during development requires the products of three essential genes, ced-9, ced-4 and ced-3 (Horvitz et al., 1994). The mammalian homologues of these genes have been extensively studied and there are multiple forms of both ced-9 and ced-3. The genes of the Bcl-2 family are the mammalian homologues of ced-9 and proteins in this group have either anti- or pro-apoptotic functions (Adams and Cory, 1998; Reed et al., 1998). The mammalian ced-3 homologues comprise a group of related cysteine proteases, termed caspases, since they cleave proteins after a specific aspartate residue. They all exist as zymogens and are cleaved and activated in response to apoptotic signals (Cohen, 1997; Cryns and Yuan, 1998; Nunez et al., 1998).
The discovery of the human homologue of CED-4, the apoptotic protease activating factor (Apaf-1) (Zou et al., 1997) confirmed the conservation of the suicide programme between nematodes and humans. Apaf-1 monomers are maintained in an inert form by both the binding of Bcl-XL (Pan et al., 1998; Hu et al., 1998a) and also by self-association of C-terminal WD-40 repeats (Hu et al, 1998b). Following an apoptotic stimulus, cytochrome c that has been released from the mitochondria binds to Apaf-1 (Li et al., 1997). At the same time dATP binds to Apaf-1 at a P loop found in the ced-4 homology region (Cardozo and Abagyan, 1998). Hydrolysis of the dATP causes oligomerization of Apaf-1 monomers into an ‘apoptosome’ consisting of up to eight Apaf-1 molecules (Zou et al., 1999; Saleh et al., 1999). This oligomerization permits binding of procaspase-9 and encourages interactions between procaspase-9 molecules, thus allowing intermolecular cleavage and activation to occur. This initiator caspase can then cleave procaspase-3, one of the effector caspases essential for cleavage of a number of key substrates during the execution phase of apoptosis.
Apaf-1 is expressed in almost all tissues studied with the exception of skeletal muscle. In these cells very little Apaf-1 mRNA can be detected by RT–PCR or Northern blot analysis and no Apaf-1 protein is observed by Western blotting (Zou et al., 1997; Burgess et al., 1999). This may be an important control point as after strenuous activity mitochondria in skeletal muscles have been shown to swell and/or degenerate (Zamora et al., 1995), which could lead to accidental triggering of the apoptotic pathway as a consequence of cytochrome c release.
Homozygous Apaf-1 knockout mice die early in embryogenesis and show reduced apoptosis leading to many morphological abnormalities (Cecconi et al., 1998; Yoshida et al., 1998). In murine Apaf-1 knockout cell lines, overexpression of c-Myc can promote tumorigenesis in a manner analogous to cell lines lacking p53 (Soengas et al., 1999). Conversely, overexpression of Apaf-1 increases the sensitivity of cells to stimulators of apoptosis such as etoposide and paclitaxel (Perkins et al., 1998).
The 5′ untranslated regions (UTRs) of both the human and mouse versions of Apaf-1 have an overall homology of 56% and have many of the features of a translationally regulated gene. They are long (578 and 585 nucleotides respectively), G-C rich (67 and 66% respectively) and contain upstream AUG codons (two in the human, one in the mouse). The single upstream AUG codon in the murine 5′ UTR is conserved in the human version and lies in a region of 70% homology approximately 40 nucleotides upstream of the physiological start site.
The high degree of secondary structure in the 5′ UTR of an mRNA can contribute to the control of translation in a number of ways. For example, regions of stable secondary structure may be inhibitory to the normal scanning mechanism of translation initiation (Pain, 1996), or they may contain an internal ribosome entry segment (IRES). IRESs were originally identified in the 5′ UTRs of picornaviral RNAs and they act to direct ribosome binding straight to the initiation codon, either directly or indirectly via other linking factors (Jackson et al., 1994, 1995; Jackson and Kaminski, 1995). Recently a number of IRESs have been identified in several eukaryotic mRNAs whose protein products are associated with cell growth or cell death including vascular endothelial growth factor (VEGF) (Stein et al., 1998; Miller et al., 1998), fibroblast growth factor-2 (FGF-2) (Vagner et al., 1995), platelet derived growth factor (PDGF) (Bernstein et al., 1997) and c-myc; (Nanbru et al., 1997; Stoneley et al., 1998).
In this study we have investigated the 5′ UTR of Apaf-1 and shown that it contains an IRES. This is the first example of an IRES to be identified in a gene whose protein product is intimately associated with the initiation of apoptosis and our data suggest that initiation of translation of Apaf-1 only occurs by internal ribosome entry and not by the more generally used scanning mechanism of translation.
The 5′ untranslated region of Apaf-1 inhibits the translation of a downstream reporter gene in vitro
The 5′ UTR of Apaf-1 was introduced upstream of the luciferase start codon at the NcoI site in the vector pSKL (Stoneley et al., 1998); this ensured that it was in the correct reading frame (Figure 1a). Capped mRNAs generated from this vector using T7 RNA polymerase were then used to prime rabbit reticulocyte lysates and the amount of luciferase generated was measured (Figure 1b). The presence of the Apaf-1 5′ UTR was found to potently inhibit the translation of the downstream luciferase gene. For example, when the lysates contained 20 ng of pSKAL RNA, 100-fold less luciferase was produced (Figure 1b). This suggested either that a region of stable secondary structure was obstructing the scanning translation initiation complex or that non-canonical factors which were not present in rabbit reticulocyte lysates were required for translation initiation of Apaf-1.
The Apaf-1 5′ UTR causes a reduction in the translation of a luciferase reporter enzyme in vivo
The Apaf-1 5′ UTR was inserted into the monocistronic luciferase reporter vector pGL3 (Figure 2a) to create the vector pGAL. Both constructs were transfected by the calcium phosphate method into HeLa cells, and luciferase activity measured (Figure 2b). In the presence of the Apaf-1 5′ UTR upstream of the luciferase reporter there was again a significant decrease in the amount of luciferase produced compared to the control. However, the effect was much less dramatic than that observed in vitro and in this case approximately fourfold less luciferase was produced (Figure 2b). This would suggest either the presence of stable secondary structure in the Apaf-1 UTR which was inhibitory to scanning or that the initiation of Apaf-1 translation is mediated by an alternative mechanism which is less efficient than scanning. To distinguish between these two possibilities, a stable hairpin (Stoneley et al., 1998) was introduced upstream of the control vector and the Apaf-1 5′ UTR to create the vectors phpL and phpAL respectively. The presence of the hairpin reduced the luciferase produced by the control vector 200-fold, but had only a small effect (1.5-fold) on the amount of luciferase produced from phpAL. This effect may have been due to the hairpin destabilizing structural elements in the Apaf1 5′ UTR. This data suggests that Apaf-1 translation can indeed be initiated by an alternative mechanism e.g. internal ribosome entry. To determine whether Apaf-1 5′ UTR contained an IRES it was inserted into the dicistronic construct pRF (formerly known as pGL3R, Stoneley et al., 1998) to give pRAF which incorporates both the firefly and sea pansy (Renilla reniformis) luciferase reporter genes (Figure 3a).
Apaf-1 5′ UTR directs internal ribosome entry in dicistronic reporter vectors
HeLa cells were transiently transfected with either pRF, pRAF or pRMmAF (the murine form of Apaf-1 5′ UTR which is 56% homologous to the human version) by the calcium phosphate method. Expression of Renilla and firefly luciferase were assayed using the ‘Stop & Glo’ assay kit (Promega) and we show that both Apaf-1 sequences can direct internal ribosome entry in the dicistronic assay. The amount of firefly luciferase produced from the human and murine Apaf-1 containing constructs was ten and 15-fold over readthrough. These values are slightly higher than those obtained from cells transfected with the HRV-IRES-containing plasmids but less than those obtained from cells transfected with plasmid constructs which harbour the c-myc-IRES (Figure 3b).
IRES activity is not due to enhanced ribosomal readthrough
The result from the dicistronic assay alone does not prove the existence of an IRES since ribosomal readthrough could be enhanced by signals in the UTR. Thus to eliminate this possibility a stable hairpin (−55 kcal/mol) was inserted upstream of the Renilla luciferase to create the plasmid vectors phpRAF and phpRmAF (Figure 4a). This hairpin was sufficient to impede ribosomal scanning, and Renilla activity produced from these two constructs dropped to 15–20% of wild type, while firefly luciferase activity directed by the Apaf-1 5′ UTRs was maintained. This demonstrates that increased readthrough cannot account for the increased translation of firefly luciferase (Figure 4b).
IRES activity is not due to the presence of a cryptic promoter or aberrant splicing
It is possible that the increase in firefly luciferase activity in the vector pRAF was due to the presence of a cryptic promoter or splicing signals in the 5′ UTR, resulting in monocistronic RNAs being produced. To show that only full length mRNA was being produced from the dicistronic constructs, cells were transfected with pGL3, pGAL, pRF and pRAF. Poly(A)+selected RNA was subjected to formaldehyde gel electrophoresis and then transferred onto nitrocellulose. The blot was then probed with 32P-labelled luciferase fragments that were generated using random prime labelling (Figure 5). Transcripts corresponding to full length pGL3, pGAL, pRF and pRAF were observed. In all cell lines a luciferase transcript of approximately 1.3 knt in length was also detected (*) which probably represents an aberrantly processed luciferase transcript. This fragment cannot account for the IRES activity observed in the vector pRAF as densitometry analysis of this luciferase fragment showed that in the vectors pRF and pRAF these transcripts are present in approximately equal amounts (0.8 : 1 respectively). However, ten times more firefly luciferase is observed in cells transfected with pRAF compared to pRF (Figure 3b). This transcript is also present in cells that have been transfected with a dicistronic vector containing the c-myc IRES (Stoneley, 1998).
Mapping the Apaf-1 IRES
To define the boundaries of the Apaf-1 IRES a series of plasmid constructs was generated containing decreasing lengths of the sequence coding for the 5′ UTR (Figure 6). The ability of these truncated sequences to promote internal ribosome entry on a dicistronic mRNA was compared to the full length 5′ UTR. Although the full length IRES is required for maximal luciferase activity, the data suggest that elements which are vital to internal ribosome entry lie between nucleotides −233 and 1, since fragments from −115−1 have only 20% of the luciferase activity of the full length 5′ UTR, yet fragments from −233−1 maintain 75% of the activity (Figure 6b).
The Apaf-1 IRES is used in a wide range of cell lines
To investigate how widely the Apaf-1-IRES is utilized, a panel of cell lines derived from different tissues, including Human cervical carcinoma (HeLa); Human liver carcinoma (HepG2); Human breast carcinoma (MCF7); Human embryonic kidney (HK293); African Green Monkey kidney (COS7); Human lung (MRC5); Human neuronal cells (SY5Y); Mouse fibroblasts (Balb/c) and Chinese Hamster ovary T (CHO-T) cells were co-transfected using either FuGene 6 or the calcium phosphate method with either pRF or pRAF (Figure 2a) and pcDNA3.1/HisB/lacZ. The expression from both Renilla and firefly luciferase cistrons was assayed and normalized to the transfection control, β-galactosidase and expressed relative to the luciferase activities of the control vector in each line (Figure 7). The human Apaf-1 IRES is most active in HeLa and HepG2 cells, and is not used in SY5Y neuronal cells or Balb/c mouse cells suggesting that these cell types lack non-canonical protein factors that are required for internal ribosome entry via the Apaf-1 IRES.
The dicistronic vector pRmAF, which contains the murine version of the Apaf 5′ UTR was also tested in the rodent cell lines Balb/c and CHO-T. Interestingly, the murine version of Apaf-1 was able to direct IRES activity in the CHO-T cells (eight times more firefly luciferase activity than the control vector) and again not in the Balb/c cells (Figure 7).
Apaf-1 5′ UTR contains an IRES
Our data show that translation initiation of Apaf-1 occurs by internal ribosome entry and this provides evidence for an IRES in an mRNA whose protein product is associated with the initiator of a caspase cascade during apoptosis. Experiments performed with monocistronic plasmid constructs that contain a stable hairpin imply that initiation of translation of Apaf-1 mRNA may only occur by internal ribosome entry. Moreover, the reduction in luciferase activity obtained when the Apaf-1 5′ UTR was placed in front of the luciferase reporter vector suggests that internal ribosome entry initiated by the Apaf1 IRES is approximately 70% less efficient than the scanning mechanism of translation. In this way, the Apaf-1 IRES differs significantly from those identified in c-myc and FGF-2 mRNAs, since it has been proposed that initiation of translation of these mRNAs can occur by cap-dependent mechanisms as well as internal ribosome entry (De Benedetti and Rhoads, 1990; De Benedetti et al., 1994; Kevil et al., 1995; West et al., 1998). The Apaf-1 IRES was active, although to differing degrees in all human cell types tested, with the exception of the neuronal line SY5Y. The differences in Apaf-1 IRES-mediated initiation of translation between the cell lines examined may be due to the levels of expression of proteins that are required for internal ribosome entry in these cells. Interestingly, the IRES is most active in cell lines of tumour origin. Most of the ability of the Apaf-1 5′ UTR to initiate internal ribosome entry resides in a 233 nt fragment and although this is a rather small IRES when compared to those of viral origin, it is a size that is consistent with those that have been observed in eukaryotic mRNAs e.g. the Bip and FGF-2 IRESs are 220 and 165 nucleotides in length respectively (Yang and Sarnow, 1997; Vagner et al., 1995).
Function of the Apaf-1 IRES
A reduction in the cap-dependent scanning mechanism of translation can be induced by a variety of cellular conditions. A reduction in translation occurs (i) during mitosis due to de-phosphorylation of eIF4E (Bonneau and Sonenberg, 1987); (ii) following cellular stress such as heat shock, osmostic shock and hypoxia which all cause a shut down of scanning by modulation of the activities of the eIF4F complex and/or eIF-2 (Panniers, 1994; Rhoads and Lamphear, 1995); (iii) during picornaviral infection due to cleavage of eIF4G by viral proteases, providing a rationale for the preferential translation of viral mRNAs due to their mechanism of internal ribosome entry (Lamphear et al., 1995; Ohlmann et al., 1996; Pestova et al., 1996) and (iv) during apoptosis where eIF4G is also cleaved, but in this case by caspase-3 (Clemens et al., 1998; Marissen and Lloyd, 1998; Morley et al., 1998). Some of the eukaryotic IRESs identified thus far are active during these situations, hence the VEGF IRES is active during hypoxia (Stein et al., 1998). We have recently shown that the c-myc-IRES is utilized during apoptosis (Stoneley et al. 2000 in press) and it has been shown that many eukaryotic IRESs are functional following Picornaviral infection (Johannes and Sarnow, 1998). Our data would suggest that the Apaf-1 IRES is used continuously, and we propose that internal ribosome entry can be used as a default pathway for protein synthesis initiation. This would also imply that it is important for a cell to be able to maintain a constant cellular level of Apaf-1 which is not all that surprising given the central role that Apaf-1 plays in apoptosis. It has recently been shown that X-linked inhibitor of apoptosis (XIAP) contains an IRES and the data suggest that inititation of translation of this mRNA also only occurs by internal ribosome entry (Holcik et al, 1999).
In summary we have shown that Apaf-1 5′ UTR contains an IRES and our data demonstrate that translation initiation of Apaf-1 occurs primarily by this mechanism.
Materials and methods
Media and serum were purchased from GIBCO BRL, Luciferase assay kits ‘Stop & Glo’ and rabbit reticulocyte lysates were purchased from Promega. Galactolight plus assay system was purchased from Tropix. All cells were purchased from the American Type culture collection. FuGene6 was purchased from Roche Molecular Biochemicals. All other chemicals were purchased from Sigma (Poole, UK).
Cell lines were grown at 37°C in Dulbecco's modified Eagle's Medium supplemented with 10% foetal calf serum (FCS), in a humidified atmosphere containing 5% CO2. SY5Y were grown in 50% HAMS F12, 50% DMEM and 10% FCS.
Calcium phosphate-mediated DNA transfection of HeLa, HepG2 and Cos-7 cells was performed essentially as described by Jordan et al. (1996). All other cell lines were transfected using FuGene6 according to manufacturer's protocols.
In vitro run-off transcription and in vitro translation
Vector DNA was linearized by restriction digestion using a site downstream of the sequence of interest. Capped transcripts were synthesized in a reaction containing 1× Transcription buffer (80 mM Hepes-KOH, pH 7.5, 24 mM MgCl2, 2 mM spermidine, 40 mM DTT), 7 mM KOH, 40 units of RNasin, 1 mM ATP, 1 mM UTP, 1 mM CTP, 0.5 mM GTP, 2 mM m7G(5′)ppp(5′)G, 1 μg of DNA template and 20 units of T7 RNA polymerase in a final volume of 50 μl. After incubation of the reaction for 1 h at 37°C, the RNA was isolated and used to prime a 12.5 μl in vitro translation reaction containing 8.25 μl of Promega rabbit reticulocyte lysate and up to 50 ng of RNA as recommended in the manufacturer's instructions.
Reporter gene analysis
The activity of firefly luciferase in lysates prepared from cells transfected with pGL3 or pGAL was measured using a luciferase reporter assay system (Promega). Light emission was measured over 10 s using an Optocomp-1 Luminometer (MGM instruments). The activity of both firefly and Renilla luciferase in cell lysates transfected with dicistronic luciferase plasmids was measured using the Dual-luciferase reporter assay system (Promega). Assays were performed according to the manufacturer's recommendations except that only 25 μl of each reagent was used. Light emission was measured in the same manner as described previously.
The activity of β-galactosidase in lysates prepared from cells transfected with pcDNA3.1/HisB/lacZ was measured using a Galactolight plus assay system (Tropix). Enzyme activity was then determined by measuring the light emission from the reaction in a luminometer, as previously described.
Northern blot analysis
Total cellular RNA and Poly(A)+ selected mRNA was prepared and analysed by Northern blotting exactly as described previously (West et al., 1995). DNA probes used for the detection of luciferase mRNA species were also as described (Stoneley et al., 1998).
This work was supported by a grant from the Wellcome Trust number 055580 (SA Mitchell). MJ Coldwell holds an MRC studentship.
internal ribosome entry segment
Apoptotic protease activating factor-1
eukaryotic initiation factor
About this article