ABSTRACT
Salivarian trypanosomes are extracellular parasites of mammals that are transmitted by tsetse flies. The procyclic acidic repetitive proteins (PARPs) are the major surface glycoproteins of the form of Trypanosoma brucei that replicates in the fly. The abundance of PARP mRNA and protein is very strongly regulated, mostly at the post-transcriptional level. The 3'-untranslated regions of two PARP genes are of similar lengths, but are dissimilar in sequence apart from a 16mer stem-loop that stimulates translation and a 26mer polypyrimidine tract. Addition of either of these PARP 3'-untranslated regions immediately downstream of a reporter gene resulted in developmental regulation mimicking that of PARP. We show that the PARP 3'-UTR reduces RNA stability and translation in bloodstream forms and that the 26mer polypyrimidine tract is necessary for both effects.
The salivarian trypanosomes are unicellular parasites that live in the extracellular fluids of mammals and are transmitted by tsetse flies. The bloodstream forms evade the immune response by means of a thick coat of variant surface glycoprotein (VSG) (1 ,2 ). When they enter the tsetse fly midgut, many morphological and biochemical changes ensue as they transform into procyclic forms. One change is the replacement of VSG by the procyclic acidic repetitive protein (PARP or procyclin) (3 ,4 ).
Trypanosome gene expression is unusual in that nearly all protein coding genes so far investigated are transcribed in a polycistronic fashion (5 ,6 ). The precursor transcripts are processed into monocistronic mRNAs by 3'-polyadenylation and by 5' addition of a capped 39 nt spliced leader. Polyadenylation is both spatially and temporally coupled to the trans splicing reaction. When trans splicing is inhibited, polyadenylation ceases (7 ). Usually several polyadenylation sites are used, scattered over a short region ~100 bases upstream of a trans splicing signal (8 -11 ). Such signals commonly include a polypyrimidine tract with one or more acceptor AG dinucleotides up to ~80 bases downstream (12 ,13 ).
There is no evidence for any regulation of RNA polymerase II-catalysed transcription in trypanosomes. Developmental regulation of mRNA levels is determined post-transcriptionally; the sequences responsible are usually located in the 3'-untranslated region (3'-UTR) of the transcripts concerned (14 -19 ). Transcription of the VSG and PARP genes is, in contrast, subject to some regulation (5 ,6 ,20 -22 ), but the RNA polymerase responsible is probably RNA polymerase I (23 -25 ). Expression of the PARP genes is regulated at several levels. Transcription is ~10-fold less active in bloodstream form trypanosomes than in procyclic forms (21 ,22 ,26 ), whereas PARP is undetectable in bloodstream forms. Indeed, expression of the antigenically invariant PARP at this stage would probably be lethal.
We have previously used a bicistronic vector bearing a CAT gene and hygromycin resistance (HYG) cassette to study developmental regulation mediated by 3'-UTRs placed downstream of the CAT gene. 3'-UTRs from genes expressed in bloodstream forms caused bloodstream form-specific CAT expression; 3'-UTRs from genes expressed in procyclic forms, such as PARP genes, caused procyclic form-specific CAT expression (17 ). Actin (ACT) mRNA levels are not developmentally regulated; CAT mRNA and CAT protein were correspondingly not regulated if the CAT gene bore an ACT 3'-UTR (17 ). Similar results have been reported for PARP and other 3'-UTRs placed downstream of CAT (18 ), luciferase (14 ) and phleomycin resistance genes (27 ,28 ).
Here we demonstrate that the sequence primarily responsible for PARP post-transcriptional regulation is a polypyrimidine tract that is conserved in all PARP gene 3'-UTRs. It acts by accelerating RNA turnover and inhibiting translation in bloodstream forms.
Bloodstream trypanosomes (MiTat 1.2) and procyclic trypanosomes (derived from bloodstream Antat 1.1) were cultured and transfected and CAT assays performed as described (17 ,29 ). Hygromycin-resistant bloodstream trypanosome populations were generated by transfecting 107 trypanosomes with 10 [mu]g plasmid DNA linearized within a tubulin targetting sequence and selected with 12.5 [mu]g/ml hygromycin and the resulting populations were analysed as soon as sufficient parasites were available (22 ). Alternatively, plasmids were linearized within an rRNA targeting sequence and cloned by limiting dilution. CAT assays were performed at least three times on each population. Results were reproducible when independent populations were generated with the same plasmid.
Plasmids for stable transfection were based on pHD 324 (17 ), a bicistronic construct containing CAT (with an ACT 3'-UTR) and hygromycin resistance (HYG) genes. Different 3'-UTRs or mutated forms of them were cloned downstream of CAT; pHD 390 (Fig. 1 ; 17 ) contains the PARP[alpha] 3'-UTR. The PARP[beta] 3'-UTR (Fig. 2 , pHD 440) was obtained by PCR with genomic DNA from procyclic AnTat 1 cells as template.
Total RNA was isolated using TRIzol reagent (Gibco Life Technologies Inc., Eggenstein). RNA from 3.3 * 106 (procyclic forms) or 1-5 * 107 (bloodstream form) cells was separated on formaldehyde gels and blotted onto Hybond N+ membrane (Amersham, Braunschweig). In a few experiments mRNA was detected using enhanced chemiluminescence (ECL, Amersham). The exposed autoradiograms (30 s-5 h) were scanned (Agfa ARCUS II) and quantitated using NIH Image 1.52. In the other studies, including all turnover experiments, 32P-labelled probes were used and bands quantitated using a phosphorimager (Molecular Dynamics). When compared, chemiluminescence and radioactive probing yielded quantitatively similar results.
For RNase protection analysis (30 ), a CAT cassette cloned into pBluscript II SK+ (Stratagene) was cleaved with NcoI and transcribed in the presence of [32P]UTP to yield a 290 base product, which contained 56 bases that were not present in the mature CAT mRNAs. After RNase digestion the 234 base protected product was detected on sequencing gels, with appropriate controls.
The relationship between RNA half-lives and abundance were calculated using the formula:
v = log 2(1/t½ + 1/tD)[RNA]
where t½ is the measured half-life in non-growing cells and tD is the mean division time of a growing population (8 h in bloodstream forms, 12 h in procyclics) (31 ). We assumed that the rate of degradation (v) of an mRNA equals its rate of synthesis.
Sequence alignments were done using align/lalign (32 ) and RNA folding assessed by MFOLD (33 ). Polyadenylation and splicing sites were determined by RT-PCR (9 ).
Trypanosoma brucei contains 8-12 PARP genes, arranged in direct tandem repeats containing two to three genes each (5 ). Figure 1 A shows a simplified map of a PARP B locus with the corresponding mRNAs immediately below (PARP A loci are similar). Each locus bears an upstream (a) and a downstream (b) gene. The genes differ somewhat in the coding region but diverge in the 3'-UTRs. The untranslated regions, including both the intergenic regions and the 3'-UTRs of the mature mRNAs, include several polypyrimidine tracts. Some of these direct trans splicing of mature PARP mRNAs or of unstable, non-coding RNA intermediates. Others can also be shown to direct trans splicing if they are placed 5' of a CAT open reading frame (9 ). Polypyrimidine tracts with known trans splicing activity are shown as filled boxes in Figure 1 . Known linked upstream polyadenylation sites are also indicated.
An alignment of all PARP 3'-UTR sequences (e.g. PARP B[alpha] and PARP B[beta], Fig. 2 ) reveals two conserved regions: a 16mer that is predicted to form a stem-loop and stimulates translation in procyclic forms (34 ) and a 26mer polypyrimidine tract (`26' in Figs 1 -4 ). Secondary structure predictions for all available PARP 3'-UTR sequences (not shown) revealed no common features apart from the 16mer stem-loop; the 26mer poly(U) were not predicted to be involved in base pairing.
To investigate the regulatory function of PARP 3'-UTRs, we used a vector designed to integrate into the tubulin locus (17 ; Fig. 1 B). Downstream of a fragment from the tubulin locus (to enable targeted integration) is an rRNA promoter, which is not developmentally regulated (22 ). Next is an ACT 5'-UTR and splicing signal, followed by a CAT gene. The 3'-UTR downstream of CAT can be exchanged using unique restriction sites; it is from the PARP B[alpha] locus in pHD 390 (Fig. 1 B). Beyond this 3'-UTR is a second 5' trans splicing signal and 5'-UTR from the ACT locus (17 ), then a hygromycin resistance (HYG) cassette. Unless otherwise noted, the plasmid was linearized, transfected into trypanosomes and hygromycin-resistant populations studied. Such populations are derived from at least 20 individual transformants containing one to three copies of the construct (17 ,22 ); their use obviates the need to perform independent copy number determinations for each plasmid tested. Transcription of CAT was probably mediated by RNA polymerase II reading through from upstream (17 ).
CAT mRNA with an ACT 3'-UTR (CAT-ACT 3', pHD 324) shows no developmental regulation (17 ). Results for the CAT-ACT 3' plasmid were therefore used as a control (100% value) throughout this study. Using the PARP[alpha] 3'-UTR, CAT mRNA and CAT protein levels were similar in procyclic forms to those seen with the ACT 3'-UTR; with the PARP[beta] 3'-UTR CAT activity in procyclic forms was 2-fold less (not shown). In bloodstream forms (Fig. 3 ) CAT activity was suppressed >100-fold and CAT mRNA levels ~10-fold by the PARP 3'-UTRs. The discrepancy between the levels of CAT-PARP 3' mRNA and the corresponding proteins indicates that the PARP 3'-UTRs not only reduce mRNA levels, but also adversely affect translation in bloodstream forms. Differentiation of bloodstream trypanosomes containing the CAT-PARP 3' transgene into procyclic forms resulted in >100-fold up-regulation of CAT activity (to about half the final, procyclic level) within 48 h (35 ).
Deletion mutagenesis was used to localize sequences in the PARP 3'-UTRs that reduce bloodstream-form expression. Progressive 5' deletions of the PARP[alpha] 3'-UTR up to position 139 (Fig. 3 A) had marginal effects on CAT activity, but somewhat elevated CAT mRNA levels. Deletion 1-158 was the first to include the entire 26mer; suddenly CAT activity rose >100-fold to 42% of the ACT 3'-UTR control. The level of CAT mRNA from this mutant was about half that seen from the control. Further deletion ([Delta]1-197) yielded an additional 1.5-2-fold increase in CAT activity. Results for the PARP[beta] 3'-UTR were quite similar (Fig. 3 B); 5' deletions that did not include the 26mer had little effect; as soon as the 26mer had gone ([Delta]1-159) CAT rose to 85%. Further deletions of the PARP[beta] 3'-UTR yielded only marginal increases in RNA or protein (Fig. 3 B).
To examine the behaviour of the 26mer in more detail, we first replaced the 26mer of the PARP[alpha] 3'-UTR with a BglII site. Expression of CAT in bloodstream trypanosomes with this plasmid, yielding CAT-PARP[Delta]26 3' RNA, was 34% of the ACT control ([Delta]26, Fig. 4 ). Returning the 26mer element to its original location (in the artificial BglII site) restored down-regulation if the 26mer was in the correct orientation ([Delta]26+26, Fig. 4 ). When the 26mer was placed immediately downstream of the CAT cassette, either upstream of PARP[Delta]26 3' ([Delta]26+p26, Fig. 4 ) or upstream of ACT 3', it did not suppress expression in bloodstream forms (ACT +p26, Fig. 4 ). When a larger segment of the PARP[alpha] 3'-UTR (117-203) was inserted between CAT and the ACT 3'-UTR, CAT activity was reduced 5-fold (ACT +p87, Fig. 4 ). Insertion of the inverted (antisense) 26mer had only minor effects ([Delta]26 +r26, Fig. 4 and data not shown).
In all constructs tested so far, polyadenylation of CAT mRNAs was directed by an ACT 5' splicing sequence downstream (Fig. 1 ). To restore the PARP[alpha] 3'-UTR to its `natural' context, we placed the complete PARP intergenic region (including the 3'-UTR) between the CAT and HYG cassettes. An equivalent construct with the [Delta]26 deletion was also tested. This change in the region downstream of the 3'-UTRs had no influence on the level of CAT expressed (not shown).
The conserved 16mer stem-loop has been shown to enhance translation in procyclic forms without influencing mRNA levels (34 ). Similar results were obtained for bloodstream forms: deletion of the 16mer from the [Delta]26 construct ([Delta]26/16) reduced CAT 3-fold, but the RNA level was unaffected (Fig. 4 ).
To see if the PARP[alpha] and PARP[beta] 26mers can act as trans splicing signals, we transferred appropriate segments of the 3'-UTRs (125-203 and 127-206 respectively, Fig. 2 ) to a 5'-UTR position, upstream of a CAT gene, and measured CAT activity after transient transfection. In the absence of a functional trans splicing signal, no CAT is produced (9 ,12 ). Both fragments were capable of directing CAT production in bloodstream and procyclic trypanosomes as efficiently as the ACT splice signal.
We next deleted from pHD 390 the segment from position 197, downstream of the 26mer in the PARP[alpha] 3'-UTR (Fig. 2 ) to the beginning of the HYG gene (Fig. 1 C, pHD 614). This forced use of the 26mer as a trans splicing signal for the HYG mRNA (Fig. 1 C). Drug-resistant bloodstream trypanosomes were easily obtained. The HYG mRNA was spliced at the AG at position 197. Polyadenylation was 74 or 57 bases upstream of the 26mer and yielded CAT RNA and CAT activities similar to those from the ACT 3'-UTR control (not shown). Thus the 26mer must be within the mature 3'-UTR to down-regulate expression in bloodstream forms.
We now mutated individual bases within the 26mer. The resulting CAT and CAT mRNA levels are shown in Figure 5 . In this series of experiments PARP[Delta]26 yielded CAT activity (black bars) and CAT mRNA (grey bars) of ~25% of the ACT 3' control. All mutations of the 26mer, even those with only three transitions, affected regulation. CAT activity increased to at least 8% and, with one exception, RNA levels were doubled, attaining values similar to those from the [Delta]26 construct (Fig. 5 ). Mutant I and III 3'-UTRs yielded similar CAT activities to, but more RNA than, the mutant VI 3'-UTR. Mutant V has the sequence of the polypyrimidine tract that lies downstream of the PARP[alpha] polyadenylation site (9 ; see Fig. 1 ). This sequence is an active splice signal and mediates accurate PARP[alpha] polyadenylation (9 ), but is not as effective as the 26mer in reducing CAT expression and mRNA. The most interesting result was for mutant VI. This 3'-UTR appeared to have specifically lost the ability to down-regulate translation; the RNA level was reproducibly the same as for the wild-type PARP[alpha] 3'-UTR, but CAT expression was at least 10-fold higher.
Post-transcriptional regulation often operates at the level of RNA stability. We therefore compared turnover of selected mRNAs in cloned transgenic bloodstream and procyclic trypanosomes (see legends to Figs 6 and 7 ). RNA synthesis was inhibited using chloroquine (36 ) or actinomycin D (18 ) and the amounts of RNA assessed by blot hybridization or RNase protection. In all plots the negative time points show cells that were not drug treated and the zero time point represents cells that received drug 0-5 min before centrifugation (total time in the presence of drug 10-15 min). In procyclic forms the half-lives of ACT and PARP mRNAs were 1.5-2.5 h (Fig. 6 A); results for CAT-ACT 3' and CAT-PARP 3' mRNAs were similar (not shown). RNAs with PARP 3'-UTRs were consistently slightly more stable than those with ACT 3'-UTRs, but the difference was not statistically significant. There was no difference between (-) and t = 0 samples.
We have shown that two PARP 3'-UTRs mediate >10-fold down- regulation of CAT mRNA levels and >100-fold down-regulation of CAT activity in bloodstream trypanosomes. The 10-fold post-transcriptional regulation of mRNA levels is consistent with the 100-fold PARP mRNA regulation observed in these trypanosomes (22 ), as PARP transcription is 5-10-fold less active in bloodstream forms than in procyclics (22 ). A conserved polypyrimidine tract (26mer) is important in this regulation. The function of the 26mer did not depend on its distance from the coding region or from the polyadenylation site, but it had to be present in the mature mRNA, as transcripts polyadenylated 5' of the 26mer were abundant and well-translated in bloodstream forms. Interactions with other sequences in the 3'-UTR were not essential, but in the PARP[alpha] 3'-UTR additional regulatory elements were present immediately downstream. The ability of the 26mer to suppress expression in bloodstream forms was dependent on context. It could not repress expression when placed at the 5'-ends of the ACT 3' or PARP[alpha][Delta]26 3'-UTRs. Perhaps, when the 26mer is moved from its normal location its function is overridden by other sequences that either exert a positive influence on the RNA level or form secondary structures that preclude 26mer function.
Although the 26mer is U-rich, it is poorly homologous to the UA-rich nonamer that is implicated in RNA decay in mammalian cells (39 ,40 ). Trypanosomes diverged so early in eukaryotic evolution that the vague resemblance between the two sequences is most likely fortuitous. Also, the first six bases (UAAUAU) of the 26mer were not required for repression ([Delta]1-139, Fig. 3 ). In contrast, all mutations of the two pyrimidine tracts within the 26mer affected activity. As few as three base changes in these portions were sufficient to increase CAT expression and CAT mRNA to levels approaching, or equivalent to, those seen when the entire 26mer was deleted (Fig. 5 ). Polypyrimidine tracts are common in trypanosome 3'-UTRs, whether or not the mRNAs are abundant in bloodstream forms, so recognition of this particular one by regulatory mechanisms must be very sequence specific.
Ehlers et al. (41 ) measured RNA turnover in trypanosomes by pulse-chase labelling with [3H]adenine. They found that total poly(A)+ RNA turned over faster in bloodstream forms than in procyclics. The estimated half-life of tubulin mRNA, which shows little developmental regulation, was 1.4 h in bloodstream forms and 7 h in procyclic forms (41 ). These values were probably artifactually long due to the large ATP pool (36 ,37 ). Using RNA synthesis inhibitors, we found that the half-life of ACT mRNA was 20 min in bloodstream forms, but 90 min in procyclics.
The levels of ACT mRNA, as a proportion of total mRNA, are approximately the same in bloodstream forms and procyclic forms (42 ). How can this be when the degradation rates differ nearly 5-fold? Firstly, we routinely obtain two to three times less RNA from bloodstream trypanosomes than from procyclic forms. Secondly, bloodstream forms grow at 37oC and procyclics at 27oC; perhaps everything: transcription, processing and degradation, is faster at the higher temperature.
Previous reports have indicated regulation of trypanosomatid RNAs through accelerated or delayed turnover. Experiments in which VSG 5'- and 3'-UTRs were linked to a CAT gene yielded results suggesting that these sequences are capable of reducing mRNA levels in procyclic forms (18 ,43 ), with some effects on RNA stability (18 ). Regulation of gene A2 mRNA in Leishmania donovani was also shown to be mediated by elements in the 3'-UTR that affected RNA stability (44 ), the data being consistent with biphasic mRNA decay kinetics.
Our results indicate that accelerated turnover contributes to the low abundance of PARP or CAT-PARP 3' mRNAs in bloodstream forms. After inhibition of mRNA synthesis, most CAT-PARP 3' mRNA was degraded within 10 min. The CAT-PARP[Delta]26 3' mRNA, in contrast, behaved like CAT-ACT 3' mRNA, indicating that the 26mer was required for rapid degradation. After 10-15 min the rates of degradation of all these RNAs were indistinguishable. The biphasic kinetics of PARP and CAT-PARP 3' mRNA degradation suggest that there are two RNA subpopulations. Perhaps most of the newly synthesized RNA is degraded in the nucleus, while the portion that escapes into the cytosol is more stable? So far our attempts to test this hypothesis have been frustrated by an inability to detect CAT-PARP 3' mRNAs after cell fractionation. In one experiment (Fig. 7 ) a possible degradation intermediate of CAT-PARP 3' mRNA was seen. We do not know the structure of this band. All mRNAs investigated showed a gradual decrease in size after inhibition of RNA synthesis. This would be consistent with shortening of the poly(A) tail prior to degradation, as is seen in other eukaryotes (45 ,46 ).
In mammalian cells and yeast binding of splicing factors to unprocessed RNAs can inhibit their export from the nucleus, leading to rapid degradation (47 ,48 ). It therefore seemed possible that the U-rich 26mer was affecting RNA abundance by this mechanism. When a segment including both the 26mer and the downstream sequence up to the next AG dinucleotide was placed upstream of an open reading frame (CAT or HYG), trans splicing activity was indeed detected. However, the trans splicing ability was not developmentally regulated and preliminary mutation results (49 ) have yielded no evidence so far for any correlation between trans splicing potentiality and the ability to reduce mRNA levels.
Bloodstream forms contain some mature PARP mRNA, but PARP is undetectable because the 26mer represses translation. Surveying the results from all mutants, an increase in CAT activity to a level significantly above background (2% and more) usually correlated with a rise in the amount of RNA to >20%; RNA levels of >35% yielded CAT activities >25%. One mutation of the 26mer abolished translational repression without affecting the level of mRNA (Fig. 5 ), suggesting involvement of separate factors (or separate specificities within one factor) that interact with the 26mer. To further investigate post-transcriptional regulation of PARP expression it will be necessary to identify and characterize these factors.
We are indebted to Elisabetta Ullu (Yale University, New Haven, CT) for suggesting the use of Sinefungin for RNA degradation studies and kindly donating some of her remaining supplies for this purpose (the drug is no longer commercially available). We thank Drs Timothy Nilson (Case-Western Reserve University, USA), Christian Tschudi (Yale University, New Haven, CT), Stephan Krieger, Martin Eilers Patrick Lorenz and Henriette Irmer (ZMBH) for discussions or helpful comments on the manuscript. Oligonucleotides were synthesized by the ZMBH oligonucleotide synthesis facility. This work was supported by the Deutsche Forschungsgemeinschaft.
*To whom correspondence should be addressed. Tel: +49 6221 546876; Fax: +49 6221 545894; Email: cclayton@sun0.urz.uni-heidelberg.de
REFERENCES