Despite mounting evidence that p53 senses and responds to physiological cues in vivo, existing knowledge regarding p53 function and target genes is largely derived from studies in cancer or stressed cells. Herein we utilize p53 transcriptome and ChIP-Seq (chromatin immunoprecipitation-high throughput sequencing) analyses to identify p53 regulated pathways in the embryonic kidney, an organ that develops via mesenchymal-epithelial interactions. This integrated approach allowed identification of novel genes that are possible direct p53 targets during kidney development. We find the p53-regulated transcriptome in the embryonic kidney is largely composed of genes regulating developmental, morphogenesis, and metabolic pathways. Surprisingly, genes in cell cycle and apoptosis pathways account for <5% of differentially expressed transcripts. Of 7,893 p53-occupied genomic regions (peaks), the vast majority contain consensus p53 binding sites. Interestingly, 78% of p53 peaks in the developing kidney lie within proximal promoters of annotated genes compared with 7% in a representative cancer cell line; 25% of the differentially expressed p53-bound genes are present in nephron progenitors and nascent nephrons, including key transcriptional regulators, components of Fgf, Wnt, Bmp, and Notch pathways, and ciliogenesis genes. The results indicate widespread p53 binding to the genome in vivo and context-dependent differences in the p53 regulon between cancer, stress, and development. To our knowledge, this is the first comprehensive analysis of the p53 transcriptome and cistrome in a developing mammalian organ, substantiating the role of p53 as a bona fide developmental regulator. We conclude p53 targets transcriptional networks regulating nephrogenesis and cellular metabolism during kidney development.
- gene expression
- kidney development
- signaling pathways
the tumor suppressor p53 protects the organism from genotoxic insults that may result in genome instability and cancer. As a transcriptional regulator, p53 plays a key role in cell-fate determination by regulating cell cycle arrest, DNA repair, apoptosis, or senescence genes, thus limiting the propagation of cells with damaged genomes (5, 61). p53 interacts with and recruits coactivator complexes, components of the basal transcriptional machinery, and chromatin modifiers (1, 8, 31, 62). p53 also represses transcription of some genes, such as PKD1, IL6, and others via recruitment of co-repressors such as Sin3A and HDACs (49, 59, 60). Recently Leonova et al. (32) showed that p53-mediated transcription regulation, along with DNA methylation, is essential in epigenetic silencing of major classes of retro-elements and satellite DNA in mice.
These direct modes of activation or repression require p53 binding to DNA. The consensus sequence of the p53 response element (RE) is highly degenerate, consisting of two decameric half-sites of 5′-RRRCWWGYYY-3′ (R = A, G; W = A, T; Y = C, T) separated by a 0–20 bp spacer (18, 43, 63). Each decamer in turn is an inverted repeat of two pentameric quarter sites.
In addition to its classical tumor suppressor role, p53 is also implicated in pathways regulating cell migration, autophagy, and metabolism, e.g., oxidative and glycolytic pathways for energy generation and glucose homeostasis (3, 4, 6, 19, 42). These findings implicate p53 in metabolic diseases from cancer to cardiovascular dysfunction. p53 is also a modulator of cellular differentiation (38). p53 deletion impairs terminal differentiation of renal epithelia, skeletal muscle cells, various hematopoietic cell-lines, thyroid cells, oligodendrocytes, neuronal maturation, axon outgrowth, and regeneration (2, 9, 37, 45, 53, 56, 57). Thus, p53 has transcriptional functions that reach far beyond its classical role as master regulator of cell cycle or apoptosis genes. The data presented in this study further reinforce this concept.
The mammalian kidney develops via reciprocal inductive signaling between two intermediate mesoderm derivatives, the ureteric bud (UB) and the metanephric mesenchyme (MM) (14, 15, 50, 52). UB outgrowth is stimulated by glial-derived neurotrophic factor that is expressed in and released from the MM. In turn, bud-derived signals potentiate the proliferation, maintenance and differentiation of the MM. Prior to UB invasion of the MM at embryonic age 10.5 (E10.5), the Osr1+ MM differentiates into two distinct/exclusive lineages, the Six2+ cap mesenchyme (CM) positioned as a “cap” around the ureteric tips and the Foxd1+/Meis1+ outer cell layers, called the cortical stroma interstitium/mesenchyme that surrounds the CM (29). Lineage tracing studies demonstrated that the CM consists exclusively of nephron progenitor cells that will differentiate into nephrons (10, 29).
We have previously shown that p53 is expressed in both UB and MM lineages in the kidney (46) and that p53-null embryos on C57BL/6 background exhibit a range of congenital abnormalities of the kidney and urinary tract such as duplicated ureters, reduced nephron numbers, and compromised nephron progenitor renewal and differentiation (46, 48). Here we use whole transcriptome analysis coupled with genome-wide profiling to map p53 occupancy in kidney development. Integration of gene expression and chromatin immunoprecipitation-high throughput sequencing (ChIP-Seq) data identified key regulatory genes and pathways in nephrogenesis as p53 targets.
MATERIALS AND METHODS
p53+/− mice on C57BL/6 background were commercially obtained (Jax). Heterozygous pairings were done to obtain +/+ and −/− litter-matched kidneys. 500 ng RNA was obtained from each of three pairs of p53+/+ and p53−/− E15.5 kidneys, amplified, labeled and the cDNA was hybridized to Agilent 4×44K Whole Mouse Genome Microarray slide from Agilent Technologies. To prevent bias from signaling detection on different colors (Cy5 red, Cy3 green), we applied dye-swap strategy to this experiment, in which identical pairs of p53+/+ and p53−/− cDNA samples were reversely labeled by Cy5 and Cy3 into two groups. Raw data were processed by GeneSpring software (13). Only genes, showing a significant (P < 0.05) differential change in expression after Benjamini and Hochberg false discovery rate (FDR) correction in the three datasets, were used for further analyses. All animal protocols utilized in this study were approved by and in strict adherence to guidelines established by the Tulane University Institutional Animal Care and Use Committee.
Quantitative real-time PCR.
Validation of microarray data was done by RT-PCR on RNA from E15.5 p53+/+ and p53−/− kidneys, for genes indicated. Exon-spanning primer-probes for TaqMan gene expression assay from Applied Biosystems were utilized. Reactions were prepared by TaqMan RNA-to-CT 1-Step Kit and amplified by Stratagene Mx3000P and Mx3005P. Expression was normalized against endogenous GAPDH mRNA levels. Experimental and biological triplicates were applied.
ChIP was performed using an antibody against p53 (Santa Cruz, SC6243X) on chromatin prepared from E15.5 C57BL/6 mice kidneys. Input DNA (without p53 immunoprecipitation) was used as background. Samples were then prepared for sequencing using Illumina's protocol. Briefly, samples were linked to adapters; library was size-selected (200–250 bp) and PCR-amplified. Sequencing was processed by Genome Analyzer 2 (GAII); the 35-nt sequence reads (“tags”) were mapped to the mouse genome using the ELAND algorithm. Only tags that map uniquely and have no more than two mismatches were used for subsequent analysis. The number of overlaid tags was significantly identified by the model-based analysis of ChIP-Seq (MACS) peak calling program; the p53 binding at each peak region (interval) was quantified as the number of tags (tag density or peak value).
All p53-bound regions identified by ChIP-Seq were subjected to degenerate consensus site search using RegionMiner in Genomatix Software Suite v2.0 (http://www.genomatix.de). According to strict stringency of matrix similarity >80% and core similarity >75% to six different matrices, MatInspector from Genomatix Software Suite v2.0 (http://www.genomatix.de/) mapped p53 binding sequences. Four matrices are different permutations of the 20–23 bp consensus and two matrices are 11–12 bp half-sites. These sequences were further regrouped into 5′-RE and 3′-RE of p53 binding sites. Sequences of 5′-RE and 3′-REs were processed by WebLogo (16, 51) to generate their patterns, respectively.
Cell culture, transient transfection, and reporter assay.
p53-null human lung carcinoma cells H1299 (ATCC), were maintained in media supplemented with 10% fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 mg/ml) at 37°C in a humidified incubator with 5% CO2. Cells were transfected with 1.0 μg of DNA/well promoter-reporter vectors along with pCMV-p53-(wild type) or pCMV-p53-(mutant) expression plasmids (0–100 ng). A modified pGL3 basic vector was generated by removing the SmaI site from the multicloning site and used in all cloning experiments, because of a reported effect (65) that the SmaI site is able to interact with p53 and induces luciferase expression of the original pGL3 basic vector (Promega). The proximal 5′-regulatory regions plus intragenic sequences corresponding to a 3,056 bp ChIP-Seq peak region [−1530 bp ∼ transcription start site (TSS) ∼ +1525 bp, Chr16: 38087556–38090611] for mouse Gsk3b, and a 1255 bp ChIP-Seq peak region (−1236 bp ∼ TSS ∼ +18 bp, Chr4: 148681788–148683042) for mouse Kif1b were amplified by PCR and cloned upstream a luciferase reporter in the modified promoterless pGL3 Basic-Luciferase reporter (Luc) construct. The 1255 bp Kif1b promoter fragment was used to generate the following truncation fragments: 693 bp (−674 bp ∼ TSS ∼ +18 bp, Chr4: 148681788–148682480) and 380 bp (−361 bp ∼ TSS ∼ +18 bp, Chr4: 148681788–148682167). Transfection was performed using the Lipofectamine Plus reagent (Invitrogen) according to the manufacturer's recommendations. Four hours after transfection, fresh medium was replaced, and cell extracts were prepared 24 h later using a reporter lysis reagent (Promega). Aliquots of cell lysate were analyzed for luciferase activity after normalization for protein content as previously described (47).
Other published resources.
p53-deficient microarrays of E8.5 mouse embryos (24) and mouse epidermal carcinoma (20), GUDMAP (21, 36), microarray for renal developmental compartments (11), and p53 ChIP-Seq on growth arrest/apoptosis-induced cancer cells (54) were compared with our p53 microarray and ChIP-Seq data for further data analyses shown in results.
p53-regulated transcriptome in the developing kidney.
We previously described renal hypoplasia in p53-null mice (46). To identify differentially expressed genes in p53-null kidneys that might be responsible for the hypoplastic phenotype, we performed gene expression microarray on E15.5 p53+/+ and p53−/− litter-matched kidneys. Although hypoplasia is evident from E12.5, we chose E15.5 embryonic age for microarray analysis because UB branching morphogenesis and nephrogenesis are both active at this stage and for ease of comparison to available expression databases (21, 36). Since the gene expression data are to be integrated with p53 ChIP-Seq data (below), we reasoned we would detect primary p53 target genes by this strategy, despite the high number of indirectly deregulated genes. By two-color dye-swap strategy, the microarrays (44,000 probes) detected 3,233 unique genes with significant changes in gene expression (P < 0.05) (Fig. 1A).
Functional annotation of differentially expressed genes using DAVID Bioinformatics Resources v6.7 (25, 26) revealed that the p53-regulated transcriptome in the developing kidney is highly diverse. The top 10 functional categories contain genes involved in protein catabolic processes, RNA metabolism/processing and DNA metabolism/repair, cell cycle, tube development and morphogenesis, and apoptosis (Table 1). Reclassification of the annotated clusters by GO_BP (biological processes) categories showed 11 overrepresented biological processes (Fig. 1B). Almost half of the differentially expressed genes in the embryonic p53−/− kidney belong to the “development and morphogenesis” and “metabolism categories.” The p53-regulated transcriptome in the embryonic kidney differs from that of p53-null embryo (24) and cancer cell (20) in the “gene transcription and regulation,” “metabolism,” and “cell cycle” categories. In the embryonic kidney, p53-regulated metabolic processes assume a more important functional role than cell cycle, apoptosis, and DNA damage response pathways.
The gene expression changes detected in microarray analysis from p53-null kidneys were validated by real-time quantitative RT-PCR. In addition to known p53 targets such as p21, Bax, Pten, and Ccnd1 (34, 44, 55, 58), we confirmed the significant changes in expression of potentially novel target genes such as nephrogenesis regulators (Pax2, Fgf8, Osr1, and Hnf1b), Wnt pathway genes (Fzd4 and Gsk3b), ciliogenesis transcription factor Rfx3, circadian periodicity gene Per1, structural histone gene Hist1h1d, and sequence-specific transcriptional repressor Bcl6 (Fig. 2).
Chromatin-binding profile of p53 in the embryonic kidney.
We analyzed the global DNA-binding profile of p53 in E15.5 mouse kidneys. The immunoprecipitated DNA was amplified and sequenced by the Illumina/Solexa protocol, and sequences aligned to the mouse genome (mm9) to identify p53-enriched genomic regions. We mapped 17.6 million unique reads to the genome. Within these p53-enriched regions, peak recognition was performed using the MACS program (66) with a P value cut-off of 1 E-6. False peak filtering was done to remove sequences enriched in the Input (non-ChIP) sample that represent false peaks. The FDR was 1.8%. MACS identified 7,893 p53-ChIP peaks (Fig. 3A). Ninety percent (7,102/7,893) of p53 peaks are associated with genes, whereas 10% of peaks are intergenic. Two-thirds of p53 peaks (67.59%, 5,335/7,893) are located within one kilobase flanking the TSS (Fig. 3B). Of these, 68% are upstream of the TSS.
Next, we compared our p53 ChIP-Seq dataset to a published p53 ChIP-Seq dataset from cancer cells (54). The MACS mapped intervals (peak regions) were classified into promoter-overlapping or nonpromoter-overlapping intervals. Interestingly, 78% (6,157/7,893) of p53-occupied intervals in the embryonic kidney are found at gene promoters, whereas only 7% (353/5,052) of p53-occupied intervals were associated with gene promoters in cancer cells (Fig. 3C).
Motif analysis of p53 binding sites in the E15.5 kidney genome.
Using the RegionMiner search program of Genomatix Software Suite v2.0 (http://www.genomatix.de), we determined that 99.59% (7,861/7,893) of p53-occupied intervals contain at least one-half of the p53 consensus sequence, thereby validating the specificity of the ChIP. To characterize whether the sequence of the pulled-down p53 motif in our ChIP-Seq data conforms to the consensus p53 binding site, six matrices of p53 binding sites and their cores were selected by the Matrix Library 8.2 of Genomatix Software Suite. Regardless of the variation in spacer length used, a single prominent motif was identified, which greatly resembles the known consensus of p53-binding sites (Fig. 3D). There was no correlation between the ChIP-Seq tag density (peak value) and expression level of p53-bound genes (Fig. 3E).
Identification of p53-target genes in the developing kidney.
p53-ChIP tracks were examined with the Integrated Genome Browser (40). Several established p53 target genes such as p21, Bad, Gadd45a, PUMA (Bbc3), PCNA, Ccnd1, Ccne1, and Mdm2 show p53 peaks corresponding to known p53-binding sites (Fig. 4). To identify novel targets of p53, we combined our gene expression array and ChIP-Seq datasets. We found 1,683 genes common to both datasets that may be primary targets of p53 transcriptional regulation. Analysis using the KEGG_Pathway function of DAVID Resources identified 40 significantly altered pathways (P < 0.05), including cancer/diseases-specific pathways, and key pathways in kidney development (e.g., Wnt signaling, focal adhesion, and VEGF signaling) (Table 2). In addition, by overlaying the common genes to canonical signaling pathways (Ingenuity Pathway Analysis), we found that several p53-regulated genes are components of the Fgf, Wnt, Bmp, and Notch pathways (Fig. 5). Deregulation of these pathways in the absence of p53 would influence cell behavior decisions between survival, proliferation, and differentiation and thus cell fate. Figure 6 shows p53 occupancy tracks of genes in each of these pathways, such as Fgfr2, Gsk3b, Bmpr2, Smad1, Notch2, and Hes1.
DAVID functional annotation clustering analysis classified 227/1,683 (∼14%) genes into development and morphogenesis pathways (Table 3). Integration of the 1,683 p53 putative targets with the transcriptome of the developing mouse nephron (11) followed by cluster analysis [Cluster 3.0 and Treeview v1.60 (17)] indicated that target genes are enriched in nephron progenitors (CM) and nascent nephron (renal vesicle, developing proximal tubule, and glomerulus) (Table 4; Fig. 7, A and B). Figure 7C shows a heat-map of p53-target nephrogenesis genes. In the CM, p53 targets the antiapoptotic factor Bcl2, which is required for progenitor survival, the ciliary gene Kif26b, and Osr1, one of the earliest transcriptional regulators of the metanephric mesenchyme. In the renal vesicle, the earliest epithelial precursor of the nephron, p53 targets FAM188A (C1qdc2), Laminin-βa (Lmnβm), Tcf23 (HNF1β), Cyclin D1 (Ccnd1), and Cadherin 6 (Cdh6). Several of these genes encode essential regulators of nephron formation and patterning. A limitation to this approach is that the gene enrichment analysis lists genes enriched in specific compartments over other compartments (11), thus Pax2, which is present in the CM, nascent nephron, and ureteric lineage and is a p53 target gene (48), is not present in the heat-map (Fig. 7C). Figure 8 depicts ChIP-Seq tracks of novel developmental renal regulators that are potential direct p53 target genes.
Figure 9 shows a group of novel p53 target genes including the periodicity gene Per1, the chromobox3 gene (Cbx3), which encodes a chromatin-regulating protein, which binds methylated histone H3K9, and Ccdc47, which encodes a calcium-binding protein involved in embryogenesis. Per1 expression decreased twofold in mutant kidneys, whereas CBX3, HSPA9, and NAIF1 expression showed lower but significant changes in gene expression. Since the expression of these genes does not appear to be cell type specific [based on GUDMAP (21, 36)], it is possible that small changes are from localized expression changes in specific cell populations. Altered CBX3 expression may contribute to secondary gene expression changes, while changes in HSPA9 and NAIF1 may impact cell proliferation and survival. Both outcomes could significantly impact renal morphogenesis.
Functional analysis of novel p53-bound target genes.
To confirm direct regulation by p53 of putative target genes from the list of p53 bound and differentially expressed genes (Table 2), reporter analysis was done on cloned promoter fragments of select genes that showed downregulation in p53-null kidneys. We selected two genes with a high ChIP-Seq tag density (>100) and tested their promoter regulation by p53. Promoter fragments overlapping p53 peak regions from Kif1b and Gsk3b were subcloned in reporter plasmids and tested for responsiveness to p53 dosage in transient transfection assays. Increasing p53 dosage activated Kif1b promoter-driven (−1236 bp ∼ TSS ∼ +18 bp) reporter gene expression in H1299 cells (Fig. 10A). However, the 3056 bp GSK3b promoter fragment (−1530 bp ∼ TSS ∼ +1525 bp) was nonresponsive to increasing p53 levels (Fig. 10B). Lack of transcription stimulation of GSK3b promoter by p53 suggests that secondary transcriptional or posttranscriptional (via miRNA) regulation is responsible for differential expression in p53-null kidneys. Alternatively, the cloned fragment is not sufficient to promote transcription in the absence of distal/proximal sites not present in the construct.
We sought to establish a p53 gene signature in embryonic kidneys to assess the transcriptional contribution of p53 to kidney development. To this end, we performed p53-ChIP-Seq on embryonic mouse kidney tissue and integrated this data with gene expression data from p53-null kidneys to detect potential direct targets of p53. This approach allowed identification of novel p53 target genes under physiological conditions during normal development.
Our data reveal that p53 has large genome coverage in the developing kidney. However, almost 80% of genes associated with p53-bound regions did not show significant changes in gene expression in null kidneys, including bona fide p53 target genes such as Bbc3 (PUMA), Bad, Mdm2 (7, 27, 39), and novel genes such as Slc39a10, Ddx17 (p72 RNA Helicase) and Rps10. It is possible that a large number of these genes are not expressed at this point in time of kidney development, either because of a lack of appropriate coactivators or from active repression. For example, p53 apoptosis targets are expected to be expressed after severe genotoxic stress, whereas p53 target renal function genes, e.g., Bdkr2 and Aqp2 (45, 47), are expressed later in time upon nephron maturation. Since the arrays and ChIP-Seq were performed on whole kidneys, it is likely that the p53-regulated genome is underestimated given the complexity of the cell types in the kidney.
Nearly 1,600–3,700 p53 binding sites are predicted to be present in the human genome (12, 28). A study by Nikulenkov et al. (41) identified ∼4,300 high-confidence p53 ChIP-Seq peaks in MCF7 cells, a human breast cancer cell line, treated with nongenotoxic small molecules nutlin-3a or RITA or a DNA damage agent 5-FU as inducers of p53 stability. The authors also found that the p53 peaks predominantly overlapped irrespective of the method used to induce p53. Another study utilizing Actinomycin D or Etoposide to induce growth arrest versus apoptosis functions of p53, respectively, in a p53+ human osteosarcoma cell-line U2OS, collectively showed over 4,500 unique p53-bound genes (54). ChIP-Seq in mouse embryonic stem (ES) cells revealed 7,749 p53 peaks (33), similar to 7,893 peaks identified in the present study that are associated with 8,515 genes in the embryonic mouse kidney. Of the p53 occupied genes, 1,435 overlap between induced p53 in U2OS cells (54) and endogenous p53 in mouse embryonic kidney, while the majority of p53-occupied genes are unique to the developing kidney, consistent with context and cell-specific function of p53. In addition to the number of genes occupied by p53 in vivo, the chromatin-binding profile of p53 in the embryonic kidney differs from that obtained in U2OS cancer cells with regard to the location of peaks. p53 enrichment occurs predominantly at promoters in the kidney compared with intergenic and nonpromoter regions in tumor cells. This fundamental difference in occupancy may be a result of differences in the cancer versus developmental epigenome allowing differential access of p53. Whether the disparity in p53 occupancy between the human (cell lines) and mouse (ES cells and embryonic kidney) data is a result of species differences or developmental stage remains to be determined.
An interesting feature of chromatin occupancy by p53 is its binding to diverse array of genes in multiple and often opposing pathways. The wide spectrum of genes with p53 occupancy suggests that p53 might play a role of a biosensor during embryonic development, a dynamic time when the cell experiences rapidly changing cues in forms of morphogen/growth factor gradients, oxygen availability, and metabolic demands. Remarkably, around half the genes that are altered in the microarray show enrichment in developmental and morphogenesis and metabolic pathways. Several genes in Wnt, Fgf, and Notch signaling pathways are altered in expression and show p53 occupancy by ChIP, marking them as potential p53 target genes. In this regard, it is conceivable that p53 is required to integrate signals from different pathways and facilitate transcriptional cross talk between them.
Kif1b (MGI:108426) encodes a kinesin-3 motor protein for transport of lysosomes to cell periphery in nonneuronal cells (35). Since lysosomal positioning coordinates mTORC1 signaling and autophagic flux in response to nutrient availability (30), regulation of Kif1b by p53 suggests another homeostatic function of p53 in response to nutrient stress. GSK3b, besides being a key regulator of glycogen metabolism, regulates canonical Wnt signaling, as well as activity of transcription factors and microtubules (23, 64). Thus p53, via GSK3b, may contribute to multiple aspects of cell behavior including migration, proliferation, and survival (23).
Interestingly, our data revealed considerably more p53 direct targets in nephron progenitors and nascent nephrons (metanephric mesenchyme and its derivatives) than other compartments (e.g., tubular or collecting system). Indeed, our studies have shown that conditional deletion of p53 from the renal progenitor population causes deleterious effects on renewal and differentiation of these cells, leading to reduced nephron endowment (Ref. 48 and J. Liu and Z. Saifudeen, unpublished data). Our findings that components of the Wnt, Fgf, Notch, and ciliary pathways are possibly direct p53 target genes fit well with the proposed role of p53 in nephron progenitor cell homeostasis. The presence of p53 at the nexus of multiple developmental pathways invokes context-dependent modulation by p53 of cell behavior and fate in response to the developmental milieu. It is interesting to speculate that p53, via transcriptional modulation of Per1, may contribute to the periodic regulation of transcription in the CM or the ureteric tree to regulate nephrogenesis and branching morphogenesis, respectively, in response to changing developmental cues.
To our knowledge, this study is the first comprehensive analysis of the p53 transcriptome and cistrome in a developing mammalian organ. The results demonstrate the widespread binding of p53 to the genome in vivo, the context-dependent differences in the p53 regulon between cancer, stress, and development, and the role of p53 as a bona fide developmental regulator of gene expression.
Research reported in this publication was supported by an Institutional Development Award from the National Institutes of Health (NIH) under Award Number P30GM-103337, by NIH Grant R01DK-62550 to S. S. El-Dahr, and by NIH Center of Biomedical Research Excellence grant support P20RR-017659 to Z. Saifudeen. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: Y.L., J.L., and Z.S. performed experiments; Y.L., N.M., D.B., and Z.S. analyzed data; Y.L. and Z.S. prepared figures; Y.L. and Z.S. drafted manuscript; Z.S. and S.S.E.-D. conception and design of research; Z.S. and S.S.E.-D. interpreted results of experiments; Z.S. and S.S.E.-D. edited and revised manuscript; Z.S. and S.S.E.-D. approved final version of manuscript.
The authors thank the Tulane Hypertension and Renal Centers of Excellence Molecular and Imaging Core and the Cancer Crusaders Next Generation Sequence Analysis Core at the Tulane Cancer Center.
- Copyright © 2013 the American Physiological Society