Cost minimization of ribosomal frameshifts

Cost minimization of ribosomal frameshifts

ARTICLE IN PRESS Journal of Theoretical Biology 249 (2007) 162–167 www.elsevier.com/locate/yjtbi Cost minimization of ribosomal frameshifts Herve´ S...

149KB Sizes 0 Downloads 34 Views

ARTICLE IN PRESS

Journal of Theoretical Biology 249 (2007) 162–167 www.elsevier.com/locate/yjtbi

Cost minimization of ribosomal frameshifts Herve´ Seligmann Department of Evolution, Systematics and Ecology, The Hebrew University of Jerusalem, Jerusalem 91404, Israel Received 1 June 2007; received in revised form 8 July 2007; accepted 9 July 2007 Available online 18 July 2007

Abstract Properties of mRNA leading regions that modulate protein synthesis are little known (besides effects of their secondary structure). Here I explore how coding properties of leading regions may account for their disparate efficiencies. Trinucleotides that form off frame stop codons decrease costs of ribosomal slippages during protein synthesis: protein activity (as a proxy of gene expression, and as measured in experiments using artificial variants of 50 leading sequences of beta galactosidase in Escherichia coli) increases proportionally to the number of stop motifs in any frame in the 50 leading region. This suggests that stop codons in the 50 leading region, upstream of the recognized coding sequence, terminate eventual translations that sometimes start before ribosomes reach the mRNA’s recognized start codon, increasing efficiency. This hypothesis is confirmed by further analyses: mRNAs with 50 leading regions containing in the same frame a start preceding a stop codon (in any frame) produce less enzymatic activity than those with the stop preceding the start. Hence coding properties, in addition to other properties, such as the secondary structure of the 50 leading region, regulate translation. This experimentally (a) confirms that within coding regions, off frame stops increase protein synthesis efficiency by early stopping frameshifted translation; (b) suggests that this occurs for all frames also in 50 leading regions and that (c) several alternative start codons that function at different probabilities should routinely be considered for all genes in the region of the recognized initiation codon. An unknown number of short peptides might be translated from coding and non-coding regions of RNAs. r 2007 Elsevier Ltd. All rights reserved. Keywords: Gene expression regulation; Alternative coding properties; Genetic diversity; Secondary structure; Alternative initiation codon.

1. Introduction How do different mRNA leading regions affect the efficiency of protein synthesis? We still lack an integrated framework to understand how variant leading regions modulate protein synthesis. A possible approach is to consider that coding properties of these non-coding leading regions may account for their disparate efficiencies. Genes do not have necessarily a unique initiation codon, but rather different codons might be used at different frequencies according to circumstances. Indeed, many genes possess alternative initiation codons in the vicinity of the recognized initiation region (Markussen et al., 1995; Yamasaki et al., 1999; He et al., 2000; Liu et al., 2000; Tucker et al., 2001; Leissring et al., 2004; Rhee et al., 2004;

Tel./Fax: +972 2 6585876.

E-mail address: [email protected] 0022-5193/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jtbi.2007.07.007

Outten and Culotta, 2004; review in Kozak, 2002a, b), and such phenomena that diversify the products synthesized from a single sequence might have adaptive importance (Kochetov et al., 2005), as well as sometimes be pathogenic (Irvin-Wilson and Chaudhuri, 2005; Webb et al., 2005). It is also known that artificially adding an upstream in-frame start codon alleviates a translational block when the original start codon is embedded in a helix (Satchidanandam and Shivashankar, 1997). The mRNAs of heat shock proteins possess particularly stable secondary structures, so that the protein is immediately expressed when sudden high temperatures melt the structure in which the initiation codon is embedded (Nagai et al., 1991a, b). Hence these phenomena suggest that interactions between mRNA secondary structure, their sequence, and the environment determine in probabilistic ways both the levels of expression and which protein variant is expressed (in terms of which initiation codon is used). This blurs the presumably clearcut boundaries of genes, a problem that is rampant in

ARTICLE IN PRESS H. Seligmann / Journal of Theoretical Biology 249 (2007) 162–167

sequence annotation, as found for reptilian and avian mitochondrial proteins (Slack et al., 2003). Analyses here test the hypothesis that start motifs located in the 50 leading region of beta galactosidase from Escherichia coli (Barrick et al., 1994), upstream of the normal initiation codon, sometimes initiate translation. The data from Barrick et al. (1994) consist of 185 activity levels for beta galactosidase, where the 12 nucleotides that precede the wild type initiation codon (the 50 leading region) have been artificially mutated for each of the 185 cases. The hypothesis that initiation codons in these mutated 50 leading regions sometimes initiate peptide synthesis, is tested by the following experimental design: if initiation codons, in any frame (frame ‘0’ determined by the ‘normal’ downstream reading frame of the gene), precede a termination (stop) codon in the same frame as the initiation codon, this should decrease the expression of the gene, if indeed translation was initiated at the start within the 50 leading region. This is because ribosomes would in this scenario start and stop protein synthesis within the 50 leading region before reaching the actual protein-coding region, decreasing measured activity levels of beta galactosidase. Confirmation of the latter prediction would suggest that the hypothesis that ribosomes sometimes translate sequences that are usually not recognized as coding, such as 50 leading regions, is a valid working hypothesis. Note that here protein activity is used as a proxy for protein expression levels, as the properties of the 50 leading region affect expression, not enzymatic activity. As a control, the activity levels recorded for the above mutants are compared with those recorded when the 50 leading region has a stop codon that precedes (in the same frame) an initiation codon. In this case, one expects an increase in enzymatic activity, because trinucleotides that form stop ‘codons’ in an off-frame context, within protein coding sequences, apparently decrease costs of protein synthesis due to ribosome slippage. This hypothesis yields a number of predictions about various phenomena, such as the structure of genetic codes, synonymous codon usages, gene expression efficiency, ribosome stability and slippage frequencies (Seligmann and Pollock, 2004); coevolution of the genetic code with frameshift proneness in polymerases (Jestin and Kempf, 1997) and with the tendency to form secondary structure (Itzkovitz and Alon, 2007). Because the experimental test of the ‘alternative initiation codon’ hypothesis uses phenomena due to off frame stop codons, first experimental confirmation of the ‘off frame stop codon’ hypothesis for the data from Barrick et al. (1994) is presented. 1.1. Experimental tests of the off frame stop codon hypothesis Part of the variation in enzymatic activity observed after experimental manipulation of the 50 leading region in two published data sets (Gheysen et al., 1982; Ozbudak et al.,

163

2002) is explained by the ‘off frame’ stop codon hypothesis: a positive correlation exists between gene expression and the number of stops in any frame in the 50 leading region (the frame was defined according to the downstream initiation codon). Table 1 presents the same analysis for the more ample data published by Barrick et al. (1994) for enzymatic activity levels of beta-galactosidase in E. coli: artificial introduction of stops in any frame in the 50 leading region increases enzymatic activity (based on 185 artificial variants of the 50 leading region): a t-test detects a significant increase (P ¼ 0.028, 1 tailed test) in enzymatic activity for variants with a stop (in any frame) in the 50 leading region, as compared to those without any stop. Note that activities do not follow a normal distribution, and hence a t-test is not the most adequate method to use, unless one applies natural logarithm transformation. This indeed increases the significance of the comparisons in Table 1. Another alternative test, the non-parametric Mann–Whitney U test for inequality of medians, does not assume normal distribution and yields therefore even more robust statistical conclusions than the previous analyses. Correlation analyses also confirm the result that the number of stops in the 50 leading region increase gene expression (Table 2), irrespectively of the method used. Here again, the data are more adequately analyzed after natural log transformation of enzymatic activity levels, and results are qualitatively confirmed if one uses the more conservative non-parametric Spearman correlation test. Further analyses show that part of the variation in enzymatic activity that remains unexplained by the number Table 1 Enzymatic activity as a function of the number of stop motifs in the 50 leading region Stops

Mean

sd

Ln

0 1 2 3

180.93 305.72 375.16 189.90

403.94 454.08 772.58 –

3.6029 4.6532 4.7224 5.2465

Comparisons 0–1 0–2 1–2

P(t) 0.028 0.127 0.36

P(t) 0.000 0.068 0.46

n 1.9294 1.7452 1.4624 –

103 74 7 1 P(U) 0.000 0.059 0.42

Comparison of mean enzymatic activity measured for mRNAs with 50 leading regions as function of the number of stop motifs in the 50 leading region in any frame. Comparisons indicate the statistical significances (P) of differences between enzymatic activity for various combinations of numbers of stop motifs. P(t) indicates significances of t-tests between means of the measured enzymatic activity (second column), and for the natural logarithmic transformation of the activity (fourth column). This transformation normalizes the distribution of the data. P(U) indicates the statistical significance of differences between medians for the same data, using Mann–Whitney non-parametric tests. The significances of the nonparametric test, which does no assumption on the distribution of the data, are indicated to show that qualitatively, results do not depend on assuming normal distribution.

ARTICLE IN PRESS H. Seligmann / Journal of Theoretical Biology 249 (2007) 162–167

164

Table 2 Correlation of enzymatic activity with the number of stops Method

Enzymatic activity

Enzymatic activity, ln

Pearson r Weighted, n Weighted, 1/sd Spearman

Corr 0.138 0.227 0.106 0.284

Corr 0.264 0.355 0.156

P 0.015 0.001 0.076 0.000

P 0.000 0.000 0.034

Correlation coefficients between enzymatic activity and the number of stop motifs in any frame in the 50 leading region. Pearson r is for the parametric correlation coefficient, the second row indicates the statistics of the regression when weighting data points according to the number of repeat experiments to estimate enzymatic activity, the third row indicates similar analyses, but weighting data points inversely proportionally to the standard deviation of the enzymatic activity measured for various repeat experiments. Spearman indicates results for the non-parametric Spearman rank correlation coefficients. Analyses for data after natural logarithmic transformation are only indicated for parametric tests, non-parametric ones are not affected by transformations. The significances of the non parametric test, which does no assumption on the distribution of the data, are indicated to show that qualitatively, results do not depend on assuming normal distribution.

of stops is due to varying degrees of accuracy in the estimation of the enzymatic activity levels. Using regression analyses, I weighted data points proportionally to the number of replicate experiments used to estimate enzymatic activity. This increased the percentage of variation in enzymatic activity explained by the number of stops from 6.9% to 12.6%. It is interesting to note that a similar regression analysis that weights data points proportionally to the inverse of the standard deviation of the enzymatic activity level measured for the different replicates decreases the explanatory power of the stop hypothesis. This suggests that the variation in enzymatic activities measured for different replicates of the same experiment are related to the presence or absence of stops, and hence are part of the biological phenomenon, rather than an estimate of measurement accuracy. It indeed makes sense that the presence or absence of stops in the 50 leading region does not only affect the level of gene expression, but probably also regulates the natural variability of gene expression. 1.2. Is the 50 leading region also sometimes translated? According to the accepted understanding of translation, the results presented in the above section make little sense: stops, in any frame, should not affect translation if they are outside the protein coding sequence (here upstream of the recognized initiation codon). This is because, usually, nucleotide triplets in 50 leading regions do not function as ‘codons’. However, analyses in Tables 1 and 2 clearly show that stop motifs (to term triplets that are not in the coding frame of protein coding sequences, Antezana and Kreitman, 1999) increase translational efficiency. Several alternative mechanisms could explain this: (a) it is possible that ribosomes start sometimes translation before they reach the recognized start codon; (b) interactions between

stop motifs and ribosomes might affect translational efficiency, at least in the open reading frame (Jemiolo et al., 1995; Pisarev et al., 2006); (c) stop motifs in that region, for the matching DNA sequence, might enhance the binding efficiency of RNA polymerases, and increase mRNA production, and ultimately, gene expression. Evidence below supports (a), notwithstanding the possibility that other mechanisms might also contribute to the positive correlation observed between numbers of stop motifs in the 50 leading region and gene expression (Tables 1 and 2). 1.3. Experimental test of translation initiation in 50 leading regions The results from Tables 1 and 2 are suggestive that parts of the 50 leading region are sometimes translated, fitting the main hypothesis that start codons in the region usually defined as the 50 leading region upstream sometimes initiate translation. The data from Barrick et al. (1994) are also adequate for testing the above prediction. Enzymatic activities measured for 50 leading regions possessing a stop preceded by a start motif were averaged and compared, as control, to averaged enzymatic activities for 50 leading regions possessing a start preceded by a stop motif. Table 3 shows that in each of the three possible frames, the mean enzymatic activity for beta-galactosidase (Barrick et al., 1994) from artificial 50 leading regions that contain a start motif followed by a stop motif is lower than the average enzymatic activity from 50 leading regions with a stop preceding a start. A series of statistical tests were applied to these data, in order to test the hypothesis while integrating different types of relevant information, and hence increase the confidence in the result. First, a t-test between average activities from mRNAs with the two types of 50 leading regions (start preceding stop, versus stop preceding start) shows that, overall, there is a tendency for 50 leading regions with a start that precedes a stop to have lower translation than those where the stop precedes the start: for untransformed enzymatic activity levels, P ¼ 0.02; for natural logarithm transformed data, P ¼ 0.013 (one tailed t tests); for the non-parametric Mann–Whitney U test, P ¼ 0.017 (one tailed test). This tendency exists also separately for such comparisons in each frame, statistically significantly so for frame +1 (raw data, P ¼ 0.012; transformed data, P ¼ 0.014) and frame +2 (raw data, P ¼ 0.031; transformed data, P ¼ 0.048). In this case, the more robust but less sensitive Mann–Whitney test did not detect significant differences between enzymatic activity levels. However, combining the statistical significances of the three tests (Sokal and Rohlf, 1994) yields P ¼ 0.0036 for the raw data, P ¼ 0.0043 for transformed data, and P ¼ 0.033 for the non-parametric Mann–Whitney test. (Calculating combined statistical significances is only appropriate when the tests to be combined are independent from each other, which is the case here, because data for the same 50 leading regions did never fit into more than one

ARTICLE IN PRESS H. Seligmann / Journal of Theoretical Biology 249 (2007) 162–167

165

Table 3 Enzymatic activity as a function of 50 leading region coding properties Frame

Start followed by stop

Stop followed by start n

Mean

sd

0 1 2

167.54 25.00 16.60

134.41 1.70 25.03

All

114.18

128.03

4.5507 3.2177 1.7432 3.8553

n

Mean

sd

7 2 2

675.48 298.41 576.65

1158.40 271.56 380.97

11

485.93

694.13

Ln transformed 0 1 2 All

Differences P(t)

P(U)

6 8 4

0.167; 0.012; 0.031;

0.216 0.059 0.083

18

0.020;

0.013

6 8 4 18

0.132 0.014 0.042 0.017

Ln transformed 1.5246 0.0679 2.6143 1.7938

7 2 2 11

5.5247 4.9291 5.7461 5.3430

1.4451 1.7477 1.7766 1.5370

Comparison of mean enzymatic activity measured for mRNAs with 50 leading regions that contain a start preceding a stop, and those with a stop preceding a start. P(t) values are for one tailed t-tests, P(U) are for Mann Whitney non-parametric tests for inequality of medians; sd indicates standard deviation, n indicates the number of mRNAs with 50 leading regions fitting in a specific treatment group. Means, sd and tests for data after natural logarithmic transformation of the enzymatic activity are also indicated. This transformation adjusts the distribution of the data to a normal distribution. The significances of the non-parametric test, which does no assumption on the distribution of the data, are indicated to show that qualitatively, results do not depend on assuming normal distribution.

0

1.4. Start codons in 5 leading regions The tests in the above section suggest that usually untranslated sequences in the 50 leading region that possess a start followed by a stop codon (both in the same frame) are sometimes translated, and hence function as short open reading frames. For these reasons also, the presence of stops in any frame in the 50 leading region increases gene expression, because it stops ‘‘unprogrammed’’ protein synthesis that might have started more upstream, making the ribosomal machinery available for interacting again with the mRNA, likely this time resulting in the ‘‘programmed’’ protein synthesis. Analyses below explore whether start codons in the 50 leading region interact with the ribosome without specifically functioning as initiation codons by plotting the activity reported for mRNAs possessing varying numbers of starts in their 50 leading region as a function of the number of starts in that region (Fig. 1). These analyses exclude all cases where stop motifs

Frame 0

300

Frame 1 Frame 2

Enzymatic activity

of the groups used for the calculations in Table 3.). Note that the accuracy of the protein activity data varies for different 50 leading regions, because in some cases, four replicates were averaged, while in others, up to 12 replicates were used. Therefore, also statistical tests weighing enzymatic activity from 50 leading region variants proportionally to the number of replicates (as indicated in Table 3 of Barrick et al., 1994) were done. This procedure integrates into the analyses the level of accuracy of the various single data, and increases statistical significances (not shown) of all the unweighted tests reported in Table 3. This indicates that part of the variation that is not explained by the hypothesis is due to measurement error, and strengthens further the validity of the results, as suggested by previous analyses (Table 2).

200

100

0

0

2 1 Start motifs in 5' leading region

3

Fig. 1. Enzymatic activity as a function of the number of start codons in all frames in the 50 leading region of beta-galactosidase (185 artificially induced variants, data from Barrick et al. (1994) in E. coli. Data are averages for all variants with a given number of start codons. None of the differences between combinations of means is statistically significant. For ‘‘frame 1’’, there were no cases with 3 starts in the 50 leading region.

were present in the 50 leading region. For starts counted according to each frame, there were similar effects on enzymatic activity; none of the effects described is statistically significant. A single start in any frame in the 50 leading region ostensibly slightly increases enzymatic activity as compared to activity measured for mRNAs with 50 leading regions that do not possess any start. Additional starts, in any frame, ostensibly decrease activity. Although none of these differences are statistically significant, they weakly suggest interactions between starts in 50 leading regions and the transcription or translation machineries. The fact that results are similar for all frames, including frame ‘0’, is suggestive that the effects of starts in 50 leading

ARTICLE IN PRESS 166

H. Seligmann / Journal of Theoretical Biology 249 (2007) 162–167

regions are usually not due to expression of the extended normal reading frame (option (a) in above section), but rather suggest that starts interact with ribosomes or polymerases (options (b) and (c), respectively, in above section). These effects, if any, seem weak, but considering further information, such as secondary structure, could control for confounding effects and enhance the phenomenon enough to be detectable. Results were qualitatively similar for raw and transformed (not shown) data. 1.5. Regulation of random and normal translation initiation and termination These analyses suggest that protein expression is lower when the 50 leading regions contain a start that precedes a stop than when the stop precedes the start. This suggests that ribosomes detach themselves from the mRNA when they reach the stop that is in the 50 leading region, if they started and terminated translation in the 50 leading region before reaching the ‘real’ initiation codon. Hence short sequences possessing start and stop in the same frame seem to be occasionally translated. Presumably, trinucleotides that form stops in the 50 leading region should not be referred to as ‘codons’. According to a strict usage of terms, trinucleotides that are not in the coding frame or outside of the coding sequence are termed ‘motifs’ (Antezana and Kreitman, 1999). However, results suggest a probabilistic approach to limits of coding sequences, and that expression efficiency is increased if ribosomal activity after frameshifts is stopped early by off frame stop motifs. The results show that within the coding frame, translational activity is regulated by off frame stop motifs, as well as in any frame in the 50 leading region. However, it is important to stress here that the hypothesis does not claim to explain all the variation in enzymatic activity described by Barrick et al. (1994) for the various artificial variants in the 50 leading region. As indicated above, other factors also account for such variation, such as the secondary structure of the initiation region. Considerable variation in enzymatic activity exists even among mRNAs whose 50 leading region variants have the same basic ‘coding’ properties as those examined here (for example, the same number of stops, or the same relative positions of stops and starts). It is likely that taking into account the secondary structure of the 50 leading region independently explains significant additional proportions of the variation in measured enzymatic activities. Results on ‘codon’ properties of trinucleotides in the 50 leading region suggest a line of research that could yield new insights on the mechanisms by which ribosome and 50 leading region interact, how protein synthesis is initiated, and how it is terminated (Ivanov et al., 2001a, b). The probability of spontaneous, accidental frameshifts by ribosomes, as known at this point, is relatively low (1:30000 amino acids, Parker, 1989). This means that about 2% of the peptides produced for a 500 amino acid

long protein are affected by at least one frameshift instance. The off frame stop codon hypothesis suggests that off-frame stops would limit the costs due to these 2%. 1.6. Do randomly translated oligopeptides have functions? According to the results presented here, an unknown number of (presumably random) short peptides are translated from coding and non-coding regions of RNAs. One should consider the possibility that this putatively large pool of peptides might routinely play adaptive, as well as pathogenic roles in some normal biological functions, especially those involving ‘learning’ and ‘memory’ such as the immune system. 1.7. The structure of the genetic codes The observations reported above that specific codons, such as starts and stops, play roles outside of the boundaries coding for a protein, and of the open reading frame, are further evidence for the highly integrated nature of the various properties of the genetic code, such as the optimization of the code for maximizing both off frame stop frequencies and the potential to form secondary structures (Itzkovitz and Alon, 2007). Similarly, the apparent coevolution of polymerase error proneness for deletions (frameshifts) with stop codons suggests that similar adaptations exist with and for other sequences with specific ‘‘special’’ functions, such as, but not limited to, the typical bacterial promoter sequence AGGAGG, which is located immediately upstream of the leading sequence mutated by Barrick et al. (1994). It is likely that further analyses could find how the small variations among the different alternative genetic codes are optimized with their associated (archaean, bacterial, eukaryote, etc.) molecular transcriptional and translational machineries, leading to a fine-tuned, integrated understanding of the molecular setup of organisms. Acknowledgments I thank Shalev Itzkovitz for encouraging comments and discussion; Leigh VanValen and Jean-Luc Jestin for thorough reviews of the manuscript. References Antezana, M.A., Kreitman, M., 1999. The nonrandom location of synonymous codons suggests that reading frame-independent forces have patterned codon preferences. J. Mol. Evol. 49, 36–43. Barrick, D., Villanueba, K., Childs, J., Kalil, R., Schneider, T.D., Lawrence, C.E., Gold, L., Stormo, G.D., 1994. Quantitative analysis of ribosome binding sites in E.coli. Nucleic Acids Res. 22, 1287–1295. Gheysen, D., Iserentant, D., Derom, C., Fiers, W., 1982. Systematic alteration of the nucleotide sequence preceding the translation initiation codon and the effects on bacterial expression of the cloned sv40 small-t antigen gene. Gene 17, 55–63.

ARTICLE IN PRESS H. Seligmann / Journal of Theoretical Biology 249 (2007) 162–167 He, Y., Armanious, M.K., Thomas, M.J., Cress, W.D., 2000. Identification of E2F-3B, an alternative form of E2F-3 lacking a conserved N-terminal region. Oncogene 19, 3422–3433. Itzkovitz, S., Alon, U., 2007. The genetic code is nearly optimal for allowing additional information within protein-coding sequences. Genome Res. 17, 405–412. Irvin-Wilson, C.V., Chaudhuri, G., 2005. Alternative initiation and splicing in dicer gene expression in human breast cells. Breast Cancer Res. 7, R563–R569. Ivanov, V.I., Beniaminov, A.D., Mikheev, A.N., Minyat, E.E., 2001a. Why termination tRNAs have not been found? Because they are hidden in the large ribosomal RNA (a hypothesis). Mol. Biol. 35, 614–622. Ivanov, V.I., Beniaminov, A.D., Mikheev, A.N., Minyat, E.E., 2001b. A mechanism for stop codon recognition by the ribosome: a bioinformatic approach. RNA 7, 1683–1692. Jemiolo, D.K., Pagel, F.T., Murgola, E.J., 1995. UGA suppression by a mutant RNA of the large ribosomal subunit. Proc. Natl. Acad. Sci. USA 92, 12309–12313. Jestin, J.L., Kempf, A., 1997. Chain termination codons and polymeraseinduced frameshift mutations. FEBS Letters 419, 153–156. Kochetov, A.V., Sarai, A., Rogozin, I.B., Shumny, V.K., Kolchanov, N.A., 2005. The role of alternative translation start sites in the generation of human protein diversity. Mol. Genet. Genomics 273, 491–496. Kozak, M., 2002a. Emerging links between initiation of translation and human diseases. Mamm. Genome 13, 401–410. Kozak, M., 2002b. Pushing the limits of the scanning mechanism for initiation of translation. Gene 299, 1–34. Leissring, M.A., Farris, W., Wu, X.N., Christodoulou, D.C., Haigis, M.C., Guarente, L., Selkoe, D.J., 2004. Alternative translation initiation generates a novel isoform of insulin-degrading enzyme targeted to mitochondria. Biochem. J. 383, 439–446. Liu, J., Prolla, G., Rostagno, A., Chiarle, R., Feiner, H., Inghirami, G., 2000. Initiation of translation from a downstream in-frame AUG codon on BRCA1 can generate the novel isoform protein DELTABRCA1(17aa). Oncogene 19, 2767–2773. Markussen, F.H., Michon, A.M., Breitwieser, W., Ephrussi, A., 1995. Translational control of Oskar generates short OSK, the isoform that induces pole plasm assembly. Development 121, 3723–3732. Nagai, H., Yuzawa, H., Yura, T., 1991a. Regulation of the heat-shock response in Escherichia coli-involvement of positive and negative cisacting elements in translational control of sigma-32 synthesis. Biochimie 73, 1473–1479.

167

Nagai, H., Yuzawa, H., Yura, T., 1991b. Interplay of 2 cis-acting messenger-rna regions in translational control of sigma-32 synthesis during the heat-shock response of Escherichia coli. Proc. Natl. Acad. Sci. USA 88, 10515–10519. Outten, C.E., Culotta, V.C., 2004. Alternative start sites in the Saccharomyces cerevisiae GLR1 gene are responsible for mitochondrial and cytosolic isoforms of glutathione reductase. J. Biol. Chem. 279, 7785–7791. Ozbudak, E.M., Thattai, I., Kurtser, E., Grossman, A.D., Van Oudenaarden, A., 2002. Regulation of noise in the regulation of a single gene. Nat. Genet. 31, 69–73. Parker, J., 1989. Errors and alternatives in reading the genetic code. Microbiol. Rev. 53, 273–298. Pisarev, A.V., Kolupaeva, V.G., Pisareva, V.P., Merrick, W.C., Hellen, C.U.T., Pestova, T.V., 2006. Specific functional interactions of nucleotides at key ()3 and (+)4 positions flanking the initiation codon with components of the mammalian 48S translation initiation complex. Genes Dev. 20, 624–636. Rhee, S., Yang, S.J., Lee, S.J., Park, D., 2004. beta Pix-b(L), a novel isoform of beta Pix, is generated by alternative translation. Biochem. Biophys. Res. Commun. 318, 415–421. Satchidanandam, V., Shivashankar, Y., 1997. Availability of a second upstream AUG can completely overcome inhibition of protein synthesis initiation engendered by mRNA secondary structure encompassing the start codon. Gene 196, 231–237. Seligmann, H., Pollock, D.D., 2004. The ambush hypothesis: hidden stop codons prevent off frame gene reading. DNA Cell Biol. 23, 701–705. Slack, K.E., Janke, A., Penny, D., Arnason, U., 2003. Two new avian mitochondrial genomes (penguin and goose) and a summary of bird and reptile mitogenomic features. Gene 302, 43–52. Sokal, R.R., Rohlf, F.J., 1994. Biometry. 3rd ed. Freeman and Company, New York. Tucker, P., Laemle, L., Munson, A., Kanekar, S., Oliver, E.R., Brown, N., Schlecht, H., Vetter, M., Glaser, T., 2001. The eyeless mouse mutation (ey1) removes an alternative start codon from the Rx/rax homeobox gene. Genesis 31, 43–53. Webb, E., Cox, J., Edwards, S., 2005. Cervical cancer-causing human papillomaviruses have an alternative initiation site for the L1 protein. Virus Genes 30, 31–35. Yamasaki, K., Weihl, C.C., Roos, R.P., 1999. Alternative translation initiation of Theiler’s murine encephalomyelitis virus. J. Virol. 73, 8519–8526.