Codon usage regulates human KRAS expression at both transcriptional and translational levels

KRAS and HRAS are highly homologous oncogenic Ras GTPase family members that are mutated in a wide spectrum of human cancers. Despite having high amino acid identity, KRAS and HRAS have very different codon usage biases: the HRAS gene contains many common codons, and KRAS is enriched for rare codons. Rare codons in KRAS suppress its protein expression, which has been shown to affect both normal and cancer biology in mammals. Here, using HRAS or KRAS expression in different human cell lines and in vitro transcription and translation assays, we show that KRAS rare codons inhibit both translation efficiency and transcription and that the contribution of these two processes varies among different cell lines. We observed that codon usage regulates mRNA translation efficiency such that WT KRAS mRNA is poorly translated. On the other hand, common codons increased transcriptional rates by promoting activating histone modifications and recruitment of transcriptional coactivators. Finally, we found that codon usage also influences KRAS protein conformation, likely because of its effect on co-translational protein folding. Together, our results reveal that codon usage has multidimensional effects on protein expression, ranging from effects on transcription to protein folding in human cells.

The RAS family of small GTPases, comprised of KRAS, HRAS, and NRAS, are important for normal development and in cancer (1)(2)(3)(4). With regards to the latter, these genes are collectively mutated in one-third of all human cancers (5). The encoded proteins share ϳ85% homology, with the primary differences lying in the last ϳ23 amino acids (6). Despite this high homology, multiple lines of evidence suggest that KRAS is expressed at low levels, and moreover, that this low expression is biologically critical to the function of the gene. Namely, KRAS mRNA levels are typically the lowest of the RAS isoforms in human tissues (7) (and see below). Furthermore, increasing the expression of the endogenous murine Kras gene results in hyperproliferation of their hematopoetic stem cells and renders the mice more resistant to a carcinogen that induces Kras-mutant lung tumors (8,9). The importance of the low expression of KRAS cannot be overstated, as this gene is essential (10), unlike NRAS and HRAS, and is mutated in one quarter of all human cancers, far more than the other two RAS genes (4). In addition, overexpression or amplification of KRAS, but not mutations in the coding sequence, is also associated with certain types of cancer (11,12). As such, elucidating the mechanism by which KRAS expression is kept low is critical to our understanding of normal and cancer biology.
Most amino acids are encoded by two to six synonymous genetic codons. Synonymous codons are used with different frequencies in all organisms, and every organism has its own preferred codon usage bias (13)(14)(15)(16). Codon usage bias has been shown to positively correlate with tRNA abundance, thus optimal codons were thought to be translated more efficiently and more accurately (16 -20). Consistent with this, strong codon usage biases have been shown to be important for the expression of highly expressed genes in different organisms, and codon optimization has been widely used to enhance heterologous protein expression (21)(22)(23)(24)(25). Therefore, codon usage can be an important determinant in gene expression. In addition, codon usage has been shown to influence the rate of translation elongation and protein structure by affecting the co-translational folding process in Escherichia coli, fungi, and insects (26 -35). In addition to its role in regulating protein translation, codon usage also has a major role in determining the level of gene expression through transcriptional and post-transcriptional processes (24, 36 -38). As such, gene codon usage has been proposed to be a code within the genetic code that can determine both gene expression levels and protein structures and therefore activity (24, 26).
Interestingly, whereas the RAS proteins share high amino acid identity, their codon usage varies widely. Human codon usage has biases for C/G at the wobble positions. HRAS is enriched in optimal codons, KRAS is enriched in rare codons and has a low codon bias score, whereas NRAS has a codon usage profile between these two (39). With regards to KRAS, we previously found that changing rare codons to common in the ectopic and/or endogenous human and/or murine KRAS increased KRAS mRNA, translation, and protein levels (39). Moreover, such changes had in vivo phenotype, namely altering the proliferation of hematopoetic stem cells or affecting various stages of tumorigenesis (9,39,40). However, the contribution of codon bias to the various steps in KRAS protein production had not been systematically examined. We now report that changing rare codons to common in KRAS increased translation efficiency and mRNA levels. Regulation of mRNA levels is a major mechanism affecting KRAS levels, but the effect was not a consequence of mRNA stability, but instead transcriptional. Moreover, codon usage also had an impact on the structure of KRAS. Thus, the rare codon bias of KRAS affects multiple aspects of gene expression process, which sheds important insights into the codon usage-mediated gene regulatory mechanisms for mammalian genes.

Codon usage strongly affects expression of KRAS
Calculation of the average codon adaptation index (CAI) 2 of human genes showed that the median CAI of human genes is 0.83. The average CAI of human HRAS is 0.87, whereas KRAS, which uses many rare codons, has an average CAI of 0.69 (39) (Fig. S1A). For example, GTG and ATC are the most preferred codons for valine and isoleucine, respectively, in the human genome. Although these optimal codons are overwhelmingly used in HRAS, they are rarely used in KRAS (Fig. S1B). Similar to our previously reported (39) data, transient transfection of an N terminally FLAG-tagged WT HRAS or KRAS cDNA (referred to as Hras and Kras constructs, respectively) in human embryonic kidney HEK-293T cells resulted in highly divergent protein expression (Fig. 1A). However, as noted above, codon bias can affect the entire process of protein production, and hence we decided to dissect the contribution of each to the ultimate end product of a functional protein. To this end, we first needed to benchmark the effect of the rare codons on KRAS expression. We thus chose three previously created versions of human KRAS cDNA for this analysis (39). The WT KRAS was used as the control for native codon usage. In opKRAS, the rarest valine codon (GTA) was changed to the most common (GTG), but isoleucine was changed from ATA to ATC. For KRAS*, the HRAS cDNA was mutated to encode for the KRAS protein (Fig.  1B). The nucleotide sequences of different codon-optimized KRAS cDNAs are shown in the Fig. S2. As expected, transient transfection of vectors encoding each of these three cDNAs into human 293T cells yielded a stepwise increase in protein levels. Codon optimization of only 18 codons in opKRAS resulted in an ϳ10-fold increase in KRAS protein level compared with the WT KRAS construct (Fig. 1C). Remarkably, Kras* produced about 100-fold more KRAS over the control ( Fig. 1C and Fig. S3C), which is similar to the difference between proteins produced from Hras and Kras cDNAs (Fig.  1A). Co-transfection of these KRAS constructs individually with an eGFP-expression construct showed that transfection efficiencies were similar for each construct (Fig. S3, A and B).
We note here that changing progressively more rare codons to common in KRAS increases KRAS protein levels, but the effect of codon usage within specific regions of KRAS had not been explored. Thus, we further manipulated the opKras sequence, generating three partially optimized versions called N-opKras, M-opKras, and C-opKras, in which only the valine and isoleucine codons encoding the N-terminal, central, or C-terminal regions were optimized ( Fig. 1B and Fig. S2). These three versions of opKras expressed KRAS protein at levels intermediate between those of the Kras and opKras constructs (Fig. 1D). We also mutated the opKras construct so that only nine rare isoleucine or nine rare valine codons (opKras-I and opKras-V) were optimized. Both opKras-I and opKras-V resulted in KRAS protein levels that were higher than those produced from the WT Kras but lower than those from opKras (Fig. 1E). Thus, the three versions of KRAS exhibited the expected effect on protein production and different synonymous codons have independent and accumulative effects on KRAS expression.

Rare codons suppress KRAS mRNA translation
Codon usage has been shown to regulate translation elongation rate and co-translational protein folding (26, 29, 41) and has been proposed to influence translation efficiency and accuracy (16 -20, 42). In agreement, we previously reported that changing rare codons to common increased KRAS mRNA in the polysome fractions (39). To determine the effect of codon usage on KRAS translation, we synthesized different versions of KRAS mRNA by in vitro transcription; mRNAs had 5Ј caps and poly(A) tails. Equal amounts of KRAS mRNA and the control luciferase mRNA were co-transfected into 293T cells and the amount of KRAS protein produced was determined. The amount of protein produced from opKRAS and KRAS* mRNAs was about 0.5-and about 4-fold higher, respectively, than that produced from the WT KRAS mRNA ( Fig. 2A). In contrast, the levels of luciferase protein were comparable (Fig. S5A). Comparison of the protein decay rates after the addition of the protein synthesis inhibitor cycloheximide revealed that codon optimization in opKRAS and KRAS* did not affect KRAS protein stability (Fig. S4). We note that the differences in protein levels using KRAS mRNA were much less than those observed in Fig. 1C, suggesting that additional mechanism(s) mediate the codon usage effect on protein production.
To further confirm the effect of codon usage on translation, we performed in vitro translation assays using 293T cell lysates. Similar fold-differences in KRAS protein production were observed for the opKRAS and KRAS* mRNAs (Fig. 2B) as were observed in the above cell-based assay. Interestingly, when the same mRNAs were translated in an extract made from budding yeast, an organism with A/T-biased codons, the trend in expression pattern was opposite. The highest level of KRAS was synthesized from the WT KRAS mRNA ( Fig. 2C and Fig. S5B). Together, these results suggest that rare codons in KRAS suppress KRAS translation in human cells. Furthermore, we generated stably transfected HEK-293T cell lines with Kras and with Kras* expression constructs and performed polysome profiling (Fig. S5C). Northern blot analysis of the polysomal fractions showed that the WT Kras mRNA was peaked in the monosome fraction (fraction number 8), whereas Kras* mRNA was enriched in the polysome fractions (fractions [15][16][17][18][19] (Fig.  S5C). Again, these data indicate that optimal codons of KRAS promote efficient mRNA translation in cells. However, we estimate a ϳ4-fold increase in translation (Fig. 2, A and B), which when benchmarked against the very high amount of protein produced by KRAS* suggest that translation is only one aspect accounting for the effect of rare codons on KRAS protein levels.

KRAS codon optimization increases mRNA level but does not affect mRNA stability
As noted above, multiple experiments support that rare codons impact KRAS protein production beyond translation. We had previously reported that ectopic KRAS* only generated an modest increase of mRNA compared with WT KRAS, as assessed by quantitative RT-PCR (39). To more accurately measure the effect of codon usage of KRAS mRNA levels we utilized Northern hybridization analysis using a probe that anneals to the common 5Ј UTR of all three ectopic KRAS mRNAs. In cells that were transfected with different codon-optimized versions of KRAS cDNA expression constructs the codon optimization resulted in about 4-and 10-fold increases in mRNA levels for opKRAS and KRAS*, respectively, compared with WT KRAS mRNA levels (Fig. 3A). These results suggest that the effect of codon usage on KRAS mRNA expression is more pronounced than first thought and plays a major role in regulating KRAS expression.
Codon usage was previously shown to affect mRNA levels by influencing mRNA stability in different organisms (27, 38,  [43][44][45][46]. To determine whether the effect of codon usage on KRAS mRNA levels is due to its effects on mRNA stability, we compared mRNA decay rates of WT and codon-optimized mRNAs after the addition of ␣-amanitin, a commonly used transcription inhibitor. Northern blotting quantifications revealed that there were no significant differences in mRNA stability (Fig. 3B). A similar result was also obtained when another transcription inhibitor, actinomycin D, was used to determine mRNA decay rates (Fig. S6). These results indicate that in addition to its effect on translation efficiency, KRAS codon usage also has a major role in mRNA levels without overtly affecting mRNA stability. Consistent with this conclusion, it was previously shown that mammalian genes with high GC contents, which is associated with more preferred codons, have higher expression levels than those with lower GC content without affecting mRNA degradation rates (37, 47,48).

Regulation of KRAS expression by codon usage
Using the RNA-seq results from the Genotype-Tissue Expression (GTEx) Portal, we compared the relative RNA levels of KRAS and HRAS in different human tissues. As shown in Fig.  3C, across all tissues examined, the KRAS mRNA levels are much lower than those of HRAS mRNA. This result is consistent with the notion that regulation of the mRNA level is a major mechanism that suppresses KRAS expression in human cells.

KRAS codon usage regulates transcription and chromatin modifications
The increase in KRAS transcript levels by codon optimization prompted us to examine KRAS transcription. Using human 293T cell lines stably transfected with a vector encoding WT KRAS, opKRAS, or KRAS*, we performed Br-UTP-coupled nuclear run-on assays to examine transcriptional by RNA polymerase II (Pol II) at the locus of interest. Because this assay quantifies the frequency of transcription initiation, the levels of newly synthesized transcripts should not be influenced by RNA stability. Codon optimization resulted in significant increases of KRAS transcription with about 10-fold higher levels of transcript from Kras* than WT Kras (Fig. 4A). Quantification of KRAS transgene copy numbers in these stably transfected cell lines indicated that they have similar transgene copy numbers.
After transcriptional initiation, the Pol II carboxyl-terminal domain (CTD) is phosphorylated at serines 2 and 5 (49, 50). Ser-5 phosphorylation of the CTD tail occurs soon after initiation, whereas Ser-2 phosphorylation of the CTD of Pol II takes place during the transcriptional elongation process. To confirm the effect of codon usage on KRAS transcription, we compared the enrichment of phosphorylated Pol II CTD on the WT and codon-optimized KRAS by chromatin immunoprecipitation (ChIP) assays. Codon optimization resulted in a significant enrichment of both Ser-2 and Ser-5 phosphorylated Pol II at the

Regulation of KRAS expression by codon usage
KRAS transgene loci (Fig. 4B). These results further indicate that codon usage impacts KRAS transcription.
To determine the mechanism by which codon usage affect KRAS transcription, we first performed histone H3 ChIP. The occupancies of histone H3 at the KRAS transgene loci showed no significant differences among three stable cell lines (Fig. 4C), suggesting that codon usage does not influence nucleosome density. We then performed ChIP assays for several histone modification marks associated with transcriptionally active chromatin. Both H3K4 trimethylation and H3K9 acetylation enrichments at the opKRAS and KRAS* was significantly increased compared with the WT KRAS transgene locus (Fig.  4D), consistent with mRNA level differences. p300 is the major histone acetyltransferase that mediates H3K9 acetylation. ChIP assays showed that the enrichment of p300 at the transgene loci was significantly higher for opKRAS cells than for the WT KRAS cells and was further increased for the KRAS* cells (Fig.  4E). These results indicate that the KRAS codon usage impacts

Regulation of KRAS expression by codon usage
transcription by affecting histone modifications and chromatin structure. Optimal codons may result in transcriptionally permissive chromatin structures to promote recruitment of transcription co-activators, such as p300.
To determine whether the effect of codon usage on transcription is a general phenomenon or is KRAS-specific, we examined the effect of codon usage on CFL2 expression. CFL2 encodes an intracellular protein that is a major component of intranuclear and cytoplasmic actin rods. Mutation of CFL2 results in human nemaline myopathy. Aside from its importance in human disease, we previously reported that CFL2 is enriched in rare codons, and that changing rare codons to common increases the amount of ectopic CFL2 protein (39). We confirmed that CFL2 protein expression is indeed greatly enhanced after codon optimization (Fig. 4F). Similar to KRAS, codon optimization also led to a 6-fold increase of CFL2 mRNA levels (Fig. 4G). ChIP assays were performed to examine the enrichment of Pol II Ser-2 and Ser-5 phosphorylation and H3K9 acetylation at the transgene loci in cells that stably expressed either the WT CFL2 or codon-optimized CFL2 (Fig.  4H). As expected, codon optimization resulted in a significant increase of enrichment of all three markers. Collectively, we conclude that codon usage also affects KRAS transcription and mRNA levels, an effect that may be applicable to other human genes enriched in rare codons.

The differential effects of codon usage on KRAS expression in different cell lines
We had previously demonstrated that ectopic expression of KRAS* generated more protein than KRAS in a variety of cell types (9, 39, 40). However, the contribution of changing rare codons to common on KRAS mRNA levels in different cells,

Regulation of KRAS expression by codon usage
especially in light of the above results, had not been undertaken.
To determine whether the effect of codon usage on KRAS mRNA levels is cell line-specific, we transfected the WT KRAS, opKRAS, and KRAS* expression constructs into two human hepatocellular carcinoma cell lines, Huh7 and HepG2, and two human breast cancer cell lines MDA-MB-231 and MCF-7. Codon optimization resulted in increases of both KRAS protein and mRNA in each of these cell lines (Fig. 5, A-D, and Fig. S7). As in 293T cells, the fold-changes of mRNA levels due to codon optimization in these cell lines were smaller than those of protein levels, suggesting that the effect of codon usage on translation efficiency is shared among these cell lines. However, it is clear that different cell lines responded differently to codon optimization. In HepG2 cells, the effects of codon usage on KRAS protein and RNA were much smaller than those in the other cell lines. Less than 10-fold more KRAS protein and less than 50% more mRNA were produced from KRAS* than from WT KRAS. In contrast, in Huh7 cells the differences were larger than those seen in HEK-293T cells. In addition, the changes of KRAS transcript levels showed a strong positive correlation with the changes in protein levels in different cell lines (coefficient r ϭ 0.93, Fig. 5E), indicating a major role for KRAS mRNA in determining KRAS protein levels. This suggests the intriguing possibility that the effect on codon bias at individual steps of protein production may be differentially regulated in different tissues. It should be noted that tRNA content can also be different in these cell lines which may affect translation efficiency (51).

Codon optimization alters KRAS protein structure
We and others have previously shown that codon usage affects the translation elongation rate in fungi and flies (26 -28), which in turn can influence protein structure during the cotranslational folding process. Codon usage has been shown to regulate protein folding in vitro and in E. coli, fungi, and Drosophila cells (26, 27, 29 -35). To determine whether codon usage influences protein folding in mammals, we compared KRAS protein structures by performing a limited trypsin digestion assay, which can indicate protein structural differences. Cell extracts of 293T cells transfected with the WT KRAS or KRAS* expression constructs were used for trypsin digestion assays. In the presence of the same concentration of trypsin, KRAS protein in the WT KRAS cell extract was much more resistant to trypsin digestion than that expressed from KRAS* (Fig. 6A).
To confirm our conclusion, we carried out thermal shift assays using extracts of 293T cells (52,53). This assay quantifies changes in thermal denaturation and aggregation temperature of a protein as a result of treatment by increasing temperatures, which results in the disappearance of the protein from the

Regulation of KRAS expression by codon usage
supernatant. Changes in denaturation and aggregation temperature are indicative of structural changes of a protein. Increasing temperatures from 4 to 52.6°C resulted in a gradual slow disappearance of KRAS from the WT KRAS extract. In contract, most of the KRAS from the KRAS* extract was precipitated and disappeared from the supernatant above 40 ºC (Fig.  6B). Together, these results suggested that codon optimization of KRAS alters structural properties of KRAS proteins.

Discussion
Unlike HRAS and NRAS, KRAS is an essential gene (10). However, the expression level of KRAS is much lower than that of HRAS in all tissues examined (Fig. 3C). As stated above, numerous experiments support the conclusion that the low level of KRAS is important for how the gene functions in normal and cancer biology of mammals, including in whole animal settings. Thus, the expression of KRas is likely normally suppressed to prevent cancer in normal cells. As such, understanding how KRAS expression is maintained at a low level is critical. One feature of this gene that contributes to the poor expression of KRAS is due to its poor codon usage. Consistent with our previous observations, we show here that codon usage has a remarkable effect on KRAS expression from cDNA constructs. Codon optimization of KRAS resulted in the up-regulation of KRAS by about 100-fold (Fig. 1). In addition, the effect of codon usage was accumulative. The number of codons optimized was correlated with levels of KRAS, and effects were not restricted to a specific region of KRAS ORF. These results suggest that the different codon usage profiles of KRAS and HRAS are the primary reason for their different protein levels. In agreement with our previous observations, this increase was attributed to increased translation and mRNA levels. At the translational level, optimal codons promote efficient translation of KRAS mRNA. This conclusion is supported by analyses of translation of KRAS mRNA constructs in cells and in vitro (Fig. 2, A and B) and by polysome profiling results that showed that codon optimization of KRAS led to the enrichment of the KRAS mRNA in the highly translated polysome fraction. We have previously shown that optimal codons are known to increase the rate of translation elongation, and rare codons can result in ribosome stalling in fungi and fly cells (26 -28). Our data suggest that codon usage has a similar effect in human cells.
Codon usage also determines KRAS mRNA levels. However, unlike in yeast and some other organisms (27, 38, 44 -46), codon usage did not have a significant influence on KRAS mRNA stability in our experiments in human cells (Fig. 3B and  Fig. S6). Instead, we found that optimal codon usage promotes KRAS transcription. This conclusion is supported by nuclear run-on and Pol II ChIP assay results (Fig. 4, A and B). In addition, we showed that codon optimization resulted in increases of H3K4me3 and H3K9ac, histone modifications that are associated with active chromatin (Fig. 4D). Furthermore, we found that codon optimization led to enrichment of histone acetyltransferase p300 at the KRAS locus. Similar effects of codon optimization were also observed for the CFL2 gene (Fig. 4,  F-H), indicating that the transcriptional effect of codon usage may be a general phenomenon for human genes. Together, these results suggest that optimal codon usage affects transcrip-tion by recruiting co-transcriptional activators such as p300, which lead to chromatin modifications that alter chromatin structure and activate transcription. Preferred codons may enhance exonic transcription factor binding and thus affecting chromatin structure and transcription (54). Some of the experiments in this study were performed transiently in plasmid transfection. It should be noted that it was previously shown that transiently transfected plasmid DNA does form chromatin with histone proteins in cells (55,56).
Consistent with our conclusion here, mammalian genes with high GC content, which is associated with more common codons, had higher expression levels than those with lower GC content without an effect on mRNA degradation rates (47,48). In addition, codon usage was shown to contribute to the balanced expression of Toll-like receptors in mammals through effects on transcription (37). Codon usage was previously shown to have a major role at the transcriptional level in Neurospora through regulation of chromatin structures (24). Therefore, a role of codon usage on transcription appears to be conserved from fungi to human. How codon usage affects chromatin structures is not known. Codon information within the ORF may be read by the transcription machinery in the form of DNA elements that favor or inhibit the recruitment of regulatory proteins that can regulate chromatin structure, resulting in suppression or activation of transcription. It should be noted that our experiment here was performed with ectopic expression from cDNA transgenes, which may have different effects from the endogenous locus. However, we have previously shown that KRAS codon optimization at the endogenous locus resulted in elevated KRAS expression (39).
Although a positive role for codon optimization on KRAS expression was observed in different cell lines, the degree of the codon usage effect differed (Fig. 5). Such differential codon usage effects may be caused by different tRNA expression profiles in different cell lines (51), which are known to influence translation. In addition, the effects may also be due to differential expression levels of the chromatin regulatory factors that mediate the transcriptional effect of codon usage.
Finally, our results show that KRAS codon usage may also affect KRAS protein structure. We and others have previously shown that codon usage regulates translation elongation in fungi and Drosophila (26 -28). Changes in elongation rate change the time available for co-translational folding thus influencing protein structures. Consistent with this, codon usage has been shown to regulate protein folding in vitro and in E. coli, fungi, and Drosophila cells (26, 27, 29 -35). Our results suggest that this is also the case in human cells. Consistent with this conclusion, a single synonymous SNP that results in a rare codon in the human MDR1 gene, which encodes a transporter, was found to result in altered drug and inhibitor interactions (57). Furthermore, codon usage has been implicated in cotranslational protein folding of cystic fibrosis transmembrane conductance regulator (CFTR), a protein that regulates transmembrane conductance, which is mutated in cystic fibrosis patients (58 -60). Taken together, our results demonstrate that codon usage influences gene expression and protein structure in human cells by multiple mechanisms. Because many human diseases are known to be associated with synonymous muta-

Regulation of KRAS expression by codon usage
tions (11,61,62), our study suggests how these mutations can contribute to disease progression without altering amino acid sequences.

Cell lines
HEK-293T, HepG2, MDA-MB-231, and MCF7 cells were maintained at 37°C in 5% CO 2 in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal bovine serum (Sigma) and 100 units/ml of penicillin and streptomycin. Huh7 cells were maintained in McCoy's 5A media (Gibco/Invitrogen) with identical supplements. For expression studies, cells were either transiently transfected with plasmids using PolyJet (SignaGen) according to the manufacturer's instructions or selected for stable expression of constructs by puromycin or neomycin antibiotic resistance. To evaluate transfection efficiency, an eGFP expression construct was co-transfected with the target plasmids. The levels of eGFP were determined by Western blotting and fluorescence microscope analyses (Fig. S3).
Plasmids pBabe-Kras, -opKras, and -Kras* expression constructs were created previously (39). The other constructs used in this study were created based on these plasmids. The cDNA sequences of different codon-optimized KRAS constructs generated in this study are shown in Fig. S2.

RNA analysis
RNA was extracted with TRIzol (Ambion) in accordance with the manufacturer's protocol. For Northern blotting analyses, equal amounts of total RNA (5 g) were loaded onto agarose gels. After electrophoresis, the RNA was transferred onto the nitrocellulose membrane (GVS North America). The membrane was probed with an RNA probe specific for 5Ј UTR of the KRAS mRNAs. The probe was labeled with [ 32 P]UTP (PerkinElmer Life Technologies) during transcription by T7 RNA polymerase (Ambion) with the manufacturer's protocol. The primer sequences used for the template amplification were Northern forward: 5Ј-CCTTAGGTCACTGGAAAGATG-3Ј, and Northern reverse: 5Ј-TAATACGACTCACTATAG-GGGTCGTCATCGTCTTTGTAGTC-3Ј.
For the RNA stability assay, cells were grown and transfected with the indicated plasmids for 2 days before the addition of actinomycin D (final concentration 10 g/ml) or ␣-amanitin (final concentration 50 g/ml), and collected at the indicated time points. After incubation at 30°C for 30 min, 6 l of stop buffer and 60 units of RNase-free DNase I were added. RNAs were isolated using TRIzol. To isolate the newly synthesized RNA, Protein G beads were added, and incubated for 2 h at 37°C. Beads were washed and RNAs were extracted using TRIzol. Newly synthesized mRNA levels were measured by quantitative real-time PCR analysis. The primer set located in the 5Ј-UTR that is shared among three KRAS transgene genes was used for quantitative RT-PCR: Kras forward: 5Ј-AGCCCTTTGTACACCCTAA-3Ј, reverse: 5Ј-GTCGTCAT-CGTCTTTGTAGTC-3Ј. The results were normalized relative to GAPDH mRNA expression levels in each sample and further normalized to mRNA levels at the endogenous Kras locus.

ChIP assay
Cells were fixed with 1% formaldehyde (Sigma) for 15 min at room temperature with shaking. Glycine (Sigma) was then added to a final concentration of 125 mM. The cross-linked cells were collected and prepared using lysis buffer (50 mM Tris, pH 8.1, 10 mM EDTA, 1% SDS, Roche complete protease inhibitor (EDTA-free)) with sonication. Equal amounts of protein were used for each immunoprecipitation reaction. Antibodies against histone H3 (ab1791), the RNA Pol II C-terminal domain (phospho-S2; ab5095), the RNA Pol II C-terminal domain (phospho-S5; ab5131), and histone H3 (trimethyl-K4, ab8580) were purchased from Abcam. Antibody against histone H3 acetyl-Lys-9 (39917) was purchased from Active Motif. Antibody against p300 (sc-48343) was purchased from Santa Cruz Biotechnology. The ChIP reaction was carried out with 2 l of antibody. Immunoprecipitated DNA was enriched using GammaBind G-Sepharose beads (GE Healthcare) and eluted using elution buffer. Purified DNA was quantified by real-time quantitative PCR. The ChIP results were first normalized by the ratio of ChIP to input DNA, then normalized to the gapdh DNA levels in each sample, and further normalized to the result from ChIP of the endogenous Kras locus.

In vitro transcription
To prepare the templates for in vitro transcription, the plasmids were linearized by NheI followed by successive phenolchloroform extraction and ethanol precipitation. The capped and poly(A)-tailed mRNA transcripts were synthesized using a HiScribe T7 quick high yield RNA synthesis kit (New England Biolabs) supplemented with 3Ј-o-Me-m7G(5Ј)ppp(5Ј)G antireverse cap structure analog (New England Biolabs) following the manufacturer's instructions. The mRNA concentrations were measured using a Nanodrop (Thermo Scientific).

In vitro translation using mammalian HEK-293T and Saccharomyces cerevisiae cell-free lysates
To prepare HEK-293T cell-free lysate, HEK-293T cells were harvested by centrifugation at 1000 ϫ g for 4 min and washed with PBS three times. Cell pellets were resuspended in 2 volumes of hypotonic buffer (10 mM HEPES-KOH, pH 7.6, 10 mM potassium acetate, 0.5 mM magnesium acetate, 5 mM DTT), and incubated on ice for 40 min to 1 h. Cells were then homogenized by 20 -30 strokes in a Dounce homogenizer on ice, and the final concentration of potassium acetate was adjusted to 50 mM. The cell extract was centrifuged at 16,000 ϫ g for 10 min at 4°C. The supernatant was aliquoted, snap frozen in liquid nitrogen, and stored at Ϫ80°C before use. To perform translation assay, 3 l of reaction mixture (20 mM HEPES-KOH, pH 7.6, 0.5 mM spermidine, 8 mM creatine phosphate, 0.2 mM GTP, 1 mM ATP, 20 M complete amino acids (Promega), 100 mM potassium acetate, 1 mM magnesium acetate, 0.13 units/l of creatine phosphate kinase, 0.2 units/l of SUPERaseIn RNase Inhibitor (Invitrogen)), 1 l of the mRNA template (180 ng), and 8 l of cell-free translation extract were used in each reaction. The reactions were incubated in a 30°C water bath for 30 min and stopped by adding SDS sample buffer, followed immediately by heating at 90°C. The samples were subsequently analyzed by Western blotting.
To prepare S. cerevisiae cell-free lysate, cells were harvested by centrifugation at 4°C for 5 min at 3,000 rpm, and resuspended in 1.5 ml of buffer A (30 mM HEPES-KOH, pH 7.6, 100 mM potassium acetate, 3 mM magnesium acetate, 2 mM DTT) with 8.5% mannitol and 0.5 mM PMSF/g of cell weight (64). Lysate were centrifuged at 4°C for 6 min at 18,000 rpm and supernatant was collected. Small molecular weight molecules are removed from the extract using Zeba Desalt Spin Columns (Pierce). Aliquots (200 l) are pipetted into 1.6-ml Eppendorf tubes, frozen with liquid nitrogen, and stored at Ϫ80°C. To perform translation assay, 7 l of translation reaction mixture (5 l of cell lysate with 1 l of 10ϫ energy mixture, 0.06 l of 10 units of creatine phosphate kinase, 0.5 l of 2 M KOAc, 0.12 l of 0.1 M Mg(OAc) 2 , 0.1 l of 1 mM amino acids mixture, and 0.1 l of SUPERase In RNase inhibitor (Life Technologies), and 0.12 l of RNase-free water), and 3 l of the mRNA template (60 ng) was used in each reaction. The reactions were incubated in a 26°C water bath for 15 min and stopped by adding SDS sample buffer, followed immediately by heating at 90°C. The samples were subsequently analyzed by Western blotting.

Polysome profiling
HEK-293T cells stably expressing Kras and Kras* were suspended in 425 l of hypotonic buffer (5 mM Tris-HCl, pH 7.5, 2.5 mM MgCl 2 , 1.5 mM KCl, Roche complete protease inhibitor (EDTA-free)) and 5 l of 10 mg/ml of CHX, 1 l of 1 M DTT, 100 units of RNasin were added, and the samples were vortexed. After 5 min, 25 l of 10% Triton X-100 and 25 l of 10% sodium deoxycholate were added, and samples were centrifuged at 13,000 ϫ g for 10 min at 4°C. The collected supernatants were then loaded onto a sucrose gradient prepared in 200 mM HEPES, pH 7.6, 1 M KCl, 50 mM MgCl 2 , 100 g/ml of CHX, Roche complete protease inhibitor (EDTA-free), and 100 units of RNasin and centrifuged at 35,000 ϫ g for 2 h at 4°C. Fractions were collected, and absorbance at 254 nm was monitored to obtain the polysome profiles. RNA samples were isolated from individual fractions using the TRIzol reagent (Invitrogen) and resolved by Northern blotting.

Trypsin sensitivity assay
Protein extracts were diluted to a total protein concentration of 2.5 g/l. A 100-l aliquot of extract was treated with trypsin at room temperature with gentle shaking. A 20-l sample was taken from the reaction at each time point (0, 5, 15, and 30 min) after addition of trypsin. Each 20-l sample was mixed with protein loading buffer, and proteins were resolved on an SDS-PAGE gel (12.5%). Western blotting was performed to examine KRAS protein levels at each time point as previously described (29). Experiments were performed side by side and protein samples were transferred to the same membrane for Western blot analysis.

Thermal shift assay
Protein extract was diluted to a total protein concentration of 2.5 g/l. Aliquots of 18 l of extract was heated for 2 min at different temperatures (C1000 Thermal Cycler PCR machine, Bio-Rad) followed by cooling for 3 min at room temperature. The lysates were then centrifuged at 15,000 ϫ g for 20 min at 4°C to separate the soluble fractions from precipitates. The amount of KRAS protein in the supernatants was then analyzed by Western blot analysis. Protein samples were transferred to the same membrane for Western blot analysis.