Functional genomics in Trypanosoma brucei identifies evolutionarily conserved components of motile flagella

Eukaryotic cilia and flagella are complex, highly conserved, microtubule-based structures that are important in a number of biological processes. Cilia and flagella drive cellular movement of unicellular eukaryotes, including several important human pathogens. Examples include multiflagellate organisms such as Giardia lamblia and Trichomonas vaginalis, which respectively, cause epidemic diarrhea and trichomoniasis, the latter being the most common, non-viral, sexually transmitted disease in the world (Rughooputh and Greenwell, 2005). Kinetoplastid parasites, which cause African sleeping sickness, Chagas disease and leishmaniasis, all depend on a single flagellum for cell motility in their insect vectors and mammalian hosts. Motile flagella also play a prominent role in the mating cycle of apicomplexan parasites that cause malaria and toxoplasmosis (Ferguson, 2002; Vlachou et al., 2006). Notably, the role of flagellar motility in protozoa is not limited to cell movement, but also contributes to nutrient uptake (Williams et al., 2006) and cell division (Brown et al., 1999; Ralston et al., 2006). For example in Trypanosoma brucei, the causative agent of African sleeping sickness, the flagellum is now recognized to be a multifunctional organelle with crucial roles in cell motility, cell morphogenesis, cytokinesis and host-parasite interactions (Hill, 2003; Kohl et al., 2003; Ralston et al., 2006). Recent work strongly suggests that flagellar motility is required for viability in the bloodstream stage of the T. brucei lifecycle (Broadhead et al., 2006; Ralston et al., 2006; Ralston and Hill, 2006). Therefore, in protozoan parasites the flagellum is essential for disease pathogenesis and parasite development.

Flagella and cilia are also important for normal human physiology and ciliary defects lead to a wide variety of diseases and developmental abnormalities (Afzelius, 2004; Pan et al., 2005; Snell et al., 2004). In humans, one of the most familiar flagellum-related disorders is male infertility, in which flagellated sperm lack effective motility (Sapiro et al., 2002). Motile cilia are also important for moving fluids across the surface of epithelial cells of the respiratory tract and female reproductive tract, as well as ependymal cells that line the ventricles of the brain and nodal cells of the developing embryo (Afzelius, 2004). The importance of these functions is evidenced by clinical features of immotile cilia syndrome (also known as primary ciliary dyskinesia, PCD), which include respiratory deficiencies, hydrocephalus and situs inversus (Afzelius et al., 2001). In addition to their role as organelles of motility, human cilia serve sensory and transport functions and loss of this functionality underlies major human diseases including retinopathies, polycystic kidney disease and Bardet-Biedl syndrome (Pan et al., 2005; Pazour and Witman, 2003; Wang et al., 2006).

The axoneme of motile cilia typically contains nine outer doublet microtubules arranged around a pair of singlet microtubules. Dynein arms protruding from one outer doublet toward the adjacent doublet provide the motive force for flagellar motility (Summers and Gibbons, 1971). Radial spokes extend inward from the outer doublets to the central pair apparatus and are required for motility (Witman et al., 1978). Connecting the outer doublets to one another are nexin links (Warner, 1976), which are thought to help maintain outer-doublet organization (Cibert, 2001; Lindemann, 2004). In its simplest form, flagellar movement can be described as the sliding of outer doublet microtubules relative to one another powered by the dynein arms (Brokaw, 1972; Satir, 1968; Shingyoji et al., 1977; Summers and Gibbons, 1971). Attachment of doublets to the basal body and to each other limits sliding and this produces a bend, which is propagated along the axoneme through the coordinated activation and deactivation of dyneins. Although dynein arms, radial spokes and the central pair apparatus have been well studied, relatively little is known about proteins responsible for the coordinated regulation of sliding and resisting forces necessary for wave propagation. Genetic and biochemical studies in Chlamydomonas reinhardtii suggest that regulation is achieved in part by the dynein regulatory complex (DRC), which acts as a reversible inhibitor of dynein (Huang et al., 1982; Piperno et al., 1992; Gardner et al., 1994). Modeling studies suggest that beat regulation also involves transverse forces supplied by the elastic nexin links in response to axonemal curvature (Cibert, 2001; Lindemann, 2004). Only one subunit of the DRC is known (Ralston et al., 2006; Rupp and Porter, 2003) and the composition of nexin links is completely unknown, therefore, identification of proteins specifically important for flagellar beat and regulation represents a major challenge for understanding flagellum function.

We previously demonstrated that the DRC is an evolutionarily conserved system for dynein regulation and that trypanin (TPN) functions as part of this system in T. brucei (Ralston et al., 2006). TPN is highly conserved among organisms with motile flagella, whereas it is absent in organisms that contain non-motile flagella or lack flagella altogether (Hill et al., 2000; Ralston et al., 2006). We reasoned that additional proteins important for motility and regulation of motility would be similarly conserved and we therefore used comparative genomics to identify genes with the same phylogenetic distribution as TPN. Comparative genomics has proven effective previously for identification of candidate flagellar and basal body genes (Avidor-Reiss et al., 2004; Li et al., 2004), although previous studies did not exclusively focus on organisms with motile flagella. Using this approach we identified 50 genes unique to organisms with motile flagella. This group of genes, referred to as T. brucei components of motile flagella (TbCMF), encode homologues to 17 previously identified flagellar proteins, three proteins that were characterized subsequent to our screen (Broadhead et al., 2006) and 30 novel proteins that have not yet been characterized in any organism. We analyzed the 30 novel genes and several known genes that had not been investigated in T. brucei. RNAi of these 41 TbCMF genes demonstrated a role in flagellar motility in the majority of cases. Ultrastructural and motility analyses identified one family of novel TbCMF proteins that appear to function as part of the nexin links. The proteins identified in this study are important for understanding flagellar motility in microbial pathogenesis and human physiology.

We performed an `in silico' screen using available genome databases to identify genes that are conserved uniquely among organisms with motile flagella (Fig. 1). Our screen led to the identification of 50 predicted T. brucei proteins that are conserved in Trypanosoma cruzi, Leishmania major, Chlamydomonas reinhardtii, Homo sapiens, Mus musculus, Ciona intestinalis and Drosophilia melanogaster, and are not found in Caenorhabditis elegans or Arabidopsis thaliana (Fig. 1, supplementary material Table S1). We refer to this group of proteins as TbCMF proteins

As expected, the set of TbCMF proteins includes homologues of proteins previously demonstrated through genetic and biochemical studies to be bona fide flagellar components, as well as proteins identified in recent genomic and flagellar proteomic analyses (supplementary material Table S2) (Avidor-Reiss et al., 2004; Broadhead et al., 2006; Li et al., 2004; Ostrowski et al., 2002; Pazour et al., 2005; Smith et al., 2005). Previously characterized flagellar proteins that met the screen criteria include radial spoke proteins (RSP3, 4 and 6), DRC proteins (TPN), protofilament ribbon proteins (Rib72 and Rib43a), an MBO2 homologue, and several axonemal dynein subunits (IC140, IC138, IC78, LC7a, LC7b and LC1). A group of predicted dynein heavy chain proteins were originally identified in the screen, but were not included in the final dataset, because our analysis did not distinguish them unambiguously from cytoplasmic dynein sequences. The TbCMF dataset was enriched for proteins containing WD repeats, leucine-rich repeats, EF hand calcium-binding domains and IQ calmodulin-binding motifs (supplementary material Table S2), which are commonly found in flagellar proteins (Li et al., 2004). These results validate the in silico screen as yielding a dataset enriched for flagellar proteins. Notably, the TbCMF dataset also includes 16 proteins not found in a proteomic analysis of extracted axonemes from T. brucei, as well as 28 proteins that were not identified in previous comparative genomic studies (Avidor-Reiss et al., 2004; Broadhead et al., 2006; Li et al., 2004; Ostrowski et al., 2002). Thirty-three TbCMF proteins had not been functionally characterized at the time of the screen. Three of these proteins have subsequently been demonstrated by RNAi to be required for normal flagellar function in T. brucei (Broadhead et al., 2006). The remaining 30 proteins have not been examined in any organism.

Since ciliary defects underlie several human diseases (Ibanez-Tallon et al., 2003), we asked whether human homologues of TbCMF genes were associated with ciliary diseases or mapped to disease loci for which the disease gene is currently unknown. As shown in supplementary material Table S2, several human TbCMF homologues fall within loci linked to ciliary-based diseases in mammals. In some cases, this connection has been noted previously. For example, TbCMF 2, TbCMF 3, TbCMF 4 and TbCMF 34 are homologous to the human Rib72/EFHC1 protein, which has been directly implicated in juvenile myoclonic epilepsy-1 (Suzuki et al., 2004). In other cases, this connection to human disease has not previously been reported, e.g. the human homologues of TbCMF 10 and TbCMF 56 encode novel proteins and are found at loci associated with retinitis pigmentosa and PCD, respectively.

To address the hypothesis that TbCMF genes are required for flagellar motility, 41 of the 50 genes were targeted for tetracycline-inducible RNAi knockdown and the resulting mutants were assessed for motility defects. We focused our attention on uncharacterized genes but also included some known flagellar genes because the vast majority of these have only been investigated in C. reinhardtii. The nine remaining genes were not studied further because they have previously been characterized in T. brucei, have been shown to have a role in motility in C. reinhardtii, or will be the focus of future work.

The first level of analysis was to examine each RNAi knockdown mutant at the whole-culture level by microscopy. When induced, most TbCMF-knockdown mutants formed clusters of multiple cells that remained physically attached. The severity of clustering and time of onset differed among the mutants and this difference was used to place each mutant into one of four distinct phenotypic classes (Fig. 2A). Twenty-four percent of mutants fell into the first class (Class 1), described as `unaffected', because after six days of induction the induced cultures remained indistinguishable from uninduced cultures. Fifteen percent of the mutants exhibited a `mild' clustering phenotype (Class 2 mutants), where small clusters of 5-10 cells appeared within 2 days post induction (dpi), and these clusters remained constant in size and remained homogenously distributed. There was no apparent growth defect associated with Class 2 mutants. Approximately 34% of TbCMF mutants exhibited a `moderate' clustering phenotype (Class 3 mutants), where small clusters appeared in the culture at 2 dpi and grew in size and increased in number over time. Class 3 mutants were viable, but grew at a reduced rate. Finally, 27% of mutants exhibited a `severe' clustering phenotype (Class 4 mutants), where visible clusters appeared at 1 dpi and increased in size and number over time. Class 4 differed from Class 3 in that these mutants were ultimately lethal. To determine whether differences in phenotypic severity were the consequence of different levels of knockdown, northern blots were performed to examine knockdown efficacy in two sample mutants from each mutant class. Clear and specific knockdown of the targeted mRNA was observed in each case (Fig. 2B). Therefore, phenotype severity is not simply a reflection of the degree of knockdown.

We previously demonstrated that flagellar motility defects can interfere with the completion of cytokinesis in T. brucei, resulting in the accumulation of multicellular clusters (Ralston et al., 2006). Therefore the clustering phenotype of TbCMF knockdown mutants supports the idea that these genes are required for normal flagellar motility. However, cytokinesis failure can also occur as a result of other primary defects (Das et al., 1994; Hammarton et al., 2005). To determine whether TbCMF mutants had true motility defects, sample mutants from each phenotype class for which mRNA knockdown was confirmed were selected for more in-depth analysis. These mutants were examined independently by sedimentation assays (Bastin et al., 1999; Ralston et al., 2006) (Fig. 3A) and direct microscopic observation (Fig. 3B, Fig. 4, and supplementary material Fig. S1). These assays were initiated before clustering became apparent in the culture to ensure that any defect observed was due to loss of motility and was not secondary to cell clustering. The class 1 mutant TbCMF 15 did not sediment significantly more than uninduced cells over the course of the assay, whereas TbCMF 3, a Class 2 mutant, showed mild sedimentation. TbCMF 46 (Class 3) and TbCMF 9 (Class 4) mutants sedimented extensively, compared with uninduced cells (Fig. 3A). The extent of sedimentation roughly correlated with the clustering phenotype of these mutants. Though only four mutants are shown in Fig. 3A, 25% of the TbCMF mutants have been tested in this quantitative assay and this correlation has held true for all mutants tested.

To examine directly how the movement and gross structure of the flagellum was affected by loss of TbCMF gene expression, knockdown mutants were examined by high-resolution differential interference contrast (DIC) microscopy. No gross differences in length or outer morphology of the flagellum were observed in any of the mutants examined. In terms of motility, uninduced cells exhibited normal three-dimensional, auger-like movement (Hill, 2003), with extremely rapid and vigorous flagellar beat as did two Class 1 mutants, TbCMF 15 and TbCMF 33 (not shown). TbCMF 3 (supplementary material Movie 1) and TbCMF 2 (not shown) also displayed normal flagellar movement, although slight irregularities may not have been detectable. In TbCMF 46 (supplementary material Movie 2) and TbCMF 49 (not shown), flagellar motility was partially restricted, because a waveform could be produced and propagated, but flagellar movement was limited to two dimensions. In TbCMF 5 and TbCMF 9, the majority of cells displayed extremely deficient flagellar movement. For TbCMF 9, there was a gradient of flagellar defects. Most of the TbCMF 9 mutant cells displayed almost complete flagellar paralysis, with only a slight twitch of the flagellum that was restricted to a segment of the flagellum and not propagated (supplementary material Movie 3). A few cells had rapidly beating flagella, although movement was only two-dimensional and did not effectively propel the cell. The phenotype of TbCMF 5 mutants (supplementary material Movie 4) was more homogenous, where motility consisted of a strong two-dimensional beat at the distal tip of the flagellum, with severely decreased amplitude at the proximal region of the flagellum. Also, the posterior end of these cells became twisted like a corkscrew, which resulted in a slow spinning motion. Thus, detailed analysis revealed that although the exact beat defects may differ between mutants within a given class, the Class 3 and Class 4 mutants examined had dramatically impaired wave propagation along the flagellum leading to complete loss of propulsive cellular movement (Fig. 3B).

To provide a quantitative assessment of propulsive movement in TbCMF mutants, we developed an assay that allowed for automated tracking of a large number of individual trypanosomes. A representative trace is shown in Fig. 4A for TbCMF 46 grown in the presence or absence of tetracycline. All uninduced cultures examined exhibited essentially the same result. In each case, the majority of uninduced cells were vigorously motile and progressed forward along a curvilinear path for 5 seconds or more. These were classified as `runners' (R). In some cases, uninduced cells were clearly motile and moved away from their point of origin, but only tumbled and did not run for five contiguous seconds. Cells with this type of movement were classified as `tumblers' (T) and were manifested in the trace as tightly wound circling lines. In rare cases, uninduced cells were immotile (I) and did not move from their point of origin. Upon knockdown, the number of runners decreased with a concomitant increase in the number of tumbling and/or immotile cells. In most of the moderate and severe mutants examined, the largest increase was in immotile cells with a moderate increase in tumblers (Fig. 4B). For the mild strains tested, the increase in tumbling cells was comparable to moderate and severe mutants. The number of immotile cells however, was roughly 2.5-fold less than seen in severe and moderate mutants. Motility trace analysis of an additional eight mutants is shown in supplementary material Fig. S1. Thus, motility analysis on 14 total mutants, together with the sedimentation assays, indicates a direct correlation between increased motility defects and increased clustering severity. This demonstrates directly that these mutants do exhibit motility defects and supports earlier reports that motility contributes to cytokinesis in T. brucei (Branche et al., 2006; Ralston et al., 2006).

To understand the molecular mechanisms behind TbCMF motility defects, mutants were examined for ultrastructural defects using transmission electron microscopy. Several mutants were examined and will be the topic of future work. Here we have focused on two class 4 mutants, TbCMF 5 and TbCMF 9, which revealed two important findings. First, no gross defects were observed in TbCMF 5 mutants (Fig. 5A-C), suggesting that the severe motility defect in this mutant may be due regulatory problems rather than gross structural abnormalities. Notably, central pair orientation, which is randomized in the central pair mutants PF16 and PF20 (Ralston et al., 2006), remained fixed in all mutants examined as it does in wild-type trypanosomes (Ralston et al., 2006). The fact that several mutants with varying degrees of flagellar beat defects have normal central pair orientation, suggests that randomized central pair orientation is not a general consequence of abnormal flagellar beat. Thus PF16 and PF20 are likely to function specifically in maintenance of central pair orientation. This conclusion is also supported by recent independent studies (Gadelha et al., 2006).

The second finding was a striking axonemal defect observed in TbCMF 9 mutants. These mutants exhibited disordered and fractured axonemes in 44% of the sections examined, which in the mildest cases showed doublets falling away or completely separated from the rest of the axoneme (Fig. 5D,E). In more severe cases, the axoneme had split and one half of the axoneme had twisted around to sit `piggyback' on the other outer doublets (Fig. 5F). This gradient of structural abnormalities correlated with the range of mutant phenotypes seen with DIC, suggesting a direct link between the extent of doublet abnormalities and the motility defect. To our knowledge, the ultrastructural defect of TbCMF 9 has not been previously reported for any flagellar mutant in any organism. As discussed below, TbCMF 9 is part of a novel family of proteins that includes two other closely related proteins in T. brucei, TbCMF 76 and TbCMF 76b, with BLASTP expect values of 10e^-59 and 10e^-69, respectively. To determine whether these proteins have similar functions as well as sequence homology, we examined axoneme ultrastructure in TbCMF-76b knockdown mutants. As shown in Fig. 5G-I, the TbCMF 76b mutant exhibited the same range of ultrastructural defects observed for TbCMF 9, including abnormal outer-doublet spacing (5G), displaced outer doublets (5H) and piggybacked hemi-axonemes (5I). These ultrastructural defects were observed in 40% of sections.

These findings suggest that TbCMF 9 and TbCMF 76b are required for maintaining the stability of the connections between outer doublet microtubules. Alternatively, misplaced doublets might be a consequence of a templating defect, i.e. doublet defects in the axoneme arising from basal body defects. We distinguished between these possibilities in two ways. First, transverse sections through the basal body and transition zone of TbCMF 9 mutants revealed normal microtubule arrangements in all cases (not shown). Second, longitudinal sections from both mutants showed outer doublet microtubules that were bent away from the main axis of the axoneme (Fig. 5J, arrows) providing direct evidence that the outer-doublet defects arise within the axoneme and are not the result of a templating defect. Together, the data demonstrate the presence of a novel family of proteins in T. brucei responsible for maintaining the structural integrity of the outer-doublet array in motile flagella.

If TbCMF 9 functions to maintain connections between outer doublet microtubules, it should be localized along the length of the axoneme. We therefore used GFP tagging to determine the subcellular localization of TbCMF 9, as well as TbCMF proteins from other mutant classes. Since overexpression can interfere with protein localization, we used the pLEW100 tet-inducible expression vector (Wirtz et al., 1999) to drive low-level expression of GFP fusion proteins. We first used GFP-tagged TPN as a control. As shown in Fig. 6A, the TPN-GFP fusion protein was expressed only in the presence of tetracycline. Importantly, the fusion protein was expressed at a level roughly equivalent to the endogenous protein and, like endogenous TPN (Hill et al., 2000), was quantitatively associated with detergent-extracted cytoskeletons (Fig. 6A). When live cells were viewed by fluorescence microscopy the TPN-GFP fusion protein was localized along the flagellum (Fig. 6C arrows). There was a low level of autofluorescence in the cell body that was also evident in uninduced cells (Fig. 6C) and untransfected cells (Fig. 6B). This autofluorescence overlapped with mitochondrial markers (not shown) and was lost upon detergent extraction (Fig. 6E). GFP alone was not specifically localized to the flagellum (supplementary material Fig. S2). The localization of TPN-GFP demonstrates for the first time that TPN is localized to the flagellum in live cells and establishes this system as a reliable means to examine flagellar protein localization in T. brucei.

We next constructed GFP fusions for TbCMF 9, as well as sample TbCMF Class 1 (TbCMF 19), Class 2 (TbCMF 40) and Class 3 (TbCMF 46) proteins. TbCMF 9, TbCMF 40 and TbCMF 46 are novel proteins of unknown function. TbCMF 19 is a T. brucei homologue of a characterized C. reinhardtii flagellar protofilament ribbon protein, Rib43a (supplementary material Table S1). In all cases TbCMF-GFP fusion proteins were localized to the flagellum in live cells (not shown) and remained stably associated with the flagellum upon detergent extraction (Fig. 6G). The few spots observed in the cytoplasm of TbCMF-9-GFP cells probably correspond to aggregated protein, because these did not occur in consistent number or placement. Western blots on whole-cell lysates, detergent-soluble proteins, and detergent-insoluble cytoskeletons demonstrated that TbCMF 9, TbCMF 19 and TbCMF 46 were quantitatively associated with the cytoskeleton (Fig. 6H). A significant fraction of TbCMF 40 was released into the detergent-soluble fraction. It is not known whether the soluble fraction represents non-flagellar protein, or protein released from the flagellar compartment. Overall, this localization data supports the proposal that the proteins identified in this study represent conserved components of motile flagella. More specifically, it supports a function for TbCMF 9 as an outer-doublet connector along the length of the axoneme.

The mechanism and regulation of flagellar motility is not fully understood, thus it is important to identify proteins important for proper function and regulation of flagellar beat. To accomplish this we have used the African trypanosome, T. brucei, as a model system to identify motility genes via comparative genomics and functional analysis. We identified 50 genes that are unique to organisms with motile flagella, of which 30 are novel. We used RNAi to investigate the function of these novel genes, as well as several others that had not been functionally characterized in T. brucei previously. The majority of these genes are required for normal cell motility. For 30 of these genes, our study represents the first functional characterization in any organism. Taken together, our analyses of motility defects, protein localization and ultrastructural defects demonstrate that we have identified a novel set of T. brucei proteins that represent conserved components of motile flagella, referred to as TbCMF proteins.

Several recent studies have used proteomic and comparative genomic approaches to identify candidate flagellar components (Avidor-Reiss et al., 2004; Broadhead et al., 2006; Li et al., 2004; Ostrowski et al., 2002; Pazour et al., 2005; Smith et al., 2005). Our effort is distinguished by focusing specifically on organisms with motile flagella and by taking advantage of the powerful tools available in T. brucei for probing the function of the candidate flagellar proteins identified. The impetus for our screen stemmed from the observation that trypanin, which is part of an evolutionarily conserved axonemal dynein regulatory system (Ralston et al., 2006), is conserved only in organisms with motile flagella. Coordinated regulation of axonemal dyneins represents one of the least understood aspects of flagellar function and we reasoned that additional genes important for axoneme motility and regulation of motility would exhibit a similar phylogenetic distribution. As motile cilia and flagella are important for human physiology and essential for viability of bloodstream-form T. brucei (Broadhead et al., 2006; Ralston and Hill, 2006), our findings have direct relevance to infectious and inherited human diseases.

The TbCMF dataset contains subunits of axonemal complexes that are established components of motile flagella including radial spokes, the dynein regulatory complex, and inner and outer dynein arms. Approximately 76% of TbCMF proteins are also represented in one or more recently completed proteomic analyses of eukaryotic flagella (Broadhead et al., 2006; Gibbons, 1963; Ostrowski et al., 2002; Pazour et al., 2005; Smith et al., 2005). We also identified several proteins not found in earlier proteomic and genomic studies, adding to the repertoire of proteins important for flagellar motility. Notably absent from the TbCMF dataset are components of the central pair apparatus, such as PF16 and PF20 (Smith and Lefebvre, 1996; Smith and Lefebvre, 1997). Using T. brucei PF16 as a query, we found a hit in the A. thaliana predicted proteome with an E-value of 10e^-15, which would have excluded this gene from our dataset. Direct protein alignments indicate that this A. thaliana protein is not likely to be a true PF16 orthologue and this reflects an inherent limitation of comparative genomics, namely that some bona fide flagellar proteins might be missed. Nonetheless, the majority of known proteins expected to be unique to motile flagella are in the TbCMF dataset, as are several novel proteins for which functional analysis demonstrated a clear role in flagellar motility.

Several TbCMF proteins are represented as families of two or more related sequences (e.g. TbCMF 2/3/4/34, TbCMF 5/6, TbCMF 40/40a and TbCMF 9/76/76b) and often the family is expanded in T. brucei. For example, there is a single TPN homologue in mammals and C. reinhardtii, but there are two paralogues in T. brucei, referred to as TPN and trypanin-related protein (TRP) (K.L.H. and J. E. Donelson, unpublished observation) (Ralston et al., 2006). Likewise, a single gene in C. reinhardtii encodes the Rib72 protein (Ikeda et al., 2003), but there are four related sequences (TbCMF 2/3/4/34) in T. brucei. Pair-wise alignments of TbCMF 9, TbCMF 76 and TbCMF 76b indicate that they are part of a family that includes two closely related proteins in C. reinhardtii and a single protein in humans. TbCMF 9, TbCMF 76 and TbCMF 76b are distantly related to TbCMF 8, which is the T. brucei homologue of MBO2 from C. reinhardtii (Segal et al., 1984). This expansion might reflect unique aspects of flagellar beat in trypanosomes (Hill, 2003) or might be a manifestation of the asymmetric architecture of the trypanosome flagellum, in which a paracrystalline rod is anchored to outer doublets 4-7 along the length of the axoneme (Cachon et al., 1988). This arrangement imposes unique constraints on systems for regulation and propagation of flagellar beat and these demands might be met by using distinct but functionally related regulatory systems on each side of the axoneme (Ralston et al., 2006).

Motile cilia are crucial for normal human development and physiology and human homologues of TbCMF genes might represent novel disease gene candidates. In support of this, many human CMF genes have been directly implicated in human disease, or map to loci that are linked to diseases that are known to be, or are suspected to be, caused by ciliary dysfunction (supplementary material Table S2). In 70% of these cases, this represents the first time that a connection has been made between this gene and a human disease. Identification of candidate disease genes is especially crucial in the case of PCD, which is genetically heterogeneous, making disease gene identification through linkage analysis extremely difficult. Indeed, the only three genes so far identified as causal in PCD were identified by specifically looking for mutations in candidate disease genes that encode axonemal dynein subunits (Ibanez-Tallon et al., 2003). Juvenile myoclonic epilepsy (JME) is caused by mutations in the flagellar protofilament ribbon protein Rib72 (Ikeda et al., 2005). Although biochemical fractionation demonstrates that Rib72 is an integral component of the axoneme (Ikeda et al., 2003; Patel-King et al., 2002), the function of Rib72 in flagellar motility has not previously been examined. Therefore, our results (Fig. 4) provide the first demonstration that Rib72 is required for flagellar beat, suggesting that JME results from defects in motility functions of the cilium.

We used RNAi to investigate the function of 41 of the 50 TbCMF genes identified. Thirty of these had not previously been characterized in any organism and our data provide the first demonstration that these proteins play a role in flagellar motility. Most TbCMF mutants (76%) accumulate as multicellular clusters that fail to complete cell separation. In at least fourteen cases where motility was examined independently by sedimentation and/or direct microscopic observations the severity of this phenotype correlates with the severity of the motility defect and thus further supports the idea that flagellar motility is required for normal cell division in T. brucei (Ralston et al., 2006). Thus, as shown by indirect and direct methods, TbCMF genes are confirmed in vivo to be functionally important for flagellar motility.

Functional analysis of TbCMF 9 and TbCMF 76b, both members of a novel protein family, demonstrates functional analogy among these proteins. Both TbCMF 9 and TbCMF 76b were identified in a recent axonemal proteomic analysis (Broadhead et al., 2006), although only TbCMF 9 was functionally characterized in that study. When TbCMF 9 (or MENG) was knocked down by RNAi, the cells exhibited reduced motility, but no apparent cytokinesis defect or loss of viability (Broadhead et al., 2006). This result differs from our findings; however, the reason for this is not presently clear. Broadhead and colleagues did not examine the extent of RNAi knockdown, thus knockdown of MENG might not have been as complete as we observed for TbCMF 9 (Fig. 2B). Both TbCMF 9 and TbCMF 76b mutants exhibit motility and cytokinesis defects and share a novel axonemal ultrastructural defect. The affected doublets are most often those not adjacent to the PFR, suggesting that attachment of microtubules to the PFR might provide enough residual stability to keep those doublets properly positioned. In fact `hot spots' for breakage appear to be between doublets 3-4 and 6-7 on either side of the PFR. Hot spots of breakage are also observed in disintegrating gill cilia after detergent and protease treatment to remove nexin links and might also reflect differential activation of specific dynein subsets (Satir and Matsuoka, 1989).

The doublet abnormalities of TbCMF 9 and TbCMF 76b can be explained if they function as part of the nexin links in maintaining the integrity of the outer doublet microtubule array. Nexins are elastic structures that link adjacent outer doublets (Warner, 1976). They provide support and organization to the outer-doublet array and contribute to transverse elastic forces postulated to help regulate dynein activity in response to microtubule sliding and curvature (Cibert, 2001; Lindemann, 2004). These structures have been identified by electron microscopy (Gibbons, 1965; Olson and Linck, 1977; Warner, 1976) and axonemal fractionation (Stephens, 1970; Stephens and Edds, 1976), although they have never been characterized directly by mutational analysis. If nexin links are absent or compromised, one predicts that, as the flagellum beats, the resulting forces would cause outer doublets to separate and fall away from the outer-doublet ring. This is precisely what was seen in TbCMF 9 and TbCMF 76b mutants. Mutants with this defect have not been observed previously and this was not observed in any of the other TbCMF mutants examined in this study. As motility becomes increasingly aberrant, forces within the structurally impaired axoneme may become strong and erratic enough to shift the loose doublets, resulting in the piggyback arrangement observed. Loss of the canonical 9+2 axonemal arrangement is expected to have a severe impact on flagellar beat. If breaks are localized to a single region, one might expect a localized beat that is not propagated along the axoneme. In more severe cases, the entire flagellum might become paralyzed, which is again what we observed in the TbCMF 9 mutant. An alternative hypothesis is that axonemal splitting results from asynchronous dynein activation (Satir and Matsuoka, 1989) and these hypotheses are not mutually exclusive.

The protein composition of nexins is unknown and several lines of evidence suggest that TbCMF 9 and TbCMF 76b are the first ever identified components of these linkages. First, nexins are prominent features of motile axonemes (Bozkurt and Woolley, 1993; Gibbons, 1963), but are not present in immotile sensory cilia of C. elegans (Perkins et al., 1986; Signor et al., 1999), thus, nexin genes are expected to be recovered in our screen. Whether nexins, or other CMF proteins are also present in non-motile cilia in organisms that assemble both motile and non-motile cilia remains to be determined and their function investigated. For instance, the human homologue of trypanin, Gas11, is expressed in mammalian cell types that contain motile cilia, non-motile primary cilia or lack cilia (Colantonio et al., 2006; Whitmore et al., 1998; Yeh et al., 2002), and appears to have been adapted to serve multiple functions in mammals (Colantonio et al., 2006). Second, aberrant arrangement of outer doublet microtubules, as well as the paralyzed and localized flagellar beat defects observed in TbCMF 9 and TbCMF 76b mutants are precisely what one would expect for nexin mutants. Indeed, the outer-doublet defect of TbCMF 9 and TbCMF 76b is reminiscent of that seen in cilia following protease and chemical treatments to solubilize the nexin links and other mechanical impediments to axoneme disintegration (Linck, 1973a; Linck, 1973b; Lindemann et al., 1992; Satir and Matsuoka, 1989). Third, nexin proteins are expected to be localized along the length of the axoneme, as we observed for TbCMF 9. Fourth, nexin links in T. brucei are retained following detergent and salt-extraction of flagellar axonemes (not shown) and biochemical fractionation demonstrates that TbCMF 9 is also retained in axonemal preparations following detergent extraction (Fig. 6H) and salt extraction (not shown). Finally, TbCMF 9 and TbCMF 76b contain domains with homology to domains found in structural maintenance of chromosomes (SMC) proteins, which have functional characteristics similar to what one would hypothesize for a nexin. SMC proteins are known for their role in the structural and functional organization of chromosomes (Hirano, 2006; Nasmyth and Haering, 2005), but also play a ciliary role, because SMC1 and SMC3 have been localized to retinal and renal cilia (Khanna et al., 2005). SMC proteins are large (110-170 kDa) and very flexible, which together with ATP binding (Hirano, 2006), could provide the elastic and dynamic functions ascribed to nexins. SMC proteins generally function as heterodimers of closely related proteins, which might explain why TbCMF 9 and TbCMF 76b are part of a gene family. Since nexins are intimately associated with the DRC (Mastronarde et al., 1992; Nicastro et al., 2006; Woolley, 1997), it would not be surprising if nexin genes exhibit the same phylogenetic distribution as DRC genes. A major finding to some recent cryoelectron studies (Nicastro et al., 2006) is that novel linkers connect outer-arm dyneins with each other and with the inner-arm dyneins and DRC. These linkers are postulated to provide a mechanism for coordinating dynein activity and it will be of interest to determine whether any of the novel CMF proteins encode components of these linkers.

Dutcher and colleagues (Li et al., 2004) recently described a group of predicted proteins that are conserved in flagellated organisms, namely C. reinhardtii, H. sapiens, M. musculus, C. intestinalis and D. melanogaster, but are not found in organisms lacking flagella, namely A. thaliana (Table S1 in supplementary material) (Li et al., 2004). This includes 81 proteins that are also not found in C. elegans, which contains non-motile sensory cilia. To identify T. brucei proteins that are restricted to organisms with motile flagella, the `best matched' accession number for these proteins was used to procure the corresponding protein sequence from the NCBI database. These were then compared (BLASTp; BLOSUM 62) (Altschul et al., 1990) against the T. brucei predicted proteome at the Gene Database (GeneDB, http://www.genedb.org/) (Hertz-Fowler et al., 2004). T. brucei hits with an expected value of 1×10^-20 or less were retained and identical entries were removed, yielding a total of 38 proteins. To supplement this dataset, we used the proteome comparison (Procom) tool developed by Stormo and colleagues (Li et al., 2005). Procom is a web-based tool that allows for rapid, automated comparison of predicted eukaryotic proteomes. T. brucei was used as the anchor organism, H. sapiens, M. musculus, D. melanogaster, L. major, T. cruzi, C. reinhardtii and C. intestinalis were used as intersection organisms, whereas C. elegans and A. thaliana were used as subtraction organisms. An expected value of 1×10^-10 was used for intersect and subtract cut-off values. This analysis yielded 82 entries from the T. brucei predicted proteome. Sequences for each of these 120 proteins (38 + 82) were obtained from GeneDB and redundant entries removed, for a preliminary total of 88 proteins. As a further refinement, we used reciprocal best BLAST to determine whether the top human homologue for each T. brucei protein returned the original T. brucei entry as the top hit when compared against the T. brucei predicted proteome. Individual protein sequences were also evaluated by direct alignment. This reduced our dataset to 46 proteins. Reciprocal BLAST analysis also identified another four T. brucei proteins as paralogues to proteins already on the list. Thus, the final dataset contains 50 proteins. The final TbCMF dataset was compared against the C. briggsiae genome using the same methods listed above and no homologues were identified. Trypanin-related protein, TRP (GeneDB ID Tb09.244.2800), identified in this screen is 28% identical and 43% similar to TPN at the amino acid level and will be described elsewhere (K.H., unpublished observation).

The gene targets for RNAi (400-600bp) were amplified by PCR from 29-13 (Wirtz et al., 1999) genomic DNA using primers specific to an RNAi target sequence identified by the Trypanofan RNAit algorithm (Redmond et al., 2003) (http://trypanofan.path.cam.ac.uk/cgi-bin/rnait.org). The TbCMF gene targets were ligated into the p2T7-Ti/B-RNAi vector, which is a tetracycline-controlled expression vector with opposing T7 promoters (LaCount et al., 2002). Inserts were verified by sequencing at the UCLA genomics center.

For GFP tagging, full-length TbCMF genes were amplified from 29-13 (Wirtz et al., 1999) genomic DNA with primers containing the appropriate restriction sites and ligated into pKH10 (N-terminal tag) or pKH12 (C-terminal tag). For C-terminal tags, TbCMF genes were cloned into the HindIII and XbaI sites at the 5′ end of the GFP ORF in pKH12. pKH12 was generated by removing the GFP cassette from pHD:HX-GFPm3 (Hill et al., 1999) as a HindIII-BamHI fragment and inserting it into the HindIII-BamHI sites of pLEW100 (Wirtz et al., 1999). For N-terminal tagging, TbCMF genes were cloned into XbaI and BamHI sites at the 3′ end of the GFP ORF in the pKH10 expression vector. pKH10 is a derivative of pLEW100 (Wirtz et al., 1999) in which the GFP ORF is flanked with HindIII and XbaI-BamHI restrictions sites at its 5′ and 3′ end, respectively. pKH10 was generated by PCR amplification of the GFP-coding sequence minus its stop codon from pHD496-GFP (Biebinger et al., 1997), using primers to introduce a 5′ HindIII restriction site and 3′ XbaI and BamHI restriction sites. The amplified GFP product was inserted into HindIII and BamHI restriction sites of pKH12. All sequences were verified by DNA sequencing at the UCLA genomics center. For GFP alone in pLEW82 (see supplementary material Fig. S1), the GFP cassette was removed from pHD:HX-GFPm3 (Hill et al., 1999) as a HindIII-BamHI fragment and inserting it into the HindIII-BamHI sites of pLEW82 (Wirtz et al., 1998) to generate pKH15.

Procyclic 29-13 cells (Wirtz et al., 1999) were used to create RNAi and GFP-tagged TbCMF strains. Cells were maintained and transfected as described previously (Hill et al., 1999; Hutchings et al., 2002) and clonal lines were obtained by limiting dilution. Knockdown mutants were induced with 0.3-1 μg/ml tetracycline. For our initial analysis, at least three clonal lines for each knockdown mutant were induced to assess clone-to-clone variability. Very little variation was observed and a single clone for each knockdown was selected for further analysis. For expression of GFP-TbCMF fusion proteins cells were induced with 1 μg/ml tetracycline for 24 hours and viewed on a Zeiss Axioskop II compound fluorescent microscope using a 63× oil objective.

RNA was isolated from induced (same dpi as in motility assays) and uninduced TbCMF strains using a Qiagen RNeasy Miniprep kit according to the manufacturer's instructions. Northern blots (Hill et al., 1991) using 5 μg total RNA were probed with ³²P-labeled DNA fragments corresponding to the region used for RNAi knockdown.

Unless otherwise stated, strains were examined at a time point before clustering became apparent in the culture, typically 24 hours post induction for Class 3 and 4 strains and 48 hours for Class 1 and 2 strains. Digital images and movies were captured using a Sony handycam and directly imported using Adobe Premiere Elements Software. Sedimentation assays were carried out essentially as described previously (Bastin et al., 1999; Ralston et al., 2006) except that OD measurements were done in triplicate and were taken every hour for 10 hours. Whole culture observations of the induced TbCMF cultures in flasks were made every 24 hours for 6 days post induction on a Zeiss Axiovert 200 inverted microscope using a 5× objective. Cells were observed using a 63× oil objective on a Zeiss Axiovert 200 inverted microscope in polyglutamate-coated slide chambers (Gadelha et al., 2005). Log-phase cells of uninduced and induced TbCMF cells were diluted to 1×10⁶ cells/ml and added to a polyglutamate-coated slide chamber (Gadelha et al., 2005) (∼40 μl) before the sides were sealed with a thin layer of Vaseline. The cells were viewed under dark-field illumination on a Zeiss Axioskop II compound microscope using a 10× objective. Approximately 30 seconds of video from separate regions on each slide was captured. Motility traces were generated using Metamorph software (Molecular Devices). Cells that were not tracked for the full 30 seconds, e.g. as a result of leaving the field or plane of focus, were not used in the analysis.

Trypanosome lysates, detergent-soluble proteins, and cell cytoskeletons were prepared as described previously (Hill et al., 2000) and western blotted (Hill et al., 1999). Primary antibody dilutions were as follows, mGFP 1:500 (Clontech 632380) and mTPN 1:5000 described previously (Ralston et al., 2006).

Transmission electron microscopy was performed as described previously (Hutchings et al., 2002). Class 3 and 4 mutants were analyzed 32-48 hours post induction, a time point at which the mRNA is clearly knocked down, and motility defects are evident.

We thank Randy Nessler (University of Iowa) for his assistance with electron microscopy. We are grateful to George Cross (Rockefeller University) for the 29-13 cell line. We also appreciate the helpful comments and discussions from Nancy Sturm. This work was supported by grants from the National Institutes of Health (R01AI52348), Ellison Medical Foundation (ID-NS-0148-03) and Beckman Young Investigator Program to K.L.H. D.M.B. is the recipient of a USPHS National Research Service Award in Microbial Pathogenesis (2-T32-AI-07323) and the UCLA Graduate Dissertation Year Fellowship for 2006-2007. K.S.R. is the recipient of a USPHS National Research Service Award in Genetics (GM07104).