Abstract
The leukemia stem cell (LSC) compartment is a complex reservoir fueling disease progression in acute myeloid leukemia (AML). The existence of heterogeneity within this compartment is well documented but prior studies have focused on genetic heterogeneity without being able to address functional heterogeneity. Understanding this heterogeneity is critical for the informed design of therapies targeting LSC, but has been hampered by LSC scarcity and the lack of reliable cell surface markers for viable LSC isolation. To overcome these challenges, we turned to the patient-derived OCI-AML22 cell model. This model includes functionally, transcriptionally and epigenetically characterized LSC broadly representative of LSC found in primary AML samples. Focusing on the pool of LSC, we used an integrated approach combining xenograft assays with single-cell analysis to identify two LSC subtypes with distinct transcriptional, epigenetic and functional properties. These LSC subtypes differed in depth of quiescence, differentiation potential, repopulation capacity, sensitivity to chemotherapy and could be isolated based on CD112 expression. A majority of AML patient samples had transcriptional signatures reflective of either LSC subtype, and some even showed coexistence within an individual sample. This work provides a framework for investigating the LSC compartment and designing combinatorial therapeutic strategies in AML.
![](http://media.springernature.com/lw685/springer-static/image/art%3A10.1038%2Fs41375-024-02358-9/MediaObjects/41375_2024_2358_Figa_HTML.png)
Similar content being viewed by others
Introduction
LSC have been tightly linked to disease progression and poor prognosis in AML, and the development of more effective therapies to eradicate LSC is an important unmet need [1,2,3,4]. However, progress has been hampered by inter- and intra-patient heterogeneity within the LSC compartment together with a limited ability to resolve intricacies within the pool of LSC [5]. Studies of the heterogeneity within the normal hematopoietic stem cell (HSC) compartment in mice and humans [6,7,8,9] provide a framework for investigating heterogeneity within the LSC pool. One of the key tools that facilitated a deeper understanding of the HSC compartment at the functional and molecular level was the identification of cell surface markers enabling purification of human HSC at nearly single cell resolution [10]. For example, immunophenotypic mapping of the HSC compartment coupled with single cell functional repopulation assays resulted in the identification of three HSC compartments: long-term (LT-), intermediate and short-term HSC with the LT-HSC subset being the most primitive [7,8,9]. Recently, we showed that LT-HSC can be further fractionated based on expression of the cell surface marker CD112 into subsets that are molecularly, transcriptionally and functionally distinct [11]. In xenograft repopulation assays, CD112-Low HSC exhibit latency, with slower engraftment kinetics compared to CD112-High HSC [12]. However, CD112-Low HSC outcompetes CD112-High HSC in long-term upon serial transplantation [11].
In contrast to the normal HSC compartment, there are few tools currently available to study the LSC compartment. The existence of heterogeneity within this compartment is well documented in leukemia [13,14,15,16,17,18,19], but prior studies have focused on genetic mutations without being able to address functional heterogeneity. More than two decades ago, we used clonal tracking approaches to show that the LSC pool comprises multiple LSC subtypes with distinct functional capacities, with some LSC more latent and slowly repopulating in xenograft assays and others exhibiting faster repopulation kinetics [20]. Given that the existence of functional heterogeneity among LSCs is likely associated with differences in therapeutic vulnerabilities, a better understanding of the basis for the observed heterogeneity is crucial for the rational design of effective therapies aimed at eradicating the entire pool of LSC [21]. However, progress has been hampered by the lack of tools for prospective isolation of distinct LSC subtypes. Although several LSC surface markers have been described [22,23,24,25], none reliably capture functional LSC across a broad spectrum of AML patients.
Despite this heterogeneity, transcriptomic and epigenetic signatures do provide evidence that there are core properties of LSC that are broadly applicable across the full spectrum of AML [1, 2, 26,27,28]. For example, the LSC17 stemness score is highly predictive of outcome across thousands of pediatric and adult AML samples, irrespective of molecular and cytogenetic subtype [2, 29,30,31]. In addition, single cell RNA-seq analysis uncovered signatures of stem, progenitor and mature AML cells that enabled the development of a hierarchy classification system that predicts drug sensitivity in AML patients [32, 33]. These findings suggest that core stemness properties of LSC that drive clinical outcomes and treatment sensitivity are shared across AML patients, and support studies to understand whether the basis for the observed functional heterogeneity within the LSC compartment may also be shared. The gold standard for studies of functional properties of LSC has been AML patient-derived xenograft (PDX) assays, as most AML cell lines do not exhibit a LSC-driven hierarchy that captures patient biology [34]. However, PDX assays are laborious, difficult to standardize and cannot be tailored for high-throughput studies. Recently, we developed a patient-derived AML model (OCI-AML22) that retains biological properties with broad clinical relevance [34]. The OCI-AML22 LSC population exhibits hallmark stemness properties that are highly prognostic across diverse independent cohorts of AML patient samples [1, 2, 29,30,31,32, 35,36,37,38,39,40]. This clinically-relevant model provides a powerful tool for mechanistic and functional studies to investigate LSC biology in the correct cellular context.
Here, we used an integrated approach combining functional assays with single-cell multiome analysis in the OCI-AML22 model to identify two distinct LSC subtypes that could be prospectively isolated. These LSC subtypes possess distinct properties such as engraftment kinetics and depth of quiescence, that are retained through serial transplantation. In addition, these subsets are differentially responding to Cytarabine, one of the main drugs used as first line therapy for AML. The majority of AML patient samples possessed either of these LSC subtypes, albeit in varying proportions attesting to the broad biological relevance of our findings and one of the subtypes was associated with poorer prognosis showing the clinical relevance of these new LSC subsets. Altogether, the ability to prospectively isolate two distinct LSC subsets provides a framework to uncover novel therapeutic angles.
Results
Single cell multiome analysis captures heterogeneity within the OCI-AML22 CD34 + CD38- cell fraction
To characterize in depth the pool of LSC in OCI-AML22, we performed single cell (sc) multiome analysis (scATAC-Seq/scRNA-Seq) on 7160 cells isolated from the LSC-enriched CD34 + CD38- fraction (in vivo LSC frequency is 1 in 200) [34] (Fig. 1A). The majority of these cells were enriched for stem cell programs when mapped onto a single cell transcriptomic map of stem and progenitor populations from healthy human bone marrow (Supplementary Fig. S1A) [41]. To characterize these cells, we generated a Uniform Manifold Approximation and Projection (UMAP) based on both RNA-Seq and ATAC-Seq data and labeled cells based on their colocalization with normal hematopoietic populations when mapped onto the bone marrow reference map [41] (Supplementary Fig. S2) Interestingly, cells labeled as HSC-MPP (multipotent progenitor) - the population most enriched for self-renewing cells in the normal hematopoietic hierarchy - mapped to two distinct regions of the UMAP representation (Fig. 1B). Other immature populations similarly mapped to these 2 regions (Supplementary Fig. S1–S4). Cells labeled as more committed progenitor-like, such as cycling progenitors or early GMP, mapped in between the regions occupied by HSC/MPP-like cells (Supplementary Fig. S4). We also evaluated the epigenetic states of individual cells within this LSC pool using chromatin accessibility signatures from highly purified cord blood populations as a reference [42]. Using chromVar to calculate the per-cell enrichment for each signature, we identified cells highly enriched for two stem cell epigenetic signatures previously identified [42] (LT/HSPC and ACT/HSPC signatures; Fig. 1C, D), as well as cells enriched for signatures of more committed populations such as monocytes or granulocytes [42] (Supplementary Fig. S5). This analysis revealed the existence of two epigenetically distinct stem-like cell clusters localizing to the two regions occupied by transcriptionally characterized stem-like cells (Fig. 1C, D).
A Experimental design. B–D UMAP representation based on both RNA-Seq and ATAC-Seq for each of the CD34 + CD38- OCI-AML22 cells that passed QC. Cells that mapped into the HSC-MPP group from Supplementary Fig. S1 are colored in red (B) Cells are colored based on the z-score value for the LT/HSPC epigenetic signature [42] (C) or based on the z-score value for the ACT/HSPC epigenetic signature from [42] (D). E TooManyCells representation [43] using RNA-Seq data from CD34 + CD38- OCI-AML22 single cells. F TooManyPeak representation [44] using ATAC-Seq data from CD34 + CD38- OCI-AML22 single cells. G Enrichment scores for the LT/HSPC signature [42] are plotted for OCI-AML22 CD34 + CD38- cells across the TooManyCells branches as indicated. Individual cells are colored based on their branch identity in the TooManyCells Tree (see E). H Enrichment scores for the LT/HSPC signature [42] are plotted for OCI-AML22 CD34 + CD38- cells across the TooManyPeaks branches as indicated. Cells are colored based on their branch identity in the ToomanyPeaks Tree (see F). UMAP representation based on both RNA-Seq and ATAC-Seq for each of the CD34 + CD38- OCI-AML22 cells that passed QC. Cells are colored based on their branch identity in the TooManycells representation (I) or based on their branch identity in the ToomanyPeaks representation (J). K Enrichment scores for the LT/HSPC signature [42] are plotted for OCI-AML22 CD34 + CD38- cells across the TooManyCells branches as indicated. Cells are colored based on their Branch identity in the TooManycells representation shown (F). L Enrichment scores for the LT/HSPC signature [42] are plotted for OCI-AML22 CD34 + CD38- cells across the TooManyPeaks branches as indicated. Cells are colored based on the branch identity in the TooManyPeaks representation.
To understand the stem cell heterogeneity within the OCI-AML22 LSC fraction, we applied the TooManyCells [43] and TooManyPeaks [44] algorithms, which build branching trees comprising cells sharing transcriptomic or epigenetic features, respectively. Each algorithm produced a diagram with 4 trunks (Fig. 1E, F), independently suggesting the existence of 4 distinct cell populations. TrunkI on both diagrams contained cells most enriched for the LT/HSPC (Fig. 1G, H) or the ACT/HSPC epigenetic signatures (Supplementary Fig. S6A and Supplementary Fig. S7A) and most depleted for epigenetic signatures of committed populations such as monocytes or granulocytes (Supplementary Fig. S6A and Supplementary Fig. S7A). Interestingly, this stem-like TrunkI could be further split into multiple branches, using TooManyCells (Fig. 1E) or TooManyPeaks (Fig. 1F), suggesting the existence of epigenetically and transcriptionally distinct LSC subgroups making up TrunkI. Labeling each cell in the UMAP representation from Fig. 1B–D based on their designated branch showed a clear allocation of individual branches associated with stemness (Branch1-3) to two distinct regions regardless of the algorithm (TooManyCells or TooManyPeaks) used (Fig. 1I, J). Cells represented in Branch 1 and Branch 2 from both trees (hereafter termed B1/2-LSC) mapped to one region, while cells belonging to Branch 3 on both trees (hereafter termed B3-LSC) mapped to another distinct region (Fig. 1I, J). Both B1/2 LSC and B3 LSC branches were more enriched for the LT/HSPC epigenetic signature compared to cells that are part of the other trunks (Trunk II-IV) as identified by either TooManyCells or TooManyPeaks (Fig. 1K, L), establishing their stem cell identity. Overall, these results demonstrate that two transcriptionally and epigenetically distinct LSC populations (B1/2-LSC and B3-LSC) coexist within the OCI-AML22 CD34 + CD38- fraction.
B1/2 and B3 LSC signatures are enriched in LSC+ fractions from AML patient samples and co-exist within individual patients
To determine the clinical relevance of LSC identified in OCI-AML22, we used a cohort of 73 AML patient samples that have been sorted based on CD34 and CD38 expression, where each fraction was functionally assessed for their ability to engraft NSG mice [2] (Fig. 2A). GSVA using the list of genes positively upregulated in TrunkI compared to the other Trunks (II-IV) showed that they were enriched in functionally-validated LSC-containing fractions (LSC + , 138 fractions) compared to LSC-depleted fractions (LSC-, 84 fractions) [2] (Supplementary Fig. S8A). In addition, GSVA enrichment scores correlated with LSC frequency (Supplementary Fig. S8B). We then investigated whether B1/2-LSC and B3-LSC could be detected within AML patient samples. To this end, we generated the B1/2 and B3 signatures where B1/2 signature is the lists of genes significantly upregulated in cells belonging to B1/2 branches versus cells belonging to the B3 branch and vice-versa. GSVA scores for B1/2 and B3 signatures calculated across the 138 LSC-containing (LSC + ) fractions [2], enabled to cluster these fractions into 4 clusters (Fig. 2B). One cluster was not enriched for either B1/2 or B3 signatures (Non-enriched LSC+ cluster); two clusters were enriched exclusively for either the B1/2 signature or the B3 signature (single clusters: B1/2 LSC+ cluster or B3 LSC+ cluster), and another cluster was enriched for both (Mixed LSC+ cluster) (Fig. 2B). Most patient samples contained at least one LSC+ fraction enriched for B1/2 or B3 signatures (n = 60/73; 81%), underscoring the broad clinical relevance of our newly developed LSC subtype signatures. To determine whether fractions were clustered based on intra-patient differences or if LSC+ fractions belonging to the same patient could be found across different clusters, and thus uncover heterogeneity within the LSC compartment of individual patients, we focused on patient samples with more than one LSC+ fraction (n = 42/73 patients). We were able to uncover 3 groups of patient samples, based on the combinations of LSC+ fractions in each individual sample. A minority of patient samples presented only LSC+ fractions from the non-enriched cluster (n = 4/42; 9.5% - Supplementary Fig. S9A), while the majority of patient samples presented LSC+ fractions belonging to one of the single clusters (either B1/2 or B3 clusters) (n = 29/42; 69%- Fig. 2C, D). Interestingly, a third group of patient samples (n = 9/42; 21%) exhibited a more complex composition of LSC + , with a combination of B1/2 and B3 signatures either present in separate fractions (Fig. 2E) or in the same fraction (Fig. 2F–I). Taken together, these data demonstrate that the two LSC subtypes (B1/2 and B3) we identified in OCI-AML22 are found in the majority of primary AML samples and can co-exist, in varying combinations, within individual patients.
A Experimental and computational scheme. B GSVA was performed for B1/2 or B3 signatures across LSC-positive fractions functionally assessed via xenograft assays from primary AML samples [2]. Scores for each signature are plotted and were used to perform hierarchical clustering using Complexe heatmap. C–I Functionally assessed LSC positive fractions for each indicated patient sample are represented in boxes colored based on the cluster each fraction was allocated in (A). LSC negative fractions are indicated in white. Not assessed fractions are not represented. Patients with fractions belonging to the B1/2 cluster from Fig. 2A (C), Patients with fractions belonging to the B3 cluster from Fig. 2B (D), Patients with fractions belonging to cluster B1/2 or cluster B3 from Fig. 2B (E), patients with fractions belonging to the Mixed cluster (F), Patients with fractions belonging to the Mixed cluster and the B1/2 cluster (G), patients with fractions belonging to the Mixed cluster and the B3 cluster (H), patient with fractions belonging to the 3 clusters enriched with either B1/2 or B3 signatures (I).
Single cell assays capture distinct colony forming and differentiation potentials within the CD34+CD38- OCI-AML22 fraction
Having identified two groups of LSC based on transcriptional and epigenetic features, we next investigated whether there was heterogeneity in the LSC compartment at the functional level. To this end, we first developed an in vitro assay to analyze the progeny produced by single OCI-AML22 CD34 + CD38- cells deposited into 96 well plates pre-seeded with MS5 stromal cells (Fig. 3A) [45]. Approximately half of the seeded CD34 + CD38- cells were able to generate a colony (Fig. 3B). We distinguished two types of clonogenic outputs (Fig. 3C–H). The first type was characterized by smaller colonies retaining CD34+ expression on >80% of cells. In contrast, the second type was characterized by larger colonies with <80% CD34+ cells. Both colony types exhibited similar expansion of the CD34 + CD38- population from the original single seeded cell despite their differential capacity to generate large numbers of downstream progenitors (Fig. 3E–H). Collectively, these data point to the existence of two clonogenic cell types within the OCI-AML22 CD34 + CD38- fraction, each able to expand this fraction but differing in their differentiation potential.
A Experimental scheme. B Pie chart indicating the percentage of CD34 + CD38- OCIAML22 cells that do not generate colonies (depicted in white) or generate a colony (depicted in yellow or red depending on the type of colonies). C For each colony type depicted in yellow or red in Fig. 3B, the absolute number of total cells is plotted in the matching color (yellow: left; red: right). D Extreme immunophenotypic profiles of the 2 types of colonies identified from single cell in vitro assay and depicted in yellow or red in (B). For each colony type depicted in yellow or red in (B), the absolute number of CD34 + CD38- cells (E), CD34 + CD38+ cells (F), CD34-CD38+ cells (G) or CD34-CD38- cells (H) generated from each single CD34 + CD38- OCI-AML22 cells co-cultured with MS5 is depicted in the matching colors (yellow: left; red: right).
Xenotransplantation assays reveal the existence of LSC with distinct repopulation kinetics and differentiation potentials
To determine if the functional heterogeneity we found in vitro could also be identified and functionally validated in vivo, we xenotransplanted OCI-AML22 CD34 + CD38- cells at limiting dilution and then complemented this with a temporal assessment of repopulation potential of individual LSCs after 8 and 12 weeks (Fig. 4A). Interestingly, more grafts were detectable both at the limiting dose (469 cells per mouse) and at the intermediate dose (1875 cells per mouse) after 12 weeks compared to 8 weeks (Fig. 4B). Given that stem cell detection requires generation of a large enough graft to be measurable, our results suggest variability in the engraftment kinetic of individual LSCs, with one LSC subtype repopulating mice faster to become detectable at 8 weeks, compared to the other. To gain a deeper understanding of the functional heterogeneity observed in vivo within the pool of OCI-AML22 LSC, we performed dimensionality reduction using UMAP of experimental parameters including the number of cells injected, engraftment level as a surrogate for repopulation capacity, and the proportion of cells expressing CD34 and CD38 as a surrogate for differentiation capacity. The UMAP analysis revealed the existence of 2 clusters (Fig. 4C). Interestingly, despite the fact that the number of injected cells per mouse was similar (Fig. 4D), the 2 clusters differed in engraftment level (Fig. 4E) and the CD34+ expression profile of the resulting graft (Fig. 4F, G). Collectively, these data suggest the existence of 2 subtypes of functional LSC in OCI-AML22: one subtype that repopulates slowly to generate grafts that are only detectable after 12 weeks, while a second subtype repopulates mice more rapidly and thereby, produces sufficient progeny for the graft to be detected as early as 8 weeks. These distinct differentiation and expansion capacities are reminiscent of the functional differences we observed in vitro.
A Schematic representation of the experiment. NSG-SGM3 mice were injected intrafemorally at limiting dilution with multiple cell doses of sorted CD34 + CD38- OCI-AML22 cells. Engraftment was assessed 8 and 12 weeks after injection. B The percentage of mice for which engraftment could be detected at 8 weeks or 12 weeks, after injection of the indicated OCI-AML22 CD34 + CD38- cell dose sorted from a bulk culture expanded for 3-4 months in vitro is plotted. C UMAP-based clustering of engraftment parameters derived from non-injected bone marrow (BM) and injected femur (RF) (parameters: total human cell counts, %CD34 expression, %engraftment level, number of injected cells per mice) of NSG-M3S mice 12 weeks after injection CD34 + CD38- OCIAML22 cells per mouse using doses from Fig. 4B. D Statistical analysis for each cluster from (C), showing the number of injected cells (D), the engraftment level (E), the percentage of CD34+ cells in the generated grafts (F). G Representative immunophenotypic profiles of grafts based on CD34 and CD38 cell surface expression (hCD45+ subgated cells are shown) for the clusters identified in (C).
CD112 enables prospective isolation of distinct LSC subsets linked to different cell cycle states
We previously characterized two populations of normal LT-HSC that can be segregated based on expression of CD112 (CD112High and CD112Low) where each population exhibits distinct repopulation kinetics linked to differences in cell cycle state [11]. In our prior study, we generated a chromatin accessibility signature consisting of peaks more accessible in the CD112High population compared to the CD112Low population (CD112High signature). To assess whether B1/2 and B3 LSC from OCI-AML22 could be distinguished by CD112 expression, we performed ChromVAR enrichment for the CD112High signature and found that B3-LSC were more enriched for this signature compared to B1/2-LSC (Fig. 5A). To examine whether the repopulation kinetics of the two LSC subtypes we identified could be explained by distinct cycling capacities, we calculated cell cycle phase scores using canonical markers of G2M or S phases [46]. Cells that are part of TrunkI from TooManyCells (Fig. 1) were the most enriched in G0/G1, suggesting they are less actively cycling, compared to cells that were found in the other Trunks (Fig. 5B). This is consistent with the fact that LSC are less proliferative than their downstream progeny [47]. To further discriminate cycling abilities within the less proliferative TrunkI cell population, we performed GSEA using 3 independent lists of genes known to be more expressed in cycling cells (G2M) compared to quiescent cells (G0/G1): LSPC-Cycle top 250 genes [33], Tirosh Cycle [46] and the Rehman Colon Diapause Down [48] signatures. B3-CD112High LSC were more enriched for these cycling signatures compared to B1/2-CD112Low LSC (Fig. 5C), suggesting that the former corresponds to the more proliferative, rapidly repopulating cells identified in our in vitro and in vivo functional assays. To determine whether B1/2 and B3 LSCs can be separated based on CD112 cell surface expression, CD34 + CD38- OCI-AML22 cells were sorted based on CD112 expression (CD112High, top 20%; CD112Low, bottom 20%; Fig. 5D). Cell cycle was assessed using classic markers (Ki67 and Hoechst) known to distinguish between G0 (Hoechstlow Ki67-) and G1 cells (HoechstlowKi67 + ). Additionally we also measured CDK6 expression, a marker of G0 exit that can further stratify G0 cells into deeply quiescent cells (CDK6-Ki67-) versus cells primed for more rapid G0 exit (CDK6 + KI67-) [7, 11]. The CD112Low fraction contained a higher proportion of Hoechstlow Ki67- and Ki67-CDK6- cells compared to the CD112High fraction (Fig. 5E, F) and generated smaller cellular output compared to the CD112 High fraction (Supplementary Fig. S10). To extend our findings to primary samples, we examined functionally defined LSC+ fractions from two AML patient samples. The CD112Low population in both cases contained a higher proportion of cells in G0 (Hoechstlow Ki67-) and a greater proportion of CDK6-Ki67- deeply quiescent cells (Fig. 5G, H) compared to CD112High cells. Together, these data demonstrate that CD112 expression provides a tool for prospective enrichment of two LSC subtypes on the basis of their distinct cell cycle states.
A ChromVAR enrichment score for the CD112-High epigenetic signature [11] in the indicated populations from TooManyCells. B UMAP representation based on both RNA-Seq and ATAC-Seq for each of the CD34 + CD38- OCI-AML22 cells that passed QC. Cells are colored based on the cell cycle phase they were allocated to using Seurat cell cycle signature [47]. C GSEA across B3-CD112High and B1/2-CD112Low for LSPC-Cycle Top 250 signature [33], Tirosh cell cycle signature [47] and the Diapause Down in colon cancer signature [49]. D Experimental scheme. OCI-AML22 CD34 + CD38- fraction was sorted based on CD112 expression level then stained for CDK6, Ki-67 and Hoechst to determine the percentage of cells in G0 (HoechstLowKi67-) (E) or the deepest quiescent cells Ki67-CDK6- cells (F) (n = 5 individual cultures, 3 different times points). Representative FACS plots are disclosed for each condition on the right (E, F). The LSC+fraction from functionally assessed primary AML samples was sorted based on CD112 expression level (top 20% :CD112-High, bottom 20% :CD112-Low) then stained for CDK6, Ki67 and Hoechst to determine the percentage of cells in G0 (HoechstLowKi67-)(G) or the deepest quiescent cells in G0 (Ki67-CDK6-) cells (H). FACS plots are disclosed for each condition for each patient sample (G, H).
Prospectively isolated LSC subtypes preserve their distinct differentiation potentials through serial repopulation and are associated with differential sensitivity to chemotherapy
To unequivocally demonstrate the link between CD112 expression, cell cycle state and repopulation ability of the two identified LSC subtypes, we injected CD112 High and CD112 Low fractions sorted from OCI-AML22 cells into NSG mice (Fig. 6A). Although both fractions were able to engraft mice, CD112Low cells gave rise to smaller grafts compared to CD112 High cells (Fig. 6B) and presented a phenotype reminiscent of the two distinct LSC outputs we found from our in vitro and in vivo studies (Fig. 6C), although only 1 mouse transplanted with CD112Low cells showed engraftment over 20%. To assess long-term repopulation and differentiation capacities of the two prospectively isolated LSC populations, we first isolated CD34 + CD38- cells from each of the primary grafts that had been generated from either CD112Low or CD112High cells and then injected them into secondary NSG-SGM3 mice (Fig. 6D). Secondary grafts generated by cells harvested from CD112Low primary grafts contained a higher proportion of CD34+ cells (Fig. 6E and Supplementary Fig. S11) compared to the secondary grafts generated by cells harvested from CD112 High primary grafts. In order to determine clinical implications linked the existence of these two distinct LSC types, we harnessed transcriptional signatures of sensitivity to chemotherapy: Cytotoxicity to Cytidine analogs [49] and Cytidine analogs transport and metabolism [49] or a signature of cancer stem cell drug sensitivity [48]. Focusing on the LSC+ functionally assessed fractions we identified as being part of LSC-B1/2 or LSC-B3 clusters (in Fig. 2A, B), we calculated GSVA score for each of the cytarabine sensitivity signatures [49] or cancer stem cell sensitivity signature [48]. The experimental and computational procedure, described in a schematic (Supplementary Fig. S12), revealed an enrichment of each of the chemotherapy sensitivity signatures in the B3-only LSC+ fractions as compared to the B1/2 only LSC+ fractions (Fig. 6F–H). This suggested that each of the LSC subtypes we identified are linked to distinct levels of resistance to chemotherapy. Cytarabine treatment using the sorted CD112high or CD112Low CD34 + CD38- population demonstrated that the CD112 Low population was more resistant to cytarabine compared to the CD112 High population (Fig. 6I, J). In line with this result, survival analysis across 2 independent cohorts of AML patients (n = 668 patients, TCGA and GSE6891) revealed a significantly worse survival for patients exclusively enriched for the B1/2 signature but not the B3 signature as compared to patients exclusively enriched for the B3 signature but not the B1/2 signature (Fig. 6K, L). Collectively, these findings establish the existence of at least two LSC subtypes that exhibit different levels of resistance to cytarabine and can be prospectively isolated based on CD112 expression, where their distinct repopulation kinetics and differentiation potentials are retained through serial transplantation (Fig. 6M).
A Experimental in vivo scheme. B Engraftment level 8 weeks after injection of 10,000 OCI-AML22 cells in NSG mice sorted as indicated on (A). C Experimental in vivo strategy for secondary experiment and representative FACS plots for each condition based on CD34 and CD38 among the CD45 + 7AAD- engrafted population. D Experimental scheme for the secondary in vivo assay. E Percentage of CD34/CD38 cells of secondary grafts generated after injection of 10,000 CD34 + CD38- cells CD112 High or CD112 Low, then sorted for CD34 + CD38- without additional selection for CD112, as indicated in (D). Mann–Whitney test, n = 9 mice. GSVA score for signatures of cytarabine sensitivity (F, G) or colon cancer stem cell sensitivity (H) was calculated on LSC+ fractions identified as being part of B1/2-LSC+ cluster or on LSC+ fractions identified as being part of B3-LSC+ cluster from Fig. 2A, B. I, J CD112 High and CD112 Low OCI-AML22 CD34 + CD38- were sorted and treated with cytarabine for 24 h at the indicated dose. Representative curve is represented (n = 1) (I) and significance is calculated at the 5 μM dose (n = 5, paired Mann–Whitney test) (J). GSVA for B1/2 or B3 signatures was calculated across TCGA (K) or GSE6891 cohorts (L). survival curves were plotted for patients enriched for B1/2 signature only (yellow curve) or enriched for B3 signature only (red curve). M Summary scheme.
Discussion
Integrating a panel of functional in vitro and xenograft assays, with single cell multiomics, we report the existence of two LSC subtypes with distinct epigenetic and transcriptional signatures. These subtypes are associated with distinct functional properties including differences in their level of quiescence and repopulation capacity that are retained through serial transplantation. The ability to generate grafts in secondary repopulation assays formally establishes their stem cell identity. Importantly, these two LSC subtypes could be prospectively purified based on CD112 cell surface expression, demonstrating that the two subtypes are distinct stem cell entities and not a result of stochastic cell cycle fluctuations within a homogenous LSC population. Our findings complement studies showing the existence of genetic subclones across primary AML samples [17, 50], as we have also recently demonstrated for the OCI-AML22 model [34], thereby bringing additional insight into LSC properties at the transcriptomic and epigenetic level.
The identification of two novel functional LSC subsets in AML that differ at the transcriptomic and epigenetic level is of particular interest as it is widely considered that transcriptional and epigenetic features are more amenable to pharmacological perturbation [51] than mutationally-driven alterations [15, 52,53,54,55,56,57,58,59,60]. Given their specific signatures, we could predict that the identified LSC subtypes may exhibit different responses to individual therapies. We previously reported a cellular hierarchy classification for AML that represents a powerful approach to predict drug response [33]. However, in this prior study we did not link any specific LSC subtype to a distinct hierarchy classification nor investigated the possibility for patients to present multiple LSC types, as no markers had been identified that could reliably separate functional LSC subsets.
We have now shown that each LSC subtype is differentially resistant to Cytarabine, a drug that is part of the first line therapy regimen for AML patients. Our findings uncover new opportunities to understand fundamental differences between LSC resistant versus LSC sensitive to chemotherapies. The ability to exploit CD112 as a means for prospective enrichment of these distinct LSC subsets now provides a crucial tool for refining our understanding of the LSC compartment, including identifying specific vulnerabilities of all the LSC subsets that might exist in an individual patient.
The OCI-AML22 model was central to our study as we were able to undertake deep characterization and purification of the distinct LSC fractions. The LSC signatures derived from the distinct LSC subtypes found in OCI-AML22 were also captured within LSC fractions extracted across a large cohort of AML patient samples; congruent with our model distinct LSC subsets can even co-exist within the same AML patient sample. Collectively, these findings highlight the utility of the OCI-AML22 model to unlock LSC properties at a functional and mechanistic level and point to the broad biological relevance of our findings. Thus, we envision that the OCI-AML22 model, combined with the potential to derive genetic sublines or increase clonal diversity via CRISPR editing and lentiviral technologies [34], will serve as a platform for interrogating inter- and intra-patient LSC heterogeneity and extracting shared intrinsic and even clone-dependent LSC vulnerabilities. Altogether, our study sets the stage to unlock new insights into the mechanisms governing stemness in leukemia and beyond. It also provides a framework to explore LSC subtypes and their specific drug sensitivities enabling a better design of combinatorial therapies to eradicate the entire pool of LSC that might be present in a given patient.
Material and methods
Ethics approval and consent to participate statement
Research involving human subjects, human material, or human data is in accordance with the Declaration of Helsinki. This research has been approved by the University Health Network (UHN) Research Ethics Board (reference: REB 01-0573-C). informed consent was obtained from all subjects, according to the procedures approved by the University Health Network (UHN) Research Ethics Board (REB 01-0573-C).
Please, refer to Supplementary Material for detailed methods.
Data availability
Processed files for multiome (scRNA-Seq/scATAC-Seq) are deposited under the NCBI’s Gene Expression Omnibus, accessible through GEO series accession GSE272187.
References
Eppert K, Takenaka K, Lechman ER, Waldron L, Nilsson B, van Galen P, et al. Stem cell gene expression programs influence clinical outcome in human leukemia. Nat Med. 2011;17:1086–93.
Ng SWK, Mitchell A, Kennedy JA, Chen WC, McLeod J, Ibrahimova N, et al. A 17-gene stemness score for rapid determination of risk in acute leukaemia. Nature. 2016;540:433–7.
van Rhenen A, Feller N, Kelder A, Westra AH, Rombouts E, Zweegman S, et al. High stem cell frequency in acute myeloid leukemia at diagnosis predicts high minimal residual disease and poor survival. Clin Cancer Res. 2005;11:6520–7.
Gentles AJ, Plevritis SK, Majeti R, Alizadeh AA. Association of a leukemic stem cell gene expression signature with clinical outcomes in acute myeloid leukemia. JAMA. 2010;304:2706–15.
Thomas D, Majeti R. Biology and relevance of human acute myeloid leukemia stem cells. Blood. 2017;129:1577–85.
Benveniste P, Frelin C, Janmohamed S, Barbara M, Herrington R, Hyam D, et al. Intermediate-term hematopoietic stem cells with extended but time-limited reconstitution potential. Cell Stem Cell. 2010;6:48–58.
Laurenti E, Frelin C, Xie S, Ferrari R, Dunant CF, Zandi S, et al. CDK6 levels regulate quiescence exit in human hematopoietic stem cells. Cell Stem Cell. 2015;16:302–13.
Notta F, Zandi S, Takayama N, Dobson S, Gan OI, Wilson G, et al. Distinct routes of lineage development reshape the human blood hierarchy across ontogeny. Science. 2016;351:aab2116.
Rodriguez-Fraticelli AE, Weinreb C, Wang S-W, Migueles RP, Jankovic M, Usart M, et al. Single-cell lineage tracing unveils a role for TCF15 in haematopoiesis. Nature. 2020;583:585–9.
Notta F, Doulatov S, Laurenti E, Poeppl A, Jurisica I, Dick JE. Isolation of single human hematopoietic stem cells capable of long-term multilineage engraftment. Science. 2011;333:218–21.
Kaufmann KB, Zeng AGX, Coyaud E, Garcia-Prat L, Papalexi E, Murison A, et al. A latent subset of human hematopoietic stem cells resists regenerative stress to preserve stemness. Nat Immunol. 2021;22:723–34.
Kaufmann KB, Zeng AGX, Coyaud E, Garcia-Prat L, Papalexi E, Laurent EMN et al. A distinct subset of human blood stem cells resists regenerative stress to preserve stemness. Soc Sci Res Netw: Rochester, NY, 2020. https://papers.ssrn.com/abstract=3612406.
Morita K, Wang F, Jahn K, Hu T, Tanaka T, Sasaki Y, et al. Clonal evolution of acute myeloid leukemia revealed by high-throughput single-cell genomics. Nat Commun. 2020;11:5327.
Schuringa JJ, Bonifer C. Dissecting clonal heterogeneity in AML. Cancer Cell. 2020;38:782–4.
Shlush LI, Zandi S, Mitchell A, Chen WC, Brandwein JM, Gupta V, et al. Identification of pre-leukemic hematopoietic stem cells in acute leukemia. Nature. 2014;506:328–33.
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;469:356–61.
Boer de B, Prick J, Pruis MG, Keane P, Imperato MR, Jaques J, et al. Prospective isolation and characterization of genetically and functionally distinct AML subclones. Cancer Cell. 2018;34:674–89.e8.
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;481:506–10.
Miles LA, Bowman RL, Merlinsky TR, Csete IS, Ooi AT, Durruthy-Durruthy R, et al. Single-cell mutation analysis of clonal evolution in myeloid malignancies. Nature. 2020;587:477–82.
Hope KJ, Jin L, Dick JE. Acute myeloid leukemia originates from a hierarchy of leukemic stem cell classes that differ in self-renewal capacity. Nat Immunol. 2004;5:738–43.
Pollyea DA, Jordan CT. Therapeutic targeting of acute myeloid leukemia stem cells. Blood. 2017;129:1627–35.
Jin L, Hope KJ, Zhai Q, Smadja-Joffe F, Dick JE. Targeting of CD44 eradicates human acute myeloid leukemic stem cells. Nat Med. 2006;12:1167–74.
Jin L, Lee EM, Ramshaw HS, Busfield SJ, Peoppl AG, Wilkinson L, et al. Monoclonal antibody-mediated targeting of CD123, IL-3 receptor alpha chain, eliminates human acute myeloid leukemic stem cells. Cell Stem Cell. 2009;5:31–42.
Majeti R, Chao MP, Alizadeh AA, Pang WW, Jaiswal S, Gibbs KD, et al. CD47 is an adverse prognostic factor and therapeutic antibody target on human acute myeloid leukemia stem cells. Cell. 2009;138:286–99.
Saito Y, Kitamura H, Hijikata A, Tomizawa-Murasawa M, Tanaka S, Takagi S, et al. Identification of Therapeutic Targets for Quiescent, Chemotherapy-Resistant Human Leukemia Stem Cells. Sci Transl Med. 2010;2:17ra9.
Laverdière I, Boileau M, Neumann AL, Frison H, Mitchell A, Ng SWK, et al. Leukemic stem cell signatures identify novel therapeutics targeting acute myeloid leukemia. Blood Cancer J. 2018;8:1–16.
Jung N, Dai B, Gentles AJ, Majeti R, Feinberg AP. An LSC epigenetic signature is largely mutation independent and implicates the HOXA cluster in AML pathogenesis. Nat Commun. 2015;6:8489.
Corces MR, Buenrostro JD, Wu B, Greenside PG, Chan SM, Koenig JL, et al. Lineage-specific and single cell chromatin accessibility charts human hematopoiesis and leukemia evolution. Nat Genet. 2016;48:1193–203.
Murphy T, Ng SWK, Zhang T, King I, Arruda A, Claudio JO, et al. Trial in progress: feasibility and validation study of the LSC17 score in acute myeloid leukemia patients. Blood. 2019;134:2682.
Duployez N, Marceau-Renaut A, Villenet C, Petit A, Rousseau A, Ng SWK, et al. The stem cell-associated gene expression signature allows risk stratification in pediatric acute myeloid leukemia. Leukemia. 2019;33:348–57.
Ng SW, Murphy T, King I, Zhang T, Mah M, Lu Z et al. A clinical laboratory-developed LSC17 stemness score assay for rapid risk assessment of acute myeloid leukemia patients. Blood Adv. 2021. https://doi.org/10.1182/bloodadvances.2021005741.
van Galen, Hovestadt P, Wadsworth V, Ii MH, Hughes TK, Griffin GK, et al. Single-cell RNA-Seq reveals AML hierarchies relevant to disease progression and immunity. Cell. 2019;176:1265–81.e24.
Zeng AGX, Bansal S, Jin L, Mitchell A, Chen WC, Abbas HA, et al. A cellular hierarchy framework for understanding heterogeneity and predicting drug response in acute myeloid leukemia. Nat Med. 2022;28:1212–23.
Boutzen H, Madani Tonekaboni SA, Chan-Seng-Yue M, Murison A, Takayama N, Mbong N, et al. A primary hierarchically organized patient-derived model enables in depth interrogation of stemness driven by the coding and non-coding genome. Leukemia. 2022;36:2690–704.
Kan WL, Dhagat U, Kaufmann KB, Hercus TR, Nero TL, Zeng AGX, et al. Distinct assemblies of heterodimeric cytokine receptors govern stemness programs in leukemia. Cancer Discov. 2023;13:1922–47.
Vujovic A, de Rooij L, Chahi AK, Chen HT, Yee BA, Loganathan SK, et al. In vivo screening unveils pervasive RNA-binding protein dependencies in leukemic stem cells and identifies ELAVL1 as a therapeutic target. Blood Cancer Discov. 2023;4:180–207.
Paczulla AM, Rothfelder K, Raffel S, Konantz M, Steinbacher J, Wang H, et al. Absence of NKG2D ligands defines leukaemia stem cells and mediates their immune evasion. Nature. 2019;572:254–9.
Raffel S, Klimmeck D, Falcone M, Demir A, Pouya A, Zeisberger P, et al. Quantitative proteomics reveals specific metabolic features of acute myeloid leukemia stem cells. Blood. 2020;136:1507–19.
Lagadinou ED, Sach A, Callahan K, Rossi RM, Neering SJ, Minhajuddin M, et al. BCL-2 inhibition targets oxidative phosphorylation and selectively eradicates quiescent human leukemia stem cells. Cell Stem Cell. 2013;12:329–41.
van Galen P, Mbong N, Kreso A, Schoof EM, Wagenblast E, Ng SWK, et al. Integrated stress response activity marks stem cells in normal hematopoiesis and leukemia. Cell Rep. 2018;25:1109–17.e5.
Zeng A. BoneMarrowMap: A single cell RNA-seq reference map of human hematopoietic development in the bone marrow, with balanced representation of hematopoietic stem and progenitor cells and differentiated populations. Github https://github.com/andygxzeng/BoneMarrowMap (accessed 20 Dec2023).
Takayama N, Murison A, Takayanagi S-I, Arlidge C, Zhou S, Garcia-Prat L, et al. The transition from quiescent to activated states in human hematopoietic stem cells is governed by dynamic 3D genome reorganization. Cell Stem Cell. 2021;28:488–501.e10.
Schwartz GW, Zhou Y, Petrovic J, Fasolino M, Xu L, Shaffer SM, et al. TooManyCells identifies and visualizes relationships of single-cell clades. Nat Methods. 2020;17:405–13.
Schwartz GW, Zhou Y, Petrovic J, Pear WS, Faryabi RB. TooManyPeaks identifies drug-resistant-specific regulatory elements from single-cell leukemic epigenomes. Cell Rep. 2021;36:109575.
Itoh K, Tezuka H, Sakoda H, Konno M, Nagata K, Uchiyama T, et al. Reproducible establishment of hemopoietic supportive stromal cell lines from murine bone marrow. Exp Hematol. 1989;17:145–53.
Tirosh I, Izar B, Prakadan SM, Wadsworth MH, Treacy D, Trombetta JJ, et al. Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science. 2016;352:189–96.
O’Reilly E, Zeinabad HA, Szegezdi E. Hematopoietic versus leukemic stem cell quiescence: challenges and therapeutic opportunities. Blood Rev. 2021;50:100850.
Rehman SK, Haynes J, Collignon E, Brown KR, Wang Y, Nixon AML, et al. Colorectal cancer cells enter a diapause-like DTP state to survive chemotherapy. Cell. 2021;184:226–42.e21.
Li L, Fridley B, Kalari K, Jenkins G, Batzler A, Safgren S, et al. Gemcitabine and cytosine arabinoside cytotoxicity: association with lymphoblastoid cell expression. Cancer Res. 2008;68:7050–8.
Klco JM, Spencer DH, Miller CA, Griffith M, Lamprecht TL, O’Laughlin M, et al. Functional heterogeneity of genetically defined subclones in acute myeloid leukemia. Cancer Cell. 2014;25:379–92.
Losman J-A, Looper RE, Koivunen P, Lee S, Schneider RK, McMahon C.et al. (R)-2-hydroxyglutarate is sufficient to promote leukemogenesis and its effects are reversible. Science. 2013;339:1621–5.
Cancer Genome Atlas Research Network, Ley TJ, Miller C, Ding L, Raphael BJ, Mungall AJ, et al. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N. Engl J Med. 2013;368:2059–74.
Figueroa ME, Lugthart S, Li Y, Erpelinck-Verschueren C, Deng X, Christos PJ, et al. DNA methylation signatures identify biologically distinct subtypes in acute myeloid leukemia. Cancer Cell. 2010;17:13–27.
Boutzen H, Saland E, Larrue C, de Toni F, Gales L, Castelli FA, et al. Isocitrate dehydrogenase 1 mutations prime the all-trans retinoic acid myeloid differentiation pathway in acute myeloid leukemia. J Exp Med. 2016;213:483–97.
Epigenetic Therapies for Cancer | NEJM. https://www.nejm.org/doi/full/10.1056/NEJMra1805035?query=TOC.
Shih AH, Abdel-Wahab O, Patel JP, Levine RL. The role of mutations in epigenetic regulators in myeloid malignancies. Nat Rev Cancer. 2012;12:599–612.
Hua X, Zhao W, Pesatori AC, Consonni D, Caporaso NE, Zhang T, et al. Genetic and epigenetic intratumor heterogeneity impacts prognosis of lung adenocarcinoma. Nat Commun. 2020;11:2459.
Chan SM, Thomas D, Corces-Zimmerman MR, Xavy S, Rastogi S, Hong W-J, et al. Isocitrate dehydrogenase 1 and 2 mutations induce BCL-2 dependence in acute myeloid leukemia. Nat Med. 2015;21:178–84.
Tovy A, Reyes JM, Gundry MC, Brunetti L, Lee-Six H, Petljak M, et al. Tissue-biased expansion of DNMT3A-mutant clones in a mosaic individual is associated with conserved epigenetic erosion. Cell Stem Cell. 2020;27:326–35.e4.
Figueroa ME, Abdel-Wahab O, Lu C, Ward PS, Patel J, Shih A, et al. Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell. 2010;18:553–67.
Acknowledgements
JED is supported by funds from the: Princess Margaret Cancer Center Foundation, Ontario Institute for Cancer Research through funding provided by the Government of Ontario, Canadian Institutes for Health Research (RN380110—409786), Canadian Cancer Society (grant #703212 (end date 2019), #706662 (end date 2025), Terry Fox New Frontiers Program Project Grant (Project# 1106), a Canada Research Chair, Princess Margaret Cancer Center, The Princess Margaret Cancer Foundation, and Ontario Ministry of Health. HB was supported by the Helena Lam Fellowship in Cancer Research Fund. JCYW is supported by funds from the Leukemia and Lymphoma Society of Canada (grant#1042891). We thank the Leukemia Tissue Bank at Princess Margaret Cancer Center/ University Health Network as the source of primary sample blood. We thank Sayyam Shah and Andy Zeng for helpful discussions and members of John Dick lab for reading the manuscript. This manuscript is dedicated to the memory of my dad.
Author information
Authors and Affiliations
Contributions
HB conceived and supervised the study, performed research, performed bioinformatic analysis and wrote the manuscript. AM performed bioinformatic analysis and edited the manuscript, CA and AO. Performed research, SB gathered datasets, ML provided funding. JCYW edited the manuscript and provided conceptual input. KBK performed research, provided conceptual input and wrote the manuscript. JED edited the manuscript, provided conceptual input, secured fundings and supervised the study.
Corresponding authors
Ethics declarations
Competing interests
JED reports receiving a commercial research grant from Celgene/BMS, and JED and JCYW have patents licensed to Trillium Therapeutics/Pfizer.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Boutzen, H., Murison, A., Oriecuia, A. et al. Identification of leukemia stem cell subsets with distinct transcriptional, epigenetic and functional properties. Leukemia (2024). https://doi.org/10.1038/s41375-024-02358-9
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41375-024-02358-9
- Springer Nature Limited