The histone code reader Spin1 controls skeletal muscle development

While several studies correlated increased expression of the histone code reader Spin1 with tumor formation or growth, little is known about physiological functions of the protein. We generated Spin1M5 mice with ablation of Spin1 in myoblast precursors using the Myf5-Cre deleter strain. Most Spin1M5 mice die shortly after birth displaying severe sarcomere disorganization and necrosis. Surviving Spin1M5 mice are growth-retarded and exhibit the most prominent defects in soleus, tibialis anterior, and diaphragm muscle. Transcriptome analyses of limb muscle at embryonic day (E) 15.5, E16.5, and at three weeks of age provided evidence for aberrant fetal myogenesis and identified deregulated skeletal muscle (SkM) functional networks. Determination of genome-wide chromatin occupancy in primary myoblast revealed direct Spin1 target genes and suggested that deregulated basic helix-loop-helix transcription factor networks account for developmental defects in Spin1M5 fetuses. Furthermore, correlating histological and transcriptome analyses, we show that aberrant expression of titin-associated proteins, abnormal glycogen metabolism, and neuromuscular junction defects contribute to SkM pathology in Spin1M5 mice. Together, we describe the first example of a histone code reader controlling SkM development in mice, which hints at Spin1 as a potential player in human SkM disease.

Spindlin1 (Spin1) is a histone code reader binding histone H3 trimethylated at lysine 4 (H3K4me3) with high affinity. [1][2][3] H3K4me3 association is enhanced by the presence of asymmetrically dimethylated arginine 8 of histone H3 4 . Spin1 is highly expressed in several types of tumors [5][6][7] and affects cell cycle, chromatin segregation, apoptosis, and transformation of cell lines, as well as tumor formation in nude mice. 6,[8][9][10][11] While these studies suggest important roles in cancer, physiological functions of Spin1 have only been subject to initial investigation. Mouse oocytes deficient for maternal Spin1 undergo normal folliculogenesis, but fail to resume meiosis. 12 Furthermore, mice with ubiquitous Spin1 ablation die shortly after birth. 12 However, tissue-restricted defects accounting for postnatal death have not been reported.

MHC MHC
metabolism, and defective NMJs contribute to SkM pathology in Spin1 M5 mice. In summary, our data reveal a severe developmental defect caused by ablation of a histone code reader in SkM and hint at Spin1 as a potential player in human SkM disease.

Results
Loss of Spin1 in SkM results in postnatal lethality. To investigate physiological functions of Spin1 in vivo, we generated ubiquitous knockout (Spin1 R26 ) mice by crossing Spin1 p/p mice with the Rosa26-Cre deleter strain 32 (Supplementary Figure 1a). Spin1 R26 mice were born, but died within one day after birth (Supplementary Figure 1b), which is in agreement with observations by others. 12 Of note, at E18.5 Spin1 R26 mice displayed dropping forelimbs ( Figure 1a) indicating a neuromuscular defect. 33 To address potential functions of Spin1 in SkM, we analyzed Spin1 protein expression in hind limb sections of control mice at E15 and after birth (P0) by immunofluorescence. At both time points, we observed intense Spin1 staining in Pax7positive myoblast precursors (Figure 1b (arrowheads)) and weaker staining in nuclei of myofibers (Figure 1c (arrowheads)). Of note, in newborn mice, Spin1 expression was undetectable in some nuclei of myofibers ( Figure 1c, bottom row (arrows)).
Next, we deleted Spin1 in myoblast precursors by crossing Spin1 p/p mice with the Myf5-Cre deleter strain 31  Homozygous Spin1 M5 mice were obtained at the expected Mendelian ratio at birth (Supplementary Figure 1h). However, about 80% of Spin1 M5 mice died within one day after birth. Newborn Spin1 M5 mice could typically be distinguished from control littermates by an abnormal posture and the absence of milk in the stomach (Figure 1f). Moreover, at E16.5, we observed dropping forelimbs for Spin1 M5 fetuses ( Figure 1g). Together, our data show that ablation of Spin1 in SkM causes early postnatal death of the majority of mice.
SkM of Spin1 M5 mice is characterized by necrosis and structural defects in non-necrotic fibers. To characterize SkM defects in Spin1 M5 mice, we inspected hematoxylin & eosin (H&E)-stained hind limb sections at different stages of development. Compared with control littermates, we observed in newborn Spin1 M5 mice loss of fibers ( Figure 2a, top row (black asterisks)) and numerous immature or degenerating fibers lacking contractile material (dashed circles). In Spin1 M5 fetuses at E16.5, we also noted fibers with irregular H&E staining (Figure 2a, middle row (dashed circles)) and at E15 enlarged fibers, which were less abundant in control samples (Figure 2a, bottom row (white asterisks)).
Electron microscopy analyses of hind limb sections revealed for newborn Spin1 M5 mice degenerating, necrotic fibers ( Figure 2b, columns I-II (demarcated by dashed lines)), defective mitochondria (Figure 2b, column III (arrows)), and abnormal glycogen accumulation (Figure 2b, column IV (asterisks)). Similar defects were detected for Spin1 M5 fetuses at E16.5 (Supplementary Figure 2). Furthermore, in nonnecrotic fibers of Spin1 M5 fetuses, we observed structural defects including a low density of contractile material and the lack of a clear M-line (Supplementary Figure 2, column III (triangles)). Together, our H&E and electron microscopy analyses uncovered necrotic and structurally defective fibers in Spin1 M5 mice. In addition, the data suggested an onset of SkM defects before E16.5.
The common E15.5/E16.5 DEGs Ankrd1 (CARP) and Ankrd2 (ARPP) encode titin-associated regulators of sarcomere function, whose expression is often deregulated in SkM disease. 35 We therefore investigated Ankrd1 and Ankrd2 protein levels in hind limb sections of E16.5 fetuses by immunofluorescence (Figures 3b and c). Most prominently, in control fetuses, expression of both proteins was restricted to the inner (prospective oxidative) part of the tibialis anterior (TA) neighboring the tibia (Figure 3c (demarcated by dashed lines)), whereas in Spin1 M5 fetuses staining was also present in the outer (prospective glycolytic) part of the TA (Figure 3c (asterisks)). In comparison, at E15 Ankrd1 and Ankrd2 staining of hind limb muscle of Spin1 M5 and control mice was similar ( Supplementary Figures 3a and b). Thus, To analyze which proportion of developmentally regulated genes is affected by the absence of Spin1, we compared the E16.5 DEGs with gene expression changes in control fetuses from E15.5 to E16.5. In control fetuses, we observed 2042 genes differentially expressed between E15.5 and E16.5 ( Figure 3d; Supplementary Table 1c). Phenotype and pathway analyses for these 2042 DEGs using WebGestalt 36 confirmed relevance of the identified genes for SkM development or function ( Supplementary Figures 3c and d). Intersection with the 193 E16.5 DEGs revealed an overlap of 125 genes ( Figure 3d). Therefore, the absence of Spin1 prevents adequate expression of a subset of genes regulated during the fetal phase of myogenesis in control mice.
To corroborate aberrant fetal myogenesis in Spin1 M5 mice, we compared the E16.5 DEGs with transcripts previously observed to be differentially expressed in embryonic or fetal myotubes. 37 Of the 27 transcripts more highly expressed in  (Figure 3e). Importantly, eight out of thirteen transcripts more highly expressed in fetal myotubes (Myh8, Tnni2, Tnnt3, Atp2a1, Casq1, Pvalb, Ckm, and Eno3) were downregulated DEGs in Spin1 M5 mice at E16.5 ( Figure 3e). Together, our data provided evidence for an impaired progression of fetal myogenesis in Spin1 M5 mice.
Identification of deregulated SkM functional networks in Spin1 M5 fetuses. To identify genes accounting for SkM defects in Spin1 M5 fetuses, we performed phenotype and pathway analyses for the E16.5 DEGs using WebGestalt 36 (Figures 4a and b). These analyses revealed terms related to SkM function, metabolism, and pathology. We grouped the genes associated with the phenotype and pathway terms according to functional networks previously proposed in systems biology analyses for SkM 38 Deregulated basic helix-loop-helix transcription factor networks account for SkM defects in Spin1 M5 fetuses. In addition to genes required for SkM structure or function, we found the myogenic bHLH transcription factor-encoding genes Myf5, Myog, Msc (musculin, MyoR), and Hes6 15,40-43 among the E16.5 DEGs (Figure 4c (network 'transcription and signaling')). Therefore, deregulated bHLH transcription factor-controlled gene expression may be an early event accounting for aberrant myogenesis in Spin1 M5 fetuses. In support of this idea, Myf5 and Msc were also among the E15.5 DEGs (Supplementary Table 1a).
To correlate gene expression with Spin1 chromatin occupancy, we performed chromatin immunoprecipitation followed by massive parallel sequencing (ChIP-seq) using primary myoblasts. ChIP-seq with Spin1-specific antibody revealed 17106 peaks, of which the majority (61.2%) was located at the promoters (transcription start site (TSS) ± 1 kb) of 12438 genes ( Despite occupying thousands of gene promoters, Spin1 apparently only regulates a subset of target genes during fetal myogenesis. To further investigate Spin1-mediated gene regulation, we intersected the E16.5 DEGs with the cistrome observed in primary myoblasts. This analysis revealed the presence of Spin1 and H3K4me3 at the promoter of 88 E16.5 DEGs (Figure 5a). A transcription factor target analysis for these 88 genes using WebGestalt 36 revealed a significant enrichment of potential Myod1 target genes followed by E12 (Tcf3) and Mef2 targets ( Figure 5c).
To validate Myod1 promoter occupancy of Spin1 target genes, we analyzed a previously deposited Myod1 ChIPseq data set for C2C12 myoblasts (GEO data set GSE36024), which uncovered Myod1 peaks at the promo- Surviving Spin1 M5 mice exhibit major defects in soleus, tibialis anterior, and diaphragm. While the majority of Spin1 M5 mice died shortly after birth, about 20% survived and reached adulthood. We next investigated the consequences of aberrant fetal myogenesis for SkM in surviving Spin1 M5 mice between 15 and 30 weeks of age, the latter representing the oldest cohort obtained. Adult Spin1 M5 mice exhibit a severe scoliosis, cachexia, and weigh only around 60% of control littermates (Figures 6a and b).
To investigate whether degenerative changes affected specific fiber types, we analyzed SkM sections by immunofluorescence. The TA of control mice exhibited typical fiber type compositions, that is, mainly type IIb fibers in white (primarily glycolytic) TA and type IIa and IIx fibers in red ). Overall, these observations argue against selective fiber type degeneration in Spin1 M5 mice, but rather suggest that certain muscles including soleus, TA, and diaphragm degenerate.
Identification of deregulated SkM functional networks in surviving Spin1 M5 mice. We next analyzed the transcriptomes of TA of Spin1 M5 and control mice at three weeks of age (P21), which revealed 1040 DEGs (P ≤ 1e − 5; fold change ≥ 1.5; Figure 7a; Supplementary Table 1d). Intersection with the E16.5 DEGs detected a highly significant overlap of 50 genes (Figure 7a) suggesting that defects in surviving Spin1 M5 mice are related to differential gene expression at early stages of disease.
Phenotype and pathway analyses for the P21 DEGs uncovered terms related to muscle function and disease, as well as metabolism (Figures 7b and c). In addition, we noted terms such as 'focal adhesion' and 'regulation of actin cytoskeleton', which were not identified at E16.5 (Figures 4a  and b). Grouping of the P21 DEGs according to systems biology classifications 38,39 revealed the same deregulated SkM networks (Figure 7d) as for the E16.5 DEGs (Figure 4c). At P21, however, the networks contained more DEGs than at E16.5, and we identified additional networks ('hypertrophy & atrophy pathways' and 'cytoskeleton'). Of note, the strong increase in deregulated genes including collagen isoforms in the network 'extracellular matrix (ECM)' (Figure 7d) correlates with fibrosis detected in the TA of adult Spin1 M5 mice (Supplementary Figure 5d). Together, this analysis provided a comprehensive disease signature for the TA of surviving Spin1 M5 mice.
Differential gene expression in surviving Spin1 M5 mice correlates with SkM disease patterns. In the final set of experiments, we aimed to correlate differential gene expression in surviving Spin1 M5 mice with SkM disease patterns. Since several P21 DEGs encode titin-associated proteins 21,22 (Figure 7d, network 'Ttn-associated'), we hypothesized similarity with diseases caused by Ttn mutations such as tibial muscular dystrophy (TMD). 44,45 Muscular dystrophy with myositis (mdm) mice (serving as a model of TMD) express a titin mutant lacking the N2A region, which binds Ankrd1, Ankrd2, and Ankrd23. Accordingly, defective Ankrd1 signaling has been implicated in TMD in mdm mice. 46 Intersection of the P21 DEGs with 75 DEGs previously reported in mdm mice 46 revealed a significant overlap of 26 genes (Figure 8a). Given that Ankrd1 and Ankrd2 are already deregulated in Spin1 M5 fetuses at E16.5 (Figures 3c and 4c), aberrant expression of titin-associated proteins may account for SkM defects in Spin1 M5 mice.
Next, we observed downregulation of genes involved in glycogen metabolism including Pygm, Pfkm, Phka1, Phkg1, Prkaa2, and Prkag3 (Figure 7d, networks 'glucose & glycogen' metabolism and 'hypertrophy & atrophy pathways'). Mutations of these genes are known to occur in glycogen storage diseases (glycogenoses). 28 Periodic acid-Schiff (PAS) staining revealed abnormal glycogen deposits in TA and soleus fibers of Spin1 M5 mice (Figure 8b, black triangles). Although these deposits are limited to individual fibers, which differs from typical glycogenoses, our observation suggests that defective glycogen metabolism contributes to SkM disease in adult Spin1 M5 mice.
Finally, we noted strong deregulation of acetylcholine receptor subunits (Chrna1, Chrnd, Chrne, Chrng) and several genes involved in excitation-contraction coupling (Figure 7d, networks 'NMJ' and 'ECC') hinting at defective neuromuscular junctions and/or excitation-contraction coupling. Therefore, we analyzed neuromuscular junctions in the diaphragm of newborn and adult Spin1 M5 mice by electron microscopy (Figure 8c). At both time points, the analysis revealed defects of the synaptic membrane (white triangles) and an abnormal appearance as well as a reduced number of synaptic vesicles (arrows) in Spin1 M5 mice (Figure 8c). Furthermore, we observed the presence of vacuoles at nerve terminals of adult Spin1 M5 mice (Figure 8c, right columns (asterisks)). These data provide evidence for neuromuscular junction damage and abnormal excitation-contraction coupling in Spin1 M5 mice.     40 Hes6 was reported to inhibit expression of Msc thereby enhancing myoblast differentiation. 43 Therefore, deregulated bHLH transcription factor networks appear to deter-mine, at least in part, aberrant gene expression in Spin1 M5 fetuses. The preferential degeneration of soleus, TA, and diaphragm in adult Spin1 M5 mice raised the question whether this is caused by fiber type-or muscle type-selective mechanisms. Currently, we favor the latter hypothesis. While degeneration of type I and type II fibers might account for soleus defects, it would not convincingly explain the severe TA damage since this muscle contains only about 1 and 10% of these fibers types, respectively. 47 Furthermore, at E15.5 or E16.5, we did not identify DEGs involved in the regulation of fiber type identity or plasticity (e.g., Six1, Six4, Eya1, Sox6). 30 One possible determinant of muscle type-selective degeneration in Spin1 M5 mice could be the deregulation of titinassociated proteins such as Ankrd1, which has been linked with TMD in mdm mice. 46 However, despite some similarity with TMD, the SkM phenotype of Spin1 M5 mice is apparently more complex. We exemplarily correlated the P21 DEGs with abnormal glycogen accumulation in individual TA and soleus fibers as well as neuromuscular junction defects in Spin1 M5 mice. Due to the high number of DEGs observed in the TA of Spin1 M5 mice at P21, these examples only provide an initial characterization of deregulated SkM networks. Compared with transcription factors, signaling molecules, SkM structural proteins, or metabolic enzymes, the roles of epigenetic regulators in SkM physiology and disease have only more recently become a focus of research. However, while epigenetic writers, erasers, and non-coding RNAs have received considerable attention, 48,49 little is known about potential functions of histone code readers in SkM. Few readers (Brd4, Ing2, Sfmbt1, Dpf3) have been implicated in myogenesis using the C2C12 cell culture model [50][51][52][53] or in zebrafish. 54 However, in these cases either knockout mouse models have not been reported, or in mice no apparent SkM phenotype was observed. [55][56][57] Thus, to the best of our knowledge, Spin1 is the first histone code reader, for which ablation in SkM has fatal consequences in mice. Together, our histological and transcriptome analyses provide insight into severe SkM defects in Spin1 M5 fetuses and surviving adult mice, hinting at Spin1 as a potential player in human SkM disease.

Materials and Methods
Mouse studies. All mice were housed in the pathogen-free barrier facility of the University Medical Center Freiburg in accordance with institutional guidelines and approved by the regional board. Mice were maintained under temperature-and humidity-controlled conditions with a 12-h light/dark cycle, free access to water, and a standard rodent chow (3807, Provimi Kliba). Animals were killed by cervical dislocation and tissues immediately processed for further analyses.
Generation and validation of Spin1 R26 and Spin1 M5 mice. The targeting strategy for generation of a conditional Spin1 allele is outlined in Supplementary Figure 1a. Details are available upon request. The targeting construct was electroporated into C57BL/6 N Tac embryonic stem (ES) cells (Taconic), and neomycin-and puromycin-resistant clones were expanded. Selected ES cells were injected into blastocysts of C57BL/6 N mice. Resulting mice were bred to Rosa26-Flp mice to remove NeoR and PuroR selection markers. Offspring harbored a conditional Spin1 allele, in which exon 4 of Spin1 was flanked by loxP sites. C57/Bl6N mice homozygous for the conditional allele (Spin1 p/p ) were crossed with the Rosa26-Cre deleter strain 32 to generate ubiquitous knockout (Spin1 R26 ) mice or with the Myf5-Cre deleter strain 31 to selectively ablate Spin1 in myoblast and brown adipocyte precursors (Spin1 M5 mice). Mice were maintained on the C57/ Bl6N background. In phenotypic analyses of Spin1 M5 mice, Spin1 +/p or Spin1 p/p littermates served as controls. Mice were genotyped by PCR amplification using GoTaq G2 Flexi DNA polymerase (Promega) of genomic DNA extracted from tail biopsies with the NucleoSpin 96 tissue kit (Macherey Nagel). The following genotyping primers were used: forward 5-ATAGGCTCTCTGGCATGGTG-3 / reverse 5-ACAGCGTGACACATCAAAGC-3 detecting the wild-type (177 bp) and the conditional (337 bp) allele and forward 5′-GGAGGAAGACACCTAATAGCACC-3′ /reverse 5′-AAGGCAAAACGGAGACAGC-3′ detecting the deleted allele (395 bp).
Efficient Spin1 depletion in myogenic cells and remaining Spin1 expression in nonmyogenic cells of Spin1 M5 mice was validated as follows. Spin1 M5 and control mice were crossed with mice harboring a green fluorescent protein/nuclear lacZ (GNZ) reporter. 58 In mice harboring the reporter, GNZ expression is Cre dependent. GNZpositive myogenic cells in Spin1 M5 mice did not express Spin1, whereas Spin1positive cells were GNZ-negative and therefore non-myogenic (Supplementary Figure 1f). Part of these Spin1-expressing, non-myogenic cells were Tcf4-positive fibroblasts 34 (Supplementary Figure 1e). Reduction of Spin1 mRNA in hind limb muscle of newborn Spin1 M5 mice was confirmed by quantitative RT-PCR (Supplementary Figure 1g).
Of note, none of the RNA-seq analyses identified Spin1 as differentially expressed gene. Examination of the normalized reads at exons 1 to 6 of the Spin1 gene explained this observation (Supplementary Table 1e). Reads were significantly decreased at exon 4 (which is excised in a Myf5-Cre-dependent manner in myogenic cells), but not at other exons in Spin1 M5 relative to control mice. Thus, in myogenic cells of Spin1 M5 mice a truncated transcript (lacking exon 4) is expressed, which results in translation of a truncated Spin1 peptide (containing 56 N-terminal amino acids lacking any known functional domain) due to the presence of a premature STOP codon in exon 5. Normal Spin1 transcript (containing exon 4) present in SkM of Spin1 M5 mice is most likely produced by non-myogenic cells ( Supplementary Figures  1e and f). Since mapped reads at all exons contribute to the overall count, differential expression of Spin1 is not detected in RNA-seq analyses.
Isolation of primary myoblasts and ChIP sequencing. Primary myoblasts were isolated from 10-day-old C57BL/6 N mice using an established preplating protocol. 64 Briefly, limb muscles were collected from front and hind leg, minced, digested for 1 h in 0.2% collagenase type I (Sigma, C0130), and filtered through 100 μm cell strainers (Falcon, 352360). Cells were collected by centrifugation, resuspended in DMEM medium (Gibco, Schwerte, Germany, 11995-065) supplemented with 10% FCS, 10% horse serum, and 1.25% chicken embryo extract (Seralab, CE-650-J), and cultivated for 1 h on 6 cm tissue culture dishes to allow for attachment of fibroblasts. Supernatants were transferred to 6 cm dishes, again incubated for one hour, and then transferred to 10 cm tissue culture dishes coated with collagen (Gibco, A10483-01). Myoblasts were cultivated for 3 to 4 days until a confluency of about 70% was reached. Cells were fixed with 1% formaldehyde for 5 minutes, quenched in glycine (1.25 M), washed with PBS buffer, collected, and snap-frozen in liquid nitrogen. Chromatin was prepared using the NEXSON procedure. 65 Briefly, nuclei were extracted by sonication with a Covaris E220 sonicator (75 W peak power, 2% duty factor, 200 cycles/burst, 60 s). Nuclei were pelleted, resuspended in 1 ml of shearing buffer, and sonicated for 12 min (140 W peak power, 5% duty factor, 200 cycles/ burst). Chromatin was diluted 1:2 in buffer H (Diagenode auto histone ChIP-seq kit (C01010022)) before ChIP. For ChIP, one tenth of the chromatin was incubated with 1 μg of H3K4me3 antibody (Diagenode, C15410003). The remaining chromatin (from 1.2 million cells) was incubated with 5 μg of SPIN1(5867) antibody. ChIP was performed using the automated platform SX-8G IP-Star (Diagenode) and the program 'ChIP indirect method'. Chromatin was incubated with antibody for 10 h followed by 3 h of incubation with protein A-conjugated beads. 1% of the original chromatin was used as input. After elution from beads, ChIP and input samples were reverse crosslinked, and DNA purified with MinElute columns (Qiagen, 28004). Libraries were prepared using the NEBNext Ultra DNA library preparation kit (NEB, E7370S) according to the manufacturer's instruction and without size selection. Adapter-ligated fragments were amplified with 12 PCR cycles and sequenced on the Illumina HiSeq 2500 platform by the Deep Sequencing Unit of the Max-Planck-Institute of Immunology and Epigenetics (Freiburg).
Paired-end reads were mapped to the mouse reference genome (mm10) using bowtie 2 66 with default parameters. Data were further analyzed using the peak finding algorithm of MACS 1.4.2 67 using input as control. All peaks with a false discovery rate 41% were excluded from further analyses. The uniquely mapped reads were used to generate genome-wide intensity profiles, which were visualized using the integrative genomics viewer (IGV). 68 Peaks were annotated and overlaps between different peak files were calculated with HOMER. 61 The genomic features (promoter, exon, intron, 5′ or 3′ untranslated region, and intergenic region) were defined and calculated using Refseq and HOMER. Myod1 chromatin association in C2C12 myoblasts was analyzed using a previously deposited GEO data set (GSE36024). Venn diagrams were generated with the help of Venny. 69 Intensity profiles for Spin1 and H3K4me3 gene promoter occupancy were analyzed with SeqMINER. 70 GEO data availability. The RNA and DNA sequencing data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus 71 and are accessible through GEO Series accession number GSE92539.
Electron microscopy. Muscle samples were fixed by immersion in 2.5% glutaraldehyde and 2.5% paraformaldehyde in cacodylate buffer (0.1 M, pH 7.4) and then washed in cacodylate buffer for further 30 min. The samples were post-fixed in 1% osmium tetroxide in 0.1 M cacodylate buffer for 1 h at 4°C, and dehydrated in an ascending ethanol gradient (50,70,90, and 100%) and propylene oxide for 30 min each. Samples were embedded in Epon 812 substitute (Sigma-Aldrich). Semi-thin sections cut at 2 μm and ultra-thin sections cut at 70 nm were contrasted with uranyl acetate and lead citrate and examined at 70 kV with a Morgagni 268D electron microscope. Images were digitally captured by Mega View III camera (Soft Imaging System).
Statistics. Transcriptome and cistrome data were analyzed as described above. Statistical significance of gene set intersections was evaluated by a hypergeometric test using the program 'R' (http://www.R-project.org) [phyper (N12-1, N1, N-N1, N2, lower.tail = FALSE) with N1 (genes in set 1), N2 (genes in set 2), N12 (genes in intersection), and N (genome size)]. The enrichment factor (R) was calculated according to R = (N × N12)/(N1 × N2). Other data are presented as the mean value or percentage change +S.D. Comparisons between two data sets were made using the two-tailed Student's t-test for parametric data and the Wilcoxon signed-rank test for nonparametric data. A P-value of less than 0.05 was considered statistically significant. Statistical significance is indicated as follows: *Po0.05, **Po0.01, ***Po0.001.

Conflict of Interest
The authors declare no conflict of interest.