Differential Expression Profiling of Microspores During the Early Stages of Isolated Microspore Culture Using the Responsive Barley Cultivar Gobernadora

In barley, it is possible to induce embryogenesis in the haploid and uninucleate microspore to obtain a diploid plant that is perfectly homozygous. To change developmental fates in this fashion, microspores need to engage in cellular de-differentiation, interrupting the pollen formation, and restore totipotency prior to engaging in embryogenesis. In this work, we used the barley cultivar Gobernadora to characterize the transcriptome of microspores prior to (day 0) and immediately after (days 2 and 5) the application of a stress pretreatment. A deep RNA-seq analysis revealed that microspores at these three time points exhibit a transcriptome of ∼14k genes, ∼90% of which were shared. An expression analysis identified a total of 3,382 differentially expressed genes (DEGs); of these, 2,155 and 2,281 DEGs were respectively identified when contrasting expression at days 0 and 2 and at days 2 and 5. These define 8 expression profiles in which DEGs share a common up- or down-regulation at these time points. Up-regulation of numerous glutathione S-transferase and heat shock protein genes as well as down-regulation of ribosomal subunit protein genes was observed between days 0 and 2. The transition from microspores to developing embryos (days 2 vs. 5) was marked by the induction of transcription factor genes known to play important roles in early embryogenesis, numerous genes involved in hormone biosynthesis and plant hormonal signal transduction in addition to genes involved in secondary metabolism. This work sheds light on transcriptional changes accompanying an important developmental shift and provides candidate biomarkers for embryogenesis in barley.

To change developmental fates, microspores need to engage in cellular de-differentiation, interrupting the transcriptional and translational activities leading to pollen formation , and restore totipotency (Elhiti et al. 2013) in view of engaging in a new developmental path, in this case embryogenesis (Hosp et al. 2007;Seifert et al. 2016). A few previous studies have characterized microspore gene expression to shed light on the mechanisms underlying microspore development in this context in barley (Vrinten et al. 1999;Maraschin et al. 2006;Muñoz-Amatriaín et al. 2006), wheat (Sánchez-Díaz et al. 2013;Seifert et al. 2016) and rapeseed (Joosen et al. 2007;Malik et al. 2007). In barley, experiments were performed either on isolated microspores (Vrinten et al. 1999;Maraschin et al. 2006) or whole anthers (Muñoz-Amatriaín et al. 2006). In wheat and rapeseed, gene expression profiling was performed on freshly and pretreated microspores in addition to embryogenic induced microspores (Joosen et al. 2007;Malik et al. 2007;Seifert et al. 2016). With the exception of Seifert et al. (2016) who used a comprehensive RNA-seq approach, all previous work was performed on candidate genes or using arrays that interrogated only a subset of all genes (from as few as 10 candidate genes to 21k). Nonetheless, these studies provided interesting glimpses into the main genes and metabolic processes involved in this striking change of developmental fate.
In barley, Maraschin et al. (2006) used a specially designed gene chip containing 1,421 ESTs isolated from the early stages of barley zygotic embryogenesis. This tool was used to analyze either freshly isolated or pretreated uninucleate microspores (following four days of heat and osmotic stress). Globally, a total of 96 differentially-expressed ESTs were identified; these coded for genes involved in the activation of protein degradation, starch and sugar hydrolysis, stress responses and cell signaling metabolisms as well as in the inhibition of programmed cell death (Maraschin et al. 2006). As for the work of Muñoz-Amatriaín et al. (2006), because it was carried out on whole anthers, it is impossible to know which transcripts were expressed in the microspores inside the anthers vs. the rest of the anther. Among closely related cereals, in wheat, Seifert et al. (2016) characterized the transcriptome of freshly isolated microspores, microspores after a pretreatment (4°for 10 days) and microspores in culture (for 8 days). These authors identified genes encoding transcription factors known to mark the induction of embryogenesis such as AINTEGUMENTA-like 5 (AIL5) and BABY BOOM (BBM). A differential gene expression analysis identified a group of up-regulated genes involved in various epigenetic metabolisms such as DNA methylation, histone methylation and histone deacetylation.
Although barley was among the first species studied for gene expression of pretreated microspores and anthers, as described above, no comprehensive study has explored the entire transcriptome of isolated barley microspores engaging in induced embryogenesis. Consequently numerous aspects still remain obscure such as the synthetic and response pathways for various plant growth regulators as well as key transcription factors. In this work, our objective was to extensively characterize the transcriptome of barley microspores prior to (day 0) and immediately after (days 2 and 5) the application of a 48h pretreatment (thermal and osmotic stress) that efficiently induces embryogenesis in the highly responsive barley cultivar Gobernadora. To investigate key metabolic pathways involved in microspores at these stages of development, deep transcriptome sequencing was used to both catalog the genes expressed as well as those that were differentially expressed.

Plant materials
Donor plants of barley (H. vulgare ssp. vulgare cv Gobernadora, a two-row spring barley) were grown in a greenhouse and uniform immature spikes containing microspores at the mid-late to late-uninucleate stage were harvested as described by . We then isolated and purified microspores at three time points: Day 0 (freshly harvested spikes), Day 2 (immediately after completion of the pretreatment) and Day 5 (after three days in culture). Day 0 microspores were isolated from freshly harvested spikes containing haploid and uninucleate microspores and the uniformity of microspores was improved using gradient centrifugation (20% maltose-mannitol; 900xg at 12°). For day 2 and 5, the spikes were subjected to a 48-h pretreatment combining thermal (26°) and osmotic (0.3M; pH at 5.34) stresses. After pretreatment, microspores were harvested and purified using gradient centrifugation (as above). A 0.5M of these isolated microspores were collected as samples of day 2 while the remaining microspores were plated on a two-layer (solidliquid) embryogenesis induction media developed by Li and Devaux (2003) and optimized by . The optimization consisted of: (i) use of 0.3 mg/l of Thidiazuron (Sigma-Aldrich, Oakville, Ontario, Canada) rather than BAP (at 1.0 mg/l), (ii) addition of 0.6 mg/l of Dicamba (Sigma-Aldrich, Oakville, Ontario, Canada), (iii) addition of 50 mg/l each of arabinogalactan (Sigma-Aldrich, Oakville, Ontario, Canada) and arabinogalactan proteins (Sigma-Aldrich, Oakville, Ontario, Canada) and (iv) removal of ovaries. Finally, to maximize the uniformity of the microspores harvested at day 5, we collected the liquid phase containing the microspores and enriched for embryogenic microspores using a 25% maltose-mannitol gradient centrifugation (300xg; 12°). Microspores were produced in four biological replicates and, after isolation, samples were immediately frozen in liquid nitrogen and kept at -80°p rior to RNA isolation.

Cellular fixation and microscopy of microspores
Freshly isolated microspores at the three time points were collected, fixed and DAPI stained for microscopy exactly as described in González-Melendi et al. (2005)  Data processing, analysis of differential gene expression and gene clustering Using Trimmomatic v0.33 (Bolger et al. 2014), raw paired-end reads were pre-processed at a Phred quality score threshold of $25 for the 39 end and trimmed reads shorter than 25 nt were discarded. Trimmed reads were aligned to the barley reference genome (Hv_IBSC_PGSB_v2. dna.toplevel; ftp://ftp.ensemblgenomes.org/pub/release-36/) using Tophat v2.1.1 (Kim et al. 2013). Using HTSeq v0.6.1p1 (Anders et al. 2014), we calculated the number of reads mapped to exons using the reference transcriptome and generated a read-count matrix. Prior to performing differential gene expression analysis, we filtered the readcount matrix and kept only genes with $5 reads per million for a minimum of 4 samples and normalized gene expression based on the TMM method using edgeR Robinson and Oshlack 2010). In edgeR, we performed a multidimensional scaling (MDS) analysis to assess the degree of uniformity among replicates of the three developmental stages. Using edgeR, we examined the overlap among genes at the three stages of microspore development to identify genes exclusively expressed at a specific stage of development and clustered them under a category named ON/OFF genes. We then used edgeR to identify differentially expressed genes (DEGs) and measured the significance of expression changes using the generalized linear model (glm) test for two developmental transitions: (i) from day 0 to day 2 and (ii) from day 2 to day 5. Results were filtered and considered as significant if both log 2 FC $ |2.0| and q-value # 0.01 were observed. Finally, DEGs were clustered in eight groups representing their expression pattern (while the expression of a gene can go up, down or remain unchanged) between days 0 to 5 (successively from day 0 to day 2 and from day 2 to day 5).

Gene functional annotation
We retrieved functional annotation of barley genes using BioMart v0.7 available on Phytozome (https://phytozome.jgi.doe.gov) for KEGG functional orthologs (KO) and Panther description. Reference orthologous genes were retrieved using the online EMBL-EBI HMMER program (https://www.ebi.ac.uk/Tools/hmmer/) with the SwissProt database restricted to the A. thaliana and O. sativa species.

Data availability
The complete set of raw and mapped RNA-seq reads were deposited in the Sequence Read Archives (SRA) under accession number SRP127768.

RESULTS AND DISCUSSION
Isolation and purification of microspores at three key stages of IMC To better understand the changes in gene expression that underlie the developmental switch that occurs during isolate microspore culture (IMC) in barley, we used the cv. Gobernadora known as one of the most responsive genotypes in anther culture (Marchand et al. 2008) and we characterized the transcriptome of microspores at days 0, 2 and 5. As can be seen in Figure 1, without and with DAPI staining, on day 0, microspores were characterized by a single nucleus positioned across the cell wall ( Figure 1a) corresponding to the late uninucleate stage known as the most embryogenic-responsive microspore (Kasha et al. 2001); on day 2, a little enlarged microspores showed a single nucleus migrating toward the center of the cells as well as a few microspores that had two nuclei ( Figure 1b) and (iii) on day 5, some star-like microspores as well as many, more advanced, multicellular structures (MCSs) exhibiting 2 to 6 nuclei were seen ( Figure 1c). No or very few damaged or dead cells were observed. Relative to the phenotypes described by Maraschin (2005), our samples can be described as highly similar to day 0, slightly more advanced at the enlarged microspore (EM) stage (day 2) and exceeded in many cases the star-like microspore (SL) stage (day 5) and mostly corresponded to the multicellular structure (MCS) stage indicating that microspores had engaged in embryogenesis. Such MCSs were typically observed 2 to 5 days later by Maraschin (2005). We suggest that the faster microspore development observed could be attributed to three factors: (i) a superior responsiveness of the genotype Gobernadora or (ii) the use of a more efficient stress pretreatment or (iii) the use of a better induction medium.
Transcriptomes at three stages of isolated microspore culture To obtain a comprehensive overview of the set of genes expressed in microspores during the early stages of IMC, we used an RNA-seq approach. Four biological replicates at each stage yielded a total of more than 279M 50-nt paired-end reads, for an average of 92M reads per cell type. To assess the degree of uniformity among replicates, we performed an MDS analysis based on genome-guided transcriptome assemblies. As can be seen in Figure 2a, we observed three highly distinctive and very tight clusters indicating that the different replicates were tremendously uniform and that each stage has a distinct transcriptomic signature. Even if some morphological heterogeneity was observed among day-5 microspore replicates, the tightest clustering of these replicates in the MDS plot suggested an enrichment of cells having a high homogeneity in their transcriptional response to embryogenesis induction.
Overall the analysis, a total of 14,367 genes was detected when summed across all three stages and the vast majority of these genes (13,037; 90.7%) were expressed at all three developmental stages ( Figure  2b), although not necessarily in equal abundance. Among individual stages, we observed an increase in the number of genes expressed in microspores from a low of 13,105 at day 0, to 13,965 at day 2 and a maximum of 14,289 at day 5. Although small (2.9%), the number of genes expressed exclusively at a specific stage increased from 11 (day 0) to 34 (day 2) to 367 (day 5). Finally, the number of genes expressed in two of the stages was also relatively small (6.4%), ranging from 24 (day 0 and day 5) to 861 (day 2 and day 5). In the next section, we describe further some of the broad metabolic pathways in which genes expressed exclusively at a stage are involved. Mainly due to their microarray (1,421 ESTs) designed to interrogate genes expressed in early barley zygotic embryogenesis, Maraschin et al. (2006) detected a lower number of transcripts (418) representing only 2.9% of the genes interrogated in our work. In wheat microspores at comparable stages, Seifert et al. (2016) reported similar results both in terms of the number of genes (15,598) and the overlap of genes expressed across all developmental stages (11,765; 90.4%). These authors also observed an increase in the number of genes expressed in microspores from developmental stage 1 (14,470; similar to day 0 in this work) to stage 2 (14,666; day 2 here), while, in contrast, they reported a slightly decrease at stage 3 (13,128; similar to day 5). They also detected a small (4.5%) number of genes expressed exclusively at a single stage: 24, 11 and 666 in stages 1, 2 and 3, respectively. In conclusion, we succeeded in isolating highly uniform populations of microspores at three phenotypically distinct stages of IMC and, while each exhibited a distinct transcriptomic profile, the number of genes underlying these different profiles via their presence or absence was rather small (dozens to at most a thousand).
Genes exclusively expressed in microspores at day 5 provided highlight About metabolism governing the embryogenesis induction To provide a comprehensive overview of genes expressed exclusively at a specific stage of development, we explored metabolisms governed by these genes. While no or too low functional annotations were available for genes exclusively expressed in microspores at day 0 (only 11 genes) and 2 (34 genes), we limited the investigation to the 367 genes specifically expressed to microspores at day 5 and considered them as potential marks of the induction of embryogenesis. Interestingly, we observed genes (Table 1)  HORVU7Hr1G111060) and WUSCHEL-RELATED HOMEOBOX 4 (WUS; HORVU5Hr1G022120). BBM and AIL5 are both AP2/ERF transcription factors known to be expressed early and to play important roles in embryo development in B. napus (Boutilier et al. 2002) and A. thaliana, respectively (Tsuwamoto et al. 2010). Orthologs of these genes were recently reported to be turn on early during microspores culture in wheat (Seifert and al. 2016). WUS is a homeobox transcription factor reported to play an essential role in maintaining cells in a state of proliferation and responsiveness to other developmental cues (Mayer et al. 1998;Gallois et al. 2002;Smertenko and Bozhkov 2014) in addition to inducing the vegetative-to-embryonic transition in A. thaliana somatic embryogenesis (Zuo et al. 2002;Smertenko and Bozhkov 2014). Together, the expression induction of these three genes in microspores at day 5 suggests the microspore commitment in the embryogenesis developmental pathway and is consistent with the phenotype previously observed.
Then, we inspected this set of 367 genes to look for other hallmarks of microspore transition not previously reported in microspore embryogenesis. The major finding was the detection of genes belonging to hormonal biosynthetic pathways in addition to genes involved in plant Figure 1 Representative samples of microspores obtained from the cv. Gobernadora at the three stages of early isolated microspore culture. (a) Mid-late to late uninucleate microspores (day 0), (b) enlarged microspores (day 2) and (c) star-like microspore to multi-cellular structure microspore phenotype (day 5). Images were captured using a Zeiss Apoptome microscope under UV laser illumination (excitation of 390/22 nm and emission of 460/50 nm) at a 20x magnification without staining (left) and with DAPI staining (right).
hormones signal transduction processes (Table 1). For hormone biogenesis (Table 1), two genes encoding the linoleate 13S-lipoxygenase enzyme (LOX8; HORVU7Hr1G050660 and HORVU7Hr1G050670) known as the first enzyme involved in the synthesis of jasmonic acid (Wasternack 2007) were detected. Similarly, a gene encoding xanthoxin dehydrogenase (ABA2; HORVU2Hr1G015140) enzyme, involved in the abscisic acid biogenesis pathway was also detected only at day 5. Of plant hormones signal transduction pathways, we detected numerous genes involved in the signal transduction mediated by the cytokinin, abscisic acid, auxin, ethylene and brassinosteroid (Table 1). Of these genes, we detected few hormonal receptors homologous to the Arabidopsis PYL1 (HORVU1Hr1G050110) and the rice ETR4 (HOR-VU6Hr1G071860) and HK3 (HORVU3Hr1G094870) genes known to respectively encode abscisic acid, ethylene and a cytokinin receptors required to mediate the hormone transduction response. In addition, numerous genes involved in the signaling cascade related to auxin were detected such as IAA15 (HORVU1Hr1G025670), IAA26 (HOR-VU5Hr1G081180), IAA20 (HORVU7Hr1G026970) genes as well as GH3.2 (HORVU1Hr1G066340) and GH3.4 (HORVU3Hr1G074230) genes known as induced by this signal transduction pathway. Finally, we detected the expression a gene homolog to the rice CYCD3-2 (HOR-VU5Hr1G050270) gene known as induced by the brassinosteroid hormone signal transduction and involved on cell division (Hu et al. 2000).
Interestingly, _ Zur et al. (2015a and2015b) reported that the endogenous level of natural hormones (auxin/cytokinin/ABA) and its balance with exogenously applied hormones can be crucial both for the yield and quality of microspore-derived embryos and they suggested that hormonal homeostasis might be one of the most important factors determining cell embryogenic competency. Also, it was demonstrated that an addition of these three hormones to microspores in culture significantly increased embryogenesis when applied a couple of hours from the beginning of microspore culture (Ahmadi et al. 2014). Since only cytokinin and auxin were added to our culture medium, our results suggest that microspores of barley cv. Gobernadora can potentially rapidly activate hormone biogenesis and mediate hormonal signal transduction to other hormones within the first days of culture. This could explain in part its high aptitude to form embryos. Together, these results suggest that the transition toward embryogenesis involves the expression of genes governing hormonal biogenesis and signal transduction pathways. These genes could potentially serve as biomarkers to compare the efficiency of induction of embryogenesis following various pretreatments or in different genetic backgrounds.

DEG analysis and clustering
In addition to genes that were only detected at a specific stage, changes occurring in the abundance of transcripts could help understand the developmental shift undergone by these microspores. In principle, when the developmental fate of the uninucleate microspore is shifted from pollen formation to embryogenesis, we would expect to see a decrease in expression of genes associated with pollen formation and an increase in genes associated with embryogenesis. Thus, differentially expressed genes (DEGs) could shed light on the key metabolic changes driving this switch. Overall, a total of 3,382 DEGs (23.5%) were significantly over-or under-expressed; of these genes, 2,155 and 2,281 DEGs, respectively, were identified when contrasting expression at day 0 and day 2, and when comparing day 2 and day 5. For both categories of DEGs, most transcripts were up-regulated (80.4% and 66.7%, respectively). Our results are similar to those reported in wheat by Seifert et al. (2016) in terms of the number of DEGs when contrasting comparable stages of microspore development (6,385 out of 15,598), but contrast markedly in terms of up-or down-regulation. Indeed, these authors reported only 33.3% and 43.0% of up-regulated genes for the two transitions. This low proportion of up-regulated genes is rather atypical of what observed here or previously reported in barley (Maraschin et al. 2006).
To facilitate the analysis of this large set of DEGs, we grouped them into eight profiles (illustrated in Figure 3). Below, we describe broad metabolic changes associated with these various clusters of genes sharing a similar expression profile based on the KEGG and Panther annotations (the complete annotation is detailed in Table S1, Table S2,  Table S3, Table S4, Table S5, Table S6, Table S7, and Table S8 corresponding to each gene clusters). Metabolisms controlled by differentially expressed genes in microspore from day 0 and 2 Of the eight clusters, three (4, 5 and 7) shared an up-regulation of genes between days 0 and 2, but only the latter two exhibited an increase occurring only at this stage. Among these, we found numerous genes belonging to four main categories: carbohydrate metabolism, hormonemodulated gene expression, proteolytic genes, cytoprotection and hormone biogenesis. These have been proposed to participate in the interruption of pollen development . Despite being reported in much greater numbers here, examples of such genes (especially among the first two categories) have already been described in barley by Maraschin et al. (2006). In what follows, we focus on genes involved in cytoprotection, proteolysis and genes related to hormone biogenesis, three previously poorly described categories.
Of genes involved in cytoprotection (Table 2), we detected nine genes coding Glutathione S-transferases (GSTs) and thirteen heat shock proteins (HSPs), both, known to be induced in response to stress and involved in cytoprotection. Specifically, GSTs act in the glutathionemediated detoxification metabolism protecting cells against reactive oxygen species (ROS) (Chen and Singh 1999;Garretón et al. 2002). This result is concordant with the findings of Maraschin et al. (2006) who reported the induction of a GST gene in microspores subjected to stress. Heat shock proteins (HSPs) are known to act in various ways to protect proteins such as by assisting in refolding, preventing aggregation or by acting as a co-chaperone (Park and Seo 2015) Previously, members of the HSP family have been reported to be highly expressed in microspores subjected to heat and starvation stresses: HSP90 in Brassica (Seguí-Simarro et al. 2003), HSP70 in Brassica (Cordewener et al. 1997;Seguí-Simarro et al. 2003) andCapsicum (Bárány et al. 2001), as well as HSP20 in tobacco (Zarsky et al. 1995). The two latter were detected in barley for the first time in this work, but with a distinct expression pattern. The 9 HSP20 genes were up-regulated from day 0 to day 2 (Figure 3; Cluster 5) while the five HSP70 genes were initially up-regulated and then down-regulated from day 2 to day 5 (Figure 3; cluster 7). In addition to HSPs, we detected a gene encoding heat shock transcription factor (HSF) known to increase the transcriptional response of HSP genes (barley homolog of the HSFA4A (HOR-VU1Hr1G081300) gene).
Some of the proteolytic genes identified encode components of the proteasome and the ubiquitin-mediated proteolysis system as well as proteases. We detected an induction of proteolytic genes homologous to components of the 26S proteasome regulatory complex such as the ATPase RPT4 (HORVU4Hr1G027260) and PSMD10 (HORVU3Hr1G023740; HORVU3Hr1G033250; HORVU7Hr1G029900) as well as the proteasome endopeptidase complex (HORVU5Hr1G109720). In addition, we detected a few genes encoding a ubiquitin-ligase (UB-E3) protein. While expressed, no changes in expression were detected for the ubiquitinactivating (UB-E1) and the ubiquitin-conjugating (UB-E2) enzymes. In addition, numerous proteases were detected such as the cathepsin B-like protease 2 (CATHB2; HORVU4Hr1G010300), the serine carboxypeptidase 1 (CBP1; HORVU3Hr1G096830), the aminopeptidase M1 (APM1; HORVU5Hr1G057330), the endopeptidase Clp (CLPR1; HOR-VU1Hr1G094480) and numerous the aspartyl proteases (14 genes) such as the HORVU3Hr1G056630 gene coding for a phytepsin protein previously reported by Maraschin and al (2006). In wheat, Seifert et al. (2016) did not report the activation of such enzymes and metabolisms; rather, they reported an over-representation of genes related to proteolysis among a cluster of down-regulated genes when microspores underwent the transition toward the third stage of development studied.
Three clusters (1, 2 and 8) shared a down-regulation of genes between days 0 and 2, but only the second exhibited a decrease occurring exclusively at this stage (genes that decrease between days 0 and 2 and stay low expressed). Among these, the most highly represented categories were genes encoding proteins involved in genetic information processing and DNA organization such as nucleosome assembly factors. For genetic information processing, many genes (Table 3) involved in translation coding the small ribosomal subunit (a total of 7 genes encoding distinct proteins) or the large ribosomal subunit (a total of n 15 genes) were observed. Down-regulation of these genes (a total of 22) coding for ribosomal components upon application of a stress is consistent with previous reports (Rashid et al. 1982, Hoekstra et al. 1992, Telmer et al. 1995. Nonetheless, Maraschin et al. (2006) detected an increased expression of the gene coding the 60S ribosomal protein L26A. While a gene coding exactly the same ribosomal protein was not detected in our experiment, our more extensive data suggests strongly a down-regulation of genes coding ribosomal proteins. In agreement with our results, Malik et al. (2007) also observed a down-regulation of several genes encoding the small and large ribosomal subunits in B. napus microspores after three days of heat and osmotic pretreatment. In addition to ribosomal protein genes, we detected a down-regulation of three homologs of the A. thaliana NOP5A (HORVU2Hr1G022140 and HORVU1Hr1G083960) and FIB1 (HOR-VU6Hr1G091860) genes, respectively coding for the nucleolar protein 56 and the rRNA 2'-O-methyltransferase fibrillarin protein (Table 3), all known to be required for 60S ribosomal subunit biogenesis. Concordantly, Maraschin et al. (2006) also observed a down-regulation of the gene coding for the fibrillarin protein in barley microspores after stress pretreatment. For the second category, we detected a total of 7 genes (Table 3) encoding all core histone components of the nucleosome: histones H2A (HORVU3Hr1G116550, HORVU4Hr1G008800, HORVU4Hr1G058940, HORVU6Hr1G092280), H2B (HORVU3Hr1G086610), H3 (HOR-VU1Hr1G020050) and H4 (HORVU1Hr1G017830). Although that we observe few microspores having initiated a nuclear division at day 2, a reduction of expression for genes encoding these proteins may reflect the arrest of the pollen formation that involve two successive mitotic division during the gametogenesis.
Metabolisms controlled by differentially expressed genes in microspore from day 2 and 5 When then expanded our investigation to the other up-regulated DEGs between days 2 and 5 (clusters 6 and 8), i.e., those that were also expressed at earlier stages but less strongly. First, numerous genes encoding enzymes involved in secondary metabolism were found, including enzymes of the phenylpropanoid pathway. For instance, we detected many genes encoding enzymes such as shikimate O-hydroxycinnamoyltransferase (HCT1; HORVU2Hr1G086380), phenylalanine ammonia-lyase (PAL2; HORVU2Hr1G038140, HOR-VU2Hr1G038120, HORVU0Hr1G016330), cinnamoyl-CoA reductase (CCR1; HORVU7Hr1G030380) and 4-coumarate-CoA ligase (4CL3; HORVU6Hr1G030390 and 4CL4; HORVU7Hr1G111130). In addition to the one mentioned above, we detected two supplementary genes encoding a linoleate 13S-lipoxygenase (homologous to LOX6 and LOX8) and a linoleate 9S-lipoxygenase (homologous to LOX4) (Table 1). In addition to these enzymes, we detected allene oxide synthase (AOS1; n HORVU5Hr1G098090 and AOS2; HORVU4Hr1G066270), allene oxide cyclase (AOC; HORVU6Hr1G081000), acyl-CoA oxidase (ACOX2; HORVU7Hr1G029110 and ACOX3; HORVU7Hr1G083490) and 12-oxophytodienoic acid reductase (OPR7; HORVU7Hr1G095960) genes, all involved in jasmonic acid synthesis. Jacquard et al. (2009) similarly reported an up-regulation of genes encoding enzymes involved in the synthesis of jasmonic acid and in the phenylpropanoid pathway in barley anther culture. In contrast, in wheat, Seifert et al. (2016) did not report the activation of such enzymes and metabolisms. It remains too early to establish direct roles for such genes and pathways as functional analyses would need to be conducted to validate their function in microspore embryogenesis.
Previously, we reported a reduction in the expression of genes encoding both ribosomal protein subunits and nucleosome assembly components during the transition from day 0 to 2. It is interesting to observe a massive increase in the expression of genes encoding these proteins as the microspore transitions from day 2 to day 5. Indeed, we observed a total of 13 genes contributing to the synthesis of the small and large ribosomal subunits with, respectively, 2 and 11 genes (Table 4). However, these ribosomal genes encode the 30S and 50S subunits that make up the 80S ribosome found in organelles such as the chloroplast and mitochondria. In addition, we detected a total of 38 genes (Table 4) encoding all core histone components of the nucleosome: histones H2A (8 genes), H2B (5 genes), H3 (13 genes) and H4 (12 genes). Since we observe that a majority of microspores have initiated a nuclear division by day 5, an increase in the expression of genes encoding these proteins may reflect that the microspore is preparing for a first cell division as a first step in the embryogenic pathway.
Some genes exhibiting the expression profile corresponding to cluster 8, successively down-and up-regulated from day 0 to day 2 and from day 2 to day 5 (Figure 3), were homologous to the rice minichromosome maintenance proteins MCM2 (HORVU1Hr1G063700), MCM3 (HOR-VU1Hr1G070110), MCM6 (HORVU1Hr1G029770) and MCM7 (HORVU5Hr1G028260). These are known components of a DNA helicase involved in DNA replication and cell cycle (Table 4). To date, minichromosome maintenance protein genes have not been reported in the microspore embryogenesis system. Nonetheless, evidence has been found that the MCM2 (Ni et al. 2009) and MCM7 (Springer et al. 2000;Holding and Springer 2002) genes were essential during the early stages of zygotic embryogenesis in A. thaliana. Interestingly, Ni et al. (2009) showed that a disruption of MCM2 gene is lethal early during embryogenesis and, by contrast, its over-expression results in an inhibition of endoreduplication. The expression profile of these genes in our system (down-then up-regulated) and the phenotype observed in our cells at day 5 (endoreduplicated cells) lead us to think that these genes might be a major regulators contributing to the induction of microspore embryogenesis.

Concluding remarks
In barley, doubled haploid technology is widely used to develop new cultivars. Despite the wide use and economic impact of this technology, little is known about the metabolisms and pathways involved. The work presented here provides a descriptive and comprehensive overview of gene expression changes in the early stages of IMC in barley. While microspores were highly distinct phenotypically, we observed that among a gene set of 14k genes expressed across all three stages, a small set of 500 genes were uniquely expressed at a single stage. These included transcription factors associated with early embryogenesis as well as genes involved in the synthesis and response to growth regulators. Among differentially expressed genes, we saw signs of cellular responses to stress (decrease in translation, increase in GSTs, HSPs, secondary metabolism and hormone biogenesis and signal transduction). Thanks to the depth of this transcriptomic analysis, we are confident in stating that this is the most extensive characterization of the barley microspore transcriptome in IMC and provides candidate biomarkers for embryogenesis in barley.