A spatial dissection of the Arabidopsis floral transcriptome by MPSS
© Peiffer et al; licensee BioMed Central Ltd. 2008
Received: 20 August 2007
Accepted: 21 April 2008
Published: 21 April 2008
We have further characterized floral organ-localized gene expression in the inflorescence of Arabidopsis thaliana by comparison of massively parallel signature sequencing (MPSS) data. Six libraries of RNA sequence tags from immature inflorescence tissues were constructed and matched to their respective loci in the annotated Arabidopsis genome. These signature libraries survey the floral transcriptome of wild-type tissue as well as the floral homeotic mutants, apetala1, apetala3, agamous, a superman/apetala1 double mutant, and differentiated ovules dissected from the gynoecia of wild-type inflorescences. Comparing and contrasting these MPSS floral expression libraries enabled demarcation of transcripts enriched in the petals, stamens, stigma-style, gynoecia, and those with predicted enrichment within the sepal/sepal-petals, petal-stamens, or gynoecia-stamens.
By comparison of expression libraries, a total of 572 genes were found to have organ-enriched expression within the inflorescence. The bulk of characterized organ-enriched transcript diversity was noted in the gynoecia and stamens, whereas fewer genes demonstrated sepal or petal-localized expression. Validation of the computational analyses was performed by comparison with previously published expression data, in situ hybridizations, promoter-reporter fusions, and reverse transcription PCR. A number of well-characterized genes were accurately delineated within our system of transcript filtration. Moreover, empirical validations confirm MPSS predictions for several genes with previously uncharacterized expression patterns.
This extensive MPSS analysis confirms and supplements prior microarray floral expression studies and illustrates the utility of sequence survey-based expression analysis in functional genomics. Spatial floral expression data accrued by MPSS and similar methods will be advantageous in the elucidation of more comprehensive genetic regulatory networks governing floral development.
The majority of genes implicated in floral development have been identified through characterization of mutants displaying severe phenotypic deviations from wild-type development. An interesting subset of these mutants is the group of homeotic floral phenotypes. In these mutants, the organs of a single whorl of the inflorescence are duplicated within another distinct whorl at the expense of the organs typically present. The premise for these mutations is explained in the classic "ABC model" of floral development for Arabidopsis [1, 2]. According to the ABC model, interactions among MADS-box transcription factors including, but not limited to APETALA 1 (AP1), AGAMOUS (AG), and APETALA 3 (AP3) are required for sepals, petals, stamens, and gynoecia development. Functional absence of any of these transcription factors results in the homeotic replacement of floral organs. The "quartet model" of protein interactions explains the genetic ABC model  and proposes the formation of five distinct whorl-specific tetrameric complexes capable of binding DNA and activating downstream genes responsible for organ development through cis-regulation at dual CArG boxes . In vitro analysis has revealed heterodimeric interactions among MADS-box transcription factors . Furthermore, in vivo interactions of homologous petunia MADS-box proteins involved in a putative ovule-defining quaternary complex were also observed .
Despite structural support for the quartet model, many regulatory aspects of this model have yet to be identified. A number of inflorescence meristem-identity genes such as LEAFY, UNUSUAL FLORAL ORGAN, LEUNIG, and CURLY LEAF have been linked to the upstream regulation of the genes encoding these quaternary complexes. However, few downstream organ-specific genes directly activated by these complexes have been identified . Moreover, downstream targets such as FRUITFUL, SPOROCYTELESS/NOZZLE and NO APICAL MERISTEM do not obey the single whorl premise of the quartet model. [7–9]. Characterization of organ-specific gene expression downstream of the putative quaternary complexes is necessary to validate its functionality and understand the nature of its targets.
Genomic approaches have become a valuable tool in characterizing organ-related gene expression and in elucidating the genetic networks of floral development at a global level. Genome-wide analyses of transcript enrichment among Arabidopsis floral organs have been performed with the aid of hybridization-based approaches such as cDNA and oligonucleotide microarrays [10–20] and represent a strong first step in spatial characterization of the floral transcriptome. However, microarray analyses and other hybridization-based approaches are subject to a number of inherent limitations, including sensitivity to RNA quantity, non-specific probe hybridization, and substantial background levels capable of masking transcripts with low expression rates . Furthermore, quantitative analysis across multiple microarrays requires the standardization and calibration of chips to ensure equivalent hybridization.
Although technical improvements are addressing several of these microarray issues, signature sequencing (such as massively parallel signature sequencing, MPSS) represents an alternative to microarrays and can overcome a number of limitations inherent to hybridization-based technologies and other conventional methods of large-scale gene expression analysis. Developed at what is now Illumina, Inc. (originally Lynx Therapeutics, Hayward, CA), MPSS reactions permit the simultaneous or parallel sequencing of 17 or 20 nucleotide "signatures" corresponding to distinct cDNA molecules from a sample [22, 23]. These expression signatures may then be matched to their corresponding sequence in the genome to delineate gene expression. The length of MPSS signatures usually permits a single match to the Arabidopsis genome and enables highly specific quantification of transcription . The background level afforded by this technology is superior to the level of transcript detection permitted by hybridization-based technologies and enables detection of transcriptions with lower expression levels  such as many transcription factors. Also, the linear normalized nature of MPSS data acquisition reduces the importance of signal standardization between cDNA libraries. Nonetheless, previously described "bad words" in MPSS sequencing reactions , as well as the absence of the necessary restriction sites within a particular cDNA, dilution of transcripts with low expression in diverse transcriptomes, and the cost-prohibitive nature of replicates are shortcomings of the technology. Therefore, integration of microarray studies and MPSS in future meta-analysis will permit improved genome-wide expression characterization, as will advances in short read DNA sequencing technologies, such as Illumina's current "sequencing-by-synthesis" (SBS) method.
In this study, we have implemented MPSS to dissect those genes enriched within the petal, stamen, gynoecium inclusive and exclusive of the ovule, and those of the sepal/sepal-petal, petal-stamen, or stamen-gynoecium. MPSS signatures were matched to genomic annotations (TAIR Version 6) and integrated into our publicly available web interface . Using the interface, expression patterns were cross-analyzed to dissect floral organ(s) of enriched expression. Wild-type inflorescences were used to characterize transcript expression from all floral organs. The mutant agamous ("ag," SALK_014999), was incorporated to delineate a complete loss of stamen and gynoecia, due to the homeotic replacement of reproductive organs by perianth whorls. Conversely, a loss of perianth whorls was obtained through the use of apetala 1–10 ("ap1"), an apetalous mutant with increased reproductive whorls and secondary apetalous inflorescences homeotically replacing most sepals. A loss of the petals and stamen expression was achieved through use of apetala 3–6 ("ap3"), a mutant in which organs of the petal and stamen whorls have been completely replaced by carpels and sepals. Although reduced carpel tissue was present, the double mutant superman-2 apetala 1–10 ("sup ap1") was preferentially selected for carpel reduction and used to characterize stamen enrichment at the expense of all other floral organs. In addition to utilizing whole-inflorescence libraries to determine organ-enriched gene expression, dissected differentiated ovules, roots, and leaves were also used to determine the ovule inclusivity of gynoecia-expressed transcripts and enrich for floral expression on a plant-wide basis.
Results and Discussion
Comparison of inflorescence expression profiles
Inflorescence transcriptome diversity, as measured by MPSS
Floral Strain or Tissue
Total Distinct Expressed Genesa
Inflorescence Enriched (%)b
Undetected within Leaf or Root (%)c
The selection of inflorescence-specific expression using a 0 TPM threshold in leaves and roots was too stringent and resulted in a high false negative rate of identification when compared to several genes with known expression patterns (data not shown). Therefore, inflorescence-enriched gene selection was accomplished by removing genes with lower expression in inflorescences than root and leaf tissues. Furthermore, only those genes expressed in inflorescences at greater than 4 TPM were included in our filtering criteria (Table 1). Using even less stringent parameters, many floral organ enriched transcripts are accurately depicted, as evident in our biological validations; however, increasing this threshold reduced the false discovery rate and increased the informatics power to accurately detect floral expression enrichment. Trends in transcript diversity across inflorescence genotypes were not significantly impacted by inflorescence enrichment. Within both raw data and filtered inflorescence enriched expression data, the highest level of transcript diversity was observed in wild type, ap1, and ap3 inflorescences and the lowest in sup ap1 and agamous (Table 1). This is likely due to the increase in reproductive organs associated in the former, and significantly decreased floral organ diversity and strictly vegetative organs in the latter. Although differences in transcriptome diversity between these groups were significant, differences in diversity among inflorescences within both groups were not statistically significant (p > 0.10).
Correlation of expression across inflorescence libraries
Library intersection based comparison (%)a
sup ap 1
Library relative based comparison (%)b
sup ap 1
The most similar expression patterns were obtained from ap1 and wild type inflorescences. The homeotic mutants ap1 and ap3 as well as ap3 and wild type were also highly correlated (Table 2). The difference between the three correlations was statistically significant (p < 0.001) and their relative rank was conserved in both the library intersection and relativity based approaches. Therefore, increases in gynoecium biomass at the expense of perianth tissue did not alter gene expression as much as the loss of reproductive organs. Moreover, the largest phenotypic differences were mirrored by the most significant differences in gene expression. The increase of stamens within the inflorescence of sup ap1 produced the greatest difference when compared to the strictly vegetative agamous inflorescences within both methods of comparison despite similar degrees of transcript diversity within the two libraries.
To identify additional alterations in expression in ap1 generated by the addition of the sup mutation and to address possible phenotypic correlations, the expression profile of sup ap1 and ap1 were compared. Although the proportion of conserved genes between sup ap1 and ap1 was greater than that of sup ap1 and wild type inflorescences, this difference was not significant (p > 0.05) within the library intersection analysis. In the library relative analysis, the similarity of gene expression among the sup ap1 and all other inflorescences was relatively low (Table 2). This suggests that increases in stamens at the expense of all other tissues altered the inflorescence expression profile more than any other mutation. Despite the substantial gene expression differences, ap1 revealed the most similar inflorescence expression pattern to sup ap1 in the library relative analysis (p < 0.001).
Identification of organ-enriched gene expression
Organ occurrence profiles
Putative organ(s) of expression
Sepal-Gynoecium, Sepal-Petal-Gynoecium, Petal-Gynoecium
Sepal-Stamen, Petal-Stamen-Gynoecium, Sepal-Stamen-Gynoecium, Sepal-Petal-Stamen, All floral organs
The ability to dissect all possible patterns of floral organ expression is limited by the diversity of the mutant libraries employed within this study (Table 3). Enriched expression within an organ or multiple organs may only be recognized if their respective organ occurrence profiles are unique. For example, sepal-enriched transcripts cannot be isolated from sepal and petal expressed transcripts, because transcripts from both spatial enrichment groups are present or up regulated within agamous and apetala3 mutants and down regulated within ap1 and sup ap1. Similar limitations were inherent within previous homeotic microarray analyses and likely led to the characterization of Apetala1, a sepal and petal expressed transcript , as expressed in a sepal-specific manner within the previous data set . In addition, organ specific prediction may be biased by additional alterations in transcription as a result of homeotic transformation. An example of this is the dependence of sepal expression on the co-existence of sepals and petals. Putative petal-enriched transcripts may also be expressed within the sepal in wild type inflorescences but not in apetala3; the mutant used to distinguish a lack of expression in the sepal and gynoecium. Bearing in mind these limitations, the presence of transcript enrichment within the organ(s) of inflorescence may only be dissected for seven instances within our MPSS analysis: sepal/sepals-petals, petals, stamens, petals-stamens, stamens-gynoecia, and gynoecia inclusive and exclusive of the ovules (see Additional file 1 for a complete list of transcripts).
A total of 138 genes were identified as possessing gynoecium-enriched expression within our analyses. This ranks in close proximity to the 161 genes characterized within the stamens and the 161 genes identified as expressed throughout both reproductive organs. The level of transcript diversity noted within the gynoecium was similar irrespective of the ovule; both gynoecium expression inclusive of the ovule and exclusive of the ovule identified a total of 69 genes. This suggests during the first twelve stages of floral development , the ovule does not express from an extensively larger subset of gynoecium-enriched genes than the stigma, style, the placenta, or carpel wall. Despite the extensive diversity noted in previous microdissection studies , only 35 genes with petal-enriched expression were identified in agamous but absent from other mutant inflorescences. Nonetheless, this was greater than the 19 transcripts identified as expressed within both the stamen and the petals. Therefore, fewer transcripts are likely to be enriched across only the petals and stamens than are specifically enriched in each organ. This contrasts with the trend noted across the reproductive organs wherein nearly equivalent numbers of genes were identified as enriched within a single organ as expressed throughout both reproductive tissues. Unlike previous comparisons that enabled dissection of enrichment within a single organ or a cluster of two organs, the 56 genes enriched within the sepal and throughout the perianth could not be dissected from each other given the mutants employed in this study. Therefore, the relative transcriptional complexity of the sepals cannot be compared to the petals. Nonetheless, it can be concluded that the number of sepal-enriched genes does not supersede those identified within any of the reproductive organs.
To further characterize the expression patterns of our organ-enriched gene sets, the temporal expression of our spatial gene sets was determined. Although little commonality was noted between independent microarray analyses of spatial Arabidopsis expression and our organ-enriched gene sets, the spatial data accrued from this study was superimposed on cumulative temporal data identified in the GENEVESTIGATOR online microarray database to delineate the temporal expression of our organ-enriched gene sets . The majority of genes with MPSS determined floral organ-enriched expression were detected in later stages of floral development after 10% of the flower buds have opened (>36 days after planting). Therefore, our organ-enrichment analyses may be skewed to detect more genes expressed near stage twelve of development, after significant biomass for all floral organs has developed.
Functional characterization of organ-enriched gene sets
Genes possessing predicted organ-enriched expression were further characterized based on gene ontology and cis-element conservation to further understand the nature of those factors underpinning organogenesis and the maintenance of specific floral organs, as well as to identify the relative impact of the quaternary complex in activating organ-enriched genes. Regulation and function were determined using several publicly available data mining tools. The abundance of organ-enriched genes with common regulators or function was compared to the mean genome-wide level of expression to identify overrepresented characteristics within genes associated with each floral organ.
Delineation of gene function enrichment was achieved using the publicly available analysis tool EasyGO . An enrichment of terms detailing specific molecular functions, as well as associations with cellular components were identified within gene sets. Few enriched terms were identified in the majority of organ-enriched gene sets, genes possessing stamen-enriched expression were enriched for lipase activity (p < 0.01) as well as presence within the endo-membrane system (p < 0.01). The lipase DEFECTIVE ANTHER DEHISCENCE has been shown to catalyze the initial step in the jasmonic acid pathway that regulates flower opening, synchronization of pollen maturation, and anther dehiscence . Therefore, it is possible additional lipase genes are involved in similar pathways to regulate stamen maturation.
To assess the role of the MADS-box proteins involved in the quaternary complex at regulating organ-enriched gene expression, several algorithms were implemented to identify overrepresented cis-elements within each set of genes. The upstream and downstream regions surrounding each gene were scanned for over-expressed motifs at 500 and 1000 bp intervals using The University of Toronto's Promoter 2 program  as well as the University of Leeds' and the Arabidopsis Gene Regulatory Information Server's, Known Cis-element Analyzer . No significantly overrepresented motifs were characterized within any of the sets of genes with floral organ-enriched expression. This confirms previous microarray functional assessments and reveals no significant enrichment of CArg boxes in organ-enriched gene sets. Therefore, much of the noted organ-enriched expression is not mediated by a single common cis-regulatory element such as the action of the putative quaternary complex on CArg boxes  during later stages of floral development. Nonetheless, the expression of over 200 miRNAs has been characterized within wild type floral inflorescence using MPSS  and it is probable post-transcriptional regulation plays a substantial role in restricting organ-enriched expression in a manner similar to that noted between miR172 and AP2 . Future analyses characterizing miRNAs in inverse expression patterns to those identified in Table 3 may enable detection of those miRNAs negatively regulating the expression of genes responsible for floral organ maintenance and development.
Validation experiments I: MPSS correspondence with microarray analyses
The data sets produced using the threshold filtration approach of MPSS data were compared to several previous microarray studies to reveal the concurrence of analyses. Previous analysis comparing homeotic mutant gene expression relative to wild type revealed significant overlap. Approximately 70% (111 genes) of those genes identified by MPSS to be stamen-enriched agreed with this previous microarray analysis; however, the number of putative stamen-enriched genes identified within the previous study was over seven-fold greater than that identified by our analyses. This difference may be due to the difference in filtering stringency and not a discrepancy in the actual number of stamen-enriched genes. MPSS analyses were set to identify only the most probable candidates for organ enrichment and likely possess a high false negative rate. Relaxed MPSS filtration criterion based only on the absence of expression within mutants lacking stamens revealed a similar number of stamen-enriched genes. However, the fraction of organ-enriched genes identified under relaxed parameters that concurred with the previous analyses was greatly reduced to approximately 26% (267 genes).
Microarray analysis from isolated floral organs  and pollen  revealed an overlap of 25 genes within the stamen-enriched data set. Moreover, all 25 stamen-enriched genes identified through the intersection of MPSS and the dissected floral organ microarray results  concur with previous microarray analyses based on comparisons of homeotic mutants . This is significantly greater than the proportion of stamen-enriched genes identified as common to both microarray studies alone (approximately 49%). Several known stamen-enriched and Apetala3 regulated genes, including Profilin5, Arabinogalactan6, Cyp703A2, Anther7, Callose Synthase5, Aborted Microspore, Taptetum1, Lipid Transfer12, and Anther27, were identified across the microarray and MPSS analyses. However, nearly half of the stamen-enriched genes identified across all studies have putative or unknown functions including genes related to a putative self-incompatibility gene, At5g26060, a putative phospholipase A2, At4g29470, involved in the jasmonic acid pathway regulating pollen maturation and several genes within the glycine-rich protein family which is known to possess high expression within the androecium.
Much less commonality was noted among organ-enriched gene sets of the gynoecium, petals, and sepal-petals (see Additional file 1). Only 15% (10 genes) of gynoecium-enriched genes as determined by MPSS were identified within previous microarray organ-enrichment studies . Nonetheless, a few genes with known gynoecium-enriched expression were characterized by MPSS that were not identified in previous microarray analyses, including three Embryo Defective and Embryo Defective-like genes. MPSS-identified petal-expressed gene sets shared only two genes with microarray analyses . Similarly, among the set of sepal/sepal-petal localized transcripts only five genes were identified as common to microarray analyses and MPSS predictions [10, 16].
The significant reduction in similarity among MPSS and microarray  predicted genes within the gynoecia and vegetative organ-enriched expression suggests comparison of homeotic mutants is less accurate in determining enrichment of genes within these organs than the stamens in one or both analyses. However, previous microarray analysis detected significantly more genes with enriched expression in the stamens than the gynoecia despite the relatively similar levels of reproductive organ transcriptome diversity noted in this study.
Validation experiments II: In situhybridizations
Within the more stringent filtering parameters used to delineate organ-enriched gene sets, the expression of a gene with unknown function, At1g54860, was characterized as stamen-enriched. In situ hybridization revealed strong expression within the microspore as well as lesser expression within the stigmatic papillae and septum [Figure 3i–j]. This implies increased the filtering parameters may increase the accuracy of analysis for organ enrichment. However, a discrepancy between stringent MPSS and in situ data was noted in the putative gynoecium-enriched At5g59810. This gene encodes a subtilase family protein and was identified strictly within the stamen in a single microspore of the tetrad by in situ hybridization [Figure 3k–l]. Nonetheless, previous publications have revealed stigma-enriched expression of At5g59810 via several microarray analyses, RNA gel blots, and in situ analysis.
Validation experiments III: GUS histochemical assays
Although not present within our MPSS filtration parameters, several genes with expression enriched in a specific mutant relative to the other floral mutants were also characterized as controls for our filter parameters and the MPSS methodology. Contrary to MPSS predictions, At1g07930, a gene with very weak expression of 5 TPM in the stamen-less ap3 mutants and no expression within the stamen-enriched sup ap1 mutant was identified within the stamens (Figure 4k). Expressed in carpel- and stamen-enriched ap1 material but absent from all other floral mutants, At1g68200 was found present within the carpel as expected (Figure 4n). The ap3 enriched At2g43100 was found to be expressed in the carpel and petals despite the nearly-undetectable expression level of only 1 TPM in the highly petal-enriched agamous mutant (Figure 4o). At3g15160 was identified by MPSS as expressed throughout all floral organs except sup ap1 at a level higher than 10 TPM; experimentally, expression was detected in the petals and gynoecia but not within the stamens (Figure 4p). GUS staining demonstrated carpel-specific activity for At3g52900, and it was also identified as highly expressed within the carpel-enriched ap3 by MPSS analysis although significant expression (9 TPM) was also noted in the carpel-devoid agamous mutant. These results suggest absence or presence of expression within a single mutant was not as accurate at detecting expression patterns in the organs of enrichment; however analysis across all four mutants improved detection abilities.
Validation experiments IV: Reverse transcription PCR validation
Validation of MPSS organ-enrichment filters was performed in five putative gynoecia expressed genes by reverse transcription PCR (Figure 5f). Expression patterns were found to correlate with those predicted by MPSS. At4g04620, a gene encoding ATG8b, a microtubule associated protein involved in autophagy demonstrated expression in the ovule as predicted by MPSS, without any expression present within sepal or petal organs. This isoform of ATG8 has been previously characterized by microarray studies  which reveal the highest level of floral expression within petal tissue, followed by stamen, with lesser amounts in sepal and carpel tissue; however, MPSS analysis reveals gynoecia enriched expression inclusive of the ovule. MPSS predictions of gynoecia enrichment were confirmed by reverse transcription PCR. No expression was detected in petal or sepal tissues, suggesting that the distinctive alternative transcript of At4g04620 is specifically expressed in the ovule (V. Pérez-España and J-Ph. Vielle-Calzada, unpublished results). In addition, At4g27860, At1g19240, and At4g14420 were also found by reverse transcription PCR to demonstrate ovule expression. No sepal- or petal-expressed transcripts were noted within these analyses. In contrast to MPSS predictions, reverse transcription PCR revealed the expression of At2g42710, a putative structural constituent of the large ribosomal subunit in both ovule and petal tissues. These many empirical analyses suggest organ enrichment was accurately identified through our analyses; however, to delineate the specificity of expression to a single organ a meta-analysis approach is necessary to further filter transcripts and attain even more robust data sets.
Our MPSS floral transcriptome analysis has dissected additional expression data to further corroborate and supplement existing spatial analyses of gene expression performed using microarrays. Numerous well-characterized genes with known expression patterns were accurately dissected as organ-enriched within our system of transcript filtration. Furthermore, the validation experiments were largely consistent with the MPSS expression data for several previously uncharacterized genes. In agreement with previous analysis , reproductive structures possess the most diverse and complex transcriptome when compared to vegetative tissue and no significant enrichment of CArG boxes was noted upstream or downstream of putative organ-enriched genes, suggesting that floral organogenesis and maintenance requires a multitude of signalling cascades as opposed to extensive direct cis-regulation by the quaternary complex. This spatial dissection of transcript expression will provide a valuable reference for future functional studies, and represents a source for developing floral organ-enriched promoters to drive transgene expression. In addition, these promoters could be coupled to a recently-described system for isolation of specific cell types by fluorescence-activated sorting , followed by more advanced methods of gene expression analysis such as Illumina's Sequencing By Synthesis; a tag-sequencing platform with ten-fold higher sensitivity than MPSS . This will enable even more detailed spatial profiling of gene expression in future floral studies.
Plant materials, tissue collection, and nucleic acid isolation
All plant materials used for MPSS and reverse transcription PCR analyses were from Arabidopsis thaliana ecotype Columbia-0. Floral inflorescences were harvested from plants grown in Pro-mix soil in a growth chamber with 16 h of light for 5 weeks at 22°C with 60% humidity. These floral tissues included inflorescence meristems as well as floral buds corresponding to the first 12 stages of development . Leaf and root tissues were obtained from the same plants grown in 16 h of light for 21 d under sterile conditions in vermiculite and perlite. An automated micro-aspirator that allows ovule isolation and harvesting was used to collect ovule samples (M. Arteaga-Vazquez, M. Arteaga-Sanchez M. and J-P. Vielle-Calzada, unpubl. results). Sepal and petal tissues utilized in reverse transcription PCR analysis were hand-dissected from ag inflorescences. All tissue samples were harvested less than 2 h after dark and frozen at -80°C prior to nucleic acid extractions. A. thaliana ecotype Columbia-0 plants utilized for transformation of promoter: GUS fusion plasmids were grown in 16 h of light for 5 weeks under the same conditions prior to floral inoculation with Agrobacterium tumefaciens.
Floral tissues utilized within the in situ hybridization validation were derived from Arabidopsis thaliana of the Landsberg erecta ecotype. This was primarily due to the increased size of floral inflorescences as compared to the Columbia-0 relative. Plants were grown in a growth chamber with light, temperature and humidity conditions similar to those implemented to grow the plants for the creation of the MPSS libraries.
RNA used for cDNA for MPSS, in situ hybridization, and RT-PCR validation was isolated using the TRIzol (Invitrogen) reagent and the manufacturer's protocol. Genomic DNA isolated for the amplification of promoter sequences was obtained using the DNeasy Minispin Column Extraction Kit (Qiagen).
Signature sequencing and genomic correspondence
MPSS was performed as previously described [22, 23]. Signatures for each floral library were produced in multiple sequencing runs and in two distinct types of sequencing reactions [22, 25]; these sequencing runs and reactions were joined to compute a single normalized abundance for each signature observed in each of the floral, root, and leaf MPSS libraries . All raw and normalized signature data have been made publicly available on the web interface . These signatures were matched to their respective loci within the A. thaliana genomic sequence. Briefly, potential MPSS signatures were computationally derived from all possible DpnII restriction sites (GATC) and 13 adjacent bases within the genome. Potential MPSS signatures located on the sense-strand corresponding to an exon, intron, exon-intron splice boundary, or present within 500 bp of the 3' end of an annotated ORF were matched with the empirically derived MPSS sequences to determine the expression level of the respective gene or pseudogene.
MPSS library filtration, floral expression cross-analysis, and sorting
All MPSS libraries implemented within this study were filtered with a "reliability" filter in order to remove potentially erroneous signatures and distinguish a subset of valid expression levels. This filter eliminates all signatures identified within only a single sequencing run across all current expression libraries. Each tissue utilized within MPSS library corresponds to a minimum of four distinct sequencing runs. Therefore, signatures not identified within any other runs are likely resultant from random MPSS sequencing errors, which have been estimated to occur at a rate of ~0.25% per base .
Once reliable MPSS expression data was accrued, libraries corresponding to floral tissue designated with "1" (organ presence) or "0" (absence) to demarcate the organ occurrence profile among the homeotic floral mutants (Table 3). In order to isolate the subset of genes expressed specifically within a given organ, the normalized signature data corresponding to each gene's expression was matched to each organ occurrence profile through the use of our publicly available library analysis (LIBAN) interface. Genes with floral organ-specific expression in a mutant lacking that specific floral organ(s) logically possesses a normalized transcription level of 0 transcripts per million assayed (TPM). In contrast, those homeotic mutants that possess the specific floral organ consequently express the transcript at a normalized rate of greater than 4 TPM and expressed at a greater rate within inflorescence libraries than leaf or root MPSS libraries.
Once a subset of genes with organ(s) specific expression was identified, they were further sorted by their level of expression within the mutants which most overexpresses the specific organ. Although this biases the data for genes of higher expression, this permitted the establishment of a relative degree of confidence in the MPSS prediction of organ-specific expression. For example, putative stamen-specific transcripts were sorted first based on their level of expression within the superman/ap1 mutant, then by their level of expression within ap1 (which also overexpresses stamens) and finally by the wild-type expression level.
In order to identify the genome-wide correlation of expression across the floral tissues assayed within this study in a relative and intersection based manner, the proportion of genes not revealing significant (p < 0.05) alterations in expression relative to the other library was normalized by the union of the two library transcriptomes. From a more qualitative threshold based prospective, genes identified as present within each library based on our threshold of 4 TPM and enrichment in the inflorescence were normalized by the union of the inflorescence libraries. Over or under expression relative to wild type was determined using a normal approximation test for difference (Z-test) in the binomial proportions of transcripts per million . This analysis was performed using Microsoft Excel and SAS JMP v. 7.0.
In situ hybridization was performed on sections of Landsberg erecta (ler) floral tissues. Tissue fixation, sectioning and embedding were performed using a modified protocol . Anti-sense probes and control sense probes were derived from the following regions of the genome: At1g33430 (5'-CGGGGAAAGCCATAATAGTGC---CAACAGCAACTGCAATG-3'); At2g19070 (5'-AAAAAGAAAGGGGGTTTGTGTT---ACTTCGGCAATGCTACTCTTG-3'); At1g54860 (5'-TGCCTATCGCTTAATTCTGCTT---GATTGTTGTGTTTTTGTGTGAA-3'); At5g59810 (5'-GACGTAAGCCCATGGTTGATGA---CGTTAGGAGTCCCATCGTCGTC-3'); At1g72290 (5'-GAGAGTAAAAACGGAGGTGGTC---CTTGAGAAAACATTGATCA-3'); At2g42940 (5'-TGTTGCAGGTACAAACTACAAA---ACAGGGACCAGATGCGATTAG-3'). These regions were amplified and cloned into pGEM-T; after which, their orientation was determined by sequencing from the T7 and SP6 promoter sites. Digoxigenin probe synthesis, antibody detection and staining were done using the manufacturer's protocol (Boehringer Mannheim). The slides were dehydrated using ethanol and xylene, and were mounted in Per-mount (Fisher Scientific). Sections were photographed through a Leitz DRB (Leica, Wetzlar, Germany) light microscope using Kodak Ektachrome 160 ASA film.
Promoter:GUS fusions and histochemical assays
To construct non-ovule promoter:GUS vectors, a ~1.5 kb promoter fragment upstream of each respective start codon was amplified from genomic DNA of A. thaliana, Col-0. Gateway Cloning Technology (Invitrogen) was implemented to insert the amplicon into pDONR221, and ultimately into the binary expression vector pK2GWFS7 wherein the promoter was used to drive the expression of the reporter gene beta-glucuronidase. Primers used in fragment amplification and plasmid recombination are: At1g20130 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTTAGGATTTATTGGTGTTTCTC-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTTGGCCACGGCTGTGGATACG-3'); At3g27025 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTTATATCTTCGATATCATCCAA-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTGAATGATTAGTTTATGAGAGA-3'); At2g42940 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTCAGCAACTCTGACAGGCACC-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTTGTTATGAATGTTGTTATATG-3'); At1g07370 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCAACAAAGTCAAACATACAGAAG-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTTTTCGTCTTAGATATTATCAG-3'); At1g27900 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTACAAAATCAAGTGGGTTATTC-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTATCGCAGAGAACACTCAAAGAACC-3'); At1g33430 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTTGTAATCATATGTTTTAGAAGC-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTCTCCGCGCCTTAGTGC-3'); At5g07550 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTGTATGCATGCGCACACAAGCC-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTTGGTGGGAAGAAGTGGGG-3'); At1g19070 (S5'-GGGGACAAGTTTGTACAAAAAAGCAGGCTCGTGAACAAAGGATTACG-3' and AS5'-GGGGACCAGTTTGTACAAGAAAGCTGGGTAACACAAACCCCCTTTCT-3'). Binary vectors were transformed into Agrobacterium tumefacieriens strain gc101. A previously described dip method  was utilized to develop transgenic Arabidopsis, as selected on Kanamycin. Two to five independent T1 plants bearing each GUS: Promoter construct were subjected to histochemical analysis for GUS activity. Transgenic and wild-type floral tissues were infiltrated using two vacuum pulses at 7 min each in GUS assay buffer (H2O, 0.1 M NaH2PO4, 10 mM Na2EDTA,. 0.5 M K3Fe(CN)6, 0.1% Triton X-100 and 0.3% 5-bromo-4-chloro-3-indoxyl-beta-D-glucuronide (X-Gluc) and incubated at 37°C for 12 h. Chlorophyll de-staining was performed with consecutive ethanol washes at 22°C. GUS stained regions were identified and fixed. Photographs were captured using an AxioVision digital camera (Zeiss) and compiled with the GNU Image Manipulation Program GIMP.
To construct the promoter: GUS vectors corresponding to ovule-specific candidate genes, promoter fragments were amplified by PCR using specific primers for each gene and inserted into a pBI101 plasmid after digestion with BamHI and HindIII (At1g27330 and At5g24420), HindIII and SalI (At1g05550), or SalI and BamHI (At5g49180 and At2g47470). Primers used for fragment amplification are: At1g27330 (S5'-GCAAGCTTGGACAGGGAAGAGAGCAT-3' and AS5'-GCGGATCCACTAGTTGCGGTCCTGAT-3'); At1g05550 (S5'-GCAAGCTTGATTGGGCCATCTCTTTTC-3' and AS5'-GCGTCGACGCTCCACCTCATCTTGAAGG-3'); At5g49180 (S5'-GCGTCGACTGGGTTTTGTTTCCTTCAGTG-3' and AS5'-GCGGATCCGTTGAACTCTCCGAGAAGG-3'); At5g24420 (S5'-GCAAGCTTGTCGTCGTCAGAGACCTTG-3' and AS5'-GCGGATCCTCTCATCGACCCAAAAGA-3'); At3g06240 (S5'-CAAAGAGCGAATTTCTCGGCTAC-3' and AS AS5'-TCTCTTGGAATCTCCGGTAGTTG-3'). For plant transformation, vectors were transferred into Agrobacterium tumefaciens strain ASE . Transformations were performed on wild-type Landsberg erecta or Col-0 by floral dip transformation . Seedlings obtained from the T1 promoter: GUS transformants were selected in MS medium with 50 μg/ml kanamycin. Dissected gynoecia were processed for GUS histology following . Microscopical observations were conducted with a DRM Leica microscope and Nomarsky optics.
Reverse transcription PCR validation experiments
cDNA synthesis was performed using 5 μg of total RNA and Superscript II Reverse Transcriptase (Invitrogen); reverse transcription PCR conditions were: 1 μg of cDNA, 1 mM forward and reverse primers (see below), 0.6 units Taq DNA polymerase (Invitrogen), 1 mL 10× PCR Buffer (Invitrogen). Primers used were: At4g04620 (S5'-AAGAGTTCCCGTGATTGTGG-3' and AS5'-AACCCGTCTTCGTCTTTGTG-3'); At1g19240 (S5'-CCTAATTGATCGGCCAGAAA-3' and AS5'-GACAAAGAAAACAGCGCACA-3'); At4g27860 (S5'-GGTGAGGAACCGAGCATAGA-3' and S5'-AGAGAAGCAACACCGCAGAT-3'); At4g14420 (S5'-TTGTCTTTGGCAGCTCATT-3' and AS5'-CCTGAGTTGCCTACCGTGTT-3'); At2g42710 (AS5'-AAGACGCAAAAGCTGGACAT-3' and 5'-GGATAACCCTTTCCCATCGT-3'). Amplification included 30 cycles of 94°C for 30 seconds, 60°C for 30 seconds, and 72°C for 32 seconds.
This research was funded in part by NSF Plant Genome Research award #0110528 (B.C.M.) and an NSF Research Experiences for Undergraduates (REU) award to Jason Peiffer. We are grateful for the bioinformatics assistance and advice of Dr. Kan Nobuta, as well as support and advice from additional members of the Meyers Lab and members of Dr. Pam Green's lab. Dan Rodriguez and Cesar Alvarez-Mejía assisted with reverse transcription PCR and ovule-specific reporter gene constructs. M.A.V. and N.S.L. were recipients of a Ph.D fellowship from Consejo Nacional de Ciencia y Tecnología (CONACyT). Research in JPVC lab was supported by Fondos Mixtos Guanajuato, CONACyT, and the Howard Hughes Medical Institute. We would like to thank Dr. Thomas Jack for providing ap3-6 seed stock, as well as the Arabidopsis Biological Resource Center at The Ohio State University for ag seed stock.
- Bowman JL, Smyth DR, Meyerowitz EM: Genetic interactions among floral homeotic genes of Arabidopsis. Development. 1991, 112 (1): 1-20.PubMedGoogle Scholar
- Coen ES, Meyerowitz EM: The war of the whorls: genetic interactions controlling flower development. Nature. 1991, 353 (6339): 31-37. 10.1038/353031a0.PubMedView ArticleGoogle Scholar
- Egea-Cortines M, Saedler H, Sommer H: Ternary complex formation between the MADS-box proteins SQUAMOSA, DEFICIENS and GLOBOSA is involved in the control of floral architecture in Antirrhinum majus. EMBO J. 1999, 18 (19): 5370-5379. 10.1093/emboj/18.19.5370.PubMedPubMed CentralView ArticleGoogle Scholar
- Honma T, Goto K: Complexes of MADS-box proteins are sufficient to convert leaves into floral organs. Nature. 2001, 409 (6819): 525-529. 10.1038/35054083.PubMedView ArticleGoogle Scholar
- Tonaco IA, Borst JW, de Vries SC, Angenent GC, Immink RG: In vivo imaging of MADS-box transcription factor interactions. Journal of experimental botany. 2006, 57 (1): 33-42. 10.1093/jxb/erj011.PubMedView ArticleGoogle Scholar
- Soltis DE, Soltis PS, Albert VA, Oppenheimer DG, dePamphilis CW, Ma H, Frohlich MW, Theissen G: Missing links: the genetic architecture of flowers [correction of flower] and floral diversification. Trends in plant science. 2002, 7 (1): 22-31. 10.1016/S1360-1385(01)02098-2. dicussion 31–24PubMedView ArticleGoogle Scholar
- Gu Q, Ferrandiz C, Yanofsky MF, Martienssen R: The FRUITFULL MADS-box gene mediates cell differentiation during Arabidopsis fruit development. Development (Cambridge, England). 1998, 125 (8): 1509-1517.Google Scholar
- Ito T, Wellmer F, Yu H, Das P, Ito N, Alves-Ferreira M, Riechmann JL, Meyerowitz EM: The homeotic protein AGAMOUS controls microsporogenesis by regulation of SPOROCYTELESS. Nature. 2004, 430 (6997): 356-360. 10.1038/nature02733.PubMedView ArticleGoogle Scholar
- Sablowski RW, Meyerowitz EM: A homolog of NO APICAL MERISTEM is an immediate target of the floral homeotic genes APETALA3/PISTILLATA. Cell. 1998, 92 (1): 93-103. 10.1016/S0092-8674(00)80902-2.PubMedView ArticleGoogle Scholar
- Wellmer F, Riechmann JL, Alves-Ferreira M, Meyerowitz EM: Genome-wide analysis of spatial gene expression in Arabidopsis flowers. The Plant cell. 2004, 16 (5): 1314-1326. 10.1105/tpc.021741.PubMedPubMed CentralView ArticleGoogle Scholar
- Yamada K, Lim J, Dale JM, Chen H, Shinn P, Palm CJ, Southwick AM, Wu HC, Kim C, Nguyen M, et al: Empirical analysis of transcriptional activity in the Arabidopsis genome. Science (New York, NY). 2003, 302 (5646): 842-846.View ArticleGoogle Scholar
- Schmid M, Davison TS, Henz SR, Pape UJ, Demar M, Vingron M, Scholkopf B, Weigel D, Lohmann JU: A gene expression map of Arabidopsis thaliana development. Nature genetics. 2005, 37 (5): 501-506. 10.1038/ng1543.PubMedView ArticleGoogle Scholar
- Rubinelli P, Hu Y, Ma H: Identification, sequence analysis and expression studies of novel anther-specific genes of Arabidopsis thaliana. Plant molecular biology. 1998, 37 (4): 607-619. 10.1023/A:1005964431302.PubMedView ArticleGoogle Scholar
- Masayuki A, Tohru A, Makoto E, Katsunori H, Chikara K, Daisuke S, Kinya T, Masao W: Identification of anther-specific genes in a cruciferous model plant, Arabidopsis thaliana, by using a combination of Arabidopsis macroarray and mRNA derived from Brassica oleracea. Sexual Plant Reproduction. 2003, 15 (5): 213-220.Google Scholar
- Zik M, Irish VF: Global identification of target genes regulated by APETALA3 and PISTILLATA floral homeotic gene action. The Plant cell. 2003, 15 (1): 207-222. 10.1105/tpc.006353.PubMedPubMed CentralView ArticleGoogle Scholar
- Ma L, Sun N, Liu X, Jiao Y, Zhao H, Deng XW: Organ-specific expression of Arabidopsis genome during development. Plant physiology. 2005, 138 (1): 80-91. 10.1104/pp.104.054783.PubMedPubMed CentralView ArticleGoogle Scholar
- Cohen BA, Mitra RD, Hughes JD, Church GM: A computational analysis of whole-genome expression data reveals chromosomal domains of gene expression. Nature genetics. 2000, 26 (2): 183-186. 10.1038/79896.PubMedView ArticleGoogle Scholar
- Wellmer F, Alves-Ferreira M, Dubois A, Riechmann JL, Meyerowitz EM: Genome-wide analysis of gene expression during early Arabidopsis flower development. PLoS genetics. 2006, 2 (7): e117-10.1371/journal.pgen.0020117.PubMedPubMed CentralView ArticleGoogle Scholar
- Birnbaum K, Shasha DE, Wang JY, Jung JW, Lambert GM, Galbraith DW, Benfey PN: A gene expression map of the Arabidopsis root. Science (New York, NY). 2003, 302 (5652): 1956-1960.View ArticleGoogle Scholar
- Zimmermann P, Hirsch-Hoffmann M, Hennig L, Gruissem W: GENEVESTIGATOR. Arabidopsis microarray database and analysis toolbox. Plant physiology. 2004, 136 (1): 2621-2632. 10.1104/pp.104.046367.PubMedPubMed CentralView ArticleGoogle Scholar
- Coughlan SJ, Meyers BC: A Comparison of Global Gene Expression Measurement Technologies in Arabidopsis thaliana. Comparative Functional Genomics. 2004, 14 (8): 1641-1653.Google Scholar
- Brenner S, Johnson M, Bridgham J, Golda G, Lloyd DH, Johnson D, Luo S, McCurdy S, Foy M, Ewan M, et al: Gene expression analysis by massively parallel signature sequencing (MPSS) on microbead arrays. Nat Biotechnol. 2000, 18 (6): 630-634. 10.1038/76469.PubMedView ArticleGoogle Scholar
- Brenner S, Williams SR, Vermaas EH, Storck T, Moon K, McCollum C, Mao JI, Luo S, Kirchner JJ, Eletr S, et al: In vitro cloning of complex mixtures of DNA on microbeads: physical separation of differentially expressed cDNAs. Proc Natl Acad Sci USA. 2000, 97 (4): 1665-1670. 10.1073/pnas.97.4.1665.PubMedPubMed CentralView ArticleGoogle Scholar
- Meyers BC, Lee DK, Vu TH, Tej SS, Edberg SB, Matvienko M, Tindell LD: Arabidopsis MPSS. An online resource for quantitative expression analysis. Plant Physiol. 2004, 135 (2): 801-813. 10.1104/pp.104.039495.PubMedPubMed CentralView ArticleGoogle Scholar
- Meyers BC, Tej SS, Vu TH, Haudenschild CD, Agrawal V, Edberg SB, Ghazal H, Decola S: The use of MPSS for whole-genome transcriptional analysis in Arabidopsis. Genome research. 2004, 14 (8): 1641-1653. 10.1101/gr.2275604.PubMedPubMed CentralView ArticleGoogle Scholar
- Meyers BC, Lee DK, Vu TH, Tej SS, Edberg SB, Matvienko M, Tindell LD: Arabidopsis MPSS. An online resource for quantitative expression analysis. Plant Physiol. 2004, 135 (2): 801-813. 10.1104/pp.104.039495.PubMedPubMed CentralView ArticleGoogle Scholar
- Smyth DR, Bowman JL, Meyerowitz EM: Early flower development in Arabidopsis. Plant Cell. 1990, 2 (8): 755-767. 10.1105/tpc.2.8.755.PubMedPubMed CentralView ArticleGoogle Scholar
- Mandel MA, Gustafson-Brown C, Savidge B, Yanofsky MF: Molecular characterization of the Arabidopsis floral homeotic gene APETALA1. Nature. 1992, 360 (6401): 273-277. 10.1038/360273a0.PubMedView ArticleGoogle Scholar
- Xin Zhou ZS: EasyGO: Gene Ontology-based annotation and functional enrichment analysis tool for agronomical species. BMC Genomics. 2007, 8: 246-10.1186/1471-2164-8-246.PubMedPubMed CentralView ArticleGoogle Scholar
- Ishiguro S, Kawai-Oda A, Ueda J, Nishida I, Okada K: The DEFECTIVE IN ANTHER DEHISCIENCE gene encodes a novel phospholipase A1 catalyzing the initial step of jasmonic acid biosynthesis, which synchronizes pollen maturation, anther dehiscence, and flower opening in Arabidopsis. Plant Cell. 2001, 13 (10): 2191-2209. 10.1105/tpc.13.10.2191.PubMedPubMed CentralView ArticleGoogle Scholar
- Toufighi K, Brady SM, Austin R, Ly E, Provart NJ: The Botany Array Resource: e-Northerns, Expression Angling, and promoter analyses. Plant J. 2005, 43 (1): 153-163. 10.1111/j.1365-313X.2005.02437.x.PubMedView ArticleGoogle Scholar
- Manfield, et al: Arabidopsis Co-expression Tool (ACT): web server tools for microarray-based gene expression analysis. Nucleic Acid Research. 2006, 504-509. 10.1093/nar/gkl204. 34 Web ServerView ArticleGoogle Scholar
- Nakano M, Nobuta K, Vemaraju K, Tej SS, Skogen JW, Meyers BC: Plant MPSS databases: signature-based transcriptional resources for analyses of mRNA and small RNA. Nucl Acids Res. 2006, 34 (suppl_1): D731-735. 10.1093/nar/gkj077.PubMedPubMed CentralView ArticleGoogle Scholar
- Honys D, Twell D: Comparative analysis of the Arabidopsis pollen transcriptome. Plant Physiol. 2003, 132 (2): 640-652. 10.1104/pp.103.020925.PubMedPubMed CentralView ArticleGoogle Scholar
- Mayfield JA, Fiebig A, Johnstone SE, Preuss D: Gene families from the Arabidopsis thaliana pollen coat proteome. Science. 2001, 292 (5526): 2482-2485. 10.1126/science.1060972.PubMedView ArticleGoogle Scholar
- Thompson AR, Doelling JH, Suttangkakul A, Vierstra RD: Autophagic nutrient recycling in Arabidopsis directed by the ATG8 and ATG12 conjugation pathways. Plant Physiol. 2005, 138 (4): 2097-2110. 10.1104/pp.105.060673.PubMedPubMed CentralView ArticleGoogle Scholar
- Birnbaum K, Jung JW, Wang JY, Lambert GM, Hirst JA, Galbraith DW, Benfey PN: Cell type-specific expression profiling in plants via cell sorting of protoplasts from fluorescent reporter lines. Nature methods. 2005, 2 (8): 615-619. 10.1038/nmeth0805-615.PubMedView ArticleGoogle Scholar
- Seo TS, Bai X, Kim DH, Meng Q, Shi S, Ruparel H, Li Z, Turro NJ, Ju J: Four-color DNA sequencing by synthesis on a chip using photocleavable fluorescent nucleotides. Proc Natl Acad Sci USA. 2005, 102 (17): 5926-5931. 10.1073/pnas.0501965102.PubMedPubMed CentralView ArticleGoogle Scholar
- Reinartz J, Bruyns E, Jing-Zhong L, Burcham T, Brenner S, Bowen B, Kramer M, Woychik R: Massively parallel signature sequencing as a tool for in-depth quantitative gene expression in all organisms. Briefings in Functional Genomics and Proteomics. 2002, 1 (1): 95-104. 10.1093/bfgp/1.1.95.PubMedView ArticleGoogle Scholar
- Meyerowitz EM: Arabidopsis thaliana. Annual review of genetics. 1987, 21: 93-111. 10.1146/annurev.ge.21.120187.000521.PubMedView ArticleGoogle Scholar
- Ming-Hsan C, Ming-Kung C, Shu-Mei P: Floral Spray Transformation Can Efficiently Generate Arabidopsis. Transgenic Research. 2000, 9 (6): 471-486. 10.1023/A:1026522104478.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.