Myeloma

Single-cell genetic analysis reveals the composition of initiating clones and phylogenetic patterns of branching and parallel evolution in myeloma

Article metrics

Abstract

Although intratumor heterogeneity has been inferred in multiple myeloma (MM), little is known about its subclonal phylogeny. To describe such phylogenetic trees in a series of patients with MM, we perform whole-exome sequencing and single-cell genetic analysis. Our results demonstrate that at presentation myeloma is composed of two to six different major clones, which are related by linear and branching phylogenies. Remarkably, the earliest myeloma-initiating clones, some of which only had the initiating t(11;14), were still present at low frequencies at the time of diagnosis. For the first time in myeloma, we demonstrate parallel evolution whereby two independent clones activate the RAS/MAPK pathway through RAS mutations and give rise subsequently to distinct subclonal lineages. We also report the co-occurrence of RAS and interferon regulatory factor 4 (IRF4) p.K123R mutations in 4% of myeloma patients. Lastly, we describe the fluctuations of myeloma subclonal architecture in a patient analyzed at presentation and relapse and in NOD/SCID-IL2Rγnull xenografts, revealing clonal extinction and the emergence of new clones that acquire additional mutations. This study confirms that myeloma subclones exhibit different survival properties during treatment or mouse engraftment. We conclude that clonal diversity combined with varying selective pressures is the essential foundation for tumor progression and treatment resistance in myeloma.

Introduction

The transformation from a normal to a cancer cell comprises a series of biological changes driven by the multistep acquisition of genetic alterations.1 During cancer progression, the founder cell undergoes successive rounds of genetic diversification and selection, resulting in a branching pattern of tumor evolution reminiscent of the Darwinian theory of natural selection.2, 3 Recent molecular cytogenetic4, 5, 6, 7 and massively parallel sequencing studies8, 9, 10, 11, 12 support these ideas and confirm that intratumor heterogeneity is an essential prerequisite for cancer evolution. This clonal architecture shifts in space,9 time and following treatment5, 8 consistent with the hypothesis that cancer is composed of genetically diverse subclones.13 These cellular fractions are heterogeneous in both their genetic makeup and biological features determining the variability in tumor progression, clinical aggressiveness and sensitivity to therapy seen in cancer.

Multiple myeloma (MM) is a good model system to test these ideas of tumor evolution because of the well-characterized transformation process from monoclonal gammopathy of undetermined significance through smoldering myeloma to MM and ultimately to plasma cell leukemia (PCL) where myeloma growth is no longer restricted to the bone marrow.14, 15 The molecular events leading to the generation of the initiating-myeloma propagating cell (iMPC) are driven either via the acquisition of a chromosomal translocation into the Ig loci or hyperdiploidy. Such founder genetic lesions are predicted to be present in all clones, whereas secondary alterations during disease progression are expected to occur in subclonal fractions. Thus, following its initial immortalization the clonal progeny of the iMPC randomly accumulate secondary genetic events, which are selected based on their contribution to clonal fitness;15 the net result is disease progression and transition between the critical thresholds defining each stage of myeloma.

One frequent secondary lesion in myeloma is deregulation of the RAS/MAPK pathway by activating mutations in KRAS/NRAS/BRAF.11, 16 On the basis of the size of subclones carrying these and other mutations, the presence of intratumor heterogeneity in myeloma has been inferred.5, 11, 17, 18 Nevertheless, it is clear that to delineate the exact phylogeny of subclonal tumor populations, ‘deep-sequencing’10, 19 or close genetic examination of single cells is required.4, 11, 20 In this study we performed single-cell analyses based on the results of whole-exome sequencing (WES) to establish the myeloma phylogeny in t(11;14) patients carrying KRAS/NRAS/BRAF mutations.

Materials and methods

Patients

This study was performed in presentation myeloma samples from patients who entered the Medical Research Council Myeloma IX (ISRCTN68454111) and XI (ISRCTN49407852) trials. The study was approved by the National Research Ethics Service. CD138-positive bone marrow plasma cells were selected to a purity >95%, as determined by cytospin, using magnetic-assisted cell sorting (Miltenyi Biotec, Bisley, Surrey, UK). Tumor DNA was extracted using the AllPrep kit (QIAGEN, Manchester, UK); DNA concentration was assessed using Pico-green (Life Technologies, Paisley, UK). Non-tumor DNA was isolated from white blood cells using the FlexiGene kit (QIAGEN). Archival methanol:acetic-acid-fixed single cells and tumor/normal DNA were obtained from six patients. Clinical, cytogenetic, interphase fluorescent in situ hybridization21, 22 and mutation data were available (Table 1). A PCL patient with samples at both presentation and relapse was included (case history, Supplementary Methods).

Table 1 General information, genomic aberration and mutations in the studied sample series

Whole-exome sequencing

WES sequencing was performed in five cases11 and the same approach using Illumina sequencers was done on sample 11/010, a PCL patient with paired samples at diagnosis/relapse and two xenograft-related myeloma samples. Across the 15 human samples included in this study (8 tumor, 7 normal matched) the median depth after deduplication was 59 × (range 36–113) with an average of 82% of the targeted exome covered at a minimum depth of 20 × (Supplementary Table S1 and Supplementary Methods).

Translocation breakpoint definition

The t(11;14) breakpoint was defined for three or six patients in an earlier study of targeted capture sequencing using the SureSelect (Agilent, Wokingham, Berkshire, UK) system by tiling RNA baits across the IGH/IGK/IGL loci.23 Specific custom-made TaqMan-labeled primers at both breakpoint ends were designed following the manufacturer’s guidelines (Life Technologies) to assess amplification in quantitative real-time PCR (qPCR; Supplementary Table S2).

Single-cell sorting

Fixed single cells from each patient were sorted in a FACSAria cell sorter (BD Biosciences, Oxford, UK) using propidium iodide nuclei staining (Supplementary Figure S1). Depending on cell availability, we sorted 73–243 single cells into DNA lysis buffer in one to three 96-well plates (84 cells/plate; Supplementary Table S3). Lymphocytes from a healthy donor were sorted as wild-type and normal copy number controls (10 cells/plate). Finally, two wells/plate were left empty to add bulk tumor and peripheral blood DNA from the same patient (positive/negative controls for mutation detection). In addition, 60 fixed lymphocytes from a healthy donor were sorted in a separate 96-well plate as reference for the definition of copy number thresholds across all interrogated genomic regions. Assay efficiency was calculated using five 2 × dilutions for the amplified DNA of six lymphocytes.

Single-cell multiplex qPCR for genotyping and copy number analysis

A novel approach for single-cell multiplex qPCR analysis was performed (Fluidigm UK, London, UK).20 Briefly, single cells were sorted into lysis buffer followed by specific (DNA) target amplification. This multiplex specific target amplification reaction comprises the simultaneous amplification of target regions of interest using TaqMan PreAmp Master Mix and assays (Life Technologies). Mutation-specific genotyping assays were custom designed following the manufacturer’s guidelines. Predesigned genotyping assays for loci in heterozygosis in all tumor samples24, 25 were used as reference (rs346172 and rs909895). Three different TaqMan copy number assays covering each chromosomal region of interest were used for copy number analysis (Supplementary Tables S2 and S4). The specific target amplification product was diluted and qPCR was performed using the 96.96 dynamic arrays and the BioMark HD (Fluidigm UK). Genotyping assays were tested in replicates (two to four), whereas each multiple copy number assay per region was used in quadruplicates to ensure enough replicates for accuracy in copy number calling.26 Translocations, mutations and copy number aberrations were assessed at the single-cell level using Fluidigm Real-Time PCR Analysis v.3.0.2 software (Fluidigm UK). To estimate copy number values, CopyCaller v.1.0 software (Life Technologies) was used and weighted means of the calculated copy number values for each experimental replicate were determined. Hierarchical clustering was achieved using Pearson’s correlation and average linkage on the Rock platform27 and customized with R. A filtering strategy for wells with low-quality DNA amplification and subclones without a minimum cell number was applied (Supplementary Figure S1 and Supplementary Methods).20

IRF4 high-resolution melting analysis and mutational screening of KRAS/NRAS

Mutational screening for IRF4 c.368A>G was performed in 454 patients. PCR followed by high-resolution melting was tested using the Type It HRM Kit (QIAGEN) on a Rotorgene Q real-time cycler (QIAGEN; Supplementary Methods). Information for KRAS/NRAS mutations was collected from 243/454 patients from WES data and sequencing analyses.11

NOD/SCID mouse transplantation

Female NOD/SCID-IL2Rγnull mice, approximately 6 weeks old, were inoculated with 1 × 106 CD138+ cells from a PCL patient at diagnosis in 20 μl complete RPMI-1640 GlutaMAX medium and monitored for myeloma development over 5 months as before.28 Mice were then culled and myeloma xenograft samples were purified for human CD138+ cells.

Genetic algorithm in the multisample WES analysis

Xenograft tumor DNA was analyzed by WES sequencing using a modified pipeline and compared with the tumor exomes from the patient samples at presentation and relapse. A genetic algorithm was implemented using the GA package for R29 to calculate the most parsimonious assignments of mutations to clonal lineages. Custom population generation, fitness, mutation and crossover functions were written (Supplementary Methods).

Results

Detection of subclonal populations

We first performed WES sequencing in a series of t(11;14) myeloma and used the frequency of mutant reads corrected with copy number information to calculate the cancer cell fraction (CCF) carrying mutations. We generated Kernel density plots to discriminate mutations present in all tumor cells (clonal, CCF>0.75) or in a fraction of tumor cells (subclonal, CCF<0.75). All patients displayed intratumor heterogeneity with individual mutations distributed at high and at low-mid CCF (Figure 1). To capture all putative clones, we selected five to nine non-synonymous mutations (non-synonymous single nucleotide variants) scattered throughout the CCF spectrum, focusing on KRAS/NRAS/BRAF and other potential driver mutations in myeloma-related genes (Supplementary Tables S1, S2 and S4). Next, we sorted single fixed cells from each patient and simultaneously tested for the presence of translocations, mutations and copy number aberrations using qPCR (Supplementary Figure S1). Our single-cell approach reported that mutation frequencies were positively correlated with those from WES sequencing (Supplementary Figure S2), and that each individual case had 2–6 (median 4) major clones at presentation. We then focused on the clonal phylogenetic relationships for each case.

Figure 1
figure1

Subclonal diversity of somatic mutations in myeloma. (af) This patient series carried a modest number of non-synonymous mutations (median 30 non-synonymous single nucleotide variants (NS-SNVs), range 27–54). Panels display Gaussian kernel density plots, which indicate the frequency of cells carrying all acquired mutations detected by WES sequencing. Variant allele frequency (CCF, x axis) is calculated by adjusting mutant allele burden by copy number of the loci mutated. Mutations may be in major (CCF>0.75) or minor (CCF<0.75) subclones. n, Number of total non-synonymous variants in the case. The y axis displays the density of mutations present at a specific CCF. Vertical lines show the clonal frequency for the mutant genes selected for single-cell genetic analysis. §, Mutations initially chosen for single-cell genetic studies, which mutation detection assays failed to work, and thus their frequencies in single-cell analyses could not be reported. See Supplementary Figure S2 for a comparison of mutation frequencies reported by WES sequencing and single-cell genetic analyses.

A pattern of linear evolution

We examined case 90 827 and distinguished 5–6 tumor subclones, wherein each mutation and genomic aberration was acquired in a stepwise linear fashion. Briefly, mutations in 5 genes were analyzed in the 89.6% (216/241) of sorted fixed single cells that passed all filtering thresholds (Figure 2a, Supplementary Figure S3a and Supplementary Table S3). The most recurrent mutation was ATM c.428A>G (100% of tumor cells) followed by KLK8 c.356A>G (94.4%), GMEB1 c.478A>T and POLE c.776G>A (44.4%), and lastly KRAS c.182A>G (24.5%) (Supplementary Figures S2d and S3b). This case had formerly shown +8q and +21q at subclonal levels,24 which we confirmed in single-cell analyses (Supplementary Figure S3c). When considering cells with mutations and copy number aberration data, 86.7% (137/158) displayed +8q and 63.9% (101/158) had +21q (Figures 2b and c). Hierarchical clustering of the filtered data allowed us to define five clones (Figure 2b) and to delineate the most plausible tumor phylogeny. Strikingly, genetic mutations and genomic aberrations were acquired in a linear sequence, where ATM/KLK8 mutations preceded +8q and +21q, which were followed by GMEB1/POLE and KRAS mutations (Figure 2c).

Figure 2
figure2

A linear evolutionary pattern can be seen in myeloma. (a) Thumbnail heatmap of a Fluidigm array used for the study of case 90 827. Each column shows the interrogated genotyping (left) or copy number (right) assay. Each row represents a sorted single cell, which DNA is interrogated for mutation presence/absence and copy number values. Rows are sorted to group cells with the same genetic pattern (clones). Bulk tumor and peripheral blood DNA from the same case, and normal donor lymphocytes, are interrogated below. Amplification intensity is color scale based, from yellow (high DNA content) to blue (low DNA content). Black means no amplification. Zoomed images to cell patterns are shown on the right. Note clone 2 has mutations in ATM/KLK8, whereas clone 4 further carries mutations in GMEB1/POLE, and so does clone 5 with an additional KRAS mutation. Copy number aberration definition is available in Supplementary Figure S3. (b) Hierarchical clustering of 158 single cells, which passed quality and thresholds criteria. Gray, no change; red, mutation; blue, gain. Cells are clustered according to their pattern for mutations and genomic aberrations (+8q/+21q) in five clones. N, cells with a normal pattern. (c) Clonal phylogeny for case 90 827. Clones evolve via a linear evolutionary pattern where mutations (colored stars) and copy number aberrations (+8q/+21q) are acquired in a stepwise process. KRAS c.182A>G (p.Q61R) is the most recent mutation and generates clone 5. Clonal frequencies are depicted in absolute numbers and percentages. See Supplementary Figures S4 and S5 for additional cases with linear phylogenies.

Two additional cases (90 468 and 90 879) displayed a similar pattern of linear evolution. However, their clonal diversity was lower than sample 90 827, as, at presentation, only 2 and 3 clones, respectively, were identified. The predominant clone in each sample accounted for most of tumor cells (83–93%). Likewise to other cases in this series, mutations in BRAF (90468) and in KRAS (90879) occurred later in tumor development (Supplementary Figures S4 and S5).

A pattern of branching evolution

The genetic architecture of the remaining three samples (90 003, 90 482 and 11/010) had 4–6 clones detected at presentation, but in these cases clones were related via a branching phylogeny (Figures 3 and 4), resembling the complex clonal substructure we described previously.11 To investigate the intraclonal heterogeneity of sample 90 482, 2 clonal (PCHD15 c.2272_2273delinsTA and TRPA1 c.2272C>T) and 3 subclonal mutations (ACAD10 c.828C>A, NRAS c.182A>G and STK24 c.1220C>T) were assessed in single-cell analyses (Figures 1c and 3a, and Supplementary Figures S2c and S6). Fluorescent in situ hybridization assays reported subclonal chr13 loss (Δchr13).21 As the STK24 locus is at 13q31.2-q32.3, we aimed to define whether Δchr13 preceded or followed STK24 mutation. Using qPCR analysis of the RB1 locus on chr13, Δchr13 was shown in 66–87% of the filtered tumor cells (Figure 3a, Supplementary Figures S2c and S6c). Combined analysis of both wild-type/mutant STK24 alleles and chr13 copy number data demonstrated that Δchr13 preceded STK24 mutation. It occurred in a subclone of clone 1 (clone 1.1), after PCDH15 and TRPA were mutated but before the acquisition of the remaining mutations and the emergence of the most recent clones (Figures 3b and c). Clone 1.1 is the ancestor for both clone 2, which mutated the STK24 allele located in the single copy of chr13, and clone 3, which instead retained the wild-type STK24 allele but acquired ACAD10 and NRAS mutations (Figure 3c). Case 90 482 demonstrated a branching evolutionary pattern where 4 subclones, with 2 common ancestors and 2 divergent more recent clones, were involved in tumor progression. These branching evolutionary patterns are a characteristic feature of myeloma and resemble those described in other cancers.4, 9, 10

Figure 3
figure3

A branching evolutionary pattern characterizes multiple myeloma. Genetic study of sample 90 482 suggests that this case evolves through a branching process. (a) Thumbnail heatmap of a Fluidigm array used for this analysis. Layout is as in Figure 2. Zoom image to cell patterns are shown (right). Note clone 2 and clone 3 have common mutations in TRPA/PCDH15 to clone 1 but further accumulate independent mutations in NRAS/ACAD10 (clone 2) or STK24 (clone 3). (b) Hierarchical clustering of the 135 filtered single cells. Gray, no change; red, mutation or wild-type allele (for STK24 wt); deep red, loss of 13q. Four tumor subclones are distinguished, with mutations present in clone 3 (NRAS and ACAD10) but not in clone 2 (STK24). N, cells with a normal and reference pattern. (c) Clonal branching phylogeny considering both mutations (colored stars) and 13q copy number data. Clone 1, which already has TRPA and PCDH15 mutations, further acquires Δchr13, producing clone 1.1. This clone generates two lineages: clone 2, with mutation in STK24, and clone 3, with wild-type STK24 and mutations in ACAD10 and NRAS. Hence, Δchr13 precedes STK24 mutation. NRAS c.182A>G (p.Q61R) mutation occurs in an independent branch and relatively recently in myeloma development. Asterisks mean PCHD15 c.2272_2273delinsTA. Clonal frequencies are depicted in absolute numbers and percentages. See Supplementary Figure S6 for further information.

Figure 4
figure4

Parallel evolution in myeloma provides evidence of the activation of the RAS/MAPK pathway as a convergent phenotype. Close examination of two cases with branching evolutionary patterns illustrates how independent clones, but originated from the same ancestor, acquire activating mutations in the same gene or in genes from the same pathway. (a and b) Single-cell analysis of case 11/010. (a) Results from the used Fluidigm array. Four columns per genotyping assay were interrogated. No copy number assays were used. Layout, as in Figure 2. It is noteworthy that clone 1 only shows positivity for t(11;14), representing the most likely founding myeloma clone in this sample. White arrows point out NRAS c.182A>G mutations (clone 3) that are recognized in different cells to those displaying KRAS c.183A>C (clones 4-6). (b) Clonal phylogeny considering t(11;14) and mutations. Divergent clonal lineages emerged from clone 2 with activating mutations in NRAS (clone 3) or KRAS (clone 4). Clone 4 further evolves independently acquiring first PCSK6 c.463G>A (clone 5) and later IRF4 c.368A>G (clone 6) as additional hits. (c and d) Single-cell analysis of case 90 003. (c) Array heatmap obtained for this case. Layout, as in a. White arrows point out that cells with KRAS c.182A>G (clone 3) do not have KRAS c.34G>C (shown in clones 4–5). (d) Tumor phylogeny reveals a common ancestor (clone 2) carrying t(11;14) and other mutations (including IRF4 c.368A>G). Two independent clonal lineages arise from clone 2 and acquire different activating mutations in KRAS: clones 3–4. The latter clone further evolves mutating C7ORF23.

The etiological role of t(11;14)

Having shown the myeloma phylogeny, we analyzed the putative founder clone as defined by the initiating genetic lesion, the chromosomal translocation t(11:14). All patients in this study had a t(11;14), which hypothetically served as a marker of the initially immortalized clone (Table 1). To test this theory, we defined the translocation breakpoint using massively parallel sequencing combined with targeted pull down of the Ig regions23 and designed translocation-detection PCR-based assays applicable to single-cell assays for cases 90 003, 90 468 and 11/010 (Supplementary Table S2). For all three cases, we demonstrated that the t(11;14) was present in 91–100% of tumor cells (Supplementary Figure S2). Remarkably, we identified an ancestral myeloma clone showing only t(11;14) in samples 11/010 and 90 468 with a frequency of 27.5% and 7%, respectively (clone 1, Figures 4a and b, and Supplementary Figure S4). None of the remaining interrogated mutations was present in these subclones, suggesting these may be the initiating founders. However, we cannot dismiss the potential presence of other mutations/alterations that we could not detect due to technical issues (Figure 1f and Supplementary Table S2) or that were not included in the analysis. Indeed, case 11/010 had 4 mutations with frequencies ranging 80–95% that were not tested at the single-cell level and would likely subdivide the clone with only t(11;14) (Figures 1f and 4a). Conversely, the ancestral clone in case 90 003 carried the t(11;14) and 4 genetic mutations (clone 1, Figures 4c and d). The exact timeline of the events leading to this clone seems untraceable, as mutations may have followed initial immortalization by t(11;14), occurred simultaneously to translocation or, to some extent, preceded t(11;14).

Parallel evolution and the driver status of RAS mutations

Three cases had double hits in the RAS/MAPK pathway either in the same gene, KRAS (90 003 and 90 879), or in different genes such as KRAS and NRAS (11/010; Table 1). To determine whether these double hits occurred in the same or in independent subclones, we interrogated both mutations together with other altered genes at the single-cell level (Figures 1a, e and f).

The first case 90 879 had KRAS c.183A>C (p.Q61H) and c.199A>C (p.M67L) mutations at similar frequencies (92% and 83%, respectively; Supplementary Figure S2e), and these followed a linear sequence. KRAS p.Q61H occurred earlier than p.M67L, which was acquired soon after clonal expansion, as clone 3 carries both mutations and is present in 83.2% of tumor cells (Supplementary Figure S5). Although the presence of both mutations seemed to improve clonal fitness due to the predominance of clone 3, the extent to which p.M67L was responsible for this benefit is uncertain. p.M67L is not a known activating mutation and may simply be a passenger mutation in a proliferative clone defined by the activating mutation p.Q61H. Conversely, the double RAS hits in the remaining cases occurred in different clones related via a branching pattern.

The second case, 11/010, had 6 subclones (Figure 4a) and we confirmed the independent acquisition of 2 KRAS/NRAS mutations. The founding clone that only carried t(11;14) (clone 1, 27.5% frequency) acquired ABCA4 c.3294C>T, and it was this clone 2 (5.0%) that generated two divergent lineages: clone 3 (40.0%) carrying both FAT c.6080T>G and NRAS c.182A>G (p.Q61R) mutations, and clone 4 (5.0%) that had KRAS c.183A>C (p.Q61H). Remarkably, the latter gave rise to a lineage where PCSK6 c.463G>A was accumulated in clone 5 (7.5%), which subsequently originated clone 6 after the IRF4 c.368A>G mutation (15.0%; Figure 4b).

Lastly, in case 90 003, we described five clones and confirmed the co-occurrence of two KRAS mutations (Figure 4c). Clonal phylogeny identified the earliest clones 1 and 2 (4.0% and 34.0%, respectively) with t(11;14) and a range of mutations, including IRF4 c.368A>G and EGR1 c.1169A>G, the latter being specific of clone 2. Two divergent clonal lineages derived from clone 2, both acquiring KRAS-activating mutations. Clone 3 accumulated KRAS c.182A>G (p.Q61R) and accounted for 26.0% of tumor cells, whereas clones 4 and 5 had KRAS c.34G>C (p.G12R) representing 24.0% and 12.0%, respectively, of tumor cells (Figure 4d). These 2 cases (90 003 and 11/010) acquired the same convergent phenotype, that is, activation of the RAS/MAPK pathway, in 2 divergent clonal lineages derived from the same clonal ancestor, which subsequently evolved independently.

The existence of this pattern of parallel tumor evolution in two patients led us to study the presence of other common genomic features. It is noteworthy that both cases showed no chromosomal genomic aberrations other than t(11;14) (data not shown), but they shared IRF4 c.368A>G (p.K123R) as an early event in 90 003 (100% cells) and as a late hit in 11/010 (15%; Figure 4 and Supplementary Figures S2a and f). As IRF4 mutations have been formerly reported in small series of MM cases,11, 16 we screened for the IRF4 c.368A>G mutation in a larger series of presenting cases (n=452). We found IRF4 c.368A>G in nine patients (2%, Figure 5a). In a subset of cases with mutation information for KRAS/NRAS (243 patients), we observed a significant association for the co-occurrence of IRF4 and KRAS/NRAS mutations (P=0.000, Figures 5b and c). This finding suggests an epistatic effect and a collaborative relationship between these two lesions.

Figure 5
figure5

Myeloma patients with recurrent IRF4 mutations also have KRAS/NRAS mutations. (a) Screening of 452 myeloma patients identified 9 patients with IRF4 p.K123R mutations. Coding IRF4 protein mutations were recurrent in the DNA binding domain (DBD), both in myeloma and in chronic lymphocytic leukemia.45 IAD, IRF association domain. (b) Correlation analysis of 243 patients with information for mutations in IRF4, KRAS and NRAS. There was a highly significant association of co-occurrence for mutations in IRF4 and KRAS/NRAS. (c) Table of nine patients with IRF4 c.368A>G mutations and the information about KRAS or NRAS mutations at a clonal or subclonal levels.

Myeloma clones are the subject of Darwinian selection

Next, we postulated that each subclone was sustained by an MPC that had the capacity for self-renewal and proliferation, and was the progeny of the iMPC. These individual MPC would compete for access to the myeloma growth niche in the bone and underlie myeloma progression. Clones are expected to respond differently to treatment and potentially to display different propagating properties in xenograft transplant models. To test this hypothesis, we compared the tumor exome of a PCL patient, previously shown to have at least four clones at diagnosis,11 at both presentation and relapse time points. In parallel, 1 × 106 CD138+ cells from the same patient at diagnosis were injected into the tibia of NOD/SCIDyc(null) mice;28 two of the engrafted myelomas were analyzed by WES sequencing (Figure 6a). Our results demonstrated the existence of a complex phylogenetic history with fluctuations in the subclonal composition at relapse and in the two engrafted myelomas when compared with the presentation sample (Supplementary Figure S7a). Only mutations present in the two patient samples (presentation/relapse) were tested in the two engrafted samples, excluding mutations specific to the engrafted myelomas to prevent any false-positive results due to mouse DNA contamination. We identified 152 single nucleotide variants in all 4 samples (Supplementary Figure S7b and Supplementary Table S5). Briefly, mutations in 74 genes, including ATM or TP53, were shared across all samples. Although 26 mutations were characteristic of the presentation sample, 42 were specific to the relapse stage. It is noteworthy that three mutations were shared between the xenograft samples and the relapse sample. We postulate that these mutations were present at diagnosis, but at undetectable levels due to them being acquired in negligible subclones.

Figure 6
figure6

Changes in clonal architecture following patient treatment and in vivo NOD/SCID IL2Rγnull transplantation. (a) Patient history (Supplementary Methods) with xenograft experiments outlined. Isolated DNA from CD138+ cells purified from paired-patient samples at presentation11 and relapse are analyzed using exome sequencing. DNA from CD138+ cells purified from two independent myeloma xenograft samples is also studied. DTPACE, dexamethasone, thalidomide, cisplatin, adriamycin, cyclophosphamide and etoposide; ASCT, autologous stem cell transplantation; CVD, cyclophosphamide, bortezomib and dexamethasone. (b) Cluster of cancer cell fractions for all 152 single nucleotide variants (SNVs) identified in the four samples. Clones are defined on the left using genetic algorithms (Supplementary Methods). A selection of genes is shown with colored squares indicating the clone in which such genes were first mutated (right). (c) Phylogenetic natural history of this PCL patient. The seven clones detected by WES sequencing and genetic algorithms are depicted with the same colors as in b. The number of new mutations and non-synonymous (NS) SNVs together with a selection of key mutated genes are shown in the transition between clones. (d) The proportion of subclonal populations, shown as percentages, fluctuated in the four analyzed samples. There were clones described at presentation that were undetected and, therefore, potentially extinguished at relapse and at the engrafted myelomas (clones 2, 4 or 7). Conversely, new clones (3 and 6) emerged at the relapse stage as a result of a sequential accumulation of mutations in the earliest clone 1, which was at undetectable levels at presentation. See Supplementary Figure S7 for further information.

We used a genetic algorithm to group mutations present in these four samples into subclonal populations, and distinguished seven clones (Figure 6b). These were related by a complex branching pattern and all had different number of mutations (Figure 6c). The earliest ancestral clone carried six mutations, two out of which were non-synonymous single nucleotide variants in DNAH14 and FAM47C (clone 1). These mutations were present in all four samples at a 100% frequency, supporting clone 1 as the phylogenetic root (Figures 6b and c). There was a remarkable fluctuation in the clonal proportion in each sample (Figure 6d and Supplementary Figure S7c).

Four clones were detected at diagnosis with populations 5 and 7 being the predominant fractions (59% and 23%, respectively). Patient treatment caused a significant population bottleneck in which clones 2, 4 and 7 were extinguished and 41 new mutations were acquired. Consequently, clones 3 and 6 emerged at relapse (17 and 33%, respectively; Figures 6b–d). These clones originated from the earliest ancestral clone 1, which was undetected at diagnosis (Figures 6b–d). These results support the idea that earlier ancestral clones may lead to relapse.

The analysis of the engrafted myelomas showed the clonal extinction of clones 2, 4 and 7, and the re-emergence of clone 1, which was present at undetectable levels at diagnosis. Both xenograft samples had similar mutation frequencies, with slight variations between clones 1 and 5 (Figure 6d). Although the genetic architecture of these engrafted myelomas may be even more complex due to the accumulation of additional mutations not tested in this analysis, our results demonstrate that three clones were outcompeted by the remaining ones both during patient treatment and xenotransplant transitions. In addition, we show that earlier clones lead to relapse or engraftment (Figure 6d and Supplementary Figures 7c). Altogether, these findings emphasize the different survival properties of myeloma clones as a consequence of restrictive population bottlenecks such as patient treatment or xenotransplantation.

Discussion

Intratumor heterogeneity has been characterized using different approaches.4, 5, 6, 7, 8, 9, 10, 11, 12, 20 In this study, we successfully combine WES sequencing and single-cell genetic analysis to unravel the myeloma intraclonal phylogeny. Nonetheless, our results, although reporting the existence of genetically variegated subclones in MM, may be an underestimate of the true complexity for a number of reasons. First, we focused our single-cell analysis on a small list of non-synonymous single nucleotide variants, which were selected for their potential to deregulate putative driver genes decided based on prior literature and/or their CCFs. For instance, the putative ancestral clone 1 in case 11/010 (Figures 4a and b), which only carries the t(11;14) and accounts for 27.5% tumor bulk, may likely harbor additional mutations not tested in our analysis. Second, clones present at very low levels (<1%) would not have been detected due to the limited number of sorted cells. Having said this, we still demonstrate the main clonal phylogenetic relationships present in MM.

In this series of t(11;14) myelomas, even at first clinical presentation when clinical symptoms are present and treatment is required, we found evidence for the persistence of the earliest MPC clone. Two cases were characterized by the presence of a subclone carrying the t(11;14) as the sole abnormality (Figures 4a and b, and Supplementary Figure S4). This observation together with the detection of the translocation in most of the interrogated cells support the etiological significance of the t(11;14). This is also in agreement with the translocation being present in monoclonal gammopathy of undetermined significance, the earliest phase of myeloma30 associated with fewer mutations than myeloma.31 Nonetheless, once the founding clone is established, the acquisition of additional mutations inevitably leads to the genetic and phenotypic variegation within the progeny of the iMPC clone. These heterogeneous iMPC descendants are characterized by the ability to self-renew and proliferate, and can be considered as units of selection in terms of a Darwinian tumor evolution.13 Recently, we described intratumor heterogeneity in both monoclonal gammopathy of undetermined significance and smoldering myeloma,31 supporting the development of genetic complexity as an early event in the preclinical phases of the disease. Such early generation of intraclonal heterogeneity seems to be a hallmark of cancer in general, with similar findings described in the transition from premalignant to malignant stages of breast32, 33 and esophageal34 carcinomas. Our results suggest that the clonal remnants of the earliest stages of the disease may persist late in the myeloma natural history.

Clonal phylogenetic relationships have important implications for both chemotherapy and targeted treatment strategies. We describe a linear pattern of myeloma progression where mutations were sequentially accumulated (Figure 2c, Supplementary Figures S4b and S5c). However, our selection of putative driver mutations may have biased our findings in favor of the description of linear patterns, as driver mutations are suggested to impart higher cell fitness, predisposing for clonal sweeps and, thus, linear phylogenies. Nevertheless, it is worth considering the potential existence of additional clonal branches, which we could simply not describe. Moreover, as ancestor clones are still present in a tumor with a linear pattern, they may eventually accumulate new mutations, hence sprouting extra phylogenetic branches. Lastly, sampling bias may also confound a linear model, with different subclonal branches potentially present at other tumor sites.7, 9, 12 Taken altogether, a unifying interpretation of the data supports a branching evolutionary pattern as being the normal route for MM progression (Figures 3c, 4b and d, and 6c), where contemporaneous clones accumulate independent genetic hits, shaping their variegated mutational and phenotypic profiles.

Under the same environment and similar selective pressures, independent but not far-related clones may acquire similar mutations conferring important growth or survival advantages, a phenomenon known as parallel evolution. We demonstrate, for the first time, clear evidence of parallel evolution at the single-cell level in myeloma, as the same genetic pathway (RAS/MAPK) is altered more than once within the same tumor but in divergent clones, which further evolve independently (Figure 4). This finding is similar to data reported in primary tumor and metastatic renal cell carcinoma.9 The paradigm of parallel or convergent phenotypic evolution may also include common genomic aberrations (for example, Δchr13) occurring in independent clones.25 This mechanism represents an evolutionary adaptation of myeloma cells to their environment, is crucial for disease progression and is compatible with the description of concomitant RAS mutations in colon cancer35 and other hematological neoplasias.36, 37 RAS mutations thus represent true driver mutations in myeloma, as any subclone with activating KRAS/NRAS mutations is able to clonally expand and compete under the same selective pressure during myeloma progression. It may represent the Achilles heel in myeloma based on the use of targeted treatment, as recently seen for BRAF.38 This raises the therapeutic question of whether pruning aggressive myeloma clonal branches could positively select for the expansion and clonal predominance of more indolent ancestor clones, thus improving outcome.

We also show the accumulation of IRF4 c.368A>G (p.K123R) with activating RAS mutations, regardless of any chronological order. IRF4 is a transcription factor critical for the control of B-cell proliferation and differentiation,39 autophagy and plasma cell death.40 IRF4 upregulation is a hallmark in myeloma41 and IRF4 expression levels may have a therapeutic value related to lenalidomide resistance.42, 43, 44 We show for the first time that patients with IRF4 recurrent mutations significantly displayed KRAS/NRAS-activating mutations (Figure 5). This is confirmed in an independent data set,16 wherein 3/3 IRF4-mutated patients had RAS mutations, and indicates a potential epistatic effect between both pathways. Although the functional role of IRF4 p.K123R remains to be elucidated, mutations in the same protein domain are similarly frequent in chronic lymphocytic leukemia45 (Figure 5a). These observations support a role for IRF4 mutations as driver lesions in myeloma.

Remnant ancestor clones, although not necessarily the primordial founding clone, have been shown to lead to relapse.3, 5, 8, 18, 31 These ancestor clones are postulated to represent MPC units of selection, critical for progression through the clinical stages of myeloma, and for relapse. Here we demonstrate that not only early, and likely founding, MPC clones remain present at diagnosis but also that different clones exhibit distinct survival and propagating abilities following patient treatment and xenotransplantation (Figure 6). The natural history of a patient and the related xenografts demonstrated a complex phylogeny comprising extinction of clones and the arising of new subclones (Figure 6c and Supplementary Figure S7c). Remarkably, early clones present at low, or even undetectable, levels at diagnosis were able to survive treatment and lead patient relapse generating new predominant clones (green, orange and gray clones, Figures 6c and d). This finding was partially reproduced in xenograft samples (green clone, Figures 6c and d). It seems clear that subclonal populations are involved in selection and competition during myeloma progression and that their fluctuation is determined by their capacity to survive the selective pressure of treatment or to survive in a different micro-environment such as the present in the xenotransplantation.

On the basis of our results, we conclude that MM is a heterogeneous disease characterized by the accumulation of mutations and genomic aberrations at clonal and subclonal levels.5, 11, 16, 17, 18, 31 We successfully combine two state-of-the-art techniques such as WES sequencing and single-cell genetic analyses to better define the phylogenetic relationships between clonal populations in myeloma at clinical presentation and relapse. We conclude that the most plausible scenario for myeloma development is through a branching evolutionary pattern, and we also describe parallel-branching evolution where two divergent clones independently acquired the same convergent phenotype (RAS/MAPK pathway activation). Clonal diversity is the crucial component mediating Darwinian selection and underlies disease progression and the development of treatment resistance in MM. These findings have substantial implications for biopsy-based prognosis and targeted-therapy13, 15, 46 strategies.

References

  1. 1

    Hanahan D, Weinberg RA . Hallmarks of cancer: the next generation. Cell 2011; 144: 646–674.

  2. 2

    Nowell PC . The clonal evolution of tumor cell populations. Science 1976; 194: 23–28.

  3. 3

    Greaves M, Maley CC . Clonal evolution in cancer. Nature 2012; 481: 306–313.

  4. 4

    Anderson K, Lutz C, van Delft FW, Bateman CM, Guo Y, Colman SM et al. Genetic variegation of clonal architecture and propagating cells in leukaemia. Nature 2011; 469: 356–361.

  5. 5

    Keats JJ, Chesi M, Egan JB, Garbitt VM, Palmer SE, Braggio E et al. Clonal competition with alternating dominance in multiple myeloma. Blood 2012; 120: 1067–1076.

  6. 6

    Merlo LM, Pepper JW, Reid BJ, Maley CC . Cancer as an evolutionary and ecological process. Nat Rev Cancer 2006; 6: 924–935.

  7. 7

    Sottoriva A, Spiteri I, Piccirillo SG, Touloumis A, Collins VP, Marioni JC et al. Intratumor heterogeneity in human glioblastoma reflects cancer evolutionary dynamics. Proc Natl Acad Sci USA 2013; 110: 4009–4014.

  8. 8

    Ding L, Ley TJ, Larson DE, Miller CA, Koboldt DC, Welch JS et al. Clonal evolution in relapsed acute myeloid leukaemia revealed by whole-genome sequencing. Nature 2012; 481: 506–510.

  9. 9

    Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E et al. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med 2012; 366: 883–892.

  10. 10

    Nik-Zainal S, Van Loo P, Wedge DC, Alexandrov LB, Greenman CD, Lau KW et al. The life history of 21 breast cancers. Cell 2012; 149: 994–1007.

  11. 11

    Walker BA, Wardell CP, Melchor L, Hulkki S, Potter NE, Johnson DC et al. Intraclonal heterogeneity and distinct molecular mechanisms characterize the development of t(4;14) and t(11;14) myeloma. Blood 2012; 120: 1077–1086.

  12. 12

    Campbell PJ, Yachida S, Mudie LJ, Stephens PJ, Pleasance ED, Stebbings LA et al. The patterns and dynamics of genomic instability in metastatic pancreatic cancer. Nature 2010; 467: 1109–1113.

  13. 13

    Greaves M . Cancer stem cells as 'units of selection'. Evol Appl 2013; 6: 102–108.

  14. 14

    Kuehl WM, Bergsagel PL . Molecular pathogenesis of multiple myeloma and its premalignant precursor. J Clin Invest 2012; 122: 3456–3463.

  15. 15

    Morgan GJ, Walker BA, Davies FE . The genetic architecture of multiple myeloma. Nat Rev Cancer 2012; 12: 335–348.

  16. 16

    Chapman MA, Lawrence MS, Keats JJ, Cibulskis K, Sougnez C, Schinzel AC et al. Initial genome sequencing and analysis of multiple myeloma. Nature 2011; 471: 467–472.

  17. 17

    Egan JB, Shi CX, Tembe W, Christoforides A, Kurdoglu A, Sinari S et al. Whole-genome sequencing of multiple myeloma from diagnosis to plasma cell leukemia reveals genomic initiating events, evolution, and clonal tides. Blood 2012; 120: 1060–1066.

  18. 18

    Magrangeas F, Avet-Loiseau H, Gouraud W, Lode L, Decaux O, Godmer P et al. Minor clone provides a reservoir for relapse in multiple myeloma. Leukemia 2013; 27: 473–481.

  19. 19

    Durinck S, Ho C, Wang NJ, Liao W, Jakkula LR, Collisson EA et al. Temporal dissection of tumorigenesis in primary cancers. Cancer Discov 2011; 1: 137–143.

  20. 20

    Potter NE, Ermini L, Papaemmanuil E, Cazzaniga G, Vijayaraghavan G, Titley I et al. Single-cell mutational profiling and clonal phylogeny in cancer. Genome Res 2013; 23: 2115–2125.

  21. 21

    Boyd KD, Ross FM, Chiecchio L, Dagrada GP, Konn ZJ, Tapper WJ et al. A novel prognostic model in myeloma based on co-segregating adverse FISH lesions and the ISS: analysis of patients treated in the MRC Myeloma IX trial. Leukemia 2012; 26: 349–355.

  22. 22

    Chiecchio L, Protheroe RK, Ibrahim AH, Cheung KL, Rudduck C, Dagrada GP et al. Deletion of chromosome 13 detected by conventional cytogenetics is a critical prognostic factor in myeloma. Leukemia 2006; 20: 1610–1617.

  23. 23

    Walker BA, Wardell CP, Johnson DC, Kaiser MF, Begum DB, Dahir NB et al. Characterization of IGH locus breakpoints in multiple myeloma indicates a subset of translocations appear to occur in pregerminal center B cells. Blood 2013; 121: 3413–3419.

  24. 24

    Walker BA, Leone PE, Chiecchio L, Dickens NJ, Jenner MW, Boyd KD et al. A compendium of myeloma-associated chromosomal copy number abnormalities and their prognostic value. Blood 2010; 116: e56–e65.

  25. 25

    Walker BA, Leone PE, Jenner MW, Li C, Gonzalez D, Johnson DC et al. Integration of global SNP-based mapping and expression arrays reveals key regions, mechanisms, and genes important in the pathogenesis of multiple myeloma. Blood 2006; 108: 1733–1743.

  26. 26

    Weaver S, Dube S, Mir A, Qin J, Sun G, Ramakrishnan R et al. Taking qPCR to a higher level: analysis of CNV reveals the power of high throughput qPCR to enhance quantitative resolution. Methods 2010; 50: 271–276.

  27. 27

    Sims D, Bursteinas B, Gao Q, Jain E, MacKay A, Mitsopoulos C et al. ROCK: a breast cancer functional genomics resource. Breast Cancer Res Treat 2010; 124: 567–572.

  28. 28

    Fryer RA, Graham TJ, Smith EM, Walker-Samuel S, Morgan GJ, Robinson SP et al. Characterization of a novel mouse model of multiple myeloma and its use in preclinical therapeutic assessment. PLoS One 2013; 8: e57641.

  29. 29

    Scrucca L . GA: A package for genetic algorithms in R. J Stat Softw 2013; 53: 1–37.

  30. 30

    Fonseca R, Bailey RJ, Ahmann GJ, Rajkumar SV, Hoyer JD, Lust JA et al. Genomic abnormalities in monoclonal gammopathy of undetermined significance. Blood 2002; 100: 1417–1424.

  31. 31

    Walker BA, Wardell CP, Melchor L, Brioli A, Johnson DC, Kaiser MF et al. Intraclonal heterogeneity is a critical early event in the development of myeloma and precedes the development of clinical symptoms. Leukemia 2013; e-pub ahead of print 2 July 2013 doi:10.1038/leu.2013.199.

  32. 32

    Park SY, Gonen M, Kim HJ, Michor F, Polyak K . Cellular and genetic diversity in the progression of in situ human breast carcinomas to an invasive phenotype. J Clin Invest 2010; 120: 636–644.

  33. 33

    Newburger DE, Kashef-Haghighi D, Weng Z, Salari R, Sweeney RT, Brunner AL et al. Genome evolution during progression to breast cancer. Genome Res 2013; 23: 1097–1108.

  34. 34

    Merlo LM, Shah NA, Li X, Blount PL, Vaughan TL, Reid BJ et al. A comprehensive survey of clonal diversity measures in Barrett's esophagus as biomarkers of progression to esophageal adenocarcinoma. Cancer Prev Res (Phila) 2010; 3: 1388–1397.

  35. 35

    Forrester K, Almoguera C, Han K, Grizzle WE, Perucho M . Detection of high incidence of K-ras oncogenes during human colon tumorigenesis. Nature 1987; 327: 298–303.

  36. 36

    Farr CJ, Saiki RK, Erlich HA, McCormick F, Marshall CJ . Analysis of RAS gene mutations in acute myeloid leukemia by polymerase chain reaction and oligonucleotide probes. Proc Natl Acad Sci USA 1988; 85: 1629–1633.

  37. 37

    Itzykson R, Kosmider O, Renneville A, Morabito M, Preudhomme C, Berthon C et al. Clonal architecture of chronic myelomonocytic leukemias. Blood 2013; 121: 2186–2198.

  38. 38

    Andrulis M, Lehners N, Capper D, Penzel R, Heining C, Huellein J et al. Targeting the BRAF V600E mutation in multiple myeloma. Cancer Discov 2013; 3: 862–869.

  39. 39

    Lu R . Interferon regulatory factor 4 and 8 in B-cell development. Trends Immunol 2008; 29: 487–492.

  40. 40

    Lamy L, Ngo VN, Emre NC, Shaffer AL 3rd, Yang Y, Tian E et al. Control of autophagic cell death by caspase-10 in multiple myeloma. Cancer Cell 2013; 23: 435–449.

  41. 41

    Shaffer AL, Emre NC, Lamy L, Ngo VN, Wright G, Xiao W et al. IRF4 addiction in multiple myeloma. Nature 2008; 454: 226–231.

  42. 42

    Lopez-Girona A, Mendy D, Ito T, Miller K, Gandhi AK, Kang J et al. Cereblon is a direct protein target for immunomodulatory and antiproliferative activities of lenalidomide and pomalidomide. Leukemia 2012; 26: 2326–2335.

  43. 43

    Zhang LH, Kosek J, Wang M, Heise C, Schafer PH, Chopra R . Lenalidomide efficacy in activated B-cell-like subtype diffuse large B-cell lymphoma is dependent upon IRF4 and cereblon expression. Br J Haematol 2013; 160: 487–502.

  44. 44

    Zhu YX, Braggio E, Shi CX, Bruins LA, Schmidt JE, Van Wier S et al. Cereblon expression is required for the antimyeloma activity of lenalidomide and pomalidomide. Blood 2011; 118: 4771–4779.

  45. 45

    Havelange V, Pekarsky Y, Nakamura T, Palamarchuk A, Alder H, Rassenti L et al. IRF4 mutations in chronic lymphocytic leukemia. Blood 2011; 118: 2827–2829.

  46. 46

    Swanton C . Intratumor heterogeneity: evolution through space and time. Cancer Res 2012; 72: 4875–4882.

Download references

Acknowledgements

We acknowledge Professor Mel Greaves for his scientific input and the Tumour Profiling Unit at the Institute of Cancer Research for their support and technical expertise in this study. This study was supported by a Cancer Research UK sample collection grant (CTAAC), grants from the National Institutes of Health Biomedical Research Centre at the Royal Marsden Hospital and a Myeloma UK programme grant (LM, CPW, RAF, DCJ and BAW). MFK is supported by a Research Fellowship by Deutsche Forschungsgemeinschaft (KA 3338/1-1). FED is a Cancer Research UK Senior Clinical Fellow.

Author information

Correspondence to G J Morgan.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies this paper on the Leukemia website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Melchor, L., Brioli, A., Wardell, C. et al. Single-cell genetic analysis reveals the composition of initiating clones and phylogenetic patterns of branching and parallel evolution in myeloma. Leukemia 28, 1705–1715 (2014) doi:10.1038/leu.2014.13

Download citation

Keywords

  • intraclonal heterogeneity
  • intratumor heterogeneity
  • multiple myeloma
  • exome sequencing
  • single-cell analysis
  • parallel evolution

Further reading