Transcriptomic Signature of the SHATTERPROOF2 ... Technologies Transcriptomic Signature of theSHATTERPROOF2 Expression Domain Reveals the Meristematic Nature of Arabidopsis Gynoecial

Breakthrough Technologies

Transcriptomic Signature of the SHATTERPROOF2Expression Domain Reveals the Meristematic Nature ofArabidopsis Gynoecial Medial Domain1[OPEN]

Gonzalo H. Villarino, Qiwen Hu, Silvia Manrique, Miguel Flores-Vergara, Bhupinder Sehra,Linda Robles, Javier Brumos, Anna N. Stepanova, Lucia Colombo, Eva Sundberg, Steffen Heber, andRobert G. Franks*

Department of Plant and Microbial Biology (G.H.V., M.F.-V., B.S., L.R., J.B., A.N.S., R.G.F.) and Department ofComputer Science and Bioinformatics Research Center (Q.H., S.H.), North Carolina State University, Raleigh,North Carolina 27606; Università degli Studi di Milano Dip. di BioScienze, Sezione di Botanica Generale,Milan, Italy 20133 (S.M., L.C.); and Department of Plant Biology, Swedish University of Agricultural Sciences,Uppsala, Sweden 750 07 (E.S.)

ORCID IDs: 0000-0001-6241-4143 (S.M.); 0000-0002-7815-5905 (L.R.); 0000-0001-8415-1399 (L.C.); 0000-0002-3978-1030 (R.G.F.).

Plant meristems, like animal stem cell niches, maintain a pool of multipotent, undifferentiated cells that divide and differentiate togive rise to organs. In Arabidopsis (Arabidopsis thaliana), the carpel margin meristem is a vital meristematic structure that generatesovules from the medial domain of the gynoecium, the female floral reproductive structure. The molecular mechanisms that specifythis meristematic region and regulate its organogenic potential are poorly understood. Here, we present a novel approach toanalyze the transcriptional signature of the medial domain of the Arabidopsis gynoecium, highlighting the developmental stagesthat immediately proceed ovule initiation, the earliest stages of seed development. Using a floral synchronization system and aSHATTERPROOF2 (SHP2) domain-specific reporter, paired with FACS and RNA sequencing, we assayed the transcriptome of thegynoecial medial domain with temporal and spatial precision. This analysis reveals a set of genes that are differentially expressedwithin the SHP2 expression domain, including genes that have been shown previously to function during the development ofmedial domain-derived structures, including the ovules, thus validating our approach. Global analyses of the transcriptomic dataset indicate a similarity of the pSHP2-expressing cell population to previously characterized meristematic domains, furthersupporting the meristematic nature of this gynoecial tissue. Our method identifies additional genes including novel isoforms,cis-natural antisense transcripts, and a previously unrecognized member of the REPRODUCTIVE MERISTEM family oftranscriptional regulators that are potential novel regulators of medial domain development. This data set provides genome-wide transcriptional insight into the development of the carpel margin meristem in Arabidopsis.

The seed pod of flowering plants develops from thegynoecium, the female reproductive structure of theflower (Seymour et al., 2013). The gynoecium generatesthe ovules (the precursors of the seeds) and developsinto the edible fruit in many fruiting species. As an

estimated two-thirds of the calories of humankind’sdiet are derived from gynoecia and seeds, the gynoe-cium is a globally vital structure (Oram and Brock,1972; Singh and Bhalla, 2007).

In the flowering plant Arabidopsis (Arabidopsisthaliana), the gynoecium is a morphologically complex,multiorgan structure with a diversity of tissues and celltypes (Sessions and Zambryski, 1995; Bowman et al.,1999; Seymour et al., 2013). The mature gynoeciumdisplays morphological and functional differentiationalong apical-basal, medio-lateral, and adaxial-abaxial(inner-outer) axes. Stigmatic and stylar tissue form atthe apex of the gynoecium, where the pollen grains arereceived and germinate. The stigma and style alsoconstitute the apical-most portion of the transmittingtract, a structure that allows the pollen tube cell andsperm cells to reach the internally located female ga-metophytes (Sessions and Zambryski, 1995; Sessions,1999; Crawford and Yanofsky, 2008; Fig. 1, A–C). Lo-cated basal to the stigmatic and stylar tissue is the ovaryportion of the gynoecium.

Ovules form within the ovary from a meristematicstructure termed the medial ridge or carpel margin

1 This work was supported by the National Science Foundation(grant no. IOS–1355019 to R.G.F. and S.H.) and the FP7–PEOPLE–2013–IRSE FRUIT LOOK program (to E.S., L.C., and R.G.F.).

* Address correspondence to [email protected] author responsible for distribution of materials integral to the

findings presented in this article in accordance with the policy de-scribed in the Instructions for Authors (www.plantphysiol.org) is:Robert G. Franks ([email protected]).

R.G.F., G.H.V., E.S., and S.H. coordinated and conceived of thestudy; A.S., J.V., and L.R. built the GAL4/pUAS-based two-component reporter system; G.H.V., M.F.-V., and B.S. performedFACS-sorting, wet-lab, and microscopy analysis; S.M. and L.C. per-formed in situ hybridization; G.H.V., Q.H., and S.H. performed RNA-seq and related bioinformatic and computational analyses; G.H.V.and R.G.F. interpreted the data and drafted the article; all authorsrevised and approved the final article.

[OPEN] Articles can be viewed without a subscription.www.plantphysiol.org/cgi/doi/10.1104/pp.15.01845

42 Plant Physiology�, May 2016, Vol. 171, pp. 42–61, www.plantphysiol.org � 2016 American Society of Plant Biologists. All Rights Reserved. www.plantphysiol.orgon May 31, 2018 - Published by Downloaded from

Copyright © 2016 American Society of Plant Biologists. All rights reserved.

http://orcid.org/0000-0001-6241-4143

http://orcid.org/0000-0002-7815-5905

http://orcid.org/0000-0001-8415-1399

http://orcid.org/0000-0002-3978-1030

mailto:[email protected]

http://www.plantphysiol.org

mailto:[email protected]

http://www.plantphysiol.org/cgi/doi/10.1104/pp.15.01845


meristem (CMM), located in medial portions of thegynoecium (Bowman et al., 1999; Alvarez and Smyth,2002; Reyes-Olalde et al., 2013; Fig. 1, A–C). Plantmeristems are analogous to animal stem cell niches, asthey maintain a set of undifferentiated cells that candivide and differentiate into numerous tissues and celltypes (Aichinger et al., 2012). Early during floral de-velopment, patterning events divide the gynoecial pri-mordium into a medial domain that contains theCMM and lateral domains that will form the walls ofthe gynoecium (Bowman et al., 1999). These domainsexpress different sets of transcriptional regulators fromearly developmental time points.

Many genes that play a role in the development of theCMM and in the generation of ovules from this struc-ture have been analyzed previously (Reyes-Olaldeet al., 2013). However, due to the complexity of thedeveloping gynoecium and the heterogeneity of thegynoecial tissues, the ability to analyze the tran-scriptomic signature of the developing CMM or evenother specific developing gynoecial structural domainshas been limited. Wynn et al. (2011) previously evalu-ated the transcriptional properties of the gynoecialmedial domain using hand-dissected gynoecial sam-ples from the seuss aintegumenta (seu ant) doublemutants that display a loss of many medial domain-derived structures, including ovules. They identified210 genes displaying reduced expression in seu antgynoecia from floral stages 8 to 10 (stages according toSmyth et al. [1990]). Many of these genes were shownvia in situ hybridization to be preferentially expressedin the developing medial domain of the wild-type gy-noecium, and several of these genes have been shownto function during the development of ovules from themedial domain (Reyes-Olalde et al., 2013). However, itis difficult with this approach to obtain samples fromgynoecia younger than stage 8 and, thus, to assay theearliest gynoecial patterning events.

An alternative approach to investigate the tran-scriptional properties of specific cellular populationsutilizes FACS of protoplasted cells to isolate specificcell populations based on patterns of gene expression.This approach has been applied successfully to theArabidopsis shoot apical meristem (SAM; Yadav et al.,2009, 2014) and roots (Birnbaum et al., 2003, 2005;Figure 1. A system for the collection of temporally and spatially re-

stricted cell populations from the Arabidopsis gynoecium. A, Micro-scopic image of a mature wild-type Arabidopsis gynoecium. The stigma(stg), style (sty), carpel valve (cv), abaxial replum (abr), gynophore (gn),and ovary (ovy) are false colored. B, False-colored confocal cross sectionof a stage 8 gynoecium. Medial and lateral domains of the Arabidopsisgynoecium are indicated. The CMM/medial ridge is false colored pink. C,False-colored stage 11 cross section. Ovules (ov), septum (s), and carpelvalves (cv) are indicated. D, Confocal microscope image of the pSHP2-YFP two-component reporter in the pAP1-AP1::GR; ap1; cal background.YFP expression from the pSHP2-YFP reporter is confined chiefly to themedial domain of the gynoecium at late stage 7/early stage 8, althoughweak, nonmedial domain expression can be detected in portions of thestamens. Sepals (Se) and stamens (St) are labeled. E, Z-stack compositethree-dimensional projection image of a gynoecium isolated from theflower at midstage 8. YFP expression from the pSHP2-YFP reporter isdetected in the medial domain and at the apex of the gynoecium. F,

Chloral hydrate image of an inflorescence of a pAP1-AP1::GR; ap1; calplant after mock treatment. Inflorescence-like meristems do not transitionto floral meristems. G, Chloral hydrate image of an inflorescence of apAP1-AP1::GR; ap1; cal plant 125 h after spray application of dexa-methasone synthetic hormone. Samples were enriched for stages 6 to 8.H, Percentage of flowers at a given stage from inflorescences used forFACS sorting. Stages 6, 7, 8p (preovules), and 8s (postovules) are indicatedon the x axis as St6, St7, St8p, and St8s, respectively. Stage 8p is before anyvisiblemorphologicalmanifestation of ovule primordia upon observationunder differential interference contrast (DIC) microscopy. Stage 8s ovuleprimordia were observed and were at ovule stage 1-I or 1-II according toSchneitz et al. (1995). I, Confocal microscopy of YFP fluorescence ofprotoplasted cells after FACS. A to C are adapted from Azhakanandamet al. (2008) with permission.

Plant Physiol. Vol. 171, 2016 43

Domain-Specific Transcriptomic Analysis

www.plantphysiol.orgon May 31, 2018 - Published by Downloaded from Copyright © 2016 American Society of Plant Biologists. All rights reserved.


Carter et al., 2013; Lan et al., 2013) as well as to devel-oping cell lineageswithin theArabidopsis leaf epidermis(Adrian et al., 2015).

Here, we developed a novel FACS-based system forthe transcriptomic analysis of a specific cellular popu-lation from the developing gynoecium, specifically thepopulation of cells expressing the transcriptional reg-ulator SHATTERPROOF2 (SHP2). SHP2 encodes aMADS domain transcription factor that is expressedearly within the developing CMM and, thus, functionsas a marker for the meristematic population of cells thatgenerate the transmitting tract and ovules (Ma et al.,1991; Savidge et al., 1995; Colombo et al., 2010; Larssonet al., 2014). In order to focus our analysis on earlystages of gynoecium development during which keypatterning events occur, we generated a SHP2 domain-specific reporter in a genetic background that allowedthe synchronization of floral development. This, cou-pled with FACS-based protoplast sorting proceduresand RNA sequencing (RNA-seq), provided a uniquetemporal and spatial precision to assay the transcriptionalsignature of the gynoecial SHP2 expression domain.

Our system provides the ability to isolate largenumbers of cells from a temporally and spatially re-stricted gynoecial domain. We apply this method toinvestigate the transcriptomic signature of the medialdomain of the gynoecium at the developmental stageswhen key patterning events and ovule initiation occur.Our analysis reveals many genes that are expressed pref-erentially within the developing medial portions of thegynoecium, including members of the REPRODUCTIVEMERISTEM (REM) family of transcriptional regu-lators (Swaminathan et al., 2008; Romanel et al.,2009). We also take advantage of strand-specificRNA-seq technology to find protein coding genesand noncoding RNAs as well as to examine isoformsand naturally occurring antisense transcripts thatare expressed preferentially in the medial domain. Thiswork complements and extends previous analyses ofmedial domain development and generates a list of po-tential novel regulators of medial domain developmentthat are strong candidates for future functional analyses.Furthermore, global analyses of the transcriptomic dataset indicate a similarity of the pSHP2-expressing cellpopulation to previously characterized meristematicdomains, further supporting the meristematic nature ofthis gynoecial tissue.

RESULTS AND DISCUSSION

FACS-Based Protoplast Sorting Allows the Collection ofthe SHP2-Expressing Cell Population from a TemporallyRestricted Inflorescence Sample

The transcriptional regulator SHP2 is preferentiallyexpressed in the medial domain of the gynoecium andin a subset of the medial domain-derived tissues(Savidge et al., 1995; Colombo et al., 2010; Larsson et al.,2014; Fig. 1, D and E). SHP2 plays an important role in

the development of the medial domain and in thespecification of ovule identity (Liljegren et al., 2000;Favaro et al., 2003; Pinyopich et al., 2003; Colomboet al., 2010; Galbiati et al., 2013). To better characterizethe molecular mechanisms of the medial domain andovule development, we sought to identify transcriptsthat are differentially expressed within the medial do-main of the Arabidopsis gynoecium relative to the restof the inflorescence. To enable this, we generated atransgenic line containing a two-component reportersystem in which a pUAS-3xYPET reporter was drivenby a pSHP2-GAL4 driver construct (see “Materials andMethods”). Throughout this article, we refer to this two-component reporter as pSHP2-YFP (YFP for yellow flu-orescent protein).

To better understand the early specification of medialand lateral gynoecial domains and in the earliest stagesof ovule primordium initiation, we focused our tran-scriptomic analysis on floral stages 6 to 8, whenthese key developmental events occur (Bowman et al.,1999). In order to increase our ability to collect a largenumber of pSHP2-YFP-expressing cells from this spe-cific bracket of developmental stages, we crossed thepSHP2-YFP reporter into an ap1 cal-based floral syn-chronization system that allows the collection of largenumbers of semisynchronized flowers at roughly thesame developmental stage (Wellmer et al., 2006;Ó’Maoiléidigh and Wellmer, 2014). The expression ofthe pSHP2-YFP reporter in the floral synchronizationsystem was largely similar to that observed in wild-type inflorescences (Ma et al., 1991; Colombo et al.,2010; Larsson et al., 2014) and was confined chiefly tothe medial domain and medial domain-derived tissues(Fig. 1, D and E). Some expression was observed innonmedial gynoecial positions. The most apparent ofthis was within the apex of the developing gynoecium,where the pSHP2-YFP reporter is detected in bothmedial and lateral positions. Additionally, expressioncould be observed in a small number of cells within thestamens (Fig. 1D) and occasionally on the edges of se-pals that appeared to have undergone a homeotictransformation toward a carpelloid fate (data notshown). Thus, the vast majority of the pSHP-YFP re-porter expression reflected the endogenous pSHP2 ex-pression domain (in the medial and apical portions ofthe gynoecium). We cannot exclude the possibility thata small minority of the expression domain may reflectectopic expression of the reporter due to the geneticbackground or transgene insertion site or limitations ofthe regulatory sequences used in the pSHP-YFP re-porter construct.

Microscopic examination of our semisynchronizedinflorescence samples indicated that flowers rangedbetween floral stage 1 and early stage 8, with a strongenrichment for floral stages 6 through early 8 (Fig. 1, GandH). Flowers that had developed beyond late stage 8were not detected in our samples. Thus, our biologicalsample is strongly enriched for transcripts that areexpressed during early patterning of the gynoeciumand the earliest stages of ovule development (initiation)

44 Plant Physiol. Vol. 171, 2016

Villarino et al.



and does not include later floral developmental stages,where SHP2 is expressed in stigma, style, and valvemargin tissues. Additionally, as the initial expression ofthe pSHP-YFP reporter is detected at late stage 5 or earlystage 6 (Larsson et al., 2014), we expect that the popu-lation of YFP-expressing protoplasts derived from thismaterial will be highly enriched with cells from thestage 6 to 8 medial domain.FACS sorting of protoplasts derived from these inflo-

rescences yielded three populations of sorted cells(collected in biological quadruplicate): YFP-positive,YFP-negative, and all-sorted cells (Supplemental Fig.S1). The all-sorted sample included all protoplasts re-covered (regardless of YFP expression) after sorting gateswere applied to remove debris and broken cells (see“Materials and Methods”). We additionally collected(also in biological quadruplicate) nonsorted samplesfrom entire nonprotoplasted inflorescences to measurethe abundance of transcripts in the biological startingmaterial before protoplast generation and FACS sorting.In order to evaluate the purity of the YFP-positive pro-toplasts during a preliminary FACS run, YFP-positivecells were resorted. Ninety-six percent of the YFP-positive cells were found to resort into the YFP-positivegate, indicating a high degree of enrichment and purityin the YFP-positive sample (Supplemental Fig. S1).Confocal microscopy also revealed an enriched popula-tion of intact YFP-positive protoplasts after FACS (Fig. 1I).We used quantitative real-time (qRT)-PCR to esti-

mate the degree of enrichment of the endogenous SHP2and NGATHA1 (NGA1) transcripts in RNA samplesderived from the YFP-positive and YFP-negative sam-ples. NGA1 is expressed in the adaxial portions of thegynoecium starting at stage 7 in a domain that partiallyoverlaps with the SHP2 expression domain (Alvarezet al., 2009; Trigueros et al., 2009) and thus provides anadditional benchmark to estimate the enrichment ofmedial domain-expressed transcripts. The normalizedlevel of the SHP2 transcript was approximately 30-foldhigher in the YFP-positive samples relative to theYFP-negative samples (P , 0.001), while the NGA1transcript was approximately 4-fold higher in theYFP-positive samples (P , 0.05). The difference in thelevels of TUBULIN6 (TUB6) was not found to be sta-tistically significant (P = 0.4) between the YFP-positiveand YFP-negative samples (Supplemental Fig. S2).

Transcriptomic Analysis of the Gynoecial SHP2 ExpressionDomain and Identification of Candidate Regulators ofGynoecial Medial Domain Development

To investigate the transcriptomic profile of thegynoecial SHP2 expression domain, we performedhigh-throughput RNA-seq from the collected proto-plasts and nonprotoplasted inflorescence samples. Weexpect that the identification of differentially expressedgenes (DEGs) between the YFP-positive and YFP-negative samples (referred to as YFP+/2) will provideinsight into the set of transcripts differentially expressed

in the gynoecial medial domain relative to the restof the inflorescence. Additionally, DEGs identified inthe all-sorted and nonsorted comparison (referred to asall-sorted/nonsorted) are expected to reveal transcriptsthat are differentially represented as a result of theprotoplasting/FACS-sorting protocol.

Two lanes of the HiSeq2500 Illumina sequencingplatform yielded 320 million raw reads, with an averageof 20 million reads per library. Nearly 11 million readswere filtered out after removing barcode adapters andlow-quality sequences. The remaining 306 million readswere aligned against The Arabidopsis Information Re-source (TAIR) 10 reference genome (Lamesch et al.,2012), with more than 90% of them successfully mappingto the genome sequence. Among the mapped reads, 244million reads mapped uniquely to only one location andwere used for subsequent analyses. A detailed break-down is shown in Supplemental Table S1.

We used three different programs to determineexpressed and differentially expressed protein-codinggenes in our data set: Cufflinks (Trapnell et al., 2012),edgeR (Robinson et al., 2010), and DESeq2 (Love et al.,2014; see “Materials and Methods”); non-protein-coding gene models were considered separately andare presented below. These open-source programs areamong the most cited RNA-seq analysis tools used forthe identification of DEGs. Various studies suggest that,among the different RNA-seq analysis tools, no singlemethod outperforms the other ones (Rapaport et al.,2013; Soneson and Delorenzi, 2013). In our study, wecompared the results of edgeR, DESeq2, and Cufflinksusing our data set (Fig. 2). Our choice of these threetools reflects significant differences in the underlyingalgorithms. Cufflinks uses isoform expression esti-mates, while edgeR and DESeq2 compute gene level-based expression estimates. Furthermore, edgeR andDESeq2 use different normalization approaches anddispersion estimates. As these three programs use dif-ferent algorithms, we expected each program to returna somewhat different set of DEGs. Here, the term DEGis used to indicate a gene whose steady-state transcriptlevel differs significantly at a false discovery rate(FDR) , 0.001 and shows a fold change of 4 or morebetween the two compared RNA samples.

We reasoned that any gene identified as a DEG by allthree programs was highly likely to be differentiallyexpressed in our data set. Thus, to identify potentialregulators of gynoecial medial domain development,a stringent criterion was used to select a subset of theYFP+/2 DEGs for downstream analysis. For a gene tobe selected from the YFP+/2 comparison, we requiredthat the transcript be identified as differentially expressedby all three independent software packages. This led tothe identification of 411 triply identified DEGs (e.g. theoverlap sections in the Venn diagram in Fig. 2B).

Additionally, we sought to confirm that these 411genes were not differentially expressed as a result of theprotoplasting and sorting procedure. We reasoned thatDEGs identified as differentially expressed between theall-sorted and nonsorted samples would characterize




http://www.plantphysiol.org/cgi/content/full/pp.15.01845/DC1






the set of genes whose expression was altered by theprotoplasting and sorting procedure. As a conservativeestimate of the protoplast-induced gene set, we identi-fied transcripts in the union set of all the nonsorted/all-sorted DEGs (the union of the three software packages)even if they were identified by only one software pro-gram (Fig. 2A). This identified 880 genes as differen-tially expressed as a result of the protoplastingprocedures. If the 411 triply identified YFP+/2 DEGstruly represent a signature of the medial domain, thenwe expect a relatively small overlap between the setof DEGs identified in the two comparisons. Only48 transcripts were found in common between theYFP+/2 DEGs and the all-sorted/nonsorted DEGs(Fig. 2C), indicating a high degree of specificity in theDEGs identified in each comparison. We then removedthese 48 transcripts from our analysis to eliminate anythat might be differentially expressed as a result of theprotoplast generation or FACS-sorting procedures,leaving 363 cleaned protein-coding DEGs from theYFP+/2 comparison (Fig. 2C). We note that this geneset is a rather conservative estimate of the set of DEGs,as some of the 48 overlapping transcripts that wereremoved may indeed by differentially expressed be-tween the YFP-positive and YFP-negative samples.Additionally, genes that display a statistically signifi-cant differential expression between the YFP-positiveand YFP-negative samples at a magnitude of less than4-fold may still be biologically significant. The ex-pression profiles of these 363 YFP+/2 DEGs are rep-resented in a heat map (Supplemental Fig. S3). Thisgene set includes 95 DEGs whose transcript levelswere higher in the YFP-positive samples (enriched)and 268 DEGs whose transcript levels were lower(depleted) in the YFP-positive samples relative to theYFP-negative samples (Supplemental Table S2).

For the 95 DEGs that were enriched in the YFP-positive sample (at a fold change . 4), we expectedmany to be preferentially expressed in the medial por-tions of the gynoecium at floral stages 6 to 8. To test this,we examined the literature to determine the expressionpatterns of members of this gene set. From the top 15 ofthe 95 YFP-positive enriched DEGs (ranked by foldchange), five have been reported previously to be pref-erentially expressed in the gynoecial medial domain viain situ or reporter gene analysis (i.e. HECATE1 [HEC1],HEC2, SHP1, SHP2, and STYLISH1 [STY1]; Ma et al.,1991; Savidge et al., 1995; Kuusk et al., 2002; Gremskiet al., 2007; Colombo et al., 2010) and three others weredescribed previously as enriched in medial domain-derived tissues in published transcriptomic datasets (i.e. AT1G66950, AT5G14180, and AT1G03720;Skinner and Gasser, 2009; Wuest et al., 2010; Table I).An additional gene from this list, CRABS CLAW(CRC), has been shown via in situ hybridization to beexpressed in portions of the medial gynoecial domainaswell as nonmedial portions of the gynoecium (Bowmanand Smyth, 1999; Azhakanandam et al., 2008). Theexpression pattern of the remaining six genes from thisgene list have not yet been assayed in the gynoecium.Thus, as predicted, the set of 95 genes enriched inthe YFP-positive sample is enriched for genes that arepreferentially expressed in the gynoecialmedial domain.

Published functional analyses of HEC1, HEC2, SHP1,SHP2, and STY1 indicate that these genes function dur-ing the development of the medial domain or medialdomain-derived tissues (Kuusk et al., 2002; Favaro et al.,2003; Pinyopich et al., 2003; Gremski et al., 2007;Colombo et al., 2010). Many other genes in the set of 95DEGs enriched in the YFP-positive sample have beenshown previously to play a role in medial domain de-velopment (e.g. NGA family members [Alvarez et al.,2009; Trigueros et al., 2009], SPT [Heisler et al., 2001], andCUC2 [Kamiuchi et al., 2014]). Other genes within thislist are interesting candidates for future functionalstudies. These include members of the REM family oftranscriptional regulators (Swaminathan et al., 2008;Romanel et al., 2009), several auxin synthesis- or signaling-related genes, such as LIKE AUXIN RESISTANT1 (LAX1[AT5G01240]; Bennett et al., 1996) and YUCCA4 (YUC4[AT5G11320]; Cheng et al., 2006), as well as transcriptionfactors regulating other developmental processes, suchas MATERNAL EFFECT EMBRYO ARREST3 (MEE3[AT2G21650]; Pagnussat et al., 2005) and GLABROUS3(AT5G41315; Payne et al., 2000).

It is important to note that the 48 DEGs that wereidentified in both the YFP+/2 and all-sorted/nonsortedcomparisons (Fig. 2C) should not be discounted as po-tential medial domain regulators. These genes may beboth preferentially expressed in the YFP-positive cellpopulation and induced in response to the protoplastingprocedure (Supplemental Table S2). Indeed, some ofthese genes, including the transcription factors HEC3and BRASSINOSTEROID-ENHANCED EXPRESSION1,have been reported to be preferentially expressed inmedial domain-derived tissues and to function in

Figure 2. Venn diagrams of DEGs using Cufflinks, edgeR, and DESeq2(FDR , 0.001 and fold change . 4). A, Venn diagram showing DEGsidentified between the all-sorted/nonsorted samples with the threeprograms used for differential expression analysis of RNA-seq expres-sion profiles. B, Venn diagram showingDEGs between YFP+/2 samplesidentified in the three programs. C, Intersectionof theDEGs (48) frombothdata sets (A and B). A total of 363 DEGs, after removing DEGs induced bythe protoplasting/FACS-sorting stress, were used for downstream analysis.


Villarino et al.






gynoecium development (Gremski et al., 2007; Crawfordand Yanofsky, 2011). However, we chose to use thecleaned set of 363 YFP+/2 DEGs for downstreamanalyses in order to reduce the likelihood of the inclu-sion of genes whose expression was significantly al-tered by the protoplasting process.

REM Family Members Are Differentially Expressed in theSHP2 Expression Domain

In order to look for enriched categories of transcrip-tion factors within the set of 363 cleaned YFP+/2DEGs(Fig. 2C), we used the online Transcription Factor En-richment Calculator (https://dgrinevich.shinyapps.io/ShinyTF). Members of the ABI3/VP1 transcriptionfactor family that includes the REM and NGA familytranscription factors were found to be statisticallyoverrepresented (Supplemental Table S7; corrected P ,9.97E-06). The REMs belong to the plant-specific B3superfamily of transcription factors, and the expressionof many REM family members is observed in meriste-matic tissues, such as the inflorescence meristem, floralmeristem, and CMM (Franco-Zorrilla et al., 2002;Swaminathan et al., 2008; Romanel et al., 2009; Wynnet al., 2011; Mantegazza et al., 2014a, 2014b). The nu-merical designations used to describe the REM familymembers in this article are taken from Romanel et al.(2009). In our study, six REMmembers were among the363 statistically significant YFP+/2 DEGs; five werefound to have enriched expression in the YFP-positivesample, while one, REM25 (AT5G09780), was ap-proximately 4-fold less abundant in the YFP-positivesample. The REM13 (AT3G46770) transcript level isenriched approximately 12-fold in the pSHP2-YFP-expressing cells. REM13 was predicted previously tobe preferentially expressed in the inner integument,

ovule primordia, and medial domain based on tran-scriptomic data (Skinner and Gasser, 2009). We em-ployed in situ hybridization to assay the expressionpattern of the REM13 transcript during gynoecial de-velopment (Fig. 3). Using a REM13 antisense probe, wedetected signal in the medial portions of the gynoeciumcorresponding to the CMM as early as stage 7. Ex-pression also was observed in the initiating ovule pri-mordia in stage 8 gynoecia and then continued to bedetected in portions of the ovules at later develop-mental stages.

REM34/ATREM1 (AT4G31610; Franco-Zorrilla et al.,2002; Romanel et al., 2009), REM36 (AT4G31620;Mantegazza et al., 2014b), and VERDANDI (VDD/REM20; Matias-Hernandez et al., 2010; Mantegazzaet al., 2014b) also displayed enriched expression levelsin the YFP-positive sample of approximately 8-, 9-, and6-fold, respectively. Published in situ hybridizationpatterns indicate enriched medial domain expressionpatterns forREM34/ATREM1 andVDD/REM20 (Franco-Zorrilla et al., 2002;Matias-Hernandez et al., 2010;Wynnet al., 2011). Additionally, the expression of AT5G60142,a previously unnamed member of the REM family, isenriched approximately 11-fold in the YFP-positivesample (Supplemental Table S3). AT5G60142 is an in-teresting candidate for functional studies that is locatedon chromosomeV in tandemwithREM11 (AT5G60140)and REM12 (AT5G60130) and shares a high degree ofsequence similarity with these two genes as well asREM13 (Romanel et al., 2009; Mantegazza et al., 2014b).We propose to designate AT5G60142 as REM46.

Gene Set Enrichment Analysis

To gain global insights into underlying biologicalmechanisms of medial domain development and

Table I. Top 15 DEGs enriched in the YFP-positive sample as ranked by fold change

Arabidopsis gene identifiers are shown in the first column, gene names (TAIR 10 annotation) are shown in the second column, and the third columnshows the reference for each available reporter line and/or in situ hybridization (NA, not applicable). Average RPKM values are indicated for eachsample (YFP_NEG = YFP negative and YFP_POS = YFP positive).

Gene Identifier Gene Name Reference for Expression

YFP_NEG Average

Expression

YFP_POS Average

Expression

Fold Expression

Difference P FDR

AT3G50330 HEC2 Gremski et al. (2007) 1.267 80.148 67.0 1.26E281 5.36E278

AT5G67060 HEC1 Gremski et al. (2007) 0.708 40.298 59.4 1.44E2105 2.44E2101

AT1G66950 ATPDR11 Wuest et al. (2010) 0.136 5.931 45.2 9.47E276 2.02E272

AT3G58780 SHP1 Ma et al. (1991) 1.678 54.233 33.9 1.06E252 7.54E250

AT5G17040 Unknown NA 0.272 7.453 28.0 1.06E229 1.24E227

AT1G06920 OFP4 NA 0.453 11.267 25.9 6.18E234 1.04E231

AT3G51060 STY1 Kuusk et al. (2002) 3.141 73.227 24.5 2.22E298 1.89E294

AT5G14180 MPL1 Skinner and Gasser (2009) 1.348 25.806 20.1 9.86E252 6.22E249

AT2G42830 SHP2 Ma et al. (1991) 1.642 31.392 20.1 1.58E274 2.99E271

AT1G03720 Unknown Skinner and Gasser (2009) 1.898 31.968 17.7 6.66E252 4.54E249

AT2G33850 Unknown NA 0.676 10.904 16.7 1.85E219 7.15E218

AT1G69180 CRC Bowman et al. (1999) 19.435 296.675 16.0 2.60E278 8.84E275

AT2G22460 Unknown NA 0.727 10.899 15.6 2.85E217 8.52E216

AT2G21650 MEE3 Pagnussat et al. (2005);Baxter et al. (2007)

1.333 17.643 13.9 1.65E216 4.37E215

AT2G24540 AFR NA 0.84 11.029 13.8 2.27E224 1.61E222




https://dgrinevich.shinyapps.io/ShinyTF





function, Gene Set Enrichment Analysis (GSEA) wasperformed for the 95 YFP-positive enriched DEGs inthe medial domain. This analysis identified 147 GeneOntology (GO) terms that were statistically overrepre-sented (P , 0.01), including gynoecium development(GO:0048467), flower development (GO:0009908), re-sponse to gibberellin (GO:0009739), and auxin homeo-stasis (GO:0010252; Fig. 4; Supplemental Table S6). ThisGSEA analysis further suggests that the set of 95 genesenriched in the YFP-positive sample function as regu-lators of medial domain development.

In contrast, when performing GSEA with DEGsidentified between the all-sorted/nonsorted samples,a different set of 304 overrepresented GO terms wereidentified, including response to stress (GO:0006950)and response to wounding (GO:0009611), suggestingthat many of the genes identified as differentiallyexpressed between the all-sorted/nonsorted samplesreflect stress-induced changes in gene expression dur-ing protoplast/FACS sorting.

The Transcriptomic Signature of the SHP2-Expressing CellPopulation Shares Commonalities with the TranscriptionalSignatures of Other Meristematic Samples

In order to gain insight into the characteristics of the363 YFP+/2DEGs identified from the SHP2 expressiondomain, we compared the expression profile of this setof genes across several different tissues. Using Spear-man rank correlation analysis, we compared our dataset with existing Arabidopsis RNA-seq transcriptomicdata sets from whole flowers (Mizzotti et al., 2014) andaerial seedling tissues (Gene ExpressionOmnibus [GEO]accession no. GSE54125) as well as from laser-capture

microdissected inflorescence meristems, floral meri-stems, and stage 3 flowers (Mantegazza et al., 2014a).In the sample-wise hierarchical clustering (Fig. 5A), thetranscriptomic profiles from the SHP2-expressing (YFP-positive) sample clustered more closely with the meri-stematic samples, while the YFP-negative and all-sortedsamples clustered more closely with the whole-flowerand whole-seedling samples. This suggests that the ex-pression signature of the YFP-positive sample is moresimilar to that of the floral and inflorescence meristemsand young flowers than it is to whole flowers or youngvegetative seedlings (Fig. 5A).

Further supporting the similarity of the SHP2-expressing domain to other meristematic samples,the expression levels of GA20OX1 (AT4G25420) andGA20OX2 (AT5G51810) were both significantly de-pleted in the YFP-positive sample relative to the YFP-negative sample (Supplemental Table S10). GA20OX1and GA20OX2 encode key biosynthetic enzymes of theplant hormone GA (Phillips et al., 1995). Levels of ex-pression of GA20OX1 and GA20OX2 are low in theSAM relative to expression in the juxtaposed youngorgan primordia, and high levels of GA synthesis in-terfere with the maintenance of meristematic fate in theSAM (Hay et al., 2002; Jasinski et al., 2005). These datasuggest that low levels of GA also may be associatedwith the meristematic nature of the CMM. Althoughnot discussed here, the expression values of genes an-notated with a role in ethylene signaling are found inSupplemental Table S10.

We additionally compared the medial domain tran-scriptional signature with data sets generated with theAffymetrix ATH1 array, allowing comparisons withtranscriptomic signatures of a variety of cell types, in-cluding vascular and meristematic cell types from the

Figure 3. The candidate medial domain regulator REM13 (AT3G46770) is expressed within the medial gynoecial domain anddeveloping ovules (ov). Results are from an RNA in situ hybridizationwithREM13 probe. A toD, Antisense probes. E, Sense strandprobe. A, Hybridization signal is detected in the CMM (adaxial portions of the medial gynoecial domain) in a stage 7 longitudinalsection. B to D, In transverse gynoecial sections, REM13 expression is detected in the ovule primordia; stage 7 (B), stage 8 (C), andstage 9 (D) gynoecia are shown. E, A stage 8 section hybridized with a REM13 sense strand probe. Bars = 50 mm.


Villarino et al.






Arabidopsis SAM isolated via FACS (Yadav et al., 2009,2014). When these additional samples are included, thehierarchical clustering dendrogram (Fig. 5B) shows thatthe YFP-positive sample is more similar to the SAM celltypes than to the vascular procambium (AtHB8) andphloem cell types (S17). This again suggests the meri-stematic character of the YFP-positive sample (Fig. 5B).One should be cautious, however, to interpret the re-sults of this (or any) cross-platform (array/RNA-seq)comparison until validated cross-platform comparisonmethods are available. To the best of our knowledge,there is no clear consensus in the literature of a standardcross-platform comparison practice (Mudge et al., 2008;Bradford et al., 2010; Guida et al., 2011; Nookaew et al.,2012). Indeed, many researchers have used both plat-forms (array/RNA-seq) in the same experiment andcompare the final results rather than find a way to di-rectly compare the two technologies (Marioni et al.,2008; Nookaew et al., 2012; Xu et al., 2013; Wang et al.,2014; Zhao et al., 2014). Here, we employ a Spearmanrank correlation, as it is less sensitive than the Pearsoncorrelation to strong outliers, makes no assumptionsabout data distribution, and does not inflate type I errorrates. This approach fits well with the data in thiswork, as samples do not cluster based on technologyplatforms but rather based on the apparent cell typesimilarities of gene reads per kilobase of transcript permillion mapped reads (RPKM) expression levels.

Transcriptomic Analysis of the SHP2 Expression DomainComplements Existing Medial Domain and CMMData Sets

Wynn et al. (2011) previously carried out a relatedtranscriptomic study and identified many genes thatwere shown via in situ hybridization to be expressedpreferentially in the developing medial domain of thewild-type gynoecium. When comparing the 95 enrichedDEGs from our RNA-seq experiment (SupplementalTable S2; Supplemental Fig. S3) with a set of 210 medial

domain-enriched genes fromWynn et al. (2011), 23 geneswere found in common (Table III). The 24% overlapof these two gene sets is significantly higher than ex-pected by chance (hypergeometric test, P = 3.153 10230;Halbritter et al., 2012). Members of the REM, HEC, andNGA gene families, as well as several auxin homeostasis-related genes, were among the set of 23 genes identifiedin both experiments (Table III).

Figure 5. The transcriptomic signature of the SHP2-expressing domainis more similar to the transcriptomes of other meristematic samples thanit is to whole flower. A, Dendrogram based on hierarchical clusteringusing Spearman rank correlation using RNA-seq (RPKM) expressionvalues from flowers and other tissues. B, Comparison of RNA-seq andAffymetrix ATH1 array samples including transcriptomic data fromwhole flower, SAM, and seedling. WT, Wild type. Data from Mizzottiet al. (2014)(1), Mantegazza et al. (2014a)(2), GEO accession numberGSE54125(3), and Yadav et al. (2009, 2014)(4) were used for compar-ison. Samples corresponding to this study are color coded red in bothdendrograms.

Figure 4. GO term overrepresentation of SHP2domain-enriched genes suggests a role for thisset of genes in floral, gynoecial, and ovule devel-opment. The BiNGO/Cytoscape representationshows overrepresented GO terms from the95 YFP+/2 DEGs displaying enriched expres-sion in the YFP-positive samples. Edges representthe parent/child relationships of the GO terms(Ashburner et al., 2000), while the color of thenodes indicates the degree of statistical signifi-cance (P , 0.01) as reported by BiNGO (Maereet al., 2005). To unclutter the figure, given thelarge number of significant GO terms, selectednodes and edges were removed from this graph-ical representation.








Reyes-Olalde et al. (2013) recently performed acomprehensive literature survey of genes that functionduring CMM development. They reported 86 protein-coding genes corresponding to transcription factors,hormonal pathways, transcriptional coregulators, andothers of widely diverse functions. Fifteen of these 86CMM developmental regulators are found within theset of 363 YFP-positive DEGs displaying a greater than4-fold expression difference (Fig. 6, gene names markedin red; hypergeometric test, P = 3.3 3 10213; Halbritteret al., 2012). Additionally, there are 32 DEGs that arestatistically significant (at FDR , 0.01), where the foldenrichment or depletion is less than 4-fold. As manybiologically significant DEGs may display less than4-fold expression difference, these genes also are indi-cated (Fig. 6, marked with triple asterisks). Thus, a totalof 47 of the 86 genes reported by Reyes-Olalde et al.(2013) also are detected as differentially expressed be-tween YFP-positive and YFP-negative samples in ouranalysis. The expression profiles of the 86 genesreported by Reyes-Olalde et al. (2013) within the medialdomain-enriched data set from this work, as well as

within data from floral meristem-enriched samples(Mantegazza et al., 2014a), are displayed in the heatmap in Figure 6 (RPKM values can be found inSupplemental Table S9).

Transcript Isoforms in the Arabidopsis Medial Domain

One utility of transcriptome analysis through RNA-seq is the identification of novel alternative splicedtranscripts, alternative transcription start sites (TSSs),and instances of isoform switching (Sims et al., 2014).To further characterize the transcriptome of the SHP2expression domain at the isoform level, we first selectedisoforms that showed a significant (a , 0.01) changein their expression between YFP+/2 samples usingCufflinks/Cuffdiff. For this analysis, we did not apply afold magnitude cutoff, thus capturing all isoforms witha , 0.01. To avoid transcripts that were affected by thecell-sorting procedure, we removed all isoforms thatshowed a significant (a, 0.01) expression level changebetween all-sorted/nonsorted samples. This resulted in

Table II. List of genes displaying isoform-switching behavior between the SHP2-positive and SHP2-negative cell populations

Genes for which the most highly expressed isoform differed between YFP+ and YFP2 samples are shown. Matches between the Cufflinks tran-scripts and the TAIR 10 genome are indicated with Class Code (CC): = for complete transcript match and j for potentially novel isoform (fragment;Trapnell et al., 2012). YFP NEG = YFP negative, YFP POS = YFP positive, NO SORT = nonsorted, and ALL SORT = all sorted.

Isoform Identifier

TSS Group

Identifier CC

Nearest Reference

Identifier

Isoform

RPKM

for YFP

NEG

Isoform

RPKM

for YFP

POS

Isoform

RPKM

for NO

SORT

Isoform

RPKM

for ALL

SORT

Isoform

Relative

RPKM for

YFP NEG

Isoform

Relative

RPKM for

YFP POS

Isoform

Relative

RPKM for

NO SORT

Isoform

Relative

RPKM for

ALL SORT

TCONS_00000095 TSS74 = AT1G02110.1(DUF630)

11.54 37.20 15.31 19.62 0.28 0.53 0.34 0.37

TCONS_00007286 TSS5864 = AT1G23340.1(DUF239)

5.46 1.34 6.93 5.17 0.75 0.40 0.71 0.74

TCONS_00014129 TSS11387 = AT2G42890.2(MEI2-LIKE2)

7.90 2.20 5.42 5.77 0.51 0.29 0.56 0.43

TCONS_00041486 TSS33793 = AT5G53050.3(a/b-hydroxylase)

3.23 0.48 1.31 2.46 0.51 0.11 0.54 0.038

TCONS_00002986 TSS2424 = AT1G43850.1 (SEU) 6.26 18.06 7.02 11.62 0. 30 0.54 0.26 0.47TCONS_00032425 TSS26386 = AT4G32250.2

(protein kinase)2.19 11.19 3.58 5.11 0.16 0.48 0.19 0.27

TCONS_00000744 TSS563 = AT1G10570.2(OTS2)

14.70 6.40 8.02 9.45 0.69 0.48 0.38 0.69

TCONS_00004152 TSS3371 = AT1G61820.1(BGLU46)

9.49 0.82 4.71 3.79 1.00 0.42 0.74 0.086

TCONS_00006607 TSS5332 = AT1G14170.1(KH domaincontaining)

10.18 2.62 5.68 5.39 0.51 0.29 0.046 0.49

TCONS_00014122 TSS11382 = AT2G42830.2(SHP2)

0.63 4.62 1.20 0.78 0.51 0.19 0.33 0.15

TCONS_00001240 TSS954 j AT1G16710.2(HAC12)

3.50 5.88 7.11 5.77 0.46 0.71 0.77 0.78

TCONS_00006659 TSS5370 = AT1G14690.2(MAP65-7)

39.97 16.72 28.41 18.98 0.62 0.41 0.69 0.53

TCONS_00039135 TSS31808 = AT5G19130.1(GPI transamidase)

3.46 17.27 13.34 12.45 0.38 0.74 0.72 0.73

TCONS_00033470 TSS27205 = AT5G06610.1(DUF620)

7.57 1.18 3.31 4.72 0.53 0.16 0.33 0.40

TCONS_00031775 TSS25869 = AT4G23660.1(PPT1)

2.56 9.59 3.23 6.70 0.35 0.76 0.29 0.69


Villarino et al.




4,555 YFP+/2 differentially expressed isoforms(Supplemental Table S8). Within this set of isoformsdifferentially expressed between the YFP+/2 sam-ples, we sought to highlight multiple-isoform genesthat showedmajor changes in the relative frequency ofindividual isoforms between the YFP-positive andYFP-negative samples. To this end, we estimated therelative frequency of each isoform as a percentage ofthe total expression for the gene. Among the 4,555significantly differentially expressed isoforms, only 52isoforms from multiple-isoform genes displayedchanges of 20% or more in their relative frequency.The major isoform (most highly expressed isoform)differed between YFP+/2 samples for only 15 genes(Table II).

Remarkably, the transcriptional coregulator SEUSS(SEU; AT1G43850), previously implicated in medialdomaindevelopment (Franks et al., 2002;Azhakanandamet al., 2008), showed a significant increase of isoformAT1G43850.1 in the YFP-positive samples, while its sec-ond isoform, AT1G43850.2, did not change significantlybetween samples. As a result, isoform 1 was the major(predominant) isoform inYFP-positive cells and isoform2was the major (predominant) isoform in the other sam-ples. The functional significance, if any, of this isoformswitching is currently unknown. The two SEU mRNAisoforms differ only in the length of their 59 untranslatedregions (UTRs; TAIR 10). However, this difference in the59 UTR raises the possibility that the SEU isoforms mightbe differentially regulated at a posttranscriptional level.

The SHP2 (AT2G42830) gene also was found in thislist of 15 genes displaying an isoform switch (Table I).The two SHP2 isoforms differ in the splice donor siteat the 59 end of intron 4 (TAIR 10). This alternativesplicing event is expected to add six nucleotides (twoamino acids) to the AT2G42830.2 isoform. To ourknowledge, the functional consequences of this alter-ation to the SHP2 protein sequence has not yet beeninvestigated. The shorter AT2G42830.1 isoform isdetected as the majority isoform in the YFP-positivesamples.

The regulation of gene expression through alterna-tive promoter usage or the use of alternative TSSs isobserved frequently in multicellular organisms (Ayoubiand Van De Ven 1996). Using the same pipeline andcriteria we employed to select differentially expressedisoforms in the YFP+/2 samples, we identified 93 iso-forms that were differentially expressed as a result ofthe use of alternative promoter/TSSs (SupplementalTable S8, promoter usage tab). Interestingly, one suchpromoter/TSS switch was found for the REVERSIBLY

Figure 6. Heat map representation of the expression profiles of previ-ously identified regulators of CMM development. Expression profiles inRPKM are for the 86 genes reported by Reyes-Olalde et al. (2013) withfunctional roles during CMM development. Transcriptional profiles fromthis study (YFP POS=YFP positive, YFPNEG=YFP negative, ALL SORT=all sorted, and NO SORT = nonsorted) as well as from Mantegazza et al.

(2014a) corresponding to flower stage 3 (FL. STAGE 3), floral meristem(FL. MERISTEM), and inflorescence meristem (IN. MERISTEM) are in-cluded. Genes color coded in red are those identified as DEGs betweenYFP-positive and YFP-negative samples (fold change . 4 and FDR ,0.001), while genes that displayed a statistically significant expressionlevel (FDR , 0.01) between YFP-positive and YFP-negative samples(regardless of their fold change) are indicated with triple asterisks.








GLYCOSYLATED POLYPEPTIDE5 (RGP5) gene (iso-form). Members of the RGP family (RGP1 and RGP2)involved in sugar metabolism are expressed in otherArabidopsis meristematic tissues, such as the root tipand the apical meristem of young seedlings (Drakakakiet al., 2006). In our work, the transcript level of RGP5isoform 2 (AT5G16510.2) in the YFP-positive sampleis 61% higher relative to the level of this isoform in theYFP-negative sample, while the level of isoform 1(AT5G16510.1) is 75% lower (Fig. 7; SupplementalTable S8).

Auxin Homeostasis and the Development of the GynoecialMedial Domain

Auxins are a class of plant hormones that regulategrowth and development (Woodward and Bartel, 2005;Sauer et al., 2013). The most common plant auxin isindole-3-acetic acid. The regulation of auxin homeo-stasis (including synthesis, response, transport, inacti-vation, and degradation) plays an essential role inpatterning the gynoecium and other lateral organs(Woodward and Bartel, 2005; Sehra and Franks, 2015).

Table III. Overlapping DEGs between the 95 DEGs from this study that display enrichment in SHP2-expressing cells and 210 DEGs fromWynn et al.(2011) displaying reduced expression in the seu ant double mutant relative to other genotypes

Annotation/Gene Symbol Gene IdentifierRNA-seq (This Study)

Array (Log2 Scaled Expression Values from

Wynn et al. [2011])

Log2 Fold

Change

Fold

Change P FDR Wild Type ant seu seu ant

HEC1 AT5G67060 5.9 59.4 1.44E-105 2.44E-101 8.887 8.355 8.030 7.292Cys proteinase AT1G03720 4.1 17.7 6.66E-52 4.54E-49 8.998 8.165 8.409 7.053Plant invertase AT3G62820 3.7 12.9 4.76E-40 1.42E-37 8.614 8.900 8.857 7.833REM13 AT3G46770 3.6 12.2 8.42E-20 3.36E-18 8.653 8.620 7.954 7.271ATCEL2 AT1G02800 3.4 10.4 7.33E-90 4.16E-86 11.927 11.466 11.182 9.948LAX1 AT5G01240 3.3 9.7 2.74E-59 2.45E-56 9.311 9.254 9.226 8.709LTP2 AT2G38530 3.3 9.5 6.42E-54 4.97E-51 11.297 10.301 10.987 10.056AMP-dependent synthetase AT1G21540 3.2 9.3 3.45E-22 1.91E-20 8.547 8.859 7.986 7.917AGO5 AT2G27880 3.0 8.0 1.20E-71 1.45E-68 10.194 9.653 9.977 8.937UGT84A2 AT3G21560 2.9 7.6 2.77E-73 4.29E-70 9.735 9.707 8.868 8.370TAA1 AT1G70560 2.8 6.9 1.34E-69 1.43E-66 9.608 9.072 9.068 7.623ATDOF5.8 AT5G66940 2.7 6.6 2.26E-39 6.53E-37 10.380 10.135 9.761 8.007VDD AT5G18000 2.6 6.0 3.71E-31 5.10E-29 8.554 7.719 8.536 7.018MCT1 AT1G37140 2.6 5.9 8.22E-10 7.74E-09 8.003 7.406 7.572 6.713Unknown AT2G41990 2.5 5.6 5.70E-39 1.56E-36 9.690 9.448 9.068 8.450Protein kinase superfamily protein AT1G74490 2.5 5.5 6.27E-37 1.40E-34 7.798 7.727 7.312 6.976NGA3 AT1G01030 2.3 5.1 4.07E-37 9.23E-35 7.543 7.666 7.214 7.007NGA1 AT2G46870 2.3 4.8 9.56E-39 2.58E-36 8.423 8.516 7.849 7.168Cystatin/monellin AT1G03710 2.3 4.8 1.64E-34 2.98E-32 10.186 10.156 9.780 8.809YUC4 AT5G11320 2.3 4.8 8.01E-21 3.70E-19 7.787 7.889 7.858 7.070Heavy metal transport AT1G56210 2.2 4.7 7.11E-49 3.46E-46 9.229 8.487 9.092 8.507ANAC098 AT5G53950 2.1 4.2 1.88E-22 1.08E-20 8.719 8.453 8.270 7.780Unknown AT1G51670 2.1 4.1 6.00E-30 7.25E-28 9.507 9.049 9.027 8.388

Figure 7. Differential expression of RGP5isoforms. A promoter/TSS switch was foundfor the RGP5 gene (AT5G16510). Isoform 2(AT5G16510.2) increases its expression in theYFP-positive domain, while isoform 1 (AT5G16510.1)of the same gene decreases its expression inthe same domain. Three asterisks indicate astatistically significant difference at FDR,0.01;ns indicates not statistically different.


Villarino et al.





The role of auxin during the development of the medialand lateral domains of the gynoecium is less clearlydefined; however, recent studies suggest that auxinhomeostasismechanisms are likely to be distinct inmedialand lateral domains (Larsson et al., 2014; Moubayidinand Ostergaard, 2014; Sehra and Franks, 2015).To better analyze auxin homeostatic mechanisms

during medial domain development, we examinedthe expression of 127 genes with an annotated functionin auxin homeostasis. Of these 127 genes, 80 wereexpressed in our data set and 60 were differentiallyexpressed at FDR , 0.01 in the YFP+/2 comparison,without applying a fold enrichment filter (SupplementalTable S10). The expression levels of TRYPTOPHANAMINOTRANSFERASE OF ARABIDOPSIS1 (TAA1) andYUC4, two genes encoding proteins in the auxin syntheticpathway, were strongly enriched (more than 4-fold) inthe YFP-positive samples, as predicted from previ-ously published expression patterns indicating enriched

expression within the medial portions of the gynoecium(Zhao et al., 2001;Cheng et al., 2006; Stepanova et al., 2008;Tao et al., 2008; Trigueros et al., 2009;Martínez-Fernándezet al., 2014). Within the PINFORMED (PIN) family ofpolar auxin transporters, the expression levels of PIN1,PIN3, and PIN7 were enriched significantly in the YFP-positive sample (Supplemental Table S10). This is con-sistentwith the reported expression patterns at the proteinlevel of these PIN transporters within the medial domainof the gynoecium (Benková et al., 2003; Blilou et al., 2005;Larsson et al., 2014; Moubayidin and Ostergaard, 2014).

Auxin induces gene expression through a family oftranscription factors calledAUXINRESPONSEFACTORS(ARFs; Woodward and Bartel, 2005). At a fold changelevel of 1.5 and FDR, 0.01, 10ARFswere enriched in theYFP-positive sample (ARF1, ARF2, ARF3/ETTIN, ARF4,ARF5, ARF6, ARF7, ARF8, ARF16, and ARF18), while noARFs were identified as depleted in the YFP-positivesample (Supplemental Table S10). Our data suggest that

Figure 8. Differential expression of TAS3 and ARF genes. Expression of the ARFs (ARF2, ARF3, ARF4) and TAS3 transcripts isshown. Expression levels of ARF2, ARF3, and ARF4 are significantly enriched in the YFP-positive sample relative to the YFP-negative sample at FDR , 0.01. Expression levels of the TAS3 genes AT5G49615 (TAS3B) and AT3G17185 (TAS3), whichnegatively regulate the expression of ARF2, ARF3, and ARF4 expression (Williams et al., 2005), are significantly reduced (FDR,0.01) in the YFP-positive sample. All error bars indicate confidence intervals as reported by Cufflinks. Three asterisks indicatestatistically significant differences at FDR , 0.01; ns indicates not statistically different.









these ARF family members may be expressed preferen-tially in the medial domain and play a role during thedevelopment of this meristematic tissue. Previous studieshave documented gynoecial developmental defects inarf3/ettin mutants (Sessions and Zambryski, 1995) as wellas in arf6 arf8 double mutants (Nagpal et al., 2005; Wuet al., 2006). Interestingly, the levels of the precursortranscripts for two TRANS-ACTING SIRNA3 (TAS3)genes (AT5G49615 and AT3G17185) were reduced sig-nificantly (FDR, 0.01) in the YFP-positive sample (Fig. 8;Supplemental Table S8). The trans-acting small interferingRNAs that are encoded by the TAS3 genes negativelyregulate the levels of ARF2, ARF3, and ARF4 transcripts(Williams et al., 2005). Thus, the enrichment of ARF2,ARF3, and ARF4 transcript levels in the SHP2 expressiondomain may be due in part to a reduction in the level ofexpression of the TAS3-encoded trans-acting small inter-fering RNAs in the medial domain.

SQUAMOSA PROMOTER-BINDING PROTEIN-LIKE3and the cis-NAT Antisense Gene AT2G33815

The SQUAMOSA PROMOTER-BINDING PROTEIN-LIKE (SPL) genes function in the regulation of thetransition from juvenile to adult growth phases and theregulation of shoot regenerative capacity (Wu andPoethig, 2006; Wang et al., 2008, 2009; Zhang et al.,2015). In our study, the expression of SPL3 (AT2G33810)was more than 4-fold lower in the YFP-positive samplerelative to the YFP-negative sample. SPL3 encodes aDNA-binding protein directly regulating APETALA1(AT1G69120), a key regulator of floral meristem identity

specification (Leal Valentim et al., 2015). Interest-ingly, the expression of the cis-NAT antisense geneAT2G33815, complementary to portions of the SPL3gene, also was reduced significantly approximately4.5-fold in the YFP-positive sample (Cufflinks data inSupplemental Table S8). In both plant and animalsystems, the coexpression of overlapping sense andantisense transcripts can lead to the generation of cis-NAT-derived small interfering RNAs that can regulategene expression at transcriptional and posttranscrip-tional levels (Zhang et al., 2015). The expression ofanother regulator of SPL3 activity, miRNA157D(AT1G48742), also was reduced significantly in the YFP-positive samples.miRNA157D reduces the translation ofthe SPL3 transcript by acting through a miRNA156/157-responsive element in the SPL3 39 UTR (Gandikotaet al., 2007; Wang et al., 2009). These data suggestthat the miRNA156/157/SPL module may act duringmedial domain development and may be regulatedby the cis-NAT antisense gene AT2G33815. A com-plete list of differentially expressed natural antisense,transposable element, and other non-protein-codingtranscripts identified as differentially expressed byCufflinks, DESeq2, and edgeR is found in SupplementalTable S8.

Protoplasting-Induced Stress Genes

While the predominant focus of this work was toperform transcriptomic analysis in medial domain-enriched cells (YFP+/2), transcripts induced by theprotoplasting and sorting process (all-sorted/nonsorted)

Figure 9. Six-way Venn diagram imageshowing detailed overlap from all the DEGdata sets. The total numbers of DEGs undereach condition and for each program areindicated in parentheses. CTR = DEGsbetween all-sorted/nonsorted and YFPs =DEGs between YFP+/2. Cuff = Cufflinks,edg = edgeR, and Des = DESeq2. Theinteractive tool can be accessed onlineusing the InteractiVenn Web tool (http://www.interactivenn.net.) and uploadingSupplemental File S1.


Villarino et al.






http://www.interactivenn.net




also were identified (see “Materials and Methods”). Tofacilitate the visualization of all samples, we generatedan interactive six-way Venn diagram using the Web-based tool InteractiVenn (Heberle et al., 2015). Byuploading Supplemental File S1 to InteractiVenn(http://www.interactivenn.net), mousing over, andclicking on the numbers in the Venn diagram, re-searchers will find gene identifiers from DEGs betweenYFP+/2 and all-sorted/nonsorted samples (three pro-grams and two comparisons). As expected, when com-paring such different types of samples (all-sorted/nonsorted and YFP+/2), few DEGs (26) overlappedacross the six samples (Fig. 9). The lack of overlapof DEGs across the entire experiment indicates that theYFP+/2 DEGs reported here are not a result of proto-plasting stress-induced processes.When comparing the protoplasting-induced gene set

from this work (all-sorted/nonsorted DEGs) with thoseinduced due to FACS-sorting methodology in SAM byYadav et al. (2009) and in roots as reported by Birnbaumet al. (2005), few DEGs were found in common (sevenacross all data sets; Supplemental Fig. S4), indicating thatdifferent tissues and/or different protoplasting tech-niques generate different sets of protoplast-inducedgene expression changes. Thus, appropriate controlsshould be included to control for condition-specificprotoplasting-induced gene expression changes.

CONCLUSION

Despite the importance of the gynoecial medial do-main in ovule development, no domain-specific tran-scriptome has been reported previously, mainly due tothe difficulty of isolating the meristematic cells fromwhich ovules are derived. In this work, we developed anovel FACS-based system using the SHP2 expressiondomain using a GAL4/pUAS-based two-componentsystem that, when combined with flower synchroni-zation and flow cytometry, allowed for the efficientisolation of medial domain cells expressing SHP2. Thequality and quantity of biological samples that can berecovered with our system enables cell type- and strand-specific RNA-seq transcriptomic analysis and opens uppossibilities for small RNA,metabolomic, and proteomicanalyses (Petersson et al., 2009; Breakfield et al., 2012;Petricka et al., 2012; Li et al., 2013; Moussaieff et al.,2013). This approach, coupled with high-throughputRNA-seq, has yielded a unique and novel snapshot ofthe gynoecial medial domain transcriptome and a set ofcandidate regulators of medial domain development forfuture functional analysis.

MATERIALS AND METHODS

Construction of pSHP2-GAL4; pUAS-3xYpetDual-Construct Lines

The SHP2 promoter fragmentwas amplified fromColumbia-0wild-type genomicDNAusing theprimersproSHP2gwF1 (59-CACCATCTCCAACGCATTGTTACG-39)

and proSHP2gwR1 (59-CATTTCTATAAGCCCTAGCTGAAG-39). This frag-ment contains the sequences from22,170 to +1 relative to the SHP2 ATG andincludes the 59 UTR, the first intron, and the first Met codon of SHP2. Thispromoter was shown previously to mimic the endogenous SHP2 expressionpattern (Colombo et al., 2010). This genomic fragment was cloned into thepENTR/D-TOPO vector (Invitrogen) to create plasmid LJ001 and then shuttledvia Gateway LR reaction (Invitrogen) into the destination vector JMA859 (i.e.pEarleygate303-GAL4) to create plasmid AAS003. Transgenic Arabidopsis(Arabidopsis thaliana) lines were created by Agrobacterium tumefaciens-mediatedtransformation of the AAS003 plasmid into the S. No. 1880 seed stock that con-tained the pGWB2-pUAS-3xYpet responder construct (see below), generating thepSHP2-GAL4; pUAS-3xYpet dual-construct line (S. No. 1896), referred to aspSHP2-YFP. The pSHP2-YFP plants were crossed to the ap1 cal1 Wellmer floralinduction system (Wellmer et al., 2006) as described below.

JMA859 (pEarleygate303-GAL4) is a modified pEarleygate303 (Earleyet al., 2006) plasmid in which the reporter was replaced by the coding se-quences from the GAL4 Saccharomyces cerevisiae transcriptional activator. Toachieve this, pEarleygate303 was cut with NcoI (New England Biolabs) andSpeI (New England Biolabs). Then, fusion PCR was used to create the insertthat fused the GAL4 sequences to the deleted portions of pEarlygate303.This required three PCRs: first PCR with primers pEarl303NcoIFor (59-TGGCCAATATGGACAACTTCT-39) and pEarl303Rev_GAL4(tale) (59-ATGGAGGACAGGAGCTTCATACACAGATCTTCTTCAGAGA-39); secondPCRwithprimersGAL4F_pEarl303(tale) (59-TCTCTGAAGAAGATCTGTGTATGA-AGCTCCTGTCCTCCAT-39) and GAL4Rev_SpeI (59-CCGGACTAGTCTACC-CACCGTACTCGTCAA-39); and then a fusion PCR joining these two fragmentsusing the external primers to amplify. The product of the fusion PCR wasdouble digested withNcoI/SpeI and ligated intoNcoI/SpeI-cut pEarleygate303.

JMA382 (pUAS-pGWB2) was created from pGWB2 (Nakagawa et al., 2007)by replacing the p35S sequences in pGWB2with pUAS sequences (HindII/XbaIsites used). A Gateway LR reaction was then used to move the 3xYpet cassettefrom JMA710 (pENTR/D-TOPO-3xYpet) into JMA382, creating vector JMA721(i.e. pGWB2-pUAS-3xYpet). Homozygous single insertion site transgenic linesharboring JMA721 were then generated (S. No. 1880).

Plant Material

In a wild-type inflorescence, cells expressing SHP2 represent a small per-centage of the total cells. Additionally, wild-type inflorescence contains a fullrange of developmental series of floral stages. The Wellmer floral synchroni-zation system (Wellmer et al., 2006) was used to maximize the amount ofgynoecial tissue from floral stages 6 to 8 (Smyth et al., 1990). TheWellmer groupkindly provided pAP1-AP1::GR; ap1; cal seeds (kanamycin resistant in theLandsberg erecta background; S. No. 1927). The pSHP2-GAL4; pUAS-3xYpetdual-construct plants (S. No. 1896) were crossed to pAP1-AP1::GR; ap1; cal.Lines homozygous for erecta, ap1, cal, and the transgenes were selected in F2 andF3 generations (generating S. No. 2060). Because of the mixed ecotype cross(Columbia-0 and Landsberg erecta), lines that were erecta homozygous mutantand gave a consistent YFP expression pattern and consistent inducibility of theAP1-GR activity were selected before the generation of protoplasts. Plants weregrown under constant light and temperature at 22°C to minimize circadiantranscriptional fluctuations. To induce flowering in the transgenic plants, 20mmof the synthetic steroid hormone dexamethasone (Sigma) in 0.015% (v/v) Silwetwas applied directly (spray application) approximately 30 d after planting(Wellmer et al., 2006). Inflorescences were collected for protoplast generationapproximately 120 h after dexamethasone-induced floral synchronization.When collecting samples for protoplast preparation, five to six inflorescenceheads were fixed for chloral hydrate clearing and DICmicroscopy to determinethe developmental stages of the flowers of the inflorescence samples. Addi-tionally, before protoplasting, whole inflorescences were collected and frozenimmediately in liquid nitrogen for analysis of the transcriptional starting stateof the nonprotoplasted tissue (nonsorted samples; see “Experimental Design”).

Experimental Design

Material for RNA samples was gathered from batches of plants grown at1-week intervals to generate biological replicates (material from each week wasconsidered as a biological replicate). To reduce variability between bioreplicatesdue to environmental heterogeneity within the growth chamber, each bio-replicate was drawn from a pool that contained plants grown within threedifferent chamber positions. Four biological replicates of each of four tissuesamples (YFP-positive, YFP-negative, all-sorted, and nonsorted samples) were








collected (16 samples total). Whole inflorescences were collected for nonsortedsamples and immediately frozen in liquid nitrogen before RNA isolation (i.e.these samples were not subjected to protoplasting or FACS sorting). The all-sorted samples represented the total population of protoplasts that come off theFACSmachine after debris and broken cells are removed based on sorting gates(Supplemental Fig. S1). The YFP-positive and YFP-negative protoplast popu-lations are processed equivalently to the all-sorted samples except that a finalFACS-sorting gate is used to divide the all-sorted protoplasts into YFP-positiveand YFP-negative samples (Supplemental Fig. S1). RNA was isolated fromthese three protoplast populations as well as from entire nonprotoplastedinflorescences (nonsorted). The YFP-positive, YFP-negative, and all-sortedsamples were prepared and collected as described below (see “ProtoplastRecovery and Cell Sorting”).

Protoplast Recovery and Cell Sorting

Protoplasts from the S. No. 2060 plants were generated according to theprotocol of Birnbaum et al. (2005), with adaptations for inflorescence plantmaterial. Inflorescences (approximately 200) were hand collected with forcepsand/or scissors and chopped with a Personna double-edge prep blade(American Safety Razor; 74-002) within a 15-min period. Cell wall polysac-charides were digested by immersing the chopped plant material in 10 mL offilter-sterilized solution B in a 50-mL Falcon tube. Solution B (preparedaccording to Birnbaum et al., 2005) is prepared from solution A (10 mM KCl,2 mM MgCl2, 0.2 M MES, and 600 mM mannitol) to which cell wall-digestingenzymes were added (final concentrations of 1.5% (w/v) Cellulase [Yakult], 1%(w/v) Pectolyase [Yakult], and 1% (w/v) Hemicellulase [Sigma]). This mixtureis then dissolved by gently swirling, covered in foil, andwarmed in awater bathat 55°C for 10 min to inactivate DNases and proteases. After cooling to roomtemperature, CaCl2 (2 mM final) and bovine serum albumin (0.1% (w/v) final)were added, and the solution was filter sterilized through a 25-mm filter.

After 1 h of incubation at room temperaturewith occasional gentle agitation,10mLof the protoplast-rich solution Bwasfiltered through a 70-mmfilter basketto a 50-mL Falcon tube. A 10-mL rinse of solution A was applied directly to thematerial left in the 70-mm filter basket to rinse through any protoplasts leftbehind. Protoplasts were spun at 500g and 10°C for 10 min; the majority of thesupernatant was removed by aspiration, being careful not to disturb the pro-toplast pellet, which is typically not tightly compacted. Protoplasts wereresuspended in 25 mL of solution A as a rinse step to remove cell wall-digestingenzymes. Protoplasts were filtered again through a 50-mm filter mesh to a newtube, adding 8 mL of solution A to again rinse through any protoplasts stuck inthe filter. Protoplasts were spun again at 500g for 10 min. The majority of thesupernatant was removed, leaving 2 mL of the protoplasts in solution after thesecond centrifugation step. Propidium iodide (5 mg mL21

final concentration)was added to the protoplasts (to allow separation of broken protoplasts), and afinal filtering step though a 30-mmmesh filter (CellTrics; Partec) was carried outbefore loading onto the FACS machine.

Flow cytometry through FACS sorting (Moflo XDP; Beckman Coulter) wasused to isolate the YFP-expressing cells from the total pool of cells. The FACSmachine was equipped with a cooling device (set to 10°C) and fitted with a100-mm nozzle. Protoplasts were sorted at a rate of up to 10,000 events persecond at a fluid pressure of 25 p.s.i. Four sorting gates were set in an effort tocollect the cleanest set of protoplasts and to eliminate debris and broken cells. Afirst gate based on size and granularity using side-scatter and forward-scatterparameters was used to select for intact protoplasts. Then, a second gate wasused to select for single cells and remove doublets. A third gate was used toselect for cells that were negative for propidium iodide signal, as broken pro-toplasts and debris are preferentially stained by propidium iodide, which isexcited by the 488-nm laser and emits at 617 nm. The total population of pro-toplasts that came off the FACS sorter machine after these gates constituted theall-sorted sample. In parallel, the YFP-positive protoplasts and YFP-negativeprotoplasts were separated into two collection tubes using the gates describedabove and one additional sorting gate based on the level of emission intensity inthe green channel (529-nm/28-nm filters). Preliminary experiments with pro-toplasts that did not express the YFP transgene were used to set this gate anddetermine the levels of autofluorescence of the protoplasts. Protoplasts werecollected directly into 14-mL tubes containing 4 mL of Trizol (Invitrogen/LifeTechnologies) and occasionally agitated during the approximately 40 min ofsorting required to collect the protoplasts. Trizol was the method of choice, as itmaintains a high level of RNA integrity during tissue homogenization whilealso disrupting and breaking down cells and cell components. In order tominimize artifactual changes to transcript levels, the entire process of cell wall

digestion, protoplast generation, and FACS sorting was kept under 3 h. Thisprocedure typically yielded between 300,000 and 500,000 YFP-positive proto-plasts. These YFP-positive protoplasts typically represented approximately0.5% of the total FACS-sorting events. On average, from four sorting trialsrepresenting four biological replicates, the number of cells collected and pro-cessed for each sample was 575,000 for the YFP-positive samples, 1,000,000 forthe YFP-negative samples, and 493,000 for the all-sorted samples.

RNA Extraction and qRT-PCR

Total RNA was extracted from sorted protoplasts collected in Trizol (keepinga 3:1 ratio of Trizol to sorted cells) and by modifying the Plant RNeasy Mini Kit(Qiagen) protocol as follows: collected cells in Trizol (4 mL total) were vortexedfor 5 min at room temperature and 1 mL of chloroform (Sigma) was added. Thesolution was vortexed again for 1 min at room temperature and centrifuged at4,000 rpm for 10 min at 4°C to separate phases; RNA from the aqueous phase(top layer) was carefully sucked up and mixed with 700 mL of Qiagen RLTbuffer (Plant RNeasyMini Kit; Qiagen) and 7 mL of B-mercaptoethanol (Sigma).A total of 500 mL of 100% ethanol was added, and solution was then transferredto a Qiagen MinElute column (Plant RNeasy Mini Kit; Qiagen) and spun in a2-mL microfuge tube for 15 s at approximately 10,000 rpm. A total of 500 mLof Qiagen RPE buffer (Plant RNeasy Mini Kit; Qiagen) was added to thespin column and spun for 15 s at approximately 10,000. A total of 750 mL of 80%(v/v) ethanol was added to the MinElute column and spun at approximately10,000 rpm for 15 s (twice) to ensure the removal of all guanidine salts that mayinhibit downstream applications. A final 5-min spin at top speed with the capoff was performed to remove trace amounts of ethanol. Total RNA was theneluted with 10 mL of RNase-free water. A second elution was performedwith another 10 mL of RNase-free water. It is worth noting that one biologicalreplicate (fourth biological replicate) from the YFP-positive protoplasts waslost at this point, leaving only three biological replicates for this tissue sampleand yielding a total of 15 samples sequenced in two lanes and used for theexperiment.

Prior to high-throughput sequencing, qRT-PCR was conducted on YFP-positive and YFP-negative samples using the 2–DDCT method as suggested bySchmittgen and Livak (2008) to assess the relative gene expression of the specificmedial domain markers SHP2 and NGA1. Total isolated RNA was quantifiedusing fluorometric quantitation (Qubit RNA Assay Kit; Life Technologies)for both YFP-positive and YFP-negative samples (approximately 100 ng). TheSuperScript III First-Strand Synthesis System (Invitrogen/Life Technologies)was used to generate complementary DNA (cDNA; diluted 1:4 prior to qRT-PCR analysis) from total RNA. The qRT-PCR assay was performed (ThermalCyclers from Applied Biosystems) using a SYBR Green mix (QuantiTect SYBRGreen PCR Kits; Qiagen). Three biological replicates of the YFP-positive andYFP-negative samples were included, and each biological replicate was assayedin triplicate. The expression levels of the ADENINE PHOSPHORIBOSYLTRANSFERASE1 (AT1G27450) gene was used for normalization.

Bar Plots

Bar plot graphs were constructed using the R packages bear (http://sourceforge.net/projects/yjlee-r-packages/files/bear/) and plyr (https://cran.r-project.org/web/packages/plyr/index.html) to calculate means, SE, andconfidence intervals and ggplot2 (https://cran.r-project.org/web/packages/ggplot2/index.html) to generate the plots.

Library Preparation and mRNA Sequencing

Total RNA isolated was quantified using fluorometric quantitation (QubitRNA Assay Kit; Life Technologies), and RNA quality was assessed using theAgilent 2100 Bioanalyzer. The RNA integrity number for the 15 samples washigher than 7.3, which is above the Illumina threshold for library construction(greater than 7). Strand-specific cDNA libraries were constructed from ap-proximately 100 ng of total RNA using aNEBUltra Directional Library Prep Kitfor Illumina (New England Biolabs). The average size of the cDNA fragmentswas approximately 250 bp. The 15 bar-coded libraries were pooled, and single-end sequencingwas performed in aHiSeq 2500 device (Illumina) withHiSeq SRCluster Kit version 4 for the flow cell and HiSeq SBS version 4 for sequencingreagents. cDNA libraries were sequenced in 125 cycles plus seven cycles formultiplexed samples. Sequencing was performed in two lanes of a flow cell; all15 libraries were sequenced twice, and the results from the two independent


Villarino et al.




http://sourceforge.net/projects/yjlee-r-packages/files/bear/

http://sourceforge.net/projects/yjlee-r-packages/files/bear/

https://cran.r-project.org/web/packages/plyr/index.html

https://cran.r-project.org/web/packages/plyr/index.html

https://cran.r-project.org/web/packages/ggplot2/index.html

https://cran.r-project.org/web/packages/ggplot2/index.html


lanes were analyzed as technical replicates. As no lane-specific effects wereobserved during data analysis, the reads from each lane were pooled for theanalysis of DEGs (see “Table Counts and Technical Replicates”).

Bioinformatics Analysis

All bioinformatics analyses were performed on a server cluster with 128gigabytes of RAM, two 8-core processors per compute node, and a UbuntuLinux-Distribution 12.04 operating system using the Simple Linux Utility Re-source Management queue management system at the Bioinformatics ResearchCenter at North Carolina State University.

Read Processing

Quality control and preprocessing of metagenomic data were performedusing FastQC software (Schmieder and Edwards, 2011). Adapters and low-quality sequences were filtered out with Ea-Utils software (Lindgreen, 2012).Readswith phred-like quality score greater than 30 and read length greater than50 bp were kept and aligned against the TAIR 10 reference genome.

Sequence Alignment to the Arabidopsis Genome

The splice junctionmapper TopHat2 (version 2.0.10; Trapnell et al., 2009)wasused to align filtered RNA-seq reads to the TAIR 10 genome (Ensembl anno-tation) downloaded from the iGenome database (http://support.illumina.com/sequencing/sequencing_software/igenome.html). Default parametersfor TopHat2were used except for strand specificity (–library-type = fr-firststrand)to match to the first strand of cDNA synthesized (antisense to the mRNA) andmaximal intron length (–I 2000), as it has been shown that the large majority ofknown introns are smaller than the selected threshold (Li et al., 2013). To alignreads solely and exclusively against TAIR 10 annotated gene models, thearguments –T (transcriptome only) and –no-novel-juncs (no novel junction)also were included. Uniquely mapped reads were extracted from theTopHat2 output binary (BAM) file using samtools (Li et al., 2009) andselecting for the NH:i:1 two-character string tag. Only uniquely mapped readswere used for downstream analysis.

Table Counts and Technical Replicates

The HTSeq: Analyzing High-Throughput Sequencing Data with Pythonsoftware (Anders et al., 2015) was used with default parameters except for thestranded = reverse mode to generate tables-counts for downstream differentialexpression analysis for the R packages edgeR (Robinson et al., 2010) andDESeq2 (Love et al., 2014).

Using edgeR, we assessed the gene level variance versus log gene expressionlevel among technical replicates (corresponding to two lanes in the flow cell of theIlluminaHiSeq 2500). A linear-dependent Poisson distributionwas observed fortechnical replicates (Supplemental Fig. S5), in accordance with several studies(Marioni et al., 2008; Anders and Huber, 2010; Robinson et al., 2010). Thus,differential gene expression analysis was performed using pooled technicalreplicates.

Gene Expression and Differential Gene Expression

Gene expression and differential gene expression analyses were carriedout using the R packages edgeR (Robinson et al., 2010) and DESeq2 (Loveet al., 2014) and the Linux-based Cufflinks program (version 2.2.1) -G option(Trapnell et al., 2012) for DEGs and transcripts. To facilitate future use ofthese data sets, all the expressed genes identified and their expression values(F/RPKM) in the YFP+/2 (Supplemental Table S3) and all-sorted/nonsorted(Supplemental Table S4) samples are included.

Filters were applied to determine if a gene was detected, abiding by thesuggestions of statisticians and bioinformaticians (Bourgon et al., 2010; Rauet al., 2013; Soneson and Delorenzi, 2013; Love et al., 2014; Pimentel et al., 2014;Seyednasrollah et al., 2015) as a means to enrich for true DEGs, to reduce type Ierror, and to improve P value adjustment. The edgeR function (Robinson et al.,2010) cpm (counts per million) was used to discard those genes whose cpmwaslower than a threshold of two reads per gene in at least three biological repli-cates, as suggested in the edgeR vignette. For Cufflinks, a minimum RPKM of 5was set for a gene to be expressed, following the criteria of Suzuki et al. (2015).According to Sims et al. (2014), 80% of genes can be accurately quantified with

FPKM/RPKM . 10. DESeq2 performs independent filtering using the resultsfunction, as described in the DESeq2 vignette (Love et al., 2014). An FDR cutoffof less than 0.01 was used to determine DEGs in all three programs. The GeneRegulatory Information Server was used to identify transcription families inthe data set (Yilmaz et al., 2011). Enriched categories of transcription factorswithin the set of cleaned 363 YFP+/2 DEGs was assessed with the onlineTranscription Factor Enrichment Calculator tool (https://dgrinevich.shinyapps.io/ShinyTF).

Venn Diagrams and Heat Maps

Venn diagrams were constructed using the R package VennDiagram (Chenand Boutros, 2011) and the Web-based tool package InteractiVenn (Heberleet al., 2015). Heat maps were produced using the R package pheatmap (http://cran.r-project.org/web/packages/pheatmap/index.html). RPKM normaliza-tion by gene length and library size values was produced using the rpkmfunction from edgeR (Robinson et al., 2010). To calculate gene length, a TAIR 10gene length list (coding sequence plus UTRs) was constructed by extractinglength information from the TAIR 10GFF filewith homemade Perl script. Geneswith multiple isoforms were collapsed, and length was calculated using thelongest one. RPKM values were then calculated for clustering purposes and tohave an intermediate point of comparison between Cufflinks, edgeR, andDESeq2. Samples were clustered (default clustering) with parameters providedin the software. The R package colorRamp (http://stat.ethz.ch/R-manual/R-devel/library/grDevices/html/colorRamp.html) was used to produce a gra-dient of color values corresponding to gene fold change values.

GSEA

GO enrichment tests were performed using the R package topGO (Alexaand Rahnenführer, 2009), with the classic algorithm (where each GO category istested independently) and the Fisher statistical test for biological processes,molecular function, and cellular component. Enrichment analysis was per-formed separately for all the genes that were differentially expressed betweenthe YFP+/2 samples and the all-sorted/nonsorted samples. Network analysisof GO terms was performed using the BiNGO (Maere et al., 2005) pluginfor Cytoscape (Shannon et al., 2003). GO terms for the 268 genes identified asdepleted in the YFP-positive sample, as well cellular component and molecularfunction for the YFP+/2 sample, can be found in Supplemental Table S6.

Dendrograms

The R Dist function was used to compute a distance matrix using theSpearman method (Spearman test rank correlation) and the R Cor function tocompute the variance of the matrix. To perform hierarchical clustering, the hclustfunction in R was used. All statistical analyses were performed in R version 3.0.2.Dendrogram plots were built using the R ape package with edge.color = blue.

Confocal Microscopy

Confocal microscopy was performed using a Zeiss LSM 710 microscopemodel (Zeiss Axio Observer Z.1) with objective type Plan-Apochromat 203/0.8M27. Z-stack intervals were set to 2 mm, and the total thickness of the stackwas 62 mm.

Chloral Hydrate Clearing and DIC Microscopy

Inflorescence samples were fixed in a solution of nine parts ethanol to one partacetic acid for 2 h at room temperature and then washed twice in 90% (v/v)ethanol for 30 min each wash. Inflorescences were transferred to Hoyer’s so-lution (70% (w/v) chloral hydrate, 4% (v/v) glycerol, and 5% (w/v) gumarabic) and allowed to clear for several hours to overnight. Samples were thendissected in Hoyer’s solution. The dissected inflorescence heads were mountedin Hoyer’s solution under coverslips and examined with DIC optics on a ZeissAxioskop 2 to determine the floral stages.

In Situ Hybridization

For in situ hybridization analysis, Arabidopsis Columbia-0 flowers werefixed and embedded in paraffin as described previously (Franks et al., 2002;




http://support.illumina.com/sequencing/sequencing_software/igenome.html

http://support.illumina.com/sequencing/sequencing_software/igenome.html






http://cran.r-project.org/web/packages/pheatmap/index.html

http://cran.r-project.org/web/packages/pheatmap/index.html

http://stat.ethz.ch/R-manual/R-devel/library/grDevices/html/colorRamp.html

http://stat.ethz.ch/R-manual/R-devel/library/grDevices/html/colorRamp.html



Wynn et al., 2011). Sections of plant tissue were probed with digoxigenin-labeled antisense and sense RNA probes (Roche). Probes corresponded tonucleotides +686 to +920 of REM13 relative to the transcriptional start site of thecoding sequence using the following oligonucleotides to amplify the template:REM13_ISH_Fwd (59-AAAATAGAACGCGCATACCG-39) andREM13_ISH_Rev(59-TCGTGAACCAAACCGTGATA-39). Hybridization and immunologicaldetection were performed as described previously (Franks et al., 2002; Wynnet al., 2011).

Illumina sequencing raw data (fastq) have been submitted to the GEOdatabase (accession no. GSE74458).

Supplemental Data

The following supplemental materials are available.

Supplemental Figure S1. Sorting gates used to select YFP samples and theresorting of the YFP-positive cells to assess sample purity.

Supplemental Figure S2. qRT-PCR enrichment of the medial domaingenes SHP2 and NGA1 and the gene TUB.

Supplemental Figure S3. Expression profiles for the 363 DEGs (foldchange . 4 and FDR , 0.001) across all four samples (YFP positive,YFP negative, all sorted, and nonsorted).

Supplemental Figure S4. Venn diagram comparison of stress-inducedgenes due to the protoplast/FACS-sorting procedure.

Supplemental Figure S5. Gene level variance versus log gene expressionlevel among technical replicates.

Supplemental Table S1. Summary of RNA-seq data.

Supplemental Table S2. DEGs from the YFP+/2 and all-sorted/nonsorted comparison.

Supplemental Table S3. All the expressed protein-coding genes identifiedwith three different programs between the YFP+/2 samples.

Supplemental Table S4. All the expressed protein-coding genes identifiedwith three different programs between the all-sorted/nonsorted sam-ples.

Supplemental Table S5. Raw high-throughput count data for the YFP+/2and all-sorted/nonsorted comparison.

Supplemental Table S6. GSEA for the YFP+/2 and all-sorted/nonsortedcomparison, including biological process, molecular function, and cellu-lar component.

Supplemental Table S7. Transcription factor families identified in theDEGs from YFP+/2 and their statistical enrichment.

Supplemental Table S8. Isoform expression, regulation of gene expressionby alternative promoters, and antisense transcripts identified byCufflinks, edgeR, and DESeq2.

Supplemental Table S9. Expression profile (RPKM) of the 86 genes de-scribed by Reyes-Olalde et al. (2013) expressed in the medial domain.

Supplemental Table S10. Hormone-related-genes present in our data set.

Supplemental File S1. Data file to upload to the Web-based tool packageInteractiVenn.

ACKNOWLEDGMENTS

We thank Sarah Schuett (CVM, Flow Cytometry Facility, North CarolinaState University [NCSU]) and the Genomic Sciences Laboratory ResearchFacility (NCSU) for library preparation and Illumina sequencing; WilliamThomson and EmilyWear (NSCU) for FACS-sorting assistance; FrankWellmerand Diarmuid O’Maoileidigh (Smurfit Institute, Trinity College of Dublin) forthe pAP1-AP1::GR; ap1; cal floral synchronization system; Colleen Doherty(NCSU) for help with Cytoscape/BiNGO analysis; Maria Angels De LuisBalaguer (NCSU) for assistance with cross-platform expression correlationapproaches; José Alonso and Ross Sozzani (NCSU) for thoughtful commentson the article; Jigar Desai, Dmitry Grinevich, and Colleen Doherty (NCSU) forhelp with the transcription factor enrichment analysis; Aureliano Bombarely(Virginia Tech) for help with GO term analysis; and Eva Johannes (CMIF,

Molecular Imaging Facility, NCSU) for laser scanning confocal microscopeassistance.

ReceivedDecember 7, 2015; acceptedMarch 14, 2016; publishedMarch 16, 2016.

LITERATURE CITED

Adrian J, Chang J, Ballenger CE, Bargmann BOR, Alassimone J, DaviesKA, Lau OS, Matos JL, Hachez C, Lanctot A, et al (2015) Transcriptomedynamics of the stomatal lineage: birth, amplification, and terminationof a self-renewing population. Dev Cell 33: 107–118

Aichinger E, Kornet N, Friedrich T, Laux T (2012) Plant stem cell niches.Annu Rev Plant Biol 63: 615–636

Alexa A, Rahnenführer J (2010) topGO: Enrichment Analysis for GeneOntology; 2010. Bioconductor package version 3.2; https://www.bioconductor.org/packages/release/bioc/html/topGO.html - AccessedDec. 2015

Alvarez J, Smyth DR (2002) CRABS CLAW and SPATULA genes regulategrowth and pattern formation during gynoecium development in Ara-bidopsis thaliana. Int J Plant Sci 163: 17–41

Alvarez JP, Goldshmidt A, Efroni I, Bowman JL, Eshed Y (2009) TheNGATHA distal organ development genes are essential for style speci-fication in Arabidopsis. Plant Cell 21: 1373–1393

Anders S, Huber W (2010) Differential expression analysis for sequencecount data. Genome Biol 11: R106

Anders S, Pyl PT, Huber W (2015) HTSeq: a Python framework to workwith high-throughput sequencing data. Bioinformatics 31: 166–169

Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, DavisAP, Dolinski K, Dwight SS, Eppig JT, et al (2000) Gene Ontology: toolfor the unification of biology. Nat Genet 25: 25–29

Ayoubi TA, Van De Ven WJ (1996) Regulation of gene expression byalternative promoters. FASEB J 10: 453–460

Azhakanandam S, Nole-Wilson S, Bao F, Franks RG (2008) SEUSSand AINTEGUMENTA mediate patterning and ovule initiation duringgynoecium medial domain development. Plant Physiol 146: 1165–1181

Baxter CEL, Costa MMR, Coen ES (2007) Diversification and co-option ofRAD-like genes in the evolution of floral asymmetry. Plant J 52: 105–113

Benková E, Michniewicz M, Sauer M, Teichmann T, Seifertová D, JürgensG, Friml J (2003) Local, efflux-dependent auxin gradients as a commonmodule for plant organ formation. Cell 115: 591–602

Bennett MJ, Marchant A, Green HG, May ST, Ward SP, Millner PA,Walker AR, Schulz B, Feldmann KA (1996) Arabidopsis AUX1 gene: apermease-like regulator of root gravitropism. Science 273: 948–950

Birnbaum K, Jung JW, Wang JY, Lambert GM, Hirst JA, Galbraith DW,Benfey PN (2005) Cell type-specific expression profiling in plants viacell sorting of protoplasts from fluorescent reporter lines. Nat Methods2: 615–619

Birnbaum K, Shasha DE, Wang JY, Jung JW, Lambert GM, Galbraith DW,Benfey PN (2003) A gene expression map of the Arabidopsis root. Sci-ence 302: 1956–1960

Blilou I, Xu J, Wildwater M, Willemsen V, Paponov I, Friml J, Heidstra R,Aida M, Palme K, Scheres B (2005) The PIN auxin efflux facilitatornetwork controls growth and patterning in Arabidopsis roots. Nature433: 39–44

Bourgon R, Gentleman R, Huber W (2010) Independent filtering increasesdetection power for high-throughput experiments. Proc Natl Acad SciUSA 107: 9546–9551

Bowman JL, Baum SF, Eshed Y, Putterill J, Alvarez J (1999) Moleculargenetics of gynoecium development in Arabidopsis. Curr Top Dev Biol45: 155–205

Bowman JL, Smyth DR (1999) CRABS CLAW, a gene that regulates carpeland nectary development in Arabidopsis, encodes a novel protein withzinc finger and helix-loop-helix domains. Development 126: 2387–2396

Bradford JR, Hey Y, Yates T, Li Y, Pepper SD, Miller CJ (2010) A comparisonof massively parallel nucleotide sequencing with oligonucleotide microarraysfor global transcription profiling. BMC Genomics 11: 282

Breakfield NW, Corcoran DL, Petricka JJ, Shen J, Sae-Seaw J, Rubio-Somoza I,Weigel D, Ohler U, Benfey PN (2012) High-resolution experimental andcomputational profiling of tissue-specific known and novel miRNAs inArabidopsis. Genome Res 22: 163–176

Carter AD, Bonyadi R, Gifford ML (2013) The use of fluorescence-activated cell sorting in studying plant development and environmen-tal responses. Int J Dev Biol 57: 545–552


Villarino et al.


















https://www.bioconductor.org/packages/release/bioc/html/topGO.html

https://www.bioconductor.org/packages/release/bioc/html/topGO.html


Chen H, Boutros PC (2011) VennDiagram: a package for the generation ofhighly-customizable Venn and Euler diagrams in R. BMC Bioinformatics12: 35

Cheng Y, Dai X, Zhao Y (2006) Auxin biosynthesis by the YUCCA flavinmonooxygenases controls the formation of floral organs and vasculartissues in Arabidopsis. Genes Dev 20: 1790–1799

Colombo M, Brambilla V, Marcheselli R, Caporali E, Kater MM, Colombo L(2010) A new role for the SHATTERPROOF genes during Arabidopsisgynoecium development. Dev Biol 337: 294–302

Crawford BCW, Yanofsky MF (2008) The formation and function of thefemale reproductive tract in flowering plants. Curr Biol 18: R972–R978

Crawford BCW, Yanofsky MF (2011) HALF FILLED promotes reproductivetract development and fertilization efficiency in Arabidopsis thaliana.Development 138: 2999–3009

Drakakaki G, Zabotina O, Delgado I, Robert S, Keegstra K, Raikhel N(2006) Arabidopsis reversibly glycosylated polypeptides 1 and 2 areessential for pollen development. Plant Physiol 142: 1480–1492

Earley KW, Haag JR, Pontes O, Opper K, Juehne T, Song K, Pikaard CS(2006) Gateway-compatible vectors for plant functional genomics andproteomics. Plant J 45: 616–629

Favaro R, Pinyopich A, Battaglia R, Kooiker M, Borghi L, Ditta G, YanofskyMF, Kater MM, Colombo L (2003) MADS-box protein complexescontrol carpel and ovule development in Arabidopsis. Plant Cell 15:2603–2611

Franco-Zorrilla JM, Cubas P, Jarillo JA, Fernández-Calvín B, Salinas J,Martínez-Zapater JM (2002) AtREM1, a member of a new family of B3domain-containing genes, is preferentially expressed in reproductivemeristems. Plant Physiol 128: 418–427

Franks RG, Wang C, Levin JZ, Liu Z (2002) SEUSS, a member of a novelfamily of plant regulatory proteins, represses floral homeotic gene ex-pression with LEUNIG. Development 129: 253–263

Galbiati F, Sinha Roy D, Simonini S, Cucinotta M, Ceccato L, Cuesta C,Simaskova M, Benkova E, Kamiuchi Y, Aida M, et al (2013) An inte-grative model of the control of ovule primordia formation. Plant J 76:446–455

Gandikota M, Birkenbihl RP, Höhmann S, Cardon GH, Saedler H,Huijser P (2007) The miRNA156/157 recognition element in the 39 UTRof the Arabidopsis SBP box gene SPL3 prevents early flowering bytranslational inhibition in seedlings. Plant J 49: 683–693

Gremski K, Ditta G, Yanofsky MF (2007) The HECATE genes regulatefemale reproductive tract development in Arabidopsis thaliana. Devel-opment 134: 3593–3601

Guida A, Lindstädt C, Maguire SL, Ding C, Higgins DG, Corton NJ,Berriman M, Butler G (2011) Using RNA-seq to determine the tran-scriptional landscape and the hypoxic response of the pathogenic yeastCandida parapsilosis. BMC Genomics 12: 628

Halbritter F, Vaidya HJ, Tomlinson SR (2012) GeneProf: analysis of high-throughput sequencing experiments. Nat Methods 9: 7–8

Hay A, Kaur H, Phillips A, Hedden P, Hake S, Tsiantis M (2002) Thegibberellin pathway mediates KNOTTED1-type homeobox function inplants with different body plans. Curr Biol 12: 1557–1565

Heberle H, Meirelles GV, da Silva FR, Telles GP, Minghim R (2015) In-teractiVenn: a web-based tool for the analysis of sets through Venndiagrams. BMC Bioinformatics 16: 169

Heisler MG, Atkinson A, Bylstra YH, Walsh R, Smyth DR (2001) SPATULA, agene that controls development of carpel margin tissues in Arabidopsis,encodes a bHLH protein. Development 128: 1089–1098

Jasinski S, Piazza P, Craft J, Hay A, Woolley L, Rieu I, Phillips A, HeddenP, Tsiantis M (2005) KNOX action in Arabidopsis is mediated by co-ordinate regulation of cytokinin and gibberellin activities. Curr Biol 15:1560–1565

Kamiuchi Y, Yamamoto K, Furutani M, Tasaka M, Aida M (2014) TheCUC1 and CUC2 genes promote carpel margin meristem formationduring Arabidopsis gynoecium development. Front Plant Sci 5: 165

Kuusk S, Sohlberg JJ, Long JA, Fridborg I, Sundberg E (2002) STY1and STY2 promote the formation of apical tissues during Arabidopsisgynoecium development. Development 129: 4707–4717

Lamesch P, Berardini TZ, Li D, Swarbreck D, Wilks C, Sasidharan R,Muller R, Dreher K, Alexander DL, Garcia-Hernandez M, et al (2012)The Arabidopsis Information Resource (TAIR): improved gene annota-tion and new tools. Nucleic Acids Res 40: D1202–D1210

Lan P, Li W, Lin WD, Santi S, Schmidt W (2013) Mapping gene activity ofArabidopsis root hairs. Genome Biol 14: R67

Larsson E, Roberts CJ, Claes AR, Franks RG, Sundberg E (2014) Polarauxin transport is essential for medial versus lateral tissue specificationand vascular-mediated valve outgrowth in Arabidopsis gynoecia. PlantPhysiol 166: 1998–2012

Leal Valentim F, Mourik S, Posé D, Kim MC, Schmid M, van Ham RC,Busscher M, Sanchez-Perez GF, Molenaar J, Angenent GC, et al (2015)A quantitative and dynamic model of the Arabidopsis flowering timegene regulatory network. PLoS ONE 10: e0116973

Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G,Abecasis G, Durbin R (2009) The Sequence Alignment/Map format andSAMtools. Bioinformatics 25: 2078–2079

Li S, Liberman LM, Mukherjee N, Benfey PN, Ohler U (2013) Integrateddetection of natural antisense transcripts using strand-specific RNAsequencing data. Genome Res 23: 1730–1739

Liljegren SJ, Ditta GS, Eshed Y, Savidge B, Bowman JL, Yanofsky MF(2000) SHATTERPROOF MADS-box genes control seed dispersal inArabidopsis. Nature 404: 766–770

Lindgreen S (2012) AdapterRemoval: easy cleaning of next-generation se-quencing reads. BMC Res Notes 5: 337

Love MI, Huber W, Anders S (2014) Moderated estimation of fold changeand dispersion for RNA-seq data with DESeq2. Genome Biol 15: 550

Ma H, Yanofsky MF, Meyerowitz EM (1991) AGL1-AGL6, an Arabidopsisgene family with similarity to floral homeotic and transcription factorgenes. Genes Dev 5: 484–495

Maere S, Heymans K, Kuiper M (2005) BiNGO: a Cytoscape plugin toassess overrepresentation of Gene Ontology categories in biologicalnetworks. Bioinformatics 21: 3448–3449

Mantegazza O, Gregis V, Chiara M, Selva C, Leo G, Horner DS, KaterMM (2014a) Gene coexpression patterns during early development ofthe native Arabidopsis reproductive meristem: novel candidate develop-mental regulators and patterns of functional redundancy. Plant J 79: 861–877

Mantegazza O, Gregis V, Mendes MA, Morandini P, Alves-Ferreira M,Patreze CM, Nardeli SM, Kater MM, Colombo L (2014b) Analysis ofthe Arabidopsis REM gene family predicts functions during flower de-velopment. Ann Bot (Lond) 114: 1507–1515

Marioni JC, Mason CE, Mane SM, Stephens M, Gilad Y (2008) RNA-seq:an assessment of technical reproducibility and comparison with geneexpression arrays. Genome Res 18: 1509–1517

Martínez-Fernández I, Sanchís S, Marini N, Balanzá V, Ballester P,Navarrete-Gómez M, Oliveira AC, Colombo L, Ferrándiz C (2014) Theeffect of NGATHA altered activity on auxin signaling pathways withinthe Arabidopsis gynoecium. Front Plant Sci 5: 210

Matias-Hernandez L, Battaglia R, Galbiati F, Rubes M, Eichenberger C,Grossniklaus U, Kater MM, Colombo L (2010) VERDANDI is a directtarget of the MADS domain ovule identity complex and affects embryosac differentiation in Arabidopsis. Plant Cell 22: 1702–1715

Mizzotti C, Ezquer I, Paolo D, Rueda-Romero P, Guerra RF, Battaglia R,Rogachev I, Aharoni A, Kater MM, Caporali E, et al (2014) SEEDSTICKis a master regulator of development and metabolism in the Arabidopsisseed coat. PLoS Genet 10: e1004856

Moubayidin L, Ostergaard L (2014) Dynamic control of auxin distributionimposes a bilateral-to-radial symmetry switch during gynoecium de-velopment. Curr Biol 24: 2743–2748

Moussaieff A, Rogachev I, Brodsky L, Malitsky S, Toal TW, Belcher H,Yativ M, Brady SM, Benfey PN, Aharoni A (2013) High-resolutionmetabolic mapping of cell types in plant roots. Proc Natl Acad SciUSA 110: E1232–E1241

Mudge J, Miller NA, Khrebtukova I, Lindquist IE, May GD, Huntley JJ,Luo S, Zhang L, van Velkinburgh JC, Farmer AD, et al (2008) Genomicconvergence analysis of schizophrenia: mRNA sequencing reveals al-tered synaptic vesicular transport in post-mortem cerebellum. PLoSONE 3: e3625

Nagpal P, Ellis CM, Weber H, Ploense SE, Barkawi LS, Guilfoyle TJ,Hagen G, Alonso JM, Cohen JD, Farmer EE, et al (2005) Auxin responsefactors ARF6 and ARF8 promote jasmonic acid production and flowermaturation. Development 132: 4107–4118

Nakagawa T, Kurose T, Hino T, Tanaka K, Kawamukai M, Niwa Y,Toyooka K, Matsuoka K, Jinbo T, Kimura T (2007) Development ofseries of Gateway binary vectors, pGWBs, for realizing efficient constructionof fusion genes for plant transformation. J Biosci Bioeng 104: 34–41

Nookaew I, Papini M, Pornputtapong N, Scalcinati G, Fagerberg L, UhlénM, Nielsen J (2012) A comprehensive comparison of RNA-Seq-basedtranscriptome analysis from reads to differential gene expression and





cross-comparison with microarrays: a case study in Saccharomycescerevisiae. Nucleic Acids Res 40: 10084–10097

Ó’Maoiléidigh DS, Wellmer F (2014) A floral induction system for thestudy of early Arabidopsis flower development. In JL Riechmann, FWellmer, eds, Flower Development. Springer, New York, pp 307–314

Oram RN, Brock RD (1972) Prospects for improving plant protein yieldand quality by breeding. Aust Inst Agr Sci J 38: 163–168

Pagnussat GC, Yu HJ, Ngo QA, Rajani S, Mayalagu S, Johnson CS,Capron A, Xie LF, Ye D, Sundaresan V (2005) Genetic and molecularidentification of genes required for female gametophyte developmentand function in Arabidopsis. Development 132: 603–614

Payne CT, Zhang F, Lloyd AM (2000) GL3 encodes a bHLH protein thatregulates trichome development in Arabidopsis through interactionwith GL1 and TTG1. Genetics 156: 1349–1362

Petersson SV, Johansson AI, Kowalczyk M, Makoveychuk A, Wang JY,Moritz T, Grebe M, Benfey PN, Sandberg G, Ljung K (2009) An auxingradient and maximum in the Arabidopsis root apex shown by high-resolution cell-specific analysis of IAA distribution and synthesis.Plant Cell 21: 1659–1668

Petricka JJ, Schauer MA, Megraw M, Breakfield NW, Thompson JW,Georgiev S, Soderblom EJ, Ohler U, Moseley MA, Grossniklaus U,et al (2012) The protein expression landscape of the Arabidopsis root.Proc Natl Acad Sci USA 109: 6811–6818

Phillips AL, Ward DA, Uknes S, Appleford NE, Lange T, Huttly AK,Gaskin P, Graebe JE, Hedden P (1995) Isolation and expression of threegibberellin 20-oxidase cDNA clones from Arabidopsis. Plant Physiol108: 1049–1057

Pimentel H, Parra M, Gee S, Ghanem D, An X, Li J, Mohandas N, PachterL, Conboy JG (2014) A dynamic alternative splicing program regulatesgene expression during terminal erythropoiesis. Nucleic Acids Res 42:4031–4042

Pinyopich A, Ditta GS, Savidge B, Liljegren SJ, Baumann E, Wisman E,Yanofsky MF (2003) Assessing the redundancy of MADS-box genesduring carpel and ovule development. Nature 424: 85–88

Rapaport F, Khanin R, Liang Y, Pirun M, Krek A, Zumbo P, Mason CE,Socci ND, Betel D (2013) Comprehensive evaluation of differential geneexpression analysis methods for RNA-seq data. Genome Biol 14: R95

Rau A, Gallopin M, Celeux G, Jaffrézic F (2013) Data-based filtering forreplicated high-throughput transcriptome sequencing experiments. Bi-oinformatics 29: 2146–2152

Reyes-Olalde JI, Zuñiga-Mayo VM, Chávez Montes RA, Marsch-Martínez N,de Folter S (2013) Inside the gynoecium: at the carpel margin. Trends PlantSci 18: 644–655

Robinson MD, McCarthy DJ, Smyth GK (2010) edgeR: a Bioconductorpackage for differential expression analysis of digital gene expressiondata. Bioinformatics 26: 139–140

Romanel EAC, Schrago CG, Couñago RM, Russo CAM, Alves-Ferreira M(2009) Evolution of the B3 DNA binding superfamily: new insights intoREM family gene diversification. PLoS ONE 4: e5791

Sauer M, Robert S, Kleine-Vehn J (2013) Auxin: simply complicated. J ExpBot 64: 2565–2577

Savidge B, Rounsley SD, Yanofsky MF (1995) Temporal relationship be-tween the transcription of two Arabidopsis MADS box genes and thefloral organ identity genes. Plant Cell 7: 721–733

Schmieder R, Edwards R (2011) Quality control and preprocessing ofmetagenomic datasets. Bioinformatics 27: 863–864

Schmittgen TD, Livak KJ (2008) Analyzing real-time PCR data by thecomparative C(T) method. Nat Protoc 3: 1101–1108

Schneitz K, Hülskamp M, Pruitt RE (1995) Wild-type ovule developmentin Arabidopsis thaliana: a light microscope study of cleared whole-mount tissue. Plant J 7: 731–749

Sehra B, Franks RG (2015) Auxin and cytokinin act during gynoecialpatterning and the development of ovules from the meristematic medialdomain. Wiley Interdiscip Rev Dev Biol 4: 555–571

Sessions A (1999) Piecing together the Arabidopsis gynoecium. TrendsPlant Sci 4: 296–297

Sessions RA, Zambryski PC (1995) Arabidopsis gynoecium structure in thewild and in ettin mutants. Development 121: 1519–1532

Seyednasrollah F, Laiho A, Elo LL (2015) Comparison of software pack-ages for detecting differential expression in RNA-seq studies. BriefBioinform 16: 59–70

Seymour GB, Østergaard L, Chapman NH, Knapp S, Martin C (2013) Fruitdevelopment and ripening. Annu Rev Plant Biol 64: 219–241

Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N,Schwikowski B, Ideker T (2003) Cytoscape: a software environment forintegrated models of biomolecular interaction networks. Genome Res13: 2498–2504

Sims D, Sudbery I, Ilott NE, Heger A, Ponting CP (2014) Sequencing depthand coverage: key considerations in genomic analyses. Nat Rev Genet15: 121–132

Singh MB, Bhalla PL (2007) Control of male germ-cell development inflowering plants. BioEssays 29: 1124–1132

Skinner DJ, Gasser CS (2009) Expression-based discovery of candidateovule development regulators through transcriptional profiling of ovulemutants. BMC Plant Biol 9: 29

Smyth DR, Bowman JL, Meyerowitz EM (1990) Early flower developmentin Arabidopsis. Plant Cell 2: 755–767

Soneson C, Delorenzi M (2013) A comparison of methods for differentialexpression analysis of RNA-seq data. BMC Bioinformatics 14: 91

Stepanova AN, Robertson-Hoyt J, Yun J, Benavente LM, Xie DY, Dole�zalK, Schlereth A, Jürgens G, Alonso JM (2008) TAA1-mediated auxinbiosynthesis is essential for hormone crosstalk and plant development.Cell 133: 177–191

Suzuki A, Matsushima K, Makinoshima H, Sugano S, Kohno T, TsuchiharaK, Suzuki Y (2015) Single-cell analysis of lung adenocarcinoma cell linesreveals diverse expression patterns of individual cells invoked by a moleculartarget drug treatment. Genome Biol 16: 66

Swaminathan K, Peterson K, Jack T (2008) The plant B3 superfamily.Trends Plant Sci 13: 647–655

Tao Y, Ferrer JL, Ljung K, Pojer F, Hong F, Long JA, Li L, Moreno JE,Bowman ME, Ivans LJ, et al (2008) Rapid synthesis of auxin via a newtryptophan-dependent pathway is required for shade avoidance inplants. Cell 133: 164–176

Trapnell C, Pachter L, Salzberg SL (2009) TopHat: discovering splicejunctions with RNA-Seq. Bioinformatics 25: 1105–1111

Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, Pimentel H,Salzberg SL, Rinn JL, Pachter L (2012) Differential gene and transcriptexpression analysis of RNA-seq experiments with TopHat and Cufflinks.Nat Protoc 7: 562–578

Trigueros M, Navarrete-Gómez M, Sato S, Christensen SK, Pelaz S,Weigel D, Yanofsky MF, Ferrándiz C (2009) The NGATHA genes directstyle development in the Arabidopsis gynoecium. Plant Cell 21: 1394–1409

Wang C, Gong B, Bushel PR, Thierry-Mieg J, Thierry-Mieg D, Xu J, FangH, Hong H, Shen J, Su Z, et al (2014) The concordance between RNA-seq and microarray data depends on chemical treatment and transcriptabundance. Nat Biotechnol 32: 926–932

Wang JW, Czech B, Weigel D (2009) miR156-regulated SPL transcriptionfactors define an endogenous flowering pathway in Arabidopsis thaliana.Cell 138: 738–749

Wang JW, Schwab R, Czech B, Mica E, Weigel D (2008) Dual effects ofmiR156-targeted SPL genes and CYP78A5/KLUH on plastochron lengthand organ size in Arabidopsis thaliana. Plant Cell 20: 1231–1243

Wellmer F, Alves-Ferreira M, Dubois A, Riechmann JL, Meyerowitz EM(2006) Genome-wide analysis of gene expression during early Arabidopsisflower development. PLoS Genet 2: e117

Williams L, Carles CC, Osmont KS, Fletcher JC (2005) A database analysismethod identifies an endogenous trans-acting short-interfering RNAthat targets the Arabidopsis ARF2, ARF3, and ARF4 genes. Proc NatlAcad Sci USA 102: 9703–9708

Woodward AW, Bartel B (2005) Auxin: regulation, action, and interaction.Ann Bot (Lond) 95: 707–735

Wu G, Poethig RS (2006) Temporal regulation of shoot development inArabidopsis thaliana by miR156 and its target SPL3. Development 133:3539–3547

Wu MF, Tian Q, Reed JW (2006) Arabidopsis microRNA167 controls pat-terns of ARF6 and ARF8 expression, and regulates both female and malereproduction. Development 133: 4211–4218

Wuest SE, Vijverberg K, Schmidt A, Weiss M, Gheyselinck J, LohrM, Wellmer F, Rahnenführer J, von Mering C, Grossniklaus U(2010) Arabidopsis female gametophyte gene expression map re-veals similarities between plant and animal gametes. Curr Biol 20:506–512

Wynn AN, Rueschhoff EE, Franks RG (2011) Transcriptomic characteri-zation of a synergistic genetic interaction during carpel margin meristemdevelopment in Arabidopsis thaliana. PLoS ONE 6: e26231


Villarino et al.



Xu X, Zhang Y, Williams J, Antoniou E, McCombie WR, Wu S, Zhu W,Davidson NO, Denoya P, Li E (2013) Parallel comparison of IlluminaRNA-Seq and Affymetrix microarray platforms on transcriptomicprofiles generated from 5-aza-deoxy-cytidine treated HT-29 coloncancer cells and simulated datasets. BMC Bioinformatics (Suppl 9)14: S1

Yadav RK, Girke T, Pasala S, Xie M, Reddy GV (2009) Gene expressionmap of the Arabidopsis shoot apical meristem stem cell niche. Proc NatlAcad Sci USA 106: 4941–4946

Yadav RK, Tavakkoli M, Xie M, Girke T, Reddy GV (2014) A high-resolution gene expression map of the Arabidopsis shoot meristemstem cell niche. Development 141: 2735–2744

Yilmaz A, Mejia-Guerra MK, Kurz K, Liang X, Welch L, Grotewold E(2011) AGRIS: the Arabidopsis Gene Regulatory Information Server, anupdate. Nucleic Acids Res 39: D1118–D1122

Zhang TQ, Lian H, Tang H, Dolezal K, Zhou CM, Yu S, Chen JH, Chen Q, LiuH, Ljung K, et al (2015) An intrinsic microRNA timer regulates progressivedecline in shoot regenerative capacity in plants. Plant Cell 27: 349–360

Zhao S, Fung-Leung WP, Bittner A, Ngo K, Liu X (2014) Comparison ofRNA-Seq and microarray in transcriptome profiling of activated T cells.PLoS ONE 9: e78644

Zhao Y, Christensen SK, Fankhauser C, Cashman JR, Cohen JD, WeigelD, Chory J (2001) A role for flavin monooxygenase-like enzymes inauxin biosynthesis. Science 291: 306–309





Documents

Transcriptomic Signature of the SHATTERPROOF2 ... Technologies Transcriptomic Signature of theSHATTERPROOF2 Expression Domain Reveals the Meristematic Nature of Arabidopsis Gynoecial