Biosynthesis and Function of the Modified DNA Base β-d-Glucosyl-Hydroxymethyluracil in Trypanosoma brucei

Abstract

β-d-Glucosyl-hydroxymethyluracil, also called J, is a modified DNA base conserved among kinetoplastid flagellates. In Trypanosoma brucei, the majority of J is present in repetitive DNA but the partial replacement of thymine by J also correlates with transcriptional repression of the variant surface glycoprotein (VSG) genes in the telomeric VSG gene expression sites. To gain a better understanding of the function of J, we studied its biosynthesis in T. brucei and found that it is made in two steps. In the first step, thymine in DNA is converted into hydroxymethyluracil by an enzyme that recognizes specific DNA sequences and/or structures. In the second step, hydroxymethyluracil is glucosylated by an enzyme that shows no obvious sequence specificity. We identified analogs of thymidine that affect the J content of the T. brucei genome upon incorporation into DNA. These analogs were used to study the function of J in the control of VSG gene expression sites. We found that incorporation of bromodeoxyuridine resulted in a 12-fold decrease in J content and caused a partial derepression of silent VSG gene expression site promoters, suggesting that J might strengthen transcriptional repression. Incorporation of hydroxymethyldeoxyuridine, resulting in a 15-fold increase in the J content, caused a reduction in the occurrence of chromosome breakage events sometimes associated with transcriptional switching between VSG gene expression sites in vitro. We speculate that these effects are mediated by the packaging of J-containing DNA into a condensed chromatin structure.

In the DNA of kinetoplastid flagellates, a fraction of thymine (Thy) is replaced by the modified DNA base β-d-glucosyl-hydroxymethyluracil (β-Gluc-HOMeUra), called J (13, 23, 55). In all kinetoplastids, J is abundantly present in telomeric repeats (55). In the parasite Trypanosoma brucei, J has also been found in other repetitive sequences such as the 50-, 70-, and 177-bp repeats and the spliced-leader RNA gene repeats (references 56 and 57 and unpublished data). The presence of J in the 70-bp repeats and variant surface glycoprotein (VSG) gene of the telomeric VSG gene expression sites correlates with silencing of these polycistronic transcription units (56). T. brucei lives within the bloodstream of its mammalian host and periodically changes its VSG coat to avoid destruction by the host immune response, a process called antigenic variation (1, 10, 12, 17). The presence of J in inactive telomeric VSG gene expression sites suggests that it might be involved in the transcriptional repression of the VSG gene expression sites and thereby might play a role in antigenic variation. This suggestion is supported by the developmental regulation of J synthesis found in T. brucei but not in other kinetoplastids (55). In the DNA of bloodstream form (BF) T. brucei, approximately 0.3 to 1% of thymine is replaced by J. In contrast, the procyclic form (PF) trypanosomes, which are present in the tsetse fly and which do not undergo antigenic variation, completely lack J (23, 55). The synthesis of J in trypanosomes stops when the BF starts to differentiate into the insect PF and the existing J is diluted out by DNA replication (6).

The specific distribution in T. brucei (4, 41, 56, 57) suggests that J is made at the DNA level by enzymes that recognize a silent chromatin structure and modify thymine in specific sequences. We hypothesized that synthesis of J involves first conversion of Thy into HOMeUra and then conversion of HOMeUra into β-d-Gluc-HOMeUra (Fig. 1) (11). This predicts that HOMeUra would be an intermediate in J synthesis, and, indeed, Gommers-Ampt et al. (22) have already noted that BF trypanosomes contain more HOMeUra than do PF trypanosomes. Subsequent work supported the hypothesis that this extra HOMeUra is involved in J biosynthesis since immunoaffinity enrichment for J-containing BF DNA enriched not only for J but also for HOMeUra, showing that HOMeUra was preferentially present in the DNA segments containing J (55). To test whether HOMeUra in DNA is a precursor in the biosynthesis of β-d-gluc-HOMeUra, we have grown trypanosomes in the presence of the nucleoside hydroxymethyldeoxyuridine (HOMedU). We found a large increase in the J content upon incorporation of HOMedU into trypanosome DNA.

FIG. 1.

Putative biosynthetic pathway for J. First, a thymidine (dT) residue in a certain context in DNA is converted into HOMedU by a DNA thymidine-7-hydroxylase. Second, HOMedU in DNA is converted into β-d-glucosyl-HOMedU (dJ) by a β-glucosyl transferase. BrdU is a thymidine analog that cannot be converted into dJ.

In an attempt to also reduce the J content, we tested the effect of other thymidine analogs on J biosynthesis. In organisms that contain 5-methylcytosine, the cytidine analog 5-azacytidine inhibits DNA methylation. When incorporated into DNA, 5-azacytidine irreversibly binds and thereby inactivates the DNA methyltransferase (31, 49, 50). Treatment of cells with 5-azacytidine results in reactivation of repressed endogenous genes or silent retroviruses (30, 33, 46). Methylcytosine has not been found in trypanosomes and other kinetoplastid flagellates (55). We have tested the effect of bromodeoxyuridine (BrdU) and other halogenated thymidine analogs on the synthesis of J in T. brucei. The bromide group of BrdU replaces the 5-methyl moiety of Thy, and BrdU can therefore not be converted into J (Fig. 1). BrdU is usually efficiently incorporated into DNA (reviewed in reference 38), and we show here that it can be used to reduce the J content of trypanosome DNA. The ability to modify the level of J in the trypanosome genome provided us with a new opportunity to test ideas about the possible function of this base. We have now examined whether varying the amount of J has an effect on the way in which VSG gene expression sites are controlled.

MATERIALS AND METHODS

Trypanosomes.

The trypanosomes were all T. b. brucei 427 (16). PF cells were grown in a semidefined medium (14), and BF cells were grown in HMI-9 medium (28) without thymidine or Serum Plus and with 20% fetal calf serum. The following cell lines were used: PF wild type (WT) (5) and PF TKN (PF cells transfected with a herpes simplex virus thymidine kinase-thymidylate kinase [TK] gene) (53) and the BF clones 221a or MiTat 1.2 (2), 3174 (contains a neomycin phosphotransferase [NEO] gene upstream and a hygromycin phosphotransferase [HYG] gene downstream of the 70-bp repeats in the active 221 expression site) (36), RP2XR (or RP2X-1R1; contains a HYG gene downstream of the promoter of the inactive 221 expression site) (47), HTK3 and HTK16 (contain a TK gene and a HYG gene downstream of the active 221a expression site promoter) (19), HNR (contains a HYG gene in the active 221 expression site and a NEO gene in the inactive VO2 expression site), and HN1 (contains a HYG gene in the inactive 221 expression site and a NEO gene in the active VO2 expression site) (15). HOMedU, BrdU, 5-iodo-2′-deoxyuridine (IdU), 5-chloro-2′-deoxyuridine (CldU), and 5-amino-2′-deoxyuridine (amino-dU) were purchased from Sigma. For incorporation of nucleoside analogs into DNA, BF cells were seeded at a density of 2.5 × 10³ per ml in the absence or presence of thymidine analogs and were harvested when the culture had grown to 1 × 10⁶ to 2 × 10⁶ cells per ml. When non-TK BF cells were used, no substantial effect on growth was found with concentrations of up to 1 mM HOMedU or 100 μM BrdU. Only at high concentrations of HOMedU (5 mM), BrdU (250 μM), or IdU (750 μM) was a reduction of the growth rate observed; the reduction was dependent on the exact growth conditions used. PF cells were seeded at a density of 1 × 10⁵ cells per ml and harvested at a density of 1 × 10⁷ to 2 × 10⁷ cells per ml. No effect on the growth rate of PF WT cells was observed with concentrations of up to 1 mM BrdU or 1 mM HOMedU.

DNA and RNA analysis.

Total genomic DNA was isolated as described previously (3). Digested or sonicated DNA was transferred to nitrocellulose by standard procedures (48). Probes were labeled with [α-³²P]dATP by random priming. A 5′ ³²P-labeled oligomer consisting of five telomeric GGGTTA repeats was used to probe for telomeric repeats. Probe fragments for β-tubulin genes, 50-bp repeats, RIME, VSG VO2, VSG 221, and HYG, have been described previously (56, 57). For isolation of poly(A)⁺ RNA, cell pellets were frozen in liquid nitrogen and poly(A)⁺ RNA was recovered with oligo(dT)₂₅ Dynabeads (Dynal). The cells were lysed as specified by the manufacturers, and 175 μl of beads was used per sample of 10⁸ cells. RNA was resuspended in 20 μl of H₂O. A 2- to 5-μl volume of RNA was mixed with loading buffer, consisting of 50% formamide and 6% formaldehyde in 1× MOPS buffer (20 mM MOPS, 5 mM sodium acetate, 1 mM EDTA [pH 7]), and heated at 80°C for 4 min. The gels were run for 1 h at 5 V per cm, and RNA was transferred to nitrocellulose by standard procedures. All the blots were washed at a final stringency of 0.3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)–0.1% sodium dodecyl sulfate at 65°C. Dot blots and Northern blots were scanned and quantitated by phosphorimager analysis (Fujix BAS 2000). Anti-J DNA immunoblottings and anti-J immunoprecipitations were performed as described previously (55, 56). ³²P-nucleotide postlabeling combined with two-dimensional thin-layer chromatography was done as described previously (22, 54). Briefly, DNA was digested to 3′-monophosphates, which were 5′-end labeled and subsequently 3′-end dephosphorylated. Chromatograms were scanned and nucleotide spots were quantitated by phosphorimager analysis. The recovery of dJ by ³²P-postlabeling is partial and varies with the batch of nucleases used; postlabeling of synthesized standards has shown that the labeling efficiency of dJ is on average 50% (54). Here the quantitation of dJ was corrected for incomplete recovery by postlabeling.

Negative selection of HTK trypanosomes for expression site switch variants in vitro.

The in vitro expression site switching experiments were done essentially as described previously (19). HTK cells were routinely cultivated in the presence of 20 μg of hygromycin per ml to prevent switching to another expression site. For switching experiments, the cells were washed in medium containing no hygromycin and used to inoculate two to five independent fresh cultures at a density of 2.5 × 10³ cells per ml, again containing no hygromycin but now in the absence or presence of 1 mM HOMedU (which had no effect on the growth rate of HTK cells) or 100 μM BrdU (which reduced the growth rate of HTK cells depending on the culture conditions used). Cells were harvested when the cultures had grown to 1 × 10⁶ to 2 × 10⁶ cells per ml and were distributed over 96-well plates at 10⁴ per well in 150 μl of medium. Negative selection against TK activity was performed by adding 20 μg of the nucleoside analog 1-[2-deoxy-2-fluoro-8-d-arabinofuranosyl]-5-iodouracil (FIAU) per ml or 5 μM nucleoside analog (E)-5-(2-bromovinyl)-2′-deoxyuridine (BVDU; Sigma). After 6 to 7 days, clonal outgrowth of the wells was scored and FIAU/BVDU-resistant (BVDU^r) trypanosome clones were tested for their sensitivity to 20 μg of hygromycin per ml (Hyg^s). Cell lines displaying a BVDU^r Hyg^s phenotype were then immediately expanded in vitro for preparing DNA. On the basis of Hyg^s, it was possible to discriminate between HTK revertants which arose due to mutation of the TK gene and those which had inactivated the 221 expression site. In trypanosomes where the TK gene is mutated, the HYG gene in the 221 expression site is still transcribed, resulting in resistance to hygromycin. Therefore, cells displaying a BVDU^r Hyg^r phenotype were discarded. Since growth in the presence of BrdU affected the survival of cells after plating, all the clones were also plated without negative selection to determine the plating efficiency, which was used to calculate the absolute switching frequency for each condition. BVDU^r Hyg^s clones were analyzed by dot blot hybridization to check for the loss of marker genes and other expression site sequences. Clones with a faint hybridization for the marker genes (present only in cells grown in the presence of BrdU and fewer than 10% of the total switchers) were regarded as polyclonal lines and were therefore excluded from further analysis.

RESULTS

HOMeUra is a precursor of J.

To test whether HOMeUra is an intermediate in the synthesis of J, we cultured trypanosomes in vitro in the presence of HOMedU and analyzed the nucleotide composition of the DNA by ³²P-postlabeling combined with two-dimensional thin-layer chromatography (2D-TLC). We first tested the levels and toxicity of this nucleoside in PF trypanosomes, which do not contain J (Fig. 2B). To optimize the incorporation of HOMedU into DNA, we used PF trypanosomes transfected with a TK gene for efficient conversion of HOMedU into HOMedUTP (53). The levels of endogenous HOMeUra measured in the DNA of PF trypanosomes are usually low and variable, and we attribute these small amounts to a cytidine deaminase activity that contaminates one of the enzymes used for postlabeling. Deamination of dC results in the formation of dU, which comigrates with HOMedU under the 2D-TLC conditions used. When PF trypanosomes were grown for 3 days in the presence of 1 mM HOMedU, the TK-expressing PF cells (TKN) contained on average 7 mol% of this analog in their DNA (Fig. 2C) and PF WT cells contained 0.4% (Table 1). The high levels of HOMedU obtained in TKN cells compared to PF WT cells show that the endogenous TK enzyme phosphorylated HOMedU less efficiently than did the viral TK enzyme. Interestingly, the incorporation of HOMedU into PF DNA resulted in the synthesis of J, showing that HOMedU can be a precursor in the synthesis of dJ (Fig. 2C), even in cells that usually do not make J. Moreover, incorporation of HOMedU and synthesis of up to 0.28% J (Table 1), which is two- to threefold the level found in BF trypanosomes, did not affect the growth rate or morphology of PF cells.

FIG. 2.

Analysis of thymidine analogs incorporated into trypanosome DNA by ³²P-nucleotide postlabeling combined with 2D-TLC. (A) Schematic representation of the positions of the deoxynucleotides (solid circles), including dJ (large arrow), HOMedU (small arrow), and BrdU (arrowhead). The ribonucleotides (open circles) and unknown products (dashed open circles) that contaminate the DNA, enzyme, and label preparations varied per experiment or batch; ribonucleotides were more abundant in small-scale DNA preparations (E and F). Trypanosomes were grown in the presence or absence of thymidine analogs. (B to F) Chromatograms representing WT PF trypanosomes (B), TKN PF trypanosomes plus 1 mM HOMedU (C), WT BF trypanosomes (D), HTK BF trypanosomes plus 1 mM HOMedU (E), and HTK BF trypanosomes plus 100 μM BrdU (F). In the absence of nucleoside analogs, WT trypanosomes and TK transformants have the same nucleotide composition (data not shown).

TABLE 1.

Moles percent values of HOMedU and dJ in DNA of trypanosomes grown in the presence of HOMedU

Clone	HOMedU concn (mM)	na	Mol% of:
			HOMedU	dJb
PF WT	0	2	0.0	0.0
	0.1	1	0.1	0.02
	1	2	0.4	0.16
PF TKN	0	2	0.1	0.0
	0.1	1	0.9	0.18
	1	2	7.1	0.28
BF WT	0	3	0.0	0.08
	1	2	0.1	0.64
	10	3	0.5	1.46
BF HTK	0	2	0.1	0.12
	1	2	0.3	1.50
	10	1	0.2	2.94

^aThe moles percent values of the total nucleotides of dJ and HOMedU were determined by phosphorimager quantitation of ³²P-postlabeling chromatograms and are expressed as the means of one to three experiments (n). 

^bThe recovery of J by postlabeling is incomplete and variable (see Materials and Methods); the presented values of dJ have been corrected for partial recovery. 

Increasing the J content in DNA by growing BF trypanosomes in the presence of HOMedU.

The effect of HOMedU incorporation on the J content in BF trypanosomes was tested in WT cells (clone 221a) and HTK cells, which contain a HYG gene and a TK gene in the active 221 VSG gene expression site (19). The replacement of dT by HOMedU in BF cells was less complete than in PF trypanosomes, and this was only slightly enhanced by expression of the viral TK gene (Fig. 2E; Table 1). In BF cells, most of the exogenous HOMedU incorporated into DNA was glucosylated, resulting in a substantial increase in the J content (Table 1). HTK BF trypanosomes grown in the presence of 1 mM HOMedU contained approximately 1.5% J, which is 10- to 15-fold higher than the endogenous level (Fig. 2; Table 1). This elevated level of J did not affect the growth rate or morphology of the cells. Concentrations of 5 to 10 mM HOMedU reduced the growth rate of BF trypanosomes, depending on the culture conditions and the cell lines used (data not shown).

In BF T. brucei, endogenous J is present in repetitive DNA and in VSG genes in silent telomeric expression sites but is absent from transcribed VSG genes and transcribed repeats in VSG gene expression sites (56). To examine the location of J in PF trypanosomes and BF trypanosomes cultivated in the presence of HOMedU, genomic DNA was sheared by sonication to fragments of 0.5 to 3 kb and then analyzed by anti-J immunoprecipitation with antisera that specifically recognize J-containing DNA (56). Immunoprecipitated DNA fragments were identified by dot-blot hybridization. From the DNA of PF trypanosomes grown in the presence of HOMedU, telomeric repeats, 50-bp repeats, VSG genes, and tubulin genes were all immunoprecipitated (Table 2), showing that J was present throughout the genome. We have previously found that the efficiency of immunoprecipitation is determined by the degree of modification (56). The differences in the efficiency of immunoprecipitation of the various sequences tested (Table 2) indicates that the degree of conversion of HOMeUra into J in PF cells might vary with the genomic location. In BF trypanosomes grown in the presence of HOMedU, J was also present in every sequence tested (Table 2). The 50-bp repeats and the silent VSG gene VO2, which were already modified, showed increased binding to antibody following incorporation of HOMedU. Remarkably, efficient immunoprecipitation was also found for tubulin genes and for the highly transcribed 221 gene and HYG gene in the active 221 VSG gene expression site. These results support the two-step model for J biosynthesis shown in Fig. 1 and indicate that the incorporation and glucosylation of exogenous HOMedU into DNA of BF trypanosomes showed no obvious sequence specificity.

TABLE 2.

Distribution of J in DNA studied by immunoprecipitation of J-containing DNA fragments

Clone	HOMedU concn (mM)	% Efficiency of anti-J immunoprecipitationa in:
		Telomeric repeats	50-bp repeats	Tubulin genes	VO2 VSG gene	221 VSG gene	HYG gene
PF WT	0	0	1	0	0	0
	1	7	11	7	28	32
PF TKN	1	9	14	2	28	35
BF 3174b	0	31	12	0	8	2	0
	1	19	17	30	52	8	9
	10	31	24	45	71	13	22
BF HTK	0	31	12	1	6	0	0
	1	15	15	28	39	16	21

^aThe efficiency of immunoprecipitation is expressed as the percentage of the total input that was bound by the antibody. Dot blots with immunoprecipitated DNA were hybridized with the probes indicated above each column. The 1% immunoprecipitation of 50-bp repeats in PF WT trypanosomes, the 1% immunoprecipitation of tubulin genes in BF HTK cells, and the 2% immunoprecipitation of the VSG 221 gene in BF 3174 cells were not reproducible and represent nonspecific (co)immunoprecipitation. 

^bBF 3174 cells contain a HYG and a NEO gene in the active 221a VSG gene expression site. 

Reducing the level of J by growing BF trypanosomes in the presence of BrdU.

Having found that incorporation of HOMedU into DNA results in an increase in the J content, we tested whether cultivation of trypanosomes in the presence of BrdU, an analog of thymidine that cannot be converted into J (Fig. 1), could decrease the level of J. After BF cells were grown for eight or nine generations in the presence of BrdU, the DNA was analyzed by ³²P postlabeling. BF trypanosomes contained up to 9.3 mol% BrdU (approximately 30% of dT replaced by BrdU) in their DNA, and this correlated with a reduction in the level of J (Fig. 2F). To determine the reduction in the level of J more precisely, we examined the genomic DNA by anti-J immunoblot analysis, which is more sensitive than ³²P postlabeling (56). We found that the J content in BF trypanosomes grown in the presence of BrdU was reduced up to 12-fold in a dose-dependent manner (Fig. 3; Table 3). The relative reduction in the J content exceeded the relative reduction in the Thy content, which suggests that BrdU inhibits the synthesis of J. By anti-J immunoprecipitation of sonicated DNA fragments, we found that the level of J was reduced in every modified sequence analyzed, showing that the reduction was not region or sequence specific (data not shown). The growth rate of cells was reduced approximately 50% by 250 μM BrdU for non-TK BF cells and by 100 μM for HTK BF cells. We do not know whether this is caused directly by BrdU or indirectly by the reduction in J content. PF trypanosomes were less sensitive to BrdU. When cultivated in the presence of 1 mM BrdU, PF trypanosomes incorporated 12.9 mol% BrdU (Table 3) without any detectable effect on the phenotype. We also tested whether IdU, CldU, or amino-dU affected the synthesis of J, but these thymidine analogs were not incorporated as efficiently as BrdU and reduced the level of J at most only threefold.

FIG. 3.

Anti-J DNA dot blot analysis. DNA of HNR and HTK BF trypanosomes (see Fig. 4 and 5) grown in the absence or presence of BrdU was loaded as a twofold dilution series onto a dot blot and incubated with rabbit anti-J antiserum. Bound antibody was detected with a sheep anti-rabbit secondary antibody conjugated to horseradish peroxidase and visualized by enhanced chemiluminescence. We have previously shown that the detection of J on immunoblots is not affected by the presence of nonmodified DNA (56). A twofold decrease in the enhanced chemiluminescence signal therefore corresponds to a twofold increase in the J content if equal amounts of DNA are loaded. After the antibodies were stripped off, DNA loading was checked by hybridization with a RIME probe (results not shown; see Materials and Methods).

TABLE 3.

Quantitation of the level of BrdU in DNA

Clone	BrdU concn (μM)	n	Mol% ofa:
			BrdUa	dJb
PF WT	0	1	0.0	0.0
	1,000	1	12.9	0.0
PF TKN	100	1	9.0	0.0
BF WT	0	2	0.0	0.14
	100	3	5.2	0.05
	250	2	7.7	0.01
BF HTK	0	2	0.0	0.12
	50	1	7.7	0.02
	100	3	9.3	0.01

^aThe moles percent of BrdU in DNA and the level of dJ in control cells were determined by ³²P postlabeling. 

^bThe recovery of J by postlabeling is incomplete and variable (see Materials and Methods); the presented values of dJ have been corrected for partial recovery. The relative decrease in the level of dJ caused by the incorporation of BrdU into DNA was quantitated by anti-J immunoblot analysis. 

Effects of BrdU and HOMedU on repression of VSG gene expression sites.

To test whether J contributes to transcriptional repression of VSG gene expression sites, we investigated the effect of incorporation of BrdU and HOMedU into DNA on promoters in silent VSG gene expression sites. For this purpose, we used HNR cells (Fig. 4A), which have a HYG gene just downstream of the active 221a VSG expression site promoter and a NEO gene just downstream of the silent VO2 expression site promoter (15). Expression of marker genes and VSG genes was studied with RNA blots (Northern hybridization). The detection limit of marker gene expression is approximately 0.1 to 0.4% of the expression of a marker gene in the active site (indicated as ≤0.4%).

FIG. 4.

Northern blot analysis of RNA from BF trypanosomes cultivated in the absence or presence of thymidine analogs. (A) BF clones HNR and HN1, which are genotypically identical but have a different active expression site (15), and clone RP2XR (47) are described in Materials and Methods. A solid flag indicates an endogenous expression site promoter, and an open flag indicates a ribosomal promoter. Transcription is indicated by a dashed line with arrowhead, and the vertical line downstream of the VSG genes indicates the chromosome end. (B) Northern blots of HN1 control cells and of HNR and BF trypanosomes grown in the absence or presence of BrdU were hybridized with the probes indicated on the left. TUB indicates β-tubulin genes. (C) Northern blots of HN1 control cells and of HNR and RP2XR trypanosomes grown in the absence or presence of HOMedU.

Growth of HNR cells in the presence of BrdU resulted in a dose-dependent derepression of the silent NEO gene (Fig. 4B). Following growth in 0, 100, or 250 μM BrdU, the expression levels of the NEO gene in the silent VO2 site were approximately ≤0.4, 0.8, and 6%, respectively, relative to expression in an active site (set at 100%). No effect was detected on repression of the telomeric VO2 gene or on expression of the active HYG gene in the 221 expression site (Fig. 4B). A similar effect of BrdU on repression was found for the silent HYG gene in clone HN1, which is genotypically identical to clone HNR but has a silent 221 expression site and an active VO2 expression site (Fig. 4A). Growth in 0, 100, and 250 μM BrdU resulted in silent HYG gene expression levels of approximately ≤0.4, 0.7, and 9%, respectively, relative to the expression of an active HYG gene (data not shown).

The effect of HOMedU was less pronounced. Growth of HNR cells in 1 mM HOMedU resulted in a 16-fold increase in the J level (data not shown) but did not affect the expression of the silent NEO gene (which was ≤0.4% of the expression of active NEO in this experiment [Fig. 4C]). Only at 5 mM HOMedU (resulting in a 19-fold increase in the J level) did the expression of the NEO gene increase to about 1%. This derepression, albeit small, was unexpected given the derepression found after incorporation of BrdU, which resulted in a reduced J content. However, a similar effect was seen with cell line RP2XR (Fig. 4A), in which the endogenous expression site promoter has been replaced by a ribosomal promoter (47). The repression of the marker gene (HYG) downstream of this ribosomal promoter has been reported to be less tight than in silent expression sites with an endogenous promoter (47). Indeed, we found that in the absence of thymidine analogs, the expression of the silent HYG gene in clone RP2XR was ∼2% of HYG gene expression from an active expression site (Fig. 4C), which is higher than the expression level of the silent HYG gene in HNR cells. Also, after growth in the presence of HOMedU, the HYG expression levels in RP2XR cells were higher than those in HNR cells: incubation of RP2XR trypanosomes in 1 mM HOMedU (6-fold increase in the J content) and 5 mM HOMedU (16-fold increase in the J content) resulted in ∼2 and 5% expression, respectively. Following the growth of clone RP2XR in the presence of BrdU, the maximum expression of the silent HYG gene was 6% (data not shown). The apparent increase in the level of HYG mRNA in Fig. 4C at 1 mM HOMedU was not significant after quantitation of the HYG and tubulin hybridizations and subsequent correction for loading.

Effects of incorporation of BrdU and HOMedU on VSG gene expression site switching.

We studied the effects of HOMedU and BrdU on VSG gene expression site switching in vitro by using the HTK trypanosomes, in which a HYG gene and a TK gene are integrated downstream of the 221a expression site promoter (Fig. 5A). These cells are sensitive to the nucleoside analogs FIAU and BVDU, which are phosphorylated by the viral TK enzyme and kill the trypanosome (see Materials and Methods). Through the lethal combination of TK and nucleoside analog, it is possible to mimic the negative selection imposed by the host immune response against a VSG antigen type. However, by selecting against expression of the TK gene close to the promoter, one selects for inactivation or loss of the complete expression site rather than for replacement of the telomeric VSG gene without affecting the transcriptional state of the expression site. HTK cells have been used previously to study expression site switching in vitro by negatively selecting trypanosomes that had inactivated the 221a expression site (19).

FIG. 5.

Effect of thymidine analogs on VSG gene expression site switching in vitro. (A) HTK cells contain HYG and TK genes downstream of the active 221 expression site promoter (flag) and are resistant to hygromycin and sensitive to BVDU (Hyg^R, BVDU^S). The dashed line with the arrowhead indicates transcription. Following negative selection, the switched trypanosome clones, which are hygromycin sensitive and BVDU resistant (Hyg^S, BVDU^R), were analyzed by DNA dot blot hybridization for the absence (−) or presence (+) of the HYG and TK (HT) genes and the VSG 221 gene (221). On the basis of the dot blot hybridization, two genotypes of expression site switch variants could be distinguished (see the text): variants that had retained the old expression site (HT/221⁺) and variants that had deleted completely the old expression site and thereby lost the marker genes and the VSG gene (HT/221−). (B and C) HTK cells were grown in the absence (−) or presence (+) of 100 μM BrdU (B) or 1 mM HOMedU (C). The relative number of switchers of each genotype is indicated on the y axis as a percentage of the total number of switched clones in the untreated control population of each panel (HT/221− plus HT/221+ = 100%). For each growth condition, two to five independent HTK cultures were put through a switch experiment, and the data shown in panels B and C represent the means and standard deviations of the switch patterns found. In total, we analyzed 44 clones (n = 3) and 54 clones (n = 5) for growth in the absence and presence of BrdU, respectively, and 14 clones (n = 2) and 68 clones (n = 5) for growth in the absence and presence of HOMedU, respectively. (n indicates the number of cultures used for each condition.) One clone with an H+T−221− genotype was found in the −BrdU control cells (not included in the diagram).

HTK cells were expanded for approximately nine generations in the absence of hygromycin and in the absence or presence of 100 μM BrdU or 1 mM HOMedU. After this period, the cells were washed and plated into medium containing FIAU or BVDU to select for trypanosomes that had silenced TK expression. Revertants that had become resistant to FIAU and BVDU due to a mutation in the TK gene could be distinguished from revertants that had ceased expression of the TK and HYG genes on the basis of resistance to hygromycin (see Materials and Methods). TK mutants were discarded. The other FIAU- and BVDU-resistant revertant clones were expanded and analyzed by DNA dot blot hybridizations.

In a typical switching experiment with HTK control cells, two types of events lead to expression site switching in vitro (19). These events are depicted in Fig. 5A. A small fraction of the switch variants silence the 221 expression site and retain the 221 VSG gene and the marker genes TK and HYG; they therefore represent classical in situ switches (HT/221⁺, Fig. 5A). Most of the remaining switch variants have complete deletions of the previously active 221 expression site and activation of another expression site (HT/221⁻, Fig. 5A). In a small fraction of these clones, loss of the old expression site occurs by replacement via gene conversion by another expression site. The significance of this switching profile for antigenic variation has been discussed by Cross et al. (19). A small fraction of the cells become resistant to BVDU and FIAU due to deletion of the marker genes from the expression site without loss of the VSG 221 gene (19). This type of event does not necessarily represent an expression site switch and has therefore been excluded from the expression site switch studies described here. We focused on the effect of HOMedU and BrdU on the two major types of expression site switch events outlined in Fig. 5A.

Growth of HTK cells in 100 μM BrdU (Table 3) before negative selection did not substantially affect the total switching frequency or the relative contribution of the two major types of switch events (Fig. 5B). In contrast, growth in 1 mM HOMedU, which caused a 10- to 15-fold increase in the J level, resulted in a 5-fold reduction in the total switching frequency. DNA analysis showed that the frequency of in situ switching without loss of expression site sequences (HT/221⁺) was unaffected but the number of events that involved loss of the previously active 221 expression site sequences (HT/221⁻) was reduced approximately ninefold (Fig. 5C). The frequency at which switched clones arose in the controls varied from 6 × 10⁻⁵ (Fig. 5B) to 1 × 10⁻⁵ (Fig. 5C), as seen previously (19). Our results indicate that a global increase in the J content results in a reduction in the occurrence of VSG gene expression site switch events that involve chromosome rearrangements.

DISCUSSION

Two-step pathway of J synthesis.

We have studied the biosynthesis of J by using thymidine analogs. Trypanosomes grown in the presence of HOMedU incorporate this nucleoside into their DNA and make more J. This suggests that the normal conversion of Thy into J in trypanosomes occurs in two steps. In the first step, a Thy in DNA is converted into HOMeUra by a putative DNA thymidine-7-hydroxylase. In the second step, HOMeUra is converted into J by a β-glucosyl transferase (Fig. 1). Formally, it cannot be excluded that HOMedU is glucosylated prior to incorporation into DNA, but the presence of endogenous J at specific locations in BF trypanosomes strongly suggests that J is a postreplicational DNA modification introduced by enzymes that recognize thymine in a certain sequence and/or chromatin structure. This is further supported by the presence of HOMeUra preferentially in areas containing J in trypanosomes that have not been grown in HOMedU (55). PF trypanosomes, which are normally devoid of J, start synthesizing this modified base when they incorporate HOMedU into their DNA. This shows that PF trypanosomes contain sufficient β-glucosyl transferase activity to glucosylate a small fraction of the incorporated HOMedU (Table 1). The absolute lack of J in PF trypanosomes is therefore most likely to be caused by the absence of DNA thymidine-7-hydroxylase activity.

In trypanosomes, but also in other unicellular organisms, HOMeUra seems relatively inert in DNA (reviewed in reference 21). In Tetrahymena, which is normally devoid of HOMeUra, the replacement of 33% of Thy by this base does not affect the growth rate or morphology (44). In addition, it is a common natural substituent of the DNA of dinoflagellates (27, 42, 43) and it replaces all thymines in the DNA of some Bacillus subtilis bacteriophages (32). The toxicity of HOMedU for mammalian cells (58) is caused by the expression of the DNA repair enzyme HOMeUra-glycosylase (9, 29). This enzyme is part of the base excision repair pathway that removes the base HOMeUra from DNA. Incorporation of exogenous HOMedU therefore results in excessive DNA repair. Mutant mammalian cells lacking HOMeUra-glycosylase incorporate exogenous HOMedU with no effect on growth, showing that the presence of HOMeUra per se is not toxic for mammals (8). HOMeUra-glycosylase activity has been found in most animals but not in lower eukaryotes (7). The high levels of HOMeUra obtained in trypanosome DNA strongly suggest that trypanosomes do not contain substantial HOMeUra-glycosylase activity either.

Altering the J content in BF trypanosomes.

BF trypanosomes incorporated less HOMedU into their DNA than did PF trypanosomes. However, they converted the majority of the exogenous HOMedU into J, resulting in up to 2.9 mol% J. The extra J was not restricted to nontranscribed repeats and silent telomeric VSG genes but was present throughout the genome and even occurred in highly transcribed sequences such as the active expression site. These results indicate that under these conditions, J is not sufficient to repress transcription or block transcription elongation. These results also demonstrate that the specific distribution of J in BF trypanosomes and absence of J from transcribed sequences is not determined by the HOMeUra-specific β-glucosyltransferase but, rather, by the DNA thymidine-7-hydroxylase that is responsible for the synthesis of endogenous HOMeUra in the absence of exogenous HOMedU. Also, in procyclic forms, the J synthesized following the incorporation of HOMedU was present throughout the genome. The semirandom distribution of J in PF trypanosomes was confirmed by isolation and postlabeling analysis of minichromosomes, which are composed mainly of 177-bp repeats and telomeric repeats. No enrichment for J or HOMeUra was seen in minichromosomal DNA of PF-TKN cells grown in HOMedU (data not shown), whereas minichromosomal DNA of WT BF trypanosomes contains a six- to sevenfold-higher level of J than does total DNA (22).

Uptake of BrdU, which cannot be converted into HOMedU or dJ, reduced the level of J in BF trypanosomes. The degree of reduction in J content (up to 12-fold) exceeded the relative reduction in Thy residues (up to 1.3-fold) and is therefore not simply a result of the inability of BrdU to serve as an acceptor for the hydroxyl and glucose moieties. We infer that BrdU inhibits the biosynthesis of J. In eukaryotes that contain 5-methylcytidine, methylation of DNA is inhibited by 5-azacytidine, an analog that cannot be converted into 5-methylcytidine. Upon incorporation into DNA, 5-azacytidine irreversibly binds to and thereby inactivates the DNA methyltransferase (31, 49). We do not know how BrdU inhibits the synthesis of J, but we expect that incorporation into DNA is required to inactivate the DNA-modifying enzymes. We did not find a significant correlation between loss of dJ and increase in the HOMedU level in the DNA of cells that had incorporated BrdU (data not shown), suggesting that synthesis of HOMedU and not its glucosylation was inhibited by BrdU.

Effects of altered levels of J on transcriptional silencing.

The presence of J in and around VSG expression sites and in silent telomeric VSG genes suggests that J is involved in the repression of transcription (4, 41, 56). We have examined the role of J in the control of VSG gene expression sites in BF trypanosomes by altering the J content of the DNA. Incorporation of BrdU into DNA resulted in a substantial derepression of marker genes in silent VSG expression sites, but the telomeric VSG genes were not derepressed. Due to the lack of unique sequences in between the VSG gene and the marker gene, we have not analyzed where in the expression site the partial derepression ends. At this stage, we do not know whether the derepression is caused by the reduction in the level of J or by other effects of BrdU on cell physiology. IdU, another halogenated thymidine analog, was also relatively efficiently incorporated into DNA (2 to 2.7%) when used at concentrations of 500 to 750 μM, but this did not substantially affect the J content or the transcriptional repression of the silent NEO gene in an inactive VSG gene expression site (data not shown). Experiments with other thymidine analogs such as CldU or NH₂dU were uninformative because these analogs were incorporated less efficiently into DNA than was BrdU (data not shown). The effect of high concentrations of BrdU on the growth rate of trypanosomes was found only with BF cells. PF cells, which lack J, were not affected by high concentrations of BrdU in the medium, resulting in high levels of BrdU in the DNA (Table 3).

In VSG gene expression site switch experiments, the reduction in the level of J and the derepression of silent expression site promoters caused by BrdU incorporation had no substantial effect on the switching frequency (Fig. 5). Whether the complete absence of J will affect expression site control remains to be verified with mutant trypanosomes that lack synthesis of J or by the identification of more potent inhibitors of the synthesis of J.

Unexpectedly, overproduction of J caused by the incorporation of HOMedU also resulted in derepression close to the promoter of silent expression sites. The effect of HOMedU, albeit small, was unexpected, given the substantial derepression associated with a decreased J level following growth in BrdU. However, as has been suggested for 5-methylcytosine, there are two mechanisms by which J might function in DNA. Replacement of Thy by J could repel factors that would normally bind, such as transcription factors or DNA polymerase. Alternatively, J might act via the recruitment of proteins that alter the structure of the chromatin and thereby affect transcriptional repression. The moderate derepression caused by HOMedU indicates that the effect of J on expression site silencing is not direct and that J might act through the recruitment of repressor proteins. If these factors are limiting, the global overproduction of J caused by the incorporation of HOMedU would result in a redistribution of the repressor proteins. This would be analogous to transcriptional silencing mechanisms in yeast and repression through methylcytosine binding proteins in higher eukaryotes. Mammalian cells have been estimated to contain approximately 6 × 10⁶ molecules of MeCP2, a methyl-CpG binding protein, while approximately 4 × 10⁷ methyl-CpG molecules are present in a typical diploid nucleus (39). MeCP2 will therefore probably not saturate its available binding sites. The reversible silencing at telomeres in the budding yeast Saccharomyces cerevisae involves recruitment of a set of proteins that are limited for transcriptional repression at the telomere (35, 37). The silencing proteins can be titrated by the introduction of extra binding sites (34, 59), whereas overexpression of the repressor protein Sir3p results in a more efficient repression of genes adjacent to a telomere (26, 45, 51). We have recently obtained evidence for the presence of J-DNA binding proteins in nuclear extracts of BF trypanosomes and of Crithidia fasciculata, which also contains J (18). If the function of J is indeed mediated by J binding proteins and if these factors are not present in large excess, the approximately 10- to 15-fold global increase in J content caused by HOMedU incorporation could result in trapping of the J binding protein to other sites in the DNA. This might result in a partial loss of function of the endogenous J present in and around expression sites, whereas introduction of J at sites that normally lack J might result in a partial gain of J function.

Our results indicate that the presence of J might strengthen transcriptional silencing of inactive expression site promoters. We think that J is most probably a consequence and not the cause of expression site silencing. The modifying enzyme recognizes Thy in a large number of unrelated sequences, but only if these sequences are close to or part of the repetitive sequences (56). This suggests that these recognition sequences are distinguished by a specific chromatin structure or subnuclear location, which is imposed before DNA modification. The presence of J in DNA might help to stabilize a repressed chromatin structure and keep inactive expression sites in a silenced state through mitosis.

J might play a role in suppression of DNA rearrangements.

The abundance of J in stretches of repetitive DNA previously led to the suggestion that J might play a role in suppression of recombination (55, 56). The results presented in Fig. 5 indicate that J might be involved in the maintenance of chromosome stability through suppression of chromosome breakage events. Incorporation of HOMedU into DNA and the subsequent 10- to 15-fold increase in J content led to a 9-fold reduction in the frequency of VSG gene expression site switching in which the previously active site is lost. In contrast, growth in BrdU resulted in a small increase in the occurrence of these events. No effect was seen on the frequency of in situ expression site switching in which no DNA rearrangements are apparent. The majority of the DNA rearrangements associated with switching result in large-scale deletions (up to 200 kbp) and involve the loss of the long array of 50-bp repeats upstream of the expression site (19). These gross DNA rearrangements have been suggested to represent background chromosome breakage events that are picked up by the powerful TK negative-selection system. Activation of a new site cannot readily occur without inactivation of the old one, and loss of the old site may help to activate the new one. The extra J present throughout the genome, and thus also upstream of the expression site, may reduce the occurrence of the double-strand DNA breaks which can result in loss of the active expression site. Whether in normal cells J is also located in the (uncharacterized) regions upstream of VSG gene expression sites remains to be determined.

How could J be involved in suppression of DNA rearrangements? If J results in the formation or stabilization of a condensed chromatin structure, it may not only repress transcription but also suppress recombination, e.g., between similar sequences on different chromosomes. Such a dual function has been suggested for the chromatin of the silent mating-type cassette region in the fission yeast Schizosaccharomyces pombe. This locus is a “cold spot” for recombination and is transcriptionally repressed (25). Furthermore, it has been found that the silencing protein Sir2p in S. cerevisiae is involved not only in transcriptional silencing at telomeres and at the mating type loci but also in suppression of recombination between the tandemly repeated rRNA genes (20, 24). DNA modification and silencing proteins might also affect chromosome stability via DNA repair pathways. It has been found that silencing proteins in S. cerevisiae facilitate double-strand break repair by nonhomologous end joining (52) and that silencing proteins interact with proteins involved in nucleotide excision repair (40).

The conservation of J among kinetoplastid flagellates, most of which are not known to undergo transcriptional silencing, shows that J has not just evolved for the control of antigenic variation but has a more general function (55, 56). The results presented here indicate that J might also play a role in the maintenance of chromosome stability in kinetoplastids. African trypanosomes might have recruited J and thereby the proteins that bind to modified DNA for strengthening transcriptional silencing of VSG gene expression sites. Such a function of J is supported by the developmental regulation of J biosynthesis that is found only in African trypanosomes but not in the other kinetoplastida (55).