Have tRNA, will travel

The processing of tRNA is deceptively complicated from the standpoint of both biochemistry and spatial organization. Every generation in yeast, ≈3–6 million tRNAs are processed (1), ≈10 times the number of rRNAs and ≈60–100 times the number of mRNAs (2). Processing of tRNAs requires ≈60 polypeptides to catalyze removal of the 5′ leader sequence, trimming of the 3′ trailer sequence, splicing of the introns that occur in ≈22% of tRNA genes, addition of nucleotides C, C, and A (CCA) to the 3′ end of tRNA, and formation of the 25 different chemical modifications that occur in different sets of tRNAs (3). The complexity of tRNA processing has been matched recently by an equally complicated set of results regarding the cellular location of tRNA processing events. It was once almost dogma that tRNA processing occurred in the nucleus, followed by export and tRNA utilization in the cytoplasm. However, essential steps of tRNA processing are now known to occur in the cytoplasm, and tRNA charging can occur in the nucleus. This issue of PNAS features the latest startling chapter in the storied travels of tRNA during the course of its life: evidence that tRNA can flow back to the nucleus from its site of action in the cytoplasm (4). This paradigm-breaking work from the laboratory of Anita Hopper explains several previously puzzling observations and raises a number of new questions.

Problems with the Classical View of tRNA Processing

The tRNA processing pathway was once thought to be direct and rather simple. Based on a classic set of microinjection experiments in Xenopus oocytes, an elegant picture was painted in which transcription began in the nucleus, followed by 5′ and 3′ end processing, then by splicing of introns for those tRNA species that had them and subsequent export of the tRNA to the cytoplasm (5). This was accompanied by the addition of multiple modifications in either the nucleus or the cytoplasm at various stages of the maturation process (6). It is now known that tRNA processing begins in the nucleolus, where tRNA genes are unexpectedly clustered despite their disparate locations in the genome (7), that early steps of tRNA processing such as 5′ leader removal occur in the same compartment (8), and that multiple other modifications occur in the nucleus, the nuclear rim, or the cytoplasm, more or less consistent with the early processing model (3).

However, several other observations suggest unusual organization of the tRNA processing machinery. One early puzzling observation was that components of the nuclear transport system were somehow linked to splicing. Thus, unspliced but end-matured tRNA species accumulate in strains with mutations in a number of nuclear pore components (9). This finding led to an adaptation of the earlier model, positing that tRNA splicing was tightly coupled to tRNA export. Coupling of splicing and export of tRNA could also conveniently explain the role of two proteins, Rna1 and Los1, mutations in which resulted in the accumulation of unspliced tRNA (10, 11). Rna1 protein was shown to be the GTPase activating protein (GAP) of the Ran cycle (12, 13), and Los1 protein was shown to be a tRNA exportin (14–16).

tRNA can flow back to the nucleus from its site of action in the cytoplasm.

A second unsettling observation was that tRNA splicing could occur before 5′ and 3′ end processing, which was difficult to reconcile with coupled splicing and export. Thus, steady-state analysis of tRNA processing in Saccharomyces cerevisiae demonstrated the existence of at least one species of tRNA that was spliced but not end-processed (17). Similarly, disruption of a nuclear retention element of Schizo-saccharomyces pombe La protein, a protein required for tRNA 3′ end processing (18), led to the accumulation of spliced tRNA precursors that had not been processed at their 5′ and 3′ ends (19). Indeed, in Xenopus oocytes, a reexamination of tRNA processing established that splicing naturally occurs before end processing under physiological conditions (20).

The demonstration that nuclear aminoacylation of tRNA was important for tRNA export added a new dimension to tRNA processing (20), giving birth to the idea that tRNA proofreading was effected by aminoacylation of mature tRNA before export. Aminoacylation also proved to be important for tRNA export in yeast, as demonstrated by analysis of tRNA synthetase mutants, amino acid starvation, and addition of inhibitors of synthetases (21, 22). An added element of tRNA quality control in the yeast nucleus is the recently discovered nuclear tRNA degradation system that acts by polyadenylation of the tRNA, followed by exonucleolytic cleavage of tRNA (23). As a consequence of these two systems, tRNA is either aminoacylated and exported or polyadenylated and degraded.

The starting point for the article by Shaheen and Hopper (4) is the surprising result of Yoshihisa et al. (24) that at least two of the four subunits of the splicing endonuclease, Sen2 and Sen54, are located on the outer mitochondrial surface. This mitochondrial location is required for functional tRNA splicing, and the unspliced tRNAs in a sen2 mutant accumulate in the cytoplasm. The finding of splicing in the cytoplasm provides a simple explanation for the nuclear accumulation of unspliced pretRNAs when nuclear export is blocked, obviating the need for coupled splicing and export. However, splicing in the cytoplasm is difficult to reconcile with the occurrence of splicing before end processing and with nuclear aminoacylation as a proofreading step before export.

Flow of tRNA to the Nucleus Resolves Some Problems

The article by Shaheen and Hopper (4) addresses the question of how the action of a cytoplasmic endonuclease can result in the accumulation of spliced tRNAs in the nucleus under conditions in which tRNA charging is expected to be compromised. Shaheen and Hopper address two possible explanations: movement into the nucleus of the endonuclease or of the tRNA. They first show that the endonuclease does not become localized to the nucleus under amino acid starvation conditions or in tRNA^Tyr synthetase mutants, the growth conditions in which mature spliced tRNAs accumulate in the nucleus. They then assess tRNA movement from the cytoplasm to the nucleus (retrograde flow) by using a mutant strain of yeast that is defective in karyogamy, the fusion of two nuclei after mating, and a marked tRNA present in only one parental strain (in this case, a functional tRNA^Glu-D from Dictyostelium). In the resulting mated strains, there are two nuclei (called heterokaryons), and only if tRNA in the cytoplasm moves to the nucleus will it be detected in both nuclei. By using this system, tRNA^Glu-D was found in both nuclei in heterokaryons cells in either of two conditions used to generate nuclear tRNA pools: lack of the tRNA exportin Los1 or amino acid starvation. This result clearly establishes the occurrence of retrograde flow of tRNAs from cytoplasm to nucleus. A similar set of results has also recently been published by Takano et al. (25).

Shaheen and Hopper (4) demonstrate further that retrograde tRNA flow is driven by the Ran cycle and a specific importin, Mtr10. Mutants in the Ran-GAP Rna1p impair retrograde flow of tRNA to the nucleus. Moreover, mutation of Mtr10p, but not of each of the eight other nonessential importins, also blocks nuclear accumulation of endogenous tRNAs during amino acid starvation. Although Takano et al. (25) differ in their conclusions regarding the role of Ran in this process, further study of the two experimental systems will undoubtedly lead to new insights.

As described by Shaheen and Hopper (4), retrograde flow of tRNA might occur for either of two reasons. First, it might be part of a constitutive tRNA processing pathway that occurs after splicing. If so, and if aminoacylation is an important proofreading step before use of tRNA, then tRNAs that are spliced presumably require not one but two export steps: one for splicing and one after retrograde flow for subsequent aminoacylation. Los1 protein is implicated in both export processes: the accumulation of unspliced pretRNAs in the nucleus of los1-Δ cells attests to its role in the first export step, and the accumulation of intronless tRNA^Glu-D in both nuclei of heterokaryons implicates it directly in the reexport step. Thus, the myriad of genetic interactions involving los1 mutants will require reexamination for their effect on both processes (15, 22, 26–28). Retrograde flow of tRNA would also presumably require a protein to escort the tRNA to the mitochondria after its emergence in the cytoplasm and perhaps back to the importin Mtr10 after splicing. Either of these escort roles might be fulfilled by La protein, based on the occurrence of spliced but 5′ and 3′ unprocessed pretRNAs in La mutants (19). A constitutive tRNA retrograde pathway would also presumably shunt nonfunctional tRNAs to the nucleus for degradation (4), providing a convenient method for disposal of tRNA by the nuclear tRNA degradation system (23).

Retrograde tRNA flow might be part of a regulatory response to physiological signals.

Second, retrograde tRNA flow might be part of a regulatory response to physiological signals such as amino acid starvation. This is an attractive model, although clearly not all tRNAs that are uncharged would end up in the nucleus, because some such tRNAs are known to act in the cytoplasm to activate Gcn2 and derepress the Gcn4 general amino acid control pathway (29). A systematic search for other mutants that accumulate spliced but 5′ and 3′ unprocessed tRNAs in the cytoplasm or processed tRNAs in the nucleus would help to delineate the nature of the pathway(s) or functions of tRNA retrograde flow. In this regard, it would be useful to have a method to follow biogenesis of tRNA in a quantitative and explicit way, because a number of conclusions about tRNA processing rely on steady-state analysis of species in various mutants.

Understanding the biochemistry of tRNA processing has led to its share of surprises, including the discovery of RNA catalysis by RNase P (30), of a new metabolite of nicotinamide adenine dinucleotide (NAD) formed during tRNA splicing (31), of multiple ligation pathways of tRNA splicing (32), of tRNA splicing enzymes that splice mRNA (33), of the only known polymerase that is template-independent and sequence-specific (34), and of an enzyme that can add a nucleotide to a polynucleotide chain to extend it in the reverse (3′→5′) direction (35). Clearly the location of tRNA processing events has had as big an impact as the biochemical steps themselves on the biology of tRNA. The articles by Shaheen and Hopper (4) and by Takano et al. (25) have gone a long way toward explaining the consequences of cytoplasmic tRNA splicing as part of the larger question of understanding the spatial flow of tRNA in the cell.