Evolutionary history of transformation from chronic lymphocytic leukemia to Richter syndrome

Abstract

Richter syndrome (RS) arising from chronic lymphocytic leukemia (CLL) exemplifies an aggressive malignancy that develops from an indolent neoplasm. To decipher the genetics underlying this transformation, we computationally deconvoluted admixtures of CLL and RS cells from 52 patients with RS, evaluating paired CLL-RS whole-exome sequencing data. We discovered RS-specific somatic driver mutations (including IRF2BP2, SRSF1, B2M, DNMT3A and CCND3), recurrent copy-number alterations beyond del(9p21)(CDKN2A/B), whole-genome duplication and chromothripsis, which were confirmed in 45 independent RS cases and in an external set of RS whole genomes. Through unsupervised clustering, clonally related RS was largely distinct from diffuse large B cell lymphoma. We distinguished pathways that were dysregulated in RS versus CLL, and detected clonal evolution of transformation at single-cell resolution, identifying intermediate cell states. Our study defines distinct molecular subtypes of RS and highlights cell-free DNA analysis as a potential tool for early diagnosis and monitoring.

Extended Data Fig. 1.. Clonal deconvolution process.

a, distinguishing RS from CLL clones after inferring subclonal composition of paired CLL and RS samples. b, inferring phylogenetic tree from cancer cell fraction using PhylogicNDT. c, sample composition d, mapping copy number variations to clones using CopyNumber2Tree.

Extended Data Fig. 2.. Phylogenetic reconstruction and somatic genomic alterations.

For each of the patient trios with WES data, the left panel shows the phylogenetic tree tracing the transformation history from CLL to RS. The magenta frame denotes the Richter clones. The middle top panel represents the subclonal composition inferred after clustering alterations with similar cancer cell fractions as previously reported4. The middle bottom panel indicates the timeline with RS and CLL sampling time and CLL therapeutic lines. (F, fludarabine; C, cyclophosphamide; R, rituximab; P, pentostatin; O/Ofa, ofatumumab; HDMP, high-dose methylprednisolone; A, alemtuzumab; Auto, autologous stem cell transplantation; CLB, chlorambucil; B, bendamustine; CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; ESHAP, etoposide, methylprednisolone, high-dose cytarabine, cisplatin; CHP, cyclophosphamide, doxorubicin, prednisone; Len, lenalidomide; Ob, obinutuzumab; idela; idelalisisb; D, dexamethasone; Adria, adriamycin). The right panel is composed of allelic fraction plots and allelic copy ratio plots showing clonal assignment of somatic copy number events to CLL and RS clones. Cases with whole genome doubling in Extended Data Fig. 2 and clonal unrelated cases in Extended Data Fig. 3.

Extended Data Fig. 3.. Phylogenetic reconstruction and somatic genomic alterations.

For each of the patient trios with WES data, the left panel shows the phylogenetic tree tracing the transformation history from CLL to RS. The magenta frame denotes the Richter clones. The middle top panel represents the subclonal composition inferred after clustering alterations with similar cancer cell fractions as previously reported4. The middle bottom panel indicates the timeline with RS and CLL sampling time and CLL therapeutic lines. (F, fludarabine; C, cyclophosphamide; R, rituximab; P, pentostatin; O/Ofa, ofatumumab; HDMP, high-dose methylprednisolone; A, alemtuzumab; Auto, autologous stem cell transplantation; CLB, chlorambucil; B, bendamustine; CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; ESHAP, etoposide, methylprednisolone, high-dose cytarabine, cisplatin; CHP, cyclophosphamide, doxorubicin, prednisone; Len, lenalidomide; Ob, obinutuzumab; idela; idelalisisb; D, dexamethasone; Adria, adriamycin). The right panel is composed of allelic fraction plots and allelic copy ratio plots showing clonal assignment of somatic copy number events to CLL and RS clones. Cases with whole genome doubling in Extended Data Fig. 2 and clonal unrelated cases in Extended Data Fig. 3.

Extended Data Fig. 4.. Putative RS driver genes.

a-x, individual protein mutation maps for selected putative Richter drivers, showing gene mutation subtype (for example, missense), position and evidence of mutational hotspots. Panels were generated by using the cBioPortal for Cancer Genomics tool.

Extended Data Fig. 5.. RS sCNAs and genomic clustering.

GISTIC2-defined recurrent copy number gains (red, left) and losses (blue, right) are visualized for focal events for RS samples (a) and RS clones (b) (RS samples with CLL events subtracted, bottom). Chromosomes are shown on the vertical axis. Green line denotes a near significant q value of 0.25 and significant events (q<0.1) are annotated in text along with putative driver genes contained within the peak (Supplementary Table 5) c, NMF clustering of RS with DLBCL (304 de novo DLBCL samples shows clonal related RS clusters separately from DLBCL and closes to DLBCL from C2. Clonal unrelated RS clusters across DLBCL subtypes and separate from RS. Samples were annotated for clonal relationship (related RS, gray, unrelated RS, black), cohort (DLBCL, light purple; RS, dark purple) and DLBCL clusters (C1, purple; C2, yellow, C3, pink, C4, blue, C5, green). d, NMF clustering of RS shows 5 distinct genomic subtypes of transformation

Extended Data Fig. 6.. Transcriptome supports distinct RS molecular subtypes.

a, Supervised clustering of transcriptome data from 36 RS patients by molecular subtype highlights differentially regulated genes in subtype 1 and 3 (Supplementary Table 8). Samples are annotated for cohort (Discovery, pink; Validation, yellow), clonal relationship (unrelated, black, related, white), and sample purity by WES (green gradient). b, Unsupervised consensus clustering of RS transcriptome data (n=36) shows 5 clusters. (Discovery, pink; Validation, yellow), RS molecular subtype (1, purple; 2, blue; 3, orange; 4, green; and 5, pink), and sample purity by WES (green gradient).c, 5 × 5 table showing association between molecular subtype of RS and unsupervised transcriptome clusters (2 sided Fisher’s exact test, P=0.038) d, Kaplan-Meier curve showing OS of clonal unrelated RS compared to clonal related RS. P value is log rank (2 sided Mantel Cox).

Extended Data Fig. 7.. Phylogenetic trees showing CLL and RS clones from WGS of paired samples.

a, Phylogenetic tree and CCF plot for 9 patients based on WGS data showing clonal related RS (magenta box). b, Phylogenetic tree and CCF plot for 2 patients based on WGS demonstrating clonally unrelated RS c, Representative phylogenetic trees and CCF plot for 3 patients from UK cohort based on WGS.

Extended Data Fig. 8.. WGS Circos plots with or without chromothripsis.

a, chromothripsis and kataegis in RS sample (Pt 42) with whole genome doubling. Circos plots showing structural variants (interchromosomal, blue; deletion, red; inversion, yellow; tandem duplication, green; long range, teal), allelic copy number (middle), rainfall plot with kataegis regions (red) and chromosomes (outside). Adjacent rainfall plots show kategis regions (C to G, red; C to T, yellow; C to A, teal) with corresponding allelic copy number fragmentation. b, Circos plots from RS WGS samples showing structural variants (interchromosomal, blue; deletion, red; inversion, yellow; tandem duplication, green and long range, teal), allelic copy number (middle), rainfall plot with kataegis regions (red) and chromosomes (outside). SVs impacting known genes and translocation partners are labeled (Supplementary Table 7k).

Extended Data Fig. 9.. Single cell processing and transcriptome analysis of RS samples at single cell resolution.

a, flow sorting strategy for RS single-cell samples. Flow sorting to separate RS and CLL cells by size for Patient 19 and Patient 41 (lymph node, LN; peripheral blood, PB; bone marrow, BM). Flow sorting viable cells for Pt 43, Pt 4 and Pt10. Representative flow plots below demonstrate CLL and RS cells were included in sorted population. b, B-cell receptor (BCR) clonotypes plotted for RS and CLL clusters on UMAP visualization. c, Representative example from patient 10 showing CNVsingle identifies malignant B cell clusters (5 and 6) separate from immune cell clusters (0,1,2,3,4,7,9). d, UMI/cell and Gene/cell plots for CLL and RS single-cell clusters. RS demonstrates higher UMI/cell (P<2.2 × 10–16 see Methods, Supplementary Table 8). e, RNA inference of directional trajectories is shown on UMAP visualization for Pts 43 and 10. f, copy number variation heatmap inferred in each cluster from scRNA-seq data using our CNVSingle algorithm for Pts 43 and 10 (Methods)

Extended Data Fig. 10.. Single-cell transcriptome and copy number analysis of RS patients.

UMAP visualization of single-cells from patient 4 (left) with associated allelic copy number ratio plot inferred by CNVsingle (top right) and RS WES (bottom right). b, UMAP visualization of CLL and RS cells from Patient 18 (left top panel) with flow-sorting annotations (right top panel). Inferred CNAs from CNVSingle (bottom panel) are shown as heatmap with CLL (green) and RS (pink) events highlighted. c, UMAP visualization of CLL and RS cells from Patient 41 (left top panel) with flow-sorting annotations (right top panel). Inferred CNAs from CNVSingle (bottom panel) with CLL (green) and RS (pink) events highlighted. d, Plasma of patient 44 shows RS specific sCNVs on chromosome 9 and 13 leading up to RS diagnosis, which are not reflected in circulating CLL e, Plasma of patient 99 at the start of CLL-directed therapy (top) and just ahead of diagnosis of RS (bottom) during CLL response. f, Chromothripsis in post-transplant RS plasma cfDNA at time of relapse (Pt 112). g, Plot showing allele frequency of RS (purple) and CLL (green) mutations in RS WES (bottom) and plasma cfDNA WES (top) for patient 5 (top) and patient 44 (bottom)

Fig. 1.. Developing an analytic framework for detecting Richter Syndrome (RS)-specific clones.

a, Disease course of 53 RS patients from CLL diagnosis in relationship to lines of therapy and sample collection. b, Disease course of 44 of 45 RS validation cohort patients from CLL diagnosis in relationship to lines of therapy and sample collection (1 patient with missing data). c, Computational schema for deciphering CLL and RS clones within RS biopsy samples. d, Inset shows labeled sample phylogenetic tree with associated sample cancer cell fraction (CCF) plot. Phylogenetic trees with CCF clustering, clonal abundance and associated patient disease course in representative clonally unrelated (e) and related (f) cases.

Fig. 2.. The landscape of putative driver mutations in RS.

a, Phylogenetic tree schema demonstrating clones comprising RS history (gray box) and RS-specific clones (magenta box) (ANC, ancestor clone; INT, CLL intermediate clone; DIV, CLL divergent clone; RS, RS clone). b-c, Somatic mutation information across the putative driver genes and recurrent somatic copy number alterations (rows) for 52 RS patients (columns) that underwent WES, ranked by frequency (right) for both (b) RS history, alterations detected in RS cells and (c) RS clones, alterations acquired at transformation. Samples were annotated for sequencing site (DFCI/Broad, red; German CLL Study Group (GCLLSG), blue; French Innovative Leukemia Organization (FILO), yellow), IGHV status (maroon, mutated; peach unmutated), and clonal relationship (black, related; white, unrelated). Light blue frequency bars adjacent to RS history represent frequency in validation cohort of each alteration (n=45) d, GISTIC2.0 plots showing arm level (right panel) and focal (left panel) amplifications (red, top) and deletions (blue, bottom) for RS samples in the combined discovery and validation cohorts (n=97). Discovery cohort GISTIC2.0 plots are located in Extended Data Figure 5. e, Frequencies of somatic alterations in CLL clones from related RS cases (n=45, dark green bars) compared to CLL driver frequencies using 2 sided exact binomial test with Benjamini-Hochberg multiple hypothesis testing correction f, RS somatic alteration frequencies (dark purple) compared to DLBCL event frequencies (light purple) from DLBCL cohorts^, using 2 sided exact binomial test with Benjamini-Hochberg multiple test correction. g, Proportion in which a recurrent driver is found as present in CLL^ANC+INT (green) or acquired in RS (purple) across 58 related cases (only drivers affecting at least 4 patients are shown) (Supplementary Table 6). * denotes P<0.05 (McNemar test, one-sided). h, Sankey plot showing trajectories from CLL driver to acquired RS driver. Only driver pairs with at least 4 co-occurrences across the cohort are displayed and tested for statistical significance (Supplementary Table 6). * denotes P<0.05 (Fisher’s exact test, two-sided) and Q < 0.4.

Fig. 3.. Tracing evolution of RS on targeted agent therapy

a, Pathways altered in CLL transformation to RS include CLL phase alterations (light green) and new drivers identified in RS (light purple). sSNV (top shading) and sCNA (bottom shading). b, Trees depicting clonal evolution of CLL to RS in seven select patients who developed RS on novel agents. Recurrent RS drivers indicated in bold. c-d, Evolution of RS from CLL showing clonal composition and absolute tumor burden over time based on serial sampling for two patients. Left panel - a phylogenetic tree with associated driver events. (Magenta square, RS clones). Right panel - relative abundance of CLL in peripheral blood by white blood cell count (1000 cells/microliter) (top) and relative abundance of RS in bottom plot (by PET/CT scan tumor metrics) with clonal evolution dynamics. Pie charts reflect composition of each sampling timepoint. (pink dotted line, sampling time; Top bar, treatment history; PB, peripheral blood; BM, bone marrow)

Fig. 4.. Molecular mechanisms underlying transformation to RS.

a, genomic classification of RS. For 97 patients (columns), 5 patterns of RS identified by consensus NMF clustering are depicted with respective somatic mutations and copy number alterations (rows). Samples are annotated for prior treatments (chemoimmunotherapy, light green; targeted agent, dark green; no prior therapy, white); IGHV status (mutated, brown; unmutated, beige; white, not determined); clonal relatedness (related, gray; unrelated, black; unknown by WES, white); and the presence of whole genome doubling (gray). Fraction genome altered per sample is shown (top). Event frequencies are indicated as blue bars on the right side for each alteration. Genes that met significance for association with a cluster by Fisher’s exact Test (Supplementary Table 7) are highlighted by cluster association (subtype 1, purple; subtype 2, blue; subtype 3, orange; subtype 4, green; subtype 5, red). b, Overall survival according to the RS genomic pattern. Kaplan-Meier curves for each subtype according to color legends. P value is from log-rank (Mantel Cox) testing. c, WES signatures for RS samples from discovery cohort (n=52). d-e, WGS signatures for CLL (c) and RS (d) clones in 10 evaluable patients. IGHV status (mutated, M; unmutated, UM) and clonal relationship (R, related; UR, unrelated) is indicated at bottom. f, Chromothripsis and kataegis in RS sample (Pt 42) with whole genome doubling. Circos plots showing structural variants (interchromosomal, blue; deletion, red; inversion, yellow; tandem duplication, green; long range, teal), allelic copy number (middle), rainfall plot with kataegis regions (red) and chromosomes (outside). Adjacent rainfall plots show kategis regions (C to G, red; C to T, yellow; C to A, teal) with corresponding allelic copy number ratio plot showing corresponding fragmentation.

Fig. 5.. Transformation to RS at single-cell resolution.

a, Heatmap of differentially expressed transcripts with FDR<0.1 and absolute log₂ fold change > 1 in analysis between paired RS and CLL samples from Pts 27, 7, 42, 20 and 24. b, Volcano plot of transcript expression changes in RS compared to CLL. Differentially expressed genes were assessed using limma-voom (Methods) in paired mode using sample read counts. logFC denotes log2FC and P-values are adjusted for multiple comparisons. Pink dots denote select relevant transcripts. c, Schema for assignment of copy number changes to single-cells to enable identification of CLL vs RS cells. d-e, Single-cell data shows transcriptional differences between RS and CLL from Pt 43 in d, and Pt 10 in e, and highlights intermediate states. Phylogenetic tree showing clonal structure of RS from WES data (top left) and UMAP visualization of RS and CLL single-cells (top middle). Heatmap representation of differential regulated genes between clusters (top right) and dot plot showing cluster expression of representative genes in dysregulated pathways (Supplementary Table 9) (purple shading, relative expression; dot size, percent of single-cell cells expressing transcript). Inferred allelic copy number from CNVsingle for each single-cell cluster (bottom) depicted adjacent to WES allelic copy number plots color-coded to show copy number events assigned to CLL and RS clones (Methods).

Figure 6.. cfDNA isolated from plasma of RS patients shows evidence of transformation.

a, Schema showing how RS specific DNA events can be identified separately from cell-free DNA and different from circulating CLL cells. b, cfDNA in RS Pt 38 shows WGD of clonally unrelated RS, which is not seen in circulating CLL disease at time of diagnosis. c, Chromothripsis is observed in cfDNA of RS patients, as demonstrated by plotting the difference between copy number state changes across the genome (Pt 32 top, Pt 5 bottom) d, Allele frequencies for RS (purple) and CLL (green) mutations found in RS WES sample (bottom) and RS plasma sample cfDNA WES (top) for patient 38 (top panel) and patient 8 (bottom panel). e, Plasma from patients shows early detection of RS. Pt 5 (top) shows RS-related WGD and chromothripsis fragmentation 162 days prior to RS diagnosis, which is not seen in corresponding co-sampled CLL cells. Plasma from Pt 20 (bottom panel) examined 181 days prior to RS shows RS-related WGD and sSCNVs which are not seen in co-sampled CLL or in lymph node biopsy taken from prior week. f, sCNAs become detectable prior to post-transplant relapse in Pt 112, as seen by plot of fraction genome altered and corresponding cfDNA samples showing emergence of new sCNVs despite continued remission of circulating and marrow CLL. g, Metrics of RS in cfDNA are plotted for RS samples leading up to diagnosis. Y axis is fragment genome altered, color scale shows presence of chromothripsis, square represents whole genome doubled (WGD) sample and purple outline indicates samples for which RS mutations were detected on WES of cfDNA. CLL samples at left of figure depict 13 samples from 5 relapsed/refractory CLL patients. RS samples (right) show 19 samples divided by time leading up to RS in 14 RS patients. Number of samples per each category is indicated on the figure by N. Dashed lines denote serial samples from same patients. Box plots show median values as horizontal line and whiskers showing maximum and minimum values with boundaries of box showing the interquartile range.