Sara P. Garcia

Robust lineage reconstruction from high-dimensional single-cell data

Single-cell gene expression data provide invaluable resources for systematic characterization of cellular hierarchy in multi-cellular organisms. However, cell lineage reconstruction is still often associated with significant uncertainty due to technological constraints. Such uncertainties have not been taken into account in current methods. We present ECLAIR (Ensemble Cell Lineage Analysis with Improved Robustness), a novel computational method for the statistical inference of cell lineage relationships from single-cell gene expression data. ECLAIR uses an ensemble approach to improve the robustness of lineage predictions, and provides a quantitative estimate of the uncertainty of lineage branchings. We show that the application of ECLAIR to published datasets successfully reconstructs known lineage relationships and significantly improves the robustness of predictions. ECLAIR is a powerful bioinformatics tool for single-cell data analysis. It can be used for robust lineage reconstruction with quantitative estimate of prediction accuracy.

In vivo CRISPR-Cas gene editing with no detectable genome-wide off-target mutations

CRISPR-Cas genome-editing nucleases hold substantial promise for human therapeutics1–5 but identifying unwanted off-target mutations remains an important requirement for clinical translation6, 7. For ex vivo therapeutic applications, previously published cell-based genome-wide methods provide potentially useful strategies to identify and quantify these off-target mutation sites8–12. However, a well-validated method that can reliably identify off-targets in vivo has not been described to date, leaving the question of whether and how frequently these types of mutations occur. Here we describe Verification of In Vivo Off-targets (VIVO), a highly sensitive, unbiased, and generalizable strategy that we show can robustly identify genome-wide CRISPR-Cas nuclease off-target effects in vivo. To our knowledge, these studies provide the first demonstration that CRISPR-Cas nucleases can induce substantial off-target mutations in vivo, a result we obtained using a deliberately promiscuous guide RNA (gRNA). More importantly, we used VIVO to show that appropriately designed gRNAs can direct efficient in vivo editing without inducing detectable off-target mutations. Our findings provide strong support for and should encourage further development of in vivo genome editing therapeutic strategies.

Cell of origin dictates aggression and stem cell number in acute lymphoblastic leukemia

Acute lymphoblastic leukemia (ALL) comprises a highly heterogeneous set of diseases that are defined by their cells of origin, stage of maturation arrest, and the underlying oncogenic driver pathways. Subclassification of lymphoid neoplasms is often based on the presumed cell of origin based on T and B progenitor gene expression. In general, TALL has a worse prognosis and requires more intensive therapies to achieve similar remission rates as their B-ALL counterparts [1, 2]. Given the common origin of these cells from committed lymphoid progenitors, investigators have suggested that the differences in clinical presentation of TALL and B-ALL are likely accounted for by the underlying differences in the oncogenic drivers of transformation. For example, T-ALLs often express oncogenic transcription factors and acquire activating NOTCH1 mutations that elevate MYC activity to drive tumor initiation and maintenance [3, 4]. By contrast, B-ALLs are more frequently driven by chromosomal rearrangements that create fusion oncogenes and have recurrent chromosomal aberrations [5]. Despite these differences, MYC is also commonly activated in human B-ALL [6] and drives leukemia initiation and growth in mouse models [7], suggesting important roles for this pathway in both T-ALL and B-ALL. It is clear that underlying oncogenic drivers exert important roles in regulating the functional characteristics of leukemia cells, including regulating the overall fraction of leukemia stem cells (LSCs) that drive continued tumor growth and relapse following therapy resistance. Yet, to date, the effect of cell lineage on influencing LSC number and self-renewal is largely unknown, accounted for in part, by lack of experimental models to address these questions. Here, we have used a zebrafish transgenic model of Mycinduced ALL to investigate the roles of cell lineage on modulating growth, aggression, and LSC frequency in T, B, and biphenotypic ALL. Previously, the rag2-Myc transgenic zebrafish model has been exploited to develop robust T-ALL models when introduced into AB strain zebrafish [8]. Moreover, findings in this model have led to significant new insights into human T-ALL [9, 10]. Using this same transgenic approach, we had previously generated leukemias in syngeneic CG1 strain zebrafish and performed large-scale cell transplantation assays to assess latency and LSC frequency differences between intra-tumoral and intertumoral clones [9]. These experiments required implanting single LSCs into large cohorts of transplant animals, thus creating monoclonal tumors (schematic shown in Fig. 1a). Each monoclonal leukemia was then assessed for latency of regrowth and LSC frequency using large-scale limiting dilution cell transplantation [9]. From these previously published studies, we uncovered that clonal heterogeneity is common in T-ALL, with individual leukemia clones exhibiting striking differences in latency, aggression, and LSC frequency, and also identified the PI3-kinase/AKT pathway as a driver of elevated growth and LSC frequency in both zebrafish and human disease [9]. Using this previously generated library of well-annotated monoclonal ALLs defined by Blackburn et al. [9], we selected leukemias that harbored high (>1%) and low numbers of LSCs (<1%, Supplemental Table 1) [9]. These leukemias are reported using the same nomenclature from Blackburn et al [9] and were subjected to bulk RNA sequencing. Principal Component Analysis (PCA) was then performed to identify molecular differences between clones. Principal Component 1 (PC1) and PC2 represent the top two gene expression profiles derived as dimensions from Principal Component Analysis. This analysis These authors contributed equally: Sowmya Iyer, Sara P. Garcia.

tp53 deficiency causes a wide tumor spectrum and increases embryonal rhabdomyosarcoma metastasis in zebrafish

The TP53 tumor-suppressor gene is mutated in >50% of human tumors and Li-Fraumeni patients with germ line inactivation are predisposed to developing cancer. Here, we generated tp53 deleted zebrafish that spontaneously develop malignant peripheral nerve-sheath tumors, angiosarcomas, germ cell tumors, and an aggressive Natural Killer cell-like leukemia for which no animal model has been developed. Because the tp53 deletion was generated in syngeneic zebrafish, engraftment of fluorescent-labeled tumors could be dynamically visualized over time. Importantly, engrafted tumors shared gene expression signatures with predicted cells of origin in human tissue. Finally, we showed that tp53del/del enhanced invasion and metastasis in kRASG12D-induced embryonal rhabdomyosarcoma (ERMS), but did not alter the overall frequency of cancer stem cells, suggesting novel pro-metastatic roles for TP53 loss-of-function in human muscle tumors. In summary, we have developed a Li-Fraumeni zebrafish model that is amenable to large-scale transplantation and direct visualization of tumor growth in live animals.

Molecularly distinct models of zebrafish Myc-induced B cell leukemia

Zebrafish models of T cell acute lymphoblastic leukemia (T-ALL) have been studied for over a decade, but curiously, robust zebrafish B cell ALL (B-ALL) models had not been described. Recently, our laboratories reported two seemingly closely-related models of zebrafish B-ALL. In these genetic lines, the primary difference is expression of either murine or human transgenic c-MYC, each controlled by the zebrafish rag2 promoter. Here, we compare ALL gene expression in both models. Surprisingly, we find that B-ALL arise in different B cell lineages, with ighm+ vs. ighz+ B-ALL driven by murine Myc vs. human MYC, respectively. Moreover, these B-ALL types exhibit signatures of distinct molecular pathways, further unexpected dissimilarity. Thus, despite sharing analogous genetic makeup, the ALL types in each model are markedly different, proving subtle genetic changes can profoundly impact model organism phenotypes. Investigating the mechanistic differences between mouse and human c-MYC in these contexts may reveal key functional aspects governing MYC-driven oncogenesis in human malignancies.

In vivo CRISPR editing with no detectable genome-wide off-target mutations

CRISPR–Cas genome-editing nucleases hold substantial promise for developing human therapeutic applications1–6 but identifying unwanted off-target mutations is important for clinical translation7. A well-validated method that can reliably identify off-targets in vivo has not been described to date, which means it is currently unclear whether and how frequently these mutations occur. Here we describe ‘verification of in vivo off-targets’ (VIVO), a highly sensitive strategy that can robustly identify the genome-wide off-target effects of CRISPR–Cas nucleases in vivo. We use VIVO and a guide RNA deliberately designed to be promiscuous to show that CRISPR–Cas nucleases can induce substantial off-target mutations in mouse livers in vivo. More importantly, we also use VIVO to show that appropriately designed guide RNAs can direct efficient in vivo editing in mouse livers with no detectable off-target mutations. VIVO provides a general strategy for defining and quantifying the off-target effects of gene-editing nucleases in whole organisms, thereby providing a blueprint to foster the development of therapeutic strategies that use in vivo gene editing.A strategy developed to define off-target effects of gene-editing nucleases in whole organisms is validated and leveraged to show that CRISPR–Cas9 nucleases can be used effectively in vivo without inducing detectable off-target mutations.

Abstract 26: RUNX1 as a transcriptional target of activated Shp2 (PTPN11) in juvenile myelomonocytic leukemia

Juvenile myelomonocytic leukemia (JMML) is an aggressive myeloproliferative neoplasm of young children. The only current curative treatment is bone marrow transplantation. Yet even with this aggressive therapy, ~50% of children still die from their disease. Somatic mutations leading to constitutive activation of the non-receptor tyrosine phosphatase Shp2 (also called PTPN11) or of RAS signaling occur in >70% of JMML cases. However, the transcription factors that act downstream of these aberrant signaling events have not been identified. We recently showed that the key myelomonocytic transcription factor RUNX1 is inactivated by src-family kinase-mediated tyrosine phosphorylation (Huang et al. Genes Dev 2012;26:1587-1601). Moreover, we demonstrated that Shp2 directly dephosphorylates RUNX1 leading to RUNX1 relative activation. We now show that overexpression of a mutant RUNX1 that is resistant to SFK-mediated tyrosine phosphorylation (RUNX1Y260F, Y375F, Y378F, Y379F, Y386F—“RUNX1-5F”), mimicking constitutive dephosphorylation, causes cytokine-independent growth of Ba/F3 cells. Overexpression of RUNX15F in murine Lin- Sca-1+ c-kit+, (LSK) bone marrow cells leads to a dramatic expansion of Gr1+, Mac-1+ cells, and CFU-M/CFU-GM in vitro and in vivo. These effects are not seen when wild-type RUNX1 or RUNX1Y260D, Y375D, Y378D, Y379FD,Y386FD (“RUNX1-5D”; mimicking constitutive RUNX1 tyrosine phosphorylation) are overexpressed. The RUNX1-5F expressing cells have increased proliferation (BrdU incorporation), decreased apoptosis, and reduced cytokine dependence. Flow sorted Gr1+Mac1+ cells from the RUNX1-5F transduced cultures express higher levels of the direct RUNX1 target genes PU.1 and cyclin D1. RUNX1 haploinsufficiency in a Shp2 LSL-D61Y conditional knock-in JMML mouse model reduces the disease phenotype, and treatment with the RUNX1 inhibitor Ro5-3335 ameliorates cytokine-independent in vitro bone marrow colony growth from this mouse model. Gene expression analysis of CD34+ cells from JMML patients versus controls combined with RUNX1 CD34+ cell ChIP-seq data shows significant enrichment for RUNX1 chromatin occupancy at differentially expressed genes compared to all expressed genes. To test whether RUNX1 is required for the myelomonocytic hyperproliferation in JMML, CD34+ peripheral blood cells from a patient with JMML due to an activating Shp2 mutation (Shp2E78G) were lentivirally transduced with doxycylcine-inducible RUNX1-5D or RUNX1-5F expression constructs and cultured under myeloid growth conditions. Upon doxycycline induction, the RUNX1-5D overexpressing cells (resistant to Shp2) exhibited at 32% reduction in BrdU incorporation. In contrast, the control RUNX1-5F expressing cells had no significant reduction in proliferation. RUNX1 is also by direct ERK mediated phosphorylation (which is downstream of RAS/MEK). We propose that RUNX1 activation is an important common downstream consequence of both activated Shp2 and RAS signaling in the pathogenesis of JMML, and that RUNX1 inhibition may represent a novel therapeutic approach for this disorder. Citation Format: Elisa Dorantes-Acosta, Hui Huang, Sara P. Garcia, Elliot Stieglitz, Mignon Loh, Guo-Cheng Yuan, Alan B. Cantor. RUNX1 as a transcriptional target of activated Shp2 (PTPN11) in juvenile myelomonocytic leukemia [abstract]. In: Proceedings of the Second AACR Conference on Hematologic Malignancies: Translating Discoveries to Novel Therapies; May 6-9, 2017; Boston, MA. Philadelphia (PA): AACR; Clin Cancer Res 2017;23(24_Suppl):Abstract nr 26.