Transcription Dependent Loss of an Ectopically Expressed Variant Surface Glycoprotein during Antigenic Variation in Trypanosoma brucei

ABSTRACT In the mammalian host, Trypanosoma brucei is coated in a single-variant surface glycoprotein (VSG) species. Stochastic switching of the expressed VSG allows the parasite to escape detection by the host immune system. DNA double-strand breaks (DSB) trigger VSG switching, and repair via gene conversion results in an antigenically distinct VSG being expressed from the single active bloodstream-form expression site (BES). The single active BES is marked by VSG exclusion 2 (VEX2) protein. Here, we have disrupted monoallelic VSG expression by stably expressing a second telomeric VSG from a ribosomal locus. We found that cells expressing two VSGs contained one VEX2 focus that was significantly larger in size than the wild-type cells; this therefore suggests the ectopic VSG is expressed from the same nuclear position as the active BES. Unexpectedly, we report that in the double VSG-expressing cells, the DNA sequence of the ectopic copy is lost following a DSB in the active BES, despite it being spatially separated in the genome. The loss of the ectopic VSG is dependent on active transcription and does not disrupt the number or variety of templates used to repair a BES DSB and elicit a VSG switch. We propose that there are stringent mechanisms within the cell to reinforce monoallelic expression during antigenic variation.

brucei genome, and the structure is broadly conserved (7), with the VSG gene flanked upstream by a block of repetitive sequences known as the 70-bp repeats and downstream by the telomere (4,5,8). Unusually the single active VSG is transcribed by RNA polymerase I (RNA Pol I) (9,10), which is normally reserved for the transcription of rRNA genes (11) at a specialized transcription compartment known as the expression site body (ESB) (9,12,13). The active BES occupies a distinct chromatin architecture, which is depleted of nucleosomes (14,15) and is enriched for highly sumoylated chromatin-associated proteins (16) and the high-mobility group box (HMGB) chromatin protein TDP1 (17,18). The recently identified VSG exclusion (VEX) complex, consisting of VEX1 and VEX2, localizes to the active BES in a transcription-dependent manner and defines the single active BES (19,20). The chromatin assembly factor, CAF-1, maintains the epigenetic status of the active BES during DNA replication, and VEX complex association with the spliced leader locus ensures sufficient processing of the VSG mRNA (19)(20)(21).
To escape detection by the host immune system, the single expressed VSG is stochastically switched. VSG gene switching can occur by an in situ transcriptional BES switch (22,23); however, the main pathway used is gene conversion (GC), which allows access to the subtelomeric archival VSGs via recombination (24). Antigenic variation is thought to be driven by DNA double-stranded breaks (DSB) that naturally accumulate in the BESs (25,26), with the 70-bp repeats critical for recombination (5,8,27,28). Indeed, disruption of factors involved in preserving genome integrity leads to elevated VSG switch frequency. Depletion of the RNases H1 and H2, which resolve RNA-DNA hybrids known as R-loops, increases replication-associated DNA damage and VSG switching (29,30). RNase H2 interacts with the histone methyl transferase DOT1b, and DOT1b deletion results in RNA-DNA hybrids forming, DNA damage, and increased recombination-driven VSG switching (31). Similarly, knockdown of ataxia telangiectasia and Rad3-related (ATR) kinase increases recombination-driven VSG switching, which is associated with elevated levels of DNA damage (32,33). The RTR complex (RECQ2-TOPO3a-RMI1) removes recombination intermediates during mitotic crossover, and loss of the RECQ2 helicase or RMI1 deficiency leads to increased VSG gene switching by recombination (34,35), while TOPO3a knockdown results in a RAD51-dependent hyperswitching phenotype (36). The telomere (37,38), telomeric chromatin (39), nuclear lamina protein NUP-1 (40), and histone variants H3.V and H4.V (6) have also been shown to influence VSG switching.
While VSG expression is strictly monoallelic, this has been disrupted in the laboratory through the selection for activation of a second BES (41,42) or by introducing a second VSG into the active BES (43)(44)(45)(46) or an rDNA locus (41,47,48). While cell growth is not disrupted in parasites with a dual coat (42)(43)(44)46), it has been proposed that simultaneous expression of two VSG genes is unstable and the induction of high levels of expression of a second VSG from a ribosomal locus results in transcriptional attenuation of the BES VSG (48). Upon selection for activation of two BESs, both VSGs colocalize to a single transcription compartment, where they undergo dynamic transcriptional switches between the two BESs, with monoallelic expression rapidly restored upon the removal of drug selection (42). The occupation of two BESs within a single ESB is thought to be energetically unfavorable (42), and attempts to select for the simultaneous activation of three BESs were unsuccessful (49).
The VSG transcript is essential for cell survival, and knockdown triggers precytokinesis cell cycle arrest (50) and global translation arrest (45). This lethal phenotype is rescued by VSG switching (50) or the expression of a second VSG (45). Consistent with this, a DSB introduced in the actively transcribed BES is a highly toxic lesion that results in over 95% cell mortality and all of the DSB survivors having switched the active VSG (26,34). Here, we asked whether the toxicity of a BES DSB could be rescued by the expression of a second VSG and if this modified the VSG switching pathway used. We established a cell line in which monoallelic VSG expression is disrupted by the stable expression of a second telomeric VSG from a ribosomal DNA (rRNA gene) locus, resulting in the presence of both the ectopic and native VSGs at the cell surface. We found that using a targeted DSB in the active BES to trigger a VSG switch, the DNA sequence of the ectopic VSG is lost, despite it being located at a distinct genomic locus. Deletion of the ectopic VSG is transcription dependent and does not affect the number or the variety of genomic templates used for BES repair. We propose that monoallelic expression is strictly reinforced during antigenic variation, preventing multiple VSG switching events occurring at once, which would compromise parasite survival.

RESULTS
Expression of a second VSG does not rescue the toxicity of a DSB at the active BES. The established VSG up cell line facilitates the induction of a specific DSB adjacent to the 70-bp repeats in BES1 (25,26), from which VSG-2 (also known as VSG221 [5]) is actively expressed (Fig. 1Ai). Tetracycline (Tet)-inducible expression of the I-SceI meganuclease cleaves an 18-bp recognition site adjacent to the 70-bp repeats upstream of the VSG in BES1 and is highly deleterious to the cell population, with only 5% of the population able to survive. Among the population of DSB survivors, there is a 5,000fold increase in the rate of VSG switching (26). Previous reports have shown that the simultaneous expression of two VSGs does not disrupt T. brucei parasite growth (42,43,46,51) and, therefore, to determine whether the expression of a second copy of a VSG gene could rescue the toxic effect of a DSB at the active BES, we introduced an ectopic copy of VSG-5 into the VSG up cell line at a ribosomal promoter using the p 5 NTMF construct (19). The VSG-5 expression construct contains a 200-bp telomere seed, and the single crossover event at the ribosomal promoter therefore generates a de novo telomere (19,52) (Fig. 1Aii). Approximately 20% of the transformed clones show high-level VSG-5 expression (Fig. S1A in the supplemental material), and the presence of both VSG-2 and VSG-5 at the cell surface was confirmed by immunofluorescence assay (IFA) (Fig. 1B). Clones expressing high levels of VSG-5 are Double EXpressing (DEX) and are referred to as VSG upDEX . Clones where VSG-5 expression was not detected are referred to as VSG upDEX_OFF . The p 5 NTMF construct contains an NPT gene for selection and, therefore, to determine whether VSG-5 expression is stable, the VSG upDEX clones were grown in the absence of NPT selection for 7 days by removal of the G418 drug from the growth medium. Protein analysis revealed that both VSG-5 and NPT are expressed in the absence of selection pressure, indicating that the expression of ectopic VSG-5 is stable (Fig. S1B). We then monitored the rate of parasite proliferation in the VSG up and VSG upDEX cell lines under both DSB-inducing and -noninducing conditions. We did not observe growth differences ( Fig. 2A, solid lines), nor an increase in survival of VSG upDEX cells ( Fig. 2A, dashed lines), indicating that the expression of a second VSG alone does not ameliorate the toxicity of a BES DSB.
The site of VSG transcription is marked by an extranucleolar pool of RNA Pol I called the ESB (9,12) and the VEX complex (19,20). VEX2 forms a single focus in the nucleus that is coincident with the actively expressed BES (20). We therefore asked whether the ectopically expressed VSG-5 was able to recruit VEX2 and whether the number of foci was consistent between the VSG up and VSG upDEX . One allele of VEX2 was natively tagged (20) and the nucleolus marked using the L1C6 antibody (53). VSG up and VSG upDEX cells were scored for one or two VEX2 foci in G 1 -and S-phase cells (Fig. 2B). For VSG up and VSG upDEX , 96% and 92% of cells, respectively, had one VEX2 focus prior to DSB induction, and this did not change following a DSB (Fig. 2B). Two active BESs have been shown to colocalize to a single ESB (42), and the identification of a single VEX2 focus in the large majority of VSG upDEX cells suggests that ectopic VSG-5 expressed from an rDNA locus occupies the same transcription compartment as BES1.
During our experiments, we observed that the VEX2 foci appeared larger in the VSG upDEX cells, which led us to carry out measurements of the length of the VEX2 focus in both cell lines. In the VSG up cell line, the average VEX2 focus length was 0.30 mm, while in the VSG upDEX cells, it was significantly larger, with an average length of 0.38 mm (P , 0.0001) (Fig. 2C). We then assessed the size of the VEX2 focus following a DSB. We found that the length of the VEX2 focus in the VSG upDEX cell line was reduced to an average length of 0.29 mm, which is not significantly different from the Ectopic VSG Loss during Antigenic Variation ® uninduced VSG upDEX and is similar to the VSG up cell line. This indicates that in the VSG upDEX cell line, the VEX2 VSG transcription compartment is larger, possibly as a result of accommodating the transcription of two VSGs, but this is not maintained following a DSB and subsequent repair. To directly test this, we performed DNA fluorescent in situ hybridization (FISH) using probes to the Pseudo gene, found in the active BES, and the NPT gene in the p 5 NTMF construct. In order to compare the position of the NPT gene, and thus the ectopic VSG, in the nucleus when actively expressed and silenced, we examined both the VSG upDEX and VSG upDEX_OFF cell lines. We found that in the VSG upDEX cells, the Pseudo and NPT foci were an average distance of 0.4 6 0.19 mm apart, which was significantly closer in proximity than the VSG upDEX_OFF cells, where the foci were 1.02 6 0.312 mm apart (Fig. 2D). We also noted that in the VSG upDEX_OFF cells, the NPT focus was seen within the nucleolus, as expected given that integration of this construct is at a ribosomal promoter (Fig. S2). This, along with the increase in the size of the VEX2 foci, indicated that the ectopic VSG-5 copy occupies the same transcription compartment as BES1 (Fig. 2D).
VSG-2 and VSG-5 are lost from the cell surface following a DSB. Given the increase and subsequent decrease in the size of the VEX2 focus in the VSG upDEX , we assessed the expression of VSG-2 and VSG-5 protein following a DSB. An I-SceI DSB was induced in the VSG upDEX cell lines in the absence of any drug selection. Protein analysis Percentage of cells with one or two VEX2 foci in the VSG up and VSG upDEX cell lines. Cells gown under either uninduced conditions or DSB-inducing conditions for 7 days (1Tet) were scored for 1 or 2 VEX2 foci. Only cells in the G 1 or S phase were counted. For the VSG up cell line, error bars represent standard deviation between a pair of technical replicates. For the VSG upDEX cell line, error bars represent standard deviation between a pair of biological clones. n . 100 cells for each condition. (C) VEX2 foci length (mm) in VSG up and VSG upDEX cell lines. Individual black circles represent a single measurement in a G 1 -or S-phase cell. 1Tet indicates growth under DSBinducing conditions for 7 days. Horizontal bar represents the mean. Significance was calculated using an unpaired t test; ****, P , 0.0001. Righthand panel shows representative IFA image of an uninduced VSG upDEX . Scale bar, 2 mm. Inset shows a Â2 to Â3 magnification of the nucleus with black line to represent an example of the measurements made. (D, Left) Quantification of the distance between Pseudo and NPT foci in the VSG upDEX and VSG upDEX_OFF cell lines. n = 12 and n = 15 cells counted for the VSG upDEX and VSG upDEX_OFF cell lines, respectively; the average value is shown with a horizontal bar 6 SD, and the significance of colocalization as ****, P , 0.0001. Images were randomized to allow for blind counting. (D, Right) Schematics of BES1 and p 5 NTMF reporter with a black bar representing the DNA FISH probe. Representative images of the DNA FISH from the VSG upDEX and VSG upDEX_OFF cell lines.
Ectopic VSG Loss during Antigenic Variation ® revealed that VSG-2 decreased over 96 h following a DSB (Fig. 3A), which is expected, as the VSG-2 gene is located downstream of the DSB site ( Fig. 1Ai) and lost following a break (26). Protein analysis of both VSG-5 and NPT also showed a decrease over the course of the induction (Fig. 3A), suggesting that VSG-5 is not maintained at the cell surface following a DSB at the active BES. We next examined parasite populations for surface VSGs by immunofluorescence assay. As it takes approximately 4.5 days to completely replace the VSG coat (46), we analyzed cells at 7 days post-DSB induction. Neither VSG-2 nor VSG-5 were detected in the population (Fig. 3B). We then wanted to assess this finding in individual subclones. For this, we generated a panel of five uninduced subclones and 20 DSB repaired subclones for each of the VSG up and VSG upDEX cell lines. DSB induction was confirmed in all induced subclones by sensitivity to puromycin. The puromycin resistance gene is located downstream of the DSB site in the VSG up cell line and is either lost or disrupted following a DSB. All induced subclones were puromycin sensitive, while all uninduced subclones were puromycin resistant, confirming I-SceI cleavage. While both VSG-2 and VSG-5 were detected at the surface of all uninduced subclones, both were lost from the surface of all 20 DSB-induced subclones (Fig. 3C). This was confirmed by fluorescence-activated cell sorting (FACS) of a population at 0 and 7 days post-DSB. In the uninduced cells, both VSG-2 and VSG-5 are detected at the surface of 99.1% of cells, and this was reduced to 6.5% of cells 7 days post-DSB induction (Fig. 3D), with a second clone behaving in a similar manner (Fig. S3). The remaining 6.5% VSG-2/VSG-5-positive cells most likely represent those not cleaved by I-SceI. This suggests that a DSB in BES1 results in the loss of VSG-5 from the cell surface, despite the VSG-5 gene being located at a distinct locus to BES1.
Ectopic VSG-5 is lost from the genome following a DSB. The concurrent reduction in VSG-2 and VSG-5 at the trypanosome surface following a BES DSB led us to ask how the VSG upDEX cells repaired the I-SceI DSB. It has been shown that a DSB in the active BES is repaired predominantly via gene conversion using the 70-bp repetitive sequences for homology (25,26,54). To map individual repair events in the VSG up and VSG upDEX cell lines, we used our cohort of subclones (5 uninduced and 20 induced survivor subclones for each cell line) and a series of established PCR assays to assess the presence or absence of several marker genes in BES1 (26,55). Using primers specific to VSG-2, we found that for both VSG up and VSG upDEX , all uninduced subclones (n = 5) were VSG-2 positive by PCR, and all induced subclones (n = 20) were VSG-2 negative ( Fig. 4A; Fig. S4A and B), supporting both the IFA and FACS data showing loss of the protein in VSG upDEX ( Fig. 3B and C). BES1 contains two blocks of 70-bp repeats; the first block is located immediately upstream of the VSG, and a second, smaller, 70-bp repeat region is found further upstream, flanked by the pseudo gene and ESAG1 (Fig. 4A) (7). Cells that are PCR positive for ESAG1 and pseudo are expected to have undergone GC using the 70-bp repeats directly upstream of VSG-2 as a homologous template. If pseudo is lost but ESAG1 preserved, repair has likely occurred by GC using the smaller block of 70-bp repeats. Subclones that have lost ESAG1, pseudo, and VSG-2 have likely undergone BES loss. Following a VSG up DSB, 75% of subclones (n = 15) retained both ESAG1 and pseudo ( Fig. 4A and Fig. S4A), in agreement with previous findings that the large block of 70-bp repeats immediately upstream of the DSB is favored for repair (25, Ectopic VSG Loss during Antigenic Variation ® 26). For VSG upDEX , the findings were notably similar, with 70% (n = 14) and 75% (n = 15) retaining pseudo and ESAG1, respectively ( Fig. 4A and Fig. S4B). Six repaired VSG upDEX subclones had lost pseudo, and 5 lost ESAG1, suggesting repair by recombination upstream of these genes or loss of the entire BES. The one induced subclone that was ESAG1 positive but pseudo negative likely repaired the DSB via gene conversion with the smaller block of 70-bp repeats upstream of pseudo gene. Importantly, these analyses demonstrate that the expression of a second VSG does not disrupt the regions of homology used within the BES to repair the DSB.
We next looked for the presence or absence of the ectopic expression construct genes VSG-5 and NPT in the VSG upDEX subclones. The T. brucei genome contains three copies of VSG-5 (6), and amplification of the ectopic copy of VSG-5 was achieved using primers specific to the procyclin 59 untranslated region (UTR) and VSG-5 (Fig. 4B). In this construct, the procyclin 59 UTR is used for processing. Unexpectedly, all VSG upDEXinduced subclones (n = 20) were negative for both VSG-5 and NPT ( Fig. 4B and Fig. S4C). This indicates that the loss of VSG-5 at the cell surface is accompanied by loss of the DNA sequence. To validate loss of these sequences, we carried out whole-genome sequencing on 2 uninduced and 4 induced VSG upDEX subclones. Reads aligning uniquely to NPT were not observed in any of the four induced subclones (Fig. S5), consistent with the findings of the PCR assay. We were unable to identify reads specifically mapping to ectopic VSG-5 using short paired-end reads. Together, these data demonstrate that the loss of VSG-5 and NPT protein is a result of loss of the DNA sequence.
The NPT resistance cassette used for integration of the VSG-5 expression construct allowed us to test whether we could retain the ectopic VSG-5 copy using G418 selection. We saw no difference between the VSG up , VSG upDEX , or VSG upDEX_OFF cells grown under DSB-inducing conditions and selection using 2 mg/mL of G418 for 7 days (Fig. S6A). We expect the VSG upDEX_OFF cells to behave similarly to the VSG up cells, as they contain the VSG-5 ectopic construct, but it is not transcribed. The presence of VSG-2 and VSG-5 at the cell surface was then examined by IFA. In the uninduced cells, 100% of the cell population expressed only VSG-2 in VSG up and VSG upDEX_OFF , and 100% of the VSG upDEX population expressed both VSG-2 and VSG-5 on their surface (Fig. S6B). Following DSB induction in VSG up , 95% of cells switched. In the VSG upDEX and VSG upDEX_OFF cells, induction and selection with G418 resulted in 96% and 3% of cell switching, respectively (Fig. S6B). We then assessed a panel of subclones to determine the efficiency of I-SceI cutting and DSB repair in the cells. Out of 30 induced subclones tested, 29 were puromycin resistant (Fig. S6C), indicating that the majority of the subclones were not subject to an I-SceI DSB. Given that the I-SceI meganuclease has over 95% efficiency of cutting (56), this suggests that by using G418 selection, we have applied a strong selective pressure for uncut cells. The single puromycin-sensitive subclone was VSG-2 and VSG-5 positive by IFA (data not shown) and has likely repaired the DSB using short regions of homology for imperfect repair, resulting in disruption of the puromycin open reading frame but preserving VSG-2 in BES1. We were unable to identify any cells that had undergone GC to repair the DSB while maintaining VSG-5 at the cell surface. This indicates that it is highly unfavorable to undergo VSG switching while a second VSG is present at the cell surface.
Loss of VSG-5 is dependent on transcription. Loss of the ectopic VSG-5 expression construct following a DSB was unexpected, and we hypothesized that its transcriptional state could be an influence. Of the 10 VSG upDEX cell lines we originally generated, only 20% expressed VSG-5 (Fig. S1); we therefore selected 2 clones where VSG-5 is silenced and VSG-2 is the singly dominant VSG expressed. We refer to this cell line as VSG upDEX_OFF (Fig. 5A and Fig. S7A). As in the VSG upDEX cell line, cell growth was not affected in VSG upDEX_OFF , demonstrating that growth is not affected by the integration of the p 5 NTMF construct (Fig. 5B). We then generated a cohort of 5 uninduced and 20 induced VSG upDEX_OFF subclones for further analysis. All induced subclones were puromycin sensitive, indicating efficient cutting by I-SceI. Using our PCR assays, we found that 95% of the induced VSG upDEX_OFF subclones had lost VSG-2, and 65% had retained the pseudo gene ( Fig. 5C and Fig. S7B), indicating that repair using the 70-bp repeats for recombination dominated. In striking contrast to the VSG upDEX cell line, all 20 VSG upDEX_OFF subclones retained both VSG-5 and NPT following DSB induction (Fig. 5C and Fig. S7B). This demonstrates that the loss of VSG-5 and NPT in the VSG upDEX cell line is transcription dependent.
The accumulation of single-stranded DNA is delayed in VSG upDEX cells. Following a DSB, resection proceeds rapidly to facilitate repair. Given the loss of both the native VSG-2 sequence and the ectopic VSG-5 sequence, we next investigated whether the timing of the DNA damage response at the active BES was disrupted. A quantitative resection assay was employed (55,57,58) that makes use of a HindIII site within VSG-2 to assess the timing of single-stranded DNA (ssDNA) accumulation at the active BES (Fig. 6A). In brief, genomic DNA was harvested at 0, 6, and 12 h post-DSB induction. The genomic DNA (gDNA) was then either digested with HindIII or mock digested. If the gDNA is single stranded (ssDNA) at the HindIII site, the restriction digest will be  Fig. 1Ai; (ii) the transcriptionally silenced p 5 NTMF construct. The ribosomal promoter upstream of VSG-5 is silenced, as indicated by a black rectangular bar. All other details as in Fig. 1Aii. (B) Cumulative growth of the VSG upDEX_OFF cell line over 96 h. (C) PCR analysis of VSG upDEX_OFF subclones. Primer binding sites for PCRs are shown as black arrows in Fig. 6A. 2Tet, n = 5; 1Tet, n = 20.
Ectopic VSG Loss during Antigenic Variation ® inhibited, and a PCR product will be amplified (Fig. 6A). If the DNA is double stranded, HindIII digestion will proceed, and the template for the PCR amplification is destroyed. Therefore, a PCR product will only be amplified if the originating gDNA is single stranded at the HindIII site. For each sample, the amount of PCR product is compared to that of an undigested "mock"-digested sample in which the template remains intact and a PCR product is amplified. For the VSG up cell lines, we found that ssDNA increased over time following a DSB at 0, 3, 6, 9, and 12 h, respectively ( Fig. 6B and C), in FIG 6 The expression of a second VSG disrupts the timing of repair at the active BES. (A) Schematic of the quantitative assay used to measure ssDNA resection. Black lines represent genomic DNA harvested from parasites at different time points following I-SceI induction. The I-SceI recognition site in BES1 is shown as a blue bar and the HindIII digestion site with a red bar. Pathway 1 shows the outcome if the DNA adjacent to the I-SceI restriction site is single stranded. HindIII digestion will not proceed, as shown with a red cross. PCR amplification of either side of the HindIII site, demonstrated with red arrows, will produce a product as the template is intact (green tick). Pathway 2 shows the outcome if the DNA adjacent to the I-SceI restriction site is double stranded. HindIII will digest (green tick), and the PCR will not yield a product, as the template has been destroyed (red cross). Mock digestion with HindIII is used as a control. Figure created with BioRender.com (B) Quantification of DNA end resection, measured in VSG up and VSG upDEX cell lines 0, 3, 6, 9, and 12 h following induction of a DSB. The VSG up error bars represent standard deviation between a pair of technical replicates, while the VSG upDEX error bars represent standard deviation between a pair of biological clones. The qPCR assay was performed in triplicate. Significance was calculated using an unpaired t test; **, P , 0.001. (C) Table summarizing the percentage of ssDNA in each cell line following induction of a DSB. agreement with previously published data (26,55). For VSG upDEX , the relative amount of ssDNA was significantly lower at 6 and 9 h post-DSB compared to VSG up (P , 0.0001) (Fig. 6B). By 12 h, the VSG upDEX cell lines' ssDNA was similar to that of VSG up at the same time point. Therefore, the initial stages of DNA resection are delayed in VSG upDEX cell line.
Access to the VSG archive is not affected in the VSG upDEX cells. The transcriptiondependent loss of VSG-5 following a VSG switch indicates that there are stringent mechanisms at play to ensure that more than one VSG is not activated during VSG switching, preventing multiple switching events from occurring at once. Taken together with the initial delay in resection, which is critical to repair pathway choice, we wanted to determine whether the cohort VSG genes expressed in the repaired cells indicated a shift in the VSG switching pathway used. To assess this, we used VSG sequencing (VSG-seq) (59) to examine the number of distinct VSG transcripts arising in the population following a VSG switch. The expressed VSGs in the VSG up and VSG upDEX cell lines were amplified and sequenced at 0 and 7 days post-DSB induction. VSG genes were considered significantly enriched when the log 2 fold change (FC) of the induced sample compared to the uninduced is .2, with a P value of ,0.05. We identified 64 VSG sequences that were significantly enriched in the VSG up cell line and 66 in the VSG upDEX cell line (Fig. 7A). We next looked at the genomic locations of the significantly enriched genes by mapping them to the T. brucei 427 genome (6) with the minichromosomal VSGs added from the VSGnome (7) (Fig. 7B). We found that for the VSG up cell line, 9 VSGs were located on the BESs, 20 on the minichromosomes, and 16 on the megabase chromosomes ( Fig. 7B and C). For the VSG upDEX cells, the results were strikingly similar, with 9, 20, and 17 VSGs located on the BESs, minichromosomes, and megabase chromosomes, respectively ( Fig. 7B and C). Therefore, the number and genomic positions of the templates used to repair a VSG upDEX DSB are analogous to that of VSG up , indicating that despite the expression and subsequent loss of VSG-5 following a BES DSB, repair at the BES continues unabated. We then used primers specific to BES1 and VSG-8 on BES12, which is silent in VSG up and VSG upDEX prior to induction (Fig. S8). No VSG-8 product was amplified in the uninduced samples, as expected, but products were amplified in both the 196-h induced samples in both VSG up and VSG upDEX (Fig. S8). This suggests that repair of a DSB in BES1 occurs using silent VSGs that are gene converted into the active BES. The multiple-banding pattern seen in the induced samples suggests gene conversion events using different lengths of 70-bp repeats for homology. Interestingly, we noted a preference for intrachromosomal repair in both cell lines. VSG-2 is expressed from chromosome 6, and 8% and 11% of enriched genes aligned to chromosome 6 in the VSG up and VSG upDEX cells, respectively. The preference for intrachromosomal repair may be associated with the proximity of the repair template to the site of the DSB.

DISCUSSION
Here, we report that the expression of a second VSG is not tolerated during a VSG switching, with stringent mechanisms to reinforce monoallelic expression in antigenic variation. We disrupted single antigen expression through the integration of an ectopic copy of VSG-5 approximately 7 kbp upstream of a de novo telomere at an rDNA promoter, which drives transcription of VSG-5 expression while avoiding competition with the native BES promoter (19,60). Using an rDNA promoter to replace a BES promoter does not disrupt allelic exclusion (61), and the ectopic copy of VSG-5 is also subject to allelic exclusion (19), with both VSG-2 and VSG-5 present at the cell surface in equal levels. Expression of VSG-2 and VSG-5 was stable, even in the absence of drug selection pressure (Fig. S1B), likely due to the de novo telomere seed, which allows for recruitment of the VEX complex and stable transcription of VSG-5.
We initially anticipated that expression of VSG-5 would be coincident with the nucleolus, given the integration of the expression construct at a ribosomal promoter. Our data suggest that the ectopic VSG-5 locus is expressed from the same position in the nucleus as the active BES and that this leads to an increase in the size of the VEX2 focus at this position. The increase in size is dependent on the expression of two VSGs, and we show, using DNA FISH, that VSG-2 and VSG-5 transcription is coincident. Indeed, two VSGs expressed from distinct BESs colocalize to a single ESB (42). It has previously been reported that ectopic expression from an rDNA locus is achievable but unable to compensate for loss of active VSG expression (41); we suggest that the recruitment of the VEX complex by the telomere is necessary for the localization of the ectopic VSG locus to the active VSG transcription compartment. Previous work by Batram et al. has shown that inducing high levels of expression of a second VSG from an rDNA spacer using a T7 RNA polymerase promoter led to initial attenuation of the actively expressed BES VSG in a process that is reversed in around 5 days (48). We suggest that in our cell line, the stable expression of VSG-2 and VSG-5 is supported by their proximity to a telomere and their ability to recruit telomeric chromatin components. In the context of this study, we have assessed only one pair of VSG genes but believe our findings would be supported using additional pairs of VSG genes.
What leads to the transcription-dependent break-induced loss of VSG-5 is unknown, but one hypothesis is that following a DSB, VSG-5 transcription increases to compensate for the loss of VSG-2. This could lead to genomic instability through the formation of R-loops (62), which, in T. brucei, have been shown to form in both the active BES (29) and at rDNA loci (30). R-loop-induced genomic instability can result in chromosome rearrangements (63) and could account for the loss of VSG-5. Alternatively, increased transcription of VSG-5 may result in clashes between the transcription and replication machinery, leading to DNA DSBs that are repaired by recombination and resulting in genetic changes (64). In the DEX cells, we observed a significant decrease in ssDNA in the first 6 to 9 h following a DSB. The proximity of the available templates has been shown to influence the frequency of recombination (6), and it is possible that after a DSB in the BES, repair is initiated using the ectopic VSG-5 gene, which is in close spatial proximity; however, limited homology results in an unsuccessful recombination event which is eventually degraded, as is seen in fission yeast (65). Recombination intermediates could temporarily stall ssDNA formation until the homology search is reestablished. Despite the delay in the timing of DSB repair, the mechanism of repair at the BES continued undisrupted with the preference for repair using the 70-bp repeats maintained and the access to the genomic archive of VSG genes unabated.
Two studies have previously reported frequent deletion of an active BES during VSG switching. Rudenko et al. used selection markers within the silent BES to activate a transcriptional switch and observed deletion of the originally active BES (66), with up to 100 kbp of telomeric sequence lost and large changes in chromosome size seen. Cross et al. reported a similar finding using a negative selectable marker for BES inactivation, where they found that the majority of switching events were coupled to the deletion of the entire previously active BES, with large changes in chromosome size also seen (67). The mechanism of chromosome truncation is unclear; however, it was proposed that they could arise from either random chromosome breakage and healing as is seen in Plasmodium falciparum (68, 69), interchromosomal recombination, or a failed recombination event that is destroyed by exonucleases (67). Both studies propose that there is a strict requirement for the inactivation of the previously active VSG prior to activation of a new one, either by a transcriptional switch or deletion of the previously active BES (66,67). The mechanisms by which expressed VSG genes are counted to ensure monoallelic expression are not fully understood; however, it has been proposed that sequences in the VSG gene and downstream sequences, including the telomeric TTAGGG repeats, which share a high degree of homology with other BESs, could act as repressive elements in a VEX-dependent manner (19). The ectopic VSG-5 expression construct used in this study contains the native VSG-2 39 UTR and associated downstream sequences and a de novo telomere which facilitates recruitment of the VEX complex, allowing it to escape VSG silencing (19). Given the close proximity of the ectopically expressed VSG-5 and native VSG-2 alleles in the nucleus, an I-SceI-induced DSB may cause a transient release of the VEX complex from both, but only recruited back to the repaired BES expressing a new VSG gene. This would leave the ectopic locus depleted of the VEX complex, impairing expression.
Here, we report that monoallelic expression is rapidly restored during a VSG switching event. We propose that this occurs in order to prevent multiple VSG switching events from occurring at once, preserving monoallelic expression. It is of note that a number of studies into VSG switching have been unable to distinguish between BES GC, in which an entire BES is duplicated and replaces the previously active one, and an in situ switch coupled to deletion of the originally active BES (35,39,(70)(71)(72). The resulting structure of the active BES is identical in both cases, and the loss of the previously active BES will only be identified by examining chromosome truncations. Therefore, it is possible that deletion of the previously active BES occurs frequently during a VSG switch; however, these events have so far gone largely undetected.
T. brucei evasion of the mammalian host is dependent on both the monoallelic expression of a single surface antigen in order to avoid clearance by the host immune system (18) and the periodic switching of the single expressed VSG species. Here, we demonstrate that monoallelic expression is maintained at an unexpected level during antigenic variation, which has important implications for our understanding of how the parasite evades immune detection.

MATERIALS AND METHODS
T. brucei cell growth and manipulation. T. brucei Lister 427 cell lines were grown in HMI-11 medium at 37.4°C (73) with 5% CO 2 , and the density of cell cultures was measured using a hemocytometer. Wild-type (WT) cell lines used in this study were T. brucei Lister 427 MITat 1.2 clone 221a (VSG-2 WT or VSG221 WT) and MITat 1.5 (VSG-5 WT, or VSG118 WT). Transformation of cell lines was carried out by centrifuging 2.5 Â 10 7 cells at 1,000 Â g for 10 min at room temperature. The cell pellet was resuspended with 10 mg linearized DNA in 100 mL warm cytomix solution (74), placed in a cuvette (0.2-cm gap), and transformed using a Nucleofector (Lonza) (X-001 function) (75). Transfected cells were recovered in 36 mL of warm HMI-11 at 37°C for 4 to 6 h, after which cells were plated out in 48-well plates with the required drug selection. G418 selection was carried out at 2 mg/mL, puromycin at 2 mg/mL, blasticidin at 10 mg/mL, and tetracycline (Tet) at 1 mg/mL. Puromycin, phleomycin, hygromycin, blasticidin, and G418 selection was maintained at 1 mg/mL. The VSG up cell line has been described previously (26). The VSG upDEX cell line was generated by transfecting the VSG up cell line with the p 5 NTMF reporter construct (19) and clones with detectable VSG-5 expression selected. The VSG upDEX_OFF cell line was generated in the same way; however, cells that do not express VSG-5 were selected.
To measure cumulative cell growth, a hemocytometer was used, and cells were diluted to 1 Â 10 5 cells/ mL every 24 h. DSB induction was carried out with 1 mg/mL tetracycline. A pair of clones was analyzed for VSG upDEX . Generation of subclones was carried out by clonogenic assays in which VSG up and VSG upDEX cells were plated out in 96-well plates. For the uninduced (2Tet) plates, 32 cells were seeded per 96-well plate, and for the induced (1Tet), 480 cells were seeded per plate and induction carried out using 1 mg/mL tetracycline for 4 to 7 days. Cells grown under induced conditions had puromycin and G418 drug selection removed. DSB induction was confirmed in all induced subclones using a puromycin-sensitivity assay in which induced subclones were grown in the absence and presence of 2 mg/mL puromycin, and only those clones that did not grow in the presence of puromycin were considered sensitive.
Immunofluorescence analysis. Immunofluorescence was carried out according to standard protocols. In brief, cells were fixed in 1% (vol/vol) formaldehyde and incubated on ice for 30 min. Fixed cells were centrifuged for 1 min at 6,000 rpm and washed with 1 mL ice-cold phosphate-buffered saline (PBS) twice. Cells were settled onto poly-L-lysine-treated slides for up to 30 min and washed 3Â for 5 min in PBS. Blocking was carried out for 15 min in 50% fetal bovine serum (FBS) in PBS, and all antibody dilutions were in 3% FBS. For visualization of internal antigens, antigen retrieval was carried out prior to blocking by incubation at 95°C for 1 min in antigen retrieval buffer (100 mM Tris, 5% urea, pH 9.5) as described in reference 19. Primary antisera used were rat a-VSG-2 (1: 50,000), rabbit a-VSG-5 (1: 50,000), rabbit a-myc (Cell Signaling; catalog no. 71D10) (1:200), and mouse a-L1C6 (1:100) (53), a gift from the Bastin laboratory. Secondary antisera used were goat a-rat rhodamine conjugated (1:100), goat a-mouse fluorescein isothiocyanate (FITC) conjugated (1:100), both Invitrogen, and goat a-rabbit Alexa Fluor 555 and goat a-mouse Alexa Fluor 488, both used at 1:2,000 dilution and provided by Thermo Fisher. Cells were mounted in Vectashield (Vector Laboratories) containing 4,6-diamidino-2-phenylindole (DAPI). DAPI-stained nuclei and kinetoplast were used as cytological markers to determine the cell cycle stage (76). Cells with one nucleus and one kinetoplast are G 1 , one nucleus and one elongated kinetoplast (IN:1eK) are S phase, one nucleus and two kinetoplasts (1N:2K) are G 2 / M, and cells with two nuclei and two kinetoplasts are postmitosis. Images were acquired using a Zeiss Axio Imager Z2 epifluorescence microscope combined with an Axiocam 506 monocamera. Images were processed using ImageJ, version 2.1.0 (77). Statistical analysis was carried out in GraphPad Prism, version 9.
Flow cytometry. Flow cytometry experiments were performed as described in reference 19. Approximately 1 Â 10 7 cells were collected for analysis for each sample. VSGs were detected using mouse a-VSG-5 (1:10,000) and rat a-VSG-2 (1:10,000) primary antibodies. Secondary antibodies used were goat a-rabbit Alexa Fluor 488 (Thermo Fisher) and goat a-rat Alexa Fluor 647 (Thermo Fisher), both at 1:2,000 dilution. Samples were analyzed on a MoFlo Astrios EQ cell sorter, and data were analyzed using FlowJo software.
DNA FISH. DNA probes were generated by PCR using standard conditions with Phusion high-fidelity DNA polymerase (Thermo Scientific) and labeled using FISH Tag DNA multicolor kit (Invitrogen). For the Pseudo gene probe, a 1,328-bp fragment was amplified with PSEUDO_FISHF (AACAGCGCCGAATTTAATGCAAT) and PSEUDO_FISHR (GTTTCGCCTTCCCATTTGCAT) primers and labeled with Alexa Fluor 488 dye (Invitrogen). For the NTP probe, a 1,200-bp fragment was amplified from the pTMF-8U2 plasmid using NTP_FW CTAACAGGCACGGAAGCCTA and NTP_RV CTCCAAAGCAAACATGCAGA primers and labeled with Alexa Fluor 555 dye. Synthesis and labeling of the amine-modified DNA were done following the manufacturer's conditions (Invitrogen) using 1 mg of purified amplification product. For hybridization, each DNA probe was used at a final concentration of 10 ng/mL in hybridization buffer (50% formamide, 10% dextran sulfate, 2Â SSC [1Â SSC is 0.15 M NaCl plus 0.015 M sodium citrate]) containing herring sperm DNA (10 mg/mL; Sigma) and yeast tRNA (10 mg/mL; Invitrogen).
FISH protocol was adapted from Budzak et al. (42). Cells were fixed directly in medium with ice-cold paraformaldehyde at a final concentration of 4% for 10 min in ice. Cells were washed three times in cold 1Â PBS buffer and settled within previously adhered Gene Frames (Thermo Scientific) on Polysine microscopy slides for 30 min. Cell permeabilization was made with 0.1% Nonidet P-40 for 10 min and washed three times in 1Â PBS. RNase treatment was carried out with 50 mg/mL RNAse A (Thermo Scientific) for 1 h at 37°C. We added 25 mL of probe mix to the slide before placing the Gene Frames coverslip; denaturation was carried out at 82°C for 5 min, followed by overnight incubation at 37°C in a wet chamber. After hybridization, slides were washed in DNA wash buffer (50% formamide, 2Â SSC) for 30 min at 37°C, followed by washes in 1Â SSC, 2Â SSC, and 4Â SSC at 50°C for 10 min each. Slides were dipped in water and air-dried, followed by counterstaining with 1 mg/mL Hoechst for 10 min in the dark. Samples were rinsed in 1Â PBS and mounted in SlowFade Gold antifade (Invitrogen). Images were acquired using a Zeiss Axio Imager Z2 epifluorescence microscope combined with an Axiocam 506 monocamera. Images were processed using ImageJ, version 2.1.0 (77). Statistical analysis was carried out in GraphPad Prism, version 9.
Quantitative resection assay. Quantification of ssDNA by qPCR was carried out using an adapted version of reference 57. In brief, DNA was harvested at 0, 3, 6, 9, and 12 h following growth in 1 mg/mL tetracycline. Five hundred nanograms of extracted DNA were digested with either HindIII or mock digested (no enzyme) overnight at 37°C. For the quantitative PCR (qPCR), primers used were VSG21b forward (AGGCCAAGAAAGCGCTAACA) and VSG21b reverse (CCACTGGCTGCTCGGATATG) (55). Luna Universal qPCR master mix (New England Biolabs) was used with 600 pM primer mix and 5 ng DNA per reaction. The PCR cycling conditions were 95°C for 3 min and then 40 cycles of 95°C for 10 s and 55°C for 30 s on a thermal cycler. The quantification cycle (DC q ) was calculated by subtracting the average C q of the mock digest from the digested C q . The percentage of ssDNA was calculated using the following formula: % resection = 100/ [(1 1 2 Dcq )/2], assuming 100% efficiency of the I-SceI meganuclease. Three technical replicates were carried out for each experiment. Statistical analysis was carried out in GraphPad Prism, version 9.
VSG sequencing. For both the VSG up and VSG upDEX cell lines, 5 Â 10 7 cells were harvested from uninduced and 7-day DSB-induced cells, and RNA was extracted using an RNeasy RNA extraction kit (Qiagen). For the VSG up cell line, experiments were carried out in duplicate, and VSG upDEX experiments were carried out in triplicate. First-strand synthesis was performed using 500 ng of RNA, SuperScript IV reverse transcriptase (Thermo Fisher), and 200 nM "All-VSG 39-UTR" primer (GTGTTAAAATATATC) (59) that binds specifically to the conserved 14-mer in the VSG 39 UTR. The product was cleaned up using AmPureXP beads (Beckman Coulter). To specifically amplify all of the VSGs expressed in a population, conserved sequences in the 59 spliced leader and 39 UTR of every VSG mRNA were used (59). VSG cDNA was amplified by PCR using 1 mg of cDNA, 0.2 mM deoxynucleoside triphosphates (dNTPs) (NEB), 1Â PCR buffer (NEB), Phusion DNA polymerase (NEB), 200 nM spliced leader (SL) forward primer (ACAGTTTCTGTACTATATTG), and 200 nM SP6-14mer reverse primer (GATTTAGGTGACACTATAGTGTTAAAATATATC) (59). The PCR conditions were 5 cycles of 94°C for 30 s, 50°C for 30 s, and 72°C for 2 min, followed by 18 cycles of: 94°C for 30 s, 55°C for 30 s, and 72°C for 2 min carried out on a thermal cycler machine. The VSG PCR products were then cleaned up using AmpPureXP beads (Beckman Coulter) according to the manufacturer's protocols. Sequencing was carried out using a minimum of 4 mg product per sample at Beijing Genomics Institute (BGI) using the BGISEQ-500 platform in paired-end mode. The number of million reads per library was 9.21 for VSG up uninduced replicate 1, 9.29 (6) with the minichromosomal VSGs added from the VSGnome (7) using Bowtie2 (80) with "-very-sensitive" parameters and BAM files generated using SAMtools (80). The overall alignment was 95.21% for VSG up uninduced replicate 1, 95.24% for VSG up uninduced replicate 2, 97.51% for VSG up induced replicate 1, 97.08% for VSG up induced replicate 2, 97.08% for VSG upDEX uninduced replicate 1, 97.83% for VSG upDEX uninduced replicate 2, 97.81% for VSG upDEX uninduced replicate 3, 98.41% for VSG upDEX induced replicate 1, 98.52% for VSG upDEX induced replicate 2, and 99.08% for VSG upDEX for VSG upDEX induced replicate 3. Reads aligning to each transcript were acquired using featureCounts (82), and EdgeR (83) was used to perform differential expression analysis on all genes. The R script used to generate the volcano and genome plots and perform differential genome analysis was adapted from reference 55.
Data availability. Whole-genome sequencing data have been deposited onto the ENA under study accession number PRJEB43910 and unique study name ena-STUDY-INSTITUT PASTEUR-24-03-2021-19:28:45: 086-149. Transcriptomic data have been deposited onto the ENA under study accession number PRJEB49263.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.