Two Functionally Divergent UDP-Gal Nucleotide Sugar Transporters Participate in Phosphoglycan Synthesis in Leishmania major*

In the protozoan parasite Leishmania, abundant surface and secreted molecules, such as lipophosphoglycan (LPG) and proteophosphoglycans (PPGs), contain extensive galactose in the form of phosphoglycans (PGs) based on (Gal-Man-PO4) repeating units. PGs are synthesized in the parasite Golgi apparatus and require transport of cytoplasmic nucleotide sugar precursors to the Golgi lumen by nucleotide sugar transporters (NSTs). GDP-Man transport is mediated by the LPG2 gene product, and here we focused on transporters for UDP-Gal. Data base mining revealed 12 candidate NST genes in the L. major genome, including LPG2 as well as a candidate endoplasmic reticulum UDP-glucose transporter (HUT1L) and several pseudogenes. Gene knock-out studies established that two genes (LPG5A and LPG5B) encoded UDP-Gal NSTs. Although the single lpg5A– and lpg5B– mutants produced PGs, an lpg5A–/5B– double mutant was completely deficient. PG synthesis was restored in the lpg5A–/5B– mutant by heterologous expression of the human UDP-Gal transporter, and heterologous expression of LPG5A and LPG5B rescued the glycosylation defects of the mammalian Lec8 mutant, which is deficient in UDP-Gal uptake. Interestingly, the LPG5A and LPG5B functions overlap but are not equivalent, since the lpg5A– mutant showed a partial defect in LPG but not PPG phosphoglycosylation, whereas the lpg5B– mutant showed a partial defect in PPG but not LPG phosphoglycosylation. Identification of these key NSTs in Leishmania will facilitate the dissection of glycoconjugate synthesis and its role(s) in the parasite life cycle and further our understanding of NSTs generally.

Protozoan parasites of the genus Leishmania must survive in two separate and harsh environments to complete their life cycle: extracellularly within the midgut of the sand fly and intracellularly within the phagolysosome of vertebrate macrophages. A variety of abundant secreted and surface glycoconjugates have been implicated in key steps of the infec-tious cycle (1). Several of the most abundant of these contain repeating phosphoglycan (PG) 2 polymers, such as lipophosphoglycan (LPG) and proteophosphoglycans (PPGs) (2). Secreted acid phosphatases also contain PGs, and other less abundant PG-containing molecules exist (3,4). Leishmania PG repeating units contain the disaccharide-phosphate Gal(␤1,4)-Man(␣1)PO 4 (referred to hereafter as the PG repeat unit) (5). PG repeats frequently contain additional sugar substitutions in different species and strains, most commonly on the Gal residue, although modifications of the Man also occur ( Fig. 1) (6).
The assembly of Leishmania PG repeating units to form LPG and PPGs occurs in the Golgi apparatus (16 -18). Nucleotide sugar transporters (NSTs) transport cytoplasmically synthesized nucleotide sugars into the endoplasmic reticulum (ER) or Golgi lumen, where they are consumed in glycosylation reactions (19 -22). A large number of NSTs have been identified, and they now constitute a well characterized family of membrane proteins (20,23). Previously, we showed that the Leishmania LPG2 gene encoded the Golgi GDP-Man transporter, one of the earliest NSTs identified, and the first multispecific NST, since it was additionally able to transport GDP-D-Ara and GDP-L-Fuc (17). lpg2 Ϫ mutants in three species of Leishmania were completely devoid of PGs (24 -26); however, the consequences to parasite virulence differed greatly. Mutant lpg2 Ϫ L. major were unable to survive within macrophages or to cause pathology (8), whereas lpg2 Ϫ L. mexicana were able to survive and induce disease (26).
Because Golgi GDP-Man transport is required for PG synthesis, we hypothesized there would be a matching requirement for UDP-Gal transport as well. Loss of UDP-Gal transport in L. major should therefore render parasites PG-deficient and provide an additional tool to study PG function. Galactose is found in other surface glycoconjugates of trypanosomatids (18), such as the glycan head groups of Type II glycosylinositolphospholipids and in glycosylphosphatidylinositol anchor side chains containing poly-N-acetyl-lactosamine (27), although this modification is not known to occur in Leishmania. The phenotype of galactose-deficient trypanosomes has been investigated previously through inactivation of UDP-glucose epimerase (28), a gene also found in Leishmania.
Since our interests concerned the role of galactosylation reactions occurring within secretory compartments, our strategy focused on inactivation of UDP-Gal transport. Thus, it was necessary first to identify the parasite's UDP-Gal transporters. Twelve candidate NST sequences were identified in the L. major genome, with five showing varying degrees of similarity to known UDP-Gal transporters, including two pseudogenes. Our data indicate that two candidates, LPG5A and LPG5B, both encode UDP-Gal transporters and are required for phosphoglycan synthesis in L. major, although their functions differ significantly. More limited studies of another candidate, HUT1L, suggest that it may be the functional homolog of the ER UDP-Glc NST, shown recently to be important to the protein folding glucosylation/reglucosylation process (29,30).
Leishmania cells were transfected by electroporation using either a low voltage (31) or high voltage protocol (34). Following transfection, cells were allowed to grow 16 -24 h in M199 medium with 10% fetal calf serum and then plated on semisolid media containing 1% Noble agar (Fisher) and the appropriate selective drugs. Individual colonies were picked and grown in liquid medium. Clones were maintained in selective medium and then removed from selection for one passage prior to experiments. Hygromycin B was from Calbiochem (now under EMD Biosciences (San Diego, CA)), G418 was from BioWhittaker (now under Cambrex Bio Science (Walkersville, MD)), puromycin powder was from Sigma, blasticidin was from Invitrogen, nourseothricin was from Werner BioAgents (Jena, Germany), and phleomycin was from InvivoGen (San Diego, CA).
[␣- 32 (2,9). The GPI-anchored form of PPG is depicted, but other forms of PPG are similar regarding the PG structures. X, linear ␤1,3-linked galactose residue(s) that branch off the Gal-Man-PO 4 repeating unit backbone (A). EtN, ethanolamine. molecular biological techniques including Southern blotting were used as described (35).
L. major Constructs-Molecular manipulations are summarized in Table S1. Briefly, ORFs were amplified from L. major DNA and cloned into the unique BamHI site of the Leishmania expression vectors pXG(NEO) (B1288) or pXG(PHLEO) (B3324). For expression, oligonucleotide primers (Table S2) added an optimal translation initiation sequence (CCACC) upstream of the ORF start codon. PCR products were directly cloned into the TA vector pGEM-TEasy TM (Promega, Madison, WI), sequenced, and then liberated with BamHI and inserted into Leishmania expression vectors. Where indicated, ORFs were liberated from pXG vectors with BamHI and cloned into the unique BamHI site of pCDNA3 (B1974 ; Table S1). pXG(NEO)-hUGT1-cHA (B5541) and pMKIT-hUGT1-cHA (B5034) were gifts from H. Segawa.
Mapping of the LPG5A Splice Acceptor Site-Mapping of splice acceptor sites was accomplished using RT-PCR (36). FV1 RNA was prepared from exponentially growing parasites using Trizol (Roche Applied Science) and used as a template to generate randomly primed cDNAs using Superscript II (Stratagene, La Jolla, CA). A universal miniexon primer (SMB936) was used in conjunction with oligonucleotide SMB1581 (Table  S2) to amplify portions of gene-specific cDNAs from the pool of randomly primed cDNAs described above, using Taq polymerase (Roche Applied Science). The single product was cloned into pGEM-TEasy TM (Promega, Madison, WI) and sequenced.
Generation of LPG5 Mutants and Add-back Lines-Targeting fragments for LPG5A and LPG5B were liberated from their respective vectors and purified prior to transfection into L. major LV39. For single-gene mutants, two rounds of transfection were done to recover homozygous mutants, and correct targeting was confirmed using Southern analysis. The names below follow the formal nomenclature for Leishmania (37). The ⌬lpg5A/⌬lpg5A clonal strain A1 was chosen for biochemical analysis and is referred to as lpg5A Ϫ . The^lpg5B/^lpg5B clonal strain A3 was chosen for biochemical analysis and is referred to as lpg5B Ϫ . To produce a strain targeted in both LPG5A and LPG5B, LPG5B was disrupted in the lpg5A Ϫ mutant. Clonal strain 2A-2, whose genotype ⌬lpg5A/⌬lpg5A/^lpg5B/^lpg5B was confirmed by Southern analysis, is referred to as lpg5A Ϫ / 5B Ϫ . Further details are discussed in supplemental materials.
Parasite strains were passed in vitro no more than six times (M1P6) before use.
Immunoblotting-Whole cell lysates from L. major promastigotes were prepared from exponential growth or stationary phase parasites in SDS sample buffer; 5 ϫ10 6 cell equivalents/ lane were separated by discontinuous SDS-PAGE on 4% stacking and 12.5% resolving gels and then electrotransferred to nitrocellulose membranes (Hybond-ECL; Amersham Biosciences). Protein loading was assessed by staining with 5% Ponceau S, 1% glacial acetic acid (w/v) prior to Western analysis. Monoclonal antibody WIC79.3 (38) was used at a 1:1000 dilution; CA7AE (39) was used at a 1:1000 dilution. Secondary antibodies conjugated to peroxidase were from Amersham Biosciences, and ECL reactions were conducted using Pierce or PerkinElmer Life Sciences chemiluminescence kits.
Fluorescent Lectin Binding-The binding of fluorescein isothiocynate-labeled Maackia amurensis agglutinin (EY Laboratories, San Mateo, CA) was assessed by fluorescence microscopy and fluorescence-activated cell sorting. Transfectants were removed from antibiotic selection for one passage prior to experiments. Cells grown on coverslips for IFA or scraped from flasks for fluorescence-activated cell sorting were washed once with phosphate-buffered saline and fixed for 1 min in 3.7% formaldehyde (v/v) in phosphate-buffered saline and incubated for 1 h at 25°C in 20 or 40 g/ml M. amurensis fluorescein isothiocyanate in phosphate-buffered saline. Cells were washed three times with phosphate-buffered saline and visualized using an Olympus AX-70 microscope (Melville, NY) or a BD Biosciences FACSCalibur.
The number of PG repeating units per LPG molecule was determined using two methods. The first method used gas chromatography-mass spectrometry (44). Briefly, the lipid anchor was cleaved by nitrous acid deamination, and the anhydromannose residue at the reducing end was further reduced with NaBH 4 to produce a single anhydromannitol residue for each LPG molecule. The LPG was hydrolyzed with 2 N trifluoroacetic acid to release neutral monosaccharides and phos-phorylated monosaccharides (mostly galactose 6-phosphate). The phosphorylated monosaccharides were removed by anion exchange chromatography, and then the neutral monosaccharides were acetylated before gas chromatography-mass spectrometry. The ratio of mannose to anhydromannitol was calculated, which was adjusted to account for the mannose present in the LPG core and capping sugars, to determine the number of PG repeats present per LPG molecule. The second method utilized capillary electrophoresis (43). Purified LPG was treated with nitrous acid, NaBH 4 , and trifluoroacetic acid as in the gas chromatography-mass spectrometry method. The released monosaccharides were directly labeled with 1-aminopyrene-3,6,8-trisulfonate, omitting acetylation, and separated on a Beckman P/ACE 5000 (Fullerton, CA). As with the gas chromatography-mass spectrometry method described above, the ratio of Man to anhydromannitol was determined to estimate the number of PG repeating units per LPG molecule. WT LPG was used as a control in all experiments.

RESULTS
Data Base Mining for NST Candidates-We used the BLAST algorithm to identify candidate NSTs in the Leishmania Genome Project data bases (45), focusing primarily on queries with proteins with experimentally confirmed NST activity. In the completed Leishmania major genome, 12 sequences showing similarity to one or more NSTs were found (Table 1). To prioritize those mostly likely to encode UDP-Gal transporters, we compared the candidate Leishmania NSTs with UDP-Gal transporters from humans, yeast, and plants as well as the GDP-Man transporter LPG2 and human YEA4 (Table 1).
Two additional potential L. major NSTs were only detected using more sensitive PSI-BLAST comparisons ( Table  1). LmjF07.0400 showed a relationship to members of solute carrier subfamily family 35F, whose functions have not been determined. LmjF36.0670 showed a relationship to members of solute carrier family 35A, including human SLC35A3, which mediates UDP-GlcNAc uptake (47,48).
Phylogenetic analysis showed that the candidates LmjF24.0360 and LmjF18.0400 associated with other functionally character- ized UDP-Gal transporters from S. pombe, plants, and mammals ( Fig. 2 and Table S3; data not shown). Although encouraging, this cluster contained transporters with specificities for a variety of UDP-sugars as well as CMP-sialic acid transporters (23), reinforcing prior observations that phylogeny is of only limited usefulness in predicting nucleotide sugar specificity in the NST family (20,21,49). Nonetheless, we used evolutionary relationship as a guide to focus on the two candidates showing the best homology to hUGT1, which we named LPG5A and LPG5B (Table 1), and carried out limited studies on the HUT1related gene LmjF22.1010, which we named HUT1L (HUT1-like).
We identified orthologs for LPG5A, LPG5B, and HUT1L in the provisional genome sequences of Leishmania infantum and Leishmania braziliensis. Interestingly, seemingly intact orthologs for pseudogenes LmjF15.0840 and LmjF15.1055 were found in L. infantum (LinJ15.0890 and LinJ20.1080, respectively) but not L. braziliensis. We did not identify additional NSTs in these species beyond orthologs of those of L. major, although this conclusion is tentative, since their genomes were not complete.
Properties of the LPG5A and LPG5B Open Reading Frames-Typical eukaryotic NSTs are about 320 -400 amino acids in length; however, the predicted LPG5A and LPG5B proteins comprised 600 and 561 amino acids, respectively (Table  1). We mapped the 5Ј end of the LPG5A mRNA using reverse transcription-PCR and the fact that all Leishmania mRNAs bear a common 39-nucleotide 5Ј leader sequence added by trans-splicing. This placed the splice acceptor at a position located 384 nucleotides 3Ј of the annotated AUG for the 600amino acid LPG5A ORF (Fig. 3A). The next AUG following this mapped splice acceptor occurred 63 nucleotides downstream (Fig. 3B), leading to a predicted LPG5A protein of 451 amino acids ( Table 1). As shown below, the 451-amino acid LPG5A ORF was functional. Although efforts to similarly map the LPG5B 5Ј end were unsuccessful, current data suggest that the 561-amino acid ORF for LPG5B is correct (discussed below).
Several transmembrane (TM) prediction algorithms suggested the occurrence of 8 -10 TM domains for both LPG5A and LPG5B, consistent with predictions for other NSTs (Table 1, Figs. 4 and S1). We incorporated previous studies of the murine CMP-sialic acid transporter, whose topology has been experimentally determined and whose sequence is closely related to UDP-Gal transporters (50). From this and sequence alignment criteria, we developed models predicting 10 transmembrane domains in the LPG5A and LPG5B proteins (Figs. 4 and S1; Fig. S2 presents an amino acid alignment with shaded predicted TM domains).
One striking feature in both LPG5A and LPG5B was the prediction of a large cytoplasmic loop (134 and 199 amino acids, respectively) between TM domains 2 and 3 (Figs. 4 and S1). For comparison, this loop is 25-26 amino acids in the mammalian UDP-Gal and CMP-Sia NSTs, and no other characterized NSTs to date bear a loop comparable with those of LPG5A or LPG5B.  (71), which incorporates input from a number of pairwise and multiple alignment procedures and generates a consensus. Following the removal of gap positions, a bootstrap consensus evolutionary tree was constructed using the neighbor-joining algorithm (1000 replicas) implemented in the MEGA 3.1 evolutionary analysis package (72). The smaller numbers at the nodes represent the bootstrap consensus percentages and are a measure of confidence. Tests of a variety of different alignment and tree algorithms did not reveal any significant differences within the major lineages depicted here (data not shown). A table of pairwise percentage identities can be found in Table S3. Due to its greater evolutionary distance, protein Lmj18.1540 was not included in this analysis. Data base mining for L. infantum and L. braziliensis suggest that orthologs for the L. major genes depicted here are present in these species as well.  (Table S2) were used to amplify LPG5A-specific cDNA from a pool of L. major randomly generated cDNAs. Lanes not relevant to this experiment were removed. One RT-PCR was done either in the absence of reverse transcriptase (ϪRT) or in the presence of reverse transcriptase (ϩRT). The std lane contains the 1KBϩ marker (Invitrogen); the asterisk indicates the 650-bp marker. B, the PCR product from A was cloned into pGEM-TEasy TM and sequenced; from this, the location of the splice acceptor was determined, and this AG is shown (in boldface type) on the genomic DNA sequence of this region. The predicted LPG5A start codon (in boldface type) is located 63 nucleotides 3Ј of this splice acceptor.
Other conserved features included several buried charged residues, including conserved Glu and Lys residues in the second predicted TM2 domains of LPG5A, LPG5B, murine CMP-Sia, and human UDP-Gal NSTs (Figs. 4, S1, and S2). Residues shown to be functionally important in the mammalian UDP-Gal and CMP-Sia (21) were conserved in LPG5A and LPG5B as well (Figs. 4, S1, and S2).
LPG5A and LPG5B Rescue the Glycosylation Defect of the UDP-Gal NST-deficient Lec8 Mutant-To test the function(s) of the three UDP-Gal transporter candidates LPG5A, LPG5B, and HUT1L, we asked whether heterologous expression of these ORFs could rescue the CHO cell mutant Lec8. This mutant is deficient in UDP-Gal transport, leading to decreased glycoprotein galactosylation, deficiency in sialylation, and a failure to react with the sialic acid (␣2,3)-galactose-specific M. amurensis agglutinin (41,51). When a functional UDP-Gal transporter is expressed in these cells, M. amurensis agglutinin binding is restored (52), a widely used assay to functionally identify and characterize UDP-Gal transporters from a variety of species (19,21,22).
M. amurensis agglutinin binding was examined in stably transfected Lec8 cells bearing constructs expressing LPG5A, LPG5B, or HUT1L. By fluorescence microscopy (Fig. 5A) and flow cytometry (Fig. 5B), LPG5A and LPG5B transfectants showed strong M. amurensis agglutinin binding, similar to that of WT CHO cells or to Lec8 cells expressing the human UDP-Gal transporter hUGT1. A caveat to the negative HUT1L result is that a specific antiserum to HUT1L was not available, precluding verification of its correct expression and/or localization in the transfectants. These data suggested that LPG5A and LPG5B probably encoded proteins with UDP-Gal transporter activity.

Generation of Null LPG5A, LPG5B, and Double Mutants-
To further study the candidate UDP-Gal transporters, we created null mutants. Leishmania are asexual and predominantly disomic, necessitating the use of two successive rounds of transfection to generate null mutants (53,54). For LPG5A, we designed constructs that would precisely replace the 451amino acid LPG5A ORF with drug resistance marker ORFs encoding hygromycin B (HYG) or puromycin (PAC) resistance (Fig. 6A). Starting with the WT LV39 strain of L. major, correct replacement of both alleles was confirmed by Southern analysis (Fig. 6C), and this mutant is termed lpg5A Ϫ .
For LPG5B, we were unable to determine the mRNA start site by RT-PCR using a variety of LPG5B-specific primers and amplification conditions (data not shown); the reasons for this are not evident to us. The assignment of the LPG5B ORF, however, was supported by functional assays (Fig. 5), sequence alignment (Fig. S2), and transfection assays showing that several shorter ORFs initiating at downstream AUGs were inactive (data not shown). Thus, for LPG5B an insertional inactivation approach was used where autonomous cassettes mediating resistance to blastocidin (BSD) and G418 (NEO) were inserted into the LPG5B ORF (Fig. 6B). Correct targeting was confirmed by Southern analysis (Fig. 6D), and this mutant was designated lpg5B Ϫ . To make the lpg5A Ϫ /5B Ϫ mutant, we inactivated LPG5B in the background of the lpg5A Ϫ mutant above. Successful alteration of both genes was confirmed by Southern analysis (Fig. 6, C and D). All mutants grew at normal rates and to normal densities in standard M199 culture media in vitro (data not shown).
LPG5A or LPG5B Is Sufficient for PG Synthesis, but the lpg5A Ϫ /5B Ϫ Mutant Lacks PGs-L. major PG repeats are often modified by side chain Gal residues, added in consecutive ␤1,3-  (50). Charged residues predicted to be buried in transmembrane domains are shaded black, and conserved residues where inactivating mutations have been mapped in other NSTs are shaded gray. Residues with a heavy outline are identical or similar with the murine CMP-Sia transporter. A large 134-amino acid cytoplasmic loop was identified between TM2 and TM3, whose full sequence is not shown.
linkages to the Gal residue within the PG repeating unit by specific Gal-transferases ( Fig. 1) (55). Gal-modified PG repeats can be recognized by the monoclonal antibody WIC79.3 (56), whereas unmodified PG repeats can be recognized by monoclonal antibody CA7AE (39). We used these in immunoblotting to probe the effects of LPG5A and LPG5B mutations on PG synthesis; similar results were obtained by immunofluorescence microscopy (data not shown).
WT extracts showed strong reactivity to WIC79.3 in the 25-100-kDa region, corresponding to LPG, whereas a much fainter reactivity was observed within the stacking gel, where the large PPGs migrate under these conditions (Fig. 7A, lane 1) (7,10). In the lpg5A Ϫ mutant, strong WIC79.3 reactivity remained, but it migrated faster with an apparent molecular mass of 20 -40 kDa (Fig. 7A, lane 2). A slight increase in WIC79.3 reactivity in the stacking gel/PPG region was also observed (Fig. 7A, lane 2). In contrast, for the lpg5B Ϫ mutant, strong WIC79.3 reactivity in the 25-100-kDa "LPG region" remained, whereas greatly elevated reactivity was observed within the stacking gel/"PPG region" (Fig. 7A, lane 3). These data suggested that loss of LPG5A primarily affected PGs associated with LPG, whereas loss of LPG5B primarily affected PGs associated with PPG.
Both the lpg5A Ϫ and lpg5B Ϫ mutant showed a small increase in CA7AE reactivity in the stacking gel/PPG region (Fig. 7B,  lanes 2 and 3 versus lane 1). This was consistent with the idea that loss of UDP-Gal NST activity leads to decreased levels of galactosylated PGs. The lpg5B Ϫ PPG showed slightly decreased mobility relative to that of lpg5A Ϫ (Fig. 7B, lanes 2 and 3). The significance of this observation is uncertain, since structural studies of the PPG were not undertaken; potentially, the CA7AE-reactive PPG population is a minor fraction of total PPGs.
In contrast to the single mutants, no reactivity with either monoclonal antibody was seen for the double lpg5A Ϫ /5B Ϫ mutant, in either the LPG or stacking gel/PPG regions (Fig. 7, A  and B, lanes 5). This phenotype was identical to that seen in the lpg2 Ϫ mutant (Fig. 7, A and B, lanes 6), which is completely deficient in PGs through loss of the GDP-Man transporter (17). Thus, lpg5A Ϫ /5B Ϫ had similarly completely lost PG synthesis. PG expression was restored by reintroduction of either LPG5A or LPG5B, singly (data not shown) or in combination (Fig. 7C).
Altered LPG Structure in lpg5A Ϫ but Not lpg5B Ϫ Mutants-The altered mobility of lpg5A Ϫ LPG (Fig. 7A, lane 2) suggested that it was structurally altered, presumably arising through decreased levels of galactosylation. This could arise through decreased numbers of PG side chain Gal residues or PG repeats or both (Fig. 1A). We thus purified LPG from WT and mutant strains and determined its structure ( Table 2). Side-chain galactosylation was assessed by depolymerizing LPG with mild acid to release the PG repeating units, which were then separated by capillary electrophoresis and quantitated. To examine the number of PG repeats in the LPG backbone of the lpg5A Ϫ mutant, we measured the ratio of mannose to anhydromanitol, the former arising from the PG repeats and the latter from the LPG core, following removal of the lipid anchor by nitrous acid deamination and reduction with NaBH 4 and hydrolysis to monosaccharides (see "Experimental Procedures").
For lpg5A Ϫ LPG, a general decrease in both types of galactosylation was seen ( Table 2). The fraction of PG repeats lacking Gal side chains was 3-4-fold higher (0.40 versus 0.09 WT in log phase), and 43% fewer Gal residues per PG repeat were added. Moreover, the number of lpg5A Ϫ PG repeats was reduced to 50 -74% of WT (13.5 and 10 versus 18.2 and 19.8 for WT log and stationary phase LPG, respectively). In contrast, LPG from the lpg5B Ϫ mutant closely resembled that of WT, in the number of side chain Gal residues/PG repeat, the fraction of unmodified PG repeats, and the total number of repeats (Table 2). These results were consistent with the immunoblot analysis (Fig. 7). Thus, loss of LPG5A, but not LPG5B, resulted in a general deficit in LPG galactosylation, affecting both side chain and PG repeat galactosylation quantitatively, to about the same extent.
Loss of LPG5B Primarily Affects PPG Modification-Although LPG was unaffected by loss of LPG5B, PPGs in the lpg5B Ϫ but not lpg5A Ϫ mutant reacted much more strongly with antibody WIC79.3 (Fig. 7A, lanes 1-3). This gain seemed contrary to the expectation that loss of a UDP-Gal transporter would lead to a deficit in some aspect of PG galactosylation, which might lead to a loss (or perhaps no change) in WIC79.3 reactivity. However, if WIC79.3 were more strongly reactive with PG repeats bearing fewer side chain Gal residues, increased reactivity could arise by an increased level of monogalactosylated PG repeats, at the expense of polygalactosylated PG repeats. The epitope specificity of WIC79.3 was previously examined by galactosyl-ligand inhibition studies (38). To assess the specificity of WIC79.3 in the context of intact LPG, a series of parasite lines expressing LPGs showing a range of side chain PG galactosylation were studied. As shown in the supplemental materials, when tested against intact LPG or PPG in Western blotting, monoclonal antibody WIC79.3 shows a strong preference for monogalactosylated PG repeats, reacting poorly with oligogalactosylated PG repeats and not at all with polygalactosylated PG repeats (Fig.  S3). This matches the finding that the LPG present in the immunogen for WIC79.3 was probably monogalactosylated LPG (38).
Thus, the immunoblotting results are explicable as follows. The much greater reactivity of PPGs with WIC79.3 in the lpg5B Ϫ mutant can be attributed to increased levels of monogalactosylated PG repeats, relative to WT PPGs that bear predominantly oligo-/polygalactosylated PG repeats. Thus, loss of LPG5B appears to result in a decrease in galactosylation, but specifically affecting PPGs and not LPG (Table  1), in contrast to loss of LPG5A, where the reverse is seen.
The lpg5A Ϫ /5B Ϫ Mutant Is Rescued by Expression of a Heterologous UDP-Gal Transporter-Attempts to demonstrate decreased UDP-Gal uptake with purified lpg5A Ϫ /B Ϫ microsomes were unsuccessful (supplemental materials). This raised the formal possibility that the PG deficiency of the lpg5A Ϫ /5B Ϫ mutant might arise through mechanisms other than decreased UDP-Gal uptake. If this idea were correct, we reasoned that expression of a distantly related heterologous UDP-Gal transporter would be unable to rescue PG synthesis. Thus, we expressed the human UDP-Gal transporter hUGT1 in the lpg5A Ϫ /5B Ϫ mutant. However, in these transfectants, WIC79.3reactive LPG was restored fully (Fig. 7D, lane 3 versus lane 1). It is highly unlikely that hUGT1 expression would rescue PG synthesis by any mechanism other than restoration of UDP-Gal uptake, and thus we conclude that the PG deficiency of the lpg5A Ϫ /B Ϫ double mutant arises from a lack of UDP-Gal transport activity.

DISCUSSION
LPG5A and LPG5B Encode UDP-Gal Transporters Required for L. major Phosphoglycan Synthesis-Twelve NST candidate sequences were identified in the L. major genome by data base mining. Of these, LPG5A, LPG5B, and HUT1L showed an evolutionary relationship to NSTs able to transport UDP-Gal in other species (Fig. 2). Since NST phylogeny does not always reflect substrate specificity (21,23,49), we tested these three candidates in functional assays. Expression of both LPG5A and LPG5B, but not HUT1L, could rescue M. amurensis agglutinin binding in UDP-Gal transport-deficient Lec8 cells (Fig. 5). Although L. major mutants inactivated in either LPG5A or LPG5B still made PGs, we found that a mutant genetically defective in both LPG5A and LPG5B lacked detectable PGs (Fig. 7) and that PG levels were restored by re-expression of either (or both) of these genes (Fig. 7) but not by overexpression of HUT1L (data not shown). We were unable to measure UDP-Gal uptake in purified Leishmania microsomes using methods FIGURE 6. Inactivation of LPG5A and LPG5B. A, allelic replacement strategy for LPG5A. The HYG/PAC targeting fragment is shown above the map of the LPG5A locus. The probe used for Southern analysis in C is indicated. B, insertional inactivation strategy used for LPG5B. In this experiment, the targeting fragment (not shown) consists of the LPG5B ORF into which autonomous BSD or NEO selectable markers were inserted (see "Experimental Procedures"). Predicted restriction fragments and the probe used for Southern blotting in D are indicated (boldface lines). In the analysis shown in panel D, the probe used for Southern analysis (see "Experimental Procedures") lay outside the targeting fragment as shown. X, Bg, Bs, H, and S, XhoI, BglII, BsrGI, HindIII, and SalI sites, respectively. C, autoradiogram following Southern analysis of the LPG5A locus in WT and mutants; DNA was digested with XhoI, and the probe used is shown in A. D, autoradiogram following Southern analysis of LPG5B locus in WT and mutants; DNA was digested with SalI, and the probe used is shown in B.
applied successfully previously to the LPG2 Golgi GDP-Man transporter, possibly due to the presence of a strong interfering galactosylation activity, whose source was not identified (supplemental material). Thus, to confirm that the loss of PG synthesis in the lpg5A Ϫ /5B Ϫ mutant arose specifically from a lack of UDP-Gal NST activity, we introduced the human UDP-Gal NST hUGT1, which completely restored PG synthesis. Together, these data indicate that LPG5A and LPG5B both encode UDP-Gal transporters, whose combined inactivation results in loss of PGs in L. major.
From the structures of known Leishmania glycoconjugates, one could predict that in addition to GDP-Man and UDP-Gal, NSTs mediating uptake of UDP-Glc, GDP-D-arabinopyranose, and UDP-galactofuranose are required. Many NSTs are multispecific and able to transport more than one nucleotide sugar, such as LPG2, which can transport guanine diphosphoarabinose and guanine diphosphofucose in addition to GDP-Man (17,23,57). Since UDP-Glc epimerase is localized to the glycosome in trypanosomatids (58), potentially UDP-Gal transport from this compartment may depend upon genes in the NST family.
The LPG5A/5B-related group of Leishmania NSTs appears to be a dynamic area of evolutionary variation, with the occurrence of at least two pseudogenes in L. major (Table 1), both of which may encode a functional gene in L. infantum (LinJ15.0890 and LinJ20.1080) but apparently not in L. braziliensis. Moreover, the NST gene families of Trypanosoma brucei and Trypanosoma cruzi appear to be as complex and diverse as seen in Leishmania (available on the World Wide Web at www.genedb.org; data not shown). Of the remaining NSTs, several showed a distant relationship to the LPG2 GDP-Man transporter (LmjF19.1490/19.1510 and LmjF30.2680), and one showed a relationship to the YEA4 group of NSTs, which includes UDP-glucuronic acid/UDP-GalNAc transporters (LmjF15.0840).
Orthogologous genes related to LmF36.0670 were restricted to Leishmania species, whereas an ortholog of LmjF07.0400 was present in T. brucei but not T. cruzi, and ones related to Lmj19.1490/19.1510 occurred in the T. cruzi but not T. brucei genomes. Potentially these NSTs could be redundant, have specificities for nucleotide sugars required uniquely by each species, or act in distinct cellular compartments. One attractive role for NSTs shared by L. major and T. cruzi, but not T. brucei, involves UDP-galactofuranose, since UDP-galactopyranose mutase, the enzyme responsible for UDP-Gal f synthesis, is absent in the T. brucei genome (59). Correspondingly, Leishmania and T. cruzi, but not T. brucei, synthesize a variety of Gal f -containing glycoconjugates (see references cited in Ref. 59).
A Probable Role for Leishmania HUT1L as an ER UDP-Glc NST-Leishmania HUT1L was of interest, since its closest relatives mediate UDP-Gal uptake in other species (30). However, heterologous expression of HUT1L did not rescue the UDP-Gal transport defect of Lec8 cells (Fig. 5). In preliminary studies, we attempted to generate null mutants for Leishmania HUT1L with the same methods successful with LPG5A and LPG5B ( Fig.  6; data not shown). Although heterozygous replacements were readily obtained, we were unable to obtain homozygous null mutants (data not shown). Instead, a variety of aneuploid and tetraploid cell lines were recovered bearing the second replacement as well as the WT gene (data not shown), a common finding in other studies in Leishmania attempting to knock out essential genes (60).
Recently, it was shown that the Arabidopsis HUT1L relative AtUTr1 encodes a UDP-Glc/Gal NST activity localized to the ER (29), as do fungal HUT1s (30). Interestingly, both HUT1L and AtUTr1 bear a C-terminal lysine-rich motif implicated in ER localization (61). Within the ER, AtUTr1 participates in the unfolded protein response, through provision of ER UDP-Glc required for the glucosylation/glucosidase cycle occurring as part of protein folding and quality control (29,62,63). Were Leishmania HUT1L similarly involved in the protein folding cycle in Leishmania, its loss could prove lethal to the cell, since unlike other organisms, the core glycans added to nascent trypanosomatid glycoproteins lack Glc residues and must necessarily be acquired in the ER (64). Preliminary data show localization of a green fluorescent protein-tagged HUT1L protein to the Leishmania ER, 3 and thus we believe that Leishmania HUT1L may likely be the parasite's ER UDP-Glc transporter. Functional studies will be required in the future to confirm the specificity of this and other novel NSTs identified here.
LPG5A and LPG5B Show Functional Divergence-Although complete ablation of PG synthesis required inactivation of both LPG5A and LPG5B, both the single lpg5A Ϫ and lpg5B Ϫ mutants made abundant PGs, albeit with different phenotypes. For the lpg5A Ϫ mutant, a structurally altered LPG was made (Fig. 7A), containing less side chain galactosylation and fewer PG repeating units ( Table 2), but alterations in PPG immunoreactivity were not detected. In contrast, the structure of LPG was unchanged in the lpg5B Ϫ mutant ( Table 2), but alterations in PPG immunoreactivity were seen (Fig. 7A). These were attributed to decreased galactosylation, leading to an increase in WIC79.3-reactive monogalactosylated PGs at the expense of oligo-or polygalactosylated PGs linked to PPG (Figs. 7 and S3).
What underlies the functional differences between LPG5A and LPG5B? Some NSTs display specific interactions with glycosyltransferases, as in the case of the mammalian Golgi UDPgalactose transporter UGT1 with UDP-galactose:ceramide galactosyltransferase (65). Building upon this model, perhaps key Gal transferases responsible for LPG versus PPG PG synthesis preferentially associate with the LPG5A and LPG5B UDP-Gal NSTs, respectively. Another possibility invokes differential localization within the secretory pathway of LPG5A and LPG5B. Current data strongly indicate that PGs incorporated into both LPG and PPG are synthesized in the parasite Golgi apparatus (16 -18, 26). Although it has been generally assumed that relevant NSTs and glycosyltransferases are colocalized in the same compartment, recent data also raise the possibility that luminal nucleotide sugars may traffic between compartments as well (66,67). Moreover, con-siderable data point to the existence of functionally divergent subcompartments within and after the eukaryotic Golgi apparatus (68).
Either model above could explain why, despite some degree of functional separation, both LPG5A and LPG5B function must be lost in order to completely ablate PG synthesis. For the association model, the coupling could be leaky, and for the compartmentalization, anterograde and/or retrograde vesicular trafficking of luminal UDP-Gal could result in limited redistribution.
The structure of the LPG5A and LPG5B proteins themselves provide some opportunities for functional divergence. Although most features closely resemble those of other known NSTs, LPG5A and LPG5B are unique in possessing a large, divergent 134 -199-amino acid insert between the second and third TM domains (Figs. 4 and S1). Moreover, the C-terminal domain of several NSTs has been implicated in targeting NSTs to the ER and/or Golgi apparatus, through dilysine retention and/or hydrophobic export motifs (66,69,70). Both LPG5A and LPG5B show potential motifs of this sort (Fig. S2). Thus, it is reasonable to propose that these or other domains contribute in some way toward divergent functions of these two NSTs. Preliminary data suggest that at least some portion of C-terminally tagged LPG5A and LPG5B proteins are found within the parasite Golgi apparatus, and efforts are under way to more finely map the cellular distribution of these two proteins and the functional roles of the unique peptide insert domains in LPG, PPG, and other secretory pathway glycoconjugates, such as the glycosylinositolphospholipids. 4 In summary, we have presented data for the existence of multiple NST candidate sequences in L. major. Pursuant to our interests in UDP-Gal transport, we found two functionally divergent UDP-Gal transporters, LPG5A and LPG5B, which both contribute to PG synthesis in L. major. These two NSTs show novel structural features and properties, studies of which are likely to increase our knowledge of the detailed Leishmania PG synthetic and secretory pathway in the future. Given the evolutionary diversity of glycoconjugates in both Leishmania and trypanosomes and their key roles during their infectious cycles, our studies further emphasize the roles of the large and diverse NST family in these important pathogens.

TABLE 2 Biochemical analysis of WT and mutant LPGs
Values shown are averages Ϯ S.D. For PG repeat compositions, n ϭ 4 for WT and n ϭ 2 for lpg5A Ϫ and lpg5B Ϫ determinations. For PG repeat unit number determinations, a single analysis was done. The gas chromatography-mass spectrometry method was used for lpg5A Ϫ , and the CE method was used for lpg5B Ϫ (see "Experimental Procedures"); both were used for WT, where they agreed well (18.2 Ϯ 1.9 and 19.8 Ϯ 0.2 for WT log and stationary phase, respectively). Statistical comparisons for the PG repeat units/LPG used these values.