Although tRNAs require CCA at their 3′ ends for amino acid attachment and correct positioning in the ribosome, CCA is not encoded in nearly all eukaryotic tRNA genes or in many archaeal and bacterial tRNA genes (
1-
4). Instead, the CCA-adding enzyme post-transcriptionally adds CCA to the 3′ ends of tRNAs and tRNA-like transcripts (
5-
8). Recent work has identified two tRNA-like small RNAs, mascRNA (MALAT1-associated small cytoplasmic RNA) and the MEN β tRNA-like small RNA, that are generated by 3′ end processing of long nuclear-retained noncoding RNAs in human and mouse cells (
9-
11) (). These RNAs are ~70% similar in sequence and are generated by enzymes involved in canonical tRNA biogenesis, including RNases P and Z. Although the long noncoding RNAs from which they are processed are expressed at roughly equal levels, the MEN β tRNA-like small RNA is below the threshold of detection by Northern blot analysis () (
12). This suggests that these tRNA-like transcripts are differentially regulated post-transcriptionally. Using a more sensitive 3′ rapid amplification of cDNA ends (RACE) PCR approach, we detected expression of the MEN β tRNA-like small RNA
in vivo (
Fig. S1). Surprisingly, nearly all sequenced transcripts ended in CCACCA or CCACC, which is not encoded in the genome. This 3′ end modification was not detected on mascRNA (
9) and thus would be consistent with CCACCA addition signaling RNA degradation.
Although the CCA-adding enzyme is thought to terminate after CCA synthesis (
7,
8), it was the most likely candidate to catalyze CCACCA addition. We thus expressed and purified His-tagged versions of the CCA-adding enzyme from human (
13),
Escherichia coli (
14), and
Sulfolobus shibatae (
15) (
Fig. S2, A to C) and confirmed that all three enzymes terminate polymerization once CCA has been added to canonical tRNAs (
Fig. S2, D to F). In contrast, CCACCA or CCACC was added to the MEN β tRNA-like small RNA
in vitro (
Fig. S3), recapitulating the
in vivo 3′ RACE results. This indicates that the CCA-adding enzyme catalyzes CCA or CCACCA addition depending on the characteristics of the RNA substrate.
To determine the sequence elements required for CCACCA addition, we investigated converting mouse mascRNA from a CCA to a CCACCA target by generating chimeric mascRNA-MEN β tRNA-like transcripts ( and
fig. S4A). Swapping the acceptor stem (Mut 1) was sufficient to convert mascRNA from a CCA to a CCACCA target
in vitro with near 100% efficiency, whereas the other arms had no effect on CCA addition ( and
fig. S4). CCA-adding enzymes from all three kingdoms of life added CCACCA to the Mut 1 transcript, despite the eukaryotic () and eubacterial enzymes (
Fig. S4C) having a different catalytic mechanism from archaeal enzymes () (
16-
20). Substituting the acceptor stems of canonical tRNAs with that of the mouse MEN β tRNA-like small RNA similarly converted those transcripts to CCACCA targets
in vitro (
Fig. S5A). This indicates that the CCA-adding enzyme monitors the acceptor stem when selecting between CCA and CCACCA addition. Furthermore, the CCA-adding enzyme functions as a CCACCA repair enzyme as it can, for example, add an additional A to a substrate ending in CCACC (
Fig. S5B).
To investigate the features in the mouse MEN β acceptor stem that signal CCACCA addition, we took advantage of the fact that this stem is poorly conserved across species ( and
fig. S6). When the mascRNA acceptor stem was swapped with that of the human MEN β tRNA-like transcript, CCACCA addition was also observed
in vitro (). The end of the acceptor stems of the mouse and human MEN β transcripts are significantly unstable due to a C-A mismatch or multiple G-U wobbles, respectively. In contrast, other MEN β homologs have significantly more stable acceptor stems and were CCA targets
in vitro. Using COS7 cells, we confirmed that the Old World monkey MEN β homolog is a CCA target and stable transcript
in vivo that is easily detected using Northern blot analysis (
Fig. S7).
By extensively mutagenizing mascRNA and canonical tRNAs, we determined that CCACCA addition requires a transcript to have an unstable acceptor stem and guanosines at the first and second positions (See SOM,
Fig. S8 to S11). Instability facilitates isomerization of the acceptor stem after the first CCA is added, allowing the Cs of the CCA sequence to base pair with G
1G
2 (). The A of the CCA sequence becomes the new discriminator base and is positioned in the enzyme's catalytic pocket, ready to undergo a second round of CCA addition using the well-established polymerization mechanisms (
8,
17-
19). As to why polymerization stops once CCACCA has been added, we propose that additional unpaired nucleotides in the bulge interfere with enzyme recognition.
Could addition of CCACCA be a common reaction for many tRNAs? Between 30 and 45 percent of tRNA genes in a given species begin with G
1G
2, indicating a strong enrichment for GG at 5′ ends ( and
fig. S12). However, it is unlikely that most of these tRNAs are subjected to CCACCA addition under normal conditions as they have stable acceptor stems. Single point mutations that cause a mismatch or an additional G-U wobble in the acceptor stems of two unrelated arginine tRNAs ( and
fig. S13) or a cysteine tRNA (
Fig. S14) were sufficient to completely convert these transcripts from CCA to CCACCA targets
in vitro. Mutant tRNAs ending in CCACCA were rapidly degraded in HeLa nuclear extracts ( and
fig. S15), indicative of an efficient decay mechanism that is distinct from cleavage in the anticodon loop (
21).
The rapid tRNA decay (RTD) pathway functions in yeast as a surveillance mechanism to degrade a subfraction of mature tRNAs (
22-
24). Originally identified as a mechanism to degrade specific hypomodified tRNAs, the RTD pathway also degrades tRNAs containing mutations that destabilize their structure (
25). We noticed that tRNAs sensitive to RTD often conform to the rules for CCACCA addition as they tend to start with GG and have slightly destabilized acceptor stems due to a single G-U wobble near their end (). We, therefore, hypothesized that CCACCA may be added to tRNAs being degraded by RTD
in vivo.
Here, we focused on tRNA
Ser(CGA) and tRNA
Ser(UGA), two highly similar species that are rapidly degraded in a
trm44-Δ tan1-Δ strain grown at high temperatures due to the lack of 2′-O-methylation of U
44 and N
4 acetylation of C
12 (
24). By combining 3′ RACE PCR and deep sequencing (
Fig. S16), we detected a low, but reproducible, level of extended CCA motifs (defined as CCAC, CCACC, or CCACCA) on tRNA
Ser(CGA) and tRNA
Ser(UGA) in a wild type strain grown at 28°C or 37°C ( and
tables S1-2). Consistent with a role for CCACCA addition in this proofreading pathway, switching the
trm44-Δ tan1-Δ strain to growth at 37°C triggers RTD of tRNA
Ser(CGA) and tRNA
Ser(UGA) and a progressive increase in transcripts ending in extended CCA motifs, which mirrors the time course of degradation (
24). Interestingly, we also detected a significant increase in tRNAs with short poly(A) tails added to their CCA ends when RTD was occurring (
Fig. S17 and tables S3-4). Deleting the 5′-3′ exonuclease Xrn1 or Met22, a phosphatase whose substrate is an Xrn1 inhibitor, prevents tRNA degradation by RTD (
23) and, likewise, significantly decreased the level of extended CCA motifs and poly(A) tails observed ().
As increasing the stability of the acceptor or T-stems of tRNA
Ser(CGA) was shown to decrease RTD sensitivity (
25), we asked if strengthening the tRNA tertiary structure would also affect CCACCA addition levels
in vivo. Changing the U
51:U
63 mismatch to a G
51:C
63 base pair near completely abolished CCACCA addition, whereas substitution of U
6:G
67 by U
6:A
67 decreased CCACCA levels by ~60%, consistent with its more modest increase in predicted stability ( and
table S2).
We, therefore, suggest that CCACCA or poly(A) addition generates a single-stranded tail that is sufficiently long to be recognized by 3′-5′ exonucleases. Indeed, tRNAs ending in CCACCA, but not CCA, were degraded
in vitro by RNase R from
E. coli (
26) (
Fig. S18A) and by yeast Rrp44, the active 3′-5′ exonuclease component of the exosome (
27) ( and
fig. S18B). Consistent with the genetics and deep sequencing, we found that yeast Xrn1 cooperates with Rrp44 to degrade tRNAs ending in CCACCA
in vitro ( and
fig. S18C). Rrp44 often pauses once it nears the double-stranded acceptor stem (
Fig. S18C) and we propose that Xrn1 recruitment allows efficient degradation of the rest of the transcript. In the absence of Xrn1, it is likely that the CCA-adding enzyme repairs the 3′ end of the tRNA.
Our results suggest a model in which the CCA-adding enzyme adds CCACCA to tRNAs with unstable structures and a 5′ terminal GG sequence, preventing aminoacylation and marking them for degradation (
Fig. S19). Many events can destabilize a tRNA, including transcriptional errors, nucleotide modifications that alter base pairing strengths, or conformational changes caused by the binding of proteins or other RNAs
in trans. In addition, some tRNAs have genomically encoded mismatches that are corrected by post-transcriptional editing events (
28,
29), a process likely proofread by the CCA-adding enzyme as the unedited transcripts are CCACCA targets (
Fig. S20). Considering that CCACCA addition is conserved across all three kingdoms of life, surveillance by the CCA-adding enzyme likely represents a universal mechanism for ensuring accurate protein synthesis.
The publisher's final edited version of this article is available at 
