An ortholog of LEAFY in Jatropha curcas regulates flowering time and floral organ development

Jatropha curcas seeds are an excellent biofuel feedstock, but seed yields of Jatropha are limited by its poor flowering and fruiting ability. Thus, identifying genes controlling flowering is critical for genetic improvement of seed yield. We isolated the JcLFY, a Jatropha ortholog of Arabidopsis thaliana LEAFY (LFY), and identified JcLFY function by overexpressing it in Arabidopsis and Jatropha. JcLFY is expressed in Jatropha inflorescence buds, flower buds, and carpels, with highest expression in the early developmental stage of flower buds. JcLFY overexpression induced early flowering, solitary flowers, and terminal flowers in Arabidopsis, and also rescued the delayed flowering phenotype of lfy-15, a LFY loss-of-function Arabidopsis mutant. Microarray and qPCR analysis revealed several flower identity and flower organ development genes were upregulated in JcLFY-overexpressing Arabidopsis. JcLFY overexpression in Jatropha also induced early flowering. Significant changes in inflorescence structure, floral organs, and fruit shape occurred in JcLFY co-suppressed plants in which expression of several flower identity and floral organ development genes were changed. This suggests JcLFY is involved in regulating flower identity, floral organ patterns, and fruit shape, although JcLFY function in Jatropha floral meristem determination is not as strong as that of Arabidopsis.


Results
Cloning and bioinformatic analysis of JcLFY in Jatropha. Full  A total of 41 LFY amino acid sequences from different species were used to construct a phylogenetic tree showing the relationship of LFY homologs ( Supplementary Fig. S2). These genes can be classified into five classes, which generally differentiate lower plants from higher plants ( Supplementary Fig. S2). Specifically, two genes from mosses form class I; ten genes from ferns form class II; five genes from gymnosperms form class III; fifteen genes from eudicots form class IV; and nine genes from monocots form class V ( Supplementary Fig. S2). JcLFY clustered together with sequences from other eudicots plants and belongs to class IV, and JcLFY has the highest identity to RcLFY from Ricinus communis ( Supplementary Fig. S2). RcLFY and JcLFY clustered in the same clade, consistent with the close evolutionary relationship between Jatropha and Ricinus communis.
Expression pattern of JcLFY in Jatropha. RT-PCR was used to investigate the expression of JcLFY across different organs and tissues. We first tested the expression of JcLFY in various tissues, revealing that this gene is highly expressed in flower buds and the transcript occurs in young leaves, flowers, fruits, and embryos. However, transcripts were not detected in roots, shoot apices, mature leaves, pedicels, and endosperms (Fig. 1A). To better analyze JcLFY transcripts, samples were collected from flowers at different developmental stages (Fig. 1B). JcLFY was highly expressed in inflorescence buds (IB1, IB2, and IB3), flower buds (FB), male flower buds (MFB), and female flower buds (FFB), but exhibited lower expression levels in bloomed male and female flowers (Fig. 1B). Overall, JcLFY was highly expressed during the early-stage inflorescence buds (e.g., IB1) and early-stage flower buds (e.g., FB1). During inflorescence and floral organ development, JcLFY expression levels decreased. Real-time qPCR agreed with the semi RT-PCR results of JcLFY expressions across flower development stages (Fig. 1C). We further analyzed JcLFY expression in different floral whorls, finding JcLFY expression higher in stamens and carpels than in sepals and petals (Fig. 1D). To determine whether JcLFY was induced by GA, we assessed JcLFY expression levels in shoot apex and flower bud tissues after application of 1 mM GA. Unexpectedly, JcLFY expression exhibited a slight decrease after GA application (Fig. 1E). Hence, JcLFY was not induced by GA in Jatropha, in contrast with Arabidopsis 34,35 and Chrysanthemum 44   The phenotypes of three independent homozygous lines from the T2 generation were examined. Relative to WT plants under LD conditions, L8, L11, and L12 bolted 9-15 days earlier, produced 2-6 fewer rosette leaves, produced 2-4 fewer branches, and were 3-17.5 cm shorter (Fig. 2B-D and Table 1). Under SD conditions, L8, L11, and L12 flowered approximately 1-2 months earlier, produced 16-36 fewer rosette leaves, and were 6.1-32.7 cm shorter (Table 2). Thus, JcLFY overexpression in Arabidopsis significantly reduced the vegetative growth period.
In transgenic plants, the primary shoots were converted into terminal flowers (Fig. 2D). The secondary shoots produced in cauline and rosette leaf axils were converted into solitary flowers. In extreme phenotypes of transgenic plants, all branches and inflorescences were replaced by solitary flowers (Fig. 2C,D). Furthermore, transgenic plants under SD conditions exhibited opposite leaves, solitary flowers in leaf axils (Fig. 2G), and two flowers within a single leaf axil (Fig. 2H). Transgenic plants L12 exhibited the highest JcLFY expression levels (Fig. 3F) and the most severe phenotypes (Fig. 2D).
To explore genes involved in the JcLFY mediated pathway, 2-week-old soil-grown transgenic 35S:JcLFY L8, L11, and L12 plants as well as WT Arabidopsis were grown under LD conditions for microarray analysis. In total, In WT plants, the primary shoot can be subdivided into a basal rosette, which contains leaves separated by short internodes, and an apical shoot with elongated internodes; the apical shoot, often referred to as an inflorescence, bears a few bracts (small stem leaves) with associated secondary shoots, as well as a potentially indeterminate number of flowers. Scale bar = 1 cm.
Scientific RepoRts | 6:37306 | DOI: 10.1038/srep37306 1,854 genes exhibited more than 2-fold changes, 122 genes exhibited more than 10-fold changes, and 22 genes exhibited more than 50-fold changes (Supplementary Table S3, and Fig. S4A). We classified 2-fold or higher expression changes in these three lines into six groups. Overall, 48 flower-related genes were detected by microarray at this threshold ( Supplementary Fig. S4); 71 phytohormone-related genes changed expression, including genes related to auxin, cytokinin, gibberellin, abscisic acid, brassinosteroids, and jasmonate; 58 stress-related genes were changed; 480 genes encoding metabolic enzymes were changed; and 474 genes of unknown function were changed (Supplementary Table S3 and Fig. S4B).
Microarray data and qRT-PCR results indicated the promotion of flowering and terminal flowers in 35S:JcLFY transgenic Arabidopsis was associated with a significant upregulation of the FM identity genes AP1, SOC1, and CAL, and the floral organ identity genes AG, AP3, SEP1, SEP2, and SEP3 ( Supplementary Fig. S3). The expression levels of these genes were highest in L12. TFL1 expression was downregulated in L12 plants ( Supplementary Fig. S3), which likely induced solitary terminal flowers. Thus, the early-flowering and terminal flower phenotypes induced by ectopic JcLFY expression in transgenic Arabidopsis were similar to those induced by AtLFY overexpression 27 . Under both LD and SD conditions, JcLFY overexpression in Arabidopsis induced early flowering.
Chromatin immuno-precipitation experiment 45 showed Arabidopsis LFY has 15 direct target genes, and each target gene was upregulated more than 2-fold. However, most of these target genes were upregulated in the present study, though the fold changes were much lower (e.g., AT5G60630, AT5G49770, At5G03790, AT5G03230, and At2g44450 in Supplementary Table S2). Some of these target genes exhibited no significant change (e.g., AT4G22780 and AT5G46660), while others were downregulated (e.g., AT3G52470 and AT3G19390) Supplementary Table S2. To examine phenotypes, we selected two independent homozygous lines in the T2 generation that exhibited high JcLFY expression (Fig. 3D,E). Under LD conditions, lines C1 and C4 bolted 8-13 days earlier and produced 2-6 fewer rosette leaves than WT plants, and bolted 18-23 days earlier and produced 9-13 fewer rosette leaves than lfy-15 mutants (Fig. 3A-E; Table 3). In transgenic plants, solitary flowers appeared in the axils of rosette and cauline leaves, and terminal flowers appeared on primary shoots (Fig. 3D,E). The lfy-15 mutant flowers were converted to inflorescences, and floral organs were abnormal and less fertile (Fig. 3C) 14 . The transgenic mutant C1 and C4 lines rescued mutant late bolting, and the conversion of flowers to inflorescences was repressed (Fig. 3D,E).
Both C1 and C4 exhibited high JcLFY expression (Fig. 3F). Promotion of flowering in the 35S:JcLFY transgenic Arabidopsis mutant was associated with a significant up-regulation of FM identity genes AP1, SOC1, and SEPs ( Supplementary Fig. S3). The branch reduction may have occurred through the repression of TFL1 by JcLFY ( Supplementary Fig. S3I). These results demonstrate that the constitutive expression of JcLFY complements increased branches (i.e., inflorescences) in later flowering stages and abnormal flowers in lfy-15 mutant; thus, JcLFY functions as a LFY homolog.    JcLFY co-suppression changed inflorescence structure and flower organ pattern. Among transgenic Jatropha generated with the 35S:JcLFY construct, we found three JcLFY co-suppressed plants. qRT-PCR results showed that JcLFY expression levels were more than 10 folds lower than WT in flower buds; L1 exhibited the lowest JcLFY expression levels ( Supplementary Fig. S5A). Such co-suppression was first reported by Napoli et al. 46 , and it has been widely reported in many transgenic plants and animals 47 . These co-suppressed transgenic Jatropha plants exhibited no late-flowering phenotypes. When regenerated plantlets were grown in the field for 4 months, flower buds emerged in both co-suppressed transgenic and control plants (Fig. 5A,C,E,G). Unlike many other species, Jatropha flowers are not subtended by small leaves called bracts (Fig. 5B,D). Jatropha flowers are composed of three concentric rings of organs: five sepals in the first, outermost whorl; five petals in the second whorl; and ten stamens (male flower) or one carpel (female flower) in the third whorl.
Inflorescences of JcLFY co-suppressed plants exhibited many bracts surrounding florets (Fig. 5F,H). One primary inflorescence branch was analyzed for its secondary inflorescence structure revealing there were more secondary inflorescence branches in co-suppressed plants. Flowers of weakly JcLFY co-suppressed plants (i.e., L20) bloomed, but all floral organs were abnormal. In male and female flowers, sepals and petals were replaced by sepal-like structures (sepaloid organs). In female flowers, stigmas were abnormal (Fig. 6G); in male flowers, only 1-2 stamens were observed, but abnormal stigmas also occurred (Fig. 6K). The first few flowers were more abnormal than later flowers; most of the early arising flowers were aborted. Such female flowers (the central flowers, specifically) have 15-20 sepaloid organs and an abnormal carpel (Fig. 7C); marginal flowers (bisexual flowers; in WT, the flower in this position is specifically male) of L20 had 15-20 sepaloid organs and a stamen fused to sepaloid organs in each flower (Fig. 7C). A cross section of the carpels of this plant revealed abnormal ovule and ovule cavity numbers, with only 0-2 observable ovules ( Fig. 7K-M). However, WT and JcLFY-overexpression plants exhibited three ovules and ovule cavities per flower (Fig. 7I,J). Since well-developed stamens are rare and ovule numbers were reduced, the fertility of weakly JcLFY co-suppressed plants was likely reduced.
Flowers of strongly JcLFY co-suppressed plants (i.e., L1) were all aborted, all flower organs were abnormal, and the flowers were smaller than those of the controls (Fig. 6D). Male and female flowers' sepals and petals were replaced by sepaloid organs. Female flower stigmas were severely abnormal (Fig. 6H); in male flowers, stamens were unobserved (Fig. 6L). We dissected the flowers of such plants, finding the female flowers (i.e., central flowers) had 20-25 sepaloid organs and abnormal carpels (Fig. 7D); the marginal flower (in WT, this is a male flower) of L1 plants had 22-28 sepaloid organs and an abnormal carpel-like organ (Fig. 7H). Cross sections of the carpels of this plant revealed abnormal ovule and ovule cavity numbers, with only 0-1 ovules in central flowers (Fig. 7N,O) and no ovules in marginal flowers (Fig. 7P). Flowers of this plant were male and female sterile.
Comparison of these JcLFY co-suppressed plants demonstrated that stronger JcLFY co-suppression was associated with more sepaloid organs and fewer stamens and ovules. This indicates that JcLFY is important in regulating floral organ development.

Altered expression of JcLFY affected fruit and seed development. To further analyze whether
JcLFY can affect fruit development, we analyzed fruit phenotypes at various developmental stage. In JcLFY co-suppressed plants, we found fruits were longer than in WT plants. Additionally, fruits were narrower, and the shapes were severely abnormal, which was maintained to maturity (Fig. 8A). Cross sections of 10-day-old fruits revealed the WT and JcLFY-overexpression plants had fruits with 3 seeds each (Fig. 8B); however, the JcLFY co-suppressed plant fruits had only 0-1 seed each, and most of the seeds aborted at an early stage (Fig. 8C). Normally, seedless-fruits are aborted at an early stage, and most developable fruits have only one seed in each fruit (Fig. 8D,E). Because of ovule abortion in co-suppressed plants, number of seeds per fruit was fewer than number of ovules per female flower (Fig. 8E).
To assess the potential of JcLFY-overexpression plants in the genetic improvement of Jatropha, we analyzed the seed number per plant, seed weight and seed oil content in T0 transgenic plants of 35S:JcLFY (Fig. 8F-H). There was no statistical difference in seed number per plant between transgenic plants and wild-types, although the average seed number per plant of 35S:JcLFY was slightly more than that of wild-type (Fig. 8F). There was also no significant change in seed weight (Fig. 8G), but oil content in seeds of 35S:JcLFY transgenic plants was significantly decreased (Fig. 8H). Because extensive variation in seed number per plant was observed among T0 transgenic plants of 35S:JcLFY (Fig. 8F), homozygous plants from the transgenic lines by self-pollination and subsequent vegetative propagation need to be used for further assessment of the overall effect of JcLFY-overexpression on oil yield of transgenic plants.

Discussion
JcLFY is an ortholog of Arabidopsis LFY. JcLFY shares high amino acid sequence similarity with other LFY proteins and most closely resembles Ricinus communis LFY. JcLFY contains conserved domains such as a proline-rich  Supplementary Fig. S1B), suggesting that JcLFY might have a similar function as LFY. We detected JcLFY transcripts in several tissues, finding the highest accumulation in flower buds (Fig. 1A). Moreover, JcLFY expression was highest in early stage inflorescence buds (IB1) and early stage flower buds (FB1) (Fig. 1B,C), implying a possible role of JcLFY in regulating Jatropha flowering.
The present study has shown that over-expressing 35S:JcLFY in Arabidopsis can inhibit vegetative growth, hence reducing branch number and heights of transgenic plants while promoting early flowering, by approximately 15 days and 2 months relative to WT plants under LD and SD conditions, respectively (Tables 1 and 2). This is quite similar to phenotypes under constitutive LFY ortholog expression in Arabidopsis 27 , Gloxinia 29 , Brassica juncea 30 , and Nicotiana tabacum 28 . Constitutive LFY expression driven by the CaMV35S promoter attenuates development of both vegetative and inflorescence phases, inducing the production of terminal flowers on Arabidopsis primary shoots 27 .
This study has shown that JcLFY overexpression in Arabidopsis produced terminal and solitary flowers (Fig. 2). These findings are similar to the phenotypic changes caused by constitutive LFY expression in Arabidopsis 27,48 . The production of terminal and solitary flowers in LFY-overexpressing plants is caused by the inhibition of TFL1 expression induced by LFY 49 . TFL1 expression was reduced in transgenic L12 plants (Supplementary Fig. S3I). Overexpression of JcLFY in the Arabidopsis lfy-15 mutant recovered the late flowering phenotype, rescued the abnormal flowers of lfy mutant plants, and repressed inflorescence development (Fig. 3B-E). This suggests that JcLFY acted as a functional homolog of LFY in Arabidopsis.
JcLFY expression was not induced by GA. GA has been generally found to strongly inhibit flowering in some woody perennial plants, such as citrus, rose, grape, and apple 36,[50][51][52] . However, GA accelerates flowering in Arabidopsis and Chrysanthemum by inducing expression of the FM identity gene LFY [34][35][36]44 . We detected JcLFY expression levels in GA-treated shoot apex and flower bud tissues at different time points, revealing that JcLFY expression levels were not induced by GA. In contrast, JcLFY expression levels decreased between 3 and 48 h after GA application (Fig. 1E). This contrasts with Arabidopsis findings, suggesting GA may play a negative function in regulating flowering in Jatropha. Recently research revealed that although GA promoted termination of vegetative development, it inhibited flower formation in Arabidopsis 36 . Ghosh et al. 9 found that paclobutrazol, a GA biosynthesis inhibitor, promotes flower initiation in Jatropha, further suggesting GA inhibits flower initiation in Jatropha. We propose that GA inhibition of flowering in Jatropha might occur via repression of JcLFY expression. This hypothesis need to be tested in future studies by determination of endogenous GA levels in different developmental stages of wild-type plants, and by analysis of phenotypic changes in flowering time in transgenic plants with altered GA biosynthesis or perception. JcLFY regulated flowering time and floral organs. LFY is a FM identity gene that plays an important role in promoting flowering in Arabidopsis 13,19 . The JcLFY-overexpressing transgenic Arabidopsis bolted two months earlier than WT plants under SD conditions (Table 2). However, JcLFY transgenic Jatropha plants took more than seven months to produce flower buds, although two months earlier than WT plants (Fig. 4). And in a subtropical area, the 35S:JcLFY transgenic Jatropha produced flowers until two or three years after plantation when no flower was found in WT plants (Supplementary Fig. S6). Compared to the 35S:LFY transgenic citrus described by Peña et al. 31 , the flowering time of transgenic Jatropha was very late; in citrus, transgenic shoots flowered just five weeks after regeneration. Therefore, the JcLFY-overexpression transgenic lines used in this work showed a slightly early flowering, but weaker phenotype than plants overexpressing the Arabidopsis LFY gene in Arabidopsis 27 and citrus 31 . The early flowering phenotype of JcLFY transgenic Jatropha was also weaker than observed in studies of transgenic Jatropha plants overexpressing the florigen gene JcFT 41,42 , in which flower buds initiated directly from transformed callus 7 weeks after in vitro culture 47 . However, the JcLFY co-suppressed transgenic Jatropha plants did not exhibit late flowering (Fig. 5). These analysis of flowering time in overexpression and co-suppression plants suggest that JcLFY may not be a key flowering promoter. In Jatropha, FMs are derived from inflorescence meristems (IMs), but FMs execute a developmental program very different from those of IMs. Thus, there must be factors that promote the determination of FMs but not IMs. JcLFY is one of these factors because co-suppression of JcLFY delayed flower formation, leading to production of more secondary inflorescence branches (Fig. 6D). Inactivation of JcLFY induced abnormal flower development in many aspects. First, the inflorescences and flowers were subtended by bracts (Fig. 5F,H). Second, most flowers were aborted, especially in the strongly co-suppressed plants. Third, the outermost floral organs were sepal-like. These sepaloid organs substantially outnumbered sepals and petals combined in WT plants (Fig. 7A-H). Fourth, the number of stamens was reduced, and the morphology of stamens was changed (Figs 6K,L and 7G,H). Fifth, carpels occurred in every flower of the co-suppressed plants, but their morphologies were abnormal, with irregular shapes and sizes (Figs 6G,H,K and  7C,D,G,H), which may result in the abnormal fruits in JcLFY co-suppressed plants (Fig. 8). Sixth, the ovules and ovule cavities were partially or completely absent in the co-suppressed plants ( Fig. 7I-P).
The flower phenotypes of JcLFY co-suppressed plants were different from those of Arabidopsis single mutants of ap1, ap2, ap3, ag, gi, or sep [53][54][55] . The co-suppressed plants produced more inflorescences, bracts and sepaloid organs, and exhibited reduced fertility, closely resembling Arabidopsis lfy mutants 14  Cloning full length JcLFY cDNA. Total RNA was isolated from Jatropha flower buds using the silica particle method 56 . cDNA synthesis was performed by using the SMART TM cDNA Library Construction Kit (Clontech, Mountain View, CA, USA) according to the manufacturer's instructions.
Sequence and phylogenetic analyses. Specific primers (XT134/XT135) were designed to obtain the full length cDNA and genomic DNA sequences of JcLFY. Genomic sequences were amplified from 30-ng samples of DNA, and 2 μ l of cDNA was used as a template for amplifying JcLFY full length cDNA sequences. The amplified PCR products were subjected to 1% agarose gel electrophoresis. Purified PCR fragments were cloned into a T&A cloning vector (Promega, Madison, Wisconsin, USA) for sequencing. Sequence alignment was carried out using Vector NTI 11 software (Invitrogen, Carlsbad, USA). To determine the amino acid sequence, the alignment results were subjected to pairwise comparisons using DNAMAN 5.0 (Lynnon Biosoft, Quebec, Canada). A phylogenetic tree based on the protein sequences was constructed with MEGA 5.0 57 . A neighbor-joining phylogenetic tree was generated with MEGA 5.0 using a Poisson model with gamma-distributed substitution rates and 1000 bootstrap replicates.

Vector construction and plant transformation.
To characterize the function of JcLFY, the coding region of JcLFY was cloned into a derived pORE R4 vector (obtained from TAIR) driven by the cauliflower mosaic virus (CaMV) 35S promoter at the SmaI and SalI sites of the pORE R4 vector. The construct was transformed into Agrobacterium tumefaciens strain EHA 105 via the freeze-thaw method 58 and transformed into WT and lfy-15 mutant Arabidopsis by the Agrobacterium tumefaciens-mediated floral dip transformation method 59 . Transgenic lines were selected on 1/2 MS medium containing 50 μ g/ml kanamycin, and transformed seedlings were further identified by JcLFY-specific PCR analysis. Verified transgenic T2 generation seedlings were transplanted into pots and grown in different light conditions for further experiments. Transformation of Jatropha with the Agrobacterium strain EHA105 carrying the same construct was performed according to the protocol described by Fu et al. 60 . All transgenic plants were confirmed using genomic PCR and RT-PCR. Semi-quantitative PCR and real-time qPCR analyses. Total RNA was isolated from different tissues of Jatropha using the silica particles method 56 . Ten RNA samples from different Jatropha tissues were obtained, including roots, young leaves, mature leaves, flower buds, flowers, shoot apices, pedicels, fruits, endosperms, and embryos. Total RNA was extracted from frozen Arabidopsis tissues derived from 15-day-old WT plants, lfy-15 mutant plants, WT transgenic plants (lines 8, 11, and 12, henceforth L8, L11, and L12), and lfy-15 mutant transgenic plants harboring JcLFY (plants C1 and C4) using TRIzol reagent (Biocentury Transgene, Shenzhen, China). Total RNA samples of transgenic and WT Jatropha were isolated from the flower buds using the silica particles method 56 . First-strand cDNA was synthesized with the PrimeScript ® RT Reagent Kit with gDNA Eraser (TAKARA, Dalian, China). The cDNA templates of first-strand cDNA were diluted 5-fold with sterilized double-distilled water. qRT-PCR was performed using SYBR ® Premix Ex Taq ™ II (TAKARA) on a Roche 480 Real-Time PCR Detection System (Roche, Mannheim, Germany). The primers employed for qRT-PCR and semi-quantitative PCR are listed in Supplementary Table S1. qRT-PCR was conducted with three independent biological replicates and three technical replicates for each sample. The data were analyzed using the 2 −ΔΔCT method described by Livak and Schmittgen 61 . RT-PCR was carried out as described by Brownie et al. 62 . The transcript levels of specific genes were normalized using Jatropha JcActin1 or Arabidopsis Actin2.
Gene array analyses. The 35S:JcLFY transgenic Arabidopsis were transplanted into soil under LD conditions. Total RNA was extracted from the aboveground tissues of 2-week-old soil-grown WT Arabidopsis and transgenic Arabidopsis using an RNeasy ® Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Total RNA was quantified, and its quality was assessed using an Agilent 2100 Bioanalyzer (Santa Clara, CA, USA). Microarray analysis was performed using the Arabidopsis (V4, 4 × 44) Gene Expression Microarray, Design ID: 021169 (Agilent) containing 43,603 Arabidopsis gene probes and 1,417 Agilent control probes. A total of at least 1 μ g purified RNA was required for gene array analyses.
Labeled cRNA probes were fragmented using fragmentation buffer and hybridized to the Arabidopsis arrays in the presence of the Gene Expression Hybridization buffer HI-RPM and blocking agent for 17 h at 65 °C with a 10-rpm rotation speed in a hybridization oven. After the 17 h incubation, the arrays were washed using low stringency wash buffer 1 at room temperature for 1 min followed by a high stringency wash using wash buffer 2 at 37 °C. The arrays were air-dried and scanned using a high-resolution array scanner (Agilent) with the appropriate settings for one-color gene expression arrays. The signal intensities were extracted from the scanned images with the aid of Feature Extraction software 10.7.1.1 (Agilent) and subjected to background subtraction and spatial detrending. The outliers and abnormal features were flagged, and the data were normalized using intra-array percentile shift normalization (minimum threshold of 75) and median-based inter-array normalization. GeneSpring GX (Agilent) was used to calculate intensity ratios and fold changes. All genes with a P-value below 0.05 and at least 2-fold expression changes were chosen for a Gene Ontology enrichment analysis. The gene array hybridization experiments were performed by Shanghai Biotechnology Corporation (Shanghai, China).