X Shirley Liu

Harvard Medical School, Boston, Massachusetts, United States

Are you X Shirley Liu?

Claim your profile

Publications (127)1631.21 Total impact

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Graphical Abstract Highlights d LSD1 is broadly associated with AR-regulated enhancers, and a subset is H3T6ph marked d LSD1 is associated with CoREST at these sites and interacts with FOXA1 d Despite coactivator function at these sites, LSD1 mediates their H3K4 demethylation d LSD1 coactivation is through demethylation of novel histone or nonhistone substrates In Brief Cai et al. show that lysine-specific demethylase 1 (LSD1), although generally a transcriptional corepressor through H3K4 demethylation, functions broadly as a coactivator for androgen receptor and interacts with FOXA1 on androgen-stimulated genes. LSD1-mediated H3K4 demethylation persists at these sites, indicating a distinct coactivator function mediated by demethylation of other substrates.
    12/2014; 9:1-10.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Lysine-Specific Demethylase 1 (LSD1, KDM1A) functions as a transcriptional corepressor through demethylation of histone 3 lysine 4 (H3K4) but has a coactivator function on some genes through mechanisms that are unclear. We show that LSD1, interacting with CoREST, associates with and coactivates androgen receptor (AR) on a large fraction of androgen-stimulated genes. A subset of these AR/LSD1-associated enhancer sites have histone 3 threonine 6 phosphorylation (H3T6ph), and these sites are further enriched for androgen-stimulated genes. Significantly, despite its coactivator activity, LSD1 still mediates H3K4me2 demethylation at these androgen-stimulated enhancers. FOXA1 is also associated with LSD1 at AR-regulated enhancer sites, and a FOXA1 interaction with LSD1 enhances binding of both proteins at these sites. These findings show that LSD1 functions broadly as a regulator of AR function, that it maintains a transcriptional repression function at AR-regulated enhancers through H3K4 demethylation, and that it has a distinct AR-linked coactivator function mediated by demethylation of other substrates. Copyright © 2014 The Authors. Published by Elsevier Inc. All rights reserved.
    Cell reports. 12/2014;
  • [Show abstract] [Hide abstract]
    ABSTRACT: Activation-induced cytidine deaminase (AID) initiates both somatic hypermutation (SHM) for antibody affinity maturation and DNA breakage for antibody class switch recombination (CSR) via transcription-dependent cytidine deamination of single-stranded DNA targets. Though largely specific for immunoglobulin genes, AID also acts on a limited set of off-targets, generating oncogenic translocations and mutations that contribute to B cell lymphoma. How AID is recruited to off-targets has been a long-standing mystery. Based on deep GRO-seq studies of mouse and human B lineage cells activated for CSR or SHM, we report that most robust AID off-target translocations occur within highly focal regions of target genes in which sense and antisense transcription converge. Moreover, we found that such AID-targeting "convergent" transcription arises from antisense transcription that emanates from super-enhancers within sense transcribed gene bodies. Our findings provide an explanation for AID off-targeting to a small subset of mostly lineage-specific genes in activated B cells. Copyright © 2014 Elsevier Inc. All rights reserved.
    Cell. 12/2014;
  • [Show abstract] [Hide abstract]
    ABSTRACT: Notch signaling has pleiotropic context-specific functions that have essential roles in many processes, including embryonic development and maintenance and homeostasis of adult tissues. Aberrant Notch signaling (both hyper- and hypoactive) is implicated in a number of human developmental disorders and many cancers. Notch receptor signaling is mediated by tightly regulated proteolytic cleavages that lead to the assembly of a nuclear Notch transcription complex, which drives the expression of downstream target genes and thereby executes Notch's functions. Thus, understanding regulation of gene expression by Notch is central to deciphering how Notch carries out its many activities. Here, we summarize the recent findings pertaining to the complex interplay between the Notch transcriptional complex and interacting factors involved in transcriptional regulation, including co-activators, cooperating transcription factors, and chromatin regulators, and discuss emerging data pertaining to the role of Notch-regulated noncoding RNAs in transcription. This article is protected by copyright. All rights reserved
    Journal of Cellular Physiology 11/2014; · 3.87 Impact Factor
  • Clifford A Meyer, X Shirley Liu
    [Show abstract] [Hide abstract]
    ABSTRACT: Next-generation sequencing (NGS) technologies have been used in diverse ways to investigate various aspects of chromatin biology by identifying genomic loci that are bound by transcription factors, occupied by nucleosomes or accessible to nuclease cleavage, or loci that physically interact with remote genomic loci. However, reaching sound biological conclusions from such NGS enrichment profiles requires many potential biases to be taken into account. In this Review, we discuss common ways in which biases may be introduced into NGS chromatin profiling data, approaches to diagnose these biases and analytical techniques to mitigate their effect.
    Nature Reviews Genetics 09/2014; · 39.79 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Tissues may adopt diverse strategies to establish specific transcriptional programs in daughter lineages. In intestinal crypts, enhancers for genes expressed in both major cell types appear broadly permissive in stem and specified progenitor cells. In blood, another self-renewing tissue, it is unclear when chromatin becomes permissive for transcription of genes expressed in distinct terminal lineages. Using chromatin immunoprecipitation (ChIP) combined with deep sequencing (ChIP-seq) to profile activating histone marks, we studied enhancer dynamics in primary mouse blood stem, progenitor, and specified cells. Stem and multipotent progenitor cells show scant H3K4me2 marking at enhancers bound by specific transcription factors in their committed progeny. Rather, enhancers are modulated dynamically and serially, with substantial loss and gain of H3K4me2, at each cellular transition. Quantitative analysis of these dynamics accurately modeled hematopoiesis according to Waddington's notion of epigenotypes. Delineation of enhancers in terminal blood lineages coincides with cell specification, and enhancers active in single lineages show well-positioned H3K4me2- and H3K27ac-marked nucleosomes and DNaseI hypersensitivity in other cell types, revealing limited lineage fidelity. These findings demonstrate that enhancer chronology in blood cells differs markedly from that in intestinal crypts. Chromatin dynamics in hematopoiesis provide a useful foundation to consider classical observations such as cellular reprogramming and multilineage locus priming.
    Genes & Development 08/2014; 28(16):1827-39. · 12.64 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Next generation sequencing was used to identify Notch mutations in a large collection of diverse solid tumors. NOTCH1 and NOTCH2 rearrangements leading to constitutive receptor activation were confined to triple negative breast cancers (TNBC, 6 of 66 tumors). TNBC cell lines with NOTCH1 rearrangements associated with high levels of activated NOTCH1 (N1-ICD) were sensitive to the gamma-secretase inhibitor (GSI) MRK-003, both alone and in combination with pacitaxel, in vitro and in vivo, whereas cell lines with NOTCH2 rearrangements were resistant to GSI. Immunohistochemical staining of N1-ICD in TNBC xenografts correlated with responsiveness, and expression levels of the direct Notch target gene HES4 correlated with outcome in TNBC patients. Activating NOTCH1 point mutations were also identified in other solid tumors, including adenoid cystic carcinoma (ACC). Notably, ACC primary tumor xenografts with activating NOTCH1 mutations and high N1-ICD levels were sensitive to GSI, whereas N1-ICD-low tumors without NOTCH1 mutations were resistant. Significance: NOTCH1 mutations, immunohistochemical staining for activated NOTCH1, and HES4 expression are biomarkers that can be used to identify solid tumors that are likely to respond to gamma-secretase inhibitor-based therapies.
    Cancer Discovery 08/2014; · 15.93 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Recurrent mutations in histone-modifying enzymes imply key roles in tumorigenesis, yet their functional relevance is largely unknown. Here, we show that JARID1B, encoding a histone H3 lysine 4 (H3K4) demethylase, is frequently amplified and overexpressed in luminal breast tumors and a somatic mutation in a basal-like breast cancer results in the gain of unique chromatin binding and luminal expression and splicing patterns. Downregulation of JARID1B in luminal cells induces basal genes expression and growth arrest, which is rescued by TGFβ pathway inhibitors. Integrated JARID1B chromatin binding, H3K4 methylation, and expression profiles suggest a key function for JARID1B in luminal cell-specific expression programs. High luminal JARID1B activity is associated with poor outcome in patients with hormone receptor-positive breast tumors.
    Cancer cell. 06/2014; 25(6):762-777.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Chromatin regulators play an important role in the development of human diseases. In this study, we focused on Plant Homeo Domain Finger protein 8 (PHF8), a chromatin regulator that has attracted special concern recently. PHF8 is a histone lysine demethylase ubiquitously expressed in nuclei. Mutations of PHF8 are associated with X-linked mental retardation. It usually functions as a transcriptional co-activator by associating with H3K4me3 and RNA polymerase II. We found that PHF8 may associate with another regulator, REST/NRSF, predominately at promoter regions via studying several published PHF8 chromatin immunoprecipitation-sequencing (ChIP-Seq) datasets. Our analysis suggested that PHF8 not only activates but may also repress gene expression.
    Scientific Reports 05/2014; 4:5008. · 5.08 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Human neurons are functional over an entire lifetime, yet the mechanisms that preserve function and protect against neurodegeneration during ageing are unknown. Here we show that induction of the repressor element 1-silencing transcription factor (REST; also known as neuron-restrictive silencer factor, NRSF) is a universal feature of normal ageing in human cortical and hippocampal neurons. REST is lost, however, in mild cognitive impairment and Alzheimer's disease. Chromatin immunoprecipitation with deep sequencing and expression analysis show that REST represses genes that promote cell death and Alzheimer's disease pathology, and induces the expression of stress response genes. Moreover, REST potently protects neurons from oxidative stress and amyloid β-protein toxicity, and conditional deletion of REST in the mouse brain leads to age-related neurodegeneration. A functional orthologue of REST, Caenorhabditis elegans SPR-4, also protects against oxidative stress and amyloid β-protein toxicity. During normal ageing, REST is induced in part by cell non-autonomous Wnt signalling. However, in Alzheimer's disease, frontotemporal dementia and dementia with Lewy bodies, REST is lost from the nucleus and appears in autophagosomes together with pathological misfolded proteins. Finally, REST levels during ageing are closely correlated with cognitive preservation and longevity. Thus, the activation state of REST may distinguish neuroprotection from neurodegeneration in the ageing brain.
    Nature 03/2014; 507(7493):448-54. · 42.35 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Cancer cells induce a set of adaptive response pathways to survive in the face of stressors due to inadequate vascularization. One such adaptive pathway is the unfolded protein (UPR) or endoplasmic reticulum (ER) stress response mediated in part by the ER-localized transmembrane sensor IRE1 (ref. 2) and its substrate XBP1 (ref. 3). Previous studies report UPR activation in various human tumours, but the role of XBP1 in cancer progression in mammary epithelial cells is largely unknown. Triple-negative breast cancer (TNBC)-a form of breast cancer in which tumour cells do not express the genes for oestrogen receptor, progesterone receptor and HER2 (also called ERBB2 or NEU)-is a highly aggressive malignancy with limited treatment options. Here we report that XBP1 is activated in TNBC and has a pivotal role in the tumorigenicity and progression of this human breast cancer subtype. In breast cancer cell line models, depletion of XBP1 inhibited tumour growth and tumour relapse and reduced the CD44(high)CD24(low) population. Hypoxia-inducing factor 1α (HIF1α) is known to be hyperactivated in TNBCs. Genome-wide mapping of the XBP1 transcriptional regulatory network revealed that XBP1 drives TNBC tumorigenicity by assembling a transcriptional complex with HIF1α that regulates the expression of HIF1α targets via the recruitment of RNA polymerase II. Analysis of independent cohorts of patients with TNBC revealed a specific XBP1 gene expression signature that was highly correlated with HIF1α and hypoxia-driven signatures and that strongly associated with poor prognosis. Our findings reveal a key function for the XBP1 branch of the UPR in TNBC and indicate that targeting this pathway may offer alternative treatment strategies for this aggressive subtype of breast cancer.
    Nature 03/2014; · 42.35 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The transcription factor SOX2 is an essential regulator of pluripotent stem cells and promotes development and maintenance of squamous epithelia. We previously reported that SOX2 is an oncogene and subject to highly recurrent genomic amplification in squamous cell carcinomas (SCCs). Here, we have further characterized the function of SOX2 in SCC. Using ChIP-seq analysis, we compared SOX2-regulated gene profiles in multiple SCC cell lines to ES cell profiles and determined that SOX2 binds to distinct genomic loci in SCCs. In SCCs, SOX2 preferentially interacts with the transcription factor p63, as opposed to the transcription factor OCT4, which is the preferred SOX2 binding partner in ES cells. SOX2 and p63 exhibited overlapping genomic occupancy at a large number of loci in SCCs; however, coordinate binding of SOX2 and p63 was absent in ES cells. We further demonstrated that SOX2 and p63 jointly regulate gene expression, including the oncogene ETV4, which was essential for SOX2-amplified SCC cell survival. Together, these findings demonstrate that the action of SOX2 in SCC differs substantially from its role in pluripotency. The identification of the SCC-associated interaction between SOX2 and p63 will enable deeper characterization the downstream targets of this interaction in SCC and normal squamous epithelial physiology.
    The Journal of clinical investigation 03/2014; · 15.39 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The main oncogenic driver in T-lymphoblastic leukemia is NOTCH1, which activates genes by forming chromatin-associated Notch transcription complexes. Gamma-secretase-inhibitor treatment prevents NOTCH1 nuclear localization, but most genes with NOTCH1-binding sites are insensitive to gamma-secretase inhibitors. Here, we demonstrate that fewer than 10% of NOTCH1-binding sites show dynamic changes in NOTCH1 occupancy when T-lymphoblastic leukemia cells are toggled between the Notch-on and -off states with gamma-secretase inhibiters. Dynamic NOTCH1 sites are functional, being highly associated with Notch target genes, are located mainly in distal enhancers, and frequently overlap with RUNX1 binding. In line with the latter association, we show that expression of IL7R, a gene with key roles in normal T-cell development and in T-lymphoblastic leukemia, is coordinately regulated by Runx factors and dynamic NOTCH1 binding to distal enhancers. Like IL7R, most Notch target genes and associated dynamic NOTCH1-binding sites cooccupy chromatin domains defined by constitutive binding of CCCTC binding factor, which appears to restrict the regulatory potential of dynamic NOTCH1 sites. More remarkably, the majority of dynamic NOTCH1 sites lie in superenhancers, distal elements with exceptionally broad and high levels of H3K27ac. Changes in Notch occupancy produces dynamic alterations in H3K27ac levels across the entire breadth of superenhancers and in the promoters of Notch target genes. These findings link regulation of superenhancer function to NOTCH1, a master regulatory factor and potent oncoprotein in the context of immature T cells, and delineate a generally applicable roadmap for identifying functional Notch sites in cellular genomes.
    Proceedings of the National Academy of Sciences 12/2013; · 9.81 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Sequencing of DNase I hypersensitive sites (DNase-seq) is a powerful technique for identifying cis-regulatory elements across the genome. We studied the key experimental parameters to optimize performance of DNase-seq. Sequencing short fragments of 50-100 base pairs (bp) that accumulate in long internucleosome linker regions was more efficient for identifying transcription factor binding sites compared to sequencing longer fragments. We also assessed the potential of DNase-seq to predict transcription factor occupancy via generation of nucleotide-resolution transcription factor footprints. In modeling the sequence-specific DNase I cutting bias, we found a strong effect that varied over more than two orders of magnitude. This indicates that the nucleotide-resolution cleavage patterns at many transcription factor binding sites are derived from intrinsic DNase I cleavage bias rather than from specific protein-DNA interactions. In contrast, quantitative comparison of DNase I hypersensitivity between states can predict transcription factor occupancy associated with particular biological perturbations.
    Nature Methods 12/2013; · 23.57 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Transcription factor activity and turnover are functionally linked, but the global patterns by which DNA-bound regulators are eliminated remain poorly understood. We established an assay to define the chromosomal location of DNA-associated proteins that are slated for degradation by the ubiquitin-proteasome system. The genome-wide map described here ties proteolysis in mammalian cells to active enhancers and to promoters of specific gene families. Nuclear-encoded mitochondrial genes in particular correlate with protein elimination, which positively affects their transcription. We show that the nuclear receptor corepressor NCoR1 is a key target of proteolysis and physically interacts with the transcription factor CREB. Proteasome inhibition stabilizes NCoR1 in a site-specific manner and restrains mitochondrial activity by repressing CREB-sensitive genes. In conclusion, this functional map of nuclear proteolysis links chromatin architecture with local protein stability and identifies proteolytic derepression as highly dynamic in regulating the transcription of genes involved in energy metabolism.
    Cell 12/2013; 155(6):1380-95. · 31.96 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The combination of ChIP-seq and transcriptome analysis is a compelling approach to unravel the regulation of gene expression. Several recently published methods combine transcription factor (TF) binding and gene expression for target prediction, but few of them provide an efficient software package for the community. Binding and expression target analysis (BETA) is a software package that integrates ChIP-seq of TFs or chromatin regulators with differential gene expression data to infer direct target genes. BETA has three functions: (i) to predict whether the factor has activating or repressive function; (ii) to infer the factor's target genes; and (iii) to identify the motif of the factor and its collaborators, which might modulate the factor's activating or repressive function. Here we describe the implementation and features of BETA to demonstrate its application to several data sets. BETA requires ∼1 GB of RAM, and the procedure takes 20 min to complete. BETA is available open source at http://cistrome.org/BETA/.
    Nature Protocol 12/2013; 8(12):2502-2515. · 8.36 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Diversified histone modifications (HMs) are essential epigenetic features. They play important roles in fundamental biological processes including transcription, DNA repair and DNA replication. Chromatin regulators (CRs), which are indispensable in epigenetics, can mediate HMs to adjust chromatin structures and functions. With the development of ChIP-Seq technology, there is an opportunity to study CR and HM profiles at the whole-genome scale. However, no specific resource for the integration of CR ChIP-Seq data or CR-HM ChIP-Seq linkage pairs is currently available. Therefore, we constructed the CR Cistrome database, available online at http://compbio.tongji.edu.cn/cr and http://cistrome.org/cr/, to further elucidate CR functions and CR-HM linkages. Within this database, we collected all publicly available ChIP-Seq data on CRs in human and mouse and categorized the data into four cohorts: the reader, writer, eraser and remodeler cohorts, together with curated introductions and ChIP-Seq data analysis results. For the HM readers, writers and erasers, we provided further ChIP-Seq analysis data for the targeted HMs and schematized the relationships between them. We believe CR Cistrome is a valuable resource for the epigenetics community.
    Nucleic Acids Research 11/2013; · 8.81 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The epithelial-to-mesenchymal transition is an important mechanism in cancer metastasis. Although transcription factors including SNAIL, SLUG, and TWIST1 regulate the epithelial-to-mesenchymal transition, other unknown transcription factors could also be involved. Identification of the full complement of transcription factors is essential for a more complete understanding of gene regulation in this process. Chromatin immunoprecipitation-sequencing (ChIP-Seq) technologies have been used to detect genome-wide binding of transcription factors; here, we developed a systematic approach to integrate existing ChIP-Seq and transcriptome data. We scanned multiple transcription factors to investigate their functional impact on the epithelial-to-mesenchymal transition in the human A549 lung adenocarcinoma cell line. Among the transcription factors tested, impact scores identified the forkhead box protein A1 (FOXA1) as the most significant transcription factor in the epithelial-to-mesenchymal transition. FOXA1 physically associates with the promoters of its predicted target genes. Several critical epithelial-to-mesenchymal transition effectors involved in cellular adhesion and cellular communication were identified in the regulatory network of FOXA1, including FOXA2, FGA, FGB, FGG, and FGL1. The implication of FOXA1 in the epithelial-to-mesenchymal transition via its regulatory network indicates that FOXA1 may play an important role in the initiation of lung cancer metastasis. We identified FOXA1 as a potentially important transcription factor and negative regulator in the initial stages of lung cancer metastasis. FOXA1 may modulate the epithelial-to-mesenchymal transition via its transcriptional regulatory network. Further, this study demonstrates how ChIP-Seq and expression data could be integrated to delineate the impact of transcription factors on a specific biological process.
    BMC Genomics 10/2013; 14(1):680. · 4.04 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Early full-term pregnancy is one of the most effective natural protections against breast cancer. To investigate this effect, we have characterized the global gene expression and epigenetic profiles of multiple cell types from normal breast tissue of nulliparous and parous women and carriers of BRCA1 or BRCA2 mutations. We found significant differences in CD44(+) progenitor cells, where the levels of many stem cell-related genes and pathways, including the cell-cycle regulator p27, are lower in parous women without BRCA1/BRCA2 mutations. We also noted a significant reduction in the frequency of CD44(+)p27(+) cells in parous women and showed, using explant cultures, that parity-related signaling pathways play a role in regulating the number of p27(+) cells and their proliferation. Our results suggest that pathways controlling p27(+) mammary epithelial cells and the numbers of these cells relate to breast cancer risk and can be explored for cancer risk assessment and prevention.
    Cell stem cell 06/2013; · 23.56 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Despite growing appreciation of the importance of long noncoding RNAs (lncRNAs) in normal physiology and disease, our knowledge of cancer-related lncRNAs remains limited. By repurposing microarray probes, we constructed expression profiles of 10,207 lncRNA genes in approximately 1,300 tumors over four different cancer types. Through integrative analysis of the lncRNA expression profiles with clinical outcome and somatic copy-number alterations, we identified lncRNAs that are associated with cancer subtypes and clinical prognosis and predicted those that are potential drivers of cancer progression. We validated our predictions by experimentally confirming prostate cancer cell growth dependence on two newly identified lncRNAs. Our analysis provides a resource of clinically relevant lncRNAs for the development of lncRNA biomarkers and the identification of lncRNA therapeutic targets. It also demonstrates the power of integrating publically available genomic data sets and clinical information for discovering disease-associated lncRNAs.
    Nature Structural & Molecular Biology 06/2013; · 11.63 Impact Factor

Publication Stats

10k Citations
1,631.21 Total Impact Points


  • 2005–2014
    • Harvard Medical School
      • • Department of Genetics
      • • Department of Biological Chemistry and Molecular Pharmacology
      Boston, Massachusetts, United States
  • 2004–2014
    • Dana-Farber Cancer Institute
      • • Department of Medical Oncology
      • • Division of Molecular and Cellular Oncology
      • • Department of Biostatistics and Computational Biology
      Boston, Massachusetts, United States
  • 2012
    • Tongji University
      Shanghai, Shanghai Shi, China
  • 2011–2012
    • Boston Children's Hospital
      • Department of Pathology
      Boston, MA, United States
    • The University of Arizona
      • School of Plant Sciences
      Tucson, AZ, United States
    • Beth Israel Deaconess Medical Center
      • Division of Hematology/Oncology
      Boston, MA, United States
  • 2010
    • Zhejiang University
      • State Key Lab of Diagnosis and Treatment of Infectious Diseases
      Hangzhou, Zhejiang Sheng, China
    • Johns Hopkins Bloomberg School of Public Health
      • Department of Biostatistics
      Baltimore, MD, United States
  • 2009
    • Université de Rennes 2
      Roazhon, Brittany, France
    • Compendia Bioscience
      Ann Arbor, Michigan, United States
  • 2008
    • Brigham and Women's Hospital
      • Department of Medicine
      Boston, MA, United States
    • University of British Columbia - Vancouver
      • Department of Statistics
      Vancouver, British Columbia, Canada
  • 2003–2008
    • Harvard University
      • Department of Statistics
      Cambridge, Massachusetts, United States
  • 2002–2008
    • Stanford University
      • Department of Genetics
      Stanford, CA, United States
  • 2006
    • Peking University
      • School of Life Sciences
      Beijing, Beijing Shi, China