Both sequence and context are important for flagellar targeting of a glucose transporter

Summary

Many of the cilia- and flagella-specific integral membrane proteins identified to date function to sense the extracellular milieu, and there is considerable interest in defining pathways for targeting such proteins to these sensory organelles. The flagellar glucose transporter of Leishmania mexicana, LmxGT1, is targeted selectively to the flagellar membrane, whereas two other isoforms, LmxGT2 and LmxGT3, are targeted to the pellicular plasma membrane of the cell body. To define the flagellar targeting signal, deletions and point mutations were generated in the N-terminal hydrophilic domain of LmxGT1, which mediates flagellar localization. Three amino acids, N95-P96-M97, serve critical roles in flagellar targeting, resulting in strong mistargeting phenotypes when mutagenized. However, to facilitate flagellar targeting of other non-flagellar membrane proteins, it was necessary to attach a larger region surrounding the NPM motif containing amino acids 81–113. Molecular modeling suggests that this region might present the critical NPM residues at the surface of the N-terminal domain. It is likely that the NPM motif is recognized by currently unknown protein-binding partners that mediate flagellar targeting of membrane-associated proteins.

Introduction

Cilia and flagella are evolutionarily conserved organelles involved in motility in a broad range of eukaryotes from single cell protozoa such as Tetrahymena and Trypanosoma to humans (Ginger et al., 2008; Roy, 2009). In addition, these organelles have been recognized to play central roles in sensing the environment and in transducing extracellular or environmental signals to the cell interior (Berbari et al., 2009; Bloodgood, 2010; Marshall and Nonaka, 2006; Pazour and Witman, 2003; Qin et al., 2005; Singla and Reiter, 2006). As a consequence of their important role in signal transduction, much effort is being devoted to determining how specific membrane proteins are selectively sorted to the cilium (Nachury et al., 2010; Pazour and Bloodgood, 2008).

Cis-acting ciliary targeting sequences have been identified in mammalian ciliary membrane proteins such as the RPxV motif in polycistin-2 (Geng et al., 2006), the Ax(S/A)xQ motif in the somatostatin receptor SSTR3 (Berbari et al., 2008a), and others discussed extensively in a recent review (Nachury et al., 2010), and these motifs are likely to interact with specific proteins that mediate ciliary targeting. A few examples of proteins involved in ciliary trafficking include: (1) the intraflagellar transport (IFT) particles that play a role in movement of TRPV channels into ciliary axonemes in sensory neurons of Caenorhabditis elegans (Qin et al., 2005), (2) Bardet–Biedl Syndrome proteins, a complex of polypeptides involved in ciliary formation, that function to target the SSTR3 somatostatin receptor to neuronal cilia in mice (Berbari et al., 2008b), and (3) a complex of polarity proteins, Crumb3–Par3–Par6–aPKC, that participates in the assembly of mammalian ciliary membranes (Fan et al., 2007).

Kinetoplastid parasites such as Trypanosoma brucei, Trypanosoma cruzi and Leishmania are flagellated protozoa that cause devastating diseases among humans and animals. For each of these parasites, there exist examples of membrane proteins that are selectively targeted to the flagellum, although the mechanisms underlying targeting remain largely uncharacterized. In T. brucei, the probable receptor adenylate cyclase ESAG4 is localized in the flagellum of the bloodstream form parasites (Paindavoine et al., 1992). In L. major, the polytopic aquaporin AQP1 is localized to the flagellar membrane (Figarella et al., 2007), as is the dually acylated SMP-1 protein (Tull et al., 2004). In addition, a recent proteomic study identified a variety of integral membrane proteins present in the flagellar membrane of T. brucei (Oberholzer et al., 2011). One example of a flagellar membrane protein where the targeting sequence has been identified is the dually acylated flagellar calcium binding protein (FCaBP) from T. cruzi. FCaBP is tethered to the cytosolic face of the membrane via dual lipid modifications, and trafficking to the flagellar membrane requires a lysine-rich sequence between amino acids 13–24 (Godsel and Engman, 1999; Maric et al., 2011). Previous work from our laboratory has also shown that the L. mexicana and L. enriettii glucose transporters, LmxGT1 (Burchmore et al., 2003) and LeGT1 (Piper et al., 1995; Snapp and Landfear, 1997), respectively, traffic selectively to the flagellar membrane. In contrast, LmxGT2 and LmxGT3, two closely related isoforms, are excluded from the flagellum and target to the pellicular plasma membrane surrounding the cell body and to the flagellar pocket. The major amino acid sequence differences between the Leishmania GT isoforms are in their cytoplasmic N-terminal domains (Burchmore and Landfear, 1998; Burchmore et al., 2003), and two separate segments within this domain of LeGT1 have been implicated previously in flagellar targeting (Nasser and Landfear, 2004). However, the specific amino acids required for flagellar targeting of LeGT1 have not been identified. The studies reported here focused upon defining a precise flagellar targeting motif using LmxGT1 as the model protein.

Results and Discussion

Identification of an NPM motif that is essential for flagellar targeting of LmxGT1

The LmxGT1 protein extends 43 amino acids upstream of the methionine residue originally annotated as the start codon (Burchmore and Landfear, 1998) and thus encompasses a 130 amino acid N-terminal domain. The new sequence is shown in supplementary material Fig. S1. Systematic deletion performed on this 130 amino acid cytosolic N-terminal domain (Fig. 1A; supplementary material Fig. S1) of LmxGT1::GFP (the LmxGT1 protein with GFP fused to the C-terminus) revealed that a segment between amino acids 90–100 is important for flagellar targeting (Fig. 1B). The Δ(84–100) deletion mutant was targeted to the plasma membrane surrounding the cell body and the flagellar pocket, but trafficking to the flagellar membrane was almost completely eliminated (Fig. 1B,C). All other deletion mutants, including Δ(53–89), exhibited strong flagellar localization.

Fig. 1.

Trafficking of LmxGT1 deletion mutants to the flagella. (A) Schematic diagram of the cytoplasmic 130 amino acid N-terminal domain of LmxGT1. Deletions are shown as dashed lines. (B) Leishmania mexicana promastigotes stained with anti-GFP antibody (green), anti-α-tubulin antibody (red) to stain the subpellicular microtubule network, and DAPI (blue). Yellow arrowheads show examples of overlapping GFP expression and tubulin. White arrowheads indicate absence of GFP signal. Scale bar: 5 µm. (C) Quantification of deletion mutants as percentage of each phenotype. Examples of each of the four targeting phenotypes are shown in the bottom panels.

Alanine scan mutagenesis of amino acids 90–100 of LmxGT1 was performed to determine which residues are critical for flagellar targeting (Fig. 2). The N95A and P96A mutants retained only ∼25–30% targeting to the flagellum, and about half of the cells containing flagellar fluorescence also showed fluorescence over the cell body membrane (Fig. 2A,C). In contrast, >95% of the wild type LmxGT1::GFP or unaffected point mutants (i.e. P94A or K98A, Fig. 2C) targeted discretely to the flagellum. The M97A mutant exhibited ∼50% flagellar targeting and ∼50% trafficking to the flagellar pocket (Fig. 2B,C). Furthermore, the triple point mutation N95A+P96A+M97A (AAA) exhibited a strong mistargeting phenotype, with approximately 85% of the cells showing fluorescence primarily in the flagellar pocket (Fig. 2B,C). All other individual alanine mutants over this region (only P94A and K98A are shown in Fig. 2) trafficked efficiently to the flagellar membrane with a phenotype undistinguishable from that of wild type LmxGT1. Overall these results identify the N95-P96-M97 sequence as critical for flagellar targeting, hereafter referred to as the NPM motif.

Fig. 2.

Trafficking of point mutants to the flagella. (A,B) Fluorescence images of point mutants. The specific point mutations are indicated above each image. Yellow arrowheads show examples of overlapping GFP expression and tubulin. White arrowheads indicate absence of GFP signal. Scale bars: 5 µm. (A) Large fields of view showing examples of mistargeting phenotypes in N95A and P96A mutants. (B) Targeting of M97A, AAA (N95A+P96A+M97A triple mutant) and K98A mutants. (C) Quantification of results in A and B. Color scheme and expression categories are as in Fig. 1C. (D) Alignment of amino acids 90–100 of GT1 in four species of Leishmania (Lmx, L. mexicana; Lmj, L. major; Lif, L. infantum; Le, L. enriettii). Gray boxes indicate complete conservation of sequences and white boxes show conservation in the three human pathogenic species.

It is notable that 50–60% of the fluorescence in the N95A, P96A, M97A mutants, and 85% in the triple mutant, is located in the flagellar pocket, the site where secretory vesicles fuse with the surface membrane (McConville et al., 2002). This result suggests that flagellar trafficking of LmxGT1 may entail sorting from the flagellar pocket into the contiguous flagellar membrane and that such sorting may depend on a protein or proteins that interact with the NPM motif. In principle, such sorting could rely upon either active forward migration out of the flagellar pocket or retention in the flagellum. Additionally, accumulation of the individual point mutants and triple mutant at the flagellar pocket could indicate that they interact in a non-productive fashion with some factor in the flagellar pocket and are thus retained in that organelle but are unable to migrate into the flagellar membrane.

Computer modeling of the NPM motif and the LmxGT1 N-terminal domain

To gain further insights into the potential role of the NPM motif, an ab initio computational model of the LmxGT1 N-terminal domain was generated using Rosetta software (Yarov-Yarovoy et al., 2006). The model predicts that the three essential NPM residues form a ‘hook’ structure at the tip of a loop separating two α-helixes (Fig. 3A,B), referred to hereafter as the predicted helix-loop-helix or HLH domain. While we do not have experimental evidence confirming the existence of this secondary structure, the model suggests that regions flanking the NPM motif would cause the critical residues to project from the surface of the protein and thus be important for flagellar targeting through interactions with potential binding partners. The NPM motif and flanking residues were tested below for their ability to target other proteins to the flagellum.

Fig. 3.

Flagellar targeting by the NPM motif and flanking sequences. (A) The structure of the N-terminal domain was predicted using Rosetta software (Yarov-Yarovoy et al., 2006). The model shows the NPM motif (red) forming a hook at the tip of the loop connecting two predicted α-helixes (green). (B) The segment of the model that corresponds to amino acids 81–113. The NPM motif (red) is represented here using stick figures. (C) Targeting of LmxGT2 chimera proteins. Yellow arrowheads show examples of overlapping GFP expression and tubulin. White arrowheads indicate absence of GFP signal. The numbers below each image indicate the percentage of cells that target to the pellicular plasma membrane (PPM), the flagellar membrane only (FM) or both membranes (FM+PMM). ‘n’ indicates the total number of cells quantified for each mutant. Not shown is the flagellar-pocket-only localization of GT1(1–130)::GT2::GFP, which is approximately 4%. Scale bar: 5 µm.

The NPM motif and flanking sequences can facilitate targeting of LmxGT2 to the flagellar membrane

To determine whether the NPM motif is sufficient to target an integral membrane protein to the flagellar membrane, this sequence was fused to the N-terminus of LmxGT2, a related glucose transporter that is normally excluded from the flagellum and traffics to the pellicular plasma membrane of the cell body (Burchmore et al., 2003).

Both wild-type LmxGT2::GFP and NPM::GT2::GFP were localized to the pellicular plasma membrane of Leishmania promastigotes (99%, n = 237; 99%, n = 323, respectively) (Fig. 3C), indicating that the NPM motif by itself is not sufficient to confer flagellar targeting. Additionally, a LmxGT2 chimera protein, PNPMKP::GT2::GFP, was generated encompassing a stretch of amino acids surrounding the NPM motif that is well conserved in GT1 from several Leishmania species (Fig. 2D). PNPMKP::GT2::GFP also localized to the pellicular plasma membrane but not the flagellum (100%, n = 297) (data not shown). In contrast, when the entire 130 amino acid N-terminal region of LmxGT1 was fused to full length LmxGT2, GT1(1–130)::GT2::GFP, this chimeric protein localized to the flagellar membrane in the majority of cells examined (82%, n = 200), with a few cells showing expression in both the flagellar and pellicular membranes (12%) (Fig. 3C). Additionally, when residues 81–113 from LmxGT1 (containing the predicted HLH domain) were fused onto the N-terminus of LmxGT2, GT1(81–113)::GT2::GFP, the chimeric protein localized to both the flagellar and pellicular plasma membrane in the majority of cells examined (88%, n = 204). The same targeting phenotypes were reproduced when the chimeras above were examined in glucose transporter null-mutants, ruling out the possibility that endogenous LmxGT1 protein contributes to the flagellar targeting of the chimeras, i.e., by dimerization of wild-type and chimeric proteins (supplementary material Fig. S2). The residual pellicular plasma membrane expression of LmxGT2 in the fusion protein containing only residues 81–113 from LmxGT1 suggests that this smaller domain is not as efficient as the full length LmxGT1 N-terminal domain in targeting LmxGT2 to the flagellar membrane. However, both the entire 130 amino acid N-terminal domain and residues 81–113 of LmxGT1 are sufficient to facilitate flagellar targeting of the non-flagellar LmxGT2 protein.

The NPM motif and flanking sequences is sufficient to target another integral membrane protein to the flagella

Because the protein sequences of LmxGT1 and LmxGT2 are highly conserved outside of the N-terminal domains (>95%), it is possible that other sequences in the protein may contribute to flagellar targeting. To eliminate this possibility, the L. donovani nucleoside transporter 1.1 (LdNT1.1) (Vasudevan et al., 2001) was also used to examine flagellar targeting. When transfected into L. donovani promastigotes, GFP::LdNT1.1 was targeted to the pellicular plasma membrane (96%, n = 228) (Fig. 4A), consistent with previous reports (Valdés et al., 2006; Vasudevan et al., 2001). However, fusion of residues 1–130 or 81–113 of LmxGT1 onto the N-terminus of GFP::LdNT1.1 robustly retargeted this protein to the flagellar membrane (99%, n = 150; 95%, n = 213, respectively) (Fig. 4A). These experiments confirm that both segments 1–130 and 81–113 of LmxGT1 are sufficient for targeting of an unrelated polytopic membrane protein to the flagella. Furthermore, the results above suggest a conserved mechanism that recognizes the LmxGT1 flagellar targeting signal also exists in L. donovani.

Fig. 4.

Targeting of other chimera proteins to the flagella. (A) Fluorescence images of LdNT1.1 chimera proteins in L. donovani. Numbers at the bottom of each panel indicate percentage of cells showing the GFP expression represented in the panel and the total number of cells quantified (n). (B) Fluorescence images of HASPB(18)::GFP (left) and HASPB(18)::GT1(81–113)::GFP (right) in L. mexicana. (C) Quantification of results in B. n>200 cells for each chimera. (D) Fluorescence images, from left to right, of GFP and of GFP fused at its N-terminus to residues 1–130 or residues 81–113 of LmxGT1 in L. mexicana. Yellow arrowheads show examples of overlapping GFP expression and tubulin. White arrowheads indicate absence of GFP signal. Scale bars: 5 µm.

Targeting of membrane-bound GFP but not cytoplasmic GFP to the flagella

Residues 1–130 and 81–113 were tested to determine whether they can target a peripheral membrane protein to the flagellum. A membrane-bound GFP was made by fusing the first 18 amino acids from the L. major HASPB protein to the N-terminus of GFP, generating HASPB(18)::GFP. The HASPB(18) sequence is sufficient to target GFP to the plasma membrane in Leishmania through dual acylation of the N-terminal glycine and cysteine residues (Denny et al., 2000) (Fig. 4B,C). While HASPB(18)::GFP localized primarily to the pellicular plasma membrane (75%, n = 220) (Fig. 4B), the addition of residues 81–113 of LmxGT1 between the HASPB and GFP sequences re-targeted this chimeric protein to the flagellar membrane almost exclusively in 50% of cells assayed (n = 211) (Fig. 4B,C). In the remaining 50% of cells, GFP signal could be detected in both the flagellar and pellicular plasma membranes (Fig. 4B,C). Curiously, a chimeric protein generated by inserting residues 1–130 of LmxGT1 between HASPB(18) and GFP failed to reach the plasma membrane and remained in the cytoplasm (not shown). We do not have a definitive explanation for this latter phenotype, but it is possible that acylation of the HASPB(18) domain is adversely affected in this chimera.

To determine whether membrane association is a prerequisite for the LmxGT1 N-terminal domain to target proteins to the flagella, either residues 1–130 or 81–113 of LmxGT1 were fused to GFP and the localization of the chimeric proteins was examined. Both chimeric proteins failed to target to the flagella (Fig. 4D). Taken together, these results suggest that the mechanism for flagellar trafficking of LmxGT1 requires membrane association, and that it may not apply for proteins that target to the flagellar axoneme or lumen.

Overview

The combination of deletion mutagenesis, site-directed mutagenesis, and chimeric constructs employed in the study establish that residues 81–113 within the N-terminal hydrophilic domain of LmxGT1 (supplementary material Fig. S1) encompasses important ‘cis-acting’ information required for flagellar targeting. The NPM motif is critical for flagellar targeting, because deletion or site-directed mutagenesis of these residues results in a strong, albeit not complete, mistargeting phenotype. Nonetheless, these residues alone are not sufficient for even partial flagellar targeting of a non-flagellar membrane protein (Fig. 3C), indicating that additional sequence from within the N-terminal domain is required for flagellar targeting. In contrast, amino acids 81–113, which encompasses the NPM motif and surrounding sequence, facilitates at least partial flagellar targeting of some membrane proteins [LmxGT2, Fig. 3C; HASPB(18)::GFP, Fig. 4B] and can fully re-target another membrane protein to the flagellum (LdNT1.1, Fig. 4A). The observation that residues 81–113 are predicted to form a HLH secondary structure that presents the critical NPM motif at the surface of the protein suggests a potential mechanism for these sequences in flagellar targeting. Specifically, the surface exposed NPM residues may mediate interaction of LmxGT1 with partner proteins that induce migration of this permease from the flagellar pocket into the flagellar membrane. Current experiments are directed toward identifying potential interacting proteins.

The critical NPM motif is not obviously related to the limited number of ciliary targeting motifs identified in other integral membrane proteins (Nachury et al., 2010). This observation is not surprising, since most of these proteins are from mammals that are highly divergent from ancient parasitic protozoa. In addition, this sequence is not related to those involved in flagellar localization of non-membrane proteins that have been identified among the kinetoplastid protozoa (Luginbuehl et al., 2010; Pullen et al., 2004), and this result would be consistent with the notion that flagellar membrane proteins traffic to that organelle by a different mechanism than non-membrane proteins. Furthermore, the NPM motif is not obviously represented within the sequence of two other membrane proteins that traffic to the flagellum in Leishmania species, the polytopic aquaporin 1 from L. major (Figarella et al., 2007) and SMP-1, a dually acylated peripheral membrane protein from Leishmania (Tull et al., 2004). Because the motif identified here is short and of unknown degeneracy, it may be difficult to identify functionally analogous sequences at the present time. Nonetheless, the identification of the NPM motif and surrounding sequence as critical for flagellar targeting further advances our understanding of the strikingly distinct subcellular trafficking of different glucose transporter isoforms in Leishmania and establishes this system as a model for understanding flagellar targeting.

Materials and Methods

Parasite culture and transfections

Wild-type and glucose transporter null-mutant (Burchmore et al., 2003) L. mexicana and wild-type L. donovani DI700 promastigotes were cultured in RPMI 1640 medium (Invitrogen, Carlsbad, CA) supplemented with 10% heat-inactivated fetal bovine serum (Thermo Scientific Hyclone, Logan, UT), 0.1 mM xanthine, and 5 µg/ml hemin. Parasite lines carrying episomal expression vectors were cultured in the same medium with 100 µg/ml G418 (Invitrogen, Carlsbad, CA). All cultures were maintained at 26°C. Leishmania promastigotes were transfected according to previously described electroporation techniques using a Bio-Rad Gene Pulser Xcell (Burchmore et al., 2003; Robinson and Beverley, 2003).

Generation of LmxGT1 mutants and chimera proteins

LmxGT1 truncations and deletion mutants were generated using standard PCR-based techniques. Single point mutations were made using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). The NPM::GT2 and PNPMKP::GT2 chimeras were generated by PCR amplification of the LmxGT2 ORF using a forward primer containing a BamHI restriction site followed by bases representing the amino acids MNPM or MPNPMKP and a reverse primer containing an EcoRV restriction site. Chimera proteins where generated by PCR amplification of specific regions from LmxGT1 and fused to LmxGT2, LdNT1.1, or GFP. All constructs were sequenced at the OHSU Core Sequencing facilities and analyzed using Sequencher software (Gene Codes Corporation, Ann Arbor, MI) to verify accuracy. Constructs were cloned into the pXG-GFP episomal expression vectors for expression of the GFP fusion protein in Leishmania (Ha et al., 1996). A full list of primers is available upon request.

Molecular markers and immunostaining

Cells were prepared as previously described (Snapp and Landfear, 1999). Antibodies, dilutions, and sources are as follow: rabbit GFP, 1∶1000 (Molecular Probes, Eugene, OR, USA); mouse α-tubulin, 1∶2000 (Sigma-Aldrich, St Louis, MO); anti-rabbit IgG Alexa Fluor 488 and anti-mouse IgG Alexa Fluor 594, 1∶1000 (Molecular Probes, Eugene, OR).

Molecular modeling using Rosetta Software

A structural model for an N-terminal segment, consisting of the 130 amino acid domain of LmxGT1, was constructed by ab initio modeling using Rosetta version 3.1. Five-thousand independent structures were predicted on a locally installed version of Rosetta and the structures were subjected to clustering analysis. The ‘best model’ presented in this manuscript represents the cluster containing the most structure predictions with the lowest standard deviation of the mean among the positions of α-carbon atoms of all residues to all other simulations within the cluster (within 3 Ångstroms). Side chains were modeled using SCWRL3.0, a program for predicting protein side-chain conformations for a given backbone. The best models were minimized using CharmM and then validated as previously described (Subbian et al., 2004). The structures were visualized using PYMOL version 0.99.

Microscopy and data analysis

Fluorescence images were captured using a Carl Zeiss Axiovert 200M fluorescence microscope and AxioVision R4.6 image capture software (Carl Zeiss Microscopy, Thornwood, NY). Images were analyzed and processed using ImageJ (NIH, Bethesda, MD) and figures were constructed using Adobe Illustrator CS3 (Adobe Corporation, San Jose, CA). For each cell line, at least 150 cells expressing GFP were examined to ensure a representative pattern.