A number of features of antigenic variation that operate at the level of the trypanosome population in the host prolong infection (11). VSG switching is divergent, and the classical course of infection includes a series of peaks of parasitemia, each peak usually consisting of a mixture of variable antigen types (VATs) (49, 68). The scope for variation is not infinite, and it is believed that trypanosomes have a finite VSG repertoire (4, 52) which, however, may be capable of temporary augmentation through the generation of mosaic genes, by recombination, during infection (3). In a process that may contribute substantially to the prolongation of infection, the repertoire is expressed in a hierarchical fashion, meaning that some VATs have a tendency to appear early in infection, some only during late infection, and the rest are expressed somewhere in between (4, 16, 32).

Key to these features is the way in which the parasite controls VSG expression. The basis of antigenic variation is the trypanosome’s estimated repertoire of more than 1,000 distinct VSG genes (VSGs). Most of these are located inside chromosomes, but there is also a large minority at telomeres of large chromosomes and minichromosomes (22, 67, 69). VSGs are expressed individually, presenting a complex gene control problem. This is solved partly by transcription being possible only at specific telomeric loci, known as bloodstream expression sites (BESs), which are polycistronic transcription units under the control of a strong promoter (37, 57). There are multiple BESs in the genome, of which only one is active at a time. Antigenic variation proceeds by VSGs sequentially coming to occupy BESs, and up to six different means of achieving this have been described (5, 13, 56). One, known as the in situ mechanism, operates by transcriptional switching between BESs, effecting a switch between the VSGs that happen to occupy those sites. Four mechanisms appear to be based on homologous recombination. Two of these involve the duplication of a silent gene into a BES and the concurrent deletion of the VSG resident at that site. Individually, these mechanisms are termed duplicative transposition (if the silent gene is located within a chromosome) and telomere conversion (if silent gene is located at the telomere), although both can be referred to as duplicative transposition. Reciprocal recombination involves the exchange of VSG sequences between two chromosomes, and mosaic gene formation involves the splicing together of segments of VSGs or pseudogenes. The sixth mechanism proposed, point mutagenesis (25), is thought to arise under conditions of laboratory selection (31). The majority of VSGs, which reside within chromosomes, can be activated only by duplication into a BES.

All these mechanisms have been observed in “monomorphic” trypanosome lines, which are adapted to continual growth in the laboratory, have lost the ability to develop through their life cycles, and have low VSG switch rates, estimated to be 10⁻⁶ to 10⁻⁷ switch/trypanosome/generation (38). These low rates have been interpreted as meaning that trypanosome antigenic variation occurs via events that are part of background genetic noise. However, “pleomorphic” lines, not adapted in this way, switch VSGs several orders of magnitude more rapidly, at an overall rate of around 10⁻² to 10⁻³ switch/trypanosome/generation (the range, within a clone, is approximately 10⁻² to 10⁻⁵) (64, 66). To account for the difference in rate between the two types of line, we have proposed that pleomorphic trypanosomes either employ a general mechanism for VSG switching that becomes diminished in monomorphic lines or apply a dedicated switch mechanism that is diminished in, or even absent from, monomorphic lines. As most VSGs are located within chromosomes, we have further proposed that this most active mechanism is based on duplicative transposition (5).

The best approach to identify the most active switch mechanisms is by the analysis of single relapses, in which an infection with a single variant relapses to a second peak from which clones are isolated and their activation mechanisms are identified. If the analysis is undertaken much later in infection, the host’s accumulating set of antibodies against different VSGs will be removing the most readily activated VSGs from the study. For monomorphic trypanosomes, single-relapse analysis has shown that about two-thirds of switch events occur by the in situ mechanism, and about one-third occur by duplicative transposition (42), and it is only later that duplicative events come to dominate (40, 47, 63). To test whether pleomorphic trypanosomes instead have a preference for duplicative transposition, we undertook a comprehensive analysis of switching in the first two relapse peaks in a single rabbit infection with a pleomorphic, fly-transmissible trypanosome line. As initial in vitro and in vivo experiments showed that the most highly switching lines are too unstable for thorough analysis, we concentrated on a line from the more stable end of the pleomorphic switch rate range. We find that duplicative transposition, most frequently involving silent telomeric VSGs, is the predominant switching mechanism.

Infections and antibody methods. Two pleomorphic bloodstream clones were derived from a Trypanosoma brucei EATRO 795 line that is tsetse fly transmissible. The ILTat 1.2 expressor clone switches VSG at about 10⁻⁵ switch/trypanosome/generation, and the ILTat 1.61c clone, derived from a single metacyclic trypanosome, switches at about 3 × 10⁻² switch/trypanosome/generation (64). Trypanosomes were grown from stabilate in an ICR mouse (Bantin & Kingman, Hull, United Kingdom) that had been immunosuppressed by cyclophosphamide treatment (Sigma, Ltd.) (25 mg · kg of body weight⁻¹) 24 h previously. Exsanguination was performed at the initial parasitemic peak by cardiac puncture into 5% sodium citrate anticoagulant in Carter’s balanced salt solution (27). Approximately 10⁶ ILTat 1.2 trypanosomes were injected intravenously into a lop-eared rabbit (designated “A”) (Bantin & Kingman) from which preinfection plasma had just been collected. The infection was allowed to progress for 30 days, and blood was sampled daily for measurement of parasitemia, preparation of plasma, and isolation of trypanosomes. Three additional infections were initiated with trypanosomes grown from the same stabilate in the same way, except that the inoculum had been treated with antibodies against all 11 VATs characterized in this study (see below). Two of these infections were in lop-eared rabbits (designated “B” and “C”), and the third was in a New Zealand White rabbit (designated “D”) (Bantin & Kingman).

On each day of the infection, 0.6 ml of blood was taken from the rabbit and injected immediately, intraperitoneally, into an immunosuppressed ICR mouse. These “amplifier” mice were then bled when the parasitemia reached a level above approximately 10^7.8 trypanosomes · ml⁻¹ (34), and trypanosome clones were isolated immediately by micromanipulation. All clones were stabilated in liquid nitrogen as soon as possible after the parasitemia reached patency. This approach to cloning trypanosomes and preparing specific antisera was based on methods developed for rapidly switching Trypanosoma vivax that minimize growth period, and therefore the extent of antigenic variation, and also permit rapid raising of specific antisera by cymelarsen (5 mg · kg body weight⁻¹) curing of the mice from which the clones had been derived (4).

To test trypanosome VAT, immune lysis was undertaken. Trypanosomes were suspended in guinea pig complement (Harlan) to 5 × 10⁶ · ml⁻¹ and were incubated in 1:20-diluted antiserum or plasma for 1 h at room temperature, after which percentage lysis was quantified by light microscopy, counting at least 100 trypanosomes. Indirect immunofluorescence on acetone-fixed thin blood films was performed as previously described (66).

RNA isolation and cDNA generation. The trypanosome clones were grown from stabilate in immunosuppressed ICR mice, and mRNA was isolated from 100 μl of blood with TRIzol (Gibco-BRL). cDNA was generated by reverse transcription (RT), and RT-PCR was performed exactly as previously described (17). PCR products were resolved on a 0.7% agarose gel and were cloned with the T-vector (Promega) or PCRScript Amp SK (+) (Stratagene) plasmids. Automated sequencing was performed on the plasmids and directly from the PCR product by the Molecular Biology Support Unit of the University of Glasgow.

Genomic DNA isolation, Southern blotting, and hybridization. Genomic DNA was prepared as described previously (12), digested with EcoRI, HindIII, or PstI (New England Biolabs), fractionated on a 0.7% agarose gel, Southern blotted onto nylon membrane (Micron Separations, Inc. or Hybond-N; Amersham International), and bound by UV irradiation. The blots were then probed with the VSG-specific cDNA, which had been excised from the plasmid, α-³²P radiolabelled with the Prime-It II kit (Stratagene), and washed to a final stringency of 0.1× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) in 0.1% sodium dodecyl sulfate at 65°C.

Pulsed-field gel electrophoresis. Genomic plugs were prepared by immobilizing live trypanosomes (5 × 10⁷ trypanosomes in 100-μl plugs of 0.7% low-melting-point agarose) and treating them with 1 mg of proteinase K · ml⁻¹ (Sigma). Gels were run on the CHEF-DR III system (Bio-Rad) using one-half of a genomic plug per lane. The 6-day general separations were run on a 1.2% agarose gel at 15°C in 0.089 M Tris-borate–0.1 mM EDTA (85 V, 1,400 to 700 s pulse time, 144 h). The 1.8-Mb region was resolved under the same conditions, except that a fixed pulse time of 600 s was used. The minichromosomal separations were run on a 1.0% agarose gel at 14°C in 0.045 M Tris-borate–0.5 mM EDTA (200 V, 20 s pulse time, 16 h).

As is usual in rabbits, the parasitemia in rabbit A over the observed 30 days was at a low level, but it yielded two visible relapse peaks, on days 7 to 9 and 11 and on days 18 to 22. Daily injections of rabbit blood into mice, starting at day 6, showed that the rabbit blood was infective on days 6 to 11, 16 to 22, and 28 to 30, corresponding to three relapse peaks. Cloning of trypanosomes from all these amplifier mice (except those injected with the day 29 and 30 rabbit samples) yielded a total of 36 clones from the first relapse peak, 36 from the second, and 5 from the third. As it was subsequently found that there were only two VATs in the 36 clones from the second peak, trypanosomes grown in mice from the stabilates of the day 17 and 21 amplifier mice were treated with antisera against those two VATs and were cloned. This led to the isolation of another six trypanosome clones from the day 17 sample and eight clones from the day 21 sample, representing two new VATs. The day 19 amplifier mouse trypanosomes were then treated in the same way with antibodies against the four VATs from the second peak, yielding five more clones, all of the same VAT. In all, 96 trypanosome clones were derived from the three peaks. Clones from within each of the first two relapse peaks were then compared by immune lysis with the antisera derived from the mice in which the clones were isolated. Within the first relapse peak, six different VATs were represented in the 36 clones, and five distinct VATs were present in the second relapse peak (Table 1). By immunofluorescence with existing reference antisera, and in some cases confirmed subsequently by DNA sequencing, 6 of these 11 VATs were recognized as previously identified VATs: ILTats 1.21, 1.22, 1.23, 1.25 (49), and 1.64 (same as GUTat 7.13) (6, 19) and ETat 1.7, described in the VAT repertoire of the distinct, but related, EATRO 3 stock (46). ILTats 1.22 and 1.64 are also expressed in the metacyclic population (19, 29). For the new, unrecognized VATs, we have assigned the codes ILTat 1.67 to 1.72.

TABLE 1 Appearance of VATs and antibodies in rabbitA

To check that the isolated VATs had been activated in the rabbit rather than in the amplifier mice, each was tested by immune lysis against the daily serum samples from rabbit A; specific antibodies should appear only after the corresponding VAT was present in the rabbit. This was found to be the case for all VATs (Table 1). To analyze the reproducibility of the appearance of VATs in different infections, the same ILTat 1.2 infecting population was grown from stabilate and then treated with antisera against all 11 VATs isolated from rabbit A in order to remove existing switch products. After washing, the surviving trypanosomes were injected into rabbits B, C, and D. Daily (or on alternate days for rabbit D) serum samples were harvested and tested by immune lysis against the collection of VATs (Table 1). This revealed that ILTats 1.25, 1.67, 1.68, and 1.69 appeared early in infection in all rabbits, ILTats 1.21 and 1.70 appeared in all infections with a greater spread in time of appearance, and ILTats 1.64, 1.23, 1.71, 1.22, and 1.72 did not appear in all rabbits. With the exception of ILTat 1.64, all the VATs of the first relapse peak elicited antibodies in all rabbits, four with a consistent timing in all animals. There was more variation in timing for the second relapse peak VATs.

VSG-specific cDNA clones were derived for all VATs by RT-PCR. As trypanosomes are variable in their expression of VSGs, the use of universal primers in this method can yield more than one VSG cDNA product, and this was the case for ILTat 1.67. Usually the different products can be distinguished from each other by inherent differences in the lengths of their coding sequences. To verify that cloned products did correspond to the major VSG, for all VSG expressors we compared lengths of cDNA clone inserts with those of their corresponding major, single, RT-PCR products. We also partially sequenced all the major RT-PCR products directly and verified sequence identity with the cDNA plasmid clones. By these criteria, all the specific VSG-encoding sequences were cloned, and all contained the complete VSG-coding sequence. The sequences of at least 400 bases from the ends of all clones were determined. BLAST searching revealed a typical VSG pattern: general conservation at the downstream end of the coding sequence, reflecting peptide sequence conservation, and considerable variation at the amino-terminal-coding end, which contains the variable epitopes of the VSG. For ILTats 1.21, 1.22, 1.23, and 1.25, our clones perfectly matched the existing database sequences in the amino-terminal-coding region, confirming the serological results. The GUTat 7.13 (ILTat 1.64) and ETat 1.7 (ILTat 1.69) VSG sequences have not been determined previously.

To determine the number of genes corresponding to each cDNA and the gene activation mechanism used in each case, genomic DNA was isolated from a representative trypanosome clone of each VAT and intact, agarose-embedded genomic DNA plugs were prepared for analysis by pulsed-field gel electrophoresis. Genomic DNAs were digested with HindIII or EcoRI, fractionated by gel electrophoresis, blotted, and probed with full-length VSG cDNAs. This revealed that the cDNA for the infecting VSG, ILTat 1.2, hybridized to four fragments representing two genes (there is a HindIII site within the coding sequence), as previously reported (70). Two fragments comigrate at 1.2 kb, and the two telomere-oriented fragments are at 6.8 to 7 kb (see Fig. 4B and D, lanes 1). In the following, the VATs are described essentially according to their sequence of appearance in the rabbit. Several features typical of VSGs are apparent in the autoradiographs presented in Fig. 1, 3, and 4. For ILTat 1.21, there is a single fragment, probably corresponding to a single gene copy, in the DNA of the infecting VAT, ILTat 1.2, and in variants which are not expressing ILTat 1.21 (Fig. 1B, lanes 1, 2, and 4). This fragment varies in length between trypanosome clones which indicates that the ILTat 1.21 gene is telomeric, the variation probably being due to changes in the length of the telomere tract lying immediately downstream of the VSG. In the trypanosome clone expressing ILTat 1.21, there is an extra fragment (Fig. 1B, lane 3), which corresponds to a classical expression-linked copy (ELC) of the gene, the product of duplicative transposition. The extra band is not due to partial digestion, because the same filter was reprobed with other VSG cDNA probes and all lanes yielded single bands in at least one probing. Reprobing with a PCR product encompassing the trypanosome RAD51 coding sequence also revealed a single band (data not shown). The conclusion is that there is a single, silent copy of the ILTat 1.21 VSG, located at a telomere, and it was activated by duplication into a BES. For ILTat 1.69, a similar result was obtained, except that there are apparently two silent, telomeric gene copies (Fig. 1A, lanes 1, 3, and 4). When activated, an ELC fragment appeared (Fig. 1A, lane 2). The ILTat 1.64 VSG, previously characterized as being single copy and telomeric (19), also displayed an ELC when activated (Fig. 1C, lanes 1 to 4).

FIG. 4 Activation of the ILTat 1.68 VSG. ILTat 1.68 and 1.2 VSG cDNA probings of pulsed-field gel electrophoresis-separated chromosome-sized DNA and HindIII-digested genomic DNA. (A) Genomic DNA from ILTats 1.2 (lane 1), 1.25 (lane 2), 1.67 (lane 3), and 1.68 (more ...)

FIG. 1 Activation of the ILTat 1.69, 1.21, and 1.64 VSGs. Genomic DNAs from ILTats 1.2 (lane 1), 1.69 (lane 2), 1.21 (lane 3), and 1.64 (lane 4) were digested with HindIII (panels A and B) or EcoRI (panel C), separated in 0.7% agarose gels, and Southern (more ...)

FIG. 3 Activation of the ILTat 1.25 VSG. (A) Genomic DNA from ILTats 1.2 (lane 1), 1.25 (lane 2), 1.67 (lane 3), and 1.68 (lane 4) was digested with HindIII, separated in a 0.7% agarose gel, Southern blotted, and probed with ILTat 1.25 VSG cDNA. (B) (more ...)

The chromosomal locations of the various VSGs were next examined by pulsed-field gel electrophoresis. Samples from trypanosomes expressing ILTats 1.2, 1.69, 1.21, and 1.64 were fractionated on the gel shown in Fig. 2A, and a Southern blot was prepared and then probed sequentially with the cDNA fragments for each of these VSGs. One of the two ILTat 1.2 genes is located on a 1.8-Mb chromosome, and the other is on a 2.1-Mb chromosome (Fig. 2B). When expressed, there is an ELC present on a 1.8-Mb chromosome (Fig. 2B, lane 1). It is notable that, in the variants expressing the ILTat 1.69, 1.21, or 1.64 VSGs, the ILTat 1.2 ELC has disappeared (Fig. 2B, lanes 2 to 4). The ILTat 1.69 silent copies reside on a large (slot) chromosome and a minichromosome (Fig. 2C). When expressed, an ELC has appeared on a 1.8-Mb chromosome (Fig. 2C, lane 2). (The size of the slot chromosome hinders its transfer to the filter, but in lane 2, some of the 1.8-Mb chromosome has remained in the slot and has transferred more efficiently. A similar 1.8-Mb chromosome effect is apparent in Fig. 2D.) The single-copy ILTat 1.21 VSG is on a minichromosome (Fig. 2D), and the expressor has an ELC on a 1.8-Mb chromosome (Fig. 2D, lane 3). The ILTat 1.64 VSG is on a 2.3-Mb chromosome (Fig. 2E), and the expressor has an ELC on a 1.8-Mb chromosome (Fig. 2E, lane 4). Combining all these data, the ILTat 1.69, 1.21, and 1.64 VSGs were activated by ELC formation, and each ELC chased an ILTat 1.2 VSG from its bloodstream expression site. Another gene activated by ELC formation encodes the ILTat 1.25 VSG, which was the earliest to appear as a switch product in rabbit A. There are two ILTat 1.25 VSGs, both of which are telomeric. These are visible at about 5.8 and 9.2 kb in Fig. 3A, lanes 3 and 4, along with some fainter, cross-hybridizing bands. The ELC is visible in the lane with DNA from the expressor (Fig. 3A, lane 2). Due to the presence of the cross-hybridizing sequences, it is difficult to interpret location in pulsed-field gel electrophoresis, although Fig. 4B, lane 2, reveals that activation of ILTat 1.25 was accompanied by loss of an ILTat 1.2 gene from the 1.8-Mb chromosome. We also investigated the activation of ILTat 1.25 in a very highly switching trypanosome, the ILTat 1.61c clone. In a mouse, this clone switched to ILTat 1.25, which we cloned, and its DNA is compared with ILTat 1.61c DNA in Fig. 3B. There is only one silent ILTat 1.25 VSG in ILTat 1.61c (Fig. 3B, lane 1), and activation yielded an ELC (Fig. 3B, lane 2). Cross-hybridizing bands present in the ILTat 1.2 trypanosome clone are also present in ILTat 1.61c.

FIG. 2 Transposition of duplicated VSGs. (A) Chromosome-sized DNA of ILTats 1.2 (lane 1), 1.69 (lane 2), 1.21 (lane 3), and 1.64 (lane 4) was separated by pulsed-field gel electrophoresis in a 1.2% agarose gel at 85 V, with a ramped pulse frequency of (more ...)

The ILTat 1.67 VSG displayed a different means of activation. By Southern analysis of HindIII-digested DNA, it appears to be a single-copy gene, and this pattern was retained in DNA from the expressor trypanosome clone (data not shown). This resembles, therefore, an in situ switch. Supporting this observation, pulsed-field gel analysis revealed that the expressor had retained both ILTat 1.2 genes (Fig. 4B, lane 3) and had not generated an extra copy of ILTat 1.67 (data not shown), so we conclude that the switch involved the silencing of the BES containing the ILTat 1.2 gene and the activation of another containing the ILTat 1.67 gene. The ILTat 1.68 VSG also displayed a partially different activation pattern. In genomic Southern analysis, its cDNA probe revealed two fragments that can be interpreted as a single, telomeric gene and a chromosome-internal gene (or a second telomeric gene with restriction sites downstream, yielding a fragment of constant length regardless of variations in telomere tract length) (Fig. 4E). Pulsed-field gel analysis shows that these are probably located on a minichromosome and a large chromosome that did not migrate from the gel slot under the electrophoretic conditions used (Fig. 4C). In these blots, the telomeric and minichromosomal bands hybridized more strongly and are likely to be the gene that gave rise to the transcribed gene copy. The expressor showed an unexpected pattern. The minichromosomal copy was lost (Fig. 4C, lane 4), and a new copy appeared on what seems to be the same 1.8-Mb chromosome as observed for the ILTat 1.69, 1.21, and 1.64 ELCs (Fig. 4C, lane 4), and the ILTat 1.2 gene on that chromosome had indeed disappeared (Fig. 4B, lane 4). No extra band appeared on the blot of digested DNA, and the only difference observed upon activation was that the telomeric fragment migrated further than in nonexpressors (Fig. 4E). Together, these observations resemble ELC formation, with a new ILTat 1.68 VSG copy chasing the ILTat 1.2 VSG from its expression site on the 1.8-Mb chromosome, but there seems also to have been another event leading to loss of the minichromosomal 1.68 VSG. Alternatively, there could have been a reciprocal recombination between those two chromosomes, followed by loss of the minichromosome.

Figure 4 also shows the ILTat 1.2 genes, each represented as an upstream 1.2-kb HindIII fragment and as a telomeric fragment at 6.8 to 7 kb in the ILTat 1.2 DNA sample (Fig. 4D, lane 1). The loss of one gene is apparent in the ILTat 1.25 and 1.68 expressors (Fig. 4D, lanes 2, 4), but both ILTat 1.2 genes are retained in the ILTat 1.67 trypanosome (Fig. 4D, lane 3), which had undergone an in situ switch.

In the second relapse peak, all five VSGs that were activated were found to use the ELC mechanism (data not shown). Four include telomeric genes, although digestion with EcoRI or HindIII revealed that ILTat 1.71 is a single-copy, apparently chromosome-internal, gene, and two of the three ILTat 1.72 genes may also be internal by the same criteria. We have observed that the ILTat 1.23 VSG has one copy on a large chromosome and another on a minichromosome (data not shown), and ILTat 1.72 also has a minichromosomal gene copy. To summarize all the activation data, of six genes switched on in the first peak, four formed ELCs, another possibly used this mechanism in conjunction with another event, and one was activated in situ. All five in the second peak formed ELCs. Telomeric VSGs were the predominant source of ELCs, and several, including the first to be detected in switching, originated from minichromosomes. We subjected ILTat 1.2 DNA to pulsed-field gel electrophoresis under conditions in which the minichromosomes fractionated, revealing several size classes (data not shown). Probing a Southern blot of this gel showed that ILTat 1.25, 1.68, 1.69, and 1.21 genes are present on individual minichromosomes of 100, 80, 90, and 55 kb, respectively (data not shown). All the gene study data are summarized in Table 2.

TABLE 2 VSG loci and activationevents

T. brucei has a considerable task in expressing its VSGs singly and in executing a switch in expression between individual genes. By restricting expression to BESs, and allowing only one BES to be fully active at any time, the task is reduced to the mechanisms by which individual genes gain residency in, and subsequently are removed from, an active BES. Many of the pioneering studies that have revealed mechanisms for VSG switching have been aided greatly by the much-diminished VSG switch rate of the standard, monomorphic trypanosome lines (38), which have been selected by continual syringe passaging and normally display only the replicative, long, slender bloodstream stage (reviewed in references 5 and 11). The consequential near homogeneity in VAT has allowed the purification and characterization of individual VSGs and their cDNAs and the unambiguous identification of gene-switching events. It is likely that the drop in overall switch rate, by 4 to 5 orders of magnitude, during laboratory adaptation is a general phenomenon, as a high level of switching has been observed in nonadapted lines of a number of different stocks (7, 8, 41), and rapid syringe passaging of pleomorphic, cloned lines routinely leads to both monomorphism and a low level of switching (7, 10, 29, 41, 64, 65). A similar difference has been reported after passaging of Borrelia (36). The present work has been undertaken to test our proposal that pleomorphic trypanosomes are geared mainly for duplicative transposition of VSGs (5).

To minimize the effect of VSG switching, so that we could isolate a reference collection of individual VATs and corresponding antisera, we applied trypanosome-handling techniques that minimize period of growth and allow rapid preparation of specific antisera (4). The most highly switching clones are too unstable for an extensive analysis of this type, and we found that the combination of high switch rate and differentiation to the nonproliferating stumpy stage, the end product of pleomorphism, prevents analysis in vitro. Hence, most of this work has used a pleomorphic, fly-transmissible line (expressing ILTat 1.2) at the lower end of the natural switch range displayed by pleomorphic trypanosomes. An added benefit of this clone is that it was the basis of an extensive, phenotypic study of VAT switching patterns in vivo (49), and a closely related clone has been studied in a similar way following fly transmission (8), providing a large background data set to the present molecular analysis. The VATs we have isolated in the first two relapse peaks include four of the six isolated as primary switch products by Miller and Turner (49), who found that ILTat 1.21 (underlining denotes VATs observed also in our study) was the most frequent product of switching (in 32 of 41 single relapses in rats) followed by 1.4 (33 of 47), 1.25 (27 of 41), 1.26 (14 of 35), 1.24 (15 of 41), 1.22 (6 of 41) and 1.23 (2 of 36). Barry and Emery (8) found that 1.21 and 1.69 were activated very early in all goats examined following fly transmission of EATRO 795; the other bloodstream ILTat VATs in the present study were not examined. In related trypanosome stocks (isolates), the ILTat 1.69 (ETat 1.7), 1.21 (ETat 1.9), and 1.22 (GUTat 7.1, ETat 1.2) VATs have been identified, and all have been shown to be activated early in infection (6, 46). ILTats 1.22 and 1.64 (GUTat 7.13) are also expressed in the metacyclic population in the tsetse fly (6, 19), and expression of the former has been studied in considerable detail (7–9, 30, 44).

It is clear, therefore, that we are analyzing some of the most frequently activated VSGs and that activation is achieved by duplicative transposition, in support of our model. As the ILTat 1.21 and 1.25 VSGs are at the head of the activation hierarchy (49), and as they are single-copy genes (ILTat 1.25 in the 1.61c trypanosome clone), they reveal preferred, productive events in VSG switching: minichromosomal, telomeric genes act as templates for duplication into the BES that is transcriptionally active. We are now cloning these two loci from a minichromosomal library for a detailed analysis of their preferred donor status. Telomeric VSGs dominated as donors, five (ILTats 1.25, 1.21, 1.69, 1.23, and 1.70) being on minichromosomes or large chromosomes, two (ILTats 1.64 and 1.22) being on large chromosomes, and two (ILTats 1.68 and 1.72) appearing to have telomeric and internal copies. Only in the second relapse peak did we detect what may have been a chromosome-internal locus acting as a donor (ILTat 1.71), but we need to map with more restriction enzymes to rule out the possibility that this is a telomeric gene with EcoRI and HindIII sites lying immediately downstream. Preferential use of telomeric genes in early duplicative switches has been reported before (42, 51, 55, 71), but most studies have followed series of switches, rather than single relapses, and have involved monomorphic lines. The prominence of minichromosomal telomeres over telomeres of larger chromosomes as donors may be due to their relative abundance in the genome (67, 69) rather than any inherent feature, although there appears also to be a hierarchical use within this subset. The reason for the preferred use of telomeres may lie in a general interactivity of chromosome ends and the fact that telomeric VSG loci share more sequence with each other than with the internal genes of the chromosome. In other organisms, telomere interactions are now being seen as having important functions, including chromosome alignment in meiosis (18, 53), and the interactions are related to the juxtaposition of telomeres at the nuclear periphery and the presence of telomere-associated proteins (39).

There was only one example (ILTat 1.67) of in situ activation. This seems not to have had a major influence in the infection, as only two ILTat 1.67 VAT clones were isolated, and the activated BES was not used as a recipient of ELCs later in the complex, first-relapse peak. It may be that in situ switching is not common in pleomorphic trypanosomes, although it should be noted that ILTat 1.67 was activated early in all the rabbits, and that other examples of this mode of activation have been observed in fly-transmitted trypanosomes (24). The repeated use of the active BES as a recipient of duplicated genes may be because its transcriptionally active state makes it a more-accessible target for recombinational events. However, we have looked only at switch products that are expressed, and there could also be duplicative transpositions into silent BESs, as observed by Myler et al. (50). The typical duplication pattern was not displayed by the ILTat 1.68 VSG, which appeared instead to have used duplication accompanied by another event. The most simple explanation is that it was copied into the active BES, converting the ILTat 1.2 gene, and there followed a loss of the entire ILTat 1.68 minichromosome or donor gene, although it is also possible that reciprocal exchange (61) had occurred between the BES and the minichromosome, followed by loss of the latter.

Programmed DNA rearrangements in eukaryotes, such as immunoglobulin gene segment splicing, mating type switching in Saccharomyces cerevisiae, and chromosomal fragmentation in ciliates, operate through specific sequences (1, 33, 35). VSGs have common flanks that are often associated with duplicative transposition. Upstream are a set of imperfect, 70- to 76-bp repeats (2, 15, 43, 62, 69). At their downstream end, VSGs have similarities reflecting carboxy-terminal amino acid conservation, there are conserved sequences in the 3′ untranslated region of the mRNA (17, 43, 47), and telomeres share a greater extent of sequence further downstream (22, 69). The limits of duplicative transposition events often map to these conserved sequences, and the process probably involves homologous recombination, not only because it resembles gene conversion, but also because disruption of RAD51 leads to a substantial decrease in duplicative switching (45a). The relatively high level of switching in pleomorphic trypanosomes may be due to general homologous recombination operating preferentially at these sequences or due to the action of a sequence-specific enzyme that initiates the process. In contrast, the lower level of VSG switching in monomorphic trypanosomes could be due to a decrease in, or loss of, either some aspect of the general homologous recombination machinery or a putative initiating activity, resulting in a dependence on background homologous recombination. Some evidence for this comes from the examination of the upstream limit of duplication events. In the few cases examined in pleomorphic trypanosomes, the upstream limit maps to the 70-bp repeats (24, 44, 62), whereas in monomorphic trypanosomes, just more than half of reported events map there (15, 23, 28, 40), the remainder mapping to sequences of probably fortuitous similarity between the donor VSG and the recipient BES (26, 40, 48, 58). Additionally, the deletion or reversal of the 70-bp region of a BES in monomorphic trypanosomes had no effect on the incidence of VSG duplications into that BES (45). There are precedents for the possible subversion of enzymes for specific roles in gene rearrangements, with immunoglobulin gene splicing apparently having exploited a transposase and the mating-type switch using a homing endonuclease, and an activity specific to the trypanosome 70-bp repeats has been proposed several times (5, 54, 59). Another possibility for VSG duplication, which accounts for the observed imprecision in the upstream limit of VSG duplication events in monomorphic trypanosomes, invokes nonspecific DNA strand breaks as initiating the gene duplication process in the VSG 5′ flank (14).

This work was supported by the Wellcome Trust, the Swedish Society for Medical Research and the Swedish Institute (NB), and an MRC studentship to N.R. J.D.B. is a Wellcome Trust Principal Research Fellow.

We are grateful to Richard McCulloch and Michael Fotheringham for discussions and critical readings of the manuscript.