Identification of ovule transcripts from the Apospory-Specific Genomic Region (ASGR)-carrier chromosome

Background Apomixis, asexual seed production in plants, holds great potential for agriculture as a means to fix hybrid vigor. Apospory is a form of apomixis where the embryo develops from an unreduced egg that is derived from a somatic nucellar cell, the aposporous initial, via mitosis. Understanding the molecular mechanism regulating aposporous initial specification will be a critical step toward elucidation of apomixis and also provide insight into developmental regulation and downstream signaling that results in apomixis. To discover candidate transcripts for regulating aposporous initial specification in P. squamulatum, we compared two transcriptomes derived from microdissected ovules at the stage of aposporous initial formation between the apomictic donor parent, P. squamulatum (accession PS26), and an apomictic derived backcross 8 (BC8) line containing only the Apospory-Specific Genomic Region (ASGR)-carrier chromosome from P. squamulatum. Toward this end, two transcriptomes derived from ovules of an apomictic donor parent and its apomictic backcross derivative at the stage of apospory initiation, were sequenced using 454-FLX technology. Results Using 454-FLX technology, we generated 332,567 reads with an average read length of 147 base pairs (bp) for the PS26 ovule transcriptome library and 363,637 reads with an average read length of 142 bp for the BC8 ovule transcriptome library. A total of 33,977 contigs from the PS26 ovule transcriptome library and 26,576 contigs from the BC8 ovule transcriptome library were assembled using the Multifunctional Inertial Reference Assembly program. Using stringent in silico parameters, 61 transcripts were predicted to map to the ASGR-carrier chromosome, of which 49 transcripts were verified as ASGR-carrier chromosome specific. One of the alien expressed genes could be assigned as tightly linked to the ASGR by screening of apomictic and sexual F1s. Only one transcript, which did not map to the ASGR, showed expression primarily in reproductive tissue. Conclusions Our results suggest that a strategy of comparative sequencing of transcriptomes between donor parent and backcross lines containing an alien chromosome of interest can be an efficient method of identifying transcripts derived from an alien chromosome in a chromosome addition line.


Background
Apomixis, asexual reproduction through seed, is widespread among flowering plant families, but low in its frequency of occurrence [1]. Different from sexual reproduction, apomictically derived embryos develop autonomously from unreduced ovular cells instead of through fertilization of a reduced egg by a sperm. Therefore, the progeny of an apomictic plant are genetically identical to the maternal plant [2,3]. This trait can be used as an advanced breeding tool in agriculture since it would enable fixation of hybrid vigor and seed propagation of desirable genotypes [4][5][6][7]. No major agriculturally important crop possesses this trait [8][9][10]. Introgression of apomixis into crops through crossing has been impeded by factors such as polyploidy and incompatibility [9]. Therefore, discovery of genetic mechanisms underlying apomixis will be crucial for manipulation of apomixis for introduction into target crops.
Apomixis has been classified into two types and three developmental pathways: gametophytic apomixis, including apospory and diplospory, and sporophytic apomixis, which is also known as adventitious embryony [2]. In sporophytic apomixis, an embryo forms directly from an ovular cell and coexists with the zygotic embryo. For gametophytic apomixis, the embryo develops from an unreduced egg in an embryo sac derived through mitosis of either a somatic nucellar cell (apospory) or the megaspore mother cell (diplospory). In apospory, meiosis either does not complete or its products degenerate while aposporous initials (AIs) develop from one or more somatic nucellar cells. Both genotypes chosen for the present study are aposporous with the trait conferred by genetic elements from Pennisetum squamulatum. Aposporous P. squamulatum has fournucleate embryo sacs that lack antipodals [10]. Apospory in this species is inherited as a dominant Mendelian trait [11] and is associated with an approximately 50 Mb, heterochromatic and hemizygous chromosomal region designated the Apospory-Specific Genomic Region (ASGR), [12,13].
Many transcriptional approaches to discover the regulatory mechanisms and downstream effects associated with apomixis in many species have been undertaken. In Brachiaria, differential display applied to apomictic and sexual ovaries at anthesis yielded two apomixis-specific fragments [14] while a study on earlier sporogenesis and gametogenesis stages identified eleven differentially expressed fragments [15]. In Paspalum notatum, three expressed sequence tags (ESTs), all highly similar in sequence, showed differential expression in flowers between apomictic and sexual F 1 individuals after apospory initiation [16]. An additional 65 genes were identified as differentially expressed between sexual and aposporous plants [17]. cDNA-AFLP analysis in Paspalum simplex yielded transcripts linked to the apomixiscontrolling locus (ACL). Many of these linked fragments showed stop codons and frameshift mutations, suggesting that they are pseudogenes [18]. cDNA-AFLP was also applied to identify apomixis candidate genes in Poa pratensis where 179 transcript-derived fragments from spikelets showed qualitative and quantitative expression differences between apomictic and sexual genotypes [19]. The full-length sequences of two genes of interest, PpSERK (SOMATIC EMBRYOGENESIS RECEPTOR-LIKE KINASE) and APOSTART were obtained and their temporal and spatial expression patterns were assessed by reverse transcription polymerase chain reaction (RT-PCR) and in situ hybridization, respectively. While neither one of these two candidate genes showed apomixis-or sexual-specific expression, quantitative differences in expression between apomictic and sexual genotypes were observed [20].
One apomixis-specific gene was identified from a Panicum maximum ovule cDNA library and shown to be expressed in both aposporous initials and embryos at four days after anthesis [21,22]. Additional genes have been identified in Panicum through microarray and quantitative RT-PCR analysis [23]. In Pennisetum ciliare, differential display and suppression subtractive hybridization were used to identify gene expression differences in ovaries of sexual and apomictic accessions [24,25]. SuperSAGE, a high-throughput differential display approach, has been used to discover several hundred transcripts with heterochronic shifts in expression between apomictic and sexual ovules at multiple stages of development [26,27].
Formation of aposporous initials is the first and most critical event for occurrence of apospory. Because the initiation of sexual and apomictic pathways likely is activated by different signals [28], understanding the molecular mechanism underlying apospory initiation can provide insight into developmental regulation and downstream signaling that results in apomixis. In order to discover candidates for regulating aposporous initial specification in P. squamulatum, we compared two transcriptomes derived from microdissected ovules at the stage of aposporous initial (AI) formation between the apomictic donor parent, P. squamulatum, and its apomictic derivative backcross 8 (BC 8 ) containing a single P. squamulatum chromosome. Initially, a P. glaucum x P. squamulatum F 1 was crossed with a P. glaucum x P. purpureum F 1 and hybrid apomictic individuals with good male fertility were selected [29]. Subsequent backcrosses with tetraploid P. glaucum [30] yielded a BC 8 line that was shown by FISH to contain only one chromosome from P. squamulatum. This single chromosome common to both apomictic BC 8 and P. squamulatum was the ASGR-carrier chromosome based on the transmission of the trait of apomixis and linked molecular markers [31]. We hypothesize that candidate genes regulating aposporous initial specification and localized to the ASGR will function in both PS26 and BC 8 at the same developmental stage and would be identical in sequence as they are related by descent.
The development and commercialization of new massively parallel sequencing platforms have made transcriptome sequencing faster and more affordable. One platform, developed by 454 Life Sciences Corporation, the 454 GS-FLX sequencer, is capable of producing 100 Mb of sequence data with an average read length of 250 bp per bead in a 7-h run [32]. Successful applications of these high-throughput sequencing technologies to transcriptome analysis have been reported [33][34][35][36][37]. Here we present expressed sequence tags (ESTs) generated by Roche 454 high-throughput sequencing technology from dissected ovule tissues staged for aposporous initial formation from two apomictic lines chosen for their common features of apospory and single shared chromosome. Alien chromosome (ASGR-carrier chromosome) expressed transcripts were identified and tested for ASGR linkage and tissue expression.

Results
Aposporous ovule-enriched cDNA samples for sequencing Ovules from PS26 and BC 8 around the stage of aposporous initial formation were manually dissected from pistils ( Figure 1). Three biological replicates of 40 ovules each were collected for both PS26 and BC 8 . The yield of total RNA from each replicate was approximately 20 ng from which 15 ng was used for one-round of T7 RNA polymerase-based RNA amplification. The average yield from one round of amplification was 90 μg. For each library, equal amounts of amplified RNA from each replicate were combined and 15 μg amplified RNA was used for ds-cDNA synthesis. The majority of the ds-cDNA synthesized from amplified RNA was distributed in a size range from 200 bp to 1000 bp ( Figure 2).

Assembly of sequences from PS26 and BC 8 aposporous ovules
Two aposporous ovule transcriptomes, one from PS26 and the other from BC 8 , were sequenced using the high-throughput 454-FLX sequencer. The PS26 transcriptome library contained 332,567 reads with an average read length of 147 base pairs (bp) and the BC 8 transcriptome library contained 363,637 reads with an average read length of 142 bp. Assembly by the Multifunctional Inertial Reference Assembly (MIRA) program [38] resulted in 33,977 contigs from the PS26 ovule transcriptome library and 26,576 contigs from the BC 8 ovule transcriptome library (Additional file 1: PS26_MIRA.fasta, Additional file 2: BC8_MIRA.fasta). The number of reads per contig ranged from 1 to 759 in PS26 assemblies and 1 to 1661 in BC 8 assemblies with the majority having less than 30 reads per assembly in both cases. The numbers of singletons in PS26 and BC 8 libraries were 176 and 78, respectively.

Blast2GO
Contigs from both transcriptome libraries were analyzed for biological functions using Blast2GO [39]. For both libraries, the use of T7 amplified RNA biased the sequencing data toward the 3' UTR region as shown by the BlastX results of the Blast2GO analysis. 5,730 PS26 contigs (~17%) and 4,833 BC 8 contigs (~18%) had hits against the nr database of NCBI with an E-value cut-off of e -06 . For both libraries, 90% of the top BlastX hits were, in order, to Sorghum bicolor, Zea mays or Oryza sativa proteins. Blast2GO was able to fully annotate 4,400 PS26 contigs and 3,692 BC 8 contigs ( Figure 3).
To obtain additional functional data from the shorter reads, a study was initiated to test whether the most significant BlastN EST_other database hit (E-value cut off of e -20 ) could be used as a surrogate longer sequencing read for the PS26/BC 8 transcripts. Approximately 55% (14,518) of the BC 8 contigs had an EST_OTHERS hit ≤e -20 . Blast2GO analysis was used for the BC 8 _EST_-OTHERS best matches and compared with Blast2GO mapping results for the 3692 annotated BC 8 contigs. The majority (84%) of the BC 8 contigs had Blast2GO mapping data identical to the corresponding BC 8 _EST_-OTHERS mapping data while only 5% of the BC 8 contigs had >50% non-matching mapping data. Given the large percentage of identical and/or highly matching mapping data, a library of PS26_EST_OTHERS was also established using the same parameters as BC 8 _EST_-OTHERS. Approximately 53% (18,028) of the PS26 contigs had an EST_OTHERS hit ≤e -20 . Blast2GO was able to fully annotate 12,462 PS26_EST_OTHERS contigs and 10,107 BC 8 _EST_OTHERS contigs.
A Fisher's Exact Test (using GOSSIP; [40]) was done to identify significant differences of expression data between the PS26 and BC 8 libraries and the PS26_EST_-OTHERS and BC 8 _EST_OTHERS libraries. At a false discovery rate (FDR) ≤0.01, 28 GO terms were identified  as different between the PS26 and BC 8 libraries (Table  1). However, when the PS26_EST_OTHERS and BC 8 _EST_OTHERS libraries were compared at FDR <0.05 (at an FDR ≤0.01 no significant results were returned), only 7 GO terms (ribosome, translation, ribosome biogenesis, ribonucleoprotein complex biogenesis, ribonucleoprotein complex, structural constituent of ribosome, cellular component biogenesis) were identified as differentially expressed between the two libraries ( Table 1).

In Silico identification of putative alien expressed transcripts
When MIRA-assembled contigs from the two libraries were analyzed by BlastN with PS26 sequences as queries and BC 8 sequences as the database, a total of 118 comparisons were obtained with 100% sequence identity across an overlapping region ≥100 bp corresponding to 115 unique contigs from the PS26 database and 116 unique contigs from the BC 8 database. The 118 PS26/ BC 8 contigs were further analyzed by aligning the corresponding PS26 and BC 8 contigs with each other, resulting in 61 inter-genotype contigs with no mismatches that were aligned. The average overlapping regions of the 61 inter-genotype contigs was 241 bp (ranging from 181 bp to 419 bp) with an average number of 28 sequence reads. The remaining PS26/BC 8 contigs, while initially identified by BlastN as having 100% identity over a region >100 bp, did not continue to share sequence similarity outside this region and therefore did not align over the whole contig.

Mapping and predicted function of putative ASGR-carrier chromosome transcripts
Up to four primer pairs per contig were used to test for linkage of the 61 contigs to the ASGR-carrier chromosome. Sequence characterized amplified region (SCAR) primer pairs were designed based on the PS26 contig sequence (Additional file 3, Table S1). After screening by PCR against PS26, IA4X (4 × P. glaucum), N37  (P. purpureum) and a small number of progeny from apomictic BC 8 segregating for mode of reproduction, 45 contigs showed specific amplification from PS26 and apomictic BC 8 but no amplification from IA4X or sexual BC 8 individuals ( Figure 4, Table 2) establishing linkage of 45 contigs to the ASGR-carrier chromosome. Singlestrand conformation polymorphism analysis (SSCP) and a CAPS screen using two to four restriction enzymes was applied to the 14 primer pairs which amplified products in both PS26 and IA4X DNA. Four additional contigs could be linked to the ASGR-carrier chromosome using SSCP analysis ( Table 2). The CAPS screen identified a HaeIII polymorphism for PS26_c2552, a transcript also mapped by SSCP.
The markers from the 49 ASGR-carrier chromosomelinked contigs were initially screened on a limited number of apomictic (4) and sexual (4) F 1 s for mapping to the ASGR. This resulted in one contig, PS26_c9369, showing tight linkage to the ASGR as the primers amplified DNAs from only apomictic F 1 s but not sexual F 1 s ( Figure 5, Table 2). The remaining primer sets did not show amplification specificity in the F 1 population; both apomictic and sexual progeny amplified.
A larger F 1 population of 22 individuals (10 apomictic and 12 sexual) was used to map the PS26_c9369 and PS26_c2552 transcripts. PS26_c2552 was mapped based on the HaeIII polymorphism found in the CAPS screen between PS26 and IA4X and also seen in the F 1 population. PS26_c2552 is unlinked to the ASGR as the CAPS polymorphism segregated 1:1 in the population but with 7 sexual and 5 apomictic individuals containing the marker. In comparison, the PS26_c9369 primers remained specific to the 10 apomictic plants and did not amplify the 12 sexual plants.
BlastX searches against NCBI databases were carried out for the 49 PS26/BC 8 ASGR-carrier chromosome linked contigs and best protein hits for 18 contigs are summarized in Table 3. Because the sequences are 3' biased, a BlastN analysis against the expressed sequence tag (EST_OTHERS) database at NCBI with the remaining 31 PS26/BC 8 contigs was done to find potential orthologs from other species. At an E-value cutoff of e -20 , 18 contigs had EST hits (Table 3). A BlastX was performed using these EST sequences to determine if tentative protein functions could be obtained, and the best hits are listed in Table 3. The remaining 13 (27%) contigs did not have hits by either BlastX or BlastN; therefore, they were considered orphan genes.
In order to generate contiguous sequence that might enhance the potential for mapping of contigs in the F 1 population and to extract a longer cDNA sequence for PS26_c9369, a cDNA library containing~300,000 phage plaques was constructed from apomictic BC 8 mature ovary and anther RNA since all 49 ASGR-carrier chromosome transcripts showed expression in these tissues by RT-PCR. Screening of the cDNA library with 27 ASGR-carrier chromosome transcript probes yielded hybridization signals for 24 probes. PCR screening with the ASGR-carrier chromosome-specific primers identified 16 ASGR-carrier chromosome clones and one clone for PS26_c9369. Additional sequence for these clones was generated.
The PS26_c9369 clone contained a 646 bp insert. BlastX analysis identified similarity to a hypothetical protein SORBIDRAFT_10g020450 (XP_002438482.1; evalue 6e -18 ) and Oryza sativa hypothetical protein OsJ_30933 [EAZ15525.1; e-value 4e -16 ] over an~155 bp region. In both sorghum and rice, the area of similarity overlapped a pfam03004: Transposase_24 domain for those proteins. The remaining PS26_c9369 clone sequence was unique. Nine primer sets were designed from nine PS26 contigs to span introns based on predicted splicing of best hits to sorghum. Five primer sets gave strong amplification of PS26 genomic DNA. These amplicons were cloned and sequenced to identify SNPs within the PS26 genomic alleles. CAPS markers could be designed for PS26_c1580 (HpyCH4IV) and PS26_c33813 (HpyCH4IV). Mapping of 4 apomictic and 4 sexual F 1 s did not show tight linkage of these contigs to the ASGR.

Expression profiles of ASGR-linked expressed transcripts by RT-PCR
RT-PCR with RNA extracted from apomictic BC 8 leaf, root, anther, and ovary tissues was completed for the 49 candidate genes mapped to the ASGR-carrier chromosome. Forty-seven were expressed in all four organ types examined ( Figure 6a). However, one putative MADSdomain containing transcription factor, corresponding to contig PS26_c33813, showed amplification only in anther and ovary tissues ( Figure 6b) and contig PS26_c10535, a putative Lon protease, showed expression in all organs except anther.

Discussion
Transcriptional profiling has been extensively used for gene discovery in plants because the absence of introns greatly enhances the information content of the data set and eases data interpretation [41][42][43]. Combined with 454 high-throughput sequencing technology, transcriptome sequencing has become an approach to understand molecular events at the gene expression level on a genome-wide scale. Many successful applications of 454 sequencing technology in transcriptome sequencing and single nucleotide polymorphism (SNP) discovery have been reported [44][45][46][47][48][49] and supported our use of this technology for ovule transcriptome sequencing.
In contrast to studies aimed at identifying genes involved in apomictic reproduction through the identification of differences between apomictic and sexual genotypes, our study compared two apomictic lines for identical transcripts. We previously reported that the ASGR is sufficient to induce apomixis in sexual pearl millet [11,12]; therefore, the trait of apomixis in BC 8 is conferred by the ASGR-carrier chromosome from PS26 [31]. In the present study, we have attempted to identify Figure 4 Examples for mapping of transcripts to the ASGRcarrier chromosome. a: amplification from PS26, N37 and apomictic BC 8 but not from IA4X or sexual BC 8 (PS26_c583: p1510/p1511). b: amplification from PS26 and apomictic BC 8 but not from IA4X, N37 or sexual BC 8 (PS26_c9369: p1514/p1515). c: amplification from PS26, IA4X, N37 and both apomictic and sexual BC 8 (no specificity; PS26_c4364: p1504/p1505). Specificity for PS26_c4364 subsequently was achieved by using a different primer pair (  candidate genes regulating the first step of apomixis, aposporous initial development, by transcriptome analysis of ovules from both PS26 and BC 8 . The ovules were collected at the stage of aposporous initial development, which ranged from no apparent apospory initials (~70%) to distinct aposporous initials observed (~30%). By pooling ovules over this range of development our objective was to minimize the chance of missing genes involved in the pathway of apomixis initiation since we would predict transcription prior to, and perhaps beyond, apospory initial formation. The two ovule transcriptomes generated had an average read length of~150 bp, shorter than the average read length of 200-300 bases for the 454 GS FLX sequencer. The shorter than expected reads could have been due to a combination of factors in preparing the samples for sequencing such as the T7-based antisense RNA amplification method, the conversion of antisense RNA to cDNA, or during the shearing process of the cDNA to prepare the sequencing library. Another possible factor is the species itself. It has been shown that the average read length can vary among different organisms due to differences in AT/GC content [32].
Even with short reads and using stringent comparison conditions to decrease the number of false positive joins between highly similar but not identical transcripts from the two species, 61 putative ASGR-carrier chromosome candidate expressed genes were identified in silico, of which 49 have confirmed linkage to the ASGR-carrier chromosome. The 3' bias of the T7 amplified transcripts helped in the design of primers to discriminate between P. squamulatum and the BC 8 pearl millet genome containing one P. squamulatum chromosome. Our sequencing strategy helped remove, at least to a chromosomal level, the difficulties associated with candidate gene identification by comparative gene expression analysis in apomictic and sexual systems which lack, due to the apomictic process, an ability to generate isogenic lines that vary only in their mode of reproduction. Primer specificity for 48 transcripts was not seen when we attempted to map SCARs to the ASGR using a F 1 population containing many P. squamulatum chromosomes. The additional sequence generated by the phage cDNA clones allowed mapping of two more transcripts in the F 1 population. Greater sequence length would be advantageous for mapping of the ASGR-carrier chromosome transcripts to the ASGR locus.
The use of the gene ontology software Blast2Go allowed comparison of both the PS26 and BC 8 libraries and the PS26_EST_OTHERS and BC 8 _EST_OTHERS libraries created by using the most significant EST_-OTHERS BlastN result as a surrogate for our sequences. The PS26 and BC 8 transcriptomes were almost identical    on a level 3 biological process comparison. While many biological GO terms showed expression level differences when comparing the PS26 and BC 8 libraries, all but seven became non-significant when the PS26_EST_-OTHERS and BC 8 _EST_OTHERS libraries were compared. Six of the transcriptional differences noted belong to genes involved in either ribosomal or translational functions. This difference may be caused by ploidy level difference of PS26 (an octoploid) and BC 8 (a tetraploid). MIRA assembly will separate alleles of genes into different contigs. More PS26 allelic transcripts for genes involved in either ribosomal or translational functions may be expressed in PS26 than in BC 8 thus leading to a higher transcript difference between the libraries. Expression analysis of the ASGR-carrier chromosome linked genes in BC 8 tissue was used to identify transcripts specific to reproductive tissue. All but two ASGR-carrier chromosome transcripts showed constitutive expression in both vegetative and reproductive tissues. The one reproduction-specific transcript (the MADS box gene) did not map to the ASGR. The transcript which could be mapped to the ASGR shows similarity to "hypothetical" proteins in both sorghum and rice containing a Transposase_24 domain. Previous sequencing of BAC clones linked to the ASGR have shown a large number of both Type I and Type II transposons at the locus [50,13]; therefore, it is not surprising that we identified an ASGR-linked transposon transcript in our study.

Conclusions
Our data show that the combination of selecting specific reproductive tissues and sequencing with 454 highthroughput sequencing technology is a promising approach for identification of genes involved in different developmental events and that a need for longer transcript contigs will be a requirement to allow for easier mapping of these transcripts. Given the rapid advancements in next-generation sequencing technologies that enable very deep sequence coverage and paired-end reads, it is likely that the fine tissue dissection requiring RNA amplification of starting materials now could be eliminated to favor longer transcript assemblies.

Plant materials
Pennisetum squamulatum (PS26; PI 319196, 2n = 56) and backcross line 8 (BC 8 )-line 58were used for ovule collection. Compared with the BC 7 line which was used in previous studies [12], the BC 8 -line 58 contains only one alien chromosome from PS26, the ASGR-carrier chromosome [31]. P. glaucum (IA4X), P. purpureum (N37), 4 apomictic and 4 sexual plants from BC 8 -line 58 (BC 8 is facultative thus it produces~18% sexually derived offspring were used for assigning the candidate transcript fragments to the ASGR-carrier chromosome. Twenty-two individuals from a segregating F 1 population between P. squamulatum and P. glaucum were used for mapping the transcript fragments to the ASGR.

RNA isolation
Young florets were dissected from small inflorescence sections whose anthers were at stages between premeiosis and prophase, as determined by acetocarmine staining of anther squashes. One group of florets was stored in RNALater ® solution (Ambion, Austin, TX, USA) at 4°C while the other group was processed for ovary clearing by methyl salicylate [51] to screen for the ovary developmental stage. Ovules from thirty cleared florets  Potential functions of the inter-genotype contigs sharing 100% identity between PS26 and BC 8 ovule transcripts which could be mapped to the ASGR-carrier chromosome based on the best hit of the contigs to protein (BlastX) or nucleotide (BlastN) sequences in NCBI databases.
were examined for each group. If the cleared sample showed AIs in less than 30% of the ovaries and the remaining ovaries were at an earlier developmental stage, then florets stored in RNALater ® solution from the same section of inflorescence were used for ovule dissection. About 40 ovules per sample were collected and total RNA was extracted from the ovules with RNAqueous ® -Micro Kit (Ambion). RNA integrity and quantity were analyzed with an Agilent 2100 Bioanalyser (Santa Clara, CA) at the Interdisciplinary Center for Biotechnology Research (ICBR) of the University of Florida.

RNA amplification and ds-cDNA synthesis for Roche 454 sequencing
With total RNA as starting material, mRNA was amplified by T7-based in vitro transcription following the manual of TargetAmp™2-Round aRNA Amplification Kit 2.0 (Epicentre, Madison, WI). Size range and quantity of the amplified mRNA were measured by both gel electrophoresis and Agilent 2100 Bioanalyser analysis. For each sample, an equal amount of amplified mRNA from the three biological replicates was pooled for ds-cDNA synthesis following the protocol developed by the Schnable lab [52]. Size-range and quantity of ds-cDNA were also analyzed by both gel electrophoresis and using the Agilent 2100 Bioanalyser before submitting the samples for sequencing. For SSCP analysis a Bio-Rad Protean II system (Bio-Rad Laboratories, Hercules, CA) was used to separate fragments in a 1 mm thick 12% non-denaturing PAGE gel with 10% glycerol. PCR product (2 μl) was mixed with 10 μl LIS loading dye (10% sucrose, 0.01% bromophenol blue, and 0.01% xylene cyanol FF), denatured at 98°C for 10 min and cooled to RT for at least 10 min. Sample (10 μl) was loaded and the gel was run in at 200 V for 20-22 hours at 25°C. Silver staining was used to detect the SSCP fragments.

Expression patterns of transcripts mapped to the alien chromosome
Total RNA was extracted from a panel of BC 8 tissues including vegetative (leaf, root), and reproductive tissues at anthesis but before pollination (anther and ovary) with QIAGEN RNeasy ® Plant Mini kit (QIAGEN, Valencia, CA) following the manufacturer's protocol. First-strand cDNA was synthesized following the manufacturer's protocol of First-strand cDNA Synthesis kit (Invitrogen, Carlsbad, CA). RT-PCR reactions were performed using primer pairs which mapped to the ASGRcarrier chromosome in a total volume of 20 μl containing 1 μl of first-strand cDNA, 1 μM of each primer, 1X PCR buffer, 1.5 mM MgCl 2 , 0.2 mM dNTPs, and 1 unit of JumpStart™ Taq DNA polymerase (Sigma, St. Louis, MO). Amplification of contaminating genomic DNA was tested by the inclusion of controls that omitted the reverse transcriptase enzyme from the cDNA synthesis reaction, e.g. no RT controls. The PCR reaction was denatured at 94°C for 5 min followed by 35 cycles of 94°C denaturation for 30 seconds, annealing for 30 seconds at respective temperatures, and 72°C extension for 1 min. RT-PCR products were separated on a 1.5% agarose gel and stained with ethidium bromide. Gel images were captured with the Molecular Imager Gel Doc XR System (Bio-Rad Laboratories).

cDNA library construction
Ovaries and anthers collected from apomictic BC 8 around anthesis but prior to fertilization were frozen in liquid nitrogen. Total RNA was extracted with the RNeasy ® Plant Mini kit (QIAGEN) and then poly A + RNA was purified from total RNA with Oligotex ® mRNA Mini kit (QIAGEN) following the manufacturer's protocols. Yield of mRNA was quantified with a Nanodrop spectrophotometer (Thermo Fisher Scientific Inc., Wilmington, DE). mRNA was used for double-stranded cDNA synthesis with ZAP-cDNA ® Synthesis Kit following the manufacturer's protocol (Stratagene, La Jolla, CA). Ligations, packaging, titering of the packaging reactions, and plaque lifts were conducted following the manufacturer's protocol of ZAP-cDNA ® Gigapack ® III Gold Cloning Kit (Stratagene).

cDNA library screening for target genes
The apomictic BC 8 ovary and anther-enriched cDNA library was screened with α-32 P labeled probes with transcripts mapping to the ASGR-carrier chromosome. The PCR fragments amplified from apomictic BC 8 genomic DNA with the primers used for assigning a fragment to the ASGR-carrier chromosome were diluted and labeled with α-32 P by PCR in a total volume of 20 μl. The labeling reaction contained~0.1 ng primary PCR fragment, 1.25 unit Jumpstart Taq DNA polymerase (Sigma), 0.25 μM of each primer, 0.5 mM dATP/ dTTP/dGTP mixture, 5 μl of α-32 P-labeled dCTP (3000 Ci/mmol) and 1 × PCR buffer (10 mM Tris-HCl, 50 mM KCl, 1.5 mM MgCl 2 ). Probes were purified by passing through homemade Sephadex G-50 (Sigma) columns, which were assembled with Ultrafree ® -MC Centrifugal Filter Units (Millipore, Bedford, MA). Prehybridization of the membranes in hybridization buffer (0.5 M sodium phosphate, 7% SDS, 1 mM EDTA, pH 8.0) containing 0.1 mg ml -1 salmon sperm DNA, which was denatured in boiling water for 10 minutes and cooled on ice before adding to the hybridization solution, was conducted at 65°C for 4 h before addition of the labeled, denatured probe. Hybridization was conducted at 65°C overnight followed by three washes at the same temperature for 30 min each with the following buffers: 1) 1 × SSC, 0.1% SDS; 2) 0.5 × SSC, 0.1% SDS; 3) 0.1 × SSC, 0.1% SDS. After the final wash, membranes were wrapped with plastic film and exposed to x-ray film overnight at -80°C prior to manually developing with Kodak ® GBX Developer and Fixer (Thermo Fisher Scientific Inc). Autoradiographs were aligned with the respective plates to recover hybridizing plaques with sterile glass pipettes. Recovered plaques were released in tubes containing 1.0 ml SM phage buffer (according to the formula in the manual of ZAP-cDNA ® Gigapack ® III Gold Cloning Kit) and 20 μl chloroform (Sigma). After overnight elution at 4°C, 1 μl SM buffer of each recovered sample was used for PCR to verify positive signals. Since the primary screening was carried out with a high density of plaque clones, the recovered positive plaques were purified after secondary and tertiary screens at much lower densities. Single plaques showing positive hybridization signals were recovered in 500 μl SM buffer with 10 μl chloroform (Sigma) at 4°C.
Sequencing and mapping of candidate cDNA clones to the ASGR In vivo excision of single plaque clones was conducted using ExAssist ® helper phage with SOLR ® strain following the protocol in the manual of ZAP-cDNA ® Gigapack ® III Gold Cloning Kit (Stratagene). Single colonies containing the pBluescript double-stranded phagemid with the cloned cDNA insert were isolated and cultured in liquid Luria-Bertani (LB) medium containing 100 μg mL -1 ampicillin at 37°C overnight. An aliquot of each culture was further grown in freeze broth containing 100 μg mL -1 ampicillin at 37°C overnight and then stored at -80°C before sending out for sequencing. Sequencing was conducted with M13 primers (Georgia Genomics Facility, Athens, GA). Vector and bad quality sequences were trimmed from the original sequences with VectorNTI Advanced 10 (Invitrogen) and primers were designed with VectorNTI using the high quality cDNA sequences. Primers were then tested with apomictic and sexual F 1 s for linkage to the ASGR as described above.