The GPI-Phospholipase C of Trypanosoma brucei Is Nonessential But Influences Parasitemia in Mice

The neo^R and hyg^R antibiotic resistance genes were constructed to a standard format. The procyclin mini-exon acceptor site and 5′ untranslated region (UTR) was followed by the coding sequence of the resistance gene and then the 3′ UTR and polyadenylation site of a T. brucei β-tubulin gene. The constructs were derived from pJF6 (Clayton et al., 1990). The chloramphenicol acetyl transferase gene was removed by HindIII–BamHI digestion and replaced, in a three-fragment ligation, by a HindIII–EcoRI fragment containing the coding sequence and an EcoRI–BglII fragment from the β-tubulin gene. The restriction sites were added to the coding sequence using PCR or adaptors.

The gene encoding GPI-PLC (PLC) and flanking DNA was subcloned from λBS2 (Carrington et al., 1989) and λBS3 as a 9.4-kb EcoRI–SalI fragment. This plasmid is referred to as p1172. An XbaI site was introduced downstream of the GPI-PLC gene, after the polyadenylation sites (see Fig. 1,a) by site-directed mutagenesis to give p1172x. The ILTar 1 serodeme (Miller and Turner, 1981) is heterozygous for an insertion type RFLP in the intergenic region upstream of the GPI-PLC gene (see Fig. 1,a). Since the constructs were originally designed to target this trypanosome line, both alleles were cloned as PCR products, and a KpnI site was introduced into each by site directed mutagenesis (see Fig. 1 a). These were subcloned back into p1172x as a ClaI–BstEII fragment, in place of the endogenous ClaI–BstEII fragment, to give two constructs, p1172kxL and p1172kxS containing the long and short RFLP allele, respectively. The trypanosome line actually used for the gene deletion (see below) is homozygous for the long RFLP.

The entire GPI-PLC gene was then removed from these plasmids as a KpnI–XbaI fragment and replaced with the neo^R expression cassette (p1172kxL) or the hyg^R cassette (p1172kxS) to give pLN and pSH, respectively. The restriction sites were added to the chimeric antibiotic resistance genes using PCR.

The PLC replacement construct (see Fig. 1 b) was designed to integrate a copy of the PLC gene into the tubulin gene cluster, which in T. brucei is comprised of an α and β tubulin gene pair tandemly arrayed 12 to 15 times. The coding sequence of the ble^R gene, including the termination codon, from pUT58 (CAYLA), was used to replace part of the coding sequence of an α tubulin gene from the initiation codon to the StuI site. In the adjacent β tubulin gene the entire coding sequence was precisely replaced with the GPI-PLC coding sequence. The encoded mRNA therefore consists of the GPI-PLC coding sequence with β tubulin 5′ and 3′ UTRs. Where necessary, restriction enzyme sites were added using PCR.

The procyclic trypanosome line used, pro G Anvers, was derived from the EATRO 1125 (Van Meirvenne et al., 1975) stock by recent passage through tsetse flies and kept in culture for only 40 d after isolation from the fly. Procyclics were cultured in SDM-79 (Brun and Schönenberger, 1979) with 10% vol/vol heat-inactivated fetal bovine serum. Bloodstream forms were grown in mice and purified from blood by DEAE-cellulose chromatography (Lanham and Godfrey, 1970). Immunosuppressed mice were prepared by X-irradiation (550–600 radons).

Preparation, gel electrophoresis, and blotting of nucleic acids were performed as described previously (Carrington et al., 1987). SDS-PAGE and Western blotting were performed using standard techniques. Primary antibodies were used at 1 to 5 μg/ml. The secondary antibodies were peroxidase-conjugated donkey anti–rabbit immunoglobulin or peroxidase-conjugated rabbit anti–mouse immunoglobulin and were used at the manufacturer's recommended concentrations (Jackson ImmunoResearch Laboratories, West Grove, PA). Detection was by chemiluminescence (Amersham, Intl., Arlington Heights, IL). Alternatively, secondary antibody was alkaline phosphatase-linked anti–rabbit immunoglobulin (Boehringer Mannheim, Indianapolis, IN), and the detection used a chromogenic substrate.

Anti–GPI-PLC was prepared by immunizing a rabbit with a λcro-β-galactosidase–GPI-PLC fusion protein. This was prepared by cloning the end-filled DraI–EcoRI fragment from pBS1, containing a GPI-PLC cDNA (Carrington et al., 1989), into the SmaI site of pEX3 (Genofit, Grand-Lancy, France). The fusion protein was insoluble and was separated from the other insoluble proteins by preparative SDS-PAGE. The antibodies were affinity purified on a glutathione-S-transferase–GPI-PLC fusion protein (gst-plc) attached to Sepharose beads. The gst-plc was made by cloning the entire GPI-PLC coding sequence into the EcoRI and BamHI sites of pGEX2T (Pharmacia Fine Chemicals, Piscataway, NJ), the restriction sites being added to the cDNA using PCR. The gst-plc was insoluble and was purified by preparative SDS-PAGE. The anti-CRD was a kind gift of Dr. Paul Englund (Johns Hopkins Medical School, Baltimore, MD; see Fig. 3 a) and Dr. Linda Allen (Cambridge University, Cambridge, UK; see other figures).

Procyclic trypanosomes were grown to a density of 6 to 8 × 10⁶ cells/ml, washed, and resuspended in ZPFM solution (Zimmerman, 1982) at a density of 4 × 10⁷ cells/ml. 300-μl aliquots were mixed with 15 μg DNA in 4-mm cuvettes and electroporated twice at 1.5 kV, 25 μF, 200 ohms in a gene pulser (Bio Rad, Hercules, CA). These cells were used to seed a 10-ml culture, and antibiotic selection was introduced 24 h after electroporation. For elimination of the PLC gene, cells were first transformed to G418 resistance, using pLN, and then the uncloned population transformed to hygromycin resistance, using pSH. Doubly resistant trypanosomes were cloned by dilution into 40% vol/vol fresh conditioned medium with antibiotic selection. A single PLC⁻ clone was converted to bloodstream form by transmission through tsetse flies. For PLC replacement, the same procyclic PLC⁻ was transformed to phleomycin resistance using pB and cloned as above before tsetse transmission.

SDS lysates were prepared by resuspending cells in SDS-PAGE sample buffer at 2 × 10⁸ cells/ml and immediately incubating at 100°C for 3 mins. Hypotonic lysates (Fig. 3 a) were prepared by resuspending cells in TES buffer (20 mM TES, pH 7.5, 140 mM NaCl, 5 mM KCl, 10 mM glucose, 1 mM EDTA, 1 mM PMSF, 50 μg/ml leupeptin) at 3 × 10⁹ cells/ml and then diluting to 2 × 10⁸ cells/ml with hypotonic dilution buffer (1 mM EDTA, 1 mM PMSF, 50 μg/ml leupeptin). The lysates were incubated for 5 min at 37°C. PI-PLC treatment involved incubation of hypotonic lysates with 4 U Bacillus cereus PI-PLC (Boehringer) per 1 × 10⁶ cells for 30 min at 37°C. In other experiments, as an alternative to hypotonic lysis, cells were lysed in 1% Triton X-100 in TES buffer at a concentration of 2 × 10⁸ cells/ml. The preparation of recombinant GPI-PLC will be described elsewhere. In all the experiments in this paper the GPI-PLC was added in sufficient quantities to completely hydrolyze the GPI anchor of the VSG in 5 min at 37°C.

These were performed using [³H]-myristyl VSG as substrate and counting the ³H released into an organic phase on phase separation (Bülow and Overath, 1986). 3 μg of VSG that contained 10,000 cpm ³H was used as substrate in a 200-μl reaction with 4 × 10⁶ cell equivalents of trypanosomes lysed in TES buffer containing 1% Triton X-100.

Trypanosomes were grown in immunosuppressed mice for 7 d, until the parasite population was predominantly stumpy in morphology. Differentiation was induced as described (Rolin et al., 1993); briefly, cells were innoculated at a density of 1 × 10⁶ cells/ml into modified DTM (Overath et al., 1986) with 15% heat-inactivated fetal calf serum and 3 mM citrate/cis-aconitate and incubated at 27°C in 4% CO₂ in air (Ziegelbauer et al., 1990).

FACS^® analysis was performed as previously described (Rolin et al., 1993). A mixture of three anti-sera (anti-AnTat 1.1, 1.2, and 1.3) was sufficient to detect the VSG using FACS^®. The anti-procyclin was the monoclonal antibody TPRI/247 (Cedarlane, Hornby, Ontario, Canada).

The different trypanosome stocks being tested were injected intraperitoneally (i.p.) into each of six matched CFLP mice using an innoculation of 1 × 10⁶ trypanosomes, unless otherwise stated. Parasite densities in subsequent infections were determined using an improved Neubauer haemocytometer after dilution of tail blood into 0.85% ammonium chloride. The minimum detectable density was 4 × 10⁴ trypanosomes/ml. Infections were monitored daily for the first 24 d and then 3–4 times a week until 90– 100 d after infection. Mice were removed from the study on exhibiting severe clinical symptoms such as inability to walk, hunched appearance, and labored breathing.

Mann-Whitney tests were used to compare parasitemias at the first peak and survival times between groups. Probability values were adjusted for ties where appropriate. A probability value of ⩽0.05 was taken to indicate a significant difference.

Trypanosomes were purified from blood and fixed in suspension with 4% paraformaldehyde in PBS for 10 min. After washing with PBS, the cells were allowed to settle onto polylysine-coated multiwell slides. The remainder of the procedure followed standard protocols; the mouse serum was diluted 1:500.

The approach taken to generate a GPI-PLC null mutant T. brucei cell line was targeted deletion of the gene by homologous recombination (Lee and Van der Ploeg, 1990; ten Asbroek et al., 1990; Eid and Sollner-Webb, 1991). There is a single PLC gene per haploid genome (Carrington et al., 1989), and since T. brucei is diploid, two successive deletions were performed using the selectable marker cassettes neo^R and hyg^R, coding for neomycin and hygromycin phosphotransferase, respectively. The targeting constructs were designed to affect removal of the entire PLC gene, including both 5′ and 3′ UTRs. The end points of the deletion were also chosen to minimize any interference with the expression of flanking genes, and the untranslated regions were removed to avoid any influence that they might exert on marker gene expression. The desired deletion resulted in the elimination of the entire PLC gene from before the dominant mini-exon addition site to beyond the polyadenylation sites, but left the processing sites of the flanking genes untouched. The gene targeting constructs are shown in Fig. 1 a, along with a map of the genome before and after a favorable recombination event. The determination of the RNA processing sites for the PLC and flanking genes will be published elsewhere. The deletions were performed in procyclic trypanosomes that do not express GPI-PLC and thus should present no phenotype and be at no selective disadvantage. Cells were subsequently induced to differentiate to a form that usually expresses the enzyme.

Before electroporation, constructs were linearized to increase the efficiency of stable integration and to favor a gene replacement (ten Asbroek et al., 1993). Trypanosomes were first transformed to G418 resistance and the uncloned population subsequently transformed to hygromycin resistance before clonal cell lines were established. Southern analysis of doubly resistant clones showed that the desired recombinational event occurred with high efficiency (14/16 events analyzed, data not shown). A Southern blot of BamHI-digested genomic DNA probed with 5′ flanking DNA (Fig. 1,a, probe B) is a way to distinguish between construct insertion and gene replacement. Such an analysis of the PLC⁻ clone selected for continued study is shown in Fig. 2, in comparison with wild-type EATRO 1125 proG and a heterozygous G418-resistant clone displaying a single PLC gene deletion. There is no hybridization of a PLC coding sequence probe to the PLC⁻ genomic DNA (Fig. 2). Absence of the PLC gene was also confirmed by PCR analysis (data not shown).

As expected, the PLC⁻ procyclic trypanosomes showed no gross alterations in morphology or growth rate. A single PLC⁻ clone was transmitted through tsetse flies. When trypanosomes were visible in saliva from a fly they were used to infect an immunosuppressed mouse. After further amplification in an irradiated mouse, stabilates were prepared.

Since PLC⁻ procyclic trypanosomes were able to differentiate to bloodstream forms it is evident that GPI-PLC activity is not necessary for these developmental steps. To confirm the absence of GPI-PLC activity, the cleavage of mfVSG to sVSG by endogenous enzymes was assayed in PLC⁻ and control cells. A Western blot was used to determine whether the CRD was exposed (Fig. 3 a). The assay would also detect any alternative activity, such as another phospholipase C or a D, capable of hydrolyzing the GPI anchor to expose the CRD.

On SDS lysis there is very rapid inactivation of GPI-PLC, and the reaction of the VSG with the anti-CRD is weak or nonexistent. Recognition of some mfVSGs has been observed previously and seems to depend on an epitope involving the galactose branch and ethanolamine phosphate bridge of the GPI anchor (Zamze et al., 1988). The varying degree of galactosylation in different VSG anchors may explain the differing reactivity of their intact form with anti-CRD antibodies. The difference in mfVSG electrophoretic mobility between the control and null cells (Fig. 3 a, compare lanes 1 and 4) suggests that the predominant VSG in the two populations is different. This is consistent with the slight recognition of the mfVSG in the SDS lysate of the PLC⁻, but not control trypanosomes.

On hypotonic lysis and incubation at 37°C, the VSG in the control trypanosomes but not PLC⁻ trypanosomes became strongly CRD positive. This confirms the absence of GPI-PLC activity in the null mutant and, furthermore, indicates that there is no alternative endogenous metabolic route to convert mfVSG to sVSG. The VSG in the PLC⁻ cells became CRD positive on incubation with exogenous B. cereus PI-PLC. This result indicates that the GPI anchor on the VSG in the PLC⁻ cells contains PI and the core oligosaccharide recognized by anti-CRD antibodies, suggesting that the anchor structure is normal.

Absence of GPI-PLC activity in the PLC⁻ trypanosomes was confirmed by assaying detergent lysates for their ability to release tritium from [³H]-myristyl VSG. Activity in the wild-type cells was 45 pmol/min/mg protein, which is similar to previously reported values (Bülow and Overath, 1986) and in null mutant cells was <0.5 pmol/ min/mg protein, the limit of reliable detection of the activity. It is not clear whether the removal of a single myristic acid from the GPI anchor by a phospholipase A would result in either VSG shedding or exposure of the CRD epitope. However, failure to release tritium into the organic phase during the standard assay argues against the presence of such a phospholipase in the PLC⁻ cells.

A further experiment was performed to investigate if any other modifications to the GPI anchor could be detected after lysis. In this case Triton X-100 lysis was used and the lysate incubated at 37°C for 20 or 40 min with or without added recombinant GPI-PLC. In the absence of added GPI-PLC, no anti-CRD binding was detected either at 20 or 40 min, similar to the previous result. Additionally, if the lysate was incubated first for 20 min in the absence of added GPI-PLC and the enzyme then was added for the second 20 min of the incubation, the same strength of anti-CRD signal was obtained (Fig. 3 b). This result means that, in the absence of GPI-PLC gene, no modifications occurred to the GPI anchor that would affect hydrolysis by GPI-PLC, such as acylation, hydrolysis by gluconases, or hydrolysis by a phospholipase D.

It has already been reported that, unlike wild-type, PLC⁻ trypanosomes do not release sVSG in stress conditions (Rolin et al., 1996).

On ingestion by tsetse flies, bloodstream trypanosomes differentiate to procyclic forms, a process that can be mimicked in vitro (Ziegelbauer et al., 1990). Since the PLC⁻ mutant was first generated as a procyclic cell line, this differentiation would complete the life cycle. Bloodstream-form cells were harvested from irradiated mice 7 d after infection when the population was predominantly stumpy in morphology and hence predisposed to synchronous differentiation (Ziegelbauer et al., 1990). A plot of cell density against time after cells were induced to differentiate is given in Fig. 4. The control trypanosomes displayed a 15-h lag before exponential growth of procyclic forms commenced. This is similar to results previously reported with stumpy-form AnTat 1.1 cells (Ziegelbauer et al., 1990; Rolin et al., 1993). The lag and subsequent growth rate of the PLC⁻ trypanosomes were indistinguishable from the control and are within the normal range expected for this cell line. The distinct morphological changes that occur in wild-type cells during this time (Rolin et al., 1993) were also apparent during differentiation of the PLC⁻ cells (data not shown).

FACS^® analysis (Roditi et al., 1989; Ziegelbauer et al., 1990) was used to follow the kinetics of appearance of cell surface procyclin and loss of VSG in the PLC⁻ trypanosomes. In preliminary experiments, the VSG expressed by the nonclonal PLC⁻ trypanosomes was investigated by immunofluorescence. Of 12 antibodies tried (anti-AnTat 1.1 to anti-AnTat 1.12), 4 (AnTat 1.1, 1.2, 1.3, and 1.6) were found to react weakly with some cells in the population. The strength of the fluorescence indicated that the VSG(s) expressed by the PLC⁻ trypanosomes was slightly cross-reactive with these antibodies as opposed to actually being one of these VSGs. A mixture of three of the antisera (AnTat 1.1, 1.2, and 1.3) was sufficient to produce a signal for the VSG that was detectable on FACS^® analysis. Fig. 5 shows that surface VSG has been completely replaced by procyclin between 7 and 9 h after initiation of differentiation. Once again, this is consistent with previous studies on this cell line (Rolin et al., 1993). Furthermore, the transient activation of adenylate cyclase, which follows VSG release during differentiation, occurs normally in the PLC⁻ trypanosomes (Rolin et al., 1996). Finally, when bloodstream forms were fed to tsetse flies, PLC⁻ trypanosomes established a typical procyclic infection in the midgut (data not shown). This indicates that the essentially normal differentiation observed in vitro also occurs in vivo.

The initial stabilates of the bloodstream form of PLC⁻ trypanosomes were produced in immunosuppressed mice and so had not interacted with the host immune system. To establish a persistent infection in an immunologically competent mouse, trypanosomes have to exchange the metacyclic VSG for a bloodstream form VSG, and this occurs in the majority of the population (Esser and Schoenbechler, 1985). Subsequently a small subset of the population undergoes antigenic variation to avoid antibody-mediated lysis. To test whether the VSG-shedding properties of GPI-PLC play an essential part in these two processes, infections were established in laboratory mice. Persistent infections lasting >50 d were readily observed with the PLC⁻ bloodstream-form population. Parasitemic profiles for six individual mice for the first 50 d after infection are given in Fig. 6. Each mouse shows the same basic pattern that is in turn broadly similar to the profiles observed previously for six cloned T. brucei stocks (Turner et al., 1995). An initial peak of parasitemia is followed by a relapse and recrudescence (‘trough') to a comparatively level plateau, punctuated by one or more further troughs.

Although it is barely credible that such an infection could persist without antigenic variation, two experiments were performed to show that VSG switching occurs in the PLC⁻ trypanosomes. A clonal line of PLC⁻ trypanosomes was first established by infecting immunosuppressed mice with single trypanosomes, and stabilates were made 10–14 d later. Three mice were then infected using a single stabilate. The first mouse was used to prepare trypanosomes for immunofluorescence during the first peak of parasitemia. The second mouse was used to prepare serum just after the first peak of parasitemia had declined, and the third mouse was used to prepare trypanosomes for immunofluorescence during the second peak of parasitemia. The results are shown in Fig. 7 a; the serum clearly recognizes trypanosomes in the first peak, but not in the second.

In the second experiment, trypanosomes from the first and a subsequent peak were used to produce detergent lysates. The triton lysates were then treated with recombinant GPI-PLC before Western blotting with anti-CRD (Fig. 7 b). Both populations are expressing readily detectable VSG(s) that have different profiles on densitometric analysis.

The course of chronic infection of mice with control trypanosomes was monitored in parallel with the PLC⁻ infections described above. Passage through a tsetse fly can greatly alter the growth characteristics of a trypanosome cell line (Turner, 1990). Therefore, the most comparable trypanosome population available was used as a control in infections. This population had been derived from the same procyclic cell line as the PLC⁻ mutant and was transmitted through tsetse flies in parallel, but had a chloramphenicol acetyl transferase-hygromycin^R gene construct integrated into the tubulin gene cluster (Berberof et al., 1995). Since isolation from tsetse flies, both of the trypanosome stocks had been amplified twice in immunosuppressed NMRI mice and stabilates prepared from the second mouse. These stabilates were used to infect immunosuppressed CFLP mice, and blood from these mice was used to establish persistent infections in matched sets of six CFLP mice. The innoculum for control and PLC⁻ trypanosomes was matched as closely as possible; both were collected while the parasitemia was rising and contained a preponderence of long slender and intermediate forms as opposed to stumpy forms. The differences between the two parasitemic profiles are illustrated in Fig. 8, which shows log₁₀ geometric mean parasite density through time. Although each infection within a group gave the same basic profile, slight variation in the timing of peaks and troughs in individual mice means that the second and subsequent troughs are evident only by large standard error bars.

A parasitemia can be described in terms of the parameters that provide the column headings in Table I (Turner et al., 1995). Such characterization provided a clear demonstration that different trypanosome stocks are intrinsically different in the parasitemias they produce in chronic infections (Turner et al., 1995), and the same approach has been adopted here to compare PLC⁻ with control trypanosomes. The results are shown in Table I a, which also summarizes the results of two additional, parallel, infections with PLC⁻ trypanosomes. These were designed to investigate whether prolonged growth in mice and innoculum size altered the parasitemia. Syringe passaging of laboratory trypanosome lines through mice is a standard method for artificially increasing the level of parasitemia achieved during infection. Thus two sets of six mice were infected with PLC⁻ trypanosomes that had undergone 22 additional syringe passages in immunosuppressed CFLP mice. One set of mice was infected with 1 × 10⁶ trypanosomes as above and a second set with 1 × 10⁷ trypanosomes.

The PLC⁻ trypanosomes show a more than threefold decrease in first peak parasite density that was significantly different (P = 0.01) compared to control cells. In this experiment, PLC⁻ trypanosome density never reached 1 × 10⁸ cells/ml during the first 20 d of infection and for most of the time the levels were <1 × 10⁷ cells/ml. The control cells showed a higher overall parasitemia, >1 × 10⁸ cells/ml for an average of 17.5% and <1 × 10⁷ cells/ml for 32.5% of the first 20 d. Repeatedly passaging the PLC⁻ trypanosomes resulted in an increased parasitemia in the first 20 d, although all parameters indicated that the parasitemia was still lower than with the control trypanosomes and still statistically significant in the case of the lower inoculum dose (P = 0.01).

The survival time can be taken as an indication of the virulence of an infection. The mean survival time of mice infected with the control trypanosomes was 28 d, and was significantly (P = 0.01) lower than in PLC⁻ trypanosome infections, where the mean survival time was 81 d, and three of the six mice were still alive at the end of the experiment on day 99. Passaging and innoculum size of the PLC⁻ trypanosomes had no significant effect on the survival time of the infected mice despite the higher overall level of parasitemia compared with the unpassaged PLC⁻ trypanosomes. Survival time remained significantly different from control infections at both inoculation size (P = 0.05 and 0.01 for 10⁶ and 10⁷ trypanosomes, respectively).

The results above clearly showed that mice infected with PLC⁻ trypanosomes had an altered parasitemic profile when compared to the best available control. To address the question as to whether this alteration was due to the absence of the PLC gene, or some other event, a copy of the PLC gene was returned to the PLC⁻ trypanosome. It was decided to place the PLC gene in the tubulin locus (to produce PLC⁺ trypanosomes) as opposed to returning it to the endogenous position, as the latter manipulation could conceivably repair any lesion in a flanking gene. The targeting construct employed ble^R as a selectable marker and is shown in Fig. 1. Southern analysis of BamHI digested DNA from one phleomycin resistant clone is shown in Fig. 2 and shows that the PLC probe hybridises to a 6.6-kb band, as expected after the desired integration into the tubulin locus.

Phleomycin-resistant clones were transmitted through tsetse flies, and the expression of GPI-PLC characterized by Western blotting and assaying mfVSG to sVSG conversion on hypotonic lysis. The PLC⁺ trypanosomes were found to contain GPI-PLC protein but, by comparison with a twofold dilution series of AnTat 1.1 cells (Fig. 9), the level of GPI-PLC in 2 × 10⁶ PLC⁺ trypanosomes was roughly equivalent to that in only 1.25 × 10⁵ AnTat1.1 cells. Thus the PLC⁺ trypanosomes contain approximately one sixteenth of the wild-type, bloodsteam-form GPI-PLC level. The presence of GPI-PLC activity was tested in two assays. The first was detergent lysis of PLC⁺ trypanosomes; in this assay the endogenous VSG became anti-CRD positive (Fig. 10). In the second assay, lysates of PLC⁺ and wild-type trypanosomes were assayed in vitro for release of tritium from [³H]-myristyl VSG. The PLC⁺ trypanosome lysates contained an activity that released 15.5 pmol/min/mg protein or 31% of wild-type activity. There is a discrepency between the estimates of the amount of GPI-PLC protein (5–10% of wild-type) and the activity (30% of wild-type). At this stage it is not clear whether this is an artefact of the enzyme assay or due to some endogenous regulatory phenomenon.

Matched sets of six mice were infected with either PLC⁻ or PLC⁺ trypanosomes and the parasitemias monitored over 93 d. The data are summarized in Table I b. The parasitemia and mouse survival data for the PLC⁻ parasites in the two experiments were slightly different as might be expected, but statistical comparison of parasitemias at first peak and survivorship showed no significant differences (P = 0.30 and P = 0.47, respectively). Low level GPI-PLC expression coincides with an increase in the parasitemia over the first 20 d, with a significant increase in first peak parasite density compared to the PLC⁻ infections (P = 0.01). In addition there is a downward trend in mean survival time to values intermediate between the PLC⁻ and wild-type trypanosomes, although this is not statistically significant.

The function of the GPI-PLC in T. brucei has been the cause of much speculation, as the enzyme has an activity that could theoretically facilitate the rapid shedding of the protective VSG coat. A GPI-PLC null mutant has been made by targeted gene deletion to test whether the ability of the GPI-PLC to hydrolyze the GPI anchor of the VSG, which is observed on both lysis and stress, is essential in vivo. The generation of the mutant was facilitated by the developmentally regulated expression of the PLC gene, as the deletion could be performed in a life cycle stage that does not express GPI-PLC. The success of the deletion strategy is in accord with previous findings that integration into the genome is the preferred mode of stable transformation in T. brucei and that integration is exclusively by homologous recombination. The high efficiency with which the desired recombinational event took place also supports the observation that constructs in which a drug-resistance marker is flanked by noncontiguous homologous sequences tend to lead to homologous replacement of the intervening sequences in the genome (Cruz and Beverley, 1990; Eid and Sollner-Webb, 1991; ten Asbroek et al., 1993).

The major find presented in this paper is that PLC⁻ trypanosomes are capable of completing an entire life cycle. PLC⁻ mutants were generated as procyclic trypanosomes and then subjected to cyclical transmission through tsetse flies. The ability of cells isolated from the salivary glands of tsetse flies to establish a chronic infection, at least in mice, suggests that the enzyme is inessential for the mechanism of switching from metacyclic to bloodstream VSG expression, or for the mechanism of antigenic variation. Finally, PLC⁻ bloodstream trypanosomes have been demonstrated to differentiate into procyclic trypanosomes both in vitro and in vivo with apparently normal kinetics.

An absence of any redundant activity that could substitute for the GPI-PLC had previously been suggested by quantitative immunoprecipitation of the mfVSG to sVSG activity in trypanosome lysates using a GPI-PLC–specific monoclonal antibody (Bülow and Overath, 1986). The inference that GPI-PLC is the only hydrolase catalyzing this conversion has been confirmed by the absence of any mfVSG-cleaving activity in PLC⁻ trypanosomes. This in turn suggests that GPI-PLC is indeed responsible for the low rate of CRD-positive VSG shedding that occurs from living trypanosomes (Bülow et al., 1989; Seyfang et al., 1990), and this could be confirmed if it becomes possible to establish PLC⁻ trypanosomes in culture. The function of this low rate of VSG shedding remains unclear. The viability of the GPI-PLC null mutant indicates that it is not an essential process in VSG turnover, although it may well affect its kinetics. If the steady turnover of surface VSG in wild-type cells (Seyfang et al., 1990) depends in part on the action of GPI-PLC, coat exchange in the PLC⁻ trypanosomes might be expected to take longer and involve a longer period of double VSG expression. The possible importance of this in determination of parasitemic profile is highlighted by a certain mathematical model of antigenic variation (Agur et al., 1989). Although a decrease in efficiency of coat exchange may contribute to the change in overall infection dynamics exhibited by the mutant PLC⁻, it cannot explain certain characteristics such as the reduced height of first parasitemic peak.

In addition to the observed hydrolysis of the VSG anchor, it has been suggested that GPI-PLC may play a role in GPI metabolism or that its primary function may be in PI hydrolysis. The experiments in this paper allow some observations to be made although they were not primarily designed to address this subject.

Cleavage of the VSG from PLC⁻ trypanosomes by bacterial PI-PLC producing CRD-positive VSG strongly suggests that the GPI anchor structure is normal and argues against a role for GPI-PLC in GPI biosynthesis. Also, preliminary analysis of the GPI biosynthetic pathway in PLC⁻ trypanosomes indicates that the normal intermediates are present (Ferguson, M., and M.-L. Güther, personal communication).

However, GPI-PLC may yet play a part in the kinetics of the biosynthetic pathway, perhaps in the catabolism of excess GPI precursors since flux through the pathway is apparently 10-fold in excess of absolute protein anchoring requirements (Masterson and Ferguson, 1991). Such a catabolic role may involve cleavage of one or more of the cytoplasmically disposed GPI precursors in the cytoplasmic leaflet of the ER and would have to be tightly controlled to preserve sufficient GPI to form VSG anchors. When GPI-PLC is expressed in an heterologous system (Leishmania major) any such regulation is absent and cells secrete gp63 in a CRD-negative form because it never receives a GPI anchor (Mensa-Wilmot et al., 1994). It is not clear to what extent these results are applicable to T. brucei.

In T. brucei, there is a dynamic equilibrium between the GPI-PLC-sensitive end-product of the biosynthetic pathway that is attached to the VSG (glycolipid A) and an acylated, GPI-PLC-resistant, form (glycolipid C; Güther et al., 1994; Güther and Ferguson, 1995). Glycolipid C is not an obligatory GPI precursor (Masterson and Ferguson, 1991) and, unlike most GPI intermediates, is predominantly localized in the lumenal leaflet of the ER (Vidugiriene and Menon, 1994). The fact that it is kept both chemically and topologically secure from GPI-PLC attack is intriguing. The role of glycolipid C is unclear, although it may act as a store of GPI anchors and hence as a buffer against fluctuating demands for GPI (Güther et al., 1994). Preliminary analysis of the GPI biosynthetic pathway in PLC⁻ trypanosomes indicates that there is no gross alteration in the glycolipid A to C ratio, or accumulation of excess GPI (Ferguson, M., and M.-L. Güther, personal communication). Although we cannot rule out a role for GPI-PLC in catabolism of GPIs in vivo, it is clear that such a role is not essential.

Any potential role of the GPI-PLC in PI metabolism is more difficult to discuss in the absence of much knowledge about signal transduction in trypanosomes. It is not possible, at this stage, to rule out a non-essential role for the GPI-PLC involving PI hydrolysis. The loss of such a function in PLC⁻ trypanosomes may be rescued by a redundant activity that does not metabolise GPI.

The second finding is that PLC⁻ trypanosomes have an altered parasitemia in laboratory mice. The main differences between the PLC⁻ and control trypanosomes are the degree of parasitemia in the early stages of a chronic infection and the survival time of infected mice. However, before an attenuated phenotype can be firmly assigned to PLC⁻ trypanosomes it should be confirmed in independently derived mutants, which are not yet available. Data presented in this paper are based on the behavior of one clone only. It should also be noted that the natural hosts for T. brucei are mammals in sub-Saharan Africa and not laboratory mice; the course of infection with PLC⁻ trypanosomes in indigenous cattle may provide a more clearly defined phenotype. In addition, confirmation should come from a full phenotype rescue, which may only be possible if PLC⁻ trypanosomes can be modified to express GPI-PLC at wild-type levels. The possible partial rescue observed in PLC⁺ trypanosomes is however encouraging.

There is variability in the parasitemia caused by different isolates of T. brucei, but any one isolate produces a consistent course of infection (Turner et al., 1995). The progress of the initial stages of parasitemia (first 20 d) observed in both control and PLC⁻ trypanosomes are within the range reported for these different trypanosome isolates (Turner et al., 1995). However, compared to the control trypanosomes, the initial stages of parasitemia are reduced in PLC⁻ trypanosomes, an effect that is partially reversed in PLC⁺ trypanosomes. How could the absence of GPI-PLC affect the initial parasitemia? Either the growth rate of the PLC⁻ trypanosomes is reduced by an elongation of the cell cycle time and/or increased rate of differentiation to the nondividing stumpy forms. Alternatively the PLC⁻ trypanosomes could be killed more efficiently by the host immune system. One possible mechanism for the latter is that on immune lysis wild-type trypanosomes release sVSG into the blood of the mouse which then acts as a decoy to anti-VSG immunoglobulins as they are produced. It is worth noting that blood with 1 × 10⁸ trypanosomes/ml contains roughly 100 μg/ml of VSG.

The second clear difference between control and PLC⁻ trypanosomes is the survival time of the infected mouse, which was unaffected by passage number and innoculum size. Although survival time showed a downward trend in PLC⁺ infections compared to PLC⁻ infections, the change was not statistically significant. It remains to be seen whether PLC⁻ trypanosomes can be modified to express wild-type levels of GPI-PLC and, if so, whether this would result in a complete return to wild-type virulence. The survival time of PLC⁻ infected mice was apparently longer than those previously observed for a range of isolates (Turner et al., 1995), although these values were obtained in an inbred mouse strain (BALB/c as opposed to CFLP in this study). The survival time, which can be used as a measure of virulence, correlates with the initial parasitemia (Turner et al., 1995), and will be influenced by the genotype and epigenetic state of the trypanosome and the host immune system (Barry and Turner, 1991). Using the data available, it is not possible to provide a simple explanation of how the absence of GPI-PLC activity might reduce the virulence of trypanosomes. However, it is known that T. brucei causes immunosuppression in the host (Askonas, 1985) and this must be a component of virulence. If GPI-PLC activity is necessary for the immunosuppression then it might be expected that PLC⁻ trypanosomes would be less virulent than the equivalent wild-type strain.

We would like to thank Peter Overath for encouragement and his helpful comments on the manuscript, Mike Turner for his help with the manuscript and the statistics, and Lucia Cardoso de Almeida for discussions and encouragement.

CRD

cross-reacting determinant

GPI

glycosylphosphatidylinositol

mf

hydrophobic membrane form

PI

phosphatidylinositol

PLC

phospholipase C

VSG

variant surface glycoprotein