NMD3 regulates both mRNA and rRNA nuclear export in African trypanosomes via an XPOI-linked pathway

Trypanosomes mostly regulate gene expression through post-transcriptional mechanisms, particularly mRNA stability. However, much mRNA degradation is cytoplasmic such that mRNA nuclear export must represent an important level of regulation. Ribosomal RNAs must also be exported from the nucleus and the trypanosome orthologue of NMD3 has been confirmed to be involved in rRNA processing and export, matching its function in other organisms. Surprisingly, we found that TbNMD3 depletion also generates mRNA accumulation of procyclin-associated genes (PAGs), these being co-transcribed by RNA polymerase I with the procyclin surface antigen genes expressed on trypanosome insect forms. By whole transcriptome RNA-seq analysis of TbNMD3-depleted cells we confirm the regulation of the PAG transcripts by TbNMD3 and using reporter constructs reveal that PAG1 regulation is mediated by its 5′UTR. Dissection of the mechanism of regulation demonstrates that it is not dependent upon translational inhibition mediated by TbNMD3 depletion nor enhanced transcription. However, depletion of the nuclear export factors XPO1 or MEX67 recapitulates the effects of TbNMD3 depletion on PAG mRNAs and mRNAs accumulated in the nucleus of TbNMD3-depleted cells. These results invoke a novel RNA regulatory mechanism involving the NMD3-dependent nuclear export of mRNA cargos, suggesting a shared platform for mRNA and rRNA export.


INTRODUCTION
The compartmentalization of nuclear DNA away from the translational apparatus in the cytoplasm of eukaryotic cells requires elaborate mechanisms to achieve the regulated export of RNA to enable gene expression (1,2). The multiple stages of nuclear export provide several potential regulatory steps to ensure that RNAs of different function can be held inactive until fully matured, localized or targeted for cytoplasmic degradation (3,4). This prevents incorrectly or incompletely processed RNAs from perturbing normal cellular functions by interfering with important cellular processes, or generating products that would be deleterious to the cell (5,6). Such RNAs can include nascent ribosomal RNA subunits, tRNAs, incorrectly processed mRNAs, nuclear RNAi targets that require trafficking to the cytoplasm for degradation, as well as bulk mRNAs. Hence, in addition to transcription, mRNA stability and translation, the export pathway for a variety of cellular RNAs provides considerable scope for the cell to control the expression and function of transcripts (4). Such regulatory pathways have been extensively studied in both yeast and mammalian systems and include complex rRNA maturation and export steps (7) and the nonsense-mediated decay (NMD) of incorrectly processed mRNAs (8).
African trypanosomes provide an interesting opportunity to dissect the pathways of gene regulation in evolutionarily divergent organisms (9,10). These organisms, important pathogens of humans and livestock in sub-Saharan Africa, diverged very early in the eukaryotic lineage and exhibit several novelties of gene expression that have attracted considerable attention. This includes their co-transcription of functionally unrelated genes as part of polycistronic transcription units (11) and their extreme reliance on posttranscriptional mechanisms of gene expression control necessitated by this genome organization (12). Particularly, mRNAs are generated from polycistronic transcription units by the coupled processes of 5 trans-splicing of a 39nt capped leader sequence and 3 polyadenylation (13,14). These RNA processing steps allow trypanosomes to exploit the transcription of protein coding genes by RNA polymerase I in order to generate the high levels of surface protein transcripts necessary for their survival or fitness in the mammalian bloodstream or tsetse fly midgut (15,16). In the tsetse midgut, the surface coat is composed of different isoforms of procyclin. These isoforms (EP procyclin, GPEET procyclin) show differential expression during the parasite's maturation in the tsetse fly (17) and are transcribed from several transcription units on different chromosomes in the parasite genome (18). These are transcribed by RNA polymerase 1 (pol I) and contain unrelated genes of unknown function, namely the procyclin-associated genes (PAGs) (19,20). Procyclin/PAG genes are believed to be transcribed in the nucleolus since procyclic form parasites do not possess a detectable expression site body, an extranucleolar site of pol I transcription for the surface antigen genes of the bloodstream form (21). Although PAG transcripts are generated from the same transcription unit as the procyclin genes, their overall mRNA abundance is much lower (19,(22)(23)(24).
The understanding of regulated gene expression in trypanosomes through the action of protein factors remains very incomplete. Nonetheless, the trypanosome genome encodes an abundance of RNA binding proteins, many of which are unique to these evolutionarily divergent organisms (12,25,26). The characterization of these kinetoplastidspecific factors has revealed a common role in mRNA stability and translation, often involved in regulating the development through the many stages that accompany progression of the parasites through their complex life cycle (27). In addition to these specific factors, trypanosomatids also encode a core set of conserved protein factors with predicted functions in gene expression. These include many well characterized components of the translational machinery, as well as proteins associated with RNA processing, degradation and nuclear export (28,29). In particular, bioinformatics analyses of the genomes of kinetoplastid parasites and other eukaryotic groups including a broad group of opisthokonta have revealed a general conservation of many components of the nuclear export machinery, which are responsible for the trafficking of cellular RNAs (rRNAs, mR-NAs, tRNAs, miRNAs, snRNAs) to the cytoplasm (28). The Ran-GTP-dependent RNA export machinery is well conserved, with Ran orthologues in Trypanosoma brucei being over 70% identical to yeast and mammals and a divergent MEX67 orthologue with bulk mRNA export function has been recently characterized (29,30). Moreover, there exists particular conservation of the rRNA export machinery, with XPO1 and TbNMD3 demonstrating over 50% similarity to their human counterparts. Other, more evolutionarily restricted components of this machinery, such as ArX1, Alb1p and PHAX, are not evident in the T. brucei genome but there are additional kinetoplastid-specific components that have been identified, most notably the developmentally regulated proteins P34/P37(31) and NOPP44/46 (32). Hence, XPO1, TbNMD3 and P34/P37 and NOPP44/46 cooperate to process and export rRNA, demonstrating the role of both highly conserved, and parasite-specific, components in this fundamentally important export pathway (33)(34)(35)(36)(37).
In this study, we have made the surprising discovery that post-transcriptional gene silencing by RNAi of trypanosome NMD3 results in a striking increase in PAG mR-NAs in procyclic form trypanosomes, which is not mediated by the effects of NMD3 depletion on transcription, or translation. Rather, our results invoke a novel regulatory step dependent upon nuclear export of the mRNA, PAG transcripts representing an mRNA cargo with particular cytoplasmic instability whose trafficking is dependent on a nuclear export pathway conventionally involved in the maturation and export of large subunit ribosomal RNAs (38).
RNAi lines of TbNMD3 and XPO1 were created using the stem loop vector pALC14 (41). Inserts were amplified by polymerase chain reaction (PCR) using primers detailed in Supplementary Table S1. For transfection pALC14 vectors were linearized with NotI prior to transfection.

Report constructs
The constitutive CAT reporter construct was based on the expression vector pHD449 (42), as described in (43). The plasmid has a truncated aldolase 3 untranslated region (UTR) providing constitutive expression in T. brucei.

Chromatin immunoprecipitation
The chromatin immunoprecipitation (ChIP) protocol was adapted from (44). Depletion of TbNMD3 was induced 3 days prior to the chromatin IP. RNAi cells were centrifuged at 2000 rpm for 10 min in a clinical centrifuge and washed once in phosphate buffered saline (PBS). The cells were resuspended in 900 l PBS containing 1% paraformaldehyde and incubated at room temperature (RT) for 8 min. To stop the reaction 100 l of 1.25 M glycine was added and cells were washed in 10 ml PBS after 5 min. The cells were lysed by resuspension in 0.25 ml of lysis buffer (50 mM Tris-HCl pH 8, 10 mM ethylenediaminetetraacetic acid (EDTA), 1% sodium dodecyl sulphate (SDS), Protease inhibitor cocktail (Sigma)) and sonicated 3 × 15 s on ice with 30 s recovery intervals on ice. The sample was transferred to a low binding eppendorf tube and rocked for 30 min at 4 • C. After centrifugation at maximum speed for 10 min, the supernatant containing chromatin-DNA complexes was saved in PCR tubes.

ChIP DNA elution
For cross-link reversal, TE buffer was removed and the beads were resuspended in elution buffer (20 mM Tris-HCl pH 7.5, 5 mM EDTA, 50 mM NaCl, 0.1% SDS, 50 g/ml proteinase K). The input chromatin-DNA was diluted 1:100 in elution buffer. These were left for 2 h at 68 • C with rocking and vortexing at various intervals. Beads were captured and the supernatant was saved in a new tube. The DNA was cleaned up using a quick PCR clean up kit (Qiagen) and eluted with 50 l dH 2 O. DNA concentration was determined using NanoDrop.

Real time PCR
To analyse PAG locus transcription it was necessary to use primers for qRT PCR able to discriminate different regions of the EP and GPEET genomic loci. Although sequence similarity in the procyclin promoter region of the EP and GPEET gene loci prevented these regions being precisely differentiated by quantitative real time (qRT)-PCR assay, the limited up-regulation of each gene upon TbNMD3 depletion meant that a specific up-regulation of PAG gene transcription would be discriminated, although the precise allele from which particular isoform transcripts were derived could not be accurately determined. Three dilutions (0.1, 0.01, 0.001 g/reaction) of input DNA samples were used for standards. qPCR was performed with 0.001 g DNA/reaction using MESA GREEN qPCR MasterMix Plus for SYBR R Assay (Eurogentec) in the ABI Prism 7000 Sequence Detection System (Applied Biosystems). The data were analysed using 7000 System SDS software v1.2 (Applied Biosystems). The obtained values were corrected with unspecific binding to rabbit serum and data were expressed as enrichment of DNA associated with immunoprecipitated H3 relative to a 1:100 dilution of input chromatin.

RNA preparation and RNAseq analysis
RNA extraction and northern blot analysis were performed as previously described (43). RNASeq data (generated at BGI, Hong Kong) were analysed as follows. Firstly, raw fastq data were checked for quality using FastQC software (http://www.bioinformatics.babraham.ac. uk/projects/fastqc/). Next, the latest version of the T. brucei brucei 927 genome and annotation was obtained from: ftp://ftp.sanger.ac.uk/pub/project/pathogens/ gff3/CURRENT/Tbruceibrucei927.gff3.gz on 19 February 2015. The gff format file was parsed to extract all gene transcripts coordinates; 8966 possessed 5 UTR regions (total: 2 032 326 bases, average 227). Readsets were then grouped by sample type: 8 419 043 read pairs were obtained for PLUS and 19 340 407 read pairs for MINUS. Each sample readset (comprising several paired end sequence sets) was aligned independently to the T. brucei brucei 927 genome sequence using bowtie2 (v2.2.4) with parameters -verysensitive -p 12 -trim5 13. The outputs were saved as sorted, indexed BAM files (using samtools v 0.2.0). Transcript coverage for each gene in each sample was then obtained using bedtools (v2.15.0) with the gene transcript coordinates extracted from the genome gff file. Counts were normalized to the aligned reads and re-expressed as log2 counts per transcript kilobase. Gene-level data were only retained if one or both of the sample tags exceeded the 10% quantile (45). Specific regions of interest (i.e. the PAG array) were selected and excised from the primary alignment files to calculate mapping statistics for that region only. RNAseq data sets have been deposited in the GEO database with accession number GSE67246.

Preparation of riboprobes and drug treatments
In each case, genes for riboprobe preparation were inserted into pGEM-T easy vector containing T7/T3 or SP6 promoter sites flanking M13 forward and reverse primer binding sites as described previously (46). For drug treatment, Procyclic TbNMD3 RNAi cells were induced with tetracycline for 3 days. A sufficient volume of 5 × 10 6 cells/ml was cultured, so that for each time point there were 5 × 10 7 cells available for RNA preparation and analysis. Two samples were prepared before any drug treatment, one of the induced and one of the uninduced cells. Induced and uninduced cells were treated the same way.
To inhibit transcription Actinomycin D (Sigma) dissolved in DMSO (dimethyl sulfoxide) was used at a final concentration of 5 g/ml. To inhibit translation Cycloheximide (CHX) dissolved in dH 2 O was used at a final concentration of 50 g/ml. To inhibit trans-splicing Sinefungin dissolved in dH 2 O was used at a final concentration 2 g/ml. Leptomycin B (LMB) was supplied pre-dissolved in anhydrous ethanol and used at a final concentration of 0.1 or 1 g/ml.

CAT assay
The CAT ELISA assay (Roche) was performed according to manufacturer's instruction and as detailed in (43). CAT concentrations were determined from the standard curve giving the highest R-square value and the sample values being in the range of the standard.

In situ hybridization
Induced and uninduced TbNMD3 RNAi cells were harvested 62 h post induction and in situ hybridized as described (47) with the following modifications: after fixation, cells were incubated in 25 mM NH 4 Cl solution for 10 min and washed 2× with PBS. PBTSG (PBS + 2% triton, 0.5% saponin, 10 mg/ml glycine) was added to cells to permeabilize the parasite membrane and quench autofluorescence before cells were spread onto slides to settle in a humidity chamber for 20 min (at RT). Slides were washed 3× in PBS prior to prehybridization in Hybe B with/without unlabelled oligo dT. Hybridization was overnight as described with Cy3-labelled oligo dT[2 ng/ml]. Slides were washed once in 4XSSC;15% formamide, twice in 4xSSC, twice in 2XSSC, incubated in DAPI and mounted as described (47).

Cell fractionation
Induced and uninduced TbNMD3 RNAi cells were harvested 60 h post induction when cell counts were identical; aliquots for 'Total' cell were taken and cells were fractionated as described (47) with the following modifications: the cleared cell lysate was placed over 1 ml of 2 M sucrose and centrifuged at 10 000g (15 min at 4 • C) to yield the S10 supernatant and P10 pellet (nuclei-enriched) fractions. This was repeated to clarify the fractions and buffer was added to equate the volume of the P10 fraction with that of the S10 fraction. Fractions were vortexed and 1/3 volumes were aliquoted for protein and 2/3 for RNA isolation. Subsequent cDNA production and qRT-PCR were as described (47). Western blots loaded 5 l 'Total' protein and 10 l 'nuclear-enriched' (P10) or 'cytosolic' (S10) fractions per lane. Anti-cPGK and anti-P0 were used at 1:200 and 1:5000 dilution, respectively, as described (48) and mouse anti-NUPI (a kind gift from Klaus Ersfeld; University of Bayreuth, Germany) was used at 1:1000 dilution on a 4-12% gradient gel (Invitrogen) with HRP-labelled anti-mouse at [1:25 000] (Sigma).

RESULTS
The T. brucei NMD3 homologue (Tb927.7.970) has been reported to be required for normal cell growth (34) and has been implicated in 60S rRNA processing and export (35), a role conserved in other eukaryotes. A role in conventional NMD is not expected for this molecule based on the precedent from other organisms (37). To investigate in more detail the biological role of NMD3 in T. brucei we initially analysed the expression profile of this gene in different life cycle stages. Specifically, mRNAs derived from two laboratory-adapted (monomorphic) slender cell populations (T. brucei s427, T. brucei EATRO 2340, these having lost the ability to generate stumpy transmission forms through serial passage) and from developmentally competent (pleomorphic) slender, intermediate and stumpy T. brucei AnTat1.1 cells were analysed by SOLEXA profiling (49), as was mRNA from s427 procyclic cells ( Figure 1A). Interestingly this revealed that although the TbNMD3 transcript abundance was similar in the monomorphic and pleomorphic slender cells, intermediate and stumpy forms demonstrated considerably enhanced levels (4.3-and 2.8-fold with respect to pleomorphic slender cells). Procyclic forms exhibited levels 1.6-fold those in pleomorphic slender forms (Figure 1A). These analyses were confirmed by northern blotting of total mRNA derived from pleomorphic slender and stumpy cells and from procyclic forms ( Figure 1B).
To investigate the consequences of TbNMD3 depletion for transcripts, RNAi lines were generated in both bloodstream and procyclic forms. In both cases the efficiency of knock down was incomplete ( Figure 1C). Quantitation of procyclic forms by RNA-Seq (see later) demonstrated induced cells had retained 44% of the TbNMD3 mRNA of uninduced cells. Nonetheless this was sufficient to generate the expected accumulation of a 5-kb large subunit rRNA precursor, this being readily detected in ethidium bromide stained total RNA from each life cycle stage (arrowed in Figure 1D). The accumulation of this rRNA processing intermediate upon TbNMD3 depletion has been previously characterized (34). This was accompanied by an arrest in procyclic form parasite cell growth after 24-48 h, matching an earlier report ((34); Supplementary Figure S1), although cells remained motile and morphologically intact for at least 72 h. Bloodstream forms exhibited only a modest decrease in growth after TbNMD3 RNAi was induced (Supplementary Figure S1), reflecting either less effective depletion in the derived cell lines or less dependence on TbNMD3 levels at this life cycle stage. In consequence, all further studies were carried out in procyclic forms.

Analysis of mRNAs regulated by TbNMD3 depletion
The enrichment of TbNMD3 mRNA levels in stumpy forms prompted us to examine, in TbNMD3-depleted procyclic forms, the mRNA levels for several genes regulated during the trypanosome developmental cycle or associated with differentiation events. Initially, we examined 14 RNA pol I and RNA pol II driven transcripts by northern blotting ( Figure 2A). These included the genes for EP and GPEET procyclin, encoding the major surface protein expressed on procyclic forms and the genes for PAG1, PAG2, PAG4 and PAG5, which are co-transcribed with the procyclin genes in the same RNA pol I driven transcription units in the trypanosome genome (24) (Supplementary Figure S2). Additionally we examined genes for mitochondrial function (PPR1; cytochrome oxidase subunit IX; Al- In each case riboprobes to the target gene were constructed and hybridized with total RNA from TbNMD3 RNAi cells, induced or not, with detection of the rRNA precursor on ethidium stained gels confirming functional TbNMD3 depletion in each case (data not shown). Figure 2A demonstrates that there was a small increase in several of the analysed transcripts (1.1-to 1.9-fold through quantitative analysis; Figure 2B). However, most striking was the elevation of PAG gene transcripts (PAG1, 2, 4, 5), these being elevated 5-15-fold upon TbNMD3 depletion (Figure 2A and B). The PAG transcripts are detected as a distribution of RNAs of various sizes whose characteristics and boundaries have been previously described in detail (22,24). In particular transcripts for PAG1 ␣ (1231 nt excluding the SL and polyA tail) , ␤ (1617 nt) and ␥ (2473 nt) are detected reflecting alternative splice sites in the PAG1 5 UTR with an additional 4.4-kb band of unknown origin also being seen, as previously reported (20) (Supplementary Figure S2). Interestingly, although there was an increase in PAG transcripts in the induced samples, the profile of the bands detected did not substantially change; rather all de-tected species of mRNA were elevated upon TbNMD3 depletion these increasing after 48 h (Supplementary Figure  S3). One further notable change in the mRNAs analysed was detected, namely the size of the TAO transcript, which was slightly smaller in the TbNMD3 RNAi induced sample (Figure 2A). Although we have not been able to establish the basis of this size change, it was very reproducible on distinct blots of individual RNA samples and, with the elevation of PAG transcripts and rRNA processing defect, was diagnostic of TbNMD3 depletion.
To generate a higher resolution and genome-wide analysis of the effects of TbNMD3 depletion in procyclic cells, polyA+ selected RNA from RNAi induced and uninduced samples was subject to RNA-Seq transcriptome sequencing. This confirmed the relative up-regulation of the PAG transcripts upon TbNMD3 RNAi (Supplementary data 1). The EP1-PAG region (from the beginning of EP1 to the end of PAG4; Figure 3) exhibited 2.45× more reads in the TbNMD3 depleted cells compared to uninduced cells taking into account the total number of reads. The abundance but not the distribution of reads was changed indicating that particular processing events were not grossly affected. A further set of genes were also upregulated or downregulated (Supplementary data 1) although the PAG family transcripts were the most obviously represented, exclud- ing VSG, ISG65, retroposon and expression site associated gene families. The quantitative correspondence with the northern blot analyses was good ( Figure 2B), revealing the strong regulation of PAG transcripts, and more limited regulation of GPEET (1.9-fold elevated) and EP (1.3-fold elevated). Genes down-regulated comprised mainly hypothetical proteins and hypothetical conserved proteins (Supplementary data 1). From the small number of genes upregulated 2-fold or more, there was a bias toward shorter transcripts perhaps related to similar observations with the depletion of TbPuf2 and some other mRNA regulators (51).

TbNMD3 depletion does not specifically enhance PAG gene transcription
The elevation of PAG transcripts upon TbNMD3 depletion raised the possibility that TbNMD3 might specifically repress the transcription of the PAG genes. In order to in-vestigate this, a ChIP approach was pursued using an antibody against histone H3 to detect co-precipitated DNA as an indicator of transcriptional activity for specific gene loci (52). ChIP analysis of the EP1/EP2 locus showed that there was reduced associated histone H3 upon TbNMD3 depletion across the PAG locus but the same result was observed for the procyclin promoter region and also for a panel of other RNA pol II transcribed genes whose mRNAs were not elevated by TbNMD3 depletion (Supplementary Figure S4A). Since histone H3 levels were unchanged in the respective cell lines (Supplementary Figure S4B) this indicated that TbNMD3 depletion could generate a slightly decreased histone H3 association at multiple loci, but that this was not specific for the PAG transcripts and so could not account for their relative elevation.

PAG regulation is mediated through the conserved 5 untranslated region
The observation that PAG transcripts were upregulated in response to TbNMD3 RNAi prompted us to investigate whether PAG regulatory signals could be responsible. Although in trypanosomatids 3 UTR sequences frequently regulate gene expression, polyadenylation occurs immediately after the stop codon of the PAG genes, with regulation dominated by 5 UTR sequences (24). Hence, a reporter construct ( Figure 4A) was generated in which the chloramphenicol acetyl transferase (CAT) gene was linked to the PAG1 5 UTR and integrated into the RNA pol IItranscribed tubulin gene locus of the TbNMD3 RNAi cell line. A construct with an aldolase 5 UTR was also generated as a control. The level of CAT expression was then analysed when TbNMD3 was depleted or not. Figure 4B demonstrates that, as with the endogenous PAG1 gene, several transcript isoforms of the CAT-PAG1 5 UTR reporter line were increased upon induction of TbNMD3 RNAi. This effect matched the response of the endogenous PAG transcripts, whereas the control reporter line with the aldolase 5 UTR showed a relatively small increase in CAT-aldolase mRNA. At the protein level, CAT-ELISA assays demonstrated that the levels of CAT protein remained unchanged in the CAT-PAG1 reporter line ( Figure 4C). Therefore, the PAG 1 5 UTR is sufficient to confer TbNMD3 responsiveness to an unrelated reporter gene in a transcription unit driven by RNA pol II rather than RNA pol I. However, this elevated transcript abundance does not generate an equivalent CAT protein increase.

PAG transcript regulation in response to inhibitors of distinct steps in gene expression
We next investigated the basis of PAG transcript regulation following TbNMD3 depletion by analysing different control steps in the gene expression pathway. Specifically we sought to block mRNA translation, transcription and RNA processing using Cycloheximide (CHX), Actinomycin D (ActD) and Sinefungin (SinF), respectively, in order to determine if any of these inhibitors would recapitulate or reverse the effect of TbNMD3 depletion on PAG transcript levels ( Figure 5 and Supplementary Figures S5 and S6). Given the effect of TbNMD3 depletion on 60S rRNA processing and hence, potentially, translation we initially anal-  Figure S5). Using Actinomycin D ( Figure 5, 'Act D'), in contrast, resulted in a loss of longer PAG transcripts and accumulation of a smaller major transcript at 1.8 kb (arrowed on Figure 5, 'ActD'), which then remained stable in both TbNMD3 RNAi induced and uninduced samples. This suggested that inhibiting transcription caused longer PAG transcripts to be processed to a shorter stable form or that longer isoforms were preferentially degraded, but this occurred irrespective of TbNMD3 levels. Finally, we analysed the effect of inhibiting trans-splicing using Sinefungin ( Figure 5, 'SinF'). Whilst no change in the profile of PAG1 RNA was detected in the induced or uninduced samples, the elevated abundance of the PAG1 transcripts in the TbNMD3 depleted cells was quickly reversed; the PAG mRNA profiles of the uninduced and induced samples were equivalent within 1 h.
Hence, inhibiting trans-splicing counteracts the enhanced levels of PAG transcripts generated by TbNMD3 depletion.

Inhibiting XPO1 recapitulates the NMD3 RNAi phenotype
Since inhibiting various steps in the gene expression pathway had not recapitulated the NMD3 effect on the PAG transcripts, we also analysed the export of nuclear RNAs. In eukaryotes, NMD3 and the nuclear export factor XPO1 interact, with NMD3 providing a chaperone to move rRNA cargos to the cytoplasm. In trypanosomes, XPO1 action has been analysed by drug inhibition using LMB (53,54) and by RNAi, revealing a role in nuclear export of the 7SL RNA, a component of the signal recognition particle (55). Therefore, we initially analysed the effect of the XPO1 inhibitor LMB at either 1 g/ml or 0.1 g/ml, these concentrations following earlier publications using this drug (53,55). Remarkably when 1 g/ml of LMB was used (Supplementary Figure S7), but not 0.1-g/ml LMB (data not shown), a recapitulation of the TbNMD3 RNAi phenotype was observed in procyclic forms. Thus, EP and GPEET Figure 5. Effect of various inhibitors of different steps in the gene expression pathway on cells induced, or not, to deplete TbNMD3. Cells in which TbNMD3 was depleted, or not, were incubated in the presence of Cycloheximide, Actinomycin D or Sinefungin. RNA samples were isolated at various time points after drug addition and then northern blots assayed for PAG1 transcript levels. In each case, the equivalent ethidium bromide stained rRNA bands are shown to indicate relative loading and RNA integrity. mRNAs were slightly elevated, whereas PAG mRNAs were more strongly upregulated (Supplementary Figure S7A). Furthermore, analysis of the TAO transcript revealed that the subtle size change in that transcript was also observed with LMB treatment (Supplementary Figure S7B).
Given that LMB may lack specificity in T. brucei or inhibit multiple targets, we sought to validate the results by depleting XPO1 in procyclic form trypanosomes by RNAi. Therefore, an XPO1 gene fragment was inserted in to the pALC14 stem loop vector and the stable cell lines selected. Analysis of XPO1 using an antibody to this protein (kindly provided by Noreen Williams, Buffalo, NY, USA) confirmed its depletion upon the induction of RNAi (Figure 6A). Functional depletion was supported by the accumulation of the 5-kb precursor of the LSU rRNA in total RNA ( Figure 6B) and the observation that the induced cells showed a strong growth phenotype over 48 h ( Figure 6C).
Having validated the XPO1 RNAi line, the consequences for the regulation of transcripts perturbed by TbNMD3 depletion were assayed. Northern blots detecting PAG1 and PAG5 and TAO RNAs were analysed in parallel from the derived TbXPO1 and TbNMD3 RNAi lines ( Figure 6D), demonstrating that the depletion of either target generated equivalent responses for each test transcript. Thus PAG1 and PAG5 were both elevated in the XPO1 line, albeit to a lesser extent than in TbNMD3 depleted cells, and TAO demonstrated the characteristic size shift previously observed with TbNMD3 depletion. This suggested that the TbXPO1 and TbNMD3 RNAi both generate equivalent mRNA changes, implicating nuclear export as the important point of control.

mRNA transcripts are enriched in the nucleus upon NMD3 depletion
To investigate export of mRNAs from the nucleus we initially analysed the distribution of polyA+ RNAs in TbNMD3 depleted cells by in situ hybridization. We predicted that if TbNMD3 was involved in mRNA export from the nucleus we would observe a concentrated signal in the nucleus, although stable cytoplasmic mRNA would also be detected, depending on the kinetics of TbNMD3 depletion and cytoplasmic mRNA turnover. To assist visualization of any changed distribution, cells were induced for TbNMD3 depletion by RNAi over 62 h, once growth was reduced but before significant cell death was observed in the population (Supplementary Figure S1). Thereafter, the distribution of polyA+ RNA was compared in the induced and uninduced populations using Cy3-OligodT, with an unlabelled OligoT blocking step providing a control for specificity of the signal. Figure 7 shows that the induced population showed ∼20% cells with a strong nuclear concentration of Cy3 OligodT signal 62 h after induction, contrasting with less than 1% in uninduced cells, where the distribution of signal was non-nuclear. Cells pre-blocked with unlabelled OligodT showed little or no signal, confirming probe specificity (Supplementary Figure S8). Since this global polyA+ mRNA analysis suggested that all mRNAs might be accumulated in the nucleus, rather than PAG transcripts specifically, we analysed the distribution of several mRNAs after cell fractionation ( Figure 8A). Specifically, nuclear and cytoplasmic fractions of parasites where TbNMD3 was depleted or not were analysed for the relative distribution of PAG1, PAG5, KMP11 and actin, with normalization to the 140-nt tail of the spliced leader mRNA. This revealed relative enrichment of all mRNAs in the nuclear fraction relative to SL RNA after TbNMD3 depletion, with PAG1 and PAG5 particularly abundant, either through preferential retention or cytoplasmic instability. To provide further insight into this, we analysed the abundance of PAG mRNAs in a MEX67 RNAi line, and in a cell line where a dominant negative (DN) N-terminal truncation of MEX67 is overexpressed in an RNAi background, this providing a stronger depletion effect (these cell lines being kindly provided by Bernd Schimanski, Bern, Switzerland) (29). Under these conditions bulk mRNA export from the nucleus is blocked (29). Figure 8B demonstrated that upon MEX67 RNAi, there was a strong (up to 8-fold) increase in abundance of the PAG transcripts similar to that seen upon NMD3 depletion, in contrast to the abundance of the control transcripts TIM, RPL30 and RPL10. Moreover this effect was strongly enhanced by combined RNAi and the expression of the dominant negative MEX67 protein ( Figure 8B, righthand panel).
Combined these results indicate NMD3 depletion causes an accumulation of mRNA in the nucleus, a consequence being increased abundance of PAG transcripts that are normally quickly degraded in the cytoplasm.

DISCUSSION
In the experiments reported here we have demonstrated that depletion of the T. brucei orthologue of NMD3, functionally verified through rRNA processing defects, results in a significant up-regulation in PAG mRNAs (28)(29)(30)(31)(32). This regulation generated ∼5to 16-fold increases, whereas only a few other cellular mRNAs were strongly elevated or decreased under the same conditions. The PAG genes are located downstream of, and cotranscribed with, the EP and GPEET genes by RNA Pol I (19). Despite this close linkage, EP and GPEET transcripts were only modestly up-regulated upon NMD3 depletion, perhaps because these transcripts are already ∼100-fold elevated over PAG transcripts.
The regulation of PAG transcripts and the effect of PAG sequences on adjacent gene transcripts are complex. PAG genes lack any significant 3 UTR sequences with polyadenylation of PAG1 occurring within 10 bases of the stop codon. Thus, regulation is believed to operate through their extensive 5 UTR sequences and coding regions, which contribute to the low abundance of the PAG transcripts despite their transcription by RNA pol I (19,22,24). Accordingly, the CAT-PAG1 5 UTR reporter construct was upregulated upon TbNMD3 RNAi, analogous to the regulation of the endogenous PAG transcripts, whereas a control mRNA was not significantly up-regulated. Hence, the  observed regulation was specific to the PAG regulatory sequence but not dependent on the genomic location or the polymerase driving their transcription.
In order to investigate the basis of the upregulation of PAG transcripts in TbNMD3 depleted cells, we analysed several steps in the gene expression pathway. Since analysis of the PAG 5 UTR has demonstrated that it can act as a transcriptional silencer for nearby genes (22) we examined the interaction between TbNMD3 depletion and transcription, specifically the histone occupancy of the PAG genes and several additional genes. Although a slight general decrease in histone occupancy was observed, this was not specific for PAG genes and was observed across a wide range of genomic loci. Hence we do not consider that transcription is a significant contributor to the effects observed. Thereafter we examined the contribution of several post-transcriptional steps to the regulation of PAG transcripts in TbNMD3-depleted cells. We investigated whether the role of TbNMD3 in rRNA maturation and translation could have resulted in a feedback upregulation of PAG mRNA, perhaps as cells attempted to compensate for reduced TbNMD3 protein levels. However, CHX treatment did not generate an equivalent effect as TbNMD3 depletion in uninduced cells. Similarly, when transcription was inhibited with Actinomycin D, the different length forms characteristic of the PAG1 transcripts were reduced to a major short transcript; however this was unrelated to TbNMD3 depletion. Finally, to investigate the effect on mRNA processing via trans-splicing, we treated cells with Sinefungin. Interestingly, the up-regulation of PAG transcripts seen upon TbNMD3 RNAi was reversed within 1 h, suggesting that this drug can counteract the depletion of TbNMD3, destabilizing the elevated PAG transcripts or preventing their accumulation.
Having ruled out a direct role for transcription, RNA processing or translation as being responsible for PAG gene upregulation, we next turned our attention to nuclear export. Given that NMD3 cooperates with XPO1/CRM1 to export rRNA (35,56), we first investigated the effects of LMB on cells (an inhibitor of XPO1/CRM1 (57)) and then XPO1 RNAi. This revealed that the phenotypes resulting from TbNMD3 depletion were recapitulated, supporting functional nuclear export as the potential mechanism for PAG mRNA regulation. An involvement of XPO1/NMD3 in PAG regulation was a surprise since these molecules are known to cooperate in the export of rRNA but not mRNA. One explanation was that the PAG transcripts might be exported for degradation via a discrete pathway. The positioning of the PAG genes is often close to the boundary of divergent or convergent transcription units or where there is a switch from RNA pol I to pol II mediated transcription (22). Although such regions have the potential to generate transcriptional overlap and resultant dsRNA, such that they might be templates for cytoplasmic RNAi mediated through the action of TbAGO1 (58), we consider this unlikely since PAG levels are not elevated in TbAGO1 deficient cells (22).
In Figure 9 we present a model that incorporates our findings based on the pathway for PAG transcript maturation and degradation. In this model, nascent PAG transcripts would be rapidly processed to a range of mRNAs of differ- Figure 9. Model for the export and degradation pathway for PAG transcripts. The explanation of the model is provided in the manuscript text. ent lengths, which are then exported via an NMD3/XPO1dependent process to the cytoplasm, where they are rapidly degraded, generating their relatively low abundance. With NMD3, XPO1 or MEX67 depletion, the newly synthesized and processed PAG transcripts would accumulate in the nucleus. When transcription is inhibited with Actinomycin D the accumulated nuclear transcripts could undergo further cycles of slower processing or differential nuclear decay, generating a short transcript form stable in the nucleus, with longer forms no longer being generated de novo. In response to inhibition of splicing by Sinefungin, PAG transcripts could be degraded in the nucleus by the nuclear exosome (59). Indeed, blocking trans-splicing may specifically activate the degradation of incompletely processed transcripts in the nucleus, thus preventing the export of deleterious aberrant mRNAs and explaining the reversal of PAG mRNA accumulation in NMD3-depleted Sinefungintreated cells. This NMD3-dependent export pathway could preferentially target a subset of mRNAs, particularly the PAGs, explaining the apparent specificity of the accumulation of this transcript family. However our results favour a bulk export pathway, with PAG mRNAs preferentially accumulating due to their high turn over necessitated by their transcription as part of the pol I driven procyclin gene locus. This general mRNA export mechanism is supported by three lines of evidence. Firstly, TbNMD3 depletion results in cells with an accumulation of polyA+ mRNA in the nucleus as detected by in situ hybridization. Secondly, qRT-PCR of nuclear and cytoplasmic mRNAs demonstrates that several transcripts become enriched in the nuclear fraction when TbNMD3 is depleted by RNAi. Finally, depletion of the bulk mRNA export factor MEX67 recapitulates the increase of PAG transcripts observed with TbNMD3 depletion (54). Combined these observations support a role for TbNMD3 in the export of RNAs from the nucleus in addition to their role in 60S rRNA export. We do not yet know if this entails formation of a shared, possibly transient, complex involving TbNMD 3/XPO1 and TbMex67 or the passing of rRNA and mRNAs between distinct complexes which can interact or interfere functionally. Nonetheless, interaction between the pathway for rRNA and mRNA processing and transport matches a shared role of the recently identified nucleolar protein NRG1 in rRNA processing and GPEET expression control (60) and is consistent with LMB treatment of T. cruzi cells that generates nuclear mRNA accumulation, albeit with specificity for some mR-NAs (54).
Our experiments suggest that trypanosomes exhibit a shared platform for the export of rRNA and mRNAs, analogous to the capacity for the bulk mRNA export MEX67-Mtr2 complex to contribute to the export of ribosomal rRNA subunits (61). Indeed MEX67 can act as a suppressor of NMD3 mutants in yeast, underscoring the potential for functional overlap between rRNA and mRNA export in eukaryotic cells. This overlap may be particularly relevant for abundant RNAs where a quantitative export is necessary for their function or targeted elimination in the cytoplasm, or be necessary in trypanosomes where bioinformatics analysis suggests the presence of a reduced set of nuclear export proteins. Moreover, such nuclear export pathways could be a control step generating the differential abundance of cotranscribed transcripts. Whether this also has a regulatory function whereby the export of individual transcripts can be tuned depending on the physiological demands on the parasite, or developmental cues, represents an exciting area of future study.