Non-optimal codon usage affects expression, structure and function of clock protein FRQ

Codon-usage bias has been observed in almost all genomes and is thought to result from selection for efficient and accurate translation of highly expressed genes. Codon usage is also implicated in the control of transcription, splicing and RNA structure. Many genes exhibit little codon-usage bias, which is thought to reflect a lack of selection for messenger RNA translation. Alternatively, however, non-optimal codon usage may be of biological importance. The rhythmic expression and the proper function of the Neurospora FREQUENCY (FRQ) protein are essential for circadian clock function. Here we show that, unlike most genes in Neurospora, frq exhibits non-optimal codon usage across its entire open reading frame. Optimization of frq codon usage abolishes both overt and molecular circadian rhythms. Codon optimization not only increases FRQ levels but, unexpectedly, also results in conformational changes in FRQ protein, altered FRQ phosphorylation profile and stability, and impaired functions in the circadian feedback loops. These results indicate that nonoptimal codon usage of frq is essential for its circadian clock function. Our study provides an example of how non-optimal codon usage functions to regulate protein expression and to achieve optimal protein structure and function. Eukaryotic circadian oscillators consist of autoregulatory circadian negative-feedback loops. In the core circadian oscillator ofNeurospora crassa, FRQ protein is a central component that functions as the circadian negative element with its partner FRH. Two transcription factors, WHITE COLLAR (WC)-1 and WC-2, form a heterodimeric complex that activates frq transcription. The FRQ–FRH complex inhibitsWC complex activity by interacting with theWC proteins. The level and stability of FRQ have a key role in setting period length, phase and clock sensitivity to environmental signals. In addition, FRQ promotes the expression of both WC proteins in an interlocked positive-feedback loop. The protein-coding genes of Neurospora exhibit strong codon bias (Supplementary Fig. 1a). The third position of almost every codon family in this filamentous fungus has the preference C.G.T.A. Codon optimization enhances the expression of a heterologous luciferase gene in Neurospora. To establish that codon-usage bias regulates protein expression, we identified themost abundantNeurospora proteins in a whole-cell extract by mass spectrometry analyses. The genes encoding the top 100 most abundant proteins (Supplementary Table 1) exhibit much stronger codon bias than the rest of the proteincoding genes (Supplementary Fig. 1b). We classified all predicted Neurospora transfer RNA genes and predicted the relative translation elongation rate for each codon on the basis of tRNA-gene copy numbers, which correlate with tRNA abundance, and the nature of anticodon–codon interactions. The most preferred codon for each amino acid is always the codon with highest predicted translation elongation rate (Supplementary Table 2). Therefore, to ensure efficient translation of abundant proteins, selection pressure favoured a bias for codons translated by highly abundant tRNA species. ManyNeurospora genes exhibit little or no codon bias (Supplementary Fig. 1a). FRQ is a low-abundance Neurospora protein. Its codon bias index (CBI; in which CBI5 0 indicates completely random codon usage) value of 0.08 indicates that frq has little codon bias (Supplementary Fig. 1b). A codon-usage graph of the frq open reading frame (ORF) shows that many regions have non-optimal usage (Fig. 1a), whereas frh has good codon usage throughout its ORF. We created two constructs inwhich the amino-terminal end (amino acids 1–164) of frq was codon optimized. In the m-frq construct, only the non-preferred codons were changed, whereas every codon was optimized in the f-frq construct. Predicted stability of RNA secondary structurewas not notably affected by the optimization (Supplementary Table 3). These constructs and the wild-type frq (wt-frq) gene were transformed individually into a frq null strain (frq). Both m-frq and f-frq strains have significantly higher levels of FRQ proteins in constant light than that of the wt-frq strain (Fig. 1b and Supplementary Fig. 2a). On the other hand, frqmRNA levels were comparable in these strains (Supplementary Fig. 2b). FRQ is known to upregulate WC protein levels. The WC-1 and WC-2 levels, however, were similar in these strains despite themuch higher levels of FRQ in the optimized strains (Fig. 1b and Supplementary Fig. 2a, c). The wt-frq construct was able to fully rescue the arrhythmic conidiation rhythm of the frq strain in constant darkness (Fig. 1c), but both of the optimized frq strains exhibited arrhythmic conidiation phenotypes; these are not due to the modest changes in the ratios of two alternatively translated FRQ forms, as either form of FRQ alone can support robust rhythms. We created two additional constructs (m1-frq andm2-frq), inwhich only the aminoor carboxy-terminal segments of the optimized region of m-frq, respectively, were optimized. The frq transformants carrying either construct exhibited long-period conidiation rhythms and had FRQ levels between those of wt-frq and m-frq strains (Fig. 1c and Supplementary Fig. 3a). These results indicate that the severe conidiation rhythm phenotypes of the m-frq and f-frq strains are due to the cumulative effect of codon optimization and are not likely due to mutation of a DNA or RNA element. Toexaminecircadianphenotypesat themolecular level,we introduced a luciferase reporter construct that is under the control of the frq promoter into wt-frq, m-frq and f-frq strains. As shown in Fig. 1d and Supplementary Fig. 3b, the robust rhythmic luciferase activity was abolished in the optimized strains. FRQ protein levels also lost molecular rhythmicity in the optimized strains (Fig. 1e, f and Supplementary Fig. 3c, d): the overall levels of FRQ were high and circadian changes in FRQabundance andphosphorylationprofilewere abolished. In addition, FRQ stayed hyperphosphorylated in constant darkness in the optimized strains. Together, these results indicate that the non-optimal codon usage of frq governs FRQ expression level and is essential for clock function.