Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Nov;23(11):1199-1211.
doi: 10.1038/s41556-021-00766-y. Epub 2021 Oct 21.

Dynamic transcriptional reprogramming leads to immunotherapeutic vulnerabilities in myeloma

Julia Frede  1   2   3 , Praveen Anand  1   2   3 , Noori Sotudeh  1   2   3 , Ricardo A Pinto  3   4   5 , Monica S Nair  1 , Hannah Stuart  1 , Andrew J Yee  2   6 , Tushara Vijaykumar  1 , Johannes M Waldschmidt  1   2   3 , Sayalee Potdar  7 , Jake A Kloeber  1 , Antonis Kokkalis  1   2   3 , Valeriya Dimitrova  2   3   7 , Mason Mann  6 , Jacob P Laubach  1   2 , Paul G Richardson  1   2 , Kenneth C Anderson  1   2 , Noopur S Raje  2   6 , Birgit Knoechel  8   9   10 , Jens G Lohr  11   12   13
Affiliations

Dynamic transcriptional reprogramming leads to immunotherapeutic vulnerabilities in myeloma

Julia Frede et al. Nat Cell Biol. 2021 Nov.

Abstract

While there is extensive evidence for genetic variation as a basis for treatment resistance, other sources of variation result from cellular plasticity. Using multiple myeloma as an example of an incurable lymphoid malignancy, we show how cancer cells modulate lineage restriction, adapt their enhancer usage and employ cell-intrinsic diversity for survival and treatment escape. By using single-cell transcriptome and chromatin accessibility profiling, we show that distinct transcriptional states co-exist in individual cancer cells and that differential transcriptional regulon usage and enhancer rewiring underlie these alternative transcriptional states. We demonstrate that exposure to standard treatment further promotes transcriptional reprogramming and differential enhancer recruitment while simultaneously reducing developmental potential. Importantly, treatment generates a distinct complement of actionable immunotherapy targets, such as CXCR4, which can be exploited to overcome treatment resistance. Our studies therefore delineate how to transform the cellular plasticity that underlies drug resistance into immuno-oncologic therapeutic opportunities.

PubMed Disclaimer

Figures

Extended Data Fig. 1
Extended Data Fig. 1. Sorting strategy.
a) Sorting strategy for myeloma cells and normal donor plasma cells with representative flow cytometry plots. CD38+CD138+ cells were sorted after EasySep enrichment of CD138+ cells from bone marrow or peripheral blood. b) Sorting strategy for immune cell subsets. CD3+, CD19+, CD14+ and CD45+Lin- cells were sorted following exclusion of dead cells and doublets. c) Sorting strategy for NK cells. CD56+CD16+ cells (Q1–3) were sorted after exclusion of dead cells, doublets, CD3+ cells and CD138+ cells.
Extended Data Fig. 2
Extended Data Fig. 2. Quality assessment and filtering of single cell RNA data.
a) Distribution of library size (i), number of detected genes (ii), percentage of counts mapping to mitochondrial genes (iii), and percentage of counts mapping to house-keeping genes (HKGs) (iv) per cell. b) Scatter plot depicting principal component analysis using the top two dimensions. The PCA was performed on the four features depicted in a) for all cells in the unfiltered dataset. The outliers, highlighted in orange, were identified using the mvoutlier package. c) The distribution of features shown in a) after filtering out cells identified as outliers. d) Scatter plot depicting expression frequency vs mean read counts per gene in the filtered dataset. e) Density plot showing the contribution of various technical factors (timepoint, run date, individual, total features, total counts, percentage of counts mapping to mitochondrial genes, and percentage of counts mapping to house-keeping genes (HKGs)) to the total variation observed in the dataset.
Extended Data Fig. 3
Extended Data Fig. 3. Characterization of scRNA-Seq dataset of primary myeloma cells.
a) Patient characteristics: Genetic aberrations detected by FISH. b) Copy number profiles of myeloma patients were inferred from scRNA expression data using InferCNV. Normal donor plasma cells served as a control. See the enlarged version in Supplementary Information. c-e) Random forest model identifying genes that best discriminate myeloma from normal plasma cells. c) Plot depicting the error vs number of trees used by random forest model on malignant (green), non-malignant (red) and combined (black) cells. d) Relative importance of each gene in the model (mean decrease in Gini coefficient). e) Confusion matrix showing classification and error rates during training of the model, for prediction on the training set (predict_train) and the test set (predict_test). f) Detailed heatmap showing classification of malignant and normal plasma cells based on CNVs, CDR3 sequence, expression of translocation targets, and genes identified by random forest (RF) model that best discriminate between normal and malignant cells, and composite malignancy score for patient MM5. g) tSNE plots showing PAGODA2 clustering highlighting cells from patient MM5 with malignancy scores ≥ 1 or < 1. h) Heatmap with relative expression of marker genes for individual patients in single myeloma and normal plasma cells.
Extended Data Fig. 4
Extended Data Fig. 4. Transcriptional diversity is increased in MM.
a) tSNE plots showing expression of selected genes not normally expressed in plasma cells (log2-transformed counts). Depicted are CD20 (MS4A1), the B Lymphoid Tyrosine Kinase BLK, usually expressed in earlier stages of B cell development and the myeloid restricted serine protease C (CTSC). b) Heatmap with lineage scores from single cell RNA-Seq derived datasets from the Human Cell Atlas in single myeloma cells from patients, single plasma cells from normal donors or single B cells. HSC, hematopoietic stem cell; CLP, common lymphoid progenitor; CMP, common myeloid progenitor; MKP, MK progenitor; ERP, ER progenitor; CD34 Gran, CD34+ Granulocyte progenitor, MixLin, mixed lineage progenitor; PreB, Pre-B cell; ProB, Pro-B cell; PC, plasma cell. c) Heatmap showing relative expression of individual genes from different lineages from the Human Cell Atlas dataset in single myeloma cells, single normal donor plasma cells or single B cells. d) UMAP colored by patient. The enlarged heatmaps in b-c are provided in Supplementary Information. e) RNA velocity estimates of single myeloma cells and normal donor plasma cells projected onto two-dimensional UMAP. Normal donor plasma cells are indicated in red. f) Cells colored based on cell cycle phase. g) Number of genes detected per cell in single myeloma vs normal donor plasma cells. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,162 cells. p < 2.2e-16 by two-sided Wilcoxon-test. h,i) Entropy is increased in myeloma vs normal donor plasma cells. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,162 cells. p < 2.2e-16 by two-sided Wilcoxon-test. j) Validation of entropy using an alternative protein-protein interaction network. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,162 cells. p < 2.2e-16 by two-sided Wilcoxon-test. k) Predicted ordering by CytoTRACE, which orders MM cells based on their developmental potential from most mature (lowest values) to most immature (highest values). Boxplots show the median and interquartile range, whiskers extend from min to max. l) tSNE plots showing expression of plasma cell lineage transcription factors XBP1, IRF4, PRDM1, FOS, POU2AF1 and ZBTB20 (log2-transformed counts) in single myeloma cells, normal donor plasma cells and B cells.
Extended Data Fig. 5
Extended Data Fig. 5. Gene regulatory network activity in different cell types.
a-d) Gene regulatory network activity for different cell types was determined from the BLUEPRINT dataset. a) Gene expression in normal plasma cells. b) Gene expression in hematopoietic stem cells (HSCs). c) Gene expression in CLP (common lymphoid progenitor) population. d) Gene expression in macrophages. e) Network layouts for normal hematopoiesis based on the BLUEPRINT dataset (top) and based on our single-cell RNA data (bottom) illustrating extensive rewiring, gain of new connections and changes in relative activity. Edges are colored based on regulon activity where high activity is indicated in red, low activity in blue. Target genes are depicted in white, transcription factors which are not among regulons are shown in green. f) Changes in regulon activity in myeloma compared to normal donor plasma cells are projected onto the network. Transcription factors with the largest difference in regulon activity in myeloma compared to normal donor plasma cells are highlighted in insets. g) tSNE plots showing regulon activity (area under the curve, AUC) of plasma cell lineage transcription factors XBP1, IRF4, PRDM1, and FOS in single myeloma and normal plasma cells.
Extended Data Fig. 6
Extended Data Fig. 6. Quality assessment and filtering of single cell ATAC data.
a) Filtering of single cell ATAC profiles based on TSS enrichment and number of unique nuclear fragments. b) Fragment length distribution of filtered scATAC profiles showing characteristic distribution with nucleosome-free region and mononucleosome peaks. c) TSS enrichment after filtering per sample. d) Number of unique fragments per sample. e,f) tSNE plot colored by clusters (e) or individual (f). g,h) tSNE plots following batch effect removal by Harmony colored by clusters (g) or individual (h). i-k) Defining cluster identities following integration with scRNA-seq data. i) tSNE plot colored by predicted cell type identities. j) tSNE plot showing cell type identities by cluster. k) Heatmap showing confusion matrix for predicted cell type identities by cluster. l-o) Tracks with aggregated ATAC profiles for each cluster for marker genes SDC1, LYZ, MS4A1 and FCGR3A, respectively. p-s) tSNE plots colored by gene scores for marker genes SDC1, LYZ, MS4A1 and FCGR3A, respectively.
Extended Data Fig. 7
Extended Data Fig. 7. Annotation of peaks from aggregated scATAC data.
a) Processing of scATAC profiles. b) Intersection of peaks in aggregate scATAC samples showing the overlap of peaks between MM and normal donor (ND) samples. c) Number of peaks significantly different by DESeq2 enriched in ND or MM (FDR ≤ 0.1, absolute log2FC ≥ 1). d) Heatmap showing differentially accessible peaks from c) in ND and MM, sorted by the ND sample. e) ChromHMM state annotation in aggregate scATAC samples. Depicted is the fraction of peaks in each of the indicated states. f) Number of peaks in ChromHMM state 13 corresponding to heterochromatin. **p = 0.0029 by two-sided t-test. g) Fraction of peaks in indicated genomic regions. h) Boxplot comparing distance to TSS in ND vs MM. ****p < 2.2e-16 by two-sided Wilcoxon test. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 46,935 peaks and 68,752 peaks, respectively. i) Heatmap showing accessibility of enhancers (n=15,748) associated with the top multi-enhancer genes, sorted by the ND sample. j) Barplot showing genes with ≥ 2 enhancer interactions by ABC model. k) Heatmap showing multi-enhancer genes by ABC model (n=16,635), sorted by the ND sample. l) Barplot showing gene set enrichment analysis for multi-enhancer genes in MM by ABC model. m) Barplot showing differential motif enrichment analysis of the single cell ATAC-Seq dataset comparing myeloma cells with normal donor plasma cells. Shown are the top 40 differentially enriched transcription factor motifs ordered by FDR. n) Venn diagram showing overlap of differentially enriched motifs and rewired transcription factors determined by DyNet algorithm.
Extended Data Fig. 8
Extended Data Fig. 8. Alternative splicing following treatment in MM.
a) Quantification of exon inclusion/exclusion ratio (percent of spliced in, psi). Volcano plot showing differential splicing at C5D1 vs screening timepoints with FDR < 0.05. b) Barplot showing gene set enrichment analysis of differentially spliced transcripts. c) UMAP colored by Louvain clusters based on calculated psi (percent of spliced in) values. d) psi-based UMAP colored by individual. e) psi-based UMAP colored by timepoint. f) Violin plots showing the single cell distribution of logit (percent spliced in) values at screening and C5D1 for selected differentially spliced transcripts. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,374 cells. **** p ≤ 0.0001 and * p ≤ 0.05 by two-sided Wilcoxon test. g-j) Miso plots visualizing splice junctions and potential exon skipping events in differentially spliced transcripts for TSC1 (g), CD200 (h) and CALU (i) as well as SLAMF7 (j). Splice probabilities are shown on the right.
Extended Data Fig. 9
Extended Data Fig. 9. Transcriptional diversity following treatment.
a) Number of genes expressed in primary MM cells before and after treatment (p<2.2e-16 by two-sided t-test). Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,374 cells. b) Bar graph showing relative proportion of cells in each cell cycle phase at screening and C5D1. c) Shown are the lineage TF regulons downregulated and upregulated upon treatment. d) Change in regulon activity of lineage TFs upon treatment. e,f) GO-term enrichment of ATAC-Seq peaks gained (e) and lost (f) in MOLP2 cells following 72h treatment with PVD. g) Gene set enrichment analysis of the top 500 genes gaining enhancers following treatment with PVD. h) Genes with ≥ 5 enhancer interactions by ABC model. i) Gene set enrichment analysis of genes gaining enhancer interactions following treatment with PVD by ABC model. j) Top rewired TFs with differentially accessible motifs upon treatment.
Extended Data Fig. 10
Extended Data Fig. 10. Surface marker expression in MM.
a) Genes upregulated upon treatment. Shown is a barplot with log2 fold change values compared to screening timepoint. b) Motif enrichment in differential ATAC-Seq peaks in MMCL MOLP2 after PVD treatment (right), comparing to untreated (left). p values are calculated using binomial test. c-e) Live cell counts following 72h of treatment with pomalidomide (Pom), bortezomib (Bor), dexamethasone (Dex) and combination of all three drugs (PVD) in MM cell lines MOLP2 (c), MM1.S (d) and OPM2 (e) as a percentage of total cell numbers. f,g) Quantification of CXCR4 surface levels by flow cytometry following treatment with pomalidomide (Pom), bortezomib (Bor), dexamethasone (Dex) and combination of all three drugs (PVD) in myeloma cell lines MOLP2 (f) and MM1.S (g). c-g) Significance was assessed using a two-sided t-test with Welch’s correction, with * p ≤ 0.05 and **** p ≤ 0.0001. Data are presented as mean +/− SD, n=3 replicates. h,i) Cell viability following treatment with CXCR4 inhibitors BKT140 (h) and Plerixafor (i) following pre-treatment with PVD in MMCL MOLP2. Data are presented as mean +/− SD, n=3 replicates. n.s. p>0.05, * p ≤ 0.05, *** p ≤ 0.001 and **** p ≤ 0.0001 by two-sided t-test.
Fig. 1:
Fig. 1:. Co-expression of multiple transcriptional signatures in MM cells.
a) Schematic illustrating experimental setup. b) T-stochastic network embedding (tSNE) of the processed single cell RNA-Seq dataset color-coded by clusters identified by PAGODA2. c) tSNE plot color-coded by individual. d) Cell type annotation based on 21 immune cell populations from the BLUEPRINT dataset. e) Classification of malignant and normal plasma cells based on CNVs, CDR3 sequence, expression of translocation targets, and genes identified by random forest (RF) model that best discriminate between normal and malignant cells. f) Composite malignancy score to distinguish myeloma cells from normal plasma cells. Cells from normal donors are aggregated to facilitate comparison. Boxplots show median and interquartile range, whiskers extend to 1.5x interquartile range (n= 1,162 cells. ns p>0.05 and **** p ≤ 0.0001 by Wilcoxon test compared to ND). g) Co-expression of multiple established myeloma gene expression classifiers in single myeloma cells.
Fig. 2:
Fig. 2:. Transcriptional heterogeneity in single myeloma cells.
a) tSNE plot colored by cell cycle phase predicted by SEURAT. b) Stacked bar plot showing the relative proportion of cells in each cell cycle phase per cluster. c) Heatmap showing expression scores for recurrent heterogeneity programs in MM patients (scores identified across 198 cell lines reflecting 22 cancer types). d) Consensus matrix depicting pairwise similarities between NMF programs ordered by hierarchical clustering. Six clusters corresponding to the six identified programs and assignment of patients are indicated on top. Functional annotation and selected marker genes are shown below. e) Heatmap showing the top 50 genes based upon the highest NMF scores selected as signature genes for each program with selected genes labeled. f) Functional enrichment (−log10 FDR) of heterogeneity programs with six annotated gene sets. g) Heatmap showing expression scores for the six heterogeneity programs identified in d). The enlarged heatmaps in c and g are provided in Supplementary Information.
Fig. 3:
Fig. 3:. Lineage infidelity - transcriptional states diverge towards immature states.
a) Heatmap showing lineage scores for selected cell types from the BLUEPRINT dataset in single myeloma cells from patients, plasma cells from normal donors or B cells. CLP, common lymphoid progenitor; CMP, common myeloid progenitor; MPP, multipotent progenitor; MEP, megakaryocyte-erythroid progenitor; PC, plasma cell. b) Heatmap of single myeloma cells from patients or normal donor plasma cells showing expression of genes corresponding to selected cell types from the BLUEPRINT dataset. The enlarged heatmaps in a-b are provided in Supplementary Information. c,d) RNA velocity estimates, root and end states projected onto UMAP. Root and endpoints are circled. e) CytoTRACE values showing distribution of differentiation states from most immature (highest values) to most mature (lowest values). f) Expression of marker gene LAMP5 correlated with immaturity (log2-transformed counts). g) Gene set enrichment analysis of genes correlated with high CytoTRACE (associated with immaturity). h,i) Signaling activity scores across single cells for MAPK (h) and PI3K (i) signaling pathways, projected on UMAP.
Fig. 4:
Fig. 4:. Differential regulon activity and transcriptional rewiring in MM underlie alternate differentiation states.
a) Gene regulatory network (GRN) constructed based on transcriptional modules present in normal hematopoiesis. Highlighted are plasma cell module and its dominating transcription factors. b) Gene expression in normal donor plasma cells (top) and myeloma cells (bottom) projected onto normal hematopoiesis network. The plasma cell module is highlighted. c) GRN constructed exclusively from myeloma cells (red) overlayed with GRN from normal hematopoietic cells (green). d) Regulons, i.e. transcription factors with their target genes, as part of network layouts in normal hematopoiesis (Blueprint, top) and myeloma (bottom). Regulons were chosen based on rewiring score (e). e, f) Transcription factors (e) or CD markers (f) ordered by rewiring score determined by the DyNet algorithm. g-k) Dependency data were downloaded from the Cancer Dependency Map (https://depmap.org/portal/). g) Lineage dependency score in MM cell lines for top rewired TFs in plasma cells. Data are presented as mean values +/− standard error. n=20 cell lines. h) Scaled dependency scores for 675 cell lines from 21 different lineages ordered by dependency for XBP1. Myeloma cell lines (plasma cell lineage) are highlighted in pink. i) Venn diagram showing overlap between lineage dependency genes and rewired TFs. Shown are TFs in overlap. j) TFs in overlap between lineage dependencies and rewired TFs ordered by lineage dependency score in MM cell lines. Data are presented as mean values +/− standard error. n=20 cell lines. i,j) Lineage TFs are highlighted in pink. k) Scaled dependency scores for 675 cell lines from 21 different lineages ordered by dependency for TAL1. Myeloma cells (plasma cell lineage) are highlighted in pink.
Fig. 5:
Fig. 5:. Transcriptional reprogramming in MM and associated changes in chromatin accessibility and enhancer rewiring.
a) Schematic illustrating experimental setup. b) tSNE plot showing clustering of single myeloma cells colored by clusters based on single cell ATAC (scATAC). c) tSNE plot showing clustering of single myeloma cells colored by patient based on single cell ATAC (scATAC). d) Number of associated enhancers per gene for genes associated with ≥ 10 enhancers in normal donor plasma cells or patient-derived myeloma cells based on association by GREAT. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n=1,020 and 1,159 genes, respectively. p=5.5e-08 by two-sided Wilcoxon test. e) Pseudotime trajectory. f) Heatmap showing correlation of gene scores with pseudotime. Pseudotime increases from left to right. g) Heatmap showing correlation of motif accessibility with pseudotime. Pseudotime increases from left to right. h) tSNE plots of single cell ATAC-Seq colored based on bias-corrected z-scores of selected rewired transcription factors that bind differentially accessible motifs. Shown are TEAD4, FOXO3, MEF2A and IRF8. i, j) Peak-to-gene-linkage determined based on integrated single cell ATAC for TEAD4 locus (i) and ELF3 locus (j). k) Model of developmental potential and transcriptional diversity in MM.
Fig. 6:
Fig. 6:. Treatment reduces developmental potential while increasing regulon activity.
a-d) Comparison of primary MM cells on treatment (C5D1) compared to screening timepoint. a) Exon inclusion/exclusion ratio (percent of spliced in, psi) quantified at screening vs. C5D1 for transcripts with low, medium or high splicing probability. Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n=1,374 cells. **** p ≤ 0.0001 by two-sided Wilcoxon test. b) CytoTRACE values in MM cells before and after treatment (p<2.2e-16 by two-sided t-test). Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,374 cells. c) Number of regulons downregulated (160/365) and upregulated (n= 205/365) upon treatment. d) Change in regulon activity upon treatment. Selected lineage and non-lineage transcription factors are highlighted. e-g) MMCL MOLP2 was treated with pomalidomide, bortezomib and dexamethasone (PVD) for 72h prior to single cell RNA sequencing. e) CytoTRACE values for cells treated with PVD or DMSO (p<2.2e-16 by two-sided t-test). Boxplots show the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 192 cells. f) Regulons downregulated and upregulated in cells treated with PVD compared to DMSO. g) Change in regulon activity in cells treated with PVD compared to DMSO. Selected lineage TF regulons are highlighted. h) ATAC sequencing was performed on MOLP2 cells after 72h of treatment with PVD or DMSO. i) Venn diagram illustrating overlap of peaks in cells treated with DMSO or PVD. j) Number of peaks gained and lost after treatment determined by DESeq2. k) Number of genes associated with ≥ 10 enhancers in cells treated with DMSO or PVD by GREAT. l) Motif enrichment in MOLP2 cells after 72h of treatment with PVD vs DMSO. Enriched TF motifs are highlighted in red.
Fig. 7:
Fig. 7:. Inducible and stable expression of putative immunotherapy targets on myeloma cells.
a) Expression of selected surface markers not found in normal plasma cells in single myeloma (MM) cells or normal donor plasma cells (ND). b) Deconvolution of surface marker expression in normal donor plasma cells, B cells and myeloma cells from patient MM1 at screening and at cycle 5 day 1 (C5D1) based on enrichment of surface marker signatures derived from BLUEPRINT dataset. c,d) CXCR4 mRNA (c) and surface protein (d) expression in single myeloma cells from patient MM1 before treatment and at C5D1. e) Co-accessibility determined based on single cell ATAC data showing the CXCR4 locus. Aggregated scATAC tracks show chromatin accessibility upstream of CXCR4 at screening and C5D1 with differential peaks highlighted in grey. IRF4 motifs in peaks and IRF4 ChIP peaks from KMS12BM are displayed below. f) Motif enrichment of IRF4 and IRF composite elements in ATAC-peaks in MMCL MOLP2 after 72h of treatment. p values calculated using binomial test. g) IRF4 regulon activity in primary MM cells (p=1.3e-05 by two-sided t-test). Boxplots represent the median and interquartile range, whiskers extend to 1.5x the interquartile range. n= 1,374 cells. h) Treatment of MOLP2 myeloma cells with dexamethasone (Dex) or combination of dexamethasone, pomalidomide, bortezomib (PVD) resulting in upregulation of CXCR4 at the cell surface. i) Quantitative analysis of CXCR4 surface protein expression changes on OPM2 myeloma cells with drug exposure by flow-cytometry (MFI = mean fluorescence intensity). Data are presented as mean +/− SD, n=3 independent experiments. Significance was assessed using an unpaired two-sided t-test with Welch’s correction; ** p ≤ 0.01. j) Drug removal resulting in downregulation of CXCR4 surface expression in MOLP2 cells. k,l) Cell viability following treatment with CXCR4 inhibitors BKT140 (k) and Plerixafor (l) following pre-treatment with PVD in MMCL OPM2. Data are presented as mean +/− SD, n=3 independent experiments. ** p ≤ 0.01 and *** p ≤ 0.001 by unpaired two-sided t-test.
See this image and copyright information in PMC

References

    1. Sharma SV et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141, 69–80, doi:10.1016/j.cell.2010.02.027 (2010). - DOI - PMC - PubMed
    1. Hangauer MJ et al. Drug-tolerant persister cancer cells are vulnerable to GPX4 inhibition. Nature 551, 247–250, doi:10.1038/nature24297 (2017). - DOI - PMC - PubMed
    1. Shaffer SM et al. Rare cell variability and drug-induced reprogramming as a mode of cancer drug resistance. Nature 546, 431–435, doi:10.1038/nature22794 (2017). - DOI - PMC - PubMed
    1. Risom T. et al. Differentiation-state plasticity is a targetable resistance mechanism in basal-like breast cancer. Nat Commun 9, 3815, doi:10.1038/s41467-018-05729-w (2018). - DOI - PMC - PubMed
    1. Verfaillie A. et al. Decoding the regulatory landscape of melanoma reveals TEADS as regulators of the invasive cell state. Nat Commun 6, 6683, doi:10.1038/ncomms7683 (2015). - DOI - PMC - PubMed

Publication types

MeSH terms