Characterization of a Novel Alanine-rich Protein Located in Surface Microdomains in Trypanosoma brucei*

Heterologous expression in COS cells followed by orientation-specific polymerase chain reaction to select and amplify cDNAs encoding surface proteins in Trypanosoma bruceiresulted in the isolation of a cDNA (∼1.4 kilobase) which encodes an acidic, alanine-rich polypeptide that is expressed only in bloodstream forms of the parasite and has been termed bloodstream stage alanine-rich protein (BARP). Analysis of the amino acid sequence predicted the presence of a typical NH2 -terminal leader sequence as well as a COOH-terminal hydrophobic extension with the potential to be replaced by a glycosylphosphatidylinositol anchor. A search of existing protein sequences revealed partial homology between BARP and the major surface antigen of procyclic forms of Trypanosoma congolense. BARP migrated as a complex, heterogeneous series of bands on Western blots with an apparent molecular mass (∼50–70 kDa) significantly higher than predicted from the amino acid sequence (∼26 kDa). Confocal microscopy demonstrated that BARP was present in small discrete spots that were distributed over the entire cellular surface. Detergent extraction experiments revealed that BARP was recovered in the detergent-insoluble, glycolipid-enriched fraction. These data suggested that BARP may be sequestered in lipid rafts.

African trypanosomes are a group of unicellular eukaryotes responsible for sleeping sickness in man and related diseases in other mammals. The life cycle of these parasites alternates between the mammalian host and the tsetse fly vector and is characterized by the expression of two major surface glycoproteins: the variant surface glycoprotein (VSG) 1 and procyclic acidic repetitive protein (PARP) during the bloodstream and insect mid-gut stages respectively. An extensive literature exists concerning the function and expression of these two surface proteins (1)(2)(3). Although these studies have provided many fascinating insights, they have tended to occlude the question of other surface proteins in trypanosomes to the extent that by comparison with other eukaryotes our knowledge of trypanosomal surface proteins in general is characterized more by paucity than fecundity. In fact, apart from VSG and PARP only three other surface proteins have been characterized in detail: the heterodimeric transferrin receptor (4 -6), a Ca 2ϩ -regulated adenylate cyclase (7,8), and the glucose transporter (9). This shortcoming has been recognized by several workers, and a variety of approaches have been employed, with varying degrees of success, to provide a more complete description of the content, function, expression, and organization of trypanosomal surface membrane proteins (10,11). First, functional assays have been used to identify receptors for lipoproteins and lytic factors (12)(13)(14)(15) as well as for several plasma membrane enzymes (16,17), but the molecular nature of the proteins responsible remains to be established. Second, nucleotide sequencing of the VSG and PARP transcription units has identified several expression site-associated genes, termed ESAGs and procyclin expression site-associated genes, which may code for plasma membrane proteins; but, with the exception of ESAGs4, 6, and 7, the function and cellular location of the proteins encoded by these genes have yet to be established unequivocally (10,11). Third, classical biochemical approaches, involving surface labeling of whole cells using biotinylation or iodination, have allowed the identification, purification, and subsequent cloning of several stage-specific, invariant surface glycoproteins (18 -22). Finally, the heterologous expression of trypanosomal cDNAs in fibroblasts followed by immunoselection with appropriated antisera has been employed to identify procyclic stage surface proteins (23).
This study extends the latter approach by including an additional round of selection by orientation-specific PCR, using the mini-exon found at the 5Ј-end of all mature transcripts in trypanosomes, to select and amplify cDNA clones encoding surface membrane proteins from bloodstream as well as procyclic forms of the cell. This strategy led to the characterization of an alanine-rich, bloodstream stage-specific protein in Trypanosoma brucei which was related to the major surface antigen of a different life cycle stage of a different species of trypanosome: the glutamate-and alanine-rich protein, termed GARP, of procyclic forms of Trypanosoma congolense. In contrast to other surface proteins in T. brucei (11), BARP appears to be present in lipid microdomains that are distributed over the entire cellular surface from the posterior end of the cell body to the tip of the flagellum.

EXPERIMENTAL PROCEDURES
Materials-All radiochemicals and nitrocellulose filters were obtained from Amersham Pharmacia Biotech. N-Glycosidase F was obtained from Boehringer Mannheim. Alkaline phosphatase-conjugated and fluorescein isothiocyanate-labeled anti-rabbit IgG were obtained from Promega and Amersham Pharmacia Biotech, respectively. Other reagents were of the highest purity available.
Preparation of Parasites and Polyclonal Antisera-Monomorphic bloodstream forms of T. brucei strain MITat 1.1 were propagated in laboratory rats and purified as described previously (16). Insect procyclic forms of T. brucei were cultivated in SDM-79 medium containing 20% fetal calf serum (24). Hyperimmune antisera to parasite antigens from each life cycle stage were raised by immunizing rabbits with either live procyclic trypanosomes (2 ϫ 10 8 cells) or with plasma membranes (ϳ 3 mg of protein) purified from bloodstream trypomastigotes (MITat 1.1) as described previously (16). Both antisera were preabsorbed against COS cells in two successive overnight incubations (1.5 ml of serum/1.4 ϫ 10 8 cells) to remove heterophylic antibodies.
Preparation of cDNA Expression Libraries and Screening with Polyclonal Antisera-cDNA libraries were constructed in pCDM8 (25) using total cellular RNA from bloodstream (MITat 1.1) and insect procyclic forms of T. brucei exactly as described in our original report (23). Libraries were transfected separately into COS cells (50 g of DNA/10 7 cells) by electroporation (960 microfarads, 750 V/cm). After 48 h in culture (RPMI 1640, 10% fetal calf serum), both sets of transfectants were lifted in PBS, pH 7.5, containing 1 mM EDTA (1 h, 37°C). The cells were then incubated (5°C, 1 h) with antisera against either purified plasma membranes from bloodstream stage parasites or whole procyclic life cycle stage parasites. In each case the antisera were used at a final dilution of 1/12 in 1 ml of PBS, pH 7.5, supplemented with 1 mM EDTA, fetal calf serum (10%, v/v), and sodium azide (0.02%, w/v). Cell suspensions were centrifuged (1,000 ϫ g, 3 min) through Ficoll 400 (3%, w/v in PBS, pH 7.5) to remove unreacted antibody before "panning" (3 h, 25°C) on bacteriological plastic dishes coated with affinity-purified goat anti-rabbit IgG, Fc fragment-specific antibody (Jackson Immunochemicals, West Grove, PA). Plasmids were recovered from the adherent surface antigen-expressing cells using the Hirt procedure, transformed into Escherichia coli MC1061p3 by electroporation, amplified, and retransfected into fresh COS cells by spheroplast fusion for two more rounds of immunoselection by antibody panning. Plasmids rescued from the third round of panning were finally amplified after transformation of E. coli to yield two "mini-libraries" enriched for either bloodstream or procyclic stage surface antigen cDNAs.
Selection for Full-length cDNAs by Mini-exon PCR-cDNA minilibraries enriched for either bloodstream or procyclic stage surface antigen cDNAs were used as templates in two separate PCRs designed to amplify full-length sense-oriented clones containing the trans-spliced mini-exon sequence. Each PCR consisted of the appropriate template DNA (100 ng) in 0.1 M Tris-Cl, pH 8.3, 0.5 M KCl, 2.5 mM MgCl 2 , the forward mini-exon primer P1EcoRI (5Ј-CCGGAATTCGTTTCTGTAC-TATATTG-3Ј) the reverse primer pCDM8KpnI (5Ј-CAGGGTACCTAG-GTATGGAAGATCCCTC-3Ј) each at 0.2 M, deoxyribonucleotide triphosphates (0.25 mM each), and Taq polymerase 2U in a final volume of 100 l. The cycle parameters were 94°C, 1 min; 42°C, 1 min; and 72°C, 1 min for 30 cycles in a Techne PHC-1A thermal cycler. Products were recovered on glass beads (Geneclean, Bio 101, U. K.), digested with EcoRI and KpnI, and subcloned into doubly digested pUC19. Inserts of positive clones were sequenced on both strands by the dideoxy chain termination method (26) using the U. S. Biochemicals Sequenase version 2.0 kit in conjunction with primer walking where required. Sequence analysis was performed using the GCG/Wisconsin software (27).
DNA and RNA Analysis-The procedures employed for the isolation of DNA and RNA as well as Southern and Northern blot hybridizations are described elsewhere (28).
Expression of BARP in E. coli and the Production of Polyclonal Antisera-A 518-base pair XmnI/PvuII fragment that represented approximately two-thirds of the BARP open reading frame was isolated and subcloned into SmaI-digested pGEX2T. The fusion construct was sequenced using flanking vector sequences as primers to ensure that the sequence derived from the BARP gene was in the correct orientation and in-frame with the GST gene. Expression of the fusion product was induced in E. coli (29), and cell lysates were prepared and analyzed by SDS-PAGE. The fusion product, which migrated as a single band at the expected molecular mass of about 42 kDa (26 kDa from GST plus 16 kDa from BARP), was clearly visible on SDS-PAGE and was purified by GST affinity chromatography. Polyclonal anti-BARP antibodies were raised in rabbits, and the IgG fraction was prepared as described previously (21).
Immunofluoresence Microscopy-Trypanosomes were washed once with PBS (136 mM NaCl, 3 mM KCl, 16 mM Na 2 HPO 4 , 3 mM KH 2 PO 4 , 40 mM sucrose, 10 mM glucose, pH 7.6) and then resuspended (2 ϫ 10 7 cells/ml) in the same buffer at 0 -4°C. The cell suspension was mixed gently by inversion several times with an equal volume (20 -25°C) of freshly prepared paraformaldehyde (6%, w/v) in PBS adjusted to pH 7.6. The suspension was incubated at room temperature for 10 min. After washing, the fixed cells were resuspended at 2 ϫ 10 7 cells/ml and applied to polylysine-coated slides. After attachment of the cells the slides were washed with PBS containing glycine (1 mg/ml) and processed as described previously (21) using primary anti-BARP antibodies (1/250 dilution) followed by fluorescein isothiocyanate-labeled anti-rabbit IgG secondary antibodies (1/500 dilution). The cells were viewed and photographed using a laser confocal microscope equipped with a digital in situ imaging system. Partial Purification of BARP by Ion Exchange Chromatography-All steps were performed at 0 -4°C unless otherwise stated. A sample of surface radioiodinated (21) cells (1 ϫ 10 8 cells) was mixed with a suspension of freshly isolated trypanosomes (10 9 cells). The mixed suspension was washed twice in 50 mM Tes buffer, pH 7.5, 150 mM NaCl containing 10 mM glucose, and the pellet was resuspended (1.1 ml) in Tes buffer without glucose but containing CHAPS (1%, w/v), 30 g/ml leupeptin, 0.2 mM phenylmethanesulfonyl fluoride, 20 M E-64, 50 M N ␣ -p-tosyl-L-lysine chloromethyl ketone, and 1 mM EDTA. After a 1-h incubation, the suspension was centrifuged at 15,000 ϫ g for 1 h. The pellet was washed once, and the combined supernatants (total volume 3.5 ml) were concentrated and the Tes buffer replaced with 30 mM imidazole buffer, pH 6.5, containing 20 mM NaCl, 1% CHAPS, and protease inhibitors by repeated ultrafiltration using Centricon 10 filtration units (Amicon). After buffer exchange was completed the extract was applied (0.25 ml/min) to a DEAE-52 column (4 ml of packed resin) which had been equilibrated with imidazole buffer. The column was washed with 5 volumes of buffer before the application of a pH gradient (pH 6.5-4.0). Fractions were collected and samples analyzed for their content of radioactivity as well as for the presence of BARP by ELISA as described previously for ISG 100 (22). Fractions containing BARP, as judged by ELISA, or the peak of the 125 I label were subjected to Western blot analysis using antibodies against VSG or BARP, and those fractions containing the heterogeneous material corresponding to BARP were pooled. These pooled fractions were concentrated, dialyzed, and the imidazole buffer replaced with Tes buffer by ultrafiltration using Centricon 10 units (Amicon). Samples of these pooled fractions were treated with N-glycosidase F essentially as described previously (21,22). Briefly, SDS and ␤-mercaptoethanol were added to the to give a final concentration of 0.05% (w/v) and 19 mM, respectively. After these additions the sample was incubated at 75°C for 3 min, allowed to cool, and then 1/5 of the volume (of the sample) of 6 ϫ deglycosylation buffer (0.3 M Na 2 HPO 4 , pH 7.6, 60 mM EDTA, N-octyl glucoside (6%, w/v), and ␤-mercaptoethanol (6%, v/v))was added plus N-glycosidase F (5 units/ ml, final) and incubated for 20 h at 37°C. These conditions have been shown to result in the complete removal of N-linked carbohydrate from several surface proteins from T. brucei (21,22).
Isolation of Detergent-insoluble Glycolipid-enriched Fractions-Freshly isolated trypanosomes were subjected to a cold non-ionic detergent extraction procedure similar to that employed previously for the preparation of detergent-insoluble glycolipid-enriched microdomain fractions from T cells (30,31). All manipulations were performed at 4°C unless otherwise indicated. The cells (2-3 ϫ 10 8 cells) were lysed on ice in 1 ml of MNE buffer (50 mM Mes, 150 mM NaCl, 5 mM EDTA, 1 mM Na 3 VO 4 , 10 g/ml aprotinin, 0.1 mM phenylmethanesulfonyl fluoride, 1 mM NaF, and 1% Triton X-100, w/v) by gentle sonication (five bursts of 5 s at 5 W). The lysate was centrifuged in an Eppendorf for 5 min at 5,000 rpm at 4°C. The supernatant (1 ml) was mixed with an equivalent volume of 80% sucrose made with MNE buffer and transferred to a 5-ml ultracentrifuge tube. This solution was carefully overlaid with 2 ml of 35% sucrose followed by 1 ml of 5% sucrose (both prepared in MNE buffer), and the tubes were placed in a cooled AH650 Sorvall rotor. After centrifugation for 16 -18 h at 38,000 rpm, the tubes were removed, and the gradient was fractionated by the sequential collection of 2-ml, 1-ml, and finally 2-ml fractions from the top to the bottom of the gradient. The upper 2-ml fraction represented the detergent-insoluble glycolipid-enriched fraction because these fractions float in these sucrose gradients and are well separated from other detergent-insoluble material such as cytoskeletal assemblies (32). The proteins present in each fraction were concentrated by the precipitation procedure of Wessel and Flugge (33).

Isolation of cDNAs Encoding Trypanosomal Surface Protein by Heterologous Expression in COS Cells-Previously
we employed heterologous expression in COS cells to identify cDNAs coding for surface proteins in procyclic forms of T. brucei (23). This approach was extended to include bloodstream forms and included a further round of selection by orientation-specific PCR amplification. A range of different amplification products was obtained from immunoselected mini-libraries from both life cycle stages of T. brucei (Fig. 1). Three of the principal PCR products from either life cycle stage were selected for further analysis (see arrowheads in Fig. 1). Southern blot analysis indicated that all of the selected products were stage-specific (results not shown). The selected products were purified and subcloned into pUC19. Sequence analysis revealed that the two smaller procyclic stage products (ϳ0.8 and 0.5 kb) coded for two isoforms of PARP, the major surface antigen of procyclic forms of T. brucei, whereas the largest product (ϳ1.4 kb) coded for another relatively abundant procyclic surface protein, termed PSSA-2 (23). In the case of the bloodstream stage amplification products, only one of the selected products (ϳ1.4 kb) contained a complete open reading frame.
Characterization of a Trypanosomal cDNA Encoding an Alanine-rich Protein, Termed BARP-Sequencing of this 1.4-kb PCR product revealed an insert of 1344 base pairs beginning with 10 nucleotides of the mini-exon primer (EBI data bank accession number Y17281). The insert contained a single open reading frame with initiation and termination codons at positions 67 and 849, respectively. This was followed by a 3Јuntranslated region of 492 nucleotides which lacked a poly(A) tail. The translated amino acid sequence predicted a polypeptide containing 261 amino acid residues (ϳ27.6 kDa) which was rich in alanine (ϳ20%) and glutamic acid (ϳ12%) with a predicted acidic pI of 4.6 and, consequently, was termed BARP for bloodstream alanine-rich protein ( Fig. 2A). Two potential sites for N-glycosylation were present in the COOH-terminal half of  (34) demonstrated that the predicted protein was hydrophilic but contained hydrophobic segments at the NH 2 -and COOH-terminal regions of the protein (Fig. 2B). The NH 2 -terminal region was predicted to function as a leader peptide and contained a possible site for proteolytic cleavage at position 18 (35), whereas the COOHterminal hydrophobic extension has the potential to act as a cleavable signal peptide for the addition of a glycosylphosphatidylinositol (GPI) anchor with glycine at position 238 predicted to be the probable addition site of the anchor (36).
A Relationship between BARP and the Major Surface Antigen of T. congolense-A search of existing protein sequences using the FASTA analysis program revealed a potential homology between BARP and the major surface protein of procyclic forms of T. congolense (Fig. 3), a protein termed GARP for glutamateand alanine-rich protein (37,38). In addition to this homology at the amino acid sequence level, the two proteins also shared a number of additional features, which suggested that these two proteins from different life cycle stages of different species of Trypanosoma may be related (see "Discussion").
Genomic Organization and Transcription of the BARP Gene-Genomic DNA from T. brucei AnTat 1.1 was singly and doubly digested with SpeI and StuI, both of which cut at a single internal site in the BARP gene (Fig. 4A), and the digested DNA was probed with the purified StuI/SpeI fragment. A band corresponding to the internal StuI/SpeI fragment (ϳ0.77 kb) was clearly visible in the double digest (Fig. 4B, lane  1). Although the presence of other additional bands in the digests suggested the existence of related genes, there was no evidence for the presence of multiple or tandemly linked copies of the BARP gene as has been reported for other surface proteins (11). Northern blot analysis of mRNA prepared from different life cycle stages of T. brucei indicated that expression of the gene encoding BARP was likely to be stage-specific (Fig.  4C). These results demonstrated the presence of a single transcript (ϳ1.5 kb) in long slender and stumpy bloodstream forms which was less abundant in procyclic forms of the parasite. Finally, the results from genomic zooblots, performed under conditions of high stringency, demonstrated that the gene for BARP was also present in the subspecies T. brucei rhodesiense and T. brucei gambiense as well as the subgenus Trypanosoma evansi but was not detected in other related species, such as T. congolense, Trypanosoma cruzi, and Crithidia fasciculata (data not shown).
Characterization of the Protein Encoded by the Gene for BARP-Polyclonal antibodies against a GST-BARP fusion protein, containing approximately 70% of the BARP open reading frame (see Fig. 4A), were employed to characterize further the protein encoded by the BARP gene (Fig. 5). Although the amino acid sequence predicted a protein with a molecular mass of 26.7 kDa, Western analysis revealed that BARP migrated as an extremely heterogeneous series of bands, with an estimated molecular mass ranging between 50 and 70 kDa. BARP comigrated to a certain extent with the VSG, and a partial cross-reaction of the anti-BARP antibodies with VSG was observed (see also Fig. 6); but overall the pattern of staining was clearly different for both proteins (compare Fig. 5, lanes 1 and 3). Similar results were obtained using different bloodstream clonal variants of the same serodeme as well as with different serodemes (data not shown). Significantly, BARP was not detected in cultured procyclic forms of the parasite, which was consistent with the Northern blot analysis (Fig. 4C). An unusual feature of the Western blots using anti-BARP antibodies was the presence of background streaks that were not observed  in the case of procyclic forms. The antibodies against BARP behaved poorly in immunoprecipitations, and it was not possible to immunoprecipitate the protein from metabolically or surface labeled cells. Nevertheless, the results from Western blots indicated that BARP was an invariant, bloodstream stage-specific protein.
Partial Purification of BARP-Because BARP was predicted to have an unusually low pI, ion exchange chromatography was employed to provide a partial purification of the protein. This approach involved the fractionation of proteins from a CHAPS extract of cells surface labeled using 125 I on DEAE-cellulose under conditions where VSG does not bind to the resin. Labeled proteins that bound to the column were eluted by the application of a linear gradient of pH (Fig. 6A). Analysis of these fractions by ELISA using the anti-BARP antibodies demonstrated the presence of two peaks. The first peak contained proteins that did not bind to the resin and represented crossreaction of the anti-BARP antibodies with the large amount of VSG present in these fractions (Fig. 6B). The second peak eluted late in the gradient at a pH between 4.6 and 4.8. Western blots of these fractions demonstrated the presence of the heterogeneous pattern observed previously for BARP in whole cells (Fig. 6B). This material did not cross-react with anti-VSG antibodies (Fig. 6B). These data clearly demonstrated that BARP was distinct from the VSG and confirmed the predicted acidic pI of the protein.
Translation of the mRNA corresponding to the BARP gene in vitro resulted in the production of a single protein product with an apparent molecular mass of 28 kDa, consistent with the predicted size from the open reading frame, as determined by metabolic labeling with [ 35 S]methionine and Western blot/autoradiographic analysis of the reticulocyte lysate (results not shown). Therefore, the unusual electrophoretic mobility of BARP on SDS-PAGE was the result of post-translational modifications of the protein. Treatment of partially purified BARP with N-glycosidase F did not significantly affect the heterogeneity or relative mobility of BARP (Fig. 6C), whereas this treatment readily removed N-linked carbohydrate from ISG 100 (22) which decreased in size from 100 to 55 kDa as reported previously (data not shown). This result indicated that Nlinked carbohydrate was unlikely to be solely responsible for the unusual migration of BARP on SDS-PAGE. Partially purified BARP, which was devoid of VSG, was, at least partially, recognized by antibodies directed against the cross-reacting determinant of the VSG (Fig. 6C). This neo-epitope is dependent on the presence of an inositol-1,2-cyclic phosphate generated by cleavage of the GPI anchor of VSG by the action of the GPI-specific phospholipase C (39) and suggested that BARP also possessed a GPI anchor that was cleaved by GPI-specific phospholipase C during detergent lysis of the cells.
Cellular Copy Number of BARP-Using an approach similar to that employed for ISG 100 (22), an estimate of the cell copy number for BARP was calculated from the data presented in Fig. 6A using a calibration curve for the ELISA response constructed using the purified GST-BARP fusion protein. This analysis indicated that approximately 650 ng of BARP was recovered in the fractions eluted from the DEAE resin by the pH gradient. Because this material was obtained from a total of 1.1 ϫ 10 9 cells, and given a predicted molecular mass of 27.6 kDa for BARP, it can be calculated that each cell contains about 13,000 copies of the protein. This value probably underestimates the true copy number of BARP because the calculation assumes total recovery of the protein and ignores the contribution of the anti-GST antibodies in the calibration curve for the ELISA response. However, even allowing for these considerations it seems likely that BARP is not an abundant protein and is present at levels comparable to those reported for ISGs and receptors (11).
Cellular Localization of BARP-Antibodies against BARP did not react with live cells, probably because of shielding by the VSG as observed for other invariant surface proteins (20 -22), but the protein was detected by indirect immunofluorescent antibody staining of fixed cells (Fig. 7A). Confocal microscopy indicated that BARP was present in small discrete spots that were distributed over the entire cellular surface from the posterior end of the cell body to the tip of the flagellum. The punctate distribution of BARP clearly differed from the uniform surface distribution of the VSG (Fig. 7B) or the flagellar pocket/lysosomal localization of ISG 100 (Fig. 7C) (22). 15 of the brightest focal areas corresponding to BARP were selected to perform an area analysis using the Lasersharp processing software (Bio-Rad) after fitting each focal area to a multi-sided polygon. This analysis predicted an average surface area of 0.08 Ϯ 0.006 m 2 (mean Ϯ S.E.) for the BARP microdomains.
The punctate distribution of BARP and the small area of the microdomains were suggestive of sequestration of the protein in lateral heterogeneities or lipid rafts (40 -43). These microdomains are enriched in glycolipids and form detergent-insoluble complexes in cold non-ionic detergents that can be separated from the detergent-soluble fraction by floatation centrifugation on sucrose gradients (44). To determine whether BARP was also associated with similar microdomains in trypanosomes the cells were lysed in cold Triton X-100, and the lysates were subjected to floatation centrifugation on a step-sucrose gradient. Western blotting revealed the presence of BARP in the top fraction from the gradient, i.e. the detergent-insoluble glycolipid-enriched fraction, whereas both VSG (GPI-anchored) and ISG 70 (transmembrane) were essentially absent from this fraction and were recovered mainly in the intermediate and bottom fractions from the gradient (Fig. 8A). Cross-reaction between the anti-BARP antibodies and the VSG in these fractions was also apparent. Silver staining of these fractions demonstrated that most of the cellular protein was recovered in the detergent-soluble fraction, i.e. at the bottom of the gradient, but a faint series of bands migrating in the region expected for BARP was visible in the top and, to a lesser extent, in the intermediate fractions from the gradient (Fig. 8B). Together these data support the view that BARP may be associated with detergentinsoluble glycolipid-enriched microdomains.

DISCUSSION
This study has provided further evidence for the efficacy of heterologous expression of trypanosomal cDNAs in COS cells as a method for the identification of genes coding for surface proteins in trypanosomes. Interestingly, all of the cDNAs isolated so far by this technique appear to be life cycle stagespecific, a finding that is in agreement with an emerging view that this specificity is likely to be a feature of most externally disposed surface proteins in trypanosomes (see Ref. 11). In addition, none of the selected cDNAs coded for the most abundant surface protein in bloodstream forms, i.e. VSG, whereas two out of three of the clones examined from procyclic forms represented isoforms of PARP, the major surface antigen during this life cycle stage. This result probably reflected the fact that VSG was processed poorly by the pathways involved in the attachment of the GPI anchor and the subsequent trafficking of VSG to the surface membrane in COS cells (45) but suggested that these restrictions do not apply to the processing of PARP.
A new cDNA was isolated by this approach, and several results suggested that this cDNA codes for a bloodstream stage-specific, alanine-rich surface protein, termed BARP, which is probably attached to the plasma membrane by a GPI anchor. First, a surface location was an explicit requirement in the cloning strategy. Second, the deduced amino acid sequence of BARP contained the two key elements typical of several different GPI-anchored surface proteins in trypanosomes (11), the presence of an NH 2 -terminal leader sequence (35) and a COOH-terminal hydrophobic extension predicted to act as a FIG. 6. Partial purification of BARP. Panel A, a CHAPS extract of surface 125 I-labeled bloodstream forms of T. brucei was subjected to ion exchange chromatography on DEAE, and the fractions were analyzed for the presence of radioiodinated proteins (CPM, f) and BARP (ELISA: antibodies at 405 nm, q) (for details, see "Experimental Procedures"). Proteins that bound to the resin were eluted by the application of a pH gradient (E). Panel B, the majority of the heterogeneous material corresponding to BARP was recovered in two peaks as determined by ELISA; samples of the fractions (3-6 and 37-40) from these peaks were subjected to Western blot analysis using antibodies against BARP and the VSG (AnTat 1.1). Panel C, fractions from the peak eluted by the pH gradient were concentrated (for details, see "Experimental Procedures"). Samples of this material were mock-treated (lane 1) or treated with N-glycosidase F (lane 2) and then probed on Western blots with antibodies against BARP. This partially purified BARP was also probed with antibodies against the cross-reacting determinant of the VSG (lane 3).
signal sequence for the addition of a GPI anchor (36). Third, BARP was solubilized by extraction with CHAPS, but apparently not by cold Triton X-100 and could be separated from VSG by ion exchange chromatography because of its low pI. The partially purified protein reacted with antibodies directed against the cross-reacting determinant of VSG, an epitope generated by cleavage of GPI anchors by GPI-specific phospholipase C. A similar cross-reacting determinant positivity has been observed for the ESAG6 component of the heterodimeric transferrin receptor in T. brucei (4) and suggests that under appropriate conditions both of these proteins can be released from the plasma membrane by the action of GPI-specific phospholipase C. Finally, both Northern and Western blots indicated that expression of BARP was restricted to bloodstream forms of the parasite.
BARP appeared to share some homology with the major surface antigen from a different life cycle stage of a different trypanosome: the glutamate-and alanine-rich protein (GARP) of procyclic forms of T. congolense (37,38). The strict correspondence was only 35%, and both proteins were rich in alanine and glutamic acid residues, so it could be argued that the relationship was more apparent than real. However, both pro- teins share a number of other features. For example, both were predicted to have a similar size (27.6 and 26.5 kDa), acidic pI (4.6 and 4.8) hydrophobicity profile and were likely to be attached to the surface membrane by a GPI anchor. In addition, BARP and GARP display a highly heterogeneous mobility on SDS-PAGE with apparent molecular masses significantly higher than predicted from their amino acid sequences. Finally, the relative positions of the cysteine residues in both proteins were almost identical, and both were likely to possess almost identical patterns of secondary structure, consisting mostly of relatively long ␣-helical regions separated by short stretches of undefined or unfolded regions, as predicted by analysis of the sequences with the PHDsec program using a neural network-based approach on the EMBL server (46,47). This analysis also predicted that the pattern of folding of BARP was similar to that observed for the NH 2 -terminal domain of the VSG (48) as well as the heterodimeric transferrin receptor in bloodstream forms of T. brucei (49), i.e. an ␣-helical rod-like structure mounted normal to the plane of the plasma membrane. Such an organization is consistent with the view that surface proteins may be constrained to adopt configurations that are either similar to or compatible with the VSG (50).
In addition to BARP and GARP, two other surface proteins, PARP and ESAG6, migrate in a highly heterogeneous fashion on SDS-PAGE with apparent molecular masses significantly higher than predicted from their amino acid sequences (3)(4)(5). All of these proteins appear to be GPI-anchored, and this electrophoretic heterogeneity may be related to the associated GPI anchor. First, the in vitro translation products of BARP, GARP, and ESAG6 did not display any such heterogeneity (4,37), which suggested that this feature was caused by post-translational modifications. Second, the possibility that N-linked carbohydrate was solely responsible for this behavior seems unlikely because treatment of BARP with N-glycosidase F did not affect significantly the heterogeneous electrophoretic mobility of the protein, as was also the case for ESAG6 (4,5), whereas such considerations do not apply to GARP and one form of PARP, both of which lack N-glycosylation sites (3,37). Third, in the case of PARP, there is evidence that the presence of a sialylated polydisperse branched polylactosamine side chain attached to the GPI anchor may be, at least partially, responsible for the unusual migration of the protein on SDS-PAGE (51)(52)(53). Moreover, when expressed in procyclic cells, VSG also appeared to acquire a complex GPI anchor and exhibited a more heterogeneous electrophoretic mobility (54). Finally, the generation of chimeras of the ESAG6 and 7 components of the transferrin receptor has shown that the heterogeneity of ESAG6 on SDS-PAGE was related to the presence of the COOH-terminal hydrophobic extension that codes for the attachment of the GPI anchor (4). Indeed, exchange of the COOH-terminal regions of ESAGs 6 and 7 resulted in a concomitant exchange of this heterogeneous electrophoretic mobility. These considerations suggest that complexity of the anchor may be at least partly responsible for the electrophoretic heterogeneity of all of these GPI-anchored proteins.
Perhaps the most unusual feature of BARP was the clustering of the protein into small microdomains that were distributed over the entire cellular surface. Moreover, in contrast to VSG or ISG 70 , both which are distributed uniformly over the cellular surface, BARP was recovered in the detergent-insoluble glycolipid-enriched fraction when cells were subjected to Triton X-100 extraction experiments performed in the cold. These features are consistent with growing evidence for the existence of lateral heterogeneities or lipid rafts in eukaryotic cells (40,41,43) and the fact that these detergent-insoluble glycolipid-enriched microdomains appear to act as organizing centers for certain GPI-anchored proteins (42,(55)(56)(57). The mean surface area of the BARP microdomains (ϳ0.08 m 2 ) was commensurate with a diameter of approximately 300 nm, which places the BARP microdomains in the middle of the range of estimates for lipid rafts (100 -500 nM) (43). Jacobson and Dietrich (43) have estimated that rafts of this size could accommodate up to a maximum of ϳ600 proteins assuming a molecular mass of 50 kDa. From the confocal image (Fig. 7A) it can be calculated that each cell contains approximately 30 BARP microdomains, which in principle could give a total of 18,000 copies of BARP/cell, assuming full occupancy of each domain, which is close to the estimated copy number of the protein (ϳ13,000). Although lipid rafts have been best characterized in higher eukaryotes, recent results suggest that similar lipid microdomains may be present in widely separate phyla (58 -61). These considerations lead to the obvious speculation that BARP is clustered in lipid rafts in T. brucei and raises the interesting question of how this protein is constrained in sur- FIG. 8. BARP is present in the detergent-insoluble glycolipidenriched fraction isolated from T. brucei. Trypanosomes were subjected to Triton X-100 extraction in the cold followed by floatation centrifugation on a step-sucrose gradient as described under "Experimental Procedures." The gradient was fractionated by the sequential collection of 2-ml, 1-ml, and finally 2-ml fractions from the top to the bottom of the gradient. The proteins present in each fraction were precipitated and resuspended in SDS-PAGE sample buffer. Lanes 1, 2, and 3 correspond, respectively, to samples from the top, middle, and bottom of the gradient; thus, lane 1 corresponds to the low density detergent-insoluble glycolipid-enriched fraction. Panel A, Western blots of the fractions using antibodies against BARP, VSG, and ISG 70 . BARP was clearly visible as heterogeneous material migrating between 50 and 70 kDa in lane 1. Cross-reaction between antibodies against BARP and the VSG present in lanes 2 and 3 was clearly visible as was proteolysis of VSG and ISG 70 probably because of the absence of leupeptin in the lysis buffer. Panel B, silver staining of the fractions. The faint staining of a heterogeneous material similar to BARP is indicated by the arrowhead in lane 1. The majority of the proteins were recovered in the lower fraction as shown by the intensity of the silver staining in lane 3. In all cases the loading of the precipitated protein from each fraction was adjusted to give approximately 10 7 cell equivalents/lane. The track at the extreme left of the panel presents the silver stain pattern of the BenchMark™ amino acid protein ladder from Life Technologies, Inc. with the 50-and 20-kDa bands indicated. The loading corresponded to 50 ng/band. face microdomains in the presence of the uniformly distributed GPI-anchored VSG. One possibility may be that this distribution is related to differences in the GPI anchor of BARP compared with the VSG. Experiments are currently under way to test this proposal.