Nonoptimal Codon Usage Is Critical for Protein Structure and Function of the Master General Amino Acid Control Regulator CPC-1

The general amino acid control response is critical for adaptation of organisms to amino acid starvation conditions. The preference to use certain synonymous codons is a universal feature of all genomes. Synonymous codon changes were previously thought to be silent mutations. In this study, we showed that the Neurospora cpc-1 gene has an unusual codon usage profile compared to other genes in the genome. We found that codon optimization of the cpc-1 gene without changing its amino acid sequence resulted in elevated CPC-1 expression, an altered protein degradation rate, and impaired protein functions due to changes in protein structure. Together, these results reveal the critical role of synonymous codon usage in regulation of CPC-1 expression and function and establish a genetic example of the importance of codon usage in protein structure.

in these genetic systems, thus confirming the physiological role of codon usage in protein folding in eukaryotic systems. Furthermore, genome-wide correlations between gene codon usage and predicted protein structures have been observed in prokaryotes and eukaryotes, suggesting that codon usage functions as a universal code to broadly modulate protein folding (33,(51)(52)(53). However, there are currently only very few genetic examples that demonstrate the robust physiological influence of codon usage on protein folding and function (19,36,38).
N. crassa has a strong codon usage bias for C/G at wobble positions (33,54), but we observed that the cpc-1 gene has an abnormal NNU-rich codon usage bias. Amino acid starvation triggers the stabilization of yeast GCN4, and we observed a similar starvationinduced stabilization of CPC-1 protein. By changing the NNU codons of cpc-1 to synonymous NNC codons, we demonstrated that the codon usage of cpc-1 is required for CPC-1 stabilization in response to amino acid starvation, and that it is critical for the CPC-1 structure and function in vivo. Together, our results demonstrate the role of codon usage in controlling CPC-1 expression and function and establish another genetic example of the importance of codon usage in protein folding.

RESULTS
Abnormal codon usage profile of cpc-1. Examination of the N. crassa cpc-1 gene revealed that it has an unusual codon usage profile. The Neurospora genome has a strong preference for NNC codons in every ADAT-related codon family (the A34 positions of their corresponding tRNAs can be converted to I34 by adenosine deaminases acting on tRNAs, known as ADATs) and for NNC/NNG codons in other codon families (33,54). In contrast, for cpc-1, NNU codons are the most preferred codons for five (Ala, Pro, Arg, Ser, and Val) of the eight ADAT-related codon families (Fig. 1A). For Leu codons of cpc-1, the normally preferred CUC codon is one of the least used codons; it has a lower usage frequency than CUU. The usage frequency of ACU, which codes for Thr, is also higher in cpc-1 than the genome average (Fig. 1A). Interestingly, the genome-preferred NNC codons are also not the preferred codons for the majority of the ADAT-related codon families in homologous cpc-1 genes in Neurospora tetrasperma, Sordaria macrospora, and Aspergillus nidulans (see Fig. S1 in the supplemental material), suggesting that nonoptimal cpc-1 codon usage is conserved. These results raised the possibility that the nonoptimal nature of the cpc-1 codon usage profile is functionally important.
Regulation of CPC-1 expression by cpc-1 codon usage. The unusual NNU-rich codon usage profile of cpc-1 suggests that it may play a biological role in regulating CPC-1 expression or function. To test this hypothesis, we created two versions of the cpc-1 ORF: cpc-1(WT) (cpc-1 wild type), in which all native codons were maintained, and cpc-1(T¡C), in which all eight ADAT-related NNU codons (except for the codons for the N-terminal 16 amino acids) were substituted synonymously with the genomepreferred NNC codons without altering the amino acid sequence (Fig. S2). Thus, cpc-1(T¡C) has more optimal codons than the wild-type (WT) cpc-1 gene. We expressed 5= epitope-tagged versions of cpc-1(WT) and cpc-1(T¡C) ORFs under the control of the ccg-1 promoter and the ccg-1 5= untranslated region (5= UTR) to exclude the effect of the uORFs and the 5= UTR of cpc-1 on translation. To minimize the potential impact of codon usage on translation initiation, the N-terminal regions (which include 3ϫFlag, an 8ϫGly linker, and the codons for the initial 16 N-terminal amino acids of CPC-1) of the two versions of cpc-1 were identical (Fig. 1B). Constructs containing the cpc-1 transgenes were individually transformed into Neurospora strain 87-3 at the targeted csr-1 locus. Homokaryotic transformant strains were cultured in 2% glucose medium with or without 3-aminotriazole (3-AT). 3-AT treatment results in amino acid starvation in Neurospora because it is a competitive inhibitor of the product of his-3 gene, which is an enzyme required for histidine biosynthesis (6,(55)(56)(57). Because of the cross-pathway control in Neurospora, depletion of one amino acid leads to a general amino acid starvation response (3).
Gene codon optimization usually results in increased mRNA and protein levels in Neurospora (27,58,59). As expected, the mRNA levels of cpc-1(T¡C) were significantly higher than those of cpc-1(WT) when the genes were expressed in the host strain cultured in 2% glucose medium with or without 3-AT (Fig. 1C). Similarly, the CPC-1 protein levels were also upregulated in the cpc-1(T¡C) strain (Fig. 1D). These results suggest that the NNU-rich codon usage profile of cpc-1 suppresses CPC-1 expression.
Codon usage of cpc-1 and culture conditions affect CPC-1 protein stability. GCN4, the ortholog of CPC-1 in S. cerevisiae, is rapidly degraded under rich nutrient conditions but is stabilized under amino acid starvation conditions, a response that contributes to GCN4 upregulation after amino acid starvation (1,10,11,60). To determine whether CPC-1 protein stability is affected by amino acid starvation and codon usage, we compared the CPC-1 turnover rates measured after the addition of the protein synthesis inhibitor cycloheximide (CHX) in the cpc-1(WT) and cpc-1(T¡C) strains grown in 2% glucose medium with and without 3-AT. 3-AT treatment resulted in marked stabilization of CPC-1 in the cpc-1(WT) strain (Fig. 2), suggesting that as, with GCN4 in yeast (1), protein stability was altered under amino acid starvation conditions. However, CPC-1 was more stable in the cpc-1(T¡C) strain than in the cpc-1(WT) strain when the strains were grown in 2% glucose medium without 3-AT. Furthermore, the stabilization of CPC-1 observed in the cpc-1(WT) strain upon 3-AT treatment was not observed in the cpc-1(T¡C) strain (Fig. 2). These results indicate that the NNU-biased cpc-1 codon usage plays an important role in regulating CPC-1 protein stability under amino acid starvation conditions. cpc-1 codon usage affects CPC-1 structure and function. The observed effect of cpc-1 codon usage on CPC-1 protein stability raised the possibility that proper cotranslational folding of CPC-1 depends on codon usage as observed for other Neurospora proteins (19,22,24). To examine this possibility, we performed a limited trypsin digestion assay to probe the structure differences of CPC-1 proteins in the cpc-1(WT) and cpc-1(T¡C) strains. The freshly isolated protein extracts of the cpc-1(WT) and cpc-1(T¡C) strains were treated with trypsin, and the levels of full-length CPC-1 were determined by Western blot analyses as a function of digestion time. When the cultures were grown in 2% glucose medium without 3-AT, CPC-1 isolated from the cpc-1(T¡C) strain was significantly more resistant to trypsin digestion than that isolated from the cpc-1(WT) strain, but it was more sensitive to trypsin digestion after 3-AT treatment (Fig. 3). These results indicate that codon usage influences CPC-1 protein structure.
Note that the ccg-1 promoter-driven cpc-1 expression did not result in its overexpression. In fact, we found that the cpc-1 mRNA level under the control of the ccg-1 promoter and 5= UTR in a cpc-1 knockout strain (cpc-1⌬) was actually much lower than the endogenous cpc-1 level in a WT strain (Fig. S3). Thus, the effect of codon usage on CPC-1 structure is not due to its overexpression.
To determine whether the structural differences caused by codon usage result in changes in protein function, we introduced the cpc-1(WT) and cpc-1(T¡C) constructs individually into the cpc-1⌬ strain. We then compared the abilities of these two constructs to rescue the growth defect of the cpc-1⌬ mutant under amino acid starvation conditions. Under normal growth conditions, the WT and cpc-1⌬ strains had similar growth rates but the cpc-1⌬ strains expressing the cpc-1(WT) or cpc-1(T¡C) had a slightly but significantly lower growth rate in constant light at room temperature (Fig. 4A). In constant light, cpc-1(WT) and cpc-1(T¡C) are constitutively expressed, and

Lyu and Liu
® cpc-1 translation is not regulated by the uORFs due to the use of the ccg-1 5= UTR in the transgene strains. The reduced growth rate in these strains is consistent with the known role of GCN4 as a repressor of protein synthesis (61). In the presence of 3-AT, the growth rate of the cpc-1⌬ strain was dramatically reduced in a race tube assay (Fig. 4B). The growth phenotype was drastically improved in the cpc-1(WT) strain, indicating a functional rescue of the cpc-1⌬ strain by the cpc-1(WT) transgene. The growth rate of the cpc-1⌬, cpc-1(T¡C) strain, however, was much lower than that of the cpc-1⌬, cpc-1(WT) strain in the presence of 3-AT (Fig. 4B), indicating that the T¡C codon usage profile changes impaired the CPC-1 protein function even though they increased the CPC-1 protein level (Fig. 1D).
To further confirm this conclusion, we compared the mRNA levels of 14 genes regulated by CPC-1 that are involved in amino acid biosynthesis (3,9). The mRNA levels of these CPC-1 target genes in the cpc-1⌬; cpc-1⌬, cpc-1(WT)⌬; and cpc-1⌬, cpc-1(T¡C) strains treated with 3-AT were determined. The mRNA levels of all the 14 genes were dramatically upregulated in the cpc-1⌬, cpc-1(WT) strain compared to those in the cpc-1⌬ strain (Fig. 4C). As expected, the mRNA level induction of these genes was reduced in the cpc-1⌬, cpc-1(T¡C) strain (Fig. 4C). Together, these results demonstrate that the NNU-biased codon usage profile of cpc-1 plays an important role in determining CPC-1 protein structure and function in vivo.

DISCUSSION
CPC-1 is the master transcription regulator of gene expression in Neurospora in response to amino acid starvation (3,7,55,62). Like the situation of its yeast ortholog GCN4, expression of CPC-1 is also translationally regulated by a mechanism involving uORFs (7,9). The translation of cpc-1 can be translationally activated by bypassing its uORFs under amino acid starvation conditions. As reported previously for GCN4 (10,11), here, we showed that 3-AT treatment triggers CPC-1 stabilization, which also contributes to CPC-1 accumulation under amino acid starvation conditions. Although the mechanism of CPC-1 degradation is not known, it is possible that, as with GCN4, amino acid starvation regulates posttranslational modification of CPC-1, which affects its degradation by the ubiquitin-proteasome pathway (11)(12)(13).
In this study, we demonstrated that the nonoptimal codon usage profile of cpc-1 has a major impact on the structure and function of CPC-1. By changing the cpc-1 codon usage from the NNU-rich profile to the NNC-rich profile typical of the Neurospora genome, we showed that codon usage is critical for CPC-1 protein structure and function. This conclusion was supported by several lines of evidence. First, the codon manipulation altered the CPC-1 protein degradation rate and abolished amino acid starvation-induced CPC-1 stabilization (Fig. 2), suggesting that codon usage can affect CPC-1 structure. Second, the codon optimization altered the sensitivity of CPC-1 to limited trypsin digestion, indicating that codon optimization affected protein structure (Fig. 3). Third, in the presence of 3-AT, CPC-1(T¡C) was less stable and more sensitive to trypsin digestion than CPC-1 (WT) was ( Fig. 2; see also Fig. 3), suggesting that codon usage-mediated structure changes of CPC-1 affected its ability to be regulated by potential posttranslational mechanisms triggered by amino acid starvation conditions. Fourth, despite the upregulation of CPC-1 protein levels in Neurospora upon codon optimization, expression of cpc-1(T¡C) did not rescue the growth defects of the cpc-1⌬ strain under amino acid starvation conditions as effectively as that of cpc-1(WT) did. Finally, the impaired CPC-1 function of the cpc-1(T¡C) strain was further indicated by the reduced induction of CPC-1 target genes in response to amino acid starvation. Note that codon optimization of cpc-1 did not cause its overexpression (see Fig. S3 in the supplemental material). Thus, our study established a critical physiological role of codon usage in regulating CPC-1 structure and function.
In addition, our analyses established another in vivo example of the influence of codon usage on protein structure. Due to the role of codon usage in regulating the rate of translation elongation, codon usage was previously proposed to influence the cotranslational protein folding process (19,22,26,28,(33)(34)(35)(36)(37)(38). However, genetic evi-dence in support of such a role of codon usage is quite limited. By studying the codon usage function of the circadian clock genes frequency in Neurospora and Period in Drosophila, we previously showed that codon usage plays an important role in affecting the structures and, therefore, the functions of these two proteins in vivo (19,36). Similarly to the frequency and Period genes, cpc-1 is enriched in nonoptimal codons. Additionally, as with FRQ and PER proteins, most regions of the CPC-1 protein are predicted to be intrinsically disordered. Our findings are consistent with the hypothesis that the cotranslational protein folding process is sensitive to codon usage-mediated translation elongation kinetics and that this process is regulated to ensure proper functioning of the proteins with intrinsically disordered domains. Further supporting this, we and others previously showed that nonoptimal codon usage correlated with predicted unstructured domains in a genome-wide manner in Neurospora and other organisms (33,52). The structure of the DNA binding domain of the yeast GCN4 was previously shown to be flexible (63)(64)(65). GCN4 exhibits a concentration-dependent ␣-helical transition: the transition of the GCN4 basic region from an unfolded to a folded conformation depends on its accessibility to DNA binding sites (65). Such properties may make it more sensitive to the cotranslational folding process.
Taken together, our results suggest that the unusual codon profile of cpc-1 represents another example of evolutionary adaption that results in its optimal protein structure and function in response to environmental changes.

MATERIALS AND METHODS
Strains and growth conditions. N. crassa strain 87-3 (bd, a) was used as the control and was further used as the host strain for the expression of various versions of cpc-1 unless otherwise specified. For the growth rate assay, the FGSC 4200 (a, WT) strain was used as the control. The cpc-1⌬ strain was obtained from the Neurospora knockout library (66). Liquid cultures were grown in 2% glucose medium (1ϫ Vogel's, 2% glucose) or in 0.1% glucose medium (1ϫ Vogel's, 0.1% glucose, 0.17% arginine). Race tube medium contained 1ϫ Vogel's, 0.1% glucose, 0.17% arginine, 50 ng ml Ϫ1 biotin, and 1.5% agar. All the strains were cultured on slants containing 1ϫ Vogel's, 2% sucrose, and 1.5% agar before various experiments were performed. All the strains were cultured under constant light at room temperature.
Plasmid constructs. For gene expression at the csr-1 locus in N. crassa, a hygromycin B resistance gene (hph) was inserted downstream of the ccg-1 promoter of a parental plasmid, Pcsr1, to create a new plasmid, Pcsr1-hyg. Pcsr1-hyg is a csr-1-targeting expression vector with an expression cassette in which P ccg-1 and hph flank the gene of interest, and this cassette is flanked by two csr-1-related fragments that serve as the double recombination sites (67). When this plasmid was transformed into N. crassa cells, it was integrated into the csr-1 gene locus by replacing csr-1 with the expression cassette by double homologous recombination. The resulting transformants were screened for both hygromycin B (200 g ml Ϫ1 ) resistance and cyclosporine (5 g ml Ϫ1 ) resistance conferred by the presence of hph and the absence of csr-1, respectively. The levels of efficiency and accuracy of this approach were very high (Ͼ90% positive transformants). In this study, two versions of cpc-1(WT) and cpc-1(T¡C) with a 3ϫFlag tag and an 8ϫGly linker at the N termini were separately introduced into the Pcsr1-hyg construct. The resulting constructs were transformed into host strains by electroporation. Homokaryon strains were obtained by microconidium purification.
Protein stability and limited trypsin digestion assays. For protein stability assay, the cycloheximide (CHX) working concentration and experimental procedures were the same as those previously described (19). For culture conditions, fresh conidia (1 week postinoculation on slants) of the host strains were cultured in 50 ml 2% glucose medium in plates at room temperature for 2 days. The cultures were cut into small discs with a diameter of 1 cm, and then the discs were transferred into flasks with the same liquid medium and were grown with orbital shaking (200 rpm) for one more day before addition of CHX (final concentration, 10 g ml Ϫ1 ). For the samples treated with 3-AT, the culture discs in 2% glucose medium were treated with 5 mM 3-AT for 8 h before sample collection. Cells were collected at the indicated time points after addition of CHX. For the limited trypsin digestion assay, the culture conditions and sample collection procedures were the same as those described above except for the addition of CHX. The working concentration of trypsin was 0.25 g/ml. Protein extraction and Western blot analyses were performed as previously described (68). Equal amounts of total proteins (100 g) were loaded into all lanes of 7.5% SDS-PAGE gels containing 37.5:1 acrylamide/bisacrylamide. The primary and secondary antibodies used for detecting the 3ϫFlag were monoclonal anti-Flag M2 antibody produced in mouse (Sigma-Aldrich, catalog no. F3165) and goat anti-mouse IgG (HϩL)-horseradish peroxidase (HRP) conjugate (Bio-Rad, catalog no. 170-6516), respectively. Densitometry was performed using Image J.
Quantitative reverse transcription-PCR (qRT-PCR) and mRNA-seq. For qRT-PCR, the sample collection procedures were the same as those described for the protein stability assay except that CHX was not added. For cultures treated with 3-AT as indicated in the figures, the liquid cultures were treated with 5 mM 3-AT for 8 h before sample collection. RNA extraction and qRT-PCR were performed as previously described (69). ␤-tubulin (NCU04054) was quantified as an internal control. Primers used for qRT-PCR are listed in Table S1 in the supplemental material. The relative mRNA levels of cpc-1 in the WT, Lyu and Liu ® September/October 2020 Volume 11 Issue 5 e02605-20 mbio.asm.org 8 cpc-1⌬, and cpc-1⌬, cpc-1(WT) strains under amino acid starvation conditions were measured by determination of their RPKM (reads per kilobase per million) values from our high-throughput mRNA sequencing (mRNA-seq) data. The mRNA sequencing libraries used in this study were generated from cultures maintained in 2% glucose medium with 5 mM 3-AT treatment for 8 h before sample collection. The sample collection procedures were the same as those described for the protein stability assay except that CHX was not added. Total RNAs were extracted using TRIzol reagents (Invitrogen) and treated with DNase (Turbo DNase; Ambion). The libraries were prepared using NEBNext Ultra kits for RNA and sequenced by an Illumina HiSeq 2000 instrument. mRNA-seq experiments were performed by Joint Genome Institute (JGI) on an Illumina NovaSeq platform. Codon manipulation and data collection from databases. The codons of cpc-1 were optimized based on N. crassa codon usage frequency data from the Codon Usage Database (https://www.kazusa .or.jp/codon/cgi-bin/showcodon.cgi?speciesϭ5141). The mutated sites for the optimized cpc-1(T¡C) are shown in Fig. S2 in the supplemental material.
Data availability. The raw and processed sequencing data have been submitted to the NCBI Gene Expression Omnibus under accession number GSE150287.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.