Genomic determinants of antigen expression hierarchy in African trypanosomes

Abstract

Antigenic variation is an immune evasion strategy used by many different pathogens. It involves the periodic, non-random switch in the expression of different antigens throughout an infection. How the observed hierarchy in antigen expression is achieved has remained a mystery^1,2. A key challenge in uncovering this process has been the inability to track transcriptome changes and potential genomic rearrangements in individual cells during a switch event. Here we report the establishment of a highly sensitive single-cell RNA sequencing approach for the model protozoan parasite Trypanosoma brucei. This approach has revealed genomic rearrangements that occur in individual cells during a switch event. Our data show that following a double-strand break in the transcribed antigen-coding gene-an important trigger for antigen switching-the type of repair mechanism and the resultant antigen expression depend on the availability of a homologous repair template in the genome. When such a template was available, repair proceeded through segmental gene conversion, creating new, mosaic antigen-coding genes. Conversely, in the absence of a suitable template, a telomere-adjacent antigen-coding gene from a different part of the genome was activated by break-induced replication. Our results show the critical role of repair sequence availability in the antigen selection mechanism. Furthermore, our study demonstrates the power of highly sensitive single-cell RNA sequencing methods in detecting genomic rearrangements that drive transcriptional changes at the single-cell level.

Fig. 1. A CRISPR–Cas9 induced DSB in the actively expressed VSG-2 induces a VSG switch.

a, Schematic map of BES1 with the cut site (dashed line) at nucleotide position 1140 of the VSG-2 CDS. b, Western blot analysis of Cas9 and γH2A protein expression in cells capable of doxycycline (dox)-inducible expression of Cas9 and an sgRNA targeting nucleotide position 1140 of VSG-2 (sgRNA VSG-2.1140). A wild-type cell line (Lister 427) and a cell line not transfected with the VSG-2 sgRNA but capable of inducible Cas9 expression (2T1^T7-Cas9) served as controls. EF1α served as a loading control (n = 1). c, BLISS coverage around the VSG-2 CDS after 4 h Cas9 induction in the cell line with sgRNA VSG-2.1140, normalized to the BLISS coverage of wild-type cells. Shown is the average of two biological replicates. The light blue box represents the VSG-2 CDS. The on-target DSB position is indicated by a dashed line. d, FACS analysis of VSG-2 expression in sgRNA VSG-2.1140 transfected cells in a time course until 4 days post-Cas9 induction. VSG-2 expression was monitored using a fluorophore-conjugated (Alexa Fluor 488) anti-VSG-2 antibody. A minimum of 10,000 events were analysed per sample. On the left panel, VSG-2 positive and VSG-2 negative cell lines are shown as controls (n = 1). Graphic in a was created using BioRender (https://biorender.com). For gel source data, see Supplementary Fig. 1a. Source data

Fig. 2. SL-Smart-seq3xpress is a highly sensitive and specific scRNA-seq approach.

a, Schematic of SL-Smart-seq3xpress library preparation. b, Comparison of the median number of genes detected by Chromium 10X (data from Briggs et al.), Smart-seq2 (data from Müller et al.) or SL-Smart-seq3xpress. The median UMI–gene count is shown as a number above a dataset and as a bold line. For the Smart-seq2 and SL-Smart-seq3xpress datasets, each dot represents an individual cell. For the Chromium 10X dataset, the line represents the median UMI–gene count. Libraries are downsampled to 75,000 reads per cell. Number of cells analysed: Chromium 10X, 8,599; Smart-seq2, 40; SL-Smart-seq3xpress, 292. c, Detected UMIs versus genes in a SL-Smart-seq3xpress library at increasing read depth. Each single cell is represented by a coloured dot. The median gene versus UMI count for each read depth is represented by the dotted lines and numbers. d, Percentage of VSG-2 transcript counts, relative to the sum of VSG-2 and VSG-13 transcript counts, in Chromium 10X and SL-Smart-seq3xpress single-cell libraries prepared from mixed populations of VSG-2 expressing (P10 cell line) and VSG-13 expressing (N50 cell line) cells. The thresholds for defining a cell as VSG-2 or VSG-13 expressing (above 80% and below 20%, respectively) are indicated by the dotted lines. Total number of cells analysed: Chromium 10X, 185; SL-Smart-seq3xpress, 185. Graphic in a was created using BioRender (https://biorender.com). Source data

Fig. 3. DSBs in the active VSG-2 CDS led to activation of telomere-adjacent VSGs.

a, Schematic map of BES1 indicating cut sites (dashed lines). b, VSG expression in single cells before and after the induction of a DSB in the VSG-2 CDS at nucleotide position 1140, as measured by SL-Smart-seq3xpress, two biological replicates (R). The total number of cells analysed per time point and replicate is as follows: 0 days R1, 312; 0 days R2, 302; 1 day R1, 308; 1 day R2, 283; 2 days R1, 147; 2 days R2, 271; 3 days R1, 305; 3 days R2, 255; 4 days R1, 289; 4 days R2, 286; 10 days R1, 336; 10 days R2, 323. c, Proportion of cells at each time point after DSB induction from b expressing a new VSG from a given genomic location. ‘Unassigned’ refers to newly activated VSGs for which the original location is not known. d, Diagram illustrating the different types of VSG switching mechanism. e, Left, percentage of cells at days 3, 4 and 10 post-Cas9 induction in each biological replicate that switched VSG expression by a given mechanism. Right, same as the left for two more biological replicates with more intermediate time points. f, Time course of transcriptional switcher selection from BES1 (VSG-2) to BES17 (VSG-13) by the addition of neomycin (+neo) and removal of puromycin (−puro) drug selection. ATAC-seq coverage on BES1, BES17 and all other BESs (middle), and transcript reads per kilobase per million mapped reads (RPKM) for their respective VSGs (right). ‘Other VSGs’ refers to VSGs located in BESs other than BES1 and BES17. Horizontal lines represent the mean of two biological replicates. g, Time course of transcriptional switcher selection as in f but from BES17 (VSG-13) to BES1 (VSG-2) by the addition of puromycin (+puro) and removal of neomycin (−neo) drug selection. Source data

Fig. 4. DSB in VSG-8 CDS leads to switching by segmental gene conversion.

a, Schematic map of the recombinant BES1 (expressing VSG-8) with the cut sites (dashed lines) on the VSG-8 CDS. b, De novo assembled VSG transcripts from individual cells after 4 days of DSB induction at nucleotide positions 609 and 1105 of the VSG-8 CDS, aligned against the VSG-8 sequence. Five homologous VSG genes or pseudogenes (potential ‘donors’) found in the T. brucei genome by BLAST search are also included on top of the alignment. Mismatched bases are coloured based on the nucleotide identity. c,d, Recombinant fragments based on de novo assembled VSG transcripts for individual cells after 4 days of DSB induction at nucleotide positions 609 (c) and 1105 (d) of the VSG-8 CDS. The potential ‘donor’ VSG and the integrated stretch length are depicted by the coloured lines. Grey extensions on the lines represent the maximal sequence length that could have been recombined (that is, stretch until the next nucleotide difference between VSG-8 and the possible donor(s)). All single-cell data are derived from two biological replicates per cut site. Source data

Fig. 5. Model of VSG selection mechanism.

Diagram summarizing how a DSB in the VSG CDS seems to be repaired in the presence or absence of a suitable repair template.

Extended Data Fig. 1. γ-H2A IFA before and after Cas9 induction.

IFA of γ-H2A expression in cells transfected with sgRNA VSG-2.1140 either before (No Cas9 expression) or 4 h after Cas9 induction (Cas9 induced). γ-H2A was detected with an anti-γ-H2A antibody. Scale bar, 10 µm; n = 2.

Extended Data Fig. 2. Optimization of the SL-Smart-seq3xpress scRNA-seq pipeline.

a. The library preparation workflows for: the Chromium 10 × 5′ pipeline (left panel); the standard Smart-seq3xpress (SS3xpress) pipeline (middle); and the tailored SL-Smart-seq3xpress (SL-SS3xpress) pipeline (right). BC, barcode; SL, spliced leader. b, Left panel: Representative TapeStation profiles of single-cell libraries (pooled and bead-purified after cDNA dilution) prepared using different oligo(dT) concentrations - 1X, 1/4X, 1/16X - relative to the concentration in the published Smart-seq3xpress protocol. Each condition was tested with at least two independent replicates, each containing between 6 and 48 cells. Right panel: TDE1 Tn5 concentrations -1X, 0.5X, 0.1X, relative to the concentration in the published Smart-seq3xpress protocol - tested to optimize tagmentation. Here, each condition contains 48 cells. c, Optimization of SL-Smart-seq3xpress oligo(dT) and TDE1 enzyme concentrations. Left panel: Transcript diversity of SL-Smart-seq3xpress libraries prepared with varying oligo(dT) and TDE1 concentrations. Transcript diversity was measured as percentage of UMI counts/per reads sequenced/single cell. Right panel: Percentage of UMI-containing reads in SL-Smart-seq3xpress libraries prepared with varying oligo(dT) and TDE1 concentrations. Each dot represents a single cell. 96 cells were sequenced for each condition tested. d, Schematic of the mechanism leading to index hopping and our bioinformatic strategy to remove index hopped reads. e, Number of UMIs detected by SL-Smart-seq3xpress in single cells (blue) or no cell wells (orange) either without (No filter) or with (With Filter) bioinformatic removal of index hopped reads. The ratio of median UMI counts (single cell/no cell) is shown as a number between the two conditions. The median UMI count is shown as a black line. Number of wells analyzed: single cells - 2214, no cells - 120. f, Comparison of the median number of UMIs detected by Chromium 10X (data from Briggs et al.) or SL- Smart-seq3xpress. Elements of this figure have been created in BioRender. Source data

Extended Data Fig. 3. Relative expression levels of VSGs and ESAGs during DSB induction at nucleotide position 1140 of the VSG-2 CDS.

a-d, Single cell transcript levels normalized to spike-in counts. Data is shown separately for each of the two biological replicates. a, Total VSG genes, b, Total non-VSG genes, c, Total BES1 ESAGs and d, Total non-BES1 ESAGs. Source data

Extended Data Fig. 4. DSBs in or just upstream of VSG-2 lead to a very similar switching profile outcome.

a, Schematic map of the terminal region of BES1 with the cut sites (dashed lines). The coordinates are relative to the promoter or to the start of the VSG-2 CDS. b, Western blot showing Cas9 and γ-H2A expression levels before and 24 h post-Cas9 induction with doxycycline. Wild-type Lister 427 cells 24 h after doxycycline addition were used as a negative control. EF1α was used as a loading control. Clones used for scRNA-seq experiments are highlighted in bold (n = 1). c, BLISS coverage tracks on BES1 after 4 h of Cas9 induction in the cell lines with sgRNA VSG-2.782 and BES1.54824, both normalized to the BLISS coverage of wild-type cells. The light blue box represents the VSG-2 CDS and the black box represents the 70-bp repeats. The on-target DSB position for each cell line is indicated by a dashed line. Shown is the average of two biological replicates. d, scRNA-seq results of the time course experiment following DSB induction at nucleotide position 782 in the VSG-2 CDS, showing the proportion of cells expressing a given VSG at each time point for each biological replicate. e, scRNA-seq results before and 4 days after DSB induction at position 54,824 of BES1, showing the proportion of cells expressing a given VSG at each time point for each biological replicate. f, For the experiments in (d) and (e), proportion of switcher cells grouped by the genomic location of the newly activated VSG. Cells expressing a VSG for which the genomic location is unknown were categorized as ‘unassigned’. For gel source data, see Supplementary Fig. 1a. Source data

Extended Data Fig. 5. Heat map of putative recombination sites in BES1.

a, Single-cell sequencing read coverage on BES1 and BES5 showing different VSG switching scenarios for cells that switched to VSG-18 expression following a DSB in the VSG-2 CDS. Read coverage is shown for 8 representative cells, 2 cells per VSG switching scenario: no switching (No switchers, cells expressing VSG-2), switching by recombination around the 70-bp repeats (Recombinant switchers (70-bp repeat)), switching by recombination between ESAGs (Recombinant switchers (between ESAGs)), and transcriptional switching (Transcriptional switchers). b and e, Heatmaps summarizing the transcriptional signal end positions on BES1 for recombinant switcher cells after a DSB at nucleotide position 1140 (b) and 782 (e) of the VSG-2 CDS. The color of each of the squares in the heatmaps represents the fraction of cells, for switchers to a given VSG (rows), for which transcriptional signal ends at the given 5 kb bin of BES1 (columns). VSGs expressed in at least 10 cells were considered. c, Heatmaps summarizing transcriptional signal end position on BES1 (columns) for recombinant switcher cells versus the transcriptional signal start on the BES containing the incoming VSG (rows), after a DSB in the active VSG-2 CDS (at nucleotide position 1140). Each square on the heatmap represents a 5 kb bin of BES1 (columns) and the BES from the incoming VSG (rows). The color of each square represents the fraction of cells for which transcriptional signal ends at that given bin of BES1 and starts at that given bin of the BES containing the incoming VSG. The heatmaps summarize the data from 122 cells switching to VSG-9, 41 cells switching to VSG-11 and 15 cells switching to VSG-13, at 4 days and 10 days after DSB induction. d, Proportion of cells that had switched VSG expression by recombination or transcriptional switch at 4 days and 10 days after DSB induction at nucleotide position 782 in the VSG-2 CDS. Cells for which the switching mechanism could not be clearly determined, are labeled as ‘undetermined’. Source data

Extended Data Fig. 6. DSBs in VSGs with homologous sequences in the genome lead to segmental gene conversion.

a, Read coverage from bulk RNA-seq data on BES1 (left panel) and BES12 (right panel) for a control VSG-2 expressing cell line and the VSG-8 expressing cell line cloned from a DSB induction experiment at nucleotide position 1140 of the VSG-2 CDS; showing that in this cell line VSG-8 was recombined and is being expressed from BES1 (shown is a representative profile of one out of three biological replicates). b, Western blot showing Cas9 and γ-H2A expression levels before and 24 h after Cas9 induction with doxycycline (Dox) for two cell lines – sgRNA VSG-8.609 (clones 1 and 2, left panel, clone 3, right panel) and sgRNA VSG-8.1105 (clones 1 and 2, right panel) - expressing different sgRNAs to generate DSBs at positions 609 and 1105 of the VSG-8 CDS, respectively. The clones used for scRNA-seq experiments are shown in bold. In the left panel, the negative control is the wild-type cell line and the positive control is the sgRNA VSG2.1140 cell line (n = 1). c, BLISS coverage tracks on BES1 after 4 h Cas9 induction in the sgRNA VSG-8.609 and sgRNA VSG-8.1105 cell lines, both normalized to BLISS coverage of wild-type cells. The light blue box represents the VSG-2 CDS and the black box represents the 70-bp repeats. The on-target DSB position for each cell line is indicated with a dashed line. Shown is the average of two biological replicates. d-f, Growth curves following Cas9-based DSB induction for cut sites at (d) nucleotide position 1140 of VSG-2 CDS, shown is the mean ± SD of three biological replicates, (e) nucleotide position 782 of VSG-2 CDS, shown is the mean ± SD of three biological replicates and (f) and nucleotide positions 609 and 1105 of VSG-8 CDS, the values are derived from a single experiment. g, Proportion of cells expressing a given VSG before and 4 days after DSB induction at nucleotide positions 609 and 1105 of the VSG-8 CDS. Data are shown for two replicates per time point. h, Western blot showing Cas9 and γ-H2A expression levels before and 24 h after Cas9 induction with doxycycline (Dox) for two cell lines – sgRNA VSG-11.519 (clones 1,2 and 3, left panel) and sgRNA VSG-11.729 (clones 1,2 and 3, right panel) - expressing different sgRNAs to generate DSBs at positions 519 and 729 of the VSG-11 CDS, respectively. Clones used for scRNA-seq experiments are shown in bold. The positive control is the sgRNA VSG2.1140 cell and the negative control is the wild-type cell line (n = 1). i, Sanger sequencing results for clones derived from a VSG-11 expressing cell line after DSB induction at nucleotide positions 519 and 729 of the VSG-11 CDS. A single clone is shown for each cut. Sanger sequences are aligned against the reference VSG-11, together with a potential ‘donor’ sequence (Tb427_000478400:pseudogene, Tb427v11 genome). For gel source data, see Supplementary Fig. 1a. Source data

Extended Data Fig. 7. FACS gating strategy for SL-Smart-seq3xpress library preparation.

The cells were gated to exclude debris (P1), doublets (P2 and P3), and dead cells (P4). Propidium iodide was used to gate for live cells. The panel shows that P4 (colored orange) was used as a final cell population for sorting into the 384-well plates.