FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313013559> ?p ?o ?g. }

• W4313013559 abstract "Article Figures and data Abstract Editor's evaluation eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Tumour heterogeneity is thought to be a major barrier to successful cancer treatment due to the presence of drug resistant clonal lineages. However, identifying the characteristics of such lineages that underpin resistance to therapy has remained challenging. Here, we utilise clonal transcriptomics with WILD-seq; Wholistic Interrogation of Lineage Dynamics by sequencing, in mouse models of triple-negative breast cancer (TNBC) to understand response and resistance to therapy, including BET bromodomain inhibition and taxane-based chemotherapy. These analyses revealed oxidative stress protection by NRF2 as a major mechanism of taxane resistance and led to the discovery that our tumour models are collaterally sensitive to asparagine deprivation therapy using the clinical stage drug L-asparaginase after frontline treatment with docetaxel. In summary, clonal transcriptomics with WILD-seq identifies mechanisms of resistance to chemotherapy that are also operative in patients and pin points asparagine bioavailability as a druggable vulnerability of taxane-resistant lineages. Editor's evaluation This study advances a novel strategy of lineage tracing coupled with single-cell transcriptomics to allow unique insights into tumor heterogeneity and the diversity of response to treatment. These analyses reveal new insights into mechanisms of taxane resistance. Overall, the study is scientifically robust and puts forward a new methodology that will be of interest to scientists as well using this technology to gain insights into the factors that inform resistance to taxane treatment in an in vivo cancer model. https://doi.org/10.7554/eLife.80981.sa0 Decision letter Reviews on Sciety eLife's review process eLife digest Cancer begins when a cell multiplies again and again to form a tumour. By the time that tumour measures a centimetre across, it can contain upwards of a hundred million cells. And even though they all came from the same ancestor, they are far from identical. The tumour's family tree has many branches, and each one responds differently to treatment. If some are susceptible to a drug the cells die, the tumour shrinks, and the therapy will appear to be successful. But, if even a small number of cancer cells survive, they will regrow, often more persistently, causing a relapse. Identifying resistant cells, their characteristics, and how to kill them has been challenging due to a lack of good animal models. One way to keep track of a cancer family tree is to insert so-called genetic barcodes into the ancestral cells. As the tumour grows, the cells will pass the barcodes to their descendants. Scientists do this by using viruses that naturally paste their genes into the cells they infect. Applying this technique to an animal model of cancer could reveal which genes allow some cells to survive, and how to overcome them. Wild, Cannell et al. developed a genetic barcoding system called WILD-seq and used it to track all the cells in a mouse tumour. The mice received the same drugs used to treat patients with breast cancer. By scanning the genetic barcodes using recently developed single cell sequencing technologies, Wild, Cannell et al. were able to identify and count each type of cancer cell and work out which genes they were using. This revealed which cells the standard treatment could not kill and exposed their genetic weaknesses. Wild, Cannell et al. used this information to target the cells with a drug currently used to treat leukaemia. The drug identified by this new genetic barcoding approach is already licensed for use in humans. Further investigation could reveal whether it might help to shrink breast tumours that do not respond to standard therapy. Similar experiments could uncover more information about how other types of tumour evolve too. Introduction Intra-tumoural heterogeneity (ITH) is thought to underlie tumour progression and resistance to therapy by providing a reservoir of phenotypically diverse clonal lineages on which selective pressures from the microenvironment or therapeutic intervention exert their effects (Bhang et al., 2015; Turajlic and Swanton, 2016). Inference of clonal composition from bulk sequencing has elucidated the breadth of ITH across tumour types and suggests that often rare pre-existing clones can resist therapy-induced killing to drive relapse (Dentro et al., 2021; Ding et al., 2012; Gerlinger et al., 2012; Jamal-Hanjani et al., 2014; Landau et al., 2013). However, such methods are limited by their inability to characterise such resistant clones beyond genotype and how their properties change over time and in response to therapy. Recently, several lineage tracing approaches have emerged that are able to link clonal identity with gene expression by utilising expressed genetic barcodes that are read-out by single cell RNA sequencing (Biddy et al., 2018; Gutierrez et al., 2021; Quinn et al., 2021; Simeonov et al., 2021; Weinreb et al., 2020; Yang et al., 2022). These powerful methods allow deconvolution of complex mixtures of clones while simultaneously providing a gene expression profile of those cells that can indicate the pathways on which they depend. However, to date in solid tumours these technologies have mostly been used to study drug response in vitro (Gutierrez et al., 2021; Oren et al., 2021) or metastatic dissemination in vivo (Quinn et al., 2021; Simeonov et al., 2021; Yang et al., 2022) and have not been utilised to study therapeutic response in immune-competent models. A thorough understanding of the biomarkers of sensitivity and mechanisms of resistance to chemotherapy is essential if we are to improve patient outcomes. Most existing combination cancer therapies are not rationally designed but were instead empirically optimised to avoid overlapping toxicities. More recently alternative therapeutic strategies have emerged including synthetic lethality, drug synergy (AlLazikani et al., 2012; ONeil et al., 2017), and collateral sensitivity (Mueller et al., 2021; Pluchino et al., 2012; Zhao et al., 2016) that aim to leverage selective vulnerabilities of tumour cells while minimising toxicity. Of particular promise is collateral sensitivity, in which as a tumour becomes resistant to one drug it comes at the cost of sensitivity to a second drug. Since many modern clinical trials occur in the context of neo-adjuvant chemotherapy, the identification of frontline therapy-induced collateral sensitivities to second line therapy would have the potential to be rapidly translated into improved outcomes for patients. Here, we describe WILD-seq (Wholistic Interrogation of Lineage Dynamics by sequencing), an accessible and adaptable platform for lineage tracing at the single-cell transcriptomic level that facilitates in vivo analysis of clonal dynamics and apply it to the study of syngeneic triple negative breast cancer (TNBC) mouse models. Our optimised pipeline ensures recurrent representation of clonal lineages across animals and samples, facilitating analysis of clonal dynamics under the selective pressure of therapeutic intervention. Importantly, analysis of response of TNBC models to frontline taxane-based chemotherapy revealed an enrichment of clones with high levels of NRF2 signaling, implicating defense against oxidative damage as a major determinant of resistance to chemotherapy. Building on the work of others (LeBoeuf et al., 2020), we show that these NRF2-high, taxane-resistant lineages are collaterally sensitive to asparagine deprivation as a result of L-asparaginase treatment and that they adapt to this second line intervention by up-regulating de novo asparagine synthesis by increasing asparagine synthetase (Asns) expression. Together these data indicate that high levels of NRF2 signaling, which is also observed in patients following neo-adjuvant chemotherapy, promotes both resistance to chemotherapy and sensitivity to asparagine deprivation and warrant the exploration of L-asparaginase as a therapeutic modality in solid tumours. Results Establishment of an expressed barcode system to simultaneously detect clonal lineage and gene expression WILD-seq uses a lentiviral library to label cells with an expressed, heritable barcode that enables identification of clonal lineage in conjunction with single cell RNA sequencing. The WILD-seq construct comprises a zsGreen transcript which harbours in its 3´ untranslated region (UTR) a barcode consisting of two 12 nucleotide variable regions separated by a constant linker (Figure 1a). Each variable region is separated from any other sequence in the library by a Hamming distance of 5 to allow for library preparation and sequencing error correction and we detected over 2.7 million unique barcodes in our vector library by sequencing after clustering based on Hamming distance. The barcode is appropriately positioned relative to the polyadenylation signal to ensure its capture and sequencing by standard oligo-dT single-cell sequencing platforms. Figure 1 Download asset Open asset Establishment of an expressed barcode system to simultaneously detect clonal lineage and gene expression from single cells in vivo. (a) Lentiviral construct design. An attenuated PGK promoter drives expression of a transcript encoding zsGreen and harboring a WILD-seq barcode sequence in the 3′ UTR. A spacer sequence and polyadenylation signal ensure that that the barcode is detectable as part of a standard oligo dT single-cell RNA library preparation and sequencing pipeline. The barcode cassette comprises 2 distinct 12 nucleotide barcode sequences separated by a constant 20 nucleotide linker region. The library of barcode sequences was designed with Hamming distance 5 to allow for sequencing error correction. (b) Schematic of WILD-seq method. Tumour cells are infected with the WILD-seq lentiviral library and an appropriate size population of zsGreen positive cells isolated, each of which will express a single unique WILD-seq barcode. This WILD-seq barcoded, heterogenous cell pool is then subjected to an intervention of interest (such as in vivo treatment of the implanted pool with a therapeutic agent) and subsequently analysed by single cell RNA sequencing using the 10x Genomics platform. An additional PCR amplification step is included that specifically enriches for the barcode sequence to increase the number of cells to which a WILD-seq barcode can be conclusively assigned. (c) scRNA-seq of in vitro 4T1 WILD-seq cell pool. UMAP plot of in vitro cultured 4T1 WILD-seq cells. Cells for which a WILD-seq clonal barcode is identified are shown as dark grey or coloured spots. Cells which belong to five selected clonal lineages are highlighted. (d) scRNA-seq of 4T1 WILD-seq tumours. UMAP plots of vehicle-treated 4T1 WILD-seq tumours generated by injecting the 4T1 WILD-seq pool into the mammary fatpad of BALB/c mice. Four independent experiments were performed each involving injection into three separate host animals. Six animals from experiments A and B received vehicle 1 (10% DMSO, 0.9% β-cyclodextrin) and six animals from experiments C and D received vehicle 2 (12.5% ethanol, 12.5% Kolliphor). (e) Clonal representation. Proportion of tumour cells assigned to each clonal lineage based on the WILD-seq barcode (n=1 for in vitro cultured cells, n=6 for tumours from NSG mice, n=12 for vehicle-treated tumours from BALB/c mice). Selected clones from the most abundant lineages are plotted. Data represents mean ± SEM. (f) Principal component analysis of clonal transcriptomes. Pseudo-bulk analysis was performed by summing counts for all tumour cells expressing the same WILD-seq clonal barcode within an independent experiment. For in vivo tumour samples each point represents the combined cells from three animals. Principal component analysis of normalized pseudo-bulk count data showed separation of samples by origin with PC1 and PC2 and separation by clonality with PC3. (g) Transcriptomic programs associated with principal components. The top/bottom 50 gene loadings of PC1, PC2, and PC3 were analyzed using Enrichr (Chen et al., 2013; Kuleshov et al., 2016; Xie et al., 2021). (h) Clonal transcriptomic signatures from vehicle-treated BALB/c tumours. An AUCell score (Aibar et al., 2017) enrichment was calculated for each clone and for each experiment by comparing cells of a specific clonal lineage of interest to all assigned tumour cells within the same experiment. All gene sets which showed consistent and statistically significant enrichment in one of the six most abundant clones across experiments are illustrated. The standard WILD-seq pipeline is illustrated in Figure 1b. A heterogeneous cell line is transduced with a barcode library at low multiplicity of infection (MOI) to ensure that each cell receives a maximum of one barcode. An appropriate size pool of barcoded clones is selected and stabilised in culture. Empirically, we have found that three separate pools each established from 250 individual clones, that are maintained separately and combined immediately prior to implantation, works well to provide effective representation of the diversity within the cell lines used herein, while also enabling recurrent representation of the same clones across animals and experiments. Once stabilised in culture, the pool of WILD-seq clones can be analysed directly by single-cell sequencing or injected into a recipient animal for in vivo tumour growth. WILD-seq single-cell sequencing libraries can be prepared using a standard oligo-dT-based protocol and addition of an extra PCR amplification step can be used to increase coverage of the barcode region and aid cell lineage assignment. We first established a WILD-seq clonal pool from the mouse 4T1 cell line, a triple negative mammary carcinoma model that can be orthotopically implanted into the mammary fat pad of a BALB/c syngeneic host, which we have previously shown to be heterogeneous with distinct sub-clones having unique biological properties (Wagenblast et al., 2015). We performed single-cell sequencing of the in vitro WILD-seq pool (Figure 1c) and in vivo tumours derived from this clonal pool (Figure 1d). Over the course of our studies, we injected multiple cohorts of mice with our WILD-seq 4T1 pool as detailed in Supplementary file 1, some of which were subjected to a specific drug regime. All tumours were harvested at humane endpoint, as determined by tumour volume unless otherwise stated and immediately dissociated for single-cell sequencing. For the purpose of characterising the baseline properties of our clones, we performed an in-depth transcriptomic analysis of all tumours from vehicle-treated animals. A WILD-seq barcode and thereby clonal lineage could be unambiguously assigned to 30–60% of cells per sample within the presumptive tumour cell/mammary epithelial cell cluster. A total of 132 different WILD-seq barcodes were observed in vitro and in total 94 different WILD-seq barcodes were observed across our in vivo tumour samples. Our in vivo tumour samples comprised both tumour cells and host cells of the tumour microenvironment including cells of the innate and adaptive immune system, enabling simultaneous profiling of the tumour and its microenvironment (Figure 1d). Clustering was performed after removal of reads mapping to the WILD-seq vector, to avoid any influence of the WILD-seq transcript on clustering, and the WILD-seq barcode assignment subsequently overlaid onto these data. The tumour cell clusters were clearly identifiable by the high expression of the barcode transcript (Figure 1d). Occasionally a barcode was observed in cells which clustered according to their transcriptome outside of the main tumour cluster. Since this could be the result of sequencing or technical error causing a mismatch between the WILD-seq barcode and the cell of origin, only barcoded cells that clustered within the main tumour/mammary epithelium cell cluster were included in our analysis. We reproducibly observed the same clonal populations across animals and independent experiments which is critical to our ability to examine the effects of different interventions and treatments (Figure 1d and e). The relative abundance of clones was similar in tumours grown in NOD scid gamma (NSG) immunodeficient and BALB/c immunocompetent mice but was drastically different to that found in the in vitro cell pool from which they were established (Figure 1e, Supplementary file 2), suggesting that clones that show greatest fitness in cell culture do not necessarily show fitness in vivo. Therefore, in vitro clonal lineage tracking experiments are likely to capture a different collection of clones and have the potential to identify sensitive or resistance clones that are not represented in vivo. Pseudo-bulk analysis of the major clonal lineages revealed that the composition of the tumour microenvironment has a dramatic effect on the transcriptome of the tumour cells for all clones (Figure 1f). Comparison of in vitro culture, tumours from NSG mice, and tumours from BALB/c mice by principal component analysis (PCA), showed clear separation of the tumour cells depending on their environment, with differences in interferon gamma signaling, TNF-alpha signaling, and cell cycle being most prominent between cells grown in vivo and in vitro (PC1, Figure 1g). Differences in gene expression between tumours growing in immunocompetent and immunodeficient hosts were related to changes in the expression of extracellular matrix proteins and changes in interferon gamma and Il-2 signaling, consistent with the differences in T-cell abundance (PC2, Figure 1g). These data highlight the importance of the host immune system in sculpting the transcriptome and provide cautionary context for the analysis of tumour gene expression in immune-compromised hosts. Although there were large differences between clonal gene expression patterns across hosts, the clones showed consistent differences in gene expression across all settings, reflective of intrinsic clonal properties, with the biggest variation in gene expression across the clones being related to their position along the epithelial-mesenchymal transition (EMT) axis (PC3, Figure 1g). In particular, Clone 679 is the most distinct and the most mesenchymal of the clones. To further characterise the major clones in our tumours, we performed gene set expression analysis using AUCell (Aibar et al., 2017) to identify pathways that are enriched in cells of a specific clonal lineage. Analysis was performed across four independent experiments each with three vehicle-treated animals and for the majority of clones we were able to identify distinct gene expression signatures that were reproducible across animals and experiments (Figure 1h, Supplementary file 4, Supplementary file 5). Simultaneous detection of changes in clonal abundance, gene expression, and tumour microenvironment in response to BET bromodomain inhibition with WILD-seq Having established that we can repeatedly observe the same clonal lineages and their gene expression programs across animals and experiments, we next sought to perturb the system. We chose the BET bromodomain inhibitor JQ1 for our proof-of-principle experiments to assess the ability of the WILD-seq system to simultaneously measure changes in clonal abundance, gene expression and the tumour microenvironment that occur following therapeutic intervention. JQ1 competitively binds to acetylated lysines, displacing BRD4 and thereby repressing transcription at specific loci. A large number of studies have indicated that BET inhibitors may be beneficial in the treatment of hematological malignancies and solid tumours including breast cancer, possibly by inhibiting certain key proto-oncogenes such as MYC (Jiang et al., 2020). Treatment of our 4T1 WILD-seq tumour-bearing mice with JQ1 caused an initial suppression of tumour growth but with only a small overall effect on time to humane endpoint (Figure 2a). Tumours treated with JQ1 or vehicle alone were harvested at endpoint, dissociated and subjected to single-cell sequencing (Figure 2b). Two independent experiments were performed, each with 3 mice per condition. Figure 2 with 1 supplement see all Download asset Open asset Simultaneous detection of changes in clonal abundance, gene expression, and tumour microenvironment in response to BET bromodomain inhibition with WILD-seq. (a) Tumour growth curves with JQ1 treatment. 4T1 WILD-seq tumours were treated with the BET bromodomain inhibitor JQ1 or vehicle from 7 days post-implantation until endpoint (n=4 mice per condition). 75 mg/kg JQ1 (dissolved in DMSO and diluted 1:10 in 10% β-cyclodextrin) 5 days/week (5 consecutive days followed by 2 days off). Data represents mean ± SEM. (b) scRNA-seq of JQ1-treated 4T1 WILD-seq tumours. UMAP plots of vehicle- or JQ1-treated 4T1 WILD-seq tumours. Combined cells from 2 independent experiments, each with 3 mice per condition are shown. Cells for which a WILD-seq clonal barcode is identified are shown as dark grey or coloured spots. Cells which belong to four selected clonal lineages are highlighted. (c) JQ1 treatment results in a reduction in Cd8+ T cells within 4T1 tumours. Cells belonging to the T-cell compartment were computationally extracted from the single cell data and reclustered. Upper panels show combined UMAP plots from experiments A and B with Cd8a expression per cell illustrated enabling identification of the Cd8+ T cell cluster. Lower panels show neighbourhood graphs of the results from differential abundance testing using Milo (Dann et al., 2022). Coloured nodes represent neighbourhoods with significantly different cell numbers between conditions (FDR <0.05) and the layout of nodes is determined by the position of the neighbourhood index cell in the UMAP panel above. Experiments A and B were analysed separately due to differences in cell numbers. (d) Differential gene expression between JQ1- and vehicle-treated tumour cells. Single cell heatmap of expression for genes which are significantly and consistently down-regulated across clonal lineages (combined fisher p-value <0.05 and mean logFC <–0.2 for both experiments).1600 cells are represented (400 per experiment/condition), grouped according to their clonal lineage. (e) Differential gene set expression between JQ1 and vehicle-treated tumour cells. Median AUCell score per experiment/condition for selected gene sets. The five clonal lineages with the highest representation across experiments are shown. (f) Clonal representation. Proportion of tumour cells assigned to each clonal lineage in experiment A based on the WILD-seq barcode (n=3 tumours per condition). Clones which make up at least 2% of the assigned tumour cells under at least one condition are plotted. The most sensitive clone 473 is highlighted in blue and the most resistant clones 93, 439, 264 are highlighted in red. Data represents mean ± SD. (g) Clonal response to JQ1-treatment. Log2 fold change in clonal proportions upon JQ1 treatment across experiments A and B. Fold change was calculated by comparing each JQ1-treated sample with the mean of the three corresponding vehicle-treated samples from the same experiment. p-value calculated by one-sample t-test vs a theoretical mean of 0. Data represents mean ± SEM. (h and i) Correlation of JQ1-response with baseline clonal transcriptomic signatures. Clonal gene set enrichment scores for vehicle-treated tumours were calculated by comparing cells of a specific clonal lineage of interest to all assigned tumour cells within the same experiment. Correlation between these scores and JQ1-treatment response (mean log2 fold change clonal proportion JQ1 vs vehicle) was then calculated for each gene set. Selected gene sets with the highest positive or negative correlation values (Pearson correlation test) are shown. A positive correlation indicates a higher expression in resistant clones, whereas a negative correlation indicates a higher expression in sensitive clones. Resistant clonal lineages identified by barcodes 93, 264, and 439 were combined for the purpose of this analysis to have enough cells for analysis within the vehicle-treated samples. We first explored whether JQ1 had any effect on the tumour microenvironment. The most striking difference we observed was a change in abundance among the cells belonging to the T-cell compartment. To analyze this further, we computationally extracted these cells from the single cell data, reclustered them and performed differential abundance testing using Milo (Figure 2c). Milo detects sets of cells that are differentially abundant between conditions by modeling counts of cells in neighborhoods of a KNN graph (Dann et al., 2022). When applied to our reclustered T-cells, Milo identified a significant decrease in abundance in cytotoxic T-cells, as identified by their expression of Cd8a and Cd8b1, following JQ1 treatment. A significant change was observed in both of our experiments although the magnitude of the effect was greater in experiment A (Figure 2c). We next examined the effect of JQ1 treatment on the transcriptome of the tumour cells. Differential expression analysis was performed for each clonal lineage and experiment independently. As expected, given its mode of action, we identified significant down-regulation of a wide range of genes with consistent changes across clonal lineages (Figure 2d, Supplementary file 6). Among the repressed genes, were a number of genes related to interferon (IFN) signaling and antigen processing and presentation (Figure 2d and e), including Gbp2 which is strongly induced by IFN gamma, the MHC class II protein, Cd74, and B2m, a component of the MHC class I complex. JQ1 has previously been reported to directly inhibit transcription of IFN-response genes (Gibbons et al., 2019; Gusyatiner et al., 2021) suggesting this may be due to the direct action of JQ1 within our tumour cells; however, JQ1-dependent changes to the tumour microenvironment may also influence these expression pathways. Our barcoded 4T1 clones showed varied sensitivity to JQ1, with treatment causing reproducible changes to clonal proportions within the tumour (Figure 2f and g, Figure 2—figure supplement 1a and Supplementary file 2). Figure 2f shows the proportions of clones in vehicle or JQ1 treated tumours from a representative experiment. We classified one of the most abundant clones, clone 473, as highly sensitive to JQ1 treatment and three clones as being resistant to JQ1 treatment, clones 93, 439, and 264 based on these clones showing consistent behavior across two independent experiments (see Figure 2—figure supplement 1a for a side-by-side comparison of experiments). These resistant clones which together make up less than 5% of the tumour in vehicle treated mice constitute on average 12.8% of the JQ1-treated tumours. To correlate the baseline transcriptomic signatures of clones with JQ1-sensitivity and resistance, we derived a log2 fold change value for each clone using data from two independent experiments, each consisting of 3 vehicle and 3 JQ1 treated animals (Figure 2g). We used these fold change values as a measure of JQ1 response to investigate transcriptomic characteristics that are correlated with sensitivity and resistance, an approach that obviates the need for binary classification of clones as sensitive and resistant and takes into account all the available data. We identified gene sets whose expression in vehicle-treated tumours was highly correlated with response (Figure 2h and i, Supplementary file 7). Interestingly, interferon signaling which is significantly attenuated in our JQ1-treated tumours is highly correlated with sensitivity to JQ1, suggesting a possible higher dependence of the sensitive clones on these pathways. Conversely resistance is associated with higher levels of unfolded protein response (UPR), in particular the IRE1 branch of the UPR (Figure 2h and i and Figure 2—figure supplement 1b), and mTOR signaling consistent with a known role of mTOR-mediated autophagy in resistance to JQ1 (Luan et al., 2019), and cytotoxic synergy between PI3K/mTOR inhibitors and BET inhibitors (Lee et al., 2015; Stratikopoulos et al., 2015). Clonal transcriptomic correlates of response and resistance to taxane chemotherapy in the 4T1 mammary carcinoma model Our studies with JQ1 exemplify the ability of the WILD-seq system to simultaneously measure in vivo the effect of therapeutic intervention on clonal dynamics, gene expression and the tumour microenvironment. However, we were interested in using our system to investigate a chemotherapeutic agent currently in use in the clinic. We therefore treated our 4T1 WILD-seq tumour-bearing mice with docetaxel as a representative taxane, a class of drugs which are routinely used to treat triple negative breast cancer patients. As with JQ1, docetaxel treatment resulted in an initial, modest reduction in tumour growth followed by relapse (Figure 3a). Comparison of vehicle and docetaxel (DTX) treated tumours revealed differential response of clonal lineages to treatment (Figure 3b, c and d, Figure 3—figure supplement 1 and Supplementary file 2). Figure 3c shows the proportions of clones in vehicle or docetaxel treated tumours from a representative experiment. We classified clone 679 as docetaxel resistant and clone 238 as docetaxel sensitive based on these clones showing consistent behaviour across independent experiments (see Figure 3—figure supplement 1a for side-by-side comparison of experiments). Figure 3 with 1 supplement see all Download asset Open asse" @default.
• W4313013559 created "2023-01-05" @default.
• W4313013559 creator A5069901066 @default.
• W4313013559 creator A5089833949 @default.
• W4313013559 date "2022-08-17" @default.
• W4313013559 modified "2023-09-26" @default.
• W4313013559 title "Decision letter: Clonal transcriptomics identifies mechanisms of chemoresistance and empowers rational design of combination therapies" @default.
• W4313013559 doi "https://doi.org/10.7554/elife.80981.sa1" @default.
• W4313013559 hasPublicationYear "2022" @default.
• W4313013559 type Work @default.
• W4313013559 citedByCount "0" @default.
• W4313013559 crossrefType "peer-review" @default.
• W4313013559 hasAuthorship W4313013559A5069901066 @default.
• W4313013559 hasAuthorship W4313013559A5089833949 @default.
• W4313013559 hasBestOaLocation W43130135591 @default.
• W4313013559 hasConcept C104317684 @default.
• W4313013559 hasConcept C150194340 @default.
• W4313013559 hasConcept C162317418 @default.
• W4313013559 hasConcept C54355233 @default.
• W4313013559 hasConcept C70721500 @default.
• W4313013559 hasConcept C86803240 @default.
• W4313013559 hasConceptScore W4313013559C104317684 @default.
• W4313013559 hasConceptScore W4313013559C150194340 @default.
• W4313013559 hasConceptScore W4313013559C162317418 @default.
• W4313013559 hasConceptScore W4313013559C54355233 @default.
• W4313013559 hasConceptScore W4313013559C70721500 @default.
• W4313013559 hasConceptScore W4313013559C86803240 @default.
• W4313013559 hasLocation W43130135591 @default.
• W4313013559 hasOpenAccess W4313013559 @default.
• W4313013559 hasPrimaryLocation W43130135591 @default.
• W4313013559 hasRelatedWork W2027663190 @default.
• W4313013559 hasRelatedWork W2919381586 @default.
• W4313013559 hasRelatedWork W2935473914 @default.
• W4313013559 hasRelatedWork W3000738194 @default.
• W4313013559 hasRelatedWork W3025961873 @default.
• W4313013559 hasRelatedWork W3037359818 @default.
• W4313013559 hasRelatedWork W3092201827 @default.
• W4313013559 hasRelatedWork W3156737041 @default.
• W4313013559 hasRelatedWork W4286500138 @default.
• W4313013559 hasRelatedWork W4386797301 @default.
• W4313013559 isParatext "false" @default.
• W4313013559 isRetracted "false" @default.
• W4313013559 workType "peer-review" @default.