Translation Initiation from Conserved Non-AUG Codons Provides Additional Layers of Regulation and Coding Capacity

Bioinformatic analyses of fungal cpc-1.While studying the regulation of cpc-1 by its uORFs in N. crassa extracts, using a construct in which the wild-type (WT) 5′ leader of cpc-1 was fused with the open reading frame of firefly luciferase (LUC), we observed a band of predicted size and a band ~20 kDa larger than predicted (Fig. 1A). This prompted a more careful examination of the mRNA 5′ leader sequence. We found that the CPC1 reading frame extended far upstream (Fig. 1B), without any in-frame stop codons, to the major mapped transcription initiation site, which is located 703 nucleotides (nt) 5′ of the predicted AUG for the main open reading frame (designated mAUG and mORF, respectively). We previously noted that N. crassa cpc-1 could hypothetically use upstream near-cognate start codons for initiation (17). We next compiled partial or complete sequences of cpc-1 homologs from 108 Pezizomycotina species: 100 sequences included the region spanning from uORF1 to a position downstream of the mAUG and were analyzed further. All homologs contain two AUG-initiated uORFs, with uORF1 spanning 3 to 6 codons and uORF2 spanning 35 to 70 codons, including their stop codons. Surprisingly, in all cases, the reading frame for CPC1 could be substantially N-terminally extended without encountering an in-frame stop. The shortest extension of the CPC1 ORF without encountering an in-frame stop codon is 160 codons in Leptosphaeria maculans. We note that some automated annotations of CPC1 homologs include this N-terminal extension (e.g., XP_001906068 , EGR46729 , and EKJ70155 ), but annotations do not resolve where initiation occurs. The presence of this feature in both Sordariomycetes and Eurotiomycetes suggests that it was present in their last common ancestor and possibly earlier; the last common ancestor of all Pezizomycotina is estimated to have lived at least 320 million years ago (33).

• Open in new tab
• Download powerpoint

FIG 1 

N-terminal extensions of Pezizomycotina CPC1. (A) cpc-1–luc mRNA produces a larger product in vitro than luc mRNA. mRNA templates were used to program N. crassa in vitro translation extracts, and [³⁵S]Met-labeled products were analyzed on a 12% NuPAGE gel. Translation of mRNA for N. crassa cpc-1–luc (in which the 5′ leader of cpc-1 plus the first three codons of its mORF are fused in frame to firefly luciferase) and that for luc were compared to a no-mRNA control. The positions of full-length firefly LUC (arrowhead) and the larger CPC1-LUC product (asterisk) are indicated. (B) Schematic diagram of the N. crassa cpc-1 mRNA. Each reading frame is on a separate line. Frame 1 specifies CPC1 (black rectangle). uORF1and uORF2 (blue rectangles) initiate from uAUG1 and uAUG2, respectively, in other reading frames. AUG codons are indicated by green bars, and stop codons are indicated by red bars. NCCs in frame 1 upstream of uORF2 are indicated by magenta bars and are numbered 1 to 8. The approximate position of the 3′-most stop codon upstream of uORF2 and in frame with the main ORF (present in Cordyceps bassiana) is indicated (dashed red bar). The number of sequences used for comparisons is shown in parentheses. Features are drawn to scale. The nucleotide sequence of the N. crassa cpc-1 5′ leader is given in Fig. S6 in the supplemental material. (C to J) Frequency WebLogo of the conservation of the initiation contexts, from −6 to +4, of all N. crassa genes initiated with AUG (C), all A. nidulans genes initiated with AUG (D), Pezizomycotina cpc-1 uAUG1 (E), Pezizomycotina cpc-1 uAUG2 (F), Pezizomycotina cpc-1 mAUG (G), Pezizomycotina NCC6 (CUG) (H), Pezizomycotina NCC7 (AUU) (I), and Pezizomycotina NCC8 (ACG) (J). Letter heights are proportional to the frequency of occurrence of each nucleotide at each position. The crucial positions −3 and +4 are indicated in red underneath the frequency plots. Data used to calculate consensus AUG initiation context for N. crassa and A. nidulans were obtained from the Transterm database (61). (K) The amino acid sequence encoded by the N. crassa mRNA in the CPC1 reading frame, starting from the 5′ end of the mRNA and ending with the first in-frame stop codon. mAUG indicates the annotated cpc-1 initiation codon; upstream NCC6 (uNCC6), uNCC7, and uNCC8 are also indicated. The approximate positions of uORF1 and uORF2 (which are in other reading frames) are indicated by blue lines. The C-terminal bZIP domain of CPC1 is indicated in red font. The regions analyzed in Fig. S3A and B are bracketed by dashed red lines. The highly conserved patch specifically marked in Fig. S3A is bracketed by a dashed orange line. The level of conservation of each residue from the alignment of homologs from 95 species (those used to construct the tree in Fig. S1A) is shown below the amino acid sequence and was generated using ClustalX2. The conservation, expressed as percent amino acid identity, is indicated on the right side of each alignment.

Based on the mechanism of translational control of S. cerevisiae GCN4, control of cpc-1 would involve ribosomes initiating at uORF1 and reinitiating at uORF2 under amino acid-sufficient conditions. When eIF2α phosphorylation levels increase in response to amino acid limitation, ribosomes would reinitiate at the downstream cpc-1 mAUG instead of uORF2. Remarkably, without exception in the Pezizomycotina, there is no stop codon in the reading frame of the mORF between the uORF2 AUG and the mAUG. The in-frame stop codon closest to uAUG2 (Cordyceps bassiana) is 101 nt upstream of it. Thus, the potential amino-terminal extensions of CPC1 are encoded upstream of uORF2 (Fig. 1B).

Initiation upstream of the uORF2 AUG could produce N-terminally extended isoforms of CPC1 whose synthesis would not be subject to inhibition by uORF2. We searched for potential start codons in this region of N. crassa cpc-1 mRNA that were in frame with the predicted mAUG. Eight NCCs fulfilling these criteria were identified—three AUC (NCC1, NCC3, and NCC4), two ACG (NCC2 and NCC8), two AUU (NCC5 and NCC7), and one CUG codon (NCC6) (Fig. 1B). We next searched for potential NCCs in similar regions of cpc-1 transcripts from all Pezizomycotina (see Fig. S1 in the supplemental material) and compared their conservation levels. Three NCCs in N. crassa showed particularly deep conservation—the closest to the CPC1 AUG, an ACG (NCC8), is perfectly conserved in 98 of 100 species; NCC7, an AUU, is conserved in 74 of 100 species (as AUU in 64 and AUC or AUA in 10); NCC6, a CUG, is conserved in 77 of 100 species (as CUG in 73 and UUG in 4) (Fig. S1 and S2). In no case was there an in-frame stop codon between these three conserved NCCs and the mAUG. The three conserved NCCs also show a clear pattern of fungal branch-specific distribution: the minority of homologs lacking both AUU and CUG NCCs clustered separately from the other homologs (Fig. S1). The other five NCCs showed sporadic conservation and were found only in species that were closely related to N. crassa. None of the N. crassa NCCs appeared conserved in the two most distant Pezizomycotina, Arthrobotrys oligospora and Tuber melanosporum (Fig. S1). However, even these two species’ CPC1 homologs contain multiple NCCs upstream of uORF2 and in frame with the mORF—6 NCCs in A. oligospora and 7 NCCs in T. melanosporum (Fig. S2C).

We next examined the conservation of the initiation contexts for the three conserved NCCs and for the uORF1 AUG (uAUG1), uORF2 AUG (uAUG2), and mAUG. The preferred initiation context in N. crassa (Fig. 1C), which is considered optimal, is similar to the preferred context in the relatively distant Aspergillus fumigatus (34) and Aspergillus nidulans (as shown in Fig. 1D). uAUG1 and uAUG2 are in conserved optimal contexts (Fig. 1E and F and S2A), consistent with their presumed roles in regulating CPC1 translation through controlling reinitiation. Conservation of mAUG context is weaker, but the consensus is still near optimal (Fig. 1G). Of the three conserved NCCs, NCC8, which is closest to the mAUG and is the most conserved, showed the highest context conservation (Fig. 1J and S2A). The consensus initiation context of NCC8 in species that we examined is nearly optimal (nucleotides −4, −3, −1, and +4 match the consensus). The most important nucleotides, A at position −3 and G at position +4, are perfectly conserved in all Pezizomycotina that have NCC8. Lower context conservation is observed for NCC7 and NCC6 (Fig. 1H and I), although their consensus initiation contexts remain nearly optimal.

One question raised by the potential N-terminal extensions of cpc-1 homologs is whether they are evolutionarily conserved at the amino acid sequence level. A plot of the amino acid conservation of Pezizomycotina CPC1 sequences relative to N. crassa sequence is shown in Fig. 1K. A highly conserved region is present near the C terminus of CPC1 (residues 430 to 500), which corresponds to the α-helix of the bZIP DNA binding domain (Fig. 1K). Excluding this, there are few highly conserved stretches, but the N-terminal extensions are as well conserved as the mORF. We examined two conserved regions in the N-terminal extension (Fig. 1K, red dashes) more closely to determine if the conservation is at the amino acid or the nucleotide level and, if it is at the amino acid level, which reading frame showed the highest conservation (Fig. S3A and B). The first conserved region examined overlaps uORF2. The proportion of synonymous substitutions was much higher in the mORF frame than in uORF2 frame (Fig. S3A). For the 5 codons showing the highest amino acid conservation in the mORF frame (Fig. 1K, orange dashed bracket), the ratio of synonymous to nonsynonymous substitutions is particularly striking. The second conserved region examined comprises 7 codons starting 16 codons upstream of the mAUG. This region also shows a high proportion of synonymous substitutions in the mORF frame compared to the other frames (Fig. S3B), indicating that its conservation occurs because of selection in the mORF frame.

The coding potential of the upstream extension was analyzed with MLOGD (35). MLOGD calculates coding potential by using the patterns of substitutions observed across a sequence alignment to compare a coding model with a noncoding model via a likelihood-ratio test. When applied in a 20-codon sliding window, MLOGD detected a positive coding signature within the CPC1 AUG-initiated ORF (as expected) and upstream throughout the extension as far 5′ as NCC6 (Fig. S4). The coding signature was weaker (but still positive) from NCC6 to around one-third of the way through uORF2. This may be a result of increased CPC1 frame synonymous site conservation in this region (Fig. S4), leading to fewer sequence variations for MLOGD to distinguish between the coding and noncoding models. The enhanced synonymous site conservation (Fig. S4) is indicative of overlapping functional elements putting extra constraints on sequence evolution in this region, likely including the initiation contexts of NCC6 to NCC8 and the overlapping uORF2. The ratio of nonsynonymous to synonymous substitutions, dN/dS, was calculated for the region between NCC8 and the CPC1 AUG using codonml (36) and found to be 0.348 + 0.033 and thus statistically significantly less than 1 (99% confidence interval, 0.26 to 0.43), indicating that the upstream extension is indeed subject to purifying selection at the amino acid level, consistent with its being a coding sequence. Since synonymous site conservation interferes with use of dS as a proxy for neutral evolution, we also calculated dN/dS for the region from 21 codons after NCC8 to the CPC1 AUG (Fig. S4), giving a dN/dS ratio of 0.285. For comparison, the dN/dS ratio for the region between the CPC1 AUG and CPC1 stop codon was found to be 0.144 + 0.012, indicating stronger purifying selection on average in the AUG-initiated CPC1 ORF than in the upstream extension. Taken together, these data indicate that the mRNA sequences specifying the N-terminal extension are under purifying selection in the mORF frame.

We next investigated the architecture of cpc-1 homologs in fungi outside the Pezizomycotina. Examination of multiple sequences from other classes within the Ascomycota, including Saccharomycotina (including S. cerevisiae GCN4) and Taphrinomycotina, showed that these cpc-1 homologs lack the analogous N-terminal extensions of the main ORF. Thus, the conserved N-terminal extension in Ascomycota is confined to Pezizomycotina. Little comparative sequence information was available to examine other fungal phyla except for Basidiomycota. Within this phylum, analysis was complicated by the presence of multiple cpc-1 paralogs in some species. Typically, the 5′ leaders of cpc-1 homologs from Basidiomycota have 3 to 4 uAUGs. These can either initiate, or exist within, the reading frames of two or three uORFs (Fig. S5). uORF1 is 4 to 7 codons long, while one of the downstream uORFs, initiated by AUG in a good context, is much longer (uORFL). Crucially, uORFL sometimes overlaps the mORF (see Fig. S5B and C). Examination of 32 cpc-1 Basidiomycota mRNA sequences (3 from Ustilaginomycotina, 27 from Agaricomycetes, and 2 from Microbotryomycetes) with identifiable uORFs revealed that, in all cases, no stop codon in frame with the mORF is present between uORF1 and the mAUG (Fig. S5A and B). In fact, no stop codon in frame with the mORF is closer than 77 nucleotides upstream of uORF1. Unlike in Pezizomycotina, where three highly conserved NCCs were identified for most N-terminal extensions, no well-conserved NCCs were identified in Basidiomycota. However, in every Basidiomycota cpc-1 homolog, several NCCs in good initiation contexts and in the same frame with the mORF are present 5′ of the apparently inhibitory uORFL. In all cases, the first NCC is located 5′ of uORF1 such that potential translation initiation at the NCC would bypass the regulatory effects of the uORFs.

We searched the 27 uORF-containing cpc-1 homologs from Agaricomycetes for conserved features. In these, there is a single conserved NCC capable of initiating translation of an N-terminal extension and this NCC is present at least 31 nucleotides 5′ of the uORF1 AUG (Fig. S5B). Although the position of this NCC is well conserved, its identity is not. In most cases, it is AUU; in others, it is UUG, AUA, or CUG (Fig. S1B). The specific identities of these NCCs appear largely specific to phylogenetic branches.

The preferred initiation context in Agaricomycetes, as determined by analyses of Coprinopsis cinerea (Fig. S5D), is similar to the context in both Pezizomycotina and mammals. Based on this, the context of the single conserved NCC in the 5′ leaders of cpc-1 homologs in Agaricomycetes appears favorable if not optimal (Fig. S5E). The putative N-terminal extensions in Agaricomycetes are shorter than in Pezizomycotina—approximately 120 versus 180 amino acids, respectively. The amino acid conservation in Agaricomycetes is also concentrated in the C-terminal region of the mORF that contains the α-helix including the bZIP DNA binding domain (red letters in Fig. S5F). Patches of substantial conservation are observed within the 50 amino acids upstream of the mORF. The most highly conserved stretch in this region was subjected to a more careful examination (Fig. S3C, red dashed line). This sequence overlaps the last, and usually longest, uORF. This analysis indicates that conservation of amino acid sequence of the N-terminal extension in the mORF frame is more important than conservation in the uORF or the third reading frame (Fig. S3C), consistent with the findings in Pezizomycotina.

A peculiar mRNA architecture exists in cpc-1 homologs in Microbotryomycetes (Fig. S5C). Even though only two uORF-containing homologs of cpc-1 were obtained in this branch of Basidiomycota, both transcripts have the same unusual feature (Fig. S5C). Unlike Ustilaginomycotina (Fig. S5A) or Agaricomycetes (Fig. S5B), no evidence was detected for the existence of a short regulatory uORF1. The 5′ end of the homolog from Leucosporidium scottii is well supported by several expressed sequence tags (ESTs), and the upstream neighboring gene is in close proximity. Thus, instead of a uORF1, a long uORF initiated by AUG in a favorable initiation context is present, which overlaps the mORF. No in-frame stop codons are seen upstream of the mAUG. A single conserved NCC can be identified upstream of the uORF start codon so that ribosomes initiating from this NCC could synthesize an N-terminally extended CPC1 isoform and completely bypass any inhibitory effects of the uORF.

Experimental analyses of N. crassa cpc-1.To investigate the effects of uORF1, uORF2, and upstream NCCs on the translation of N. crassa CPC1 in cell extracts, the 5′ leader of cpc-1, including the first two codons of the mORF, was fused in frame to firefly luciferase (cpc-1–luc, designated wild type [WT] [Fig. S6]). The functions of initiation codons identified by bioinformatics approaches were tested by mutational analyses of this construct. A UAA mutation (designated UAA) was introduced in frame with, and 12 nt upstream of, the mAUG to terminate translation and therefore truncate translation products that initiated from upstream NCCs. The start codon of uORF1 was mutated to AAA (ΔuORF1), that of uORF2 was mutated to ACA (ΔuORF2), and that of the mORF was mutated to CTC (ΔmAUG), to eliminate their initiation activity.

The functions of uORF1 and uORF2 were examined by mutating their start codons separately or together. First, we examined these mutations in constructs containing the UAA mutation to look specifically at luciferase synthesis from the mAUG. Luciferase synthesis was measured by enzyme activity assay and by labeling with [³⁵S]Met (Fig. 2A). Compared to a construct containing both uORFs, the ΔuORF1 mutation diminished translation of the mORF as indicated by a reduced level of luciferase activity (15%) and decreased production of [³⁵S]Met-labeled polypeptides (compare constructs 1 and 2, Fig. 2A). The ΔuORF2 mutation increased translation from the mAUG approximately 2.9-fold (compare constructs 1 and 3, Fig. 2A). For the ΔuORF1 ΔuORF2 double mutant, the synthesis of luciferase increased (compare constructs 1 and 4, Fig. 2A), but this increase was less than for ΔuORF2 alone. These data suggest that reinitiation occurs after translation of uORF1, that translation of uORF2 is inhibitory, and that a fraction of ribosomes that translate uORF1 reinitiate at uORF2. In the absence of uORF1 and uORF2, synthesis of luciferase is lower than in the absence of uORF2 alone. This could be explained if the NCCs are used more efficiently in the absence of uORF1 (see below).

• Open in new tab
• Download powerpoint

FIG 2 

Contribution of cpc-1 uORF1 and uORF2 to the regulation of translation from the mAUG in N. crassa cell extracts. (A) Effects of eliminating cpc-1 uORF1 and uORF2 on translation from the mAUG. Constructs (numbered 1 to 4) contained the UAA stop codon (red bar) to eliminate translation from upstream in-frame NCCs and the indicated mutations to uORF start codons to eliminate initiation from them (uORF1 AUG to AAA and/or uORF2 AUG to ACA). Capped and polyadenylated mRNA (6 ng) was used to program N. crassa translation reaction mixtures (10 µl). LUC activity produced from mRNAs 2 to 4 obtained after 30 min of incubation at 26°C was calculated relative to the activity produced from mRNA 1. Mean values and standard deviations from three independent experiments, each performed in triplicate, are given as normalized relative light units (RLU). In addition, [³⁵S]Met-labeled translation products from translation reactions programmed with mRNAs 1 to 4 or with no mRNA were analyzed on 12% NuPAGE gels, and a representative gel is shown. The position of radiolabeled LUC produced from the mAUG is indicated. (B) Toeprint analysis indicates reinitiation following translation of cpc-1 uORF1 but not uORF2. cpc-1–luc mRNA (60 ng) was used to program 20-µl N. crassa cell-free translation reaction mixtures. WT mRNA containing the wild-type cpc-1 5′ leader and the mRNAs used in panel A were analyzed in parallel along with controls. Reaction mixtures were incubated at 26°C min with cycloheximide (CYH) added either prior to incubation (T₀) or after 10 min of incubation (T₁₀) as indicated. Radiolabeled primer CPC101 was used to examine ribosomes at uORF1 and uORF2; primer ZW4 was used to examine ribosomes at the mORF. The original data from which the toeprint signals were excised are shown in Fig. S7. (C) Discriminating translation from N. crassa cpc-1 NCCs and mAUG in vitro. Capped and polyadenylated mRNA (6 ng) was used to program N. crassa translation reaction mixtures (10 µl) with the indicated constructs. Firefly luciferase activity from each mRNA obtained after 30 min of incubation at 26°C was calculated relative to synthesis from the WT construct. Mean values and standard deviations from three independent experiments, each performed in triplicate, are plotted.

In earlier studies on S. cerevisiae GCN4, we used toeprint analyses to demonstrate reinitiation following uORF1 but not uORF4 translation in S. cerevisiae extracts (12). We adapted a similar approach to examine cpc-1 uORF1 and uORF2 in N. crassa extracts. Adding cycloheximide (CYH) to reaction mixtures at time zero (T₀) allows toeprint mapping of initiation codons where 80S ribosomes first initiate translation following initial scanning. Adding CYH at 10 min of incubation of translation reaction mixtures (T₁₀) allows toeprint mapping of initiation sites in the steady state, for example, at additional sites where ribosomes have reinitiated. At T₀ and T₁₀, ribosomes are seen at the uORF1 AUG start codon; mutation to AAA eliminated this signal (Fig. 2B). This is expected since the uORF1 AUG is in an optimal initiation context. At T₀, a reduced toeprint signal is seen at the uORF2 AUG relative to the signal at the uORF1 AUG. When the uORF1 AUG is mutated, the uORF2 AUG signal increased substantially; mutation of the uORF2 AUG to ACA eliminated this signal. These data indicate that most ribosomes initiate at uORF1 but, when it is absent, they scan to uORF2. At T₀, a relatively low signal was observed at the mORF AUG except when uORF1 and uORF2 AUGs were mutated, as expected from scanning. When CYH was added at T₁₀, the most dramatic change in signal was an increase at the mAUG in the ΔuORF2 construct. This increase of the mAUG was not seen in the ΔuORF1 construct or the ΔuORF1 ΔuORF2 construct. These data are consistent with ribosomes reinitiating at the mAUG following uORF1 translation in vitro. They suggest that uORF1 and uORF2 of N. crassa cpc-1 function similarly to uORF1 and uORF4, respectively, of S. cerevisiae GCN4.

We next compared luciferase activities obtained from constructs with and without the introduced in-frame UAA stop codon to examine translation from NCCs upstream of the mAUG (Fig. 2C). The production of luciferase decreased when the UAA was present, indicating that polypeptides with luciferase activity were produced using NCCs upstream of the mORF (compare constructs 1 and 3, constructs 2 and 4, and constructs 3 and 6 in Fig. 2C). The UAA mutation decreased luciferase synthesis in the presence or absence of uORF2 (compare constructs 1 and 2 and constructs 3 and 4 [Fig. 2C]). Elimination of uORF2 resulted in overall increased luciferase synthesis as expected from its proposed inhibitory role for initiation at mAUG. As expected, NCCs and the mAUG have separate roles in initiation; combining ΔmAUG and UAA mutations yielded no detectable luciferase (Fig. 2C).

Interestingly, ΔmAUG showed a relatively small decrease in luciferase activity compared to WT (ΔmAUG, 71% of WT; compare constructs 1 and 5, Fig. 2C). This observation is consistent with the differences between the WT and the UAA constructs (UAA, 22% of WT; compare constructs 1 and 3, Fig. 2C). These data indicate that upstream NCCs are used to initiate translation efficiently in vitro. While this is the case, we did not identify NCCs by toeprint analyses (Fig. S7). Possibly, translation initiation is distributed among multiple NCCs, reducing signals at individual NCCs.

The eight NCCs identified bioinformatically (Fig. S6) were individually eliminated, and the consequences were examined by [³⁵S]Met labeling in N. crassa and wheat germ extracts (Fig. 3 and S7). These mutations were also combined with the UAA mutation so that the resulting polypeptides produced from upstream initiation could be better resolved by SDS-PAGE. Elimination of NCC1, NCC2, NCC3, or NCC4 did not yield any detectable differences compared to UAA (Fig. S8, lanes 3 to 7). In contrast, elimination of NCC5, -6, -7, or -8 resulted in disappearance of specific truncated polypeptides (Fig. S8, lanes 8 to 11 and 3, and Fig. 3, lanes 8 to 11 and 3), indicating that NCC5 to NCC8 initiated translation in N. crassa and wheat germ systems. When NCC8 (ACG) was changed to AUG, the signal in the corresponding band increased as expected (lanes 12 and 3, Fig. S8, and lanes 8 and 3, Fig. 3).

• Open in new tab
• Download powerpoint

FIG 3 

Evidence that N. crassa NCC5 to NCC8 initiate translation in vitro. Synthetic RNAs (60 ng) for the indicated constructs were used to program 10 µl of cell-free translation systems from N. crassa (left) or wheat germ (right). Reaction mixtures were incubated for 30 min at 26°C. Radiolabeled products were analyzed on 12% NuPAGE gels. Open circles, translation products eliminated upon mutation of NCC5 to NCC8; the product predicted to be initiated from NCC8 also increased when NCC8 was changed to AUG (lane 8). Arrowhead, position of mAUG-initiated translation product (mORF). Brackets, translation products larger than the mORF produced in the absence of an in-frame UAA stop codon.

Cell translation extracts programmed with CPC1-LUC (WT) produce polypeptides migrating more slowly than luciferase synthesized from an mRNA specifying LUC alone or a CPC1-LUC mRNA with the UAA mutation (Fig. 1A, 3, and S8). When NCC5, NCC6, NCC7, and NCC8 were eliminated together in the absence of the UAA mutation, polypeptides larger than luciferase were still observed, although the amount was reduced compared to WT (Fig. 3, compare lanes 10, 11, and 3). This suggests that other upstream codons in the cpc-1 upstream region can be used to initiate polypeptide synthesis. This was observed in the presence or absence of uORF1 (lanes 10 and 11, Fig. 3).

To examine the roles of upstream NCCs in translation in vivo in N. crassa, strains containing N. crassa codon-optimized luciferase fused in frame with wild-type or mutated cpc-1 5′ sequences were constructed. Three independent transformants containing each construct were used to measure LUC activity and LUC mRNA levels. We examined WT, UAA, and ΔmAUG strains and the ΔmAUG UAA double mutant. Luciferase activity was measured and normalized to reporter mRNA levels to account for the small differences in luciferase mRNA levels observed. Expression levels of WT and UAA reporters were similar (in Fig. 4). Luciferase activity from the ΔmAUG construct was much lower, but this activity was higher than that for the ΔmAUG UAA construct. For the ΔmAUG construct, higher luciferase activity was observed than for the ΔmAUG UAA double mutant. Thus, although the amount of luciferase activity derived from upstream NCCs was less than 1% of activity from the mAUG in vivo, detectable luciferase was nevertheless observed (compare constructs 3, 1, and 2 in Fig. 4 and compare constructs 5, 1, and 3 in Fig. 2C). Possibly, NCCs were not used as efficiently in vivo as in vitro. Alternatively, N-terminally extended luciferases are less stable or less active in vivo, but we have not investigated this further.

• Open in new tab
• Download powerpoint

FIG 4 

Discriminating translation from N. crassa cpc-1 NCCs and mAUG in vivo. Constructs 1 to 4 were placed at the N. crassa his-3 locus (three independent transformants of each). LUC activities were measured, and values were plotted relative to WT. Black bars, LUC activities normalized to total extracted protein; gray bars, LUC activities normalized first to total protein and then to luc mRNA/cox-5 mRNA levels. Mean values and standard deviations for all measurements are derived from three independent experiments, each using all independent transformants.

For further investigation of translational activity in the region of cpc-1 mRNA upstream of mAUG, data from ribosome profiling experiments in N. crassa were examined. Ribosome profiling provides snapshots of genome-wide in vivo translation by deep sequencing which is amenable to quantification. Cells were grown for 24 h in the dark, and ribosome profiling data were collected and analyzed as described in Materials and Methods. As shown previously, ribosome footprint data can be used to determine the frame in which a particular region of mRNA is being translated (29, 37, 38). The ribosome footprints for the cpc-1 transcript (Fig. 5) show that uORF1 and uORF2 are heavily translated under these conditions. The frame information obtained with protected fragments of at least 28 nt using a 15-nt offset to the ribosome A site agrees with the predictions that, relative to the CPC1 frame, uORF1 is in frame 2 and uORF2 is in frame 3. The main CPC1 coding region contains ribosomes in the predicted reading frame (frame 1) as expected. Importantly, ribosome footprints in the 5′ region of the transcript outside uORF1, uORF2, and CPC1, especially between uORF2 and CPC1, were all in the CPC1 frame. This is consistent with the in vitro data showing in-frame translation upstream of the main CPC1 coding region. Furthermore, ribosome footprint data, with certain preparation protocols, show accumulation of footprints at AUG and non-AUG start codons (20, 29). This is also the case in the data set that was used for the present analysis of N. crassa. Accumulated footprints were observed at uAUG1 and uAUG2 and at four of the eight predicted NCCs (Fig. 5).

• Open in new tab
• Download powerpoint

FIG 5 

Ribosome profiling evidence for translation of the cpc-1 N-terminal extension. The ribosome-protected fragment count (cutoff at 50) for each position 5′ to 3′ along the cpc-1 mRNA is shown on top. Peaks that can be assigned to specific initiation codons, as labeled in Fig. 1B and K, are indicated in red. A schematic map of the ORF organization of cpc-1, drawn to scale, is shown below the ribosome-protected fragment count. The uORFs are represented by blue rectangles. The eight in-frame NCCs upstream of uORF2 are represented by magenta bars. The mORF is represented by a black rectangle. The reading frame of each feature relative to the mORF is indicated. The reading frame information derived from the ribosome-protected fragments for 5 specific regions along the cpc-1 mRNA, as well as the tabulated data for all N. crassa coding sequences (CDSs), is shown at the bottom. The tabulation is done by scoring the first nucleotide of each ribosome-protected fragment relative to the reading frame of the annotated coding sequence (first to last ribosome A-site codons), using a 15-nucleotide 5′ offset—the proportions of total fragments mapped to frame 1 are shown by green bars, those mapped to frame 2 are shown by red bars, and those mapped to frame 3 are shown by yellow bars. The number of reads used to tabulate the data in each histogram is indicated as “N.”

As controls, we examined ribosomes in the 5′ leaders of arg-2 (NCU07732) and eif5 (NCU00366), which are two other N. crassa mRNA transcripts that contain uORFs (31, 39). In addition to the ribosome footprints in the main ORFs that preferentially corresponded to the predicted reading frame, ribosomes are observed in the uORFs (Fig. S9). The frame information of the footprints in the main ORFs of the uORFs matches the predictions based on the gene model of the corresponding mRNAs. The footprints in the 5′ leaders outside the uORFs appear not to be spurious. Positions represented by more than 2 footprints correspond to near-cognate start codons. For example, each of the six larger peaks 5′ of the first uORF of eIF5 precisely (i.e., with 1-nt resolution) matches six NCCs—AUC, UUG, CUG, GUG, UUG, and UUG, respectively.

Although the ribosome profiling experiment described above provided strong evidence for in vivo translation of the region upstream of the mAUG in the CPC1 frame, it did not and could not address the question of whether translation of this region increases or decreases under stress conditions. The answer is important to address the question of whether the N-terminal extension is part of the translational regulation of cpc-1 or whether it merely provides a constitutive alternative isoform of cpc-1. To address this question, a second ribosome profiling experiment was performed. In it, N. crassa cells were grown in the presence or absence of 3-amino-1,2,4-triazole (3AT), which induces histidine starvation. Compared to untreated cells, as expected, 3AT cells showed elevated density in the mORF compared to uORF2, with the ratio of the ribosome footprint counts in the two regions changing from 0.57 in untreated cells to 1.66 in 3AT-treated cells. However, the ratio of the ribosome footprint count in the region between uORF2 and the mORF relative to the footprint count in the mORF remains nearly unchanged—0.19 in untreated cells to 0.21 in 3AT-treated cells. This result is consistent with the idea that amino acid starvation induces translation at the non-AUG codons of cpc-1 responsible for initiation of the N-terminal extension. Consistent with this notion, the ratio of ribosome footprint count in the region between uORF1 and uORF2, where non-AUG initiation of the N-terminal extension must occur, to the footprint count in uORF2 increases in 3AT-treated cells compared to control cells—from 0.11 to 0.24.