The exosome targets a huge variety of RNA substrates, but in most cases it remains unclear how RNAs are specifically targeted to the distinct enzymatic activities associated with the complex. To better understand exosome targeting in budding yeast, we performed highly sensitive in vivo RNA-protein crosslinking studies (CRAC), coupled with deep sequencing, on the exosome-associated nucleases Rrp44 and Rrp6, two structural subunits Rrp41 and Csl4, and the exosome cofactor Trf4.
Increased relative crosslinking to pre-tRNA was seen for the Rrp44-exo mutant relative to wild-type Rrp44. This strongly indicates that the highly structured pre-tRNAs are substrates for the exonuclease activity of Rrp44, which become “stuck” in the mutant. In contrast, pre-mRNAs and mRNAs, which are expected to be relatively unstructured compared to Pol III transcripts, were under-represented in Rrp44-exo data sets. Northern analysis revealed the accumulation of pre-tRNAs in Rrp44-exo strains, whereas levels of mature tRNAs were not clearly affected. This is consistent with reduced surveillance, rather than impaired processing of pre-tRNAs (Wlotzka et al., 2011
). Pre-tRNAs did not clearly accumulate in rrp6Δ
single mutant strains (data not shown and (Copela et al., 2008
)), but combinatorial loss of Rrp6 and Trf4 strongly amplified the accumulation of pre-tRNAs relative to the absence of Trf4 alone (Copela et al., 2008
). We predict that Rrp6 plays a major role in pre-tRNA surveillance in vivo, but this is redundant with the exonuclease activity of Rrp44 and the core exosome.
Other Pol III transcripts, 5S rRNA, U6 snRNA and scR1 were also preferentially crosslinked to Rrp44-exo, as well as to Rrp6 and Trf4 (Wlotzka et al., 2011
). This suggests that Rrp44 and Rrp6 directly cooperate to degrade these RNAs, aided by the TRAMP complex. Supporting this idea, the 3′ truncated form of the 5S rRNA (5S
) seen in Rrp44-exo strains also accumulated in strains lacking Trf4 or Rrp6 (Kadaba et al., 2006
) and when interactions between Rrp6 and the core exosome were impaired (Callahan and Butler, 2010
). Rrp44-exo strains accumulated mature and 3′ extended (up to 3nt) spliceosomal U6 snRNA, together with 3′ truncated forms of the 5S rRNA (5S
) and scR1 (scR1
) (Copela et al., 2008
; Kadaba et al., 2006
). Oligoadenylated fragments derived from these RNAs were strongly enriched among Rrp44-exo targets, consistent with Trf4 crosslinking (this study and Wlotzka et al., 2011
), strongly indicating that these are surveillance rather than processing targets.
Together, the data suggest that wild-type cells produce excess Pol III transcripts, which are normally turned over by Rrp44 and other nuclear 3′ exonucleases including Rrp6 (Callahan and Butler, 2008
; Copela et al., 2008
; Kadaba et al., 2006
; Schneider et al., 2007
). Recent transcriptome-wide tiling microarrays and pulse-chase labeling of pre-tRNAs indicate that more than 50% of tRNA gene transcription fails to generate mature, functional tRNAs (Gudipati et al., 2012
). A major pathway of exosome-mediated pre-tRNA turnover that competes with tRNA maturation would be consistent with our crosslinking results. Persistent binding of pre-tRNAs to the exosome in the absence of Rrp44 exonuclease activity very likely contributes to the impaired growth and RNA processing in Rrp44-exo strains. The recent finding that ~10% of patients suffering from multiple myeloma carry Rrp44-exo mutations (Chapman et al., 2011
) suggests that either increased synthesis of RNA Pol III products, or the resulting impaired RNA surveillance can induce malignant transformation in human cells.
Nuclear pre-mRNAs and cytoplasmic mRNAs are both targets for the core exosome, whereas the activity of Rrp6 is predicted to be specific for the nuclear RNAs (reviewed in (Houseley and Tollervey, 2009
)). However, these species cannot readily be distinguished in short sequence reads, other than by the presence of the intron. The clearest distinction is therefore the comparison between intron-exon boundaries (IE), which must be part of the unspliced pre-mRNA, and exon-exon boundaries (EE), which can only be present in the spliced mRNA. Among Rrp6 targets, IE hits were around 2 fold more numerous than EE hits, strongly supporting a role for Rrp6 in pre-mRNA surveillance. Consistent with this, analysis of the distribution of Rrp6 reads across spliced genes shows clear enrichment over introns. Cluster analyses of mRNAs showing preferential enrichment in the Rrp6 data sets identified spliced pre-mRNAs but, surprisingly, also found many ribosome synthesis factors. These mRNAs may undergo a significant level of nuclear degradation, possibly as a regulatory mechanism.
Fully functional Rrp44 showed a lower level of sequences over IE boundaries and lower total read coverage over introns, however, both were very substantially increased in the Rrp44-exo mutant. This indicates that Rrp44 is normally actively engaged in degradation of unspliced or partially spliced pre-mRNAs, but these are rapidly and efficiently cleared with little time for crosslinking. Rrp44 showed a high level of crosslinking at the 5′ ends of pre-mRNAs and preferential binding to 5′ splice sites relative to 3′ splice sites. Degradation by the exosome is dependent on cofactors, which must bind 5′ to the complex. Increased crosslinking in the 5′ region may therefore reflect loss of these cofactors leading to slowed degradation.
The Rrp44 sequence coverage over the exons of genes that contain long introns (mainly ribosomal protein genes) was strikingly higher than over genes with shorter introns. This is in agreement with the observation that pre-mRNAs with long introns are preferentially stabilized by loss of Rrp44 function (Gudipati et al., 2012
), clearly showing that these are more subject to degradation by the exosome. Whether this is related to the regulated splicing reported for ribosomal protein pre-mRNAs (Pleiss et al., 2007
) remains to be determined.
Other Pol II transcripts that are largely degraded in the nucleus include CUTs and SUTs. These transcript classes each showed similar sequence coverage for core exosome and Rrp6. SUTs were designated as “stable un-annotated transcripts” based on a lack of apparent stabilization in the absence of Rrp6 (Xu et al., 2009
). However, the similar crosslinking patterns of CUTs and SUTs, and recent microarray analyses in exosome mutant strains (Gudipati et al., 2012
), indicate that their degradation pathways are more closely related than their names suggest.
Close functional interactions between Rrp44 and Rrp6 presumably help explain why strains lacking Rrp44 exonuclease activity survive. Although the CTD domain of the Rrp44-exo protein may be tightly and non-productively associated with substrates, the endonucleolytic activity in the N-terminal PIN domain (NTD) of Rrp44 presumably remains competent to cleave these RNAs, providing free 3′ ends for Rrp6 and other exonucleases. Consistent with this model, the split-CRAC data revealed that exonuclease and endonuclease activities of Rrp44 usually act together to degrade RNA substrates. Mapping of the relative binding sites of the Rrp44 NTD and CTD regions combined with analyses of oligoadenylated substrates, leads us to propose a model (D) for the role of the PIN domain in releasing stalled exosome substrates. In Rrp44-exo strains, RNAs will be degraded inefficiently, but will still be released from the core exosome by PIN domain cleavage and presented to Rrp6 or other nucleases. In the Rrp44-endo-exo double mutant these substrates may be permanently bound to Rrp44, leading to the accumulation of gridlocked exosome complexes and non-functional RNAs in the cell.
In contrast to pre-tRNAs, other highly structured RNAs that were strongly crosslinked to Rrp6 often showed very few hits in Rrp41 and Csl4 data sets, suggesting that they interact only with Rrp44 and Rrp6, with little or no contact to the remaining core exosome. This was unexpected because in vitro data indicated that many substrates are channeled to Rrp44 through the catalytically inert exosome barrel (Bonneau et al., 2009
). Instead, the in vivo crosslinking data on structured RNAs suggest the use of an alternative entry site to the Rrp44 catalytic center, without contacts to the exosome barrel. Such an alternative entry site can be fitted onto the Rrp44-Rrp41-Rrp45 crystal structure (Bonneau et al., 2009
). We therefore hypothesize that at least some in vivo substrates are not threaded through the exosome channel. Instead, they could be docked to Rrp44 from the outside of the complex, aided by tethering to Rrp6 and other exosome cofactors. The basis for this distinction remains unclear, but a long (~33 nt) single-stranded region is required to access the exonuclease domain of Rrp44 via the exosome lumen, whereas much shorter single-stranded regions would be sufficient for direct access to the catalytic sites of Rrp44 or Rrp6.
Despite the apparent cooperation of Rrp44 and Rrp6 on many nuclear surveillance substrates, the comparison of crosslinking sites on individual core exosome subunits with Rrp6 also revealed substrates only enriched in Rrp6 data sets, revealing core-independent Rrp6 functions. One such example is the prominent Rrp6 peak in the 5′-half of the mature 18S rRNA (B), which also coincides with a peak of crosslinking by Trf4 (Wlotzka et al., 2011
). This corresponds to an RNA polymerase I pause site, at which R-loop formation leads to RNase H cleavage of the nascent transcript (El Hage et al., 2010
). We conclude that Rrp6 and the TRAMP complex degrade the cotranscriptionally truncated Pol I primary transcript independently of the core exosome. Rrp6 was reported to localize to the rDNA, interacting with the Nrd1/Nab3 heterodimer and the transcription elongation factors Spt4 and Spt5 (Leporé and Lafontaine, 2011
). We therefore speculate that Rrp6 is specifically recruited to the elongating Pol I to survey nascent rRNA transcripts.