Clonal evolution of hematopoietic stem cells after autologous stem cell transplantation

Abstract

The impact of exogenous stressors, such as cancer chemotherapies, on the genomic integrity and clonal dynamics of normal hematopoiesis is not well defined. We conducted whole-genome sequencing on 1,276 single-cell-derived hematopoietic stem and progenitor cell (HSPC) colonies from ten patients with multiple myeloma treated with chemotherapies and six normal donors. Melphalan treatment significantly increased the mutational burden, producing a distinctive mutation signature, whereas other chemotherapeutic agents had minimal effects. Consequently, the clonal diversity and architecture of post-treatment HSPCs resemble those observed in normal elderly individuals, particularly through the progression of oligoclonal hematopoiesis, thereby suggesting that chemotherapy accelerates clonal aging. Integrated phylogenetic analysis of matched therapy-related myeloid neoplasm samples traced their clonal origin to a single-HSPC clone among multiple competing clones, supporting a model of oligoclonal to monoclonal transformation. These findings underscore the need for further systematic research on the long-term hematological consequences of cancer chemotherapy.

Fig. 1. Overview of study design and cohort.

a, Diagram illustrating the experimental workflow. b, Summarized PID and ND demographics, specific chemotherapeutic agents encountered by the analyzed HSPCs, the quantity of colonies evaluated and the occurrence of t-MN. Bright red boxes denote patients who developed t-MNs. c, Oncoplot of driver mutations and CNAs identified in at least one colony per patient and normal donor. The numeric value within each box represents the count of distinct mutations identified within the same gene. PID, patient ID; ND, normal donor. Panel a created with BioRender.com.

Fig. 2. Analysis of somatic mutations and their signatures in HSPC colonies.

a, Scatter plot of somatic SNVs per colony against age, with the red line and area representing the normal mutation rate and 95% CI from six normal donor samples. b, 95% CI of normal mutation rates estimated from other studies and ours. c, Three distinct SNV signatures deduced from the sequencing of 1,032 colonies. SBS-C is closely related, with a 90% cosine similarity, to the SBS-MM1 signature. d, A stacked bar chart illustrating the frequency distribution of SBS1, SBS5 and SBS-C signatures across all SNVs in individual colonies. e, Phylogenetic trees for PID0003 and PID0004, incorporating mutation signatures. SBS-C is apparent only in mutations acquired later in the patients’ lives.

Fig. 3. Ultrametric phylogenetic trees constructed from postchemotherapy HSPCs.

These trees are based on SNVs identified in individual colonies, excluding those associated with the SBS-C signature. The trees are further detailed at the bottom, indicating the presence of driver mutations (top row) and CNAs (bottom row). a, Trees corresponding to samples that underwent cytotoxic chemotherapy treatments. b, Trees for samples treated with noncytotoxic chemotherapeutic agents. y.o., years old.

Fig. 4. Assessment of clonal diversity in postchemotherapy HSPCs.

a, The Simpson’s evenness index of post-treatment HSPCs from our cohort compared to indices from internal normal control and normal individuals as reported in ref. . b, Comparison of the Simpson’s evenness index between post-treatment HSPCs (n = 12) versus age-matched normal HSPCs (n = 6, samples from ref. and internal normal control combined). An unpaired two-sided t test was performed to assess statistical significance between normal donors and treated HSPCs. Data are presented as mean ± 95% CIs, *P < 0.05. c, Comparison of Simpson’s evenness index between HSPCs treated with cytotoxic chemotherapy (n = 7), noncytotoxic chemotherapy (n = 5) and age-matched normal control (n = 4). P value is obtained from unpaired one-sided t tests. Data are presented as mean ± 95% CIs. Asterisk indicates FDR < 0.05. d, The simulation illustrates the projected Shannon diversity index over time for a population of 100,000 HSPCs, modeled with the Moran model. Each violin plot at each timepoint represents the results of 100 independent simulations of the model. The black line represents scenarios where acquired mutations do not affect fitness; the red line includes some mutations conferring a selective advantage; the blue line indicates the introduction of chemotherapy at approximately ages 35–40 years. e, The simulation displays the Shannon diversity index over time, based on a random sampling of 100 HSPCs from a pool of 100,000, considering the emergence of chemotherapy-resistant mutations (for example, TP53 and PPM1D). FDR, false discovery rate.

Fig. 5. Clone-specific analysis of mutational rate and telomere length.

a, Distribution of SNVs across individual colonies plotted against age, with colonies categorized according to the presence of driver mutations. b, Distribution of telomere length across individual colonies plotted against age, with colonies categorized according to the presence of driver mutations. c, Comparison of mutation rate (SBS1 + SBS5 counts per year) between clades with or without driver mutations and WT colonies. Statistical significance was assessed using an unpaired t test. Single-asterisk indicates FDR < 0.05 and double-asterisk indicates FDR < 0.01. The definition of clades without driver mutations is described in Supplementary Fig. 5. Clades with fewer than three colonies are not shown. Box plots show the median (line), upper and lower quartiles (box) ±1.5× interquartile range (whiskers). d, Comparison of telomere length between clades with or without driver mutations and WT colonies. Statistical significance was assessed using unpaired t test. Single-asterisk indicates FDR < 0.05 and double-asterisk indicates FDR < 0.01. The definition of clades without driver mutations is described in Supplementary Fig. 5. Clades with fewer than three colonies are not shown. Box plots show the median (line), upper and lower quartiles (box) ±1.5× interquartile range (whiskers). For both c and d, PID0002—clade A (n = 12), clade B (n = 7), clade C (n = 8), TET2 p.M496fs (n = 67), TP53 c.-29 + 1G > T (n = 19), n.d. (n = 20); PID0003—PPM1D p.R458X (n = 7), TET2 p.S710fs (n = 10), TP53 p.G245D (n = 4), TP53 p.H179D (n = 28), n.d. (n = 43); PID0004 (first ASCT)—PPM1D p.C478X (n = 16), n.d. (n = 13); PID0004 (second ASCT)—PPM1D p.C478X (n = 40), PPM1D p.R552X (n = 6), PPM1D p.S468X (n = 6), PPM1D p.W427X (n = 3), n.d. (n = 16); PID0005 (second ASCT)—DNMT3A p.D765G (n = 4), TP53 p.E258K (n = 6), TP53 p.R248W (n = 10), TP53 p.R273H (n = 3), n.d. (n = 86); PID0006—DNMT3A p.L547H (n = 4), clade A (n = 11), clade B (n = 4), n.d. (n = 67); PID0008—TP53 p.M237I (n = 5), TP53 p.R110P (n = 4), n.d. (n = 72); PID0010—clade A (n = 10), clade B (n = 12), clade C (n = 8), n.d. (n = 55). e, Scatter plot correlating mutation rate (SBS1 + SBS5 counts per year) and telomere length in each colony. Spearman correlation analysis was performed to assess the relationship between these variables. The shaded region around the regression line indicates the 95% CI for the regression estimate. f Assessment of the contribution of specific mutation signatures on individual driver mutations detected in treated colonies. Mutations are segregated based on the clade expansion. All SBS-C-related mutations were found in colonies with no clade expansion. n.d., colonies with no driver mutations.

Fig. 6. Phylogenetic relationships between post-treatment HSPCs and corresponding t-MN samples.

a, Integrated phylogenetic tree for PID0002 highlighting the MRCA pinpointed to a clone with concurrent TET2, TP53 mutations and 17p LOH. b, Integrated phylogenetic tree for PID0005 where MRCA was identified in TP53 and DNMT3A-mutated clone. c, Integrated phylogenetic tree for PID0006 where the MRCA is identified within a clone possessing a TP53 mutation. d, Integrated phylogenetic tree for PID0008 with the MRCA traced to a TP53-mutated clone. e, Integrated phylogenetic tree for PID0010 showing the MRCA located at a branching point preceding the acquisition of a U2AF1 mutation.

Extended Data Fig. 1. Visual summary of chemotherapy exposure for mobilized PBSCs.

a, Diagram depicting the chemotherapy exposure of PBSCs from patients undergoing a single ASCT, which includes agents administered during induction therapy and any mobilization regimen. b, For patients receiving two ASCTs, this figure outlines the additional exposure of second-timepoint PBSCs to chemotherapies administered during induction, the first mobilization, any maintenance or salvage therapies post-first ASCT, and the mobilization for the second ASCT. Notably, in both scenarios, PBSCs do not encounter chemotherapies applied during the conditioning phase of the transplant. PBSCs analyzed with single-HSPC colony sequencing are specified. The schematics in a and b are created with BioRender.com.

Extended Data Fig. 2. SBS-C signature profiles and driver mutation links in melphalan-treated HSPC clades.

a, The pairwise association plot displays the cosine similarity metrics between pairs of mutation signatures, with the strength of associations represented by color intensity. The actual mutational signature plots for SBS-MM1 and SBS-C are also shown. b, Box plots comparing the proportion of SBS-C signatures among clades with or without driver mutations (n.d.) in PID0003 and PID0004. None of the comparisons showed statistical significance. Box plots show the median (line), upper and lower quartiles (box) ±1.5× interquartile range (whiskers). PID0003: PPM1D p.R458X (n = 7), TET2 p.S710fs (n = 10), TP53 p.G245D (n = 4), TP53 p.H179D (n = 28), n.d. (n = 43), PID0004: PPM1D p.C478X (n = 40), PPM1D p.R552X (n = 6), PPM1D p.S468X (n = 6), PPM1D p.W427X (n = 3), n.d. (n = 16). n.d., no driver mutations.

Extended Data Fig. 3. Phylogenetic trees of HSPCs from six normal donors’ PBSCs.

Ultrametric phylogenetic trees of HSPCs collected from six normal donors of variable ages.

Extended Data Fig. 4. Temporal phylogenetic analysis of HSPC colonies from PID0004 and PID0005.

a,b, This figure depicts the phylogenetic trees integrating colonies sampled at two different timepoints. For PID0004, the two time points are separated by an interval of 3 years (a), while for PID0005 the interval is 15 years (b). Colonies collected at the first time point are represented in red, while those from the second timepoint are denoted in blue, illustrating the clonal evolution over time.

Extended Data Fig. 5. Oncoplot depicting the mutation landscape of 108 HSPC colonies harboring TP53 mutations.

Accompanying driver mutations and CNAs are also displayed. The top bar graph enumerates the total SNVs identified in each colony, while the bottom bar graph presents the estimated telomere lengths across these colonies.

Extended Data Fig. 6. Clone-specific analysis of mutation rate and signatures using genetically engineered mouse model.

a, The experimental schema. Trp53 or Ppm1d mutant cells were transplanted with WT cells in 1:9 ratio. After the engraftment, recipient mice were treated with vehicle or cisplatin. Single-HSPC colonies were generated from bone marrow cells. Each colony was genotyped and analyzed by WGS with 30× depth to detect somatic mutations. The schematics in panel a are created with BioRender.com. b, Bar plot comparing the number of somatic SNVs between WT and Trp53 mutant cells with or without cisplatin treatment. Statistical significance was assessed using an unpaired, two-sided t test. c, Bar plot showing the mutation signature between WT and Trp53 mutant cells with or without cisplatin treatment. d, Bar plot comparing the number of somatic SNVs between WT and Ppm1d mutant cells with or without cisplatin treatment. Statistical significance was assessed using an unpaired, two-sided t test. e, Bar plot showing the mutation signature between WT and Ppm1d mutant cells with or without cisplatin treatment. f, Bar plot describing the numbers of colonies with or without CNAs. Statistical significance was assessed using a two-sided chi-square test. n.s., not significant.

Extended Data Fig. 7. Integrated phylogenetic trees illustrating the clonal evolution of post-treatment HSPCs and corresponding t-MN samples, with a focus on cases where the most recent clonal ancestor (MRCA) could not be determined.

The phylogenetic tree for PID0001 (a), PID0004 (b), PID0007 (c) and PID0009 (d), all displaying the clonal structure without a discernible MRCA.

Extended Data Fig. 8. Relative contribution of the four mutational signatures in individual t-MN samples.

SBS-C was detected in PID0004, PID0007, PID0008 and PID0009. PID0004 received melphalan during induction therapy. PID0008 bone marrow contained myeloma cells, therefore, SBS-C might be derived from myeloma cells treated with melphalan. For PID0007 and PID0009, the exposure to melphalan only occurred during the conditioning therapy, suggesting that t-MN might have arisen from residual bone marrow cells exposed to melphalan conditioning. PID0008 also contained SBS-C, however, the sample also contained persistent myeloma cells, making it difficult to discern whether SBS-C is from t-MN genome or myeloma cell genome.

Extended Data Fig. 9. Illustration of two hypothesized models for the development of t-MNs.

Model 1 depicts the scenario where t-MNs originate from HSCs that have been mobilized and subsequently transplanted. Model 2 represents the alternative pathway where t-MNs develop from residual HSCs remaining in the bone marrow, which were not mobilized and thus are subject to high-dose melphalan conditioning regimens. These schematics are created with BioRender.com.

Extended Data Fig. 10. Clonal evolution of t-MNs.

Illustrative representation of clonal evolution from postchemotherapy HSPCs to t-MNs (PID0008). The figure portrays the parallel evolution of multiple HSPC clones, each harboring different TP53 and DNMT3A driver mutations of varying clonal sizes, interspersed with wild-type (WT) cells. It highlights one clone that develops biallelic TP53 alterations and acquires additional copy number alterations, ultimately undergoing selective clonal expansion with the transformation to t-MNs.