Exosome cofactor hMTR4 competes with export adaptor ALYREF to ensure balanced nuclear RNA pools for degradation and export

Abstract

The exosome is a key RNA machine that functions in the degradation of unwanted RNAs. Here, we found that significant fractions of precursors and mature forms of mRNAs and long noncoding RNAs are degraded by the nuclear exosome in normal human cells. Exosome‐mediated degradation of these RNAs requires its cofactor hMTR4. Significantly, hMTR4 plays a key role in specifically recruiting the exosome to its targets. Furthermore, we provide several lines of evidence indicating that hMTR4 executes this role by directly competing with the mRNA export adaptor ALYREF for associating with ARS2, a component of the cap‐binding complex (CBC), and this competition is critical for determining whether an RNA is degraded or exported to the cytoplasm. Together, our results indicate that the competition between hMTR4 and ALYREF determines exosome recruitment and functions in creating balanced nuclear RNA pools for degradation and export.

Introduction

The process of eukaryotic gene expression involves many distinct surveillance mechanisms. RNAs aberrantly processed or that fail to assemble into export‐competent ribonucleoprotein complexes (RNPs) are detected and degraded in the nucleus by the surveillance machines. One of the major nuclear RNA surveillance machines is the exosome complex, which has both 3′–5′ exo‐ and endo‐ribonuclease activities (Mitchell et al, 1997; Allmang et al, 1999b; Schneider et al, 2007; Lebreton et al, 2008).

The yeast exosome functions in a wide range of processes including processing of rRNAs, snoRNAs, and snRNAs; degradation of RNA processing intermediates; degradation of various aberrantly processed RNAs as well as normal pre‐mRNAs and pre‐tRNAs; mRNA turnover; surveillance of aberrant mRNAs; regulating degradation of stable and unstable long ncRNAs such as CUTs and SUTs; and transcription termination of backtracked RNA polymerase II (Allmang et al, 1999a; Zanchin & Goldfarb, 1999; Bousquet‐Antonelli et al, 2000; van Hoof et al, 2000; Libri et al, 2002; Das et al, 2003; Kadaba et al, 2004; Milligan et al, 2005; Wyers et al, 2005; Reis & Campbell, 2007; Wang et al, 2008; Lemieux et al, 2011; Gudipati et al, 2012; Schneider et al, 2012; Lemay et al, 2014). Relatively, functions of the human exosome are less well characterized. Most studies have been focused on its roles in the degradation and/or transcription termination of promoter upstream transcripts (PROMPTs) and enhancer RNAs (eRNAs), processing of rRNAs, and degradation of pre‐mRNAs containing snoRNA‐ or miRNA‐coding introns, processed transcripts, and hyperadenylated intronless mRNAs, as well as mRNA surveillance in the cytoplasm (Schneider et al, 2007; Preker et al, 2008; Lubas et al, 2011, 2013, 2015; Andersen et al, 2013; Bresson & Conrad, 2013; Hallais et al, 2013; Andersson et al, 2014; Meola et al, 2016).

The full in vivo function of the exosome requires many cofactors. The cofactor Mtr4p is essential for almost all the yeast nuclear exosome activities. Mtr4p alone can affect exosome functions (e.g., in the 3′ processing of 5.8S rRNAs); however, most of its activities are carried out in the context of the TRAMP complex (TRAMP), which consists of Mtr4p, the noncanonical polyA polymerase Trf4p/Trf5p, and the zinc knuckle proteins Air1p/Air2p (LaCava et al, 2005; Houseley & Tollervey, 2006). Trf4p/Trf5p functions in adding a short polyA tail to exosome target RNAs that is thought to facilitate exosome recruitment (LaCava et al, 2005; Vanacova et al, 2005; Wyers et al, 2005). Cofactors of the human exosome are much more complicated. In the nucleolus, hMTR4 associates with PAPD5 and ZCCHC7, homologues of yeast Trf4p/Trf5p and Air1p/Air2p, to form the putative human TRAMP that is involved in the adenylation of rRNA degradation intermediates and selective degradation of viral RNAs (Shcherbik et al, 2010; Lubas et al, 2011; Molleston et al, 2016). In the nucleoplasm, hMTR4 interacts with the zinc knuckle protein ZCCHC8 and the RNA‐binding protein RBM7 to form the NEXT complex (NEXT) (Lubas et al, 2011), which functions in the degradation of PROMPTs and 3′ end extended forms of snRNAs and snoRNAs (Lubas et al, 2011; Andersen et al, 2013; Hallais et al, 2013; Hrossova et al, 2015). Further, NEXT associates with the cap‐binding complex (CBC) and functions in linking transcription termination to exosomal degradation (Andersen et al, 2013; Hallais et al, 2013), and in degrading pre‐mRNAs containing snoRNA‐ or miRNA‐coding introns (Lubas et al, 2015). The most recent work revealed the polyA tail exosome targeting (PAXT) connection, which comprises hMTR4, ZFC3H1, and the nuclear polyA‐binding protein PABPN1, is involved in the degradation of long processed transcripts (Meola et al, 2016).

Accumulating evidence indicates that extensive nuclear mRNA degradation occurs when mRNA export is blocked. In yeast, mutations in THO subunits and Sub2, components of the mRNA export complex TREX, result in rapid mRNA degradation that requires the exosome as well as TRAMP (Rougemaille et al, 2007; Saguez et al, 2008). Likewise, in Drosophila, nuclear mRNAs are extensively degraded when mRNA export is blocked by depletion of NXF1, p15, or UAP56 (Herold et al, 2003). In mammalian cells, the level of the export‐defective β‐globin cDNA transcript was significantly elevated upon exosome inactivation (Bresson & Conrad, 2013), raising the possibility that mRNAs with export defect can be detected and degraded by the human exosome. However, how the exosome recognizes these export‐defective RNAs remains to be investigated.

In this work, analysis of nuclear RNAs by RNA‐seq revealed that significant portions of spliced and intronless coding and long noncoding RNAs are nuclear exosome targets in RNA export‐competent human cells. We subsequently investigated the mechanism for exosome recruitment and found that hMTR4 specifically recruits the exosome to its target RNAs. This specificity is achieved through the competition between hMTR4 and ALYREF for interacting with ARS2, a CBC component. Further, our data suggest that this competition ensures balanced nuclear RNA pools for degradation and export.

Results

Genome‐wide study of the exosome function in nuclear RNA degradation

To define the in vivo targets of the human nuclear exosome, we performed stranded RNA‐seq using rRNA‐depleted RNAs isolated from the nuclei of HeLa cells depleted of control, the exosome core component hRRP40, or hMTR4 (Fig 1A). Western analysis data revealed that hRRP40 and hMTR4 were knocked down to ~10%, and confirmed purities of the nuclear fractions, using tubulin as a cytoplasmic marker, and UAP56, hMTR4 as well as hRRP6 as nuclear markers, respectively (Fig 1B and C). We generated ~60, 52, and 40 million 100‐nt tags from control‐, hRRP40‐, or hMTR4‐depleted cells, respectively. RNA‐seq reads were distributed into general genomic categories: mRNAs, long noncoding RNAs (lncRNAs) as well as PROMPTs and eRNAs, short ncRNAs (miRNAs, snoRNAs, tRNAs, snRNAs), repetitive elements, pseudogenes, and “others” (Fig 1D). In all of the three samples, most sequencing reads were from mRNAs, short ncRNAs, and long noncoding RNAs/PROMPTs/eRNAs.

Figure 1. Genome‐wide study of in vivo targets of the human nuclear exosome.

• A
  A diagram of the RNA‐seq experimental approach.
• B
  Western blotting to examine the knockdown efficiencies of hRRP40 and hMTR4. Tubulin was used as a loading control. Different amounts of control knockdown samples were loaded to estimate the knockdown efficiencies.
• C
  Western blotting to examine the purity of nuclear fractions. Nuclear proteins UAP56, hMTR4, and hRRP6 and the cytoplasmic protein tubulin served as the nuclear and cytoplasmic markers, respectively. N, nucleus; C, cytoplasm. The asterisk indicates a nonspecific band that is detected by the hRRP6 antibody.
• D
  The distribution of reads derived from RNA‐seq libraries and mapped to the indicated RNA classes. Each category represents RNAs unique to that category and nonoverlapping with previous categories. In hRRP40 knockdown cells, Fisher's exact test, lncRNA, adjusted P = 0, odds ratio = 1.10; repetitive elements, adjusted P = 0, odds ratio = 1.47; and others, adjusted P = 0, odds ratio = 1.37. In hMTR4 knockdown cells, Fisher's exact test, lncRNA, adjusted P = 0, odds ratio = 1.18; repetitive elements, adjusted P = 0, odds ratio = 1.75; others, adjusted P = 0, odds ratio = 2.06; and short ncRNA, adjusted P = 0, odds ratio = 1.53.
• E
  The Venn diagrams depict the overlap of accumulated snRNAs in hRRP40 and hMTR4 knockdown cells. The numbers in Venn diagrams show snRNAs whose RPM were elevated more than 1.5‐fold in hRRP40 or hMTR4 knockdown relative to control knockdown cells. Note that only sequencing reads mapping to the gene bodies were computed. The percentages of snRNAs that increased more than 1.5‐fold in hRRP40 and hMTR4 knockdown are shown. Statistical analysis P‐value was used to measure the overlapping of genes regulated by hRRP40 and hMTR4 and performed using Fisher's exact test by R language.
• F–L
  Same as (E), except that instead of snRNAs, accumulated snoRNAs, PROMPTs, eRNAs, intron‐containing mRNAs, intronless mRNAs, intron‐containing lncRNAs, and intronless lncRNAs are shown, respectively. Note that relative to mRNAs, more lncRNAs were accumulated in both hRRP40 (26% for mRNAs and 50% for lncRNAs according to exon changes) and hMTR4 (26% for mRNAs and 57% for lncRNAs according to exon changes) knockdown cells. In hMTR4 knockdown, compared to intron‐containing mRNAs, relatively more intronless mRNAs were accumulated (according to exon reads change, compare 43 to 26%, Fisher's exact test P < 0.0001, odds ratio = 1.67).

Source data are available online for this figure.

We computed gene expression based on the number of exonic or intronic reads mapping to each gene, normalized by the total number of reads (RPM) and spike‐in controls (see Materials and Methods for details; Dataset EV1). Consistent with previous studies, significant portions of snRNAs (45; 64%), snoRNAs (27; 57%), PROMPTs (90; 91%), and eRNAs (43; 55%) were accumulated more than 1.5‐fold in hRRP40 and hMTR4 knockdown cells (Allmang et al, 1999a; Preker et al, 2008; Gudipati et al, 2012; Andersen et al, 2013; Lubas et al, 2013, 2015; Ntini et al, 2013; Andersson et al, 2014) (Fig 1E–H). We randomly picked six of these RNAs for validation by RT–qPCRs (Appendix Fig S1). Given that exosomal degradation of these RNAs has been extensively studied and is not the focus of this study, we did not carry out further analysis on them.

Significantly, we found that noteworthy portions of mRNAs and lncRNAs derived from intron‐containing genes were accumulated more than 1.5‐fold in hRRP40 (26% for mRNAs and 50% for lncRNAs according to exon changes) and hMTR4 (26% for mRNAs and 57% for lncRNAs according to exon changes) knockdown cells (Fig 1I and K). Compared to intron‐containing ones, more intronless mRNAs, but not lncRNAs, were accumulated in hMTR4 knockdown cells (according to exon reads change, compare 43 to 26%; Fig 1I–L). Note that in each RNA category, at least 60% of the accumulated RNAs in hRRP40 knockdown were overlapped with those in hMTR4 knockdown cells. Together, these data suggest that precursors and mature forms of mRNAs and lncRNAs might be targets of the nuclear exosome in normal human cells. This study will mainly focus on these two categories of RNAs.

Both precursors and mature forms of mRNAs and lncRNAs are human nuclear exosome targets

To validate our RNA‐seq data, we randomly selected 30 RNAs, either protein coding or noncoding, intron‐containing, or intronless, and examined their levels in the nuclei of control‐, hRRP40‐, and hMTR4‐depleted cells by quantitative or semiquantitative RT–PCRs. Nuclear RNAs from three biological replicates and primers specifically amplifying precursors or mature forms were used (Figs 2 and EV1; Appendix Fig S2). Among tested RNAs, at the precursor and mature form levels, 92 and 98% were consistent with RNA‐seq results, indicating that our RNA‐seq data are reliable. According to the changes in the levels of precursors and mature forms, these RNAs can be divided into five categories (Figs 2 and EV1; Appendix Fig S2). The first category includes RNAs showing increased levels in both precursors and mature forms (Figs 2A and EV1; Appendix Fig S2). These RNA precursors are possibly exosome targets, whereas the elevated levels of mature forms might result either from the processing of stabilized precursors or from directly stabilized mature RNAs. The second category includes RNAs showing increased levels only in precursors, but not in mature forms (Figs 2B and EV1; Appendix Fig S2). The third category includes RNAs showing increased levels only in the mature forms (Figs 2C and EV1; Appendix Fig S2). Probably, spliced RNAs, but not precursors, in this category are exosome targets. The fourth category is formed by significantly accumulated intronless RNAs (Figs 2D and EV1; Appendix Fig S2). Finally, the fifth category represents RNAs whose levels in both precursors and spliced forms are not significantly elevated (Figs 2E and EV1). We estimated the percentage of exosome targets falling in each category according to their changes in exon–intron/intron–exon and exon–exon reads for intron‐containing genes or the single exon for intronless genes (Fig 2F). Together, these data suggest that significant fractions of precursors and mature forms of mRNAs and lncRNAs are targets of the human nuclear exosome.

Figure 2. Five categories of mRNAs based on level change upon exosome knockdown.

1. The levels of both precursor and mature forms are elevated. (Top) RNA‐seq signal of TCTA is shown as an example. The blue and purple lines mark the position of PCR products for pre‐mRNAs and spliced mRNAs, respectively. The intron regions are boxed in color. Numbers to the left show the RPM. (Bottom) RT–qPCRs to specifically amplify pre‐mRNAs and spliced mRNAs in the nuclei of control, hRRP40, and hMTR4 knockdown cells. The bars show RNA levels relative to 18S rRNA. Error bars, standard deviations (n = 3). Statistical analysis was performed using Student's t‐test. *P < 0.05, **P < 0.01.
2. Same as (A), except that in this category, the levels of pre‐mRNAs increase with no apparent change in those of spliced mRNAs.
3. Same as (A), except that in this category, the levels of spliced mRNAs increase with no apparent change in those of pre‐mRNAs.
4. Same as (A), except that in this category, the levels of intronless mRNAs significantly increase.
5. Same as (A), except that in this category, the levels of neither precursors nor mature forms significantly increase.
6. The distribution of exosome target mRNAs and lncRNAs falling in each category.

Figure EV1. Deep‐sequencing signals of mRNAs that were examined in Fig 2 .

1. Deep‐sequencing signals of exosome targets in each category are shown. (I) The levels of both precursor and mature forms are elevated. (II) The levels of pre‐mRNAs increase with no apparent change in those of spliced mRNAs. (III) The levels of spliced mRNAs increase with no apparent change in those of pre‐mRNAs. (IV) The levels of intronless mRNAs significantly increase. (V) The levels of neither precursors nor mature forms significantly increase.
2. Illustration of the position of RT–qPCR products on the indicated gene. The blue and purple lines show the PCR product of pre‐mRNAs and spliced mRNAs, respectively.

We noted that for some RNAs, that is, DNAJC30 and NUDFAF3 (Figs 2D and EV1A), sequencing reads were mainly mapped to a part of them that shows exosome sensitivity. For these genes, we mostly used primers to amplify the unapparently detected parts to confirm that these parts were also stabilized (Figs 2D and EV1; Appendix Fig S2). To further distinguish whether a part or the full length of the mRNAs was stabilized upon exosome knockdown, we designed multiple pairs of primers amplifying both detected and undetected parts of the DNAJC30 and NDUFAF3 mRNAs. RT–qPCR data showed that for both of these mRNAs, similar changes occurred to different parts of the mRNAs (Fig EV2), suggesting that full length of these two mRNAs was stabilized. However, the possibility that only a part of transcripts are sensitive to exosome knockdown remains for RNAs partially detected by RNA‐seq. Some of the stabilization could have resulted from transcription start site change, especially when the changes are mainly detected at the 5′ of the RNAs. Premature transcription termination might have caused the 3′ specific sensitivity to exosome knockdown.

Figure EV2. Full length of the DNAJC30 and NDUFAF3 mRNAs might be stabilized in hRRP40 or hMTR4 knockdown cells.

(Top) Deep‐sequencing signals of DNAJC30 and NDUFAF3 are shown. The plot of RNA‐seq data for DNAJC30 is the same as that in Fig 2D. The blue lines mark the position of qPCR products. (Bottom) RT–qPCRs to examine the nuclear levels of different parts of the mRNAs relative to the 18S rRNA in control, hRRP40, and hMTR4 knockdown cells. Statistical analysis was performed using Student's t‐test. Error bars represent standard deviations from biological repeats (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001.

It was possible that for some particular genes, the increased RNA levels might be due to enhanced transcription. However, RNAPII ChIP‐seq data revealed that among ~15,000 RNAPII binding peaks, only < 30 peaks were significantly increased in hRRP40 or hMTR4 knockdown, relative to those in control cells (Fig EV3 and Datasets EV2‐EV5 and EV7). Interestingly, 29.6 and 8.6% of binding peaks were reduced in hRRP40 and hMTR4 knockdown cells, respectively (Fig EV3; Datasets EV6 and EV8). Around 70% of the genes with reduced RNAPII binding in hRRP40 knockdown cells showed no significant increase in nuclear RNA levels (Fig EV3C), suggesting that the post‐transcriptional impact of hRRP40 might have been underestimated. The reason for these reduced RNAPII binding upon exosome inactivation remains unclear. Nevertheless, these data indicate that the increased nuclear RNA levels are not results of enhanced transcription.

Figure EV3. Transcription elevation does not account for the increased levels of most exosome targets.

1. The pie chart represents populations of genes for which the RNAPII binding was enhanced, remained unchanged, or was reduced.
2. Examples of RNAPII ChIP‐seq data for genes belonging to different categories.
3. The distribution of genes with reduced RNAPII binding in hRRP40 knockdown cells falling in each category according to their level change.

Identification of nuclear accumulated polyA RNAs in hMTR4 knockdown cells

Considering that most mRNAs and lncRNAs are polyadenylated, to confirm the nuclear accumulation of mRNAs and lncRNAs, we carried out another set of RNA‐seq for nuclear polyA RNAs from control and hMTR4 knockdown cells (Fig 3A). To assess experimental variations, this analysis was performed in triplicate. hMTR4 knockdown efficiency and nuclear fraction purities were confirmed by Western analyses (Fig 3B and Appendix Fig S3A). In both control and hMTR4 knockdown cells, most of the sequencing reads were mapped to mRNAs and lncRNAs/PROMPTs (Fig 3C). In total, 6179 RNAs were upregulated more than 1.5‐fold upon hMTR4 depletion in at least two of the three replicates. Significantly, more than 55% of these RNAs overlapped with significantly upregulated ones detected by nuclear total RNA‐seq (Fig 3D and Dataset EV9). Screenshot of an exemplified mRNA in each category is shown in Appendix Fig S3B–F). We note that the populations of accumulated mRNAs in our polyA RNA‐seq data are less than those in rRNA‐depleted RNAs (Figs 3E and 1I, compare 14 to 26%). One possible explanation is that the stabilized 5′ RNA degradation intermediates cannot be detected in the polyA RNA‐seq, but can be identified in rRNA‐depleted nuclear RNA‐seq. Consistent with rRNA‐depleted nuclear RNA‐seq data, relatively more lncRNAs (50%) were accumulated than mRNAs (14%) in hMTR4 knockdown cells (Fig 3E–H).

Figure 3. Identification of nuclear accumulated mRNAs and lncRNAs in hMTR4 knockdown cells.

• A
  A diagram of the RNA‐seq experimental approach.
• B
  Western blotting to examine hMTR4 knockdown efficiency and purities of nuclear fractions. UAP56 and tubulin were used as nuclear and cytoplasmic markers, respectively. N, nucleus; C, cytoplasm.
• C
  The distribution of reads derived from RNA‐seq libraries and mapped to the indicated RNA classes. Each category represents RNAs unique to that category and nonoverlapping with previous categories. Calculated mean of mapped reads in triplicate data, Fisher's exact test, lncRNAs, adjusted P = 0, odds ratio = 1.15; repetitive elements, adjusted P = 0, odds ratio = 1.66; and others, adjusted P = 0, odds ratio = 1.42.
• D
  The Venn diagrams depict the overlap of accumulated RNAs (> 1.5‐fold) detected by nuclear total RNA‐seq with those detected by nuclear polyA RNA‐seq in hMTR4 knockdown cells. Fisher's exact test P < 0.0001, odds ratio = 2.90.
• E
  The Venn diagrams depict the accumulated intron‐containing mRNAs in hMTR4 knockdown cells. The numbers in Venn diagrams show exon or intron reads that are elevated more than 1.5‐fold in hMTR4 knockdown cells. The percentages of intron‐containing mRNAs accumulated more than 1.5‐fold in hMTR4 knockdown according to exon or intron changes are shown.
• F–H
  Same as (E), except that instead of accumulated intron‐containing mRNAs, intronless mRNAs, intron‐containing lncRNAs, and intronless lncRNAs are shown, respectively. Note that relatively more lncRNAs are accumulated than mRNAs in hMTR4 knockdown cells (Fisher's exact test P < 0.0001, odds ratio = 3.46 according to exon change).
• I
  Box and whisker plots showing hMTR4 sensitivity (log2 (hMTR4 siRNA knockdown RPM/control RPM)) of pre‐mRNAs, spliced mRNAs, and intronless mRNAs. Note that this analysis was done to the all detected genes in RNA‐seq. Plot‐whisker: min to max. The difference between spliced mRNA and intronless mRNA is statistically significant (Wilcoxon test. ***P < 0.001).
• J
  Same as (I), except that pre‐lncRNAs, spliced lncRNAs and intronless lncRNAs are shown. The difference between spliced and intronless lncRNAs is statistically insignificant (Wilcoxon test. P = 0.38).

Source data are available online for this figure.

To estimate the changes occurring to precursors and mature forms of mRNAs and lncRNAs, we selectively computed reads mapping to exon–intron/intron–exon and exon–exon junctions, respectively. Compared to spliced mRNAs, accumulation was more apparent with intronless mRNAs upon hMTR4 knockdown (Fig 3I), suggesting that intronless mRNAs are less stable in the nucleus compared to spliced mRNAs. In contrast, no significant difference in RNA regulation between spliced and intronless lncRNAs was detected (Fig 3J), suggesting that splicing might not significantly enhance nuclear lncRNA stability. Considering that lncRNAs tend to have fewer introns than mRNAs and the functional coupling between splicing and export (Luo & Reed, 1999; Masuda et al, 2005; Valencia et al, 2008), they possibly have reduced export efficiencies compared to mRNAs and therefore greater sensitivity to nuclear exosome.

Nuclear exosome degrades cDNA transcripts that are export‐defective

Accumulating studies reported that extensive nuclear mRNA degradation occurs when mRNA export is blocked (Herold et al, 2003; Rougemaille et al, 2007; Saguez et al, 2008). This raises the possibility that the mRNA export pathway might compete with the mRNA degradation pathway. Our data showing that lncRNAs are more sensitive than mRNAs, and intronless mRNAs are more sensitive than spliced mRNAs, to exosome knockdown are consistent with this possibility. Many transcripts derived from cDNA plasmids that do not contain introns, that is, β‐globin and Smad cDNA transcripts, cannot efficiently recruit the mRNA export factors and are retained in the nucleus (Luo & Reed, 1999; Masuda et al, 2005; Valencia et al, 2008). A previous study reported that the level of the β‐globin cDNA transcript was remarkably elevated upon exosome knockdown (Bresson & Conrad, 2013). This study further supports the possible competition between exosome‐mediated mRNA degradation and mRNA export. To further investigate this possibility, we compared the effect of exosome knockdown on the levels of β‐globin and Smad cDNA transcripts with those of the corresponding spliced mRNAs that could be efficiently exported to the cytoplasm. We found that the levels of both β‐globin and Smad cDNA transcripts, but not the corresponding spliced mRNAs, significantly increased in hRRP40‐ or hMTR4‐depleted cells (Fig 4A–D). Although the possibility that cDNA transcripts are degraded for other reasons remains, these data, together with our RNA‐seq data, suggest that the exosome, together with hMTR4, functions in the degradation of inefficiently exported mRNAs.

Figure 4. hMTR4 functions in recruiting the exosome to its targets.

1. Export‐defective β‐globin cDNA transcript is a nuclear exosome target. RT–qPCRs to examine the level of β‐globin cDNA transcript in control, hRRP40, and hMTR4 knockdown cells. The relative level of β‐globin cDNA transcript to the transfection control HSPA1A, which is not an exosome target, is quantified and indicated in the graph.
2. Export‐proficient β‐globin spliced mRNA is not a nuclear exosome target. Same as (A), except that β‐globin pre‐mRNA reporter construct was used.
3. Export‐defective Smad cDNA transcript is a nuclear exosome target. Same as (A), except that the Smad cDNA reporter construct was used.
4. Export‐proficient Smad spliced mRNA is not a nuclear exosome target. Same as (A), except that Smad pre‐mRNA reporter construct was used.
5. hMTR4 preferentially associates with the β‐globin cDNA transcript. β‐globin spliced or cDNA reporter construct was transfected into Flag‐hMTR4 stable expression cells. 6 h post‐transfection, RNAs immunoprecipitated by ALYREF, Flag, and Myc antibodies were used for RT–qPCR analysis. The relative level of the β‐globin mRNA to GAPDH was quantified and indicated in the graph. Note that primers specifically amplifying the β‐globin spliced mRNA were used for PCRs.
6. hMTR3 preferentially associates with the β‐globin cDNA transcript. Same as (E), except that Flag‐hMTR3 stable expression cells were used and IPs were carried out using the Flag and the Myc antibodies.
7. hMTR4 is required for efficient association of the exosome with its targets. β‐globin cDNA reporter construct was co‐transfected with tRNA construct into control or hMTR4 siRNA‐treated Flag‐hMTR3 stable expression cells. The relative level of β‐globin cDNA transcript to tRNA present in the immunoprecipitate was quantified and indicated in the graph.
8. The exosome is not required for efficient association of hMTR4 with exosome targets. Same as (G), except that control or hRRP40 siRNA‐treated Flag‐hMTR4 stable expression cells were used.

Data information: (A–H) Error bars represent standard deviations from biological repeats (n = 3). Statistical analysis was performed using Student's t‐test. *P < 0.05, **P < 0.01. Source data are available online for this figure.

To examine whether hMTR4 functions in exosome‐mediated mRNA degradation in the context of TRAMP or NEXT complexes, we depleted TRAMP components, PAPD5 and ZCCHC7, as well as NEXT components, RBM7 and ZCCHC8, to < 10% at the protein level (Fig EV4A–C). As shown in Fig EV4D, the level of β‐globin cDNA transcript was apparently affected by none of these knockdowns, suggesting that these complexes might not be generally involved in exosome‐mediated mRNA degradation. In support of this possibility, knockdown of TRAMP or NEXT components did not lead to significant accumulation of five randomly picked exosome target mRNAs (Fig EV4E and F).

Figure EV4. TRAMP or NEXT might not be generally involved in the degradation of most mRNAs and lncRNAs.

1. Western blotting results show that TRAMP and NEXT components were efficiently knocked down.
2. Western blotting to estimate knockdown efficiencies of TRAMP components. HeLa cells expressing Flag‐ZCCHC7 were transfected with ZCCHC7, PAPD5, or control siRNA. 72 h post‐transfection, Western blotting was carried out to examine the exogenously expressed Flag‐ZCCHC7 or the endogenous PAPD5. Tubulin is used as a loading control. Different amounts of cell lysates of control knockdown cells were loaded to estimate the knockdown efficiencies.
3. Same as (B), except that Flag‐RBM7 expression plasmid was transfected to NEXT component knockdown cells.
4. β‐globin cDNA reporter together with the HSPA1A control plasmid was transfected into control‐, hMTR4‐, PAPD5‐, ZCCHC7‐, RBM7‐, or ZCCHC8 siRNA‐treated HeLa cells. 12 h post‐transfection, total RNAs were extracted followed by RT–qPCRs. The graph shows relative level of β‐globin mRNA to the HSPA1A mRNA. Error bars represent standard deviations from biological repeats (n = 3). Statistical analysis was performed using Student's t‐test. **P < 0.01.
5. RT–qPCRs to examine the levels of indicated mRNAs in HeLa cells treated with control, hMTR4, PAPD5, ZCCHC7, and PAPD5/ZCCHC7 siRNA. The relative levels of indicated RNAs to 18S rRNA are shown in the graph. Statistical analysis was performed using Student's t‐test. Error bars represent standard deviations from biological repeats (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001.
6. Same as (E), except that HeLa cells were treated with control, hMTR4, RBM7, ZCCHC8, or RBM7/ZCCHC8 siRNA.

Source data are available online for this figure.

hMTR4 is required for efficient association of the exosome with its targets

To investigate how the exosome specifically recognizes export‐defective RNAs, we examined the associations of the exosome, hMTR4 as well as an mRNA export adaptor, ALYREF, with β‐globin cDNA transcript and spliced mRNA. Flag‐hMTR4 stable expression cells, in which the level of Flag‐hMTR4 is near‐endogenous (Appendix Fig S4A), were transfected with β‐globin cDNA or spliced mRNA construct, followed by RNA immunoprecipitations (IPs) with the Flag and ALYREF antibodies. The Myc antibody was used as a negative control. Consistent with a previous study (Masuda et al, 2005), ALYREF efficiently associated with spliced mRNA, but not the cDNA transcript (Fig 4E). In contrast, the Flag antibody immunoprecipitated significantly more cDNA transcript than the spliced mRNA. These differential associations of ALYREF and hMTR4 with mRNAs were not due to altered protein IP efficiencies (Fig 4E). These results indicate that hMTR4 preferentially associates with the mRNAs that cannot efficiently recruit mRNA export factors. Similarly, using Flag‐hMTR3 stable expression cells, we found that Flag‐hMTR3 also preferentially associates with β‐globin cDNA transcript, relative to the corresponding spliced mRNA (Fig 4F and Appendix Fig S4B).

To investigate whether hMTR4 is required for the recruitment of the exosome to its targets, we compared the association of hMTR3 with β‐globin cDNA transcript in control‐ and hMTR4‐depleted cells. As shown in Fig 4G, this association was significantly reduced in hMTR4 knockdown cells. In contrast, hRRP40 depletion did not significantly affect the association of Flag‐hMTR4 with β‐globin cDNA transcript (Fig 4H). Together, these data indicate that hMTR4 plays a key role in exosome recruitment to its targets.

hMTR4 competes with ALYREF for interacting with ARS2

The role of hMTR4 in exosome recruitment raised the possibility that it might function in recognizing exosome targets. Based on the result that hMTR4 preferentially associates with RNAs that ALYREF does not bind, we speculate that hMTR4 and ALYREF might compete for associating with some proteins bound on RNAs. To test this possibility, we carried out IPs using the Flag or the ALYREF antibody from RNased nuclear extract prepared from Flag‐hMTR4 stable expression cells or regular HeLa cells, respectively. Mass spectrometry analysis identified ARS2, a noncanonical CBC component, as a common protein associating with hMTR4 and ALYREF (Fig 5A and B). Note that the hMTR4‐ARS2 association was previously identified using a similar approach and the association of ALYREF with CBP80, a canonical CBC component, was also reported (Cheng et al, 2006; Lubas et al, 2011; Andersen et al, 2013). To confirm the associations of hMTR4 and ALYREF with ARS2, we carried out IPs from HeLa nuclear extract using antibodies to hMTR4, ARS2, and ALYREF. Western analysis revealed that both hMTR4 and ALYREF were co‐immunoprecipitated with ARS2, and ARS2 was present in the immunoprecipitates of hMTR4 and ALYREF (Fig 5C). In contrast, ALYREF and hMTR4 did not co‐IP each other (Fig 5C). These data indicate that ARS2 interacts with both hMTR4 and ALYREF.

Figure 5. Both hMTR4 and ALYREF interact with ARS2 in vivo and in vitro .

1. The list of hMTR4‐associating proteins. IPs with the Flag or a control antibody were carried out from the RNased nuclear extract of the Flag‐hMTR4 stable expression cells. Proteins specifically identified in the Flag immunoprecipitate by mass spectrometry are listed. Unique, matched queries; total, matched peptides; MW, molecular weight. Note that ARS2 is detected in the Flag immunoprecipitate.
2. The list of ALYREF‐associating proteins. IPs with the ALYREF or a control antibody were carried out from RNased HeLa nuclear extract. Proteins specifically detected in the ALYREF immunoprecipitate by mass spectrometry are listed. Note that ARS2 was detected in the ALYREF immunoprecipitate.
3. (Left) ARS2 interacts with hMTR4 in vivo. IPs with the ARS2 or the hMTR4 antibody, or IgG were carried out from the RNase‐treated HeLa nuclear extract followed by Western blotting. 3% of the inputs were loaded. (Middle) ARS2 and ALYREF interact in vivo. IPs with the ARS2, ALYREF, and Myc antibody were carried out from the RNase‐treated HeLa nuclear extract followed by Western blotting. 10% of the inputs were loaded. (Right) hMTR4 and ALYREF have no interaction in vivo. IPs with the ALYREF, hMTR4, and IgG antibody were carried out from the RNase‐treated HeLa nuclear extract followed by Western blotting. 3% of the inputs were loaded.
4. ALYREF directly interacts with ARS2 in vitro. Purified GST and GST‐ARS2 proteins were used for pulling down purified MBP‐ALYREF or MBP in the presence of RNase A. Proteins pulled down were separated by SDS–PAGE followed by Coomassie staining. 37.5% of the inputs were loaded.
5. hMTR4 interacts with ARS2 in vitro directly. GST‐ARS2 and the negative control GST‐eIF4A3 were used for pull‐down of purified MBP‐hMTR4 or MBP in the presence of RNase A. Proteins pulled down were separated by SDS–PAGE, followed by Coomassie staining and Western blotting. 37.5% of the inputs were loaded.

Source data are available online for this figure.

To test whether these interactions are direct, we carried out pull‐downs using GST‐ARS2 purified from sf9 insect cells as well as MBP‐ALYREF and MBP‐hMTR4 purified from E. coli. GST, GST‐eIF4A3, and MBP were used as negative controls. Coomassie‐stained gels showed that MBP‐ALYREF was pulled down by GST‐ARS2, but not GST itself (Fig 5D). Similarly, Western blotting result shows that MBP‐hMTR4 was pulled down by GST‐ARS2, but not GST‐eIF4A3 (Fig 5E). These data indicate that the interactions of hMTR4 and ALYREF with ARS2 are direct.

We next asked whether hMTR4 competes with ALYREF for associating with CBC. For this, we overexpressed ALYREF, or the control DDX3, and carried out IPs using antibodies to ARS2 and hMTR4. As expected, in control cells, ARS2 and hMTR4 reciprocally co‐immunoprecipitated each other (Fig 6A). In contrast, in ALYREF‐overexpressing cells, only background level of hMTR4 was detected in the ARS2 immunoprecipitate (Figs 6A and EV5A). Likewise, significantly less ARS2 was co‐immunoprecipitated with hMTR4 in these cells. When an antibody against the canonical CBC component, CBP80, was used for IPs, similar results were obtained (Fig 6B). These data indicate that hMTR4 competes with ALYREF for associating with CBC in vivo. To further validate this competition, we next used purified proteins for in vitro pull‐downs. As shown in Fig 6C, the presence of MBP‐ALYREF, but not MBP, blocked the interaction between GST‐ARS2 with MBP‐hMTR4. Together, we conclude that hMTR4 directly competes with ALYREF for interacting with ARS2.

Figure 6. hMTR4 competes with ALYREF for associating with ARS2 and RNAs.

• A
  hMTR4 competes with ALYREF for associating with ARS2 in vivo. HeLa cells overexpressing Flag‐DDX3 (control) or Flag‐ALYREF were used for IPs in the presence of RNase A with the ARS2, hMTR4, or Myc antibody, followed by Western blotting. 3% of the inputs were loaded.
• B
  Same as (A), except that instead of the ARS2 antibody, the CBP80 antibody was used.
• C
  hMTR4 competes with ALYREF for interacting with ARS2 in vitro. Pull‐down of MBP‐hMTR4 with GST‐ARS2 was carried out in the presence of MBP or MBP‐ALYREF under RNase‐treated condition.
• D, E
  RNA‐IP analysis showed that the association of hMTR4 with RNAs was weakened by ALYREF overexpression. IPs were carried out with the hMTR4 antibody from Flag‐DDX3 (control) or Flag‐ALYREF‐overexpressing HeLa cells. The immunoprecipitates were used for RT–qPCRs to detect immunoprecipitated RNAs (D) and Western analysis to detect immunoprecipitated hMTR4 (E). Statistical analysis was performed using Student's t‐test. Error bars, standard deviations (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001.
• F
  Exosome target RNAs were accumulated in ALYREF‐overexpressing cells. Flag‐DDX3 or Flag‐ALYREF expression plasmid was transfected into HeLa cells followed by RT–qPCRs to detect levels of indicated RNAs. The relative levels of indicated RNAs to 18S rRNA were quantified and indicated in the graph. Statistical analysis was performed using Student's t‐test. Error bars, standard deviations (n = 3). *P < 0.05.
• G, H
  Genome‐wide effect of ALYREF overexpression on the association of hMTR4 with RNAs. IPs with the hMTR4 antibody were carried out from control or Flag‐ALYREF overexpression HeLa cells. The immunoprecipitates were used for Western analysis (G) and deep sequencing (H). The experiment was biologically repeated for three times. (G) Western blotting with the hMTR4 antibody to detect the IP efficiencies. (H) (Left graph) According to the change in hMTR4 association upon ALYREF overexpression, the RNAs are grouped into four classes. The percentages of RNAs falling in each class are shown. (Right graph) RNAs in the < 0.5 class were further grouped into three classes according to their level change detected by nuclear total RNA‐seq shown in Fig 1.
• I
  Screenshots of two replicates of hMTR4 RIP‐seq signals of DNAJC30 and H‐AS1 are shown.
• J
  Western blotting to examine the purities of nuclear fractions prepared from control and ALYREF overexpression cells. UAP56 and tubulin serve as the nuclear and cytoplasmic markers, respectively. N, nucleus; C, cytoplasm.
• K
  Genome‐wide effect of ALYREF overexpression on nuclear RNA levels. According to nuclear rRNA‐depleted RNA‐seq data shown in Fig 1, the mRNAs and lncRNAs that are accumulated more than 1.5‐fold in the nuclei of ALYREF overexpression cells were grouped into three classes.
• L
  Screenshots of RNA‐seq signals of PPP1R15A and HSP90AA1 mRNAs are shown as examples as exosome targets that are stabilized in the nucleus upon ALYREF overexpression. Replicates of nuclear RNA‐seq signals in control‐ and ALYREF‐overexpressing cells are shown above the gene profile, and those in control, hRRP40, and hMTR4 knockdown cells are shown in below.

Source data are available online for this figure.

Figure EV5. The competition between nuclear degradation and mRNA export.

• A
  Western analysis to examine the overexpression of Flag‐ALYREF.
• B
  According to the change in nuclear levels upon ALYREF overexpression, the mRNAs and lncRNAs are grouped into three classes. The percentages of RNAs falling in each class in each replicate are shown.
• C
  The overlap of mRNAs and lncRNAs that are accumulated in the nucleus more than 1.5‐fold upon ALYREF overexpression in two biological replicates.
• D
  Western analysis to examine the overexpression of Flag‐hMTR4.
• E
  Western blotting to examine UAP56 and hRRP40 knockdown efficiencies. Tubulin serves as the loading control.
• F, G
  Blocking mRNA export enhances hMTR4 recruitment to exosome target RNAs. Cntl/hRRP40 or UAP56/URH49/hRRP40 siRNA‐treated HeLa cells were used for IPs using the hMTR4 antibody. The immunoprecipitates were subjected to Western analysis to detect hMTR4 (F) and RT–qPCRs to detect immunoprecipitated RNAs (G). Error bars, standard deviations (n = 3). Statistical analysis was performed using Student's t‐test. *P < 0.05, **P < 0.01.

Source data are available online for this figure.

The competition between hMTR4 and ALYREF maintains balanced nuclear RNA pools for degradation and export

We reasoned that if the competition between hMTR4 with ALYREF is important for the specific exosome recruitment, ALYREF overexpression should reduce the association of hMTR4 with exosome targets. Indeed, in ALYREF‐overexpressing cells, significantly less exosome target RNAs were enriched with hMTR4, accompanied with apparently elevated levels of these RNAs in the cells (Figs 6D–F). To examine the generality of these observations, we sequenced hMTR4‐associating RNAs (three biological replicates) as well as nuclear rRNA‐depleted RNAs (two biological replicates) in control and ALYREF overexpression cells (Fig 6G–L, Datasets EV10 and EV11). These data revealed that upon ALYREF overexpression, the hMTR4 associations with 18% of detected mRNAs and lncRNAs were reduced to < 0.5‐fold (Fig 6H). Importantly, among these RNAs, around 45% were upregulated more than 1.5‐fold, and around 25% were upregulated 1‐ to 1.5‐fold, in hRRP40 or hMTR4 knockdown cells (Fig 6H). Two replicates of the exemplified RNAs, DNAJC30 and H‐AS1, are shown as examples (Fig 6I). For nuclear rRNA‐depleted RNA‐seq, in each biological replicate, more than 14% of mRNAs and lncRNAs were accumulated more than 1.5‐fold (Fig EV5B). Among 764 RNAs that showed significantly increased levels (> 1.5‐fold) in both replicates, ~52% were accumulated more than 1.5‐fold, and ~27% were accumulated 1‐ to 1.5‐fold upon exosome or hMTR4 KD (Figs 6K and EV5C). PPP1R15A and HSP90AA1 mRNAs are shown as examples for exosome targets that are stabilized upon ALYREF overexpression (Fig 6L). These data indicate that at the genome‐wide level, upon ALYREF overexpression, the reduced hMTR4 association and nuclear accumulation mainly occur to exosome targets. Together, these data support the view that the competition between ALYREF and hMTR4 is important for exosome recruitment. To further investigate this competition, we examined how the bindings of ALYREF with exosome targets are affected upon hMTR4 overexpression. Significantly, when hMTR4 was overexpressed, we reproducibly observed reduced binding of ALYREF with exosome target RNAs (Figs 7A and B, and EV5D).

Figure 7. hMTR4 functions in controlling balanced nuclear RNA pools for degradation and export.

• A, B
  RNA‐IP analysis shows the association of ALYREF with RNAs is weakened upon hMTR4 overexpression. HEK293 cells were infected with lentivirus expressing Flag‐DDX3 (control) or Flag‐hMTR4. IPs were carried out using the ALYREF antibody. The immunoprecipitates were used for RT–qPCRs to detect immunoprecipitated RNAs (A) and Western analysis to detect immunoprecipitated ALYREF (B). Statistical analysis was performed using Student's t‐test. Error bars, standard deviations (n = 3). *P < 0.05, ***P < 0.001.
• C
  Western blotting to examine ALYREF/THOC2 knockdown efficiency. Tubulin serves as the loading control.
• D, E
  Co‐knockdown of ALYREF and hTHO enhances hMTR4 recruitment to exosome target RNAs. Control or ALYREF/THOC2 siRNA‐treated HeLa cells were used for IPs with the hMTR4 antibody. The immunoprecipitates were subjected to RT–qPCRs to detect immunoprecipitated RNAs (D) and Western analysis to detect hMTR4 (E). Statistical analysis was performed using Student's t‐test. Error bars, standard deviations (n = 3). *P < 0.05, **P < 0.01.
• F
  Western blotting to examine the purity of the cytoplasmic fractions. CBP80 and tubulin were used as nuclear and cytoplasmic markers, respectively. C, cytoplasm; W, whole cell.
• G
  RNAs from cytoplasmic fractions prepared in (F) were used for RT–qPCRs to examine the cytoplasmic levels of indicated exosome target RNAs. The relative levels of indicated RNAs to 18S rRNA are quantified and indicated in the graph. Statistical analysis was performed using Student's t‐test. Error bars, standard deviations (n = 3). *P < 0.05, **P < 0.01.
• H
  Targeting ALYREF to RNAs prevents exosome activity and/or recruitment. (Top) Illustration of the β‐globin reporter construct that contains MS2‐binding site. (Bottom) RT–qPCRs to examine the levels of β‐globin cDNA‐6MS2 co‐transfected with Flag‐MS2‐MBP or Flag‐MS2‐ALYREF expression constructs in normal HeLa cells (left graph) or in control and hMTR4 knockdown cells (right graph). Error bars, standard deviations (n = 3). Statistical analysis was performed using Student's t‐test. **P < 0.01.
• I
  Model for the role of hMTR4 in exosome recruitment and in maintaining balance nuclear RNA pools for degradation and export. (Left) In normal cells, hMTR4 competes with ALYREF for associating with the CBC complex bound on RNAs and specifically recruits the exosome to RNAs failing to packed in export‐competent RNPs (see the details in the Discussion section). (Middle) In cells with mRNA export defect, hMTR4 gains access to more RNAs that are subject to degradation in the nucleus. (Right) In hMTR4 downregulated cells, nuclear RNA degradation is blocked and ALYREF, properly together with other mRNA export factors, associates with more RNAs that are subsequently exported to the cytoplasm.

Source data are available online for this figure.

Previous studies showed that blocking mRNA export led to extensive mRNA degradation (Herold et al, 2003; Rougemaille et al, 2007; Saguez et al, 2008). This might possibly be due to enhanced hMTR4/exosome association. It is known that ALYREF and THO, core components of TREX, play redundant roles in the association with CBC and the mRNA export receptor NXF1 (Viphakone et al, 2012; Chi et al, 2013). Taking into account such redundancy, we co‐knocked down ALYREF and THOC2 (Fig 7C). Note that knockdown of THOC2 results in the downregulation of the whole THO complex (Chi et al, 2013). As shown in Fig 7D, the binding of hMTR4 with exosome targets was significantly elevated in these co‐knockdown cells. This enhanced binding was also observed with UAP56 knockdown that substantially blocked mRNA export and resulted in inefficient ALYREF recruitment (Chi et al, 2013; Fig EV5E–G). These data suggest that the competition between hMTR4 and ALYREF controls nuclear RNA pools for degradation and export.

To further investigate this, we next examined the cytoplasmic levels of exosome target RNAs in hMTR4 knockdown cells. Purities of the cytoplasmic fractions were confirmed by Western blotting (Fig 7F). Importantly, cytoplasmic accumulation of exosome target RNAs was detected in hMTR4 knockdown cells (Fig 7G). Finally, to test the competition between ALYREF and hMTR4 in a more direct way, we targeted ALYREF to β‐globin cDNA transcript that is otherwise preferentially bound with and degraded by hMTR4/exosome (Fig 4E). Such targeting stabilized β‐globin cDNA transcript and knockdown of hMTR4 did not have further effect (Fig 7H). These data thus directly demonstrate the competition between ALYREF and hMTR4 for target RNAs. Taken all the data together, we conclude that the competition between hMTR4 and ALYREF functions in ensuring balanced nuclear RNA pools for export and degradation.

Discussion

In this study, we show that the human nuclear exosome functions in degrading a subset of mRNAs and lncRNAs. We investigated how the exosome is recruited to these target RNAs. Our data indicate that hMTR4 competes with ALYREF for interacting with CBC/RNAs and selectively recruits the exosome to RNAs that cannot efficiently form export‐competent RNPs. In cells, the RNA fate can be switched between degradation and export depending on the competition between hMTR4 and ALYREF.

How the exosome specifically recognizes its target RNAs remains unclear. In yeast, two main models have been proposed. One model is that exosome cofactors function in recognizing specific structure/sequences in exosome substrates (Vasiljeva & Buratowski, 2006). However, most exosome substrates are unlikely to share common structural or sequence features. In an alternative model, the exosome binds nascent transcripts “by default” and kinetic competition between maturation and degradation processes determines whether an RNA molecule is an exosome target (Gudipati et al, 2012). However, indiscriminative binding of the exosome with all RNA molecules does not seem economic, as in most cases the binding would be nonfunctional.

Our study here leads to a model for target recognition by the human nuclear exosome (Fig 7I). In this model, hMTR4 competes with ALYREF for interacting with CBC bound on RNAs. Upon maturation, ALYREF is recruited to the mature mRNAs via the direct interaction with ARS2 and interactions of ALYREF with other mRNP factors stabilize its binding on mRNAs. If an mRNA could efficiently recruit both ALYREF and these ALYREF‐facilitating factors, it is then efficiently exported to the cytoplasm. If, in any case, ALYREF cannot be recruited in time, or its binding cannot be stabilized by other factors, hMTR4 then gains access to ARS2 bound on an RNA and consequently recruits the exosome. Thus, our data are more consistent with the yeast “kinetic competition” model with an important distinction. In our model, the human exosome is specifically recruited to its targets and this specificity is obtained through the competition between hMTR4 and ALYREF. Considering that hMTR4 competes with ALYREF for binding with CBC/RNAs, how do cells ensure specific recruitment of hMTR4 to “export‐defective” RNAs, but not export‐sufficient RNAs? As previously reported, we found that the protein abundance of ALYREF is ~5 times more than that of hMTR4 in HeLa cells (Kocher et al, 2014; Appendix Fig S5). As most of hMTR4 is concentrated in the nucleolus, the nucleoplasmic concentration of hMTR4 must be even less than ALYREF. We speculate that the higher abundance of ALYREF might be important to ensure that hMTR4 can only bind mRNAs failing to efficiently recruit ALYREF, but not to normal RNAs because of limitation of ALYREF. Given the local high abundance of ALYREF, how does hMTR4 effectively compete with it? It is known that efficient ALYREF recruitment not only relies on the 5′ cap/CBC, but also depends on other TREX components as well as proper splicing and polyadenylation (Luo et al, 2001; Cheng et al, 2006; Johnson et al, 2011; Chi et al, 2013). Thus, for mRNAs that were not efficiently and/or properly processed (i.e., the β‐globin cDNA transcript), or in cells that other TREX components are downregulated (i.e., UAP56 knockdown cells), despite its high local concentration, ALYREF still cannot be efficiently recruited, and hMTR4 subsequently gets chance to bind and recruits the exosome for degradation.

All mRNAs presumably recruit CBC in their early biogenesis. However, as mentioned above, the cap/CBC itself is not sufficient to recruit ALYREF. On mRNAs derived from intron‐containing genes, splicing is required for efficient ALYREF recruitment (Masuda et al, 2005 and Fig 4). Distinct pre‐mRNA might have different splicing efficiencies that might lead to different efficiencies in recruiting mRNA export factors. For intronless mRNAs, specific export‐promoting cis‐elements are required that might have varied abilities in recruiting ALYREF (Lei et al, 2011, 2013; Chi et al, 2014). In addition, element‐mediated recruitment might not be as efficient as splicing‐dependent ALYREF recruitment. In support of these possibilities, we observed greater accumulation of intronless mRNAs compared to spliced mRNAs when the exosome function is inhibited. Thus, the difference in splicing kinetics or the strength of cis‐acting elements might cause ALYREF recruitment to different degrees on different transcripts.

Pre‐mRNAs with splicing and polyadenylation defects are also degraded by the nuclear exosome (Bousquet‐Antonelli et al, 2000; Milligan et al, 2005). Thus, it is possible that pre‐mRNA‐binding proteins and 3′ end processing factors also compete with hMTR4 for associating with CBC. hnRNPC was reported to interact with CBC on pre‐mRNAs. It is possible that hMTR4 and hnRNPC compete for binding with CBC on pre‐mRNAs and determines whether a pre‐mRNA is processed or degraded. A recent study reported that CLP1, a 3′ processing factor, also associates with ARS2 (Hallais et al, 2013). It would be interesting to investigate whether hnRNPC, CLP1, or other factors are involved in competition with hMTR4. The U snRNA export adaptor, PHAX, also associates with the CBC (Ohno et al, 2000; Andersen et al, 2013; Hallais et al, 2013). It is possible that PHAX competes with hMTR4 and determines whether an snRNA is exported or degraded (Muller‐McNicoll & Neugebauer, 2014). Indeed, during the revision of this work, the competition between PHAX and hMTR4, in the context of NEXT, was reported (Giacometti et al, 2017).

Our data indicate that both the qualities and the quantities of nuclear RNAs are controlled by the exosome. Up to date, the roles of the nuclear exosome in controlling levels of normal mature RNAs, especially mRNAs, have not been reported. We found that hMTR4 plays a key role in regulating nuclear RNA pools for degradation and export. This regulation might provide the cells an additional layer of regulation on gene expression and might be especially critical in certain tissue or developmental stages when the RNA pools require to be precisely controlled. Except for exosomal‐dependent functions, hMTR4 might have non‐exosome‐dependent functions, as a significant population of genes was upregulated in the hMTR4 knockdown, but not in hRRP40 knockdown. In mice, SKIV2L2 (the mouse MTR4 homologue) is highly expressed in round spermatids that need to undergo massive mRNA elimination to ensure cytoplasmic exclusion (Osman et al, 2011). It would be interesting to study the function of SKIV2L2 in controlling nuclear RNA pools in spermatids.

Materials and Methods

Cell culture, transfections and RNAi

HeLa and HEK293 cells were cultured in DMEM supplemented with 10% FBS (Biochrom). Lipofectamine 2000 (Invitrogen) was used for DNA transfection. For establishment of Flag‐hMTR4 and Flag‐hMTR3 stable expression cell lines, HAGE‐Flag‐hMTR4 or HAGE‐Flag‐hMTR3 plasmid together with psPAX2 and pMD2.G plasmids were co‐transfected into HEK293 cells. The media containing viruses were harvested after 48 h and added to HeLa cells, and then single green fluorescence cell was sorted by Flow Cytometer (AriaII, BD). For RNAi, siRNA transfection was carried out with Lipofectamine 2000 (Invitrogen) or Lipofectamine RNAiMax (Invitrogen). The siRNA targeting sequences are shown in Appendix Table S1. It is of note that both UAP56 and its homologue URH49 must be knocked down to observe a robust export block (Kapadia et al, 2006).

RNA isolation, reverse transcription, and PCR analysis

Total RNAs were extracted using TRIzol (Invitrogen). RNAs were treated with the RNase‐free RQ DNase I (Promega) for 2 h at 37°C, and cDNAs were synthesized from 1 μg of RNAs with random primer using M‐MLV reverse transcriptase (Promega). For semiquantitative PCRs, ³²P‐dATP was added in PCRs where indicated. For quantitative PCRs, cDNAs were amplified using GoTaq qPCR Master Mix (Promega) according to the manufacturer's protocol. Primer sequences are listed in Appendix Table S2.

Nuclear and cytoplasmic RNA preparation

1 × 10⁷ HeLa cells were washed with ice‐cold PBS and suspended in hypotonic buffer (10 mM HEPES, pH 7.9/1.5 mM MgCl₂/10 mM KCl/0.2 mM PMSF/0.5 mM DTT) and incubated for 10 min on ice. The swollen cells were dounced followed by centrifugation. The supernatant was collected as the cytoplasmic extract and the packed nuclear volume (PNV) was estimated. The nuclei were re‐suspended slowly in 1/2 PNV of low salt buffer (20 mM HEPES, pH 7.9/1.5 mM MgCl₂/20 mM KCl/0.2 mM EDTA/25% glycerol/0.2 mM PMSF/0.5 mM DTT) followed by adding 1/2 PNV of high salt buffer (20 mM HEPES, pH 7.9/1.5 mM MgCl₂/1.4 M KCl/0.2 mM EDTA/25% glycerol/0.2 mM PMSF/0.5 mM DTT) and mixing quickly. The mixture was rotated for 30 min at 4°C followed by centrifugation and the supernatant is the nuclear extract. Total RNAs of nuclear or cytoplasmic extract were extracted using TRIzol (Invitrogen).

Protein immunoprecipitations

For IPs from nuclear extract followed by mass spectrometry, 300 μl of HeLa nuclear extract was incubated under splicing condition in the presence of RNase A (50 ng/μl) and then incubated with 1 mL of IP buffer and 40 μl of antibody‐crosslinked beads overnight at 4°C. For IPs followed by Western blotting, 75 μl of HeLa nuclear extract was incubated under splicing condition in the presence of RNase A (50 ng/μl) and then incubated with 420 μl of IP buffer and 20 μl of antibody‐crosslinked beads for overnight at 4°C. For IPs from whole‐cell lysate, 1 × 10⁷ of HeLa cells transfected with indicated plasmids were harvested and re‐suspended with 1 ml of 1× TBS (Tris‐buffered saline)/0.5% Triton X‐100. After sonication and centrifugation, lysates were treated with 50 ng/μl of RNase A for 20 min at 37°C and incubated with the appropriate antibody for 2 h followed by rotation with nProteinA Sepharose (GE) at 4°C for an additional 2 h. The samples were washed three times with the same buffer and eluted with SDS loading buffer for Western blot analysis.

RNA immunoprecipitations

For RNA IPs from Flag‐hMTR4 stable expression cells, nuclear extracts were prepared from cells transfected with plasmids. 75 μl of HeLa nuclear extract was incubated under splicing condition and then incubated with 420 μl of IP buffer and 20 μl of antibody‐crosslinked beads for 2 h at 4°C. Half of the immunoprecipitates were analyzed by Western blotting. The other half was treated with proteinase K and RNAs were recovered by phenol/chloroform extraction and ethanol precipitation. For RNA IPs from Flag‐hMTR3 stable expression cells, 12 h after DNA transfection, cells were re‐suspended in 1 ml of NET‐2 buffer (50 mM Tris–HCl (pH 7.4), 150 mM NaCl, 0.1% NP‐40, 0.2 mM PMSF), followed by sonication and centrifugation. The lysates were incubated with the indicated antibodies for 2 h, followed by rotation with nProtein A Sepharose (GE) for another 2 h at 4°C. The immunoprecipitates were washed three times with the NET‐2 buffer.

For IPs from control‐, DDX3‐, and ALYREF‐overexpressing HeLa cells, cells grown on 10‐cm dishes were transfected with Flag‐DDX3 and Flag‐ALYREF plasmids (14 μg/dish), respectively. For IPs from hMTR4‐overexpressing HEK293 cells, cells grown on 10‐cm dishes were infected with lentivirus expressing Flag‐DDX3 (control) or Flag‐hMTR4; 24 h later, cells were harvested. For IPs from control/hRRP40 and UAP56/hRRP40 knockdown, cells were transfected with siRNAs for 48 h. For IPs from control and ALYREF/THOC2 knockdown, cells were transfected with siRNAs for 72 h. The rest of the experiment was the same as that for IPs from Flag‐hMTR3 stable overexpression cells.

GST pull‐downs

For each pull‐down reaction, 8 μg of GST or GST‐ARS2 and 8 μg of MBP‐tagged proteins were added to 20 μl of Glutathione Sepharose 4 Fast Flow beads in pull‐down buffer (1× PBS/0.1% Triton/0.2 mM PMSF/protease inhibitor). The mixtures were rotated for 5 h at 4°C and beads were washed with 1× PBS/0.1% Triton for three times and eluted with SDS loading buffer. Samples were separated by SDS–PAGE for Coomassie blue stain or Western blot analysis. For GST‐ARS2 pull‐down in the presence of excess MBP or MBP‐ALYREF, 8 or 24 μg of MBP‐ALYREF or MBP protein was added.

Sequencing

For rRNA‐depleted RNA sequencing, RNA was isolated from nuclear fraction with TRIzol (Invitrogen); 5 μg of total RNA was depleted with rRNA and stranded cDNA libraries were generated with TruSeq Stranded Total RNA Sample Prep Kit (Illumina) according to the manufacturer's instructions. For nuclear polyA RNA sequencing, 5 μg of total nuclear RNA was used for polyA RNA selection and stranded cDNA libraries were generated with TruSeq Stranded Total RNA Sample Prep Kit (Illumina) according to the manufacturer's instructions. For RNA associated with hMTR4 sequencing, stranded cDNA libraries were generated with TruSeq Stranded mRNA Sample Prep Kit (Illumina) according to the manufacturer's instructions. For details, the rRNA‐depleted total RNA samples were fragmented using divalent cations at elevated temperature and reverse transcribed with random hexamers to obtain double‐stranded cDNA fragments; instead of dTTP, dUTP was used in second‐strand cDNA synthesis. cDNA fragments were end‐repaired and 5′ end‐phosphorylated. After adding “A” bases to the 3′ ends, Illumina adaptor oligonucleotides were ligated to the cDNA fragments, followed by PCR amplification and bead purification. The libraries were ~270 bp. The barcode cDNA libraries were then individually loaded onto flow cells for cluster generation (version 3) after quantification with Qubit and Agilent Bioanalyzer 2100, and sequenced on an Illumina HiSeq 2000 using a single‐read protocol of 100 cycles with v3 chemistry at CAS‐MPG Partner Institute for Computational Biology Omics Core, Shanghai, China, or sequenced on an Illumina HiSeq X Ten using a double‐read protocol of 100 cycles at CLOUD HEALTH, Shanghai, China.

Accession numbers

The data associated with the manuscript are available under accession numbers GSE77641.

Author contributions

JF, BK, and HC conceived the study. JF, BK, and HC designed experiments. JF, BK, KW, GFW, BKC, LTW, XYC, SYW, and HZ performed experiments. XDW, ZSH, and GHL analyzed data. ZBS and ZCZ provided recombinant hMTR4. SC performed MS analysis. JF, BK, and HC wrote the paper. HC and GHL supervised the project.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Dataset EV1

Dataset EV2

Dataset EV3

Dataset EV4

Dataset EV5

Dataset EV6

Dataset EV7

Dataset EV8

Dataset EV9

Dataset EV10

Dataset EV11

Source Data for Expanded View and Appendix

Review Process File

Source Data for Figure 1

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 7

The EMBO Journal (2017) 36: 2870–2886