Pre-existing stem cell heterogeneity dictates clonal responses to acquisition of cancer driver mutations

16 17 Cancer cells display wide phenotypic variation even across patients with the same 18 mutations. Differences in the cell of origin provide a potential explanation, but these 19 assays have traditionally relied on surface markers, lacking the clonal resolution to 20 distinguish heterogeneous subsets of stem and progenitor cells. To address this 21 challenge, we developed STRACK, an unbiased framework to longitudinally trace 22 clonal gene expression and expansion dynamics before and after acquisition of 23 cancer mutations. We studied two different leukemia driver mutations, Dnmt3a-24 R882H and Npm1cA, and found that the response to both mutations was highly 25 variable across different stem cell states. Specifically, a subset of differentiation-26 biased stem cells, which normally become outcompeted with time, can efficiently 27 expand with both mutations. Npm1c mutations surprisingly reversed the intrinsic bias 28 of the clone-of-origin, with stem-biased clones giving rise to more mature malignant 29 states. We propose a clonal “reaction norm”, in which pre-existing clonal states 30 dictate different cancer phenotypic potential. 31 32

Single-cell, cancer initiation, cell-of-origin, lineage tracing, Dnmt3a, 33 Npm1c, clonal hematopoiesis, myeloid malignancies 34 35 36 Highlights: 37 - Single cell tracing of cancer initiation at the clonal level (STRACK). 38 - Ex vivo expansion cultures sustain intrinsic and heritable HSC heterogeneity. 39 - Premalignant mutations enhance the self-renewal of high-output stem cells, 40 increasing their survival probability. 41 - Transforming mutations reprogram low-output stem cell fates to more mature 42 malignant states. 43 44 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint

45 46 Cancer cells display striking phenotypic variation, within and across patients, yet the 47 origins of this variation are still unclear (Lenz et al. 2022). Since cancer is a clonal 48 disorder, arising from a single cell, researchers have long hypothesized that 49 phenotypic heterogeneity could be a consequence of the cell type that acquires the 50 driver mutations (Blanpain 2013; Visvader 2011). The “cell-of-origin” model 51 transformed the cancer field, leading to transformational discoveries across various 52 tumors and cell types (Baggiolini et al. 2021; Rajbhandari et al. 2023). However, 53 these classic cell-of-origin studies have several limitations. Firstly, they usually 54 induce mutations at a population level, lacking information about individual cell 55 heterogeneity. Second, they rely on reporter genes or surface-markers to isolate the 56 cell population of interest, and are thus biased by the tools and prior knowledge of 57 the model system. Finally, there are currently no methods that capture, with high 58 resolution, the pre-existing states (and fates) of the single cells that give rise to the 59 malignancy after mutagenesis. Solving these limitations could be critical to deepen 60 our understanding of cancer initiation mechanisms. 61 62 The cell-of-origin hypothesis has been extensively characterized in cancer types with 63 few driver mutations, such as myeloid malignancies. Depending on whether 64 mutations are introduced in the hematopoietic stem cells (HSCs), at the top of the 65 hematopoietic hierarchy, or in the more mature myeloid progenitors (MPs), 66 researchers have consistently shown differences in the resulting phenotypes (A. V. 67 Krivtsov et al. 2013; SanMiguel, Eudy, Loberg, Miles, et al. 2022; Stavropoulou et al. 68 2016; Cai et al. 2020; Zeisig et al. 2021; Taussig et al. 2010; Cozzio et al. 2003; 69 Huntly et al. 2004; Andrei V. Krivtsov et al. 2006; George et al. 2016). However, 70 recent single-cell sequencing studies have elucidated that stem and progenitor cell 71 populations are highly heterogeneous and cannot be simply dissected through 72 surface markers (Paul et al. 2016; Giladi et al. 2018). Furthermore, we and others 73 have shown that even the HSCs, at the top of the hierarchy, are long-term biased at 74 the level of both state and function, with a multiplicity of fate-imprinted clonal 75 hierarchies likely co-existing in the bone marrow (Rodriguez-Fraticelli et al. 2020; L. 76 Li et al. 2023; Meng et al. 2023; Jindal et al. 2023; Perié et al. 2015; Wang et al. 77 2022; Wagner and Klein 2020; Weinreb et al. 2020; Rodriguez-Fraticelli et al. 2018; 78 Tian et al. 2021; Naik et al. 2013; Dykstra et al. 2007; Wilson et al. 2015). Yet, due to 79 the lack of high-resolution cell-of-origin techniques, the functional significance of 80 stem cell heterogeneity in tumor initiation remains poorly understood (Haas, Trumpp, 81 and Milsom 2018), leaving researchers to rely solely on inference (Tong et al. 2021). 82 83 Here, we present a system called STRACK (Simultaneous Tracking of Recombinase 84 Activation and Clonal Kinetics), that precisely addresses these knowledge gaps, 85 unbiasedly linking pre-existing stem cell states (and intrinsic fates) with their potential 86 cancer states and fates. STRACK takes advantage of defined primary stem cell 87 culture systems to explicitly minimize the confounding effect of extrinsic variables 88 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint and solely focus on intrinsic determinants. To this end, STRACK combines long-term 89 ex vivo PVA-based expansion cultures, which can sustain and expand HSCs and 90 myeloid progeny for weeks (Wilkinson et al. 2019; Che et al. 2022), mouse models 91 carrying different Cre/Flp-inducible leukemia mutations (Loberg et al. 2019), a new 92 palette of LARRY expressed barcode libraries to track clones (Weinreb et al. 2020), 93 and a sister-cell clone splitting strategy (Weinreb et al. 2020; Tian et al. 2021). This 94 unique combination of methods allowed us to sample the system longitudinally and 95 obtain a dense clonal and transcriptional landscape for the same set of clones, with 96 and without mutations. 97 98

99 100 PVA-based expansion cultures maintain heterogeneous and deterministic clonal 101 behaviors 102 103 In order to characterize HSC clonal behaviors in PVA-based ex vivo expansion 104 cultures, we profiled thousands of HSC clones through a 27-day protocol using 105 single-cell lineage tracing and RNA profiling. For this, we genetically labeled ~10,000 106 long-term hematopoietic stem cells (Lineage- Sca-1+ cKit+ CD48- CD150+ CD201+ 107 or “E-SLAM”) with new LARRY barcoding libraries expressing the mT-Sapphire 108 fluorescent protein (Figure S1A). Labeled HSCs were expanded across multiple 109 wells for 27 days, sorted and randomly sampled for scRNAseq as depicted in the 110 schematic at day 7, 14 and 27 (Figure 1A and Supplementary Table 1 ~see 111 methods)(Weinreb et al. 2020). 112 113 Most of the cells profiled on day 7 expressed markers of HSCs ( Procr, Hlf, Mecom) 114 or MPPs ( Cd48)(Figure S1B). Starting on day 14, but mostly at day 27, we found a 115 continuum of differentiating cell states that we annotated using marker genes to 116 seven major cluster groupings: granulocyte monocyte progenitors (GMP), 117 megakaryocyte progenitors (Mk), Erythrocyte progenitors (Ery), Basophil progenitors 118 (Ba), Monocyte progenitors (Mono) and Neutrophil progenitors (Neu)(Figure 1B,C 119 and S1B - markers detailed on Supplementary Table 2). Even at day 27, we could 120 still annotate thousands of cells as HSCs, confirming their expansion within these 121 cultures (Figure 1C). This result suggested progressive differentiation and self-122 renewal from the initial pool of stem cells as previously reported (Wilkinson et al. 123 2019; Che et al. 2022). 124 125 Next, we leveraged LARRY barcoding to assess the clonal dynamics and cell fate 126 choices of expanded HSCs over time. We found that expansion cultures gradually 127 lose their clonality, despite initiating from highly pure EPCR+ HSCs, in line with 128 recent reports (Figure 1D)(Zhang et al. 2024). Sister-cell splitting across independent 129 wells confirmed the preferential expansion of specific clones, which correlated 130 across wells higher than expected based on a null distribution obtained from a 131 sampling simulation (Figure 1D). To describe the mechanisms leading to clonal 132 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint selection, we visualized all clones detected in both D7 and D14 timepoints using 133 “clone x state” heatmaps, which are colored based on the state-bias in each 134 individual clone (Figure 1E). We observed a striking heterogeneity in lineage fate 135 biases across clones even at D27 (Figure 1F, Figure S1C). More surprisingly, fate-136 biased HSCs showed the same fate-associated gene expression programs, which 137 were similar to those previously identified in transplant or developmentally-traced 138 native hematopoiesis (Figure S1D-E) (Rodriguez-Fraticelli et al. 2020; L. Li et al. 139 2023). These results indicate the preservation of clonal HSC heterogeneity and their 140 transcriptional programs ex vivo. 141 142 We next compared the HSC clones that persist and expand until D27 (termed “high-143 fitness clones”) with the clones that are detected early but do not expand enough to 144 be detected at later time points (termed “low-fitness clones”). Analysis of their fate 145 properties indicated that outcompeted low-fitness clones had more rapid contribution 146 to mature states (at D14) yet similar absolute numbers of HSCs, pointing towards 147 intrinsic qualities (and not quantities) as the cause of fitness differences (Figure 1E, 148 Figure S1F). To assess the early transcriptional differences associated with these 149 differences in fitness, we used longitudinal retrospective state-fate analysis, 150 comparing the HSC cell states at day 7 based on their fitness differences at day 27 151 (Figure 1G-H and Supplementary Table 3). High-fitness HSCs exhibited enriched 152 expression of markers associated with self-renewal ( Procr) and HSC-identity 153 (Mecom, Ly6a, Hlf), as well as non-conventional retinoic signaling (Rarb ), 154 extracellular-matrix (Sdc4, Mmp16), synapses (Dlg2, Ncam2) and actin cytoskeleton 155 regulation (Fmnl2, Gimp, Palld)(Figure 1H). High-fitness HSC clones also expressed 156 higher levels of low-output and Mk-biased HSC signatures, as well as Skeletal 157 morphogenesis signatures (e.g. Tcf15, Myof) (Figure 1I and and Supplementary 158 Table 4). To validate these findings during expansion, we used the Tcf15-Venus 159 mouse model, which enriches highly self-renewing HSCs (Rodriguez-Fraticelli et al. 160 2020). We sorted 500 CD45.1 wild-type E-SLAM HSCs and co-cultured them with 161 either Tcf15 high or Tcf15 low E-SLAM HSCs (from a CD45.2 background). Tcf15 high 162 HSCs consistently expanded whereas Tcf15 low cells were relatively outcompeted 163 after 20 days in expansion cultures (Figure 1J). Taken together, these analyses 164 indicate that ex vivo expansion cultures display a broad range of stable and dynamic 165 fate behaviors, including intrinsic differences in fitness, which allows us to study how 166 different stem cell clones respond upon acquisition of initiating cancer mutations. 167 168 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 169 Figure 1. State-fate analysis in ex vivo HSC expansion cultures. 170 A) Experimental design for state-fate analysis in ex vivo expansion cultures. 171 B) UMAP of integrated data from day 7, day 14 and day 27 sampling timepoints. 172 C) Distribution of cell state proportions across cluster groups in each sampling timepoint. 173 D) Observed versus expected number of clones based on a stochastic sampling model. Right, 174 spearman correlation of sister-cell clone sizes across split independent wells (observed, yellow; 175 blue, expected). 176 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint E) State-bias heatmaps showing clones (rows) and clusters/states (columns), colored by the intra-177 clonal fraction in each cluster/state. Clones shown are all those detected in both day 7 and day 178 14. Clones are separated into two heatmaps (top - bottom), depending on whether the clone is 179 detected (in at least 2 cells) at day 27. 180 F) UMAP showing example clonal behaviors at day 27 from different experiments/replicates. 181 G) Scheme of the interpretation of results based on clonal groupings. 182 H) Volcano plot for day 7 HSCs comparing low-vs-high fitness clonal groups. Selected genes are 183 highlighted. Markers associated with each group are shown in green (high-fitness) or red (low-184 fitness). 185 I) GSEA of hallmarks, GO-terms and HSC signatures for the differential gene expression results of 186 high versus low-fitness HSCs at day 7. 187 J) Validation experiment using competitive ex vivo HSC expansion of Tcf15-Venushigh versus Tcf15-188 Venuslow cells. 189 190 Sister-cell analysis and state-fate landscapes for Dnmt3a-R878H mutagenesis 191 192 To investigate how different cancer driver mutations influence stem cell fates, we 193 developed a second set of lentiviral libraries constitutively expressing Cre-194 recombinase and a fluorescent reporter or the mock fluorescent reporter alone (Cre-195 P2A-mScarlet and mScarlet) and combined them with a conditional mouse model 196 carrying a Cre-dependent Dnmt3a-R878H mutation, which is the mouse homolog of 197 Dnmt3a-R882H, one of the most frequent driver mutations in acute myeloid 198 leukemias (Loberg et al. 2019; Guryanova et al. 2016). We isolated HSCs from male 199 and female mice, transduced them with differently-indexed T-Sapphire LARRY 200 libraries, and then, on day 7, we profiled a part of the cells and split the remainder 201 into a Cre or a mock labeling reaction, and then these were further split into separate 202 wells that continued expansion independently(Figure 2A). The system allowed state-203 fate analysis for both wild-type and mutant clones arising from sister HSCs, which 204 we termed scTRACK (simultaneous cell tracking of recombinase activation and 205 clonal kinetics). 206 207 Similar to previous reports, we observed increased expansion of Dnmt3a R878H 208 mutant cells (from here on R878H cells) in competitive cultures (Figure S2A), but did 209 not identify major differences in their states (at the population level) compared to 210 wild-type (WT) controls from the same mouse line (Figure S2B). After performing 211 clonal analysis, we observed that HSC clones expanding significantly more with 212 R878H also tended to have the largest clone sizes (Figure 2B). We then compared 213 the clones that could be detected in both day 7 as well as day 27 in both WT and 214 R878H cells by plotting their behaviors using state-bias heatmaps (Figure 2C). Sister 215 WT/R878H clones displayed remarkably similar behaviors, even with a mutation and 216 20 days after splitting. Still, we noticed that most clones gained relatively more HSCs 217 upon R878H mutation in comparison with the WT (Figure 2D), resulting in a net drop 218 in the clonal output activity (Figure S2C). In some rare but notable cases, high-output 219 multilineage clones even completely lost their output activity in the presence of the 220 R878H mutation (Figure 2E), indicating that high-fitness stem-cells can be 221 reprogrammed by the Dnmt3a mutation to gain self-renewal at the expense of 222 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint differentiation. Thus, while these clones possess an inherent capacity for expansion 223 irrespective of mutational status, the mutation drives further stemness capacity. 224 225 Considering these effects, we wondered whether the mutation could be altering the 226 fitness capacity of intrinsically low-fitness clones. Comparing the clonality of R878H 227 cultures with sampling simulations confirmed that R878H cultures maintained a 228 relatively more polyclonal pool, suggesting that additional clones, normally 229 outcompeted in the WT setting, persist only upon activation of the R878H mutation 230 (Figure 2F). We next compared the clones detected only in R878H cells (mutation-231 dependent) with those detected only in WT or in both WT and R878H conditions 232 (mutation-independent) (Figure 2G). We plotted state-bias heatmaps for day 27 233 states for each clone, split into groups based on their detection in WT and/or R878H 234 cultures (Figure 2H). Notably, mutation-dependent R878H clones at day 27 behaved 235 similarly compared to mutation-independent R878H clones, with increased HSC bias 236 and a relatively larger size compared to WT-only clones. We next used retrospective 237 state-fate analysis and compared the transcriptomes of day 7 HSCs based on their 238 R878H mutation dependency at day 27 (Figure 2I). While we could not identify 239 unique markers, gene set enrichment analysis (GSEA) of R878H-dependent versus 240 independent clones indicated negative enrichment of self-renewal and high-fitness 241 signatures, suggesting their origin in low-fitness HSCs (Figure 2J, top panel). 242 Interestingly, post-mutation R878H-dependent HSCs displayed positive enrichment 243 of high-fitness and self-renewal signatures (compared with mutation-independent 244 clones), indicating the potent stemness reprogramming capacity of these driver 245 mutations (Figure 2J, bottom panel) We further confirmed this by quantifying the 246 single-cell fitness scores of R878H-dependent and independent HSCs at day 27 247 (Figure S2D). In addition to changes in HSC output bias, differential expression 248 analysis and GSEA revealed that R878H HSCs (and MPPs) displayed reduced 249 expression of early response genes, suggesting dampened inflammatory responses 250 as an additional mechanism for their competitive expansion, in line with recent 251 studies in clonal hematopoiesis (Figure S2E-F)(Serine Avagyan and Zon 2023; 252 Jakobsen et al. 2023; S. Avagyan et al. 2021). Together, these results highlight the 253 differential effect of this cancer driver mutation across different stem cell clones. 254 While the R878H mutation can mildly enhance the stemness and expansion 255 properties of high-fitness stem cell clones, it can reprogram the fates and states of 256 low fitness stem cell clones, allowing them to survive and expand in ex vivo 257 expansion cultures. 258 259 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 260 Figure 2. Sister-cell state-fate landscape of Dnmt3a-R878H mutagenesis. 261 A) Experimental design for sister cell state-fate analysis in R878H mutagenesis. 262 B) Waterfall plot showing log2 fold-change in clone size proportion for the same set of clones with 263 and without the R878H mutation. 264 C) State-bias heatmap of clones observed in both wt and R878H cultures at day 27. Log2 fold-265 change in HSC bias is shown on the right. Every row corresponds (in both heatmaps) to a single 266 HSC clone. 267 D) Waterfall plot showing log2 fold-change in HSC bias (inverse of output activity) for the same set of 268 clones comparing R878H-mutant versus WT. 269 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint E) UMAP of sister cell wild-type and mutant fates at day 27. The example shows a high-output clone 270 that displays extreme low-output behavior with the R878H mutation. 271 F) Difference between observed and expected number of clones at day 27. 272 G) Schematic showing grouping of clones as mutation-dependent and mutation-independent. 273 H) State-bias heatmaps of clones at day 27 separated into 4 clone-groups, depending on whether 274 they are detected in both WT or R878H cultures (top heatmaps), WT-only or R878-only (bottom 275 heatmaps). Clone size proportion is indicated in the right column. The top bar-plot indicates the 276 proportion of clones with barcodes detected in each state (one column for each clone-group). 277 I) UMAP of subsetted HSCs at day 7. Clones detected at day 27 are highlighted (mutation-278 dependent, top; mutation-independent, bottom). 279 J) GSEA of hallmarks and HSC signatures for the comparison of R878H-dependent versus R878H-280 independent clones (at day 7, top; at day 27, bottom) 281 282 The Flt3-Cre model recapitulates Dnmt3a-R878H mutagenesis in low-fitness HSCs 283 in vivo 284 285 Next, we sought to validate the role of the Dnmt3a R878H mutation in low-fitness 286 HSCs using a complementary approach. We took advantage of a modified version of 287 the Flt3-switch approach, which normally labels developmentally restricted HSCs 288 (Beaudin et al. 2016; Stonehouse et al. 2024). We speculated that we could identify 289 low-fitness adult HSCs by combining the Flt3-Cre allele with the LSL-TdTomato 290 reporter (tdTom), which is easier to recombine compared to mTmG or LSL-EYFP. 291 We characterized the model at steady state in young mice using single-cell RNAseq 292 profiling of tdTom+ and tdTom- LSK (Lin–c-Kit+Sca-1+) and LKs (Lin–c-Kit+ Sca1-293 )(Figure 3A,B). We verified that tdTom+ cells populated most clusters, while tdTom- 294 were mostly restricted to HSC and MkP clusters, including bridging cells that suggest 295 a direct Mk-restricted pathway (Figure 3C and S3A-C)(Haas et al. 2015; Yamamoto 296 et al. 2018; Meng et al. 2023; Carrelha et al. 2018; Rodriguez-Fraticelli et al. 2018; 297 Morcos et al. 2022). Approximately ~50% of E-SLAM HSCs were labeled with 298 tdTom+, suggesting that the Flt3-Cre tdTom model allows to separate the Mk-299 restricted and non-Mk-restricted hierarchies (Figure 3D). We next performed 300 differential gene expression analysis on tdTom+ and tdTom- HSCs (Figure 3E). We 301 found that tdTom+ cells showed relative downregulation of stem cell-associated 302 genes like Mecom and positive enrichment for low-fitness and high output 303 signatures, as well as IFNa signaling, and IL6/JAK/STAT3 signaling (Figure 3F). 304 Thus, Flt3-Cre tdTom+ HSCs transcriptionally resemble low-fitness and high-output 305 HSCs. 306 307 To test the functional consequences of acquiring Dnmt3a R878H mutations in low-308 fitness HSCs, we generated Flt3-Cre; LSL-TdTomato; Dnmt3a-fl-R878H/+ mice and 309 verified the specific mutagenesis in this tdTom+ HSC population by genotyping PCR 310 (Figure S3D). Next, we compared the single-cell transcriptional landscape of R878H 311 tdTom+ with wild-type tdTom+ hematopoiesis (Figure 3B). We observed a relative 312 loss of myeloid cells and expansion of erythroid progenitors in the R878H, which is in 313 line with results obtained in the whole-marrow Mx1-Cre R878H model, suggesting 314 this phenotype can arise without mutagenesis in the most primitive tdTom- HSC 315 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint compartment (Figure S3E). We then subsetted and re-clustered tdTom+ HSCs. 316 R878H tdTom+ cells were relatively enriched in cluster c0, which expresses higher 317 levels of stemness regulators and quiescence-associated genes (Figure 3G,H). In 318 line with this, R878H tdTom+ HSCs showed upregulation of stemness genes 319 compared to WT tdTom+ HSCs (Hlf , Mecom), and GSEA showed relative 320 enrichment of low output, self-renewal and high-fitness signatures (Figure 3I,J). 321 Finally, we mixed 500 Flt3-Cre tdTom+ HSCs (R878H or wt) with 500 (age-matched) 322 CD45.1 wild-type HSCs using ex vivo expansion cultures (Figure 3K). We observed 323 a ~2-fold increased expansion of R878H tdTom+ HSCs over wild-type cells after 20 324 days, indicating the improved competitive advantage of R878H tdTom+ HSCs 325 (Figure 3L). Altogether, these results validate our findings in ex vivo expansion 326 cultures and demonstrate that the Dnmt3a-R878H mutation acquired in vivo within a 327 low-fitness HSC population is sufficient to reprogram their transcriptional state and 328 competitive behavior. 329 330 331 Figure 3. Dnmt3a-R878H rewires stem cell fitness in the Flt3-Cre HSC model. 332 A) Experimental design for analysis of Flt3-Cre HSC models with 10x scRNAseq 3’. 333 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint B) UMAP showing Flt3-Cre wt and Flt3-Cre R878H hematopoietic landscapes, with cells colored by 334 TdTom expression. Annotated cluster groups are indicated. 335 C) Barplot of an exemplary sample showing cluster distribution of TdTom+ and TdTom- cells. 336 D) Scheme showing interpretation of Flt3-Cre model results. 337 E) Volcano plot showing differential expression analysis of Flt3-Cre tdTom+ versus tdTom- HSCs. 338 F) GSEA showing hallmarks and HSC signatures for Flt3-Cre tdTom+ versus tdTom- HSCs. 339 G) UMAP of subsetted HSCs showing the distribution density. The dotted line marks the border 340 between cluster 0 (higher stemness markers) and clusters 1 and 2 (higher cell-cycle and 341 inflammatory response). 342 H) Barplot showing relative enrichment in cluster 0 versus clusters 1 and 2. 343 I) Volcano plot showing differential expression analysis of Flt3-Cre tdTom+ R878H-mutant versus 344 Flt3-Cre tdTom+ wild-type HSCs. 345 J) GSEA showing hallmarks and HSC signatures for Flt3-Cre tdTom+ R878H-mutant versus Flt3-346 Cre tdTom+ wild-type HSCs. 347 K) Experimental design for CD45.1 competition experiment using ex vivo expansion cultures. 348 L) Proportion of tdTom+ cells at day 20 comparing R878H versus wild-type. 349 350 Sister-cell analysis in Npm1c mutagenesis reveals clone-specific origins of mature 351 versus primitive malignant states 352 353 Npm1c mutations are another frequent initiating alteration in AML and have been 354 shown to endow stemness potential to mutant cells (Uckelmann et al. 2020; 355 SanMiguel, Eudy, Loberg, Young, et al. 2022; Brunetti et al. 2018). To investigate 356 the effects of Npm1c at clonal resolution, we developed a third set of lentiviral 357 libraries constitutively expressing EGFP alongside Flpo recombinase or alone, and 358 we then performed scTRACK using the same mouse model (Figure 4A). As with 359 Dnmt3a, we again observed a very heterogeneous but significant expansion of 360 Npm1cA mutant cells in competitive ex vivo expansion cultures (Figure 4B and S4A). 361 This expansion was accompanied by the enrichment of Npm1c cells in an HSC-like 362 state, which became conspicuous only at day 27, 20 days post-mutation (Figure 4C, 363 S4B-C). Compared to WT cultures, Npm1c cultures maintained almost perfect 364 clonality, losing less than 25% of the clones expected based on the stochastic 365 sampling model (Figure 4D). Based on prior findings, we speculated that contribution 366 from non-HSC clones (e.g. GMPs) could explain these results, but tracing back the 367 sister-cell states of these clones at day 7 confirmed that most clones still originated 368 in HSCs (Figure 4E). Based on our experience with Dnmt3a, we decided to classify 369 HSCs as Npm1c-independent or -dependent and compared their origins at day 7. 370 Npm1c-dependent HSCs showed a low fitness score at day 7, which became 371 reversed at day 27, highlighting the potent fitness-programming effect of the Npm1c 372 mutation (Figure S4D). Npm1c mutant cells showed expected gene expression 373 changes compared to WT cells, including increased expression of HoxA cluster 374 genes, proteasome and ribosomal components, and stemness markers (Figure S4E-375 I and Supplementary Table 3). To quantify clonal changes in fate behaviors in 376 response to Npm1c-mutagenesis, we compared sister cell clones that had been 377 profiled in both WT and Npm1c cultures at day 27 (Figure 4F-G). We noticed that 378 Npm1c mutation tended to expand HSCs and reduce output activity and My bias, but 379 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint this was highly variable across clones. Surprisingly, sister clones with low-output and 380 high HSC content in the WT setting displayed more mature states and decreased the 381 proportion of HSC-like cells after acquiring the Npm1c mutation. Conversely, sister 382 clones with high-output properties in the WT showed the most primitive and 383 differentiation-blocked behaviors in the context of the Npm1c mutation (Figure 4H 384 and S4J). This response was highly heritable, with independently-mutated sisters of 385 the original WT clone displaying similar behavior (Fig S4K). We next compared the 386 gene expression profiles of clones that became more mature with Npm1c mutation 387 (“HSC-decreased”) with clones that became more primitive (“HSC-increased”) 388 (Figure 4I and Supplementary Table 3). HSC-decreased Npm1c clones expressed 389 higher levels of mature malignant cell markers, including GMP genes ( Plac8, Mpo) 390 as well as various genes involved in AML function (Zeb2, Plzf)(H. Li et al. 2017; Ono 391 et al. 2013). Intriguingly, mature-like malignant clones maintained expression of 392 markers of their clone-of-origin, such as Itsn1 or Mir99ahg , both of which are low-393 output/high-fitness HSC-markers. We complemented this analysis by comparing 394 Npm1c clones based on the fitness score of their HSC of origin (at day 7), which 395 revealed similar results (Figure S4L-M). Finally, we evaluated these clone-specific 396 Npm1c signatures in human AML samples that were previously classified as 397 “mature” or “primitive” (Mer et al. 2021; Naldini et al. 2023) . Across two different 398 cohorts, bulk or scRNAseq, we observed that “HSC-decreased” Npm1c clonal 399 signatures were recapitulated in mature AML patients, suggesting an HSC-biased 400 clonal origin in these leukemias (Figure 4J and S4N). To summarize, Npm1c 401 mutations in low-output HSCs result in more mature malignancies, whereas 402 mutations in high-output HSCs result in more primitive and aggressive expansions of 403 malignant cells, contrary to our expectations. Together, our results point to a unique 404 clone-of-origin mechanism for leukemia phenotypic heterogeneity, with pre-existing 405 HSC states acting as a non-genetic substrate for the emergent properties of the 406 malignant disease. 407 408 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 409 Figure 4. Pre-existing stem cell states-fates determine unique properties in malignant Npm1c 410 clones. 411 A) Experimental design for sister cell state-fate analysis in Npm1c mutagenesis. 412 B) Waterfall plot showing log2 fold-change in clonal proportion for the same set of clones with and 413 without the Npm1c mutation. 414 C) Integrated UMAP showing Npm1c mutant cells (all clones combined) at day 14 (7 days after 415 mutagenesis) and day 27 (20 days after mutagenesis), colored by cluster groups. 416 D) Difference between observed and expected number of clones at day 27. 417 E) Barplot showing distribution of day 7 states for clones observed at day 27 (shown separately for 418 wt or Npm1c). The distribution of all day 7 states is shown for comparison. 419 F) Boxplot showing sister-cell clonal behavior changes at day 27. Data is expressed as log2 fold 420 change (Npm1c vs wt) in the indicated measurement (for all clones observed at day 27 in both 421 Npm1c and wt cultures). 422 G) State-bias heatmap of clones observed in both wt and Npm1c cultures at day 27. Clone size in 423 each culture as well as log2 fold-change in HSC bias is shown on the right. 424 H) UMAP of sister cell wild-type and mutant fates at day 27. The example shows a high-output clone 425 that displays primitive leukemic behavior with the Npm1c mutation. 426 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint I) Volcano plot results of differential gene expression analysis comparing Npm1c clones with HSC-427 increased versus Npm1c clones with HSC-decreased. Example markers are highlighted. 428 J) GSEA of the Npm1c “HSC-decreased” signature (based on markers) performed on the list of 429 mature vs. primitive NPM1 AML samples based on clustering of AML bulk-RNAseq data (Mer et 430 al. 2021) 431 432

433 434 Cancer heterogeneity presents a significant challenge to oncologists. Phenotypic 435 heterogeneity (plasticity, memory and noise) impacts critical therapeutic aspects of 436 cancer biology, such as treatment resistance and clonal dominance. Recently, 437 researchers have shown that phenotypic heterogeneity could be traced back to non-438 genetic variation at the level of individual cancer cells (Fennell et al. 2021; Goyal et 439 al. 2023). The findings we present here further suggest that cancer heterogeneity 440 may be partly influenced by subtle differences in the stem cell state of origin, leading 441 to distinct responses even upon acquisition of identical cancer driver mutations. 442 Classic cell-of-origin studies were limited by pre-existing tools and knowledge of cell 443 type specific markers, which others and we have revealed to be insufficient to 444 dissect the underlying heterogeneity of hematopoietic cell-types (Rodriguez-Fraticelli 445 et al. 2020; Weinreb et al. 2020; Tian et al. 2021). Our systematic approach, 446 combining sister-cell clone splitting with precise activation of cancer driver alleles, 447 empowers a conceptual and methodological shift from the traditional cell-of-origin 448 model to a more nuanced clone-of-origin paradigm, which embodies both lineage 449 and state information. 450 451 At the population level, we observed that the Dnmt3a R878H mutation (R882H in 452 humans) appeared to enhance the fitness and expansion of HSCs, consistent with 453 prior studies (Loberg et al. 2019; Guryanova et al. 2016). Compared to their wild-454 type counterparts, R878H HSCs also exhibited reduced inflammatory signaling, 455 which might add to their competitive edge, as recent CH studies in zebrafish, mice 456 and humans have suggested (S. Avagyan et al. 2021; Serine Avagyan and Zon 457 2023; Schwartz et al. 2024; Jakobsen et al. 2023). However, at the individual clone 458 level, R878H mutation had a more pronounced effect on high-output/low-fitness 459 HSCs, which typically differentiate quickly and are outcompeted in expansion 460 cultures, but, upon R878H mutation, gain enhanced stemness and fitness. Similarly, 461 the Npm1c mutation consistently activated Pbx/Hoxa cluster expression across all 462 clones, in line with its well described role in activating these developmental genes 463 (Uckelmann et al. 2020; Brunetti et al. 2018; SanMiguel, Eudy, Loberg, Miles, et al. 464 2022). Yet, the individual clonal responses to this mutation varied greatly; 465 unexpectedly, mutant sister cells from low-output HSCs tended to exhibit a more 466 mature-like hierarchy. Previous studies in MLL-rearranged AML models have 467 suggested that cancer cells inherit pre-existing functionalities of the cell of origin 468 (George et al. 2016; Stavropoulou et al. 2016). However, our Npm1c results 469 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint completely upset the notion that these state-biases should be maintained post 470 mutation. 471 472 Our research also highlights the dynamic nature of these phenotypic 473 transformations, which became apparent only 20 days post-mutagenesis, suggesting 474 a role for cell divisions or secreted factors in culture media. Past work has also 475 elucidated the role of genetic predispositions (Weinstock et al. 2023; Bick et al. 476 2020) and extrinsic regulation in the expansion of initiating (pre)-malignant clones 477 (SanMiguel, Eudy, Loberg, Young, et al. 2022; Hormaechea-Agulla et al. 2021; 478 Schwartz et al. 2024). In combination with our results, we would like to propose here 479 the concept of a 'clonal reaction norm,' where clonal lineage, genetic background, 480 driver mutations, and environmental factors collectively determine the fate and 481 properties of a resulting cancer clone. Notably, our different Npm1c clones exhibited 482 characteristics of various human NPM1 disease subtypes, indicating that 483 understanding these norms for different cancer driver genes may have an impact on 484 diagnostics and personalized treatments. 485 486 Finally, our experimental design should also be amenable to combinatorial and 487 sequential mutagenesis. Future studies will focus on analyzing additional driver 488 genes and systems beyond hematopoiesis. Looking ahead, we anticipate the 489 development of new mouse models that can reproducibly generate tumors from 490 clone-specific origins, offering a more accurate representation of the non-genetic 491 tumor diversity that characterizes human patients to advance precision medicine 492 efforts. 493 494 495 Study limitations 496 497 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint These studies have been performed using mice, due to the accessibility of precision 498 mutagenesis that can be achieved using genetically-engineered Cre/Flp-conditional 499 mouse alleles. In the future, it is possible that Prime Editing or similar CRISPR-500 based precision editing technologies achieve the efficiencies to enable similar 501 approaches in primary human HSCs (Geurts et al. 2021). 502 Our state-fate analysis uses ex vivo expansion cultures, to maximize our capacity to 503 maintain and track many stem cell clones longitudinally, in separate environments, 504 with and without mutations. We initially attempted to use stem cell transplantation to 505 track cancer and wild-type clones, but these assays were underpowered, tracking 506 too few clones, which is possibly due to mutant cell competition or niche 507 heterogeneity. Our defined primary stem cell culture system is powerful as a proof of 508 principle, but future technological implementations should address the role of niches 509 and extrinsic components, which may take advantage of organoids or co-culture 510 systems (Sommerkamp et al. 2021; Frenz-Wiessner et al. 2024; Khan et al. 2023). 511 Finally, our state-fate analysis relies on sister-cell splitting and clonal analysis, with 512 sister states/fates serving as a proxy. This is particularly necessary due to the fact 513 that our power relies on measuring 3 entities (mutant fate, wild-type fate and state) 514 across thousands of cells. Novel technologies (such as Live-seq) are emerging to 515 allow same-cell tracing without destruction, and may eventually have the capacity to 516 perform STRACK-type studies at scale (Chen et al. 2022). 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 542 543 544 545 546 SUPPLEMENTARY FIGURES 547 548 Supplementary Figure 1. LARRY barcoding highlights a variety of clonal HSC behaviors in 549 HSC expansion cultures 550 A) Principle for LARRY barcoding and state-fate analysis. Expressed barcodes can be captured 551 efficiently with scRNAseq profiling methods. Then, single-cell indexed cDNA libraries are split in 552 two parts: one for amplifying the whole transcriptome (GEX library) and one for amplifying the 553 clonal barcode specifically (LARRY library). These libraries are sequenced together and then 554 demultiplexed to feed to the CloneRanger pipeline. 555 B) UMAP showing scaled (min-to-max) expression for various cluster markers. 556 C) Integrated UMAP showing similar clonal behaviors in WT clones split across both wells (well 1 - 557 blue; well 2 - yellow). 558 D) Pearson correlation comparing sister cell pairs (cells from the same clone) versus shuffled or non-559 sister pairs at day 7 and day 27. * p<0.001 (permutation test) 560 E) GSEA of HSC signatures in D27 high-output HSCs versus low-output HSCs. 561 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint F) Percentage of clones with an HSC at D14 (comparing high versus low fitness) 562 563 564 Supplementary Figure 2. Differences in R878H effects at the population or clonal level. 565 A) Percentage of Dnmt3a-R878H mutant or wild-type cells after 20 day ex vivo expansion culture. 566 HSCs were isolated from Dnmt3a-LSL-R878H mice and transduced with mock or Cre 567 lentivectors. Then, 500 Cre mutant cells (or 500 mock cells) were co-cultured with 500 CD45.1 568 cells. Cre or wt mutant cells were measured at day 20.**** p<0.001 (two-sample t-test, n=9) 569 B) UMAP showing cluster annotations for Dnmt3a-R878H cells at different timepoints. 570 C) Boxplots showing changes in clonal behaviors (R878H versus WT) for the same clones measured 571 in both cultures at day 14 and day 27. 572 D) Boxplots showing low-fitness HSC signature score differences across R878H-dependent and 573 independent HSCs (at day 7 and day 27) ***p<0.001 (wilcoxon test) 574 E) Volcano plot showing differential expressed genes comparing R878H vs WT HSCs or MPPs (at 575 day 27). 576 F) GSEA of significant MsigDB hallmarks and GO terms in D27 R878H v WT HSCs. 577 578 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 579 Supplementary Figure 3. Flt3-Cre model labels different stem cell compartments and 580 trajectories. 581 A) UMAP showing HSPC cluster groupings as profiled from Flt3-Cre LSL-tdTom mice. 582 B) UMAP of subsetted HSC, MPP and MkP clusters, showing tdTom+ (red) and tdTom- (gray) cells. 583 A trajectory of tdTom- cells suggests evidence for negative labeling of a direct Mk pathway from 584 the most primitive HSCs. 585 C) UMAP of subsetted HSC/MPP/MkPs showing scaled expression of different HSC, MPP and MkP 586 markers. 587 D) Genotyping of R878H mutation in Flt3-Cre Dnmt3a-LSL-R878H LSL-tdTom model. Cells were 588 sorted and genotyped as in Loberg et al. 589 E) Erythroid bias in Dnmt3a-LSL-R878H mutant tdTom+ cells, compared to WT. Plot shows ratio 590 between the proportions of the Ery to GM clusters in each replicate. 591 F) Subclustering of subsetted tdTom+ HSCs showing 3 distinct clusters. Cluster 0 shows enrichment 592 in stemness markers (e.g. Mecom). 593 594 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 595 Supplementary Figure 4. Additional analyses on Npm1c malignant clones. 596 A) Percentage of Npm1c mutant or wild-type cells after 20 day ex vivo expansion culture. HSCs 597 were isolated from Npm1-FSF-cA mice and transduced with mock or Flpo lentivectors. Then, 500 598 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint Flpo mutant cells (or 500 mock cells) were co-cultured with 500 CD45.1 competitor cells. Flpo or 599 wt mutant cells were measured at day 20 by FACS.**** p<0.001 (two-sample t-test, n=9). 600 B) Example histogram of EPCR expression at day 27 in WT, R878H and Npm1c cultures. 601 C) Annotation of LSK and LK states in Npm1c versus wild-type clones at day 27. 602 D) Volcano plot showing differences between Npm1c and wild-type HSCs at day 14. 603 E) Volcano plot showing differences between Npm1c and wild-type HSCs at day 27. 604 F) Volcano plot showing differences between Npm1c and wild-type MPPs at day 27. 605 G) Volcano plot showing differences between Npm1c and wild-type GMPs at day 27. 606 H) GSEA of significant HSC signatures, MsigDB hallmarks and GO terms in D27 Npm1c v WT 607 HSCs. 608 I) Ucell scores for the Low-fitness signature at day 7 and day 27 for WT-only (Mutation-609 independent), R878H or Npm1c HSCs (Mutation-dependent). 610 J) UMAP of sister cell wild-type and mutant fates at day 27. This additional example shows a high-611 output clone that displays a more primitive leukemic behavior with the Npm1c mutation. 612 K) UMAP of sister cell mutant fates (2 different mutant subclones of the same pre-mutation WT 613 clone) at day 27. This example shows low-output primitive leukemic behavior for both subclones 614 detected. 615 L) Volcano plot showing differences between Npm1c clones derived from high-fitness (q1 - quartile 616 1) versus low-fitness (q4 - quartile 4) HSCs at day 7. Notice enrichment of stemness genes in q4-617 derived clones. 618 M) Volcano plot showing differences between WT clones derived from high-fitness (q1 - quartile 1) 619 versus low-fitness (q4 - quartile 4) HSCs at day 7. Notice enrichment of stemness genes in q1-620 derived clones. 621 N) GSEA of the Npm1c “HSC-decreased” signature performed on mature vs. primitive NPM1 AML 622 samples as profiled based on AML scRNAseq data (Naldini et al. 2023). Samples were assigned 623 to each group based on CD96 expression (primitive marker) and CD14 expression (mature 624 marker). 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 647 648 649

650 651 I.S. was supported through the European Union's Horizon 2020 research and 652 innovation program under the Marie Skłodowska-Curie grant agreement No 945352. 653 654 D.F.P was supported through the Marie Skłodowska-Curie grant agreement No 655 101109276. 656 657 A.R.F is supported by the Cris Foundation Excellence Award (PR_EX_2020-24), the 658 ERC Starting Grant MemOriStem (101042992), the Spanish National Research 659 Agency (PID2020-114638RA-I00), the Agencia de Gestio d’Ajuts Universitaris i de 660 Recerca (AGAUR, 2017 SGR 1322), and the CERCA Program/Generalitat de 661 Catalunya. A.R.F. acknowledges support from the Institut Catalá de Recerca i 662 Estudis Avançats (ICREA), the Ministry of Science Ramon y Cajal Fellowship, and 663 the LaCaixa Junior Fellows Incoming Fellowship. 664 665 The authors would like to acknowledge the technical assistance of David Fernández, 666 José Ignacio Pons, and Freddy Monteiro from the Functional Genomics Core Facility 667 at IRB Barcelona (single-cell library preparations and sequencing). We also 668 acknowledge assistance from the Flow Cytometry and Cell Sorting Core Facility at 669 the University of Barcelona (CCIT-UB) and from the facilities from the Parc Cientific 670 de Barcelona (PCB). The authors also wish to thank mentors and colleagues for 671 various helpful discussions. Illustrations were created with Biorender. 672 673 Author Contributions: 674 675 I.S. performed molecular biology, library preparations, cell culture, and animal 676 experiments. D.F.P. generated the analytic pipelines and performed bioinformatic, 677 single-cell sequencing, and statistical analyses. A.R.F and P.S. assisted in the 678 generation of the bioinformatic pipeline and analysis. A.R.F., I.S., and D.F.P 679 conceptualized the project design. A.R.F. designed the LARRY libraries. I.S. 680 generated all the vectors and produced the lentiviral libraries with assistance from 681 lab members. I.S. and D.F.P. wrote the manuscript with help from A.R.F. All authors 682 provided feedback and input to finalize the manuscript text. 683 684 Declaration of Interest: 685 A.R.F. is an advisor for Retro Bio. The authors declare no competing interests. 686 687 688 689 690 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 691 692 693 RESOURCE AVAILABILITY 694 695 Lead contact 696 Further information and requests for resources and reagents should be directed to 697 and will be fulfilled by the lead contact, Alejo E. Rodriguez-Fraticelli alejo.rodriguez-698 [email protected] 699 700

availability 701 Plasmids generated by this study are available upon request and will be deposited to 702 Addgene and the European Plasmid Repository. 703 704 Data and code availability 705 Code and data objects are available at figshare (DOI: 706 10.6084/m9.figshare.25822948). Raw and processed single-cell RNAseq, LARRY 707 barcoding and TotalSeq™ data generated for this study will be released at Gene 708 Expression Omnibus under the accession number GSE266232. 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 735

736 737 Mice and Animal Guidelines 738 739 All procedures involving animals adhered to the pertinent regulations and guidelines. 740 Approval and oversight for all protocols and strains of mice were granted by the 741 Institutional Review Board and the Institutional Animal Care and Use Committee at 742 Parque Científico de Barcelona under protocol CEEA-PCB-22-001-ARF. The study 743 follows all relevant ethical regulations. Mice were kept under specific pathogen-free 744 conditions for all experiments. 745 746 Hematopoietic stem cell isolation 747 748 Following euthanasia, bone marrow was harvested from the femur, tibia, pelvis, and 749 sternum through mechanical crushing, ensuring the retrieval of most HSCs. The 750 collected bone marrow cells were then sieved through a 100- μ m strainer and 751 cleansed with a cold 'Easy Sep' buffer containing PBS with 2% fetal bovine serum 752 (FBS), followed by lysis of red blood cells using RBC lysis buffer (Biolegend, Catalog 753 no. 420302). At first, mature lineage cells were selectively depleted through the 754 Lineage Cell Depletion Kit, mouse (Miltenyi Biotec, Catalog no. 130-110-470), while 755 the resulting Lin- (lineage-negative) fraction was then enriched for c-Kit expression 756 using CD117 MicroBeads (Miltenyi Biotec, Catalog no: 130-091-224). These cKit-757 enriched cells were washed, blocked with FcX and stained with following 758 fluorescently labeled antibodies: APC anti-mouse CD117, clone ACK2 (Biolegend 759 catalog no. 105812), PE/Cy7 anti-mouse Ly6a (Sca-1) (Biolegend, catalog no. 760 108114); Pacific Blue anti-mouse Lineage Cocktail (Biolegend, catalog no. 133310); 761 PE anti-mouse CD201 (EPCR) (Biolegend, catalog no. 141504); PE/Cy5 anti-mouse 762 CD150 (SLAM) (Biolegend, catalog no. 115912); APC/Cyanine7 anti-mouse CD48 763 (Biolegend, catalog no. 103432). 764 765 HSC ex-vivo expansion cultures 766 767 Ex-vivo cultures of HSCs were done under self-renewing F12-PVA-based conditions 768 as described previously (Wilkinson et al. 2019). To this end, cell-culture activated 96-769 well flat-bottom plates were coated with a layer of 100 ng/ml fibronectin (Bovine 770 Fibronectin Protein, CF Catalog: 1030-FN) for 30 minutes at room temperature. 771 Following the sorting process, HSCs were transferred into 200 µl of complete HSC 772 media supplemented with 100ng/ml recombinant mouse TPO and 10ng/ml 773 recombinant mouse SCF (PeproTech Recombinant Murine TPO Catalogue Number: 774 315-14; PeproTech Recombinant Murine SCF, Catalogue Number: 250-03) and 775 grown at 37°C with 5% CO2. During lentiviral library transduction, the first media 776 change took place 24 hours post-transduction. All other protocol steps followed the 777 guidelines provided in (Wilkinson et al. 2020). 778 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 779 Construction of lentiviral pLARRY vectors 780 781 The construction of barcoded libraries was executed by a previously established 782 protocol (https://www.protocols.io/view/barcode-plasmid-library-cloning-4hggt3w). 783 First, the T-Sapphire, Scarlett, or EGFP coding sequences, and the EF1a promoter 784 sequences were PCR amplified from pEB1-T-Sapphire, pmScarlet_NES_C1, and 785 pLARRY-EGFP with primers homologous to the vector insertion site in a custom 786 synthetic lentiviral plasmid backbone (Vectorbuilder, Inc) using Gibson assembly 787 (Gibson Assembly® Master Mix, NEB, Ref. E2611L). For recombinase lentivirus 788 libraries, iCre or Flpo recombinase was PCR amplified together with EGFP and 789 Scarlett with primers homologous to the vector insertion site in a custom synthetic 790 lentiviral plasmid backbone and cloned using Gibson assembly. After magnetic-bead 791 purification, ligated vectors were then transformed into NEB10-beta electroporation 792 ultracompetent E.coli cells (NEB® 10-beta Electrocompetent E. coli, NEB, 793 Ref.C3020K) and grown overnight on LB plates supplemented with 50 μ g/mL 794 Carbenicillin (Carbenicillin disodium salt, Thermo Scientific Chemicals Ref. 795 11568616). Colonies were scrapped using LB medium and pelleted by 796 centrifugation. Plasmid maxipreps were performed using the Endotoxin-Free Plasmid 797 Maxi Kit (Macheray Nagel), following the manufacturer's protocol. pEB1-T-Sapphire 798 was a gift from Philippe Cluzel (Addgene plasmid 103977). pLARRY-EGFP was a 799 gift from Fernando Camargo (Addgene plasmid 140025). pmScarlet_NES_C1 was a 800 gift from Dorus Gadella. Additional reagent details are in Supplementary Table 5. 801 802 Barcode lentivirus library generation and diversity estimation 803 804 To barcode pLARRYv2 plasmids and generate a library, first a spacer sequence 805 flanked by EcoRV restriction sites was cloned into the plasmid after the WPRE 806 element of the vector. Custom PAGE-purified single-strand oligonucleotides with a 807 pattern of 20 random bases and surrounded by 25 nucleotides homologous to the 808 vector insertion site were synthesized by IDT DNA Technologies (Supplementary 809 Table 5). The assembly of these components and subsequent purification steps 810 were carried out through Gibson assembly (Gibson Assembly® Master Mix, NEB, 811 Ref. E2611L). Six electroporations of the bead-purified ligations were performed into 812 NEB10-beta E.coli cells (NEB® 10-beta Electrocompetent E. coli, New England 813 BiolabsEB, Ref.C3020K) utilizing a Gene Pulser electroporator (Biorad). 814 Subsequently, after transformation, the cells were incubated at 37 degrees for 1 hour 815 at 220 rpm. Post-incubation, the transformed cells were plated in six large LB-816 ampicillin agar plates overnight at 30ºC. Colonies from all six plates were collected 817 by scraping with LB-ampicillin and then grown for an additional 2h at 225 rpm and 818 30/i2 °C. Cultures were pelleted by centrifugation, and plasmids were isolated using 819 the Endotoxin-Free Plasmid Maxi Kit (Macheray-Nagel), following the manufacturer's 820 protocol. For estimating diversity, barcode amplicon libraries were prepared by PCR 821 amplification of the lentiviral library maxiprep using flanking oligonucleotides carrying 822 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint TruSeq read1 and read2 adaptors using 10 ng of the library (Supplementary Table 823 5). We used the minimal number of cycles that we could detect by qPCR to avoid 824 PCR amplification bias (10-12 cycles). After bead purification, 10 ng of the first PCR 825 product was used as a template for a second PCR to add Illumina P5 and P7 826 adaptors and indexes (Supplementary Table 5). Two independent PCRs were 827 sequenced on an Illumina NovaSeq 6000 S4 platform (Novogene UK) to confirm 828 diversity after correction of errors through collapsing with a Hamming distance of 4. 829 After collapsing, libraries were confirmed to contain at least 50 million different 830 barcodes, with enough diversity for uniquely labeling up to 100,000 HSCs with a 831 minimal false-positive rate. Lentivirus production and HSPC transduction were 832 performed as described in (Weinreb et al. 2020). 833 834 Single-cell encapsulation and library preparation for sequencing 835 836 For scRNA sequencing and subsequent plating, cells were pipetted up and down 837 gently a few times to be dissociated into single cells and transferred to a 1.5 mL 838 microtube. The well was then washed with prewarmed PBS to collect all the possible 839 remaining cells. Cells were then concentrated by centrifugation at 800 g for 8 840 minutes. Washed cells were then blocked with FcX, and stained with the E-SLAM 841 stem cell antibodies panel, to confirm expansion of E-SLAM cells. Live cells were 842 then sorted based on fluorescent reporter expression. Part of the sample as 843 specified in text was then taken for constructing a single-cell library using Chromium 844 Single Cell 3’ Reagent Kits (v3) following the manufacturer’s guidelines (10X 845 Genomics). The remaining part was then plated back for further expansion in culture. 846 To minimize the impact of batch effects on sequencing, we multiplexed different 847 conditions leveraging the unique barcode pattern of our libraries together with 848 Biolegend TotalSeq™ anti-mouse hashing antibodies (Supplementary Table 5), 849 enabling the simultaneous preparation of libraries representing all experimental 850 conditions in a single reaction for each day of sampling. 851 Following the reverse transcription of mRNA and first-strand cDNA amplification, 100 852 ng of the cDNA libraries were used as templates to amplify LARRY barcodes by 853 nested PCR similar to the protocol described in (Weinreb et al. 2020). The first PCR 854 used forward primer (Pre-Enrichment forward) CTG AGC AAA GAC CCC AAC GAG 855 AA together with the corresponding 10x Genomics dual index TruSeq reverse primer 856 using the following programs 1, 98 C, 3 min; 2, 98 C, 20 s; 3, 58 C, 15 s; 4, 72 C, 20 857 s; 5, repeat steps 2–4 08 times; 6, 72 C, 3 min; 7, 4 C, hold. The PCR products were 858 then purified with a 0.8:1 ratio of Ampure XP beads. Purified PCR products were 859 then subjected to a second PCR using the forward primer (Trueseq_LARRY) GTG 860 ACT GGA GTT CAG ACG TGT GCT CTT CCG ATC TGC TAG GAG AGA CCA TAT 861 GGG ATC and the corresponding 10x dual index Truseq reserve primer, following 862 program 1, 98 C, 3 min; 2, 98 C, 20 s; 3, 58 C, 15 s; 4, 72 C, 20 s; 5, repeat steps 2–863 4 08 times; 6, 72 C, 3 min; 7, 4 C, hold. The final PCR products were then purified by 864 a 0.8:1 ratio of Ampure XP bead: PCR products, were indexed using the 10x dual 865 index TruSeq kit, and sequenced using Illumina NovaSeq or NextSeq. 866 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 867 scRNA-seq data processing and calling of lineage barcodes 868 869 Generation of single-cell matrices for gene expression and LARRY lineage barcodes 870 was performed using cloneRanger, an in-house developed pipeline 871 (https://github.com/dfernandezperez/cloneRanger) to process 10XGenomics single-872 cell RNA-seq together with LARRY barcoding. The pipeline is based on 873 Snakemake(Köster and Rahmann 2012) and the use of Docker/singularity 874 containers to allow for reproducibility and easy deployment of the code. 875 876 For each sample, fastq files from gene expression (GEX), LARRY and TotalSeq™ 877 tags were processed using cellranger v7.0.0 with default parameters. However, 878 since cellranger only collapses barcodes that are 1 hamming distance apart, prior to 879 the execution of cellranger, fastq files containing LARRY barcodes were processed 880 using UMICollapse(Liu 2019). This allowed us to collapse all barcodes which were 4 881 hamming distance units apart or less, similar to the procedure used by (Weinreb et 882 al. 2020; Rodriguez-Fraticelli et al. 2020). In particular, the UMICollapse was 883 executed with the following parameters: “fastq -k 4 --tag” . Finally, in order to run 884 cellranger in feature barcode mode with LARRY and TotalSeq™ sequences, we 885 created a reference library csv file by extracting all detected LARRY barcodes across 886 all collapsed fastq files, together with TotalSeq™ sequences. A reference library file 887 was created for each individual sample and given as input to cellranger, executed 888 with default parameters. All the code and steps performed by the pipeline are 889 available in the cloneRanger github page. 890 891 A Seurat(Hao et al. 2024) object containing single-cell count matrices from GEX, 892 LARRY and TotalSeq™ counts was created with the function Read10X from Seurat. 893 Finally, cell doublets were removed with scDblFinder(Germain et al. 2021) using 894 default parameters and TotalSeq™ sequences were demultiplexed with the function 895 hashedDrops from the DropletUtils R package(Lun et al. 2019) with default 896 parameters. 897 898 The assignment of LARRY barcodes to individual cells was performed by 899 cloneRanger similarly to (Weinreb et al. 2020; Rodriguez-Fraticelli et al. 2020): first, 900 we generated a filtered LARRY matrix by removing barcode UMIs that were 901 sustained by less than 5 sequencing reads (this information is stored in the 902 molecule_info.h5 file generated by cellranger). Then, we further filtered the LARRY 903 matrix by removing all barcodes with less than 4 UMIs. After filtering, barcodes were 904 assigned to individual cells as following: (a) for cells in which only one barcode was 905 detected after filtering, that barcode was assigned, (b) for cells in which more than 906 one barcode was detected post-filtering, the top barcode with higher UMI counts was 907 assigned and (c) for cells in which there were ties in the top barcode, no barcode 908 was assigned. Our barcode calling strategy was developed to minimize mixing cells 909 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint from different clones at the expenses of having higher chances to split real clones 910 into subclones. 911 912 Single-cell integration, clustering and annotation 913 914 To integrate scRNA-seq samples, we applied the IntegrateLayers Seurat v5 915 workflow to test multiple integration algorithms (sample time point was used as batch 916 variable): Harmony, Reciprocal PCA, Canonical Correlation Analysis and Joint PCA. 917 After supervising the results from all algorithms, we decided to use Reciprocal PCA 918 to integrate the single-cell GEX matrices. We followed a standard Seurat pipeline 919 with some minor modifications. Raw counts were normalized with the function 920 NormalizeData and the top 3000 variable genes were selected. From those, we 921 removed all ribosomal and mitochondrial genes, as well as genes that correlated 922 with cell cycle genes ( Ube2c, Hmgb2, Hmgn2, Tuba1b, Ccnb1, Tubb5, Top2a, 923 Tubb4b, pearson cor of 0.1 or more), as performed by (Weinreb et al. 2020). Then, 924 filtered variable genes were used to compute the top 50 reciprocal-PCA 925 components. The kNN graph was computed using the function FindNeighbors 926 setting the number of neighbors to 20, which was extended to 30 for the generation 927 of UMAP components. Clusters were generated with the function FindClusters with a 928 resolution of 0.3 and the Louvain algorithm. 929 930 Annotation of cell types was performed using known gene markers from the 931 literature. A summary of the main markers for every cluster are shown in 932 Supplementary Figure 1B and the whole list of markers for every cluster, computed 933 with FindAllMarkers from Seurat, are listed in Supplementary Table 2. 934 935 Quantification and classification of HSC clonal behaviors 936 937 Clone x state heatmaps from Figures 1, 2, and 4 were generated as follows: for 938 every clone, we computed the number of cells detected across every cluster (cell 939 type), generating a clone-by-cluster matrix A. In A, each Aij represents the number of 940 cells from the cluster j detected in the clone i . After generating A, in order to account 941 for cell type abundance heterogeneity, we column-normalized the matrix by the total 942 number of cells from each cell type, generating a B matrix. Finally, to compare clonal 943 fates between clones of different sizes, we row-normalized the B matrix to obtain, for 944 every clone, the fraction of cells present in each cell type (intra-clone fraction score). 945 HSC clonal behaviors were quantified as described in (Rodriguez-Fraticelli et al. 946 2020). [explain briefly]. To determine statistically significant biased clones (clones 947 representing a higher proportion of a specific cluster than expected from random cell 948 sampling) we applied a Fisher’s exact test as described in (Biddy et al. 2018) 949 accounting for clone size. 950 951 952 Sister cell similarity analysis 953 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 954 To calculate the sister cell similarity scores shown in Supplementary Figure 1D, we 955 subsetted individually the cells corresponding to each timepoint (Day 7 and D27) and 956 proceeded as follows: The top 2000 variable genes were selected and, as described 957 above, ribosomal, mitochondrial and genes correlating with cell cycle genes were 958 filtered out. These filtered variable genes were scaled and used as input to calculate 959 the top 70 principal components (PCs). The cell-by-PC matrix (obtained with the 960 function Embeddings from Seurat) was used as input for the R cor function selecting 961 pearson as the correlation metric. This procedure generated a cell-by-cell similarity 962 matrix that was split into: (a) all pairs of sister cells, (b) all pairs of non-sister cells, 963 and (c) all pairs of sister cells in which the barcode label was previously shuffled. To 964 assess the statistical significance between the average Perason coefficient of these 965 3 groups, a permutation test with 1000 simulations was performed. Briefly, the 966 average correlation of each group was compared to a random distribution of sister 967 similarity scores generated by shuffling the larry barcodes prior to the generation of 968 the cell-by-cell similarity matrix across 1000 iterations. 969 970 Stochastic sampling model of clonal selection 971 972 To calculate the expected clonality across our experimental time course, assuming 973 that all clones have equal fitness, we developed a null clonal selection model based 974 on sampling with a binomial distribution. This model recapitulates the different 975 sampling events (sample splitting, well splitting, sampling of cells for scRNA-seq, 976 fraction of cells encapsulated in library preparation) and measured cell expansion 977 (from D7 to D14 and D14 to D27). It makes two key assumptions for simplicity: no 978 cell death (based on our limited observation of apoptotic-like events during culture) 979 and similar proliferation probabilities for all cell types. We applied sequential 980 sampling calls with replacement (using the R function sample(x, replace = TRUE)) to 981 model, from the initial distribution of clones sizes detected at D7, the following steps 982 -using empirical data for each individual well replicate-: 1) fraction of cells lost in 983 encapsulation for single-cell RNAseq at D7, 2) fraction of cells split into WT and 984 mutant samples at D7, 3) transduction efficiency of secondary LARRY barcoding, 3) 985 fraction of cells split into different wells, 4) cell expansion from D7 to D14, 5) fraction 986 of cells used for D14 sequencing, 6) cell expansion from D14 to D27, 7) fraction of 987 cells sampled for D27 sequencing and 8) fraction of cell recovery (encapsulation) 988 from D27 library preparation. The output of the model is a list of the expected clones 989 detected at D27 and their corresponding sizes. We ran 1000 iterations of the model, 990 from which we calculated an average expected clone size. The distribution of clones 991 sizes from the model was also used to calculate the expected clone size correlation 992 between wells at D27. 993 994 Single-cell differential gene expression and signature scores 995 996 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint All gene differential expression analyses, unless otherwise specified, were 997 performed using MAST (Finak et al. 2015) within the function findMarkers from 998 Seurat using the replicate information as the only latent variable. This was done in 999 order to mask differences in gene expression between male and female cells, which 1000 corresponded to replicate 1 and replicate 2, respectively. 1001 1002 Gene set enrichment analysis 1003 1004 For each comparison, we created pre-ranked lists based on log2 fold-change 1005 differences in gene expression obtained from scRNAseq or bulkRNAseq analysis. 1006 These pre-ranked lists were analyzed using gseapy version 1.1.1 (Fang, Liu, and 1007 Peltz 2023). For gene sets, we used signatures shown on Supplementary Table 4 1008 (HSC signatures) or MsigDB gene sets (mouse hallmarks or gene-ontology terms). 1009 Data from AML patients were obtained from previous studies (Mer et al. 2021; 1010 Naldini et al. 2023). For the scRNAseq dataset, patients were categorized as 1011 “mature” or “primitive” based on the ratio of % cells expressing CD14 versus % cells 1012 expressing CD96. Patients with CD14/CD96 ratio >1 were labeled as “mature”, while 1013 CD14/CD96 ratio <1 were labeled as “primitive”. For the bulkRNAseq dataset, we 1014 used the differential expression list output in the available code (comparing the 1015 “mature” patient cluster with the “primitive” patient cluster). 1016 1017 Statistical methods 1018 1019 Statistical analysis was performed using the tests as indicated throughout the text. 1020 Generally, Wilcoxon rank-sum tests were used for statistical significance except 1021 where indicated. 1022 1023 1024 1025 SUPPLEMENTARY INFORMATION 1026 1027 Supplementary Table 1. Summary statistics of clones. This table contains the 1028 summary statistics for each library and replicate in the study, including number of 1029 clones, average size, number of cells and percentage of the culture sampled. 1030 Supplementary Table 2. Cluster markers and groupings. This table contains the 1031 markers used to annotate the clusters and group them into annotations. 1032 Supplementary Table 3. Differential expressed gene analyses. This table 1033 contains all the differential gene expression analysis results. Different comparisons 1034 are under different tabs. 1035 Supplementary Table 4. Clonal HSC signatures used. This table contains the list 1036 of genes used for gene set enrichment analysis. 1037 Supplementary Table 5. Reagents tables. This table contains the lists of reagents 1038 used, including genotyping primers and LARRY barcode sequences. 1039 1040 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 1041 1042 1043

1044 Avagyan, Serine, and Leonard I. Zon. 2023. “Clonal Hematopoiesis and 1045 Inflammation - the Perpetual Cycle.” Trends in Cell Biology 33 (8): 695–707. 1046 Avagyan, S., J. E. Henninger, W. P. Mannherz, M. Mistry, J. Yoon, S. Yang, M. C. 1047 Weber, J. L. Moore, and L. I. Zon. 2021. “Resistance to Inflammation Underlies 1048 Enhanced Fitness in Clonal Hematopoiesis.” Science 374 (6568): 768–72. 1049 Baggiolini, Arianna, Scott J. Callahan, Emily Montal, Joshua M. Weiss, Tuan Trieu, 1050 Mohita M. Tagore, Sam E. Tischfield, et al. 2021. “Developmental Chromatin 1051 Programs Determine Oncogenic Competence in Melanoma.” Science 373 1052 (6559): eabc1048. 1053 Beaudin, Anna E., Scott W. Boyer, Jessica Perez-Cunningham, Gloria E. 1054 Hernandez, S. Christopher Derderian, Chethan Jujjavarapu, Eric Aaserude, 1055 Tippi MacKenzie, and E. Camilla Forsberg. 2016. “A Transient Developmental 1056 Hematopoietic Stem Cell Gives Rise to Innate-like B and T Cells.” Cell Stem 1057 Cell 19 (6): 768–83. 1058 Bick, Alexander G., Joshua S. Weinstock, Satish K. Nandakumar, Charles P. Fulco, 1059 Erik L. Bao, Seyedeh M. Zekavat, Mindy D. Szeto, et al. 2020. “Inherited 1060 Causes of Clonal Haematopoiesis in 97,691 Whole Genomes.” Nature 586 1061 (7831): 763–68. 1062 Biddy, Brent A., Wenjun Kong, Kenji Kamimoto, Chuner Guo, Sarah E. Waye, Tao 1063 Sun, and Samantha A. Morris. 2018. “Single-Cell Mapping of Lineage and 1064 Identity in Direct Reprogramming.” Nature 564 (7735): 219–24. 1065 Blanpain, Cédric. 2013. “Tracing the Cellular Origin of Cancer.” Nature Cell Biology 1066 15 (2): 126–34. 1067 Brunetti, Lorenzo, Michael C. Gundry, Daniele Sorcini, Anna G. Guzman, Yung-Hsin 1068 Huang, Raghav Ramabadran, Ilaria Gionfriddo, et al. 2018. “Mutant NPM1 1069 Maintains the Leukemic State through HOX Expression.” Cancer Cell 34 (3): 1070 499–512.e9. 1071 Cai, Sheng F., S. Haihua Chu, Aaron D. Goldberg, Salma Parvin, Richard P. Koche, 1072 Jacob L. Glass, Eytan M. Stein, et al. 2020. “Leukemia Cell of Origin Influences 1073 Apoptotic Priming and Sensitivity to LSD1 Inhibition.” Cancer Discovery 10 (10): 1074 1500–1513. 1075 Carrelha, Joana, Yiran Meng, Laura M. Kettyle, Tiago C. Luis, Ruggiero Norfo, 1076 Verónica Alcolea, Hanane Boukarabila, et al. 2018. “Hierarchically Related 1077 Lineage-Restricted Fates of Multipotent Haematopoietic Stem Cells.” Nature 1078 554 (7690): 106–11. 1079 Che, James L. C., Daniel Bode, Iwo Kucinski, Alyssa H. Cull, Fiona Bain, Hans J. 1080 Becker, Maria Jassinskaja, et al. 2022. “Identification and Characterization of in 1081 Vitro Expanded Hematopoietic Stem Cells.” EMBO Reports 23 (10): e55502. 1082 Chen, Wanze, Orane Guillaume-Gentil, Pernille Yde Rainer, Christoph G. Gäbelein, 1083 Wouter Saelens, Vincent Gardeux, Amanda Klaeger, et al. 2022. “Live-Seq 1084 Enables Temporal Transcriptomic Recording of Single Cells.” Nature 608 1085 (7924): 733–40. 1086 Cozzio, Antonio, Emmanuelle Passegué, Paul M. Ayton, Holger Karsunky, Michael 1087 L. Cleary, and Irving L. Weissman. 2003. “Similar MLL-Associated Leukemias 1088 Arising from Self-Renewing Stem Cells and Short-Lived Myeloid Progenitors.” 1089 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint Genes & Development 17 (24): 3029–35. 1090 Dykstra, Brad, David Kent, Michelle Bowie, Lindsay McCaffrey, Melisa Hamilton, 1091 Kristin Lyons, Shang-Jung Lee, Ryan Brinkman, and Connie Eaves. 2007. 1092 “Long-Term Propagation of Distinct Hematopoietic Differentiation Programs in 1093 Vivo.” Cell Stem Cell 1 (2): 218–29. 1094 Fang, Zhuoqing, Xinyuan Liu, and Gary Peltz. 2023. “GSEApy: A Comprehensive 1095 Package for Performing Gene Set Enrichment Analysis in Python.” 1096 Bioinformatics 39 (1). https://doi.org/10.1093/bioinformatics/btac757. 1097 Fennell, Katie A., Dane Vassiliadis, Enid Y. N. Lam, Luciano G. Martelotto, Jesse J. 1098 Balic, Sebastian Hollizeck, Tom S. Weber, et al. 2021. “Non-Genetic 1099 Determinants of Malignant Clonal Fitness at Single-Cell Resolution.” Nature, 1100 December, 1–7. 1101 Finak, Greg, Andrew McDavid, Masanao Yajima, Jingyuan Deng, Vivian Gersuk, 1102 Alex K. Shalek, Chloe K. Slichter, et al. 2015. “MAST: A Flexible Statistical 1103 Framework for Assessing Transcriptional Changes and Characterizing 1104 Heterogeneity in Single-Cell RNA Sequencing Data.” Genome Biology 16 1105 (December): 278. 1106 Frenz-Wiessner, Stephanie, Savannah D. Fairley, Maximilian Buser, Isabel Goek, 1107 Kirill Salewskij, Gustav Jonsson, David Illig, et al. 2024. “Generation of Complex 1108 Bone Marrow Organoids from Human Induced Pluripotent Stem Cells.” Nature 1109 Methods, February. https://doi.org/10.1038/s41592-024-02172-2. 1110 George, Joshy, Asli Uyar, Kira Young, Lauren Kuffler, Kaiden Waldron-Francis, 1111 Eladio Marquez, Duygu Ucar, and Jennifer J. Trowbridge. 2016. “Leukaemia 1112 Cell of Origin Identified by Chromatin Landscape of Bulk Tumour Cells.” Nature 1113 Communications 7 (July): 12166. 1114 Germain, Pierre-Luc, Aaron Lun, Carlos Garcia Meixide, Will Macnair, and Mark D. 1115 Robinson. 2021. “Doublet Identification in Single-Cell Sequencing Data Using 1116 scDblFinder.” F1000Research 10 (September): 979. 1117 Geurts, Maarten H., Eyleen de Poel, Cayetano Pleguezuelos-Manzano, Rurika Oka, 1118 Léo Carrillo, Amanda Andersson-Rolf, Matteo Boretto, et al. 2021. “Evaluating 1119 CRISPR-Based Prime Editing for Cancer Modeling and CFTR Repair in 1120 Organoids.” Life Science Alliance 4 (10). 1121 https://doi.org/10.26508/lsa.202000940. 1122 Giladi, Amir, Franziska Paul, Yoni Herzog, Yaniv Lubling, Assaf Weiner, Ido Yofe, 1123 Diego Jaitin, et al. 2018. “Single-Cell Characterization of Haematopoietic 1124 Progenitors and Their Trajectories in Homeostasis and Perturbed 1125 Haematopoiesis.” Nature Cell Biology 20 (7): 836–46. 1126 Goyal, Yogesh, Gianna T. Busch, Maalavika Pillai, Jingxin Li, Ryan H. Boe, 1127 Emanuelle I. Grody, Manoj Chelvanambi, et al. 2023. “Diverse Clonal Fates 1128 Emerge upon Drug Treatment of Homogeneous Cancer Cells.” Nature 620 1129 (7974): 651–59. 1130 Guryanova, Olga A., Kaitlyn Shank, Barbara Spitzer, Luisa Luciani, Richard P. 1131 Koche, Francine E. Garrett-Bakelman, Chezi Ganzel, et al. 2016. “DNMT3A 1132 Mutations Promote Anthracycline Resistance in Acute Myeloid Leukemia via 1133 Impaired Nucleosome Remodeling.” Nature Medicine 22 (12): 1488–95. 1134 Haas, Simon, Jenny Hansson, Daniel Klimmeck, Dirk Loeffler, Lars Velten, Hannah 1135 Uckelmann, Stephan Wurzer, et al. 2015. “Inflammation-Induced Emergency 1136 Megakaryopoiesis Driven by Hematopoietic Stem Cell-like Megakaryocyte 1137 Progenitors.” Cell Stem Cell 17 (4): 422–34. 1138 Haas, Simon, Andreas Trumpp, and Michael D. Milsom. 2018. “Causes and 1139 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint Consequences of Hematopoietic Stem Cell Heterogeneity.” Cell Stem Cell 22 1140 (5): 627–38. 1141 Hao, Yuhan, Tim Stuart, Madeline H. Kowalski, Saket Choudhary, Paul Hoffman, 1142 Austin Hartman, Avi Srivastava, et al. 2024. “Dictionary Learning for Integrative, 1143 Multimodal and Scalable Single-Cell Analysis.” Nature Biotechnology 42 (2): 1144 293–304. 1145 Hormaechea-Agulla, Daniel, Katie A. Matatall, Duy T. Le, Bailee Kain, Xiaochen 1146 Long, Pawel Kus, Roman Jaksik, Grant A. Challen, Marek Kimmel, and 1147 Katherine Y. King. 2021. “Chronic Infection Drives Dnmt3a-Loss-of-Function 1148 Clonal Hematopoiesis via IFNγ Signaling.” Cell Stem Cell 28 (8): 1428–42.e6. 1149 Huntly, Brian J. P., Hirokazu Shigematsu, Kenji Deguchi, Benjamin H. Lee, Shinichi 1150 Mizuno, Nicky Duclos, Rebecca Rowan, et al. 2004. “MOZ-TIF2, but Not BCR-1151 ABL, Confers Properties of Leukemic Stem Cells to Committed Murine 1152 Hematopoietic Progenitors.” Cancer Cell 6 (6): 587–96. 1153 Jakobsen, Niels Asger, Sven Turkalj, Andy G. X. Zeng, Bilyana Stoilova, Marlen 1154 Metzner, Murtaza S. Nagree, Sayyam Shah, et al. 2023. “Selective Advantage 1155 of Mutant Stem Cells in Clonal Hematopoiesis Occurs by Attenuating the 1156 Deleterious Effects of Inflammation and Aging.” bioRxiv. 1157 https://doi.org/10.1101/2023.09.12.557322. 1158 Jindal, Kunal, Mohd Tayyab Adil, Naoto Yamaguchi, Xue Yang, Helen C. Wang, 1159 Kenji Kamimoto, Guillermo C. Rivera-Gonzalez, and Samantha A. Morris. 2023. 1160 “Single-Cell Lineage Capture across Genomic Modalities with CellTag-Multi 1161 Reveals Fate-Specific Gene Regulatory Changes.” Nature Biotechnology, 1162 September. https://doi.org/10.1038/s41587-023-01931-4. 1163 Khan, Abdullah O., Antonio Rodriguez-Romera, Jasmeet S. Reyat, Aude-Anais 1164 Olijnik, Michela Colombo, Guanlin Wang, Wei Xiong Wen, et al. 2023. “Human 1165 Bone Marrow Organoids for Disease Modeling, Discovery, and Validation of 1166 Therapeutic Targets in Hematologic Malignancies.” Cancer Discovery 13 (2): 1167 364–85. 1168 Köster, Johannes, and Sven Rahmann. 2012. “Snakemake--a Scalable 1169 Bioinformatics Workflow Engine.” Bioinformatics 28 (19): 2520–22. 1170 Krivtsov, Andrei V., David Twomey, Zhaohui Feng, Matthew C. Stubbs, Yingzi Wang, 1171 Joerg Faber, Jason E. Levine, et al. 2006. “Transformation from Committed 1172 Progenitor to Leukaemia Stem Cell Initiated by MLL-AF9.” Nature 442 (7104): 1173 818–22. 1174 Krivtsov, A. V., M. E. Figueroa, A. U. Sinha, M. C. Stubbs, Z. Feng, P. J. M. Valk, R. 1175 Delwel, et al. 2013. “Cell of Origin Determines Clinically Relevant Subtypes of 1176 MLL-Rearranged AML.” Leukemia 27 (4): 852–60. 1177 Lenz, Guido, Giovana R. Onzi, Luana S. Lenz, Julieti H. Buss, Jephesson A. Dos 1178 Santos, and Karine R. Begnini. 2022. “The Origins of Phenotypic Heterogeneity 1179 in Cancer.” Cancer Research 82 (1): 3–11. 1180 Li, Hubo, Brenton G. Mar, Huadi Zhang, Rishi V. Puram, Francisca Vazquez, 1181 Barbara A. Weir, William C. Hahn, Benjamin Ebert, and David Pellman. 2017. 1182 “The EMT Regulator ZEB2 Is a Novel Dependency of Human and Murine Acute 1183 Myeloid Leukemia.” Blood 129 (4): 497–508. 1184 Li, Li, Sarah Bowling, Sean E. McGeary, Qi Yu, Bianca Lemke, Karel Alcedo, 1185 Yuemeng Jia, et al. 2023. “A Mouse Model with High Clonal Barcode Diversity 1186 for Joint Lineage, Transcriptomic, and Epigenomic Profiling in Single Cells.” Cell 1187 186 (23): 5183–99.e22. 1188 Liu, Daniel. 2019. “Algorithms for Efficiently Collapsing Reads with Unique Molecular 1189 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint Identifiers.” PeerJ 7 (December): e8275. 1190 Loberg, Matthew A., Rebecca K. Bell, Leslie O. Goodwin, Elizabeth Eudy, Linde A. 1191 Miles, Jennifer M. SanMiguel, Kira Young, et al. 2019. “Sequentially Inducible 1192 Mouse Models Reveal That Npm1 Mutation Causes Malignant Transformation 1193 of Dnmt3a-Mutant Clonal Hematopoiesis.” Leukemia 33 (7): 1635–49. 1194 Lun, Aaron T. L., Samantha Riesenfeld, Tallulah Andrews, The Phuong Dao, Tomas 1195 Gomes, participants in the 1st Human Cell Atlas Jamboree, and John C. 1196 Marioni. 2019. “EmptyDrops: Distinguishing Cells from Empty Droplets in 1197 Droplet-Based Single-Cell RNA Sequencing Data.” Genome Biology 20 (1): 63. 1198 Meng, Yiran, Joana Carrelha, Roy Drissen, Xiying Ren, Bowen Zhang, Adriana 1199 Gambardella, Simona Valletta, Supat Thongjuea, Sten Eirik Jacobsen, and 1200 Claus Nerlov. 2023. “Epigenetic Programming Defines Haematopoietic Stem 1201 Cell Fate Restriction.” Nature Cell Biology 25 (6): 812–22. 1202 Mer, Arvind Singh, Emily M. Heath, Seyed Ali Madani Tonekaboni, Nergiz Dogan-1203 Artun, Sisira Kadambat Nair, Alex Murison, Laura Garcia-Prat, et al. 2021. 1204 “Biological and Therapeutic Implications of a Unique Subtype of NPM1 Mutated 1205 AML.” Nature Communications 12 (1): 1054. 1206 Morcos, Mina N. F., Congxin Li, Clara M. Munz, Alessandro Greco, Nicole Dressel, 1207 Susanne Reinhardt, Katrin Sameith, et al. 2022. “Fate Mapping of 1208 Hematopoietic Stem Cells Reveals Two Pathways of Native Thrombopoiesis.” 1209 Nature Communications 13 (1): 4504. 1210 Naik, Shalin H., Leïla Perié, Erwin Swart, Carmen Gerlach, Nienke van Rooij, Rob J. 1211 de Boer, and Ton N. Schumacher. 2013. “Diverse and Heritable Lineage 1212 Imprinting of Early Haematopoietic Progenitors.” Nature 496 (7444): 229–32. 1213 Naldini, Matteo Maria, Gabriele Casirati, Matteo Barcella, Paola Maria Vittoria 1214 Rancoita, Andrea Cosentino, Carolina Caserta, Francesca Pavesi, et al. 2023. 1215 “Longitudinal Single-Cell Profiling of Chemotherapy Response in Acute Myeloid 1216 Leukemia.” Nature Communications 14 (1): 1285. 1217 Ono, Ryoichi, Masahiro Masuya, Hideaki Nakajima, Yutaka Enomoto, Eri Miyata, 1218 Akihide Nakamura, Satomi Ishii, et al. 2013. “Plzf Drives MLL-Fusion-Mediated 1219 Leukemogenesis Specifically in Long-Term Hematopoietic Stem Cells.” Blood 1220 122 (7): 1271–83. 1221 Paul, Franziska, Ya ’ara Arkin, Amir Giladi, Diego Adhemar Jaitin, Ephraim 1222 Kenigsberg, Hadas Keren-Shaul, Deborah Winter, et al. 2016. “Transcriptional 1223 Heterogeneity and Lineage Commitment in Myeloid Progenitors.” Cell 164 (1-2): 1224 325. 1225 Perié, Leïla, Ken R. Duffy, Lianne Kok, Rob J. de Boer, and Ton N. Schumacher. 1226 2015. “The Branching Point in Erythro-Myeloid Differentiation.” Cell 163 (7): 1227 1655–62. 1228 Rajbhandari, Nirakar, Michael Hamilton, Cynthia M. Quintero, L. Paige Ferguson, 1229 Raymond Fox, Christian M. Schürch, Jun Wang, et al. 2023. “Single-Cell 1230 Mapping Identifies MSI Cells as a Common Origin for Diverse Subtypes of 1231 Pancreatic Cancer.” Cancer Cell 41 (11): 1989–2005.e9. 1232 Rodriguez-Fraticelli, Alejo E., Caleb Weinreb, Shou-Wen Wang, Rosa P. Migueles, 1233 Maja Jankovic, Marc Usart, Allon M. Klein, Sally Lowell, and Fernando D. 1234 Camargo. 2020. “Single-Cell Lineage Tracing Unveils a Role for TCF15 in 1235 Haematopoiesis.” Nature 583 (7817): 585–89. 1236 Rodriguez-Fraticelli, Alejo E., Samuel L. Wolock, Caleb S. Weinreb, Riccardo 1237 Panero, Sachin H. Patel, Maja Jankovic, Jianlong Sun, Raffaele A. Calogero, 1238 Allon M. Klein, and Fernando D. Camargo. 2018. “Clonal Analysis of Lineage 1239 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint Fate in Native Haematopoiesis.” Nature 553 (7687): 212–16. 1240 SanMiguel, Jennifer M., Elizabeth Eudy, Matthew A. Loberg, Linde A. Miles, Tim 1241 Stearns, Jayna J. Mistry, Michael J. Rauh, Ross L. Levine, and Jennifer J. 1242 Trowbridge. 2022. “Cell Origin-Dependent Cooperativity of Mutant Dnmt3a and 1243 Npm1 in Clonal Hematopoiesis and Myeloid Malignancy.” Blood Advances 6 1244 (12): 3666–77. 1245 SanMiguel, Jennifer M., Elizabeth Eudy, Matthew A. Loberg, Kira A. Young, Jayna J. 1246 Mistry, Kristina D. Mujica, Logan S. Schwartz, Timothy M. Stearns, Grant A. 1247 Challen, and Jennifer J. Trowbridge. 2022. “Distinct Tumor Necrosis Factor 1248 Alpha Receptors Dictate Stem Cell Fitness versus Lineage Output in Dnmt3a-1249 Mutant Clonal Hematopoiesis.” Cancer Discovery 12 (12): 2763–73. 1250 Schwartz, Logan S., Kira A. Young, Timothy M. Stearns, Nathan Boyer, Kristina D. 1251 Mujica, and Jennifer J. Trowbridge. 2024. “Transcriptional and Functional 1252 Consequences of Oncostatin M Signaling on Young Dnmt3a-Mutant 1253 Hematopoietic Stem Cells.” Experimental Hematology 130 (February): 104131. 1254 Sommerkamp, Pia, François E. Mercier, Adam C. Wilkinson, Dominique Bonnet, and 1255 Paul E. Bourgine. 2021. “Engineering Human Hematopoietic Environments 1256 through Ossicle and Bioreactor Technologies Exploitation.” Experimental 1257 Hematology 94 (February): 20–25. 1258 Stavropoulou, Vaia, Susanne Kaspar, Laurent Brault, Mathijs A. Sanders, Sabine 1259 Juge, Stefano Morettini, Alexandar Tzankov, et al. 2016. “MLL-AF9 Expression 1260 in Hematopoietic Stem Cells Drives a Highly Invasive AML Expressing EMT-1261 Related Genes Linked to Poor Outcome.” Cancer Cell 30 (1): 43–58. 1262 Stonehouse, Olivia J., Christine Biben, Tom S. Weber, Alexandra Garnham, Katie A. 1263 Fennell, Alison Farley, Antoine F. Terreaux, et al. 2024. “Clonal Analysis of Fetal 1264 Hematopoietic Stem/progenitor Cell Subsets Reveals How Post-Transplantation 1265 Capabilities Are Distributed.” bioRxiv. 1266 https://doi.org/10.1101/2024.02.19.579920. 1267 Taussig, David C., Jacques Vargaftig, Farideh Miraki-Moud, Emmanuel Griessinger, 1268 Kirsty Sharrock, Tina Luke, Debra Lillington, et al. 2010. “Leukemia-Initiating 1269 Cells from Some Acute Myeloid Leukemia Patients with Mutated Nucleophosmin 1270 Reside in the CD34(-) Fraction.” Blood 115 (10): 1976–84. 1271 Tian, Luyi, Sara Tomei, Jaring Schreuder, Tom S. Weber, Daniela Amann-1272 Zalcenstein, Dawn S. Lin, Jessica Tran, et al. 2021. “Clonal Multi-Omics 1273 Reveals Bcor as a Negative Regulator of Emergency Dendritic Cell 1274 Development.” Immunity 54 (6): 1338–51.e9. 1275 Tong, Jingyuan, Ting Sun, Shihui Ma, Yanhong Zhao, Mankai Ju, Yuchen Gao, Ping 1276 Zhu, et al. 2021. “Hematopoietic Stem Cell Heterogeneity Is Linked to the 1277 Initiation and Therapeutic Response of Myeloproliferative Neoplasms.” Cell 1278 Stem Cell 28 (4): 780. 1279 Uckelmann, Hannah J., Stephanie M. Kim, Eric M. Wong, Charles Hatton, Hugh 1280 Giovinazzo, Jayant Y. Gadrey, Andrei V. Krivtsov, et al. 2020. “Therapeutic 1281 Targeting of Preleukemia Cells in a Mouse Model of NPM1 Mutant Acute 1282 Myeloid Leukemia.” Science 367 (6477): 586–90. 1283 Visvader, Jane E. 2011. “Cells of Origin in Cancer.” Nature 469 (7330): 314–22. 1284 Wagner, Daniel E., and Allon M. Klein. 2020. “Lineage Tracing Meets Single-Cell 1285 Omics: Opportunities and Challenges.” Nature Reviews. Genetics 21 (7): 410–1286 27. 1287 Wang, Shou-Wen, Michael J. Herriges, Kilian Hurley, Darrell N. Kotton, and Allon M. 1288 Klein. 2022. “CoSpar Identifies Early Cell Fate Biases from Single-Cell 1289 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint Transcriptomic and Lineage Information.” Nature Biotechnology 40 (7): 1066–1290 74. 1291 Weinreb, Caleb, Alejo Rodriguez-Fraticelli, Fernando D. Camargo, and Allon M. 1292 Klein. 2020. “Lineage Tracing on Transcriptional Landscapes Links State to Fate 1293 during Differentiation.” Science 367 (6479). 1294 https://doi.org/10.1126/science.aaw3381. 1295 Weinstock, Joshua S., Jayakrishnan Gopakumar, Bala Bharathi Burugula, Md 1296 Mesbah Uddin, Nikolaus Jahn, Julia A. Belk, Hind Bouzid, et al. 2023. “Aberrant 1297 Activation of TCL1A Promotes Stem Cell Expansion in Clonal Haematopoiesis.” 1298 Nature 616 (7958): 755–63. 1299 Wilkinson, Adam C., Reiko Ishida, Misako Kikuchi, Kazuhiro Sudo, Maiko Morita, 1300 Ralph Valentine Crisostomo, Ryo Yamamoto, et al. 2019. “Long-Term Ex Vivo 1301 Haematopoietic-Stem-Cell Expansion Allows Nonconditioned Transplantation.” 1302 Nature 571 (7763): 117–21. 1303 Wilkinson, Adam C., Reiko Ishida, Hiromitsu Nakauchi, and Satoshi Yamazaki. 2020. 1304 “Long-Term Ex Vivo Expansion of Mouse Hematopoietic Stem Cells.” Nature 1305 Protocols 15 (2): 628–48. 1306 Wilson, Nicola K., David G. Kent, Florian Buettner, Mona Shehata, Iain C. Macaulay, 1307 Fernando J. Calero-Nieto, Manuel Sánchez Castillo, et al. 2015. “Combined 1308 Single-Cell Functional and Gene Expression Analysis Resolves Heterogeneity 1309 within Stem Cell Populations.” Cell Stem Cell 16 (6): 712–24. 1310 Yamamoto, Ryo, Adam C. Wilkinson, Jun Ooehara, Xun Lan, Chen-Yi Lai, Yusuke 1311 Nakauchi, Jonathan K. Pritchard, and Hiromitsu Nakauchi. 2018. “Large-Scale 1312 Clonal Analysis Resolves Aging of the Mouse Hematopoietic Stem Cell 1313 Compartment.” Cell Stem Cell 22 (4): 600–607.e4. 1314 Zeisig, Bernd B., Tsz Kan Fung, Magdalena Zarowiecki, Chiou Tsun Tsai, Huacheng 1315 Luo, Boban Stanojevic, Claire Lynn, et al. 2021. “Functional Reconstruction of 1316 Human AML Reveals Stem Cell Origin and Vulnerability of Treatment-Resistant 1317 MLL-Rearranged Leukemia.” Science Translational Medicine 13 (582). 1318 https://doi.org/10.1126/scitranslmed.abc4822. 1319 Zhang, Qinyu, Rasmus Olofzon, Anna Konturek-Ciesla, Ouyang Yuan, and David 1320 Bryder. 2024. “Ex Vivo Expansion Potential of Murine Hematopoietic Stem Cells 1321 Is a Rare Property Only Partially Predicted by Phenotype.” eLife 12 (March). 1322 https://doi.org/10.7554/eLife.91826. 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint 1337 1338 1339 1340 1341 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 17, 2024. ; https://doi.org/10.1101/2024.05.14.594084doi: bioRxiv preprint