Flagellum ontogeny in trypanosomes studied via an inherited and regulated RNA interference system

Trypanosomes are motile protozoans responsible for a variety of diseases in a wide range of organisms. One representative of this group, Trypanosoma brucei causes African sleeping sickness and these parasites develop in the bloodstream of humans and other mammals as extracellular organisms. Trypanosomes diverged very early in eukaryotic evolution, and exhibit many unique features, both at the cellular (Clayton et al., 1995; Gull, 1999) and molecular level (Vanhamme and Pays, 1995; Lee and Van der Ploeg, 1997). In particular, expression of protein-coding genes is markedly different from other eukaryotes. Genes are transcribed in a polycistronic unit that is then processed by trans-splicing and poly-adenylation. The few promoters characterised so far all seem to use RNA polymerase I and show no significant similarities with other eukaryote promoters (Lee and Van der Ploeg, 1997). RNA polymerase II promoters have not been identified and intergenic sequences do not have promoter activity (Haüsler and Clayton, 1996). In addition, control of gene expression mostly takes place at the post-transcriptional level where the 3′ untranslated region (UTR) plays a critical role for RNA stability and translation.

Considerable progress in understanding parasite biology has been accomplished recently thanks to the development of transfection systems (Clayton, 1999). In T. brucei, DNA with linearised ends corresponding to trypanosome sequence is specifically inserted in the genome at this site by homologous recombination (ten Asbroek et al., 1990; Lee and Van der Ploeg, 1990; Eid and Sollner-Webb, 1991). Such homologous recombination facilitates double gene knock-out approaches in these diploid organisms, however, in many studies of phenotype and gene function, a flexible antisense strategy would be a useful alternative method. Unfortunately, expression of antisense RNA from transfected DNA does not have significant effect in trypanosomes (Tovar and Fairlamb, 1996; Clayton, 1999). Two noticeable exceptions have been published recently. First, expression of RNA antisense of the Leishmania A2 gene (a stage-regulated gene) from an episome resulted in a considerable decrease in the amount of both its mRNA and its protein, reducing the infectivity of the parasite (Zhang and Matlashewski, 1997). These authors suggest that the high number of repeated sequences within the A2 gene might explain their success. Second, our transformation of T. brucei with the aim of expressing antisense RNA of the PFRA gene yielded a single clonal mutant called snl-1 (Bastin et al., 1998, 1999a). This mutant grew normally but exhibited ablation of the PFRA mRNA and PFRA protein, leading to severe alteration of the paraflagellar rod (PFR), a unique lattice-like structure present inside the flagellum of trypanosomes (Bastin et al., 1996a; Maga and Lebowitz, 1999; Bastin and Gull, 1999), and cell paralysis (Bastin et al., 1998, 1999a). The PFRA gene (Schlaeppi et al., 1989) encodes one of the 2 major PFR proteins, the other one, PFRC, shares 60% identity with PFRA (Deflorin et al., 1994), but its expression is barely modified in the snl-1 mutant (Bastin et al., 1998).

Understanding the molecular mechanisms involved in the production of such phenotypes is essential since it may reveal important features about gene expression in Kinetoplastid parasites.

Our original experiments showed that the effectiveness of the molecular events in the snl-1 clone appeared related to the expression of antisense RNA from within the PFRA gene cluster. Expression from distant sites in other clones obtained in these experiments produced little or no effect (Bastin et al., 1998, 1999a). Recently, the discovery of a phenomenon called RNA interference (RNAi) has extended conventional views about antisense mechanisms (Fire et al., 1998). Microinjection of a mixture of in vitro synthesised sense and antisense RNA in Caenorhabditis elegans worms is much more efficient in blocking the expression of different genes than antisense (or sense) alone. This presence of double-stranded (ds) RNA leads to a specific, potent and rapid degradation of the corresponding mRNA (Montgomery et al., 1998). Mutants incapable of RNAi have recently been identified and indicate the existence of a complex pathway (Cogoni and Macino, 1999a,b; Tabara et al., 1999; Ketting et al., 1999; Grishok et al., 2000). RNAi initiated by dsRNA has been demonstrated in several organisms (reviewed by Fire, 1999) including trypanosomes (Ngô et al., 1998). Consideration of the genotype of the paralysed snl-1 mutant suggested that insertion of a strong promoter in the opposite orientation from normal polycistronic transcription of the PFRA genes, would provide conditions for in situ expression of sense and antisense RNA. This would permit the capability for local formation of dsRNA, leading to an RNAi effect on the transcripts from downstream genes in that PFRA cluster and those from the homologous cluster in this diploid organism. Here we test and verify this hypothesis by experiments which lead to the generation of a new cell line (snl-2). This expresses an RNA containing linked copies of sense and antisense PFRA, capable of forming a dsRNA, from a tetracycline-inducible promoter. Induction of expression of this PFRA dsRNA reproduced PFRA ablation, severe alteration of the PFR assembly and cell paralysis. We show this inducible RNAi system to be fast-acting, heritable and reversible. These properties in the snl-2 mutant have allowed us to reveal novel aspects of targeting, construction, turn-over and hierarchy mechanisms operating in forming and mature flagella.

Procyclic T. brucei (strain 427) were grown at 27°C in semi-defined medium containing 10% fetal calf serum. Plasmids pHD430 (Wirtz and Clayton, 1995), pHD451 (Biebinger et al., 1997) and pαT6451 (Bastin et al., 1998) have been described. The pαT6451 was linearised through the unique NsiI restriction site for insertion in the procyclin locus, or through NotI for insertion in the rDNA spacer locus or through BsmI for insertion within one of the PFRA genes. The 3′ UTR of the last PFRA gene (529bp, nt 4509 to 5039; Schlaeppi et al., 1989) was PCR amplified from genomic DNA using 5′-GGAAGCTT-GATTGTGTACTGTAATTG-3′ as forward primer (HindIII site underlined) and 5′-GGGGTACCAACACGGATATATACG-3′ (KpnI site underlined) as reverse primer, cloned in a pBluescript T vector and cut out with NotI and KpnI for insertion in the corresponding sites of pαT6451, generating the plasmid p3′PFRAαT6451. This plasmid was linearised through the unique BbrPI restriction site for insertion in the 3′UTR of the last PFRA gene of a cluster. The pααPFRA430 was assembled by ligating the nearly full-length copy of the PFRA gene (nt 78-1800) cut out from the pPFRATAGSK (Bastin et al., 1996b) with HindIII and BamHI in the plasmid pPFRATAG430 (Bastin et al., 1999b), an inducible trypanosome expression vector containing the Ty-1 epitope tagged copy of the full length PFRA gene, also digested with HindIII (nt 1 of the PFRA) and BamHI (nt 22 of the epitope tag sequence). Therefore, the pααPFRA430 plasmid contains the last 1722 bp from the PFRA gene in the antisense orientation, 12 bp from the epitope tag sequence in the antisense orientation, 12 bp from the epitope tag sequence in the sense orientation and the last 1722 bp of the PFRA gene in the sense orientation. This plasmid was linearised through the unique EcoRV restriction site for insertion in the rDNA spacer or through the unique SalI site for insertion in the actin locus. DNA was purified from cesium chloride gradients or Qiagen resin. Plasmids were linearised by single restriction digest and introduced in trypanosomes by electroporation (Beverley and Clayton, 1993). Resistant cells were selected as described (Bastin et al., 1999b). For inducible expression of PFRA dsRNA, the pααPFRA430 plasmid was integrated in the opposite strand of the rDNA spacer (known to be transcriptionally silent) in trypanosomes expressing the tet-repressor (Wirtz and Clayton, 1995; Bastin et al., 1999b). Transformants were subcloned by limiting dilution in the absence of tetracycline. For screening, wells were duplicated and grown either with or without 1 μg of tetracycline per ml for 24 hours. The clone exhibiting best control of expression was subcloned a second time and named snl-2. For induction experiments, snl-2 were grown at 3×10⁶ cells per ml and incubated with 1 μg tetracycline per ml for the indicated periods of time. For reversibility experiments, snl-2 trypanosomes induced for 48 hours were washed 3 times and resuspended in fresh medium without tetracycline.

Monoclonal antibodies L8C4 (recognising PFRA) (Kohl et al., 1999) and TAT-1 (anti-tubulin, Woods et al., 1989) were used as hybridoma supernatants. For immunofluorescence, trypanosomes were spread on poly-L-lysine coated slides, fixed in methanol and processed as described (Sherwin et al., 1987). Slides were examined with a Zeiss Axioskop or a Leica DMRXA microscope. Images were captured using a cooled slow scan CCD camera and processed in Adobe Photoshop. For immunoblotting, 10⁷ trypanosomes were washed twice in PBS, resuspended in Laemmli buffer in the presence of proteases inhibitors and boiled for 10 minutes before SDS-PAGE. Proteins (10 μg per lane) were transferred to nitrocellulose membranes and stained with India ink before processing. Final detection was carried out using ECL according to manufacturer’s instruction (Amersham).

Our original Southern blot analysis of genomic DNA from snl-1 trypanosomes showed that the pαT6451 plasmid was not inserted at the expected targeted procyclin locus but was present within one of the PFRA alleles (Bastin et al., 1998). This was confirmed by extensive Southern blotting using different probes such as the PFRA gene, the hygromycin-resistance gene and the pGem backbone plasmid (not shown). The plasmid had presumably been integrated at this site by a rare recombination event (Fig. 1A). This view was strengthened by analysis of the other drug-resistant transformants from this original experiment since they all showed the expected integration at the procyclin locus. None of these exhibited the paralysed phenotype or lack of PFR structure. Trypanosomes are diploid and the insertion in snl-1 had occurred at only one allele, the other one being unchanged as compared to wild-type genomic DNA (Fig. 1A). Such an insertion results in the introduction of 2 promoters in the opposite orientation relative to the PFRA genes, therefore providing the capability for production of 2 PFRA antisense RNA molecules (Fig. 1A).

We were able to reproduce the snl-1 genotype and its paralysis phenotype by deliberately inserting the pαT6451 plasmid within one of the 2 PFRA genomic clusters. This was achieved by linearisation in the PFRA sequence at the BsmI unique restriction site. Repeated attempts to accomplish ablation of PFRA expression and a paralysis phenotype by antisense expression from distant sites in the genome such as the procyclin or rDNA gene loci (Fig. 1B) were unsuccessful. Similarly no paralysis phenotype resulted from expression of antisense immediately downstream of the PFRA gene cluster (Fig. 1C). All the experiments suggested that the effect was only produced by insertion into the PFRA gene cluster, with promoters operating in the opposite direction to endogenous transcription (insertion of a plasmid similar to pαT6451 but with promoters integrated in the same direction as PFRA transcription did not produce PFRA ablation, not shown). This insertion could lead to in situ concomitant and overlapping production of sense and antisense PFRA RNA, likely to form PFRA dsRNA and hence produce the RNAi effect that has been demonstrated in several species (reviewed by Fire, 1999).

To address our hypothesis of the importance of in situ, overlapping, production of sense and antisense PFRA RNA, we constructed the pααPFRA430 plasmid, able to express an RNA containing linked sense and antisense copies of the PFRA gene, potentially forming a dsRNA in a hairpin structure (Fig. 1D). To control the expression of this RNA, we used a tetracycline-inducible promoter (Wirtz and Clayton, 1995). In these conditions the modified trypanosome promoter will lead to regulatable transcription of the downstream gene dependent on the presence of tetracycline. The pααPFRA430 plasmid, linearised with EcoRV, was transfected into trypanosomes expressing the tet-repressor and inserted in the inverted rDNA spacer. Resistant populations were cloned in the absence of tetracycline and these cells exhibited the classic aspect of wild-type trypanosomes with normal growth and motility. However, when cultures were grown in the presence of tetracycline, they still divided normally but exhibited serious motility defects similar to those observed for the snl-1 mutant. This was the case for all the transformants obtained with the pααPFRA430 plasmid (more than 50 clones from 4 different transfection experiments). Consequently, these cells were called snl-2. The paralysed phenotype was also observed in transformants where the pααPFRA430 plasmid, linearised with SalI, was inserted downstream of the actin gene. Reduction in motility was also observed in the related parasite L. mexicana after double gene knock-out of the PFRA homologue (Santrich et al., 1997; Maga et al., 1999). In contrast, transfection with the control plasmid pPFRAPFRA430 (2 tandem sense PFRA copies) yielded transformants that all behaved exactly like wild-type trypanosomes in both induced and non-induced conditions.

Immunofluorescence with the anti-PFRA antibody L8C4 (Kohl et al., 1999) performed on snl-2 trypanosomes grown without tetracycline showed the normal bright staining of the flagellum in all cells (Fig. 2A), whereas all the cells from induced samples exhibited no signal at all. This indicates the absence of a normal PFR. In contrast, expression levels of other proteins such as tubulin, procyclin or protein components of the basal body, the nucleus and the flagellum attachment zone filament were not modified (data not shown). Immunoblot analysis with L8C4 showed that PFRA was totally absent in snl-2 samples induced for 48 hours, in contrast to the non-induced situation (Fig. 2B). Northern blots were carried out on total RNA extracted from snl-2 trypanosomes induced for 0, 4 or 72 hours (Fig. 2C). The PFRA mRNA was barely visible in samples induced for only 4 hours and was not detectable at all after 3 days of induction, when the paralysis phenotype was extreme. A sense probe failed to detect any PFRA antisense molecule and none of the probes could detect an RNA corresponding to the expected total mass of the PFRA dsRNA.

The paralysis phenotype was exhibited by the snl-2 cells for at least 3 months (>300 generations), as long as they were maintained in the presence of tetracycline. In protein expression studies, the tetracycline-repressor system allows titration of the amounts of RNA and consequent protein by modulating the tetracycline concentration in the medium (Wirtz and Clayton, 1995). We grew snl-2 cells for 48 hours in the presence of incremental series of tetracycline concentrations (Fig. 2D). When grown with 1ng-1μg of tetracycline per ml, all the snl-2 cells were paralysed and the PFRA protein could not be detected on a western blot (lanes 1-4). An abrupt change was observed at 0.1 ng per ml, where most cells appeared to be paralysed, but with a minority of swimming trypanosomes. Only a weak PFRA band was detected on the immunoblot (lane 5) and immunofluorescence of such populations revealed a mixture of brightly stained cells with others that fail to produce any signal at all (not shown). Only a few ‘intermediates’ or weakly labelled cells were identified. Finally, when snl-2 were incubated with 0.01 ng of tetracycline per ml or less, all the cells behaved like wild-type and exhibited the normal amount of PFRA (lanes 6-10), as observed in non-induced samples (lane 11). Therefore, it appears that there is a critical concentration effect operating in this RNAi induction system above which a maximum phenotype is accomplished.

In trypanosomes, the flagellum is attached along the length of the cell body. Our previous studies of the trypanosome cell cycle show that the old flagellum remains in place whilst the new one is constructed, always at the posterior end of the cell (Sherwin and Gull, 1989a). PFR construction in the new flagellum starts at around 0.52 of the unit cell cycle (Woodward and Gull, 1990) and closely follows the flagellar axoneme in wild-type trypanosomes. DAPI staining shows that the kinetoplasts (mitochondrial DNA) are segregated first, followed by mitosis in the nucleus. Hence, a trypanosome which has 2 kinetoplasts and 2 nuclei (or 2K2N) will be close to the end of its 8.5 hour cell cycle and will also possess both an old and a new flagellum (Sherwin and Gull, 1989a). So, if one views the 2K2N cells within the population at various points during an induction experiment, one can assess the effect of PFRA dsRNA expression on PFRA protein levels, localisation and the consequences for flagellar construction.

snl-2 cells induced to express the PFRA dsRNA were sampled over a 24 hour period and were processed for immunofluorescence with the anti-PFRA antibody (Fig. 3A). In each sample, we concentrated on the 2K2N cells since these provided an indicator, in the same trypanosome, of the effect of PFRA disappearance on new and existing flagella. The RNAi effect on PFRA expression was already distinguishable after only 2 hours of induction. At this time, 2K2N cells showed a lack of PFRA staining at the distal tip of their new flagellum (Fig. 3A). This was more pronounced in such cells after 4 hours of induction (compare with 0 hour in Fig. 3A). 2K2N trypanosomes in populations induced for 8 hours showed an absence of PFRA in their new flagellum. In all cases, PFRA present in the old flagellum was not noticeably affected.

Western blot analysis was performed on such samples and probed with the anti-PFRA antibody L8C4 (Fig. 3B). In non-induced cells, a strong band was detected that only showed a slight reduction after 8 hours of induction (approx. one cell cycle period), since PFRA protein synthesised before the induction was still present in the old flagella of many cells. A considerable difference was observed after further cell divisions and at 24 hours only a weak band remained. After induction for a week, no PFRA band could be detected in snl-2 samples, in contrast to the original snl-1 mutant where a small amount of PFRA was still present.

To see whether RNAi effects were reversible, snl-2 trypanosomes were induced to express PFRA dsRNA for 48 hours, then tetracycline was removed from the medium and the paralysed cells were washed extensively (Fig. 4A). Low amounts of PFRA could already be detected 2 days after removal of the inducer, but motility was not restored at that stage. One day later, the amount of PFRA was similar to that observed in the non-induced control and cells displayed normal motility. To make sure that such results were not due to the emergence of revertant cells, this population was grown in the presence of tetracycline and this again led to PFRA ablation and cell paralysis (Fig. 4A). Thus the snl-2 trypanosomes represent a conditional mutant, lacking PFRA and being paralysed when RNAi is induced whilst having a wild-type phenotype under non-induced conditions. Such a system allows us to ask what constraints are placed on building portions of a flagellum and what dependency relationship exist? Is construction of a PFR only possible when the flagellar axoneme is being elaborated or can the complex PFR structure be assembled in mature flagella which lack such a component? To address these issues, we washed tetracycline out of an induced, paralysed, snl-2 culture and viewed the cells by immunofluorescence microscopy. We examined trypanosomes at all stages of the cell cycle at time intervals over a 72 hour period. As before, cells having an old and a new flagellum were particularly informative. Essentially, we asked whether PFRA returned to both flagella or was only incorporated in the new flagellum. The answer appears to be the former (Fig. 4B). We did not at any time point see cells with one bright flagellum and one unstained flagellum. Rather, regardless of the individual cell to cell variation as the population gained PFRA and motility, both old and new flagellum within each cell showed essentially the same level of staining (Fig. 4B). This experiment also shows that PFRA protein can access equally both a mature and an elongating flagellum.

Our work strongly suggests that overlapping expression of sense and antisense PFRA RNA is required for successful ablation of gene expression in trypanosomes. In contrast, expression of antisense PFRA RNA alone did not lead to PFRA ablation, even when produced immediately downstream of the PFRA genes. These latter results are similar to reports of expression of large amounts of antisense RNA of the trypanothion reductase gene in Trypanosoma cruzi and in T. brucei that in each case did not affect significantly protein expression (Tovar and Fairlamb, 1996; Clayton, 1999).

In contrast, in the snl-1 mutant, the insertion of the pαT6451 plasmid results in the integrated promoter potentially driving the transcription of two antisense PFRA RNA molecules lacking either a valid 3′ UTR, or both 5′ and 3′ UTR (Fig. 1A). In addition, we assume that the endogenous, distant promoter 5′ to this PFRA gene cluster, responsible for normal transcription, is still active. This will result in polycistronic transcription of sense RNA from a total of 5 PFRA genes, however, the second gene would lack a valid 3′UTR and the third gene a 5′UTR (Fig. 1A). In such conditions, these RNAs might not be processed immediately and could anneal with each other or with normal PFRA sense RNA and form PFRA dsRNA. The presence of dsRNA has been shown to lead to potent and specific degradation of the corresponding mRNA in several organisms (Fire, 1999) including trypanosomes (Ngô et al., 1998).

In the snl-2 cell line, expression of an RNA molecule containing both sense and antisense of the PFRA gene produces even more efficient PFRA ablation than the original snl-1. Such an RNA is highly likely to form a stable dsRNA, that could initiate RNAi even when expressed from genomic sites distant from the endogenous PFRA genes. It now remains to be seen whether the methodology could be applied to other genes in trypanosomes. In the related parasite L. donovani, expression of antisense RNA of the A2 gene produced successful A2 protein expression ablation (Zhang and Matlashewski, 1997). However, in these experiments, the construct used to express the A2 antisense RNA was flanked by 2 copies of the A2 3′UTR in inverted orientation. With hindsight it is likely that these are able to form A2 dsRNA and to initiate RNAi.

Our development of an inducible RNAi system that is particularly effective at massively reducing target protein expression reveals interesting aspects of the RNAi phenomenon. First, the system is very stable giving long term control whilst inducer is present (>300 generations). Second, our studies show that is very fast acting, and a reversible phenomenon. This latter result suggests that RNAi consequences are not irreversible modifications at the DNA levels as initially suggested (Wagner and Sun, 1998). In addition, a critical level of PFRA dsRNA appears to be needed to produce the full molecular and cellular phenotype of the snl-2 mutant. This threshold effect could be significant when taken together with evidence pointing towards amplification/enzymatic steps in RNAi operation (Fire et al., 1998). This amplification step is also underlined by the recent discovery of the central role of a putative RNA-dependent RNA polymerase in mutants disabled in RNAi (Cogoni and Macino, 1999a; Smardon et al., 2000).

The inducible RNAi system for PFRA acts as an effective conditional PFRA mutant allowing the interrogation of flagellar ontogeny. Under normal circumstances, the PFR is constructed at the same time as the axoneme in the new flagellum. Our studies of how the snl-2 mutant population reacquires PFRA show that PFR proteins can access both an existing as well as a growing flagellum. This extends and clarifies our earlier observation (Bastin et al., 1999b) that an epitope tagged PFRA protein was preferentially located in the new flagellum. Given these new results, it appears that flagellar proteins can access equally both old and new flagella but are found preferentially in the new because of the assembly process. Furthermore reacquisition of a full PFR by old flagella in snl-2 suggests that such flagella still maintain an intraflagellar transport and assembly system (Cole et al., 1998; Pazour et al., 1998, 1999; Porter et al., 1999; Rosenbaum et al., 1999).

Moreover, the full PFR is reacquired by old and new flagella equally, showing that full flagellum construction is not necessarily required to be concomitant with original axoneme construction. Whilst we have studied the relationships between two diverse flagellum structures, these results are reminiscent of how different components of the axoneme are reacquired. In Chlamydomonas, various authors have shown that specific components can be added to existing flagella, so rescuing the mutant phenotypes (Lefebvre and Rosenbaum, 1986).

The fast-acting RNAi action on PFRA expression whilst the cell is constructing a new flagellum reveals interesting aspects of flagellum assembly in trypanosomes. First, the pattern of inhibition of PFR assembly in the new flagellum shows that assembly is polar and occurs at the distal end. This is similar to the pattern of axoneme assembly (Witman, 1975; Sherwin and Gull, 1989b; Johnson and Rosenbaum, 1992). The PFR in the old flagellum is essentially unchanged by PFRA ablation suggesting a high level of stability and low turn-over. This result confirms in a novel manner that found by ourselves when we studied patterns of insertion of epitope tagged PFRA proteins into new and old flagella (Bastin et al., 1999b).

Finally, this effective, fast-acting, heritable and reversible RNAi system has proven to be extremely valuable for dissecting aspects of flagellum ontogeny in trypanosomes. Given the diploid nature of this and other parasites, this approach is likely to be of major use in studies as diverse as revealing essential cell functions and target verification in drug development.

We thank Klaus Ersfeld and Paul McKean for critical reading of the manuscript and Christine Clayton for providing the plasmids pHD430 and pHD451. This work was supported by a Wellcome Trust Programme and Equipment Grant. K.E. was funded by a BBSRC Research Master’s studentship and L.K. by an International Wellcome Fellowship.