Rab28 function in trypanosomes: interactions with retromer and ESCRT pathways

Membrane trafficking requires temporal and spatial coordination to ensure that molecules reach their correct destinations. In animals, fungi and multicellular plants, this system is highly flexible and facilitates responses to environmental and developmental changes. Rab (Ras-related proteins in brain) GTPases regulate membrane trafficking and function in many transport steps (Seabra et al., 2002; Behnia and Munro, 2005). In the GTP-bound form, Rabs are deployed to membranes and recruit effector molecules involved in vesicle formation, cytoskeletal-dependent transport, tethering and fusion with destination organelles (Grosshans et al., 2006). Rabs are regulated by multiple proteins, including GTPase-activating proteins (GAPs) and guanine exchange factors (GEFs), making them pivotal for coordinating membrane transport (Stenmark, 2009; Brighouse et al., 2010).

Rab28 is a divergent member of the Rab family, with lower sequence identity against canonical Rab proteins (31–33%) than is typical (>40%) (Brauers et al., 1996; Pereira-Leal and Seabra, 2001), and also with little homology beyond the GTP-binding site and divergence within phosphate-binding regions. Conserved residues in Rab switch/interswitch regions bind many proteins, including GAPs, GEFs and effectors (Eathiraj et al., 2005; Ostermeier and Brunger, 1999; Delprato and Lambright, 2007; Merithew et al., 2001); significantly, the WDTAGQE motif in Rab28 is diverged to WDIGGQT. Further, the switch domain in human RAB28 undergoes a greater conformational change between GTP and GDP states than other Rab proteins (Lee et al., 2008). Two Rab28 splice variants that differ at their C-termini are present in mammals. The short isoform is ubiquitously expressed and the long isoform restricted to testes (Brauers et al., 1996). Human RAB28 contains an S[D/E][D/E]E motif within the unstructured N-terminal domain, which binds BARD1, an interaction partner of the breast cancer gene product BRCA1 (Irminger-Finger et al., 2006). However the role of Rab28 in endomembrane trafficking remains undefined.

Multivesicular bodies (MVBs) are late endosomal sorting compartments. Ubiquitylated proteins destined for lysosomal degradation are concentrated into intralumenal budding vesicles by the endosomal sorting complex required for transport (ESCRT complex), whereas resident proteins and those destined for retrograde transport are retained within the limiting membrane (Felder et al., 1990; van Deurs et al., 1993; Wollert and Hurley, 2010). The retromer VPS26–VPS29–VPS35 subcomplex selects cargo destined for the Golgi complex into outwardly budding tubules generated by sorting nexins (SNXs) (Arighi et al., 2004; Seaman, 2004). Retromer and ESCRT complex activities must be tightly coordinated but the regulatory mechanism(s) responsible are unknown.

Saccharomyces cerevisiae lacks Rab28, so we selected Trypanosoma brucei as an alternative model. T. brucei is experimentally tractable and a member of the Euglenozoa, a group suggested to be close to the eukaryotic evolutionary root, with a well-characterized endomembrane system (Cavalier-Smith, 2010; Field and Carrington, 2009). Trypanosomes have an ordered organellar anatomy with endocytosis and exocytosis restricted to an invagination of the plasma membrane, the flagellar pocket. Clathrin-mediated endocytosis is the sole uptake mechanism and endocytosis is developmentally regulated (Langreth and Balber, 1975; Morgan et al., 2001; Allen et al., 2003; Hung et al., 2004; Jeffries et al., 2001).

We demonstrate that T. brucei Rab28 partially colocalizes with vacuole protein sorting 23 (Vps23) and that it participates in endocytic transport pathways. Using RNA interference (RNAi) we find that Rab28 mediates maintenance of the Golgi complex and maintains expression levels and locations of retromer and ESCRT complex subunits. These data suggest Rab28 functions in coordinating late endocytic events.

We examined the role of the T. brucei orthologue of mammalian Rab28 in membrane transport. T. brucei offers an attractive system in which to study this Rab protein on account of a streamlined endocytic system coupled to a high level of definition, together with extensive evidence that Rab orthologues maintain broadly similar functions across deep evolutionary time (Field and Carrington, 2009; Brighouse et al., 2010). T. brucei Rab28 (Tb927.6.3040) was initially identified by comprehensive screening of the trypanosome genome for Ras- and Rab-like small GTPases (Berriman et al., 2005; Ackers et al., 2005); T. brucei Rab28 shares 49% identity and 58% similarity to H. sapiens RAB28 and extensive similarity to orthologues in other taxa, notably within the C-terminal hypervariable domain. Rab28 is widely distributed across the Eukaryota, despite secondary losses resulting in the absence of Rab28 from Plantae, Fungi and Amoebozoa, and therefore Rab28 is dispensable in certain organisms (Lumb and Field, 2011). To examine T. brucei Rab28 expression in trypanosomes we initially analyzed mRNA levels; quantitative real time PCR (qRT-PCR) confirmed Rab28 transcripts in both bloodstream form (BSF) and procyclic form (PCF) trypanosomes, suggesting a role throughout the life cycle (Fig. 1A).

To determine subcellular location, T. brucei Rab28 was fused to an N-terminal haemagglutinin (HA) or YFP-epitope tag and ectopically expressed in BSF cells. Production of the respective chimeras, TbRab28HA (31 kDa) and TbRab28YFP (51 kDa), of the correct molecular weight were verified by western blotting (Fig. 1B). Indirect immunofluorescence analysis (IFA) on cells expressing TbRab28HA and TbRab28YFP detected discrete puncta in the cytoplasm posterior to the nucleus and anterior to the kinetoplast. No such staining was seen in non-transfected cells (Fig. 1D). Rab28-positive structures replicated following kinetoplast segregation and were partitioned between daughter cells (Fig. 1E). To verify the location of T. brucei Rab28, we raised antibodies against a GST::T. brucei Rab28 fusion protein in rabbits. The specificity of affinity-purified antibody was validated by western blot, and IFA recapitulated the distribution of tagged Rab28 proteins in BSF cells (Fig. 1C,F). This antibody proved to be highly labile and hence could not be used in subsequent analyses. However, the distinct location, highly similar to endogenous T. brucei Rab28 for both the HA and YFP chimeras, argued strongly for a location of Rab28 between the kinetoplast and nucleus. When equivalent Rab28 chimeras were expressed in PCFs, the localization was essentially indistinguishable from BSF, which suggested that the location of T. brucei Rab28 is maintained between developmental stages (Fig. 1G).

The region between the nucleus and kinetoplast in trypanosomes contains the flagellar pocket, endosomes, the lysosome and the Golgi complex, a crowded region that makes fine discrimination between membraneous subcompartments challenging (Field and Carrington, 2009). However, the location of T. brucei Rab28 was consistent with association with one or more of these compartments, especially the endosomes and lysosome (Field et al., 1998; Jeffries et al., 2001; Gabernet-Castello et al., 2009; Leung et al., 2008; Alexander et al., 2002).

We attempted to identify the subcellular location of T. brucei Rab28 more clearly using a combination of wide field and confocal immunofluorescence microscopy, the latter to eliminate potential colocalization in the x-y plane but distinct location in the z-axis (Fig. 2A). Cells ectopically expressing Rab28 were co-stained with antibodies against a validated panel of markers for trypanosome intracellular compartments. We found little colocalization between T. brucei Rab28 and the clathrin heavy chain, despite very close apposition of the membranes. T. brucei Rab28 appeared distinct from Rab5A, Rab5B and Rab11, with essentially no colocalization, despite a presence on structures that are extremely close to each other. Furthermore, T. brucei Rab28 and GRASP are completely distinct. Therefore, we conclude that Rab28 has essentially no steady-state presence on the Golgi complex, clathrin-containing compartments, or early and recycling endosomes.

There was some colocalization with p67, suggesting that T. brucei Rab28 might have a presence on lysosomal or pre-lysosomal membranes (Kelley et al., 1999). Furthermore, Rab28 partly colocalizes with HA-tagged Vps23, an ESCRT I component (Leung et al., 2008; Katzmann et al., 2001; Babst et al., 2000), and to a lesser extent with the retromer subunit, Vps26 (Shorter et al., 1999; Seaman et al., 1998). In summary, these data suggest that T. brucei Rab28 locates close to multiple endosomal constituents but specifically associates more substantially with Vps23-positive structures. By analogy to mammalian cells, these data suggest a presence for T. brucei Rab28 on structures similar to late endosomes and pre-lysosomes. To exclude potential overexpression artefact we also tagged T. brucei Rab28 at the N-terminus in situ, i.e. in the endogenous genomic location (Kelly et al., 2007) (Fig. 2B). In situ T. brucei Rab28 also demonstrated association with Vps23, which was essentially indistinguishable from the ectopically expressed Rab28 localization (Fig. 2A). Higher definition assignment of T. brucei Rab28 is not possible using confocal microscopy due to the extremely close apposition of the organelles and the small volume of the trypanosome cytoplasm in this region.

RNAi is highly efficient in trypanosomes, and especially resistant to off-target effects (Koumandou et al., 2008). Moreover co-suppression requires high sequence similarity, and specific knockdown of trypanosomatid genes sharing over 60% identity has been achieved without off-target effects (Bastin et al., 2000); T. brucei Rab28 is 27% identical to its closest paralogue, Rab23. A plasmid allowing tetracycline (TET)-inducible expression of Rab28 double-stranded RNA was used to generate a T. brucei Rab28^RNAi line. Knockdown was validated by both northern blot and western blot on a Rab28^RNAi mutant ectopically expressing TbRab28HA. Expression of Rab28 mRNA and protein was rapidly suppressed by ~90% after 24 hours, whereas BiP levels remained unchanged, indicating that RNAi was specific (Fig. 3A,B).

A significant and reproducible impact on proliferation of BSF cells was obtained within 48 hours of inducing T. brucei Rab28^RNAi cells (Fig. 3C), indicating that Rab28 is required for normal cellular function. By day 3, the propagation rate of induced cells had decreased by 50%, and remained at this level for the duration of the experiment. The proliferation defect was not due to a cell cycle block, suggesting a more specific functional mechanism whereby proliferation was reduced while cell morphology remained normal (data not shown). The absence of a BigEye phenotype (Allen et al., 2003; Hall et al., 2004) caused by engorgement of the flagellar pocket due to blockade in endocytic membrane uptake, suggests that T. brucei Rab28 does not function in bulk uptake of membrane or early endocytosis, consistent with the absence of Rab28 from early endosomal compartments. Rab28 is scored as non-essential in BSF cells using a whole genome scan approach but, interestingly, does appear to be essential in PCFs (Alsford et al., 2011).

As 90% of the total surface glycoprotein in trypanosomes is the variant surface glycoprotein (VSG), and hence VSG is the overwhelming mannose-containing surface macromolecule, uptake of the mannose-binding lectin concanavalin A (ConA) mainly reports on VSG endocytosis (Overath et al., 1994; Cross, 1996; Mehlert et al., 2002; Allen et al., 2003). To address a possible role for Rab28 in surface protein endocytosis, we monitored ConA accumulation in T. brucei Rab28^RNAi cells. Quantitative fluorescence microscopy clearly showed that accumulation of fluorescein isothiocyanate (FITC)–ConA was compromised in T. brucei Rab28 knockdown cells (Fig. 4A). To verify this observation, cells were co-stained with antibodies to p67. ConA normally accumulates in the lysosome of trypanosomes (Brickman et al., 1995), but RNAi against T. brucei Rab28 prevented delivery of ConA to the lysosome in 80% of cells, resulting in retention in a pre-lysosomal compartment (Fig. 4B,C). These data suggest defective endocytic transport as a consequence of Rab28 suppression and a failure to deliver cargo to the lysosome.

Because ConA is not a physiologically relevant ligand, and the lectin essentially reports bulk glycoprotein flow (Overath and Engstler, 2004), we selected the transferrin receptor (which in trypanosomes is a GPI-anchored heterodimer) to examine receptor-mediated endocytosis (Steverding et al., 1995). Accumulation of Alexa-Fluor-633-labelled transferrin, measured using fluorescence activated cell sorting (FACS), revealed a dramatic 67% decrease in transferrin accumulation in T. brucei Rab28^RNAi cells (Fig. 4D), suggesting a requirement for Rab28 in receptor-mediated uptake and indicating a general impact of Rab28 depletion on endocytosis.

Unlike the extensive recycling of transferrin observed in mammalian cells, transferrin is extensively degraded within the trypanosomal endosomal system by T. brucei catB, a cathepsin B-related protease (Mackey et al., 2004; Grab et al., 1992). The resulting peptide digestion products are exported via a Rab11-dependent recycling pathway (Steverding et al., 1995; Pal et al., 2003). To investigate whether there is a role for T. brucei Rab28 in this process, loss of accumulated Alexa-Fluor-633–transferrin was followed in knockdown cells. The fluorophore survives endosomal (catB) exposure, making this a suitable method for analyzing the recycling pathway (Pal et al., 2003). Fluorescence decay kinetics were essentially identical in control and Rab28 knockdown cells (Fig. 4E), suggesting that T. brucei Rab28 does not play a major role in re-export of transferrin-derived peptides.

T. brucei are sensitive to trypanosome lytic factor (TLF), an innate immune factor in human serum (Hager et al., 1994). TLF is a component of the HDL fraction and manifests trypanolytic activity at the lysosome (Perez-Morga et al., 2005; Raper et al., 2001; Pays et al., 2006). TLF undergoes receptor-mediated endocytosis, traversing the flagellar pocket, early endosomes and late endosome prior to reaching the lysosome (Vanhollebeke et al., 2008). We examined TLF sensitivity after T. brucei Rab28 RNAi to analyze terminal endocytic events and lysosomal delivery.

Cells were incubated in varying concentrations of normal human serum. Silencing T. brucei Rab28 significantly decreased the sensitivity of trypanosomes to TLF. For example, after 4.5 hours in 10% human serum, cell survival in Rab28^RNAi cells was increased by 41% (P<0.005) compared with control cells (Fig. 4F). The effect was specific to T. brucei Rab28^RNAi cells, as silencing of a second late endosomal protein, Vps23 (Leung et al., 2008) failed to elicit this effect (Fig. 4G). These data suggest that depletion of Rab28 significantly impairs trafficking of TLF to the lysosome, consistent with the defect in ConA lysosomal delivery.

The data above indicate a role in late endocytosis for T. brucei Rab28, consistent with partial colocalization with Vps23, a component of ESCRT I (Babst et al., 2000). A possible role for T. brucei Rab28 in turnover of internalized surface proteins was examined. Invariant surface glycoproteins 65 and 75 (ISG65 and ISG75) are abundant type I cell surface trans-membrane proteins possessing multiple cytoplasmic lysine residues and undergo ubiquitin-dependent degradation (Chung et al., 2004; Leung et al., 2011).

ISG75 turnover was examined in T. brucei Rab28^RNAi cells following cycloheximide treatment. After 4 hours, ISG75 levels are reduced by 50% in non-induced cells, but in Rab28-depleted cells, ISG75 levels are reduced less than 10% (Fig. 5A). This suggests that T. brucei Rab28^RNAi cells cannot efficiently turnover ISG75 and that Rab28 has a possible role in sorting at the late endosome, where ubiquitylated proteins are presumably targeted to the lysosome. To confirm this, the levels of intracellular ISG75 were monitored by immunofluorescence. A highly significant 50% increase (P<0.0001) in intracellular ISG75 levels was observed in T. brucei Rab28^RNAi cells relative to non-induced cells, correlating with decreased ISG75 turnover (Fig. 5B,E). Importantly, at steady state, total protein levels of ISG75 were unchanged in Rab28-depleted cells, ruling out a general increase to ISG75 levels (Fig. 5C). Similarly, increased intracellular levels of ISG65 were observed in induced cells compared with non-induced cells, with an even more pronounced effect than ISG75 (Fig. 5D,E). We were unable to detect any significant change in ISG65 or ISG75 levels on the surface of the cells by FACS analysis, but as we did not have an accurate estimate of relative internal and external pool sizes, this might reflect changes in a comparatively small intracellular pool (data not shown). Previous analysis demonstrated that, in common with higher eukaryotes, turnover of ubiquitylated proteins is dependent on the ESCRT pathway (Chung et al., 2008), so we propose that the increase in intracellular ISG65 and ISG75 following Rab28 depletion is due to failure to progress to the lysosome, similar to the effects of Rab28 RNAi on both ConA and TLF.

To further delineate pathways requiring Rab28 expression, markers of intracellular compartments were monitored using both western blotting and IFA, to examine both expression levels and the location of the marker proteins. No obvious changes in distribution or total protein levels of clathrin heavy chain (CHC) or Rab5A were observed in T. brucei Rab28^RNAi cells (Fig. 6A,C), which suggests that Rab28 is not directly involved in formation or distribution of clathrin-coated vesicles or early endosomes, and is consistent with the localization of Rab28.

However, structural alterations to recycling compartments marked by Rab11 were observed in T. brucei Rab28^RNAi cells. A gradual expansion of Rab11 staining was seen over 4 days, but this was not accompanied by an increase to Rab11 protein levels (Fig. 6A,C). Expansion of Rab11 localization required 4 days to become apparent but Rab28 was silenced within 24 hours, which suggests that progressive disruption to recycling endosomes is probably a secondary effect, possibly as a consequence of the pre-lysosomal blockade to traffic and/or re-routing of endocytic cargo through the recycling pathway. Additionally, disordered compartments might result from defective targeting of components essential for maintaining structural integrity of recycling endosomes.

Surprisingly, Rab28 knockdown had dramatic effects on Golgi complex morphology (Fig. 6A). Silencing Rab28 resulted in fragmentation of Golgi membranes as monitored by GRASP immunoreactivity. Control cells in G1 possess a maximum of three GRASP spots, whereas 15% of T. brucei Rab28^RNAi cells displayed more than five GRASP spots after 24 hours. After 48 hours of RNAi, complete dispersal of the Golgi complex was observed in 30% of T. brucei Rab28^RNAi cells (Fig. 6B). This observation was verified with a second Golgi complex marker, Rab1 (Dhir et al., 2004). In T. brucei Rab28^RNAi cells, Rab1-positive membranes also expanded, providing further evidence that Rab28 expression is required to retain Golgi architecture, which suggests that T. brucei Rab28 participates in retrograde trafficking from late endosomes to the Golgi complex. Importantly, perturbation of Golgi architecture occurred with kinetics more similar to Rab28 knockdown, and is therefore a more direct consequence of Rab28 loss than alterations in recycling endosomes. Clearly, disruption of the Golgi complex will impact multiple trafficking pathways and possibly accounts for the expansion of Rab11 endosomes.

Most intracellular VSG is associated with the Golgi complex and the endoplasmic reticulum (ER) (Webster and Grab, 1988). In permeabilized cells, residual surface staining of VSG was seen, together with ER and Golgi intracellular pools. However, in Rab28-depleted cells, intracellular VSG appeared diffuse compared with control cells (Fig. 6A), which is also consistent with Golgi disruption. These data suggest that a focus of intracellular VSG (i.e. the Golgi complex) became dispersed over time.

Finally, a mild perturbation in lysosome morphology was observed in T. brucei Rab28^RNAi cells, with expansion of p67 staining (Fig. 6A). This effect was moderate because most (~70%) T. brucei Rab28^RNAi cells contained morphologically normal lysosomes, and indicates that, although p67 is faithfully targeted in cells depleted of Rab28, targeting of other, so far unidentified lysosomal structural components might be affected.

To further examine the effect of T. brucei Rab28 RNAi on subcellular structures we analyzed cells by transmission electron microscopy (TEM). After 36 hours of RNAi against Rab28, the majority of induced cells lacked discernible stacked Golgi complexes (Fig. 7C,D). Where present, membranes in the region of the cell normally containing the Golgi complex exhibited anomalous cisternal structures (Fig. 7B). Significantly, a detectable Golgi was completely absent at 96 hours post-induction. In 15 sections from unperturbed cells we found five stacks (i.e. in 33% of sections) but we were unable to observe a single stacked Golgi complex in over 50 sections from the T. brucei Rab28^RNAi cells (not shown and Fig. 7E). Furthermore, we infrequently detected structures with a high proportion of intralumenal vesicles (Fig. 7E), as previously been reported when AP-1 was silenced in trypanosomes, which might correspond to multivesicular bodies or other stress-induced structures (Allen et al., 2007). We also observed the appearance of highly vesiculated structures in close proximity to the flagellar pocket; these structures might represent residual material derived from the Golgi complex, post-Golgi transport intermediates and/or endosomal structures (Fig. 7E). By contrast, the subpellicular array of microtubules, flagellum attachment zone and flagellum all retained normal architecture, which signifies a specific defect to endomembrane trafficking networks. These data are fully consistent with the effects seen at the Golgi complex and recycling endosomes observed using light microscopy.

In mammalian cells, retromer mediates trafficking from endosomes to the Golgi complex (Seaman, 2004; Arighi et al., 2004; Belenkaya et al., 2008). T. brucei possesses orthologues of the cargo-recognition retromer subcomplex (Vps26, Vps29 and Vps35) and one sorting nexin (Vps5), suggesting conserved function (Koumandou et al., 2011).

Vps26-associated fluorescence was reduced by 75% in T. brucei Rab28^RNAi cells and was accompanied by decreased protein levels, indicating that Rab28 expression influences Vps26 protein stability and/or synthesis (Fig. 8A–C). To determine whether Rab28 acts upstream or downstream of Vps26, a T. brucei Vps26^RNAi cell line expressing TbRab28HA was generated. Rab28 levels were monitored by western blot after induction of RNAi against Vps26. Rab28 levels remained unchanged in T. brucei Vps26^RNAi cells, indicating that Rab28 functions upstream of Vps26 (Fig. 8H).

To further investigate involvement of Rab28 in late endosomal trafficking, the effects of Rab28 knockdown on ESCRT I were analyzed. Vps23 was undetectable by IFA in T. brucei Rab28^RNAi cells (Fig. 8D), a result reflected in total protein levels in lysates, which were reduced by ~70% (Fig. 8E). To assess whether Rab28 knockdown resulted in a general impact on ESCRT I and the expression of Vps28, a second epitope-tagged ESCRT I subunit was monitored (Leung et al., 2008). A similar decrease was observed, representing an ~80% reduction in total protein levels (Fig. 8D–F). These data demonstrate that the stability of the ESCRT I complex as a whole was compromised by silencing Rab28. T. brucei Vps23 was silenced in cells containing TbRab28HA. Rab28 levels were unaltered in Vps23-depleted cells, suggesting that T. brucei Rab28 functions upstream of Vps23 (Fig. 8G). To characterize interactions between Rab28 and Vps23 and Vps26, T. brucei Vps23 was silenced and the HA-tagged Vps26 (TbVps26HA) levels monitored. Although some increase in Vps26 was found (Fig. 8G), Vps26 knockdown failed to elicit in trans effects on TbVps23HA (Fig. 8H). Hence, there is no evidence for in trans suppression between T. brucei Vps26 and Vps23 expression, suggesting that these effects are mediated via loss of Rab28.

Rab proteins regulate recognition, fusion and fission events in membrane transport. In endocytosis, maturation of early into late endosomes is marked by Rab5 to Rab7 conversion (Rink et al., 2005) and a switch in the repertoire of Rab effectors and other proteins on the endosomal membrane in a coordinated process. Overall, however, the endocytic system is very complex and involves additional Rab proteins, including Rab4, Rab11 and Rab21. We have analyzed the function of trypanosome Rab28, which appears to also mediate endocytic traffic, including endocytosis of transferrin, turnover of trans-membrane domain endocytic cargo glycoproteins and lysosomal delivery. Furthermore, although T. brucei Rab28 was required for proliferation, knockdown did not perturb cell cycle progression nor result in a global blockade of membrane uptake from the trypanosome endocytic organelle, the flagellar pocket. Together with partial colocalization with Vps23, we suggest that T. brucei Rab28 predominantly localizes to late endocytic compartments.

Knockdown of T. brucei Rab28 resulted in a pre-lysosomal transport block, implicating Rab28 in sorting material prior to lysosomal delivery. An in trans decrease in expression levels of ESCRT I (Vps23 and Vps28) and retromer complex (Vps26) proteins on Rab28 knockdown suggests a functional connection between Rab28, ESCRT and retromer. The effects on ESCRT are fully consistent with the observed partial protection of ISG65 and ISG75, as both require ubiquitylation for turnover and, hence, are probably sorted by the ESCRT system (Chung et al., 2008; Leung et al., 2011). The effects on the Rab11 recycling compartment suggest increased cycling of cargo between endosomes, as reported upon silencing of Vps26 and Vps23 in mammalian cells (Seaman, 2004; Razi and Futter, 2006; Raiborg et al., 2008). The partial protection of ISG75 is probably due to incomplete silencing of T. brucei Rab28, as found earlier for T. brucei Vps23 (Leung et al., 2008), although we cannot exclude Rab28-independent pathways. Significantly, T. brucei Rab28 probably functions upstream of Vps23 and Vps26, as knockdowns of the latter did not affect Rab28 expression. The clear complexity of trafficking within this region of the trypanosome endomembrane system suggests that analysis of the roles that Rab5 and Rab11 play in ISG transport will be highly informative, especially as our understanding of these pathways has advanced somewhat since the original descriptions of these pathways (Field and Carrington 2009).

Disruption of Golgi complex morphology by Rab28 knockdown was unexpected because Rab28 does not localize at Golgi membranes, but a similar kinetic profile for disruption of Rab1, Rab11 and GRASP location argues for a fairly comprehensive impact affecting both the cis and trans faces as well as the stack itself. Disruption of sphingolipid biosynthesis through inhibition of serine palmitoyltransferase has a similar effect on the Golgi complex, but the effect is relatively mild (Fridberg et al., 2008). However, disruption of Golgi morphology upon depletion of T. brucei Rab28 parallels Vps26 knockdown in mammalian cells and trypanosomes (Seaman, 2004; Koumandou et al., 2011). The absence of conventional retromer cargo from trypanosomes means that direct investigation of retromer function is difficult, but the complex probably mediates similar functions in both systems (Koumandou et al., 2011). Gross morphological aberrations at the Golgi complex potentially result through depletion of SNAREs or other factors required for correct homotypic fusion due to defective retrieval from late endosomes, i.e. a ‘traffic jam’.

Although mechanisms coordinating retrograde transport with endosomal maturation are well understood, precisely how the functions of the sorting complexes involved in retrograde and anterograde transport are integrated remains elusive. Specifically, both retromer and ESCRT participate in transport at late endosomal compartments, but their very distinct functions necessitate coordination. T. brucei Rab28 could be required for maintaining structural boundaries at endosomes or for correct assembly of ESCRT and/or retromer complexes, and potentially in controlling turnover of ubiquitylated cargo, which is probably a major route for degradation of surface proteins in trypanosomes (Leung et al., 2011). Depletion of T. brucei Rab28 resulted in extensive morphological defects, suggesting that Rab28 is required to maintain endosome structure. Direct interactions between T. brucei Vps23 and Rab28 were not detected by co-immunoprecipitation (data not shown), but it is possible that interaction is mediated via Rab28 effector proteins for example and that Rab28 is itself controlled by the relevant GAP and GEF proteins. Furthermore, ESCRT 0 recruits ESCRT I in mammalian cells (Bache et al., 2003; Katzmann et al., 2003), but ESCRT 0 is absent from most organisms and the primary mechanism of ESCRT recruitment in trypanosomes is unclear (Leung et al., 2008). Significantly, ESCRT 0 depletion in mammalian cells failed to abolish epidermal growth factor receptor degradation and only inhibits Vps23 membrane association by ~50%, which suggests the presence of alternative recruitment mechanisms (Razi and Futter, 2006; Bache et al., 2003; Katzmann et al., 2003).

Clearly, species retaining Rab28 must have distinct mechanisms for coordinating ESCRT and retromer function compared with Rab28-deficient species. Importantly, trypanosome Rab28 is 80% identical to human Rab28 in the switch and interswitch regions, and these regions are key in determining effector interactions (Delprato and Lambright, 2007; Eathiraj et al., 2005; Merithew et al., 2001). Together with observations that orthologues generally retain similar functions across evolution (Brighouse et al., 2010), this suggests that T. brucei Rab28 is probably functionally homologous to H. sapiens RAB28 and provides a mechanism allowing sorting and subsequent degradation of ubiquitylated cargo in the absence of the ESCRT 0 subunit, Hrs (Razi and Futter, 2006; Bache et al., 2003; Hammond et al., 2003). Moreover, increased autophagosome frequency is observed when ESCRT complexes are inactivated, suggesting that ESCRT is involved in autophagosome–lysosome fusion events (Doyotte et al., 2005; Filimonenko et al., 2007; Lee et al., 2007). Silencing H. sapiens RAB28 dramatically inhibits autophagosome formation (Nicole McKnight and Sharon Tooze, Cancer Research UK, London, personal communication), also implicating H. sapiens RAB28 in generation of autophagosomes, possibly through regulation of ESCRT function. ESCRT, retromer and autophagocytosis are essential for cellular homeostasis, and malfunctions in these systems contribute to multiple pathologies (Hu et al., 2006; Tanida et al., 2005; Kirschbaum and Yarden, 2000; Vaccari and Bilder, 2005; Moberg et al., 2005; Thompson et al., 2005). There is a highly significant correlation between familial breast cancer and a single-nucleotide polymorphism directly upstream of H. sapiens RAB28 (Yang et al., 2008), which is consistent with a role for Rab28 in sorting of ubiquitylated proteins and, specifically, mitogenic receptors.

Bloodstream form (BSF) and procyclic form (PCF) Lister 427 strains were maintained in HMI-9 and SDM79 medium, respectively, as described (Field and Field, 1997). Single marker BSF (SMB) cells (Wirtz et al., 1999) were maintained with 2.5 μg/ml neomycin. Cells in exponential growth, <1×10⁶/ml and 3–8×10⁶ cells/ml for BSF and PCF cells, respectively, were used for all experiments.

A 581 bp fragment from position 109–690 of the T. brucei Rab28 ORF (Tb927.6.3040) was amplified using primers 5′-GCGACAGTTCAGACTCAGAAAA-3′ and 5′-CACTGCGCATTTACCCTTCT-3′ (Redmond et al., 2003) and cloned into p2T7^TABlue. An AMAXA Nucleofector II was used to transfect SMB cells with NotI-digested plasmid. Clones were selected and maintained in the presence of 5 μg/ml hygromycin and 2.5 μg/ml geneticin. For ectopic expression T. brucei Rab28 was amplified from genomic DNA using primers 5′-CCAGAAGCTTCTAGTAGCGACAGTTCAGACTCAG-3′ and 5′-GCATGGATCCCTACATCACTGCGCATTTAC-3′ and cloned into pHD1034 (Quijada et al., 2002) or pXS519 (Hall et al., 2004) containing N-terminal HA or YFP for expression in BSF and PCF cells, respectively. BSF cells were transfected with 5 μg linearized pHD1034-T. brucei Rab28 and clones maintained in 2.5 μg/ml geneticin and 0.2 μg/ml puromycin. PCFs were transformed as described (Hall et al., 2004). Plasmids allowing ectopic expression of T. brucei Vps23HA, Vps28FLAG and Vps26HA were as described (Leung et al., 2008; Koumandou et al., 2011).

To express T. brucei Rab28 at the endogenous locus, the 5′-end ORF was generated by PCR using the following primers: Rab28 GTag_F2 (5′-CTCAAGCTTTGATGAGTAGCGACAGTTCAGACTC-3′) and Rab28 GTag_ R3 (5′-CGCGATATCGTCGCTGCCAACTGCGTGAAG-3′). The PCR product was first cloned into pGEM-T Easy plasmid vector (Promega) and subsequently subcloned with HindIII and EcoRV into plasmid vector p3077 (a gift from Mark Carrington, University of Cambridge, UK) (Kelly et al., 2007) that generates an N-terminal 4× TY1 epitope-tagged fusion protein. Plasmid was linearized using the unique SphI site within the 5′-end ORF prior to transfection into cells. For immunofluorescence assay, detection was achieved using BB2 monoclonal antibody (a gift from Keith Gull, University of Oxford, UK) at 1:10 dilution and at 1:50 dilution for western blot analysis.

RNAi was induced with 1 μg/ml TET. Silencing of T. brucei Rab28 was monitored by northern and western blot. BSF proliferation was monitored in triplicate by inoculating 1×10⁴ cells/ml daily in fresh medium. Cell densities were measured with a Coulter Z1 Counter (Beckman). All assays were performed at 36 hours unless otherwise stated.

The T. brucei Rab28 ORF, without the prenylation motif, was amplified from T. brucei genomic DNA using primers 5′-CCTAGGATCCCCCGACAGTTCAGACTCAGAAAAAAGG-3′ and 5′-CCTAGAATTCTTCCAAATTTAGCGCATCTTCAGG-3′ and ligated in-frame with GST into pGEX6p2 and transformed into BL21(DE3) Escherichia coli. GST::T. brucei Rab28 was expression-induced with 0.5 mM isopropyl β-D-thiogalactoside (IPTG). Recombinant protein was affinity-purified on glutathione Sepharose-4B (GE Healthcare). Purity of isolated GST::T. brucei Rab28 was estimated at ⩾95% by SDS-PAGE and Coomassie Blue staining. Antiserum was raised in rabbits against the full-length fusion protein, immunizing four times with a total of 2 mg protein in Freund's complete adjuvant (Covalab). GST::T. brucei Rab28 was coupled to CNBr–Sepharose 4B (Sigma) for affinity purification.

10⁸ BSF and 5×10⁷ PCF cells were harvested at 4°C and RNA was extracted using RNeasy (Qiagen). Complementary DNA was generated from 2 μg RNA using SuperscriptTM II RNase H reverse transcriptase (Invitrogen). qRT-PCR was performed using iQ-SYBRGreen Supermix on a Mini Opticon (BioRad). β-tubulin, expressed at similar levels in BSF and PCF cells (Diehl et al., 2002) was used to normalize. qRT primers were Rab28 forward, 5′-GCGACAGTTCAGACTCAGAAAA-3′; Rab28 reverse, 5′-CACTGCGCATTTACCCTTCT-3′; β-tubulin forward, 5′-CAAGATGGCTGTCACCTTCA-3′; and β-tubulin reverse, 5′-GCCAGTGTACCAGTGCAAGA-3′.

Northern blotting was carried out using standard procedures. RNA was extracted from 10⁸ cells, and 3.5 μg RNA separated by gel electrophoresis prior to transfer to N-Hybond membrane. T. brucei Rab28 and ribosomal RNA (rRNA)-specific probes were labelled with [α-³²P]dCTP (3000 Ci/mmol) and quantified by phosphorimaging.

Cells were harvested, washed twice in ice-cold PBS and heated for 5 minutes at 94°C in 2× SDS-PAGE loading buffer. 10⁷ cells per lane were resolved on 12.5% SDS–polyacrylamide gels. Proteins were electrophoretically transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore), blocked and processed following standard procedures. Polyclonal rabbit anti-T. brucei Rab28 was used at 1:500, polyclonal rabbit anti-T brucei BiP serum (kind gift of James D. Bangs, University of Winsconsin, WI) 1:10,000, polyclonal rabbit anti-ISG65 serum 1:5000, polyclonal rabbit anti-ISG75 serum 1:5000 (both from Mark Carrington, University of Cambridge, UK), polyclonal rabbit anti-T. brucei Rab5A serum (Field et al., 1998) 1:1000, polyclonal rabbit anti-T. brucei Rab11 serum (Jeffries et al., 2001) 1:1000, polyclonal rabbit anti-T. brucei CHC serum (Morgan et al. 2001) 1:1000, polyclonal rabbit anti-GFP serum (a gift of Michael Rout, Rockefeller University, New York, NY) 1:10,000, monoclonal anti-β-tubulin (Chemicon) 1:20,000 and monoclonal anti-HA (Roche) 1:5000. Incubations with anti-IgG rabbit horseradish peroxidase conjugates (Sigma) were performed at 1:20,000 in Tris-buffered saline containing Tween-20 and non-fat powdered milk. Detection was by chemiluminescence with luminol (Sigma) on BioMaxMR film (Kodak). For densitometry, fluorographs were scanned at 16-bit gray scale, and exposures selected where the signal was unsaturated. Exposures in figures usually represent overexposed versions of data used for quantification. Quantification was done using NIH ImageJ software (http://rsbweb.nih.gov/ij/).

Cycloheximide (Sigma) was added to log-phase cultures at a final concentration of 100 μg/ml.

Log-phase cells were fixed in 3% paraformaldehyde in PBS on ice and adhered to poly-L-lysine slides (Sigma). Cells were permeabilized with 0.1% Triton-X-100, washed and blocked in 20% foetal calf serum (FCS), incubated with primary antibodies for 1 hour and washed, secondary antibodies applied at 1:1000 for 1 hour, and mounted with Vectashield containing DAPI (Vectalabs). Image acquisition was performed with a Nikon Eclipse E600 epifluorescence microscope fitted with a Hamamatsu CDD digital camera and Metamorph (Molecular Devices), and processed with Photoshop (Adobe). Confocal microscopy images were acquired with a Leica TCS-NT confocal microscope with a 100× 1.4 NA objective. Images were processed with Huygens deconvolution software (Scientific Volume Imaging) and Adobe Photoshop. All images were taken under non-saturating conditions. All quantification was done using identical exposures using raw data within Metamorph. Antibodies were used at the following dilutions: T. brucei Rab28, 1:500; T. brucei Rab5A, 1:200; T. brucei Rab11, 1:200; T. brucei CHC, 1:750; T. brucei Rab1, 1:100; p67, 1:500; VSG-221, 1:1000; GRASP, 1:200; ISG100, 1:5; ISG75 1:5000; ISG65 1:1000; and antibodies to HA and GFP at 1:2000 and 1:2500, respectively.

1×10⁸T. brucei Rab28^RNAi cells were washed with 0.1 M HEPES pH 7.0, 0.98% sodium chloride and fixed for 4 hours on ice in 2% glutaraldehyde, 2 mM calcium chloride, 0.1 M PIPES buffer pH 7.4. Then, 100 μl 33% H₂O₂ was added to each 10 ml of fixative immediately prior to use. After three washes in 0.1 M HEPES pH 7.0, the pellet was processed for electron microscopy. Sections were viewed using a Philips CM100 electron microscope (FEI-Philips) operated at 80 kV.

Cells were harvested, washed in serum-free media and resuspended at 5×10⁵ cells/ml in serum-free media containing 1% BSA. Parasites were equilibrated at 37°C for 15 minutes before addition of FITC-conjugated Concanavalin A (ConA) or transferrin–Alexa Fluor 633 conjugate (Molecular Probes) to final concentrations of 5 μg/ml or 25 μg/ml, respectively. Aliquots of 2×10⁶ cells were removed, and the uptake of fluorophore quenched by addition of ice-cold PBS. Samples were washed at 4°C to remove excess probe. ConA samples were prepared for immunofluorescence and transferrin samples were fixed in 1% formaldehyde on ice for 10 minutes before flow cytometry on a Cyan LX-FACS (DakoCytomation). Cell-associated fluorescence from 50,000 cells was measured for Alexa Fluor 633 at 665 and 720 nm and analysed with Summit (Cytomation).

Cells were washed to remove excess free serum components and resuspended at 5×10⁵ cells/ml in serum-free media containing 1% BSA. Transferrin–Alexa Fluor 633 was added at 25 μg/ml and cells incubated at 37°C for 1 hour to allow intracellular accumulation. Cells were washed at 4°C to remove excess fluorophore and returned to 37°C. Cells were fixed and taken for FACS analysis.

Blood was taken from fasted healthy volunteers and allowed to clot for 1 hour at 37°C before cooling at 4°C. Serum was separated by centrifuging at 10,000 g for 20 minutes at 4°C and heat-inactivated by incubating at 55°C for 30 minutes. Sera were stored at −80°C. Cells were harvested, washed in serum-free media and resuspended in media supplemented with 0–20% human serum. Total serum concentration was maintained at 20% with FCS. Cells were incubated at 37°C and cell density determined using a haemocytometer.

We thank James Bangs, Mark Carrington and Keith Gull for various antibodies, V. Lila Koumandou for tagged cell lines and members of the Field laboratory for discussions, donation of serum and support.