Osmotic stress-induced somatic embryo maturation of coffee Coffea arabica L., shoot and root apical meristems development and robustness

Somatic embryogenesis (SE) is the most important plant biotechnology process for plant regeneration, propagation, genetic transformation and genome editing of coffee, Coffea arabica L. Somatic embryo (SEs) conversion to plantlets is the principal bottleneck for basic and applied use of this process. In this study we focus on the maturation of SEs of C. arabica var. Typica. SEs conversion to plantlet up to 95.9% was achieved under osmotic stress, using 9 g/L gelrite, as compared with only 39.34% in non-osmotic stress. Mature SEs induced in osmotic stress developed shoot and root apical meristems, while untreated SEs were unable to do it. C. arabica regenerated plants from osmotic stress were robust, with higher leaf and root area and internode length. To understand a possible regulatory mechanism, gene expression of key genes of C. arabica, homologous to sequences in the Arabidopsis thaliana genome, were analyzed. A set of two component system and cytokinin signaling-related coding genes (AHK1, AHK3, AHP4 and ARR1) which interact with WUSCHEL and WOX5 homedomains and morphogenic genes, BABY-BOOM, LEC1, FUS3 and AGL15, underwent significant changes during maturation of SEs of C. arabica var. Typica. This protocol is currently being applied in genetic transformation with high rate of success.


Results
SE induction of C. arabica var. Typica. Direct SE was induced from leaf explants of C. arabica var. Typica with a high potential of propagation, plantlet conversion and genetic transformation. Three months were required to induce the development of SEs to be ready for stable genetic transformation 8 . Briefly, this protocol was followed: leaf explants of coffee cultured in callus induction medium (CIM medium, Van Boxtel et al. 14 ) provided proembryogenic masses (PEM) obtained after one month with 93% efficiency of cultivated explants. SEs development was achieved when PEM were subcultured in SE-P medium after two months in culture, with an efficiency ranging from 82 to 100% (Fig. 1A, B).
SEs maturation is enhanced under osmotic stress. SEs maturation was more efficient in osmoticstress (SE-M9) (see Material and Methods) in contrast with non-osmotic (SE-M3) (see Material and Methods) solid medium and under light conditions according to the plantlet conversion obtained (Table 1). In the first stage of SEs maturation, secondary SE was induced from the base of initial SEs under osmotic-stress in 83.7% compared to 16.3% of non-stress medium (Fig. 1C). Secondary SEs at early torpedo stage, symmetric and having a suspensor-like structure were separated and individually subcultured to the corresponding fresh medium. 19,000 secondary SEs from osmotic stress (SE-M9 medium) and 6192 SEs derived from non-osmotic medium (SE-M3 medium) were used. Secondary SEs produced in osmotic stress (SE-M9 medium) were characterized by a prominent suspensor-like structure at the early cotyledonary stage, well developed cotyledons, parallels and symmetric, while in non-osmotic medium (SE-M3 medium) secondary SEs produced less prominent suspensor-like structures and altered cotyledon shape (fused-circular) (Fig. 1D, E). At the end of the second stage of maturation, SEs produced in SE-M3 and SE-M9 differed in their phenotype. SEs size at cotyledonary stage was 3.85 ± 0.43 mm in osmotic stress (SE-M9) and 6.0 ± 0.33 mm in non-osmotic stress (SE-M3) (Fig. 1D, E; Table 1). SEs cotyledon size derived from osmotic stress medium (SE-M9) ranged from medium (60%) to small (40%), according to Etienne et al. 1 , contrasting with non-osmotic (SE-M3) that induced large and misshapen (55%) and medium (45%) in size (Table 1). SEs became green in color after three weeks in SE-M9 medium and four weeks in SE-M3 medium. When mature SEs were subcultured individually in SE-G medium, germination was clearly observed after 2 weeks of culture in both treatments. SEs conversion to plantlet was 95.9% achieved in osmoticstress medium (SE-M9) (18,050 plantlets derived from 19,000 secondary mature SEs) compared to 39.3% in non-osmotic (SE-M3) (2477 plantlets derived from 6192 secondary mature SEs) ( Table 1). Differences in root and shoot length were evident after 15, 30 and 45 days respectively of conversion in SE-G medium (Fig. 1F, H).

Robustness of plants induced by osmotic stress.
Plantlets derived from SEs under osmotic and nonosmotic treatments were subcultured to petri dishes containing SE-G medium for further growth and development. After 5 months in SE-G medium, plantlets were transferred to soil conditions. At this time, the architec- (F) Plantlet conversion of SEs developed from SE-M3 non-osmotic medium cultured in SE-G medium after 30, 45 and 60 days from left to right, respectively. Bar represents 5 mm. (G) Regenerated plants after 5 months in SE-G, derived from SE-M3 non-osmotic medium. Small root 3.7 cm length and 8 leaves can be appreciated in these plants. Bar (Fig. 1I). These plants were successfully established in soil-containing pots and incubated in a growth chamber. Survival rate was 100% in plants derived from osmotic stress medium while 90% was obtained from non-osmotic stress. After eight months in soil, as expected from previous morphology, more robust plants developed from SE-M9 osmotic medium. An increase in rooting and leaf area as well as internodes length in plants derived from osmotic stress was observed compared to the non-osmotic treatment. Plants derived from osmotic-stress developed a root area of 105.82 ± 1.06 a cm 2 compared to 11.56 ± 0.08 b cm 2 from non-osmotic stress ( Fig. 2A,B). Root length was significantly different: 47.3 ± 0.9 a cm were recorded in osmotic stress derived plants while in non-osmotic stress only 23.2 ± 1.24 b cm were induced ( Table 2). The number of lateral roots from SEs derived from SE-M9 osmotic medium were 18.3 ± 0.13 a as compared with 4.7 ± 0.4 b in SE-M3 non-osmotic medium. Similarly, lateral root length showed 16.27 ± 9.13 a cm in SE-M9 osmotic medium compared with 12.1 ± 4.7 a cm from SE-M3 nonosmotic medium ( Table 2).
In regard to leaves, plants derived from osmotic stress developed an average of 22.0 ± 4.0 a leaves, a leaf area of 36.31 ± 3.2 a cm 2 , a length of 13.3 ± 1.33 a and a width of 4.3 ± 0.5 a . In contrast, plants derived from  Effect of osmotic stress on SAM and RAM in mature SEs. SEs at cotyledonary stage grown under osmotic stress and cytokinins developed shoot and root apical meristems in 100% of analyzed samples, while untreated SE did not formed it Fig. 3). Histological analysis of mature SEs induced by osmotic stress revealed that morphology of root apical meristems (RAM) are tiered or closed eudicot, typical of the Rubiaceae family to which C. arabica belongs to 19 . Three-tiered differential cells: stella, cortex and columella, were observed in the RAM. Columella cells with 12X18 μm, stella cells 27X5.8 μm, cortex cells 12X18 μm, and quiescent center (QC) 11.76 μm (Fig. 3B, D). On the other hand, embryos derived from non-osmotic medium, apical meristems were absent ( Fig. 3A, C).  Table S2).

Differentially expressed key factors involved in
According to the gene expression analysis of somatic embryo maturation of C. arabica, a molecular model is presented, which indicates how the upregulation of type A response regulators (ARR7-ARR15) and histidinekinase 3 (AHK3) under non-osmotic treatment inhibits SAM and RAM development and consequently somatic Table 2. Morphological differences between plants of C. arabica var. Typica regenerated from mature SEs produced in SE-M9 (osmotic medium) and SE-M3 (non-osmotic) medium after eight months. Differences in variables among treatments were tested using factorial analysis of variance (ANOVA). Values are means ± standard deviation. In each row, means followed by different letters are significantly different (P < 0.05). SE-M3 non-osmotic medium (control) and SE-M9 osmotic medium. Software ImageJ. www.nature.com/scientificreports/ embryo maturation. In this treatment, genes involved in drought tolerance and seed maturation AHK1, AHP4 and ARR1 were repressed, blocking the activation of genes necessary for stem cell maintenance (WUS-WOX5) and master regulators of embryogenesis (LEC1-FUS3-BBM-AGL15) (Fig. 5). The opposite effect under osmotic stress treatment (SE-M9), where AHK1, AHP4 and ARR1 were upregulated allowing activation of key SE genes such as BBM, LEC1, MP, and FUS3 leading to acquire embryogenic competence, and robustness in regenerated plants (Fig. 5). Further analysis of each component is required to clearly elucidate this developmental phenomenon.

Discussion
SE induced from leaf explants has been the most widely used target tissue in coffee micropropagation and stable genetic transformation 1 . Published protocols have shown that SE induction and proliferation is time consuming, ranging from 9 to 20 months until they can be used for stable genetic transformation 8,20,21 . In most of these reports, SE regeneration to plants in C. arabica has been shown to be more difficult and time consuming than in C. canephora P. As noted by several authors, auxins negatively affect SE development in coffee [14][15][16] . Our strategy was to simulate natural processes occurring during embryo development in coffee seeds. We hypothesized that osmotic stress and the presence of cytokinins, are both necessary for a correct SEs development. This is supported by cytokinin signaling, as it has shown that plays a critical role during root and stem cell niche formation allowing RAM system initiation in SE 17,18 . Our qRT-PCR analyses revealed differential expression of key factors allowing the enhanced plant regeneration in SE-M9 treatment in comparison to the SE-M3 (Fig. 4, Table S1). According to the gene network devised in STRING with a high confidence threshold (0.700) based on C. arabica homologous genes present in Arabidopsis (Fig. S2) and transcriptional gene expression analysis by qPCR, SE maturation is related to 3 main processes: stem cell maintenance, cytokinin signaling, and embryogenesis (Fig. 5). Under these conditions, the coding genes involved in the two-component signaling system and cytokinin signaling, such as histidine-kinase1 (AHK1), it was upregulated in SE-M9 treatment. AHK1 has functions as osmosensor, improvement for drought tolerance in Arabidopsis [22][23][24] and plays a unique role in the regulation of desiccation processes during seed maturation 25 . AHK1 feeds phosphates to the phosphorelay-integrated histidine phosphotransfer protein     (Figs. 4, 5). WOX5 is a transcription factor which may be involved in the differentiation and maintenance of the stem cells in the RAM 17,37,38 .
The auxin responsive factor, ARF5, was upregulated in SE-M9 (osmotic stress) treatment and downregulated in normal conditions (Fig. 4). ARF5 mediates embryo axis formation and vascular tissue differentiation [39][40][41] . Arabidopsis mutants in arf1, arf5 and arf7 displayed a significant reduction in capacity and efficiency of SEs maturation 42 . Otherwise, overexpression of ARF5 results in null SEs development 42 . García-Gómez et al. 43 , proposed a dynamic regulatory network model to understand the interaction between auxin and cytokinin signaling pathways and some transcriptional regulators important in the root apical meristem (RAM) of A. thaliana. Among genes that were analyzed, ARF5 was showed to be essential for WOX5 activity.
The master regulator of SE, BBM, is a transcription factor of the AP2/ERF family, which was upregulated in SE-M9 (Figs. 4, 5). BBM promotes cell proliferation, differentiation and morphogenesis, specifically during embryogenesis. BBM-induced embryogenesis relies on transcriptional activation of LEC1, LEC2, FUS3 and AGL15 genes 44 . Members the AP2 gene family (cytokinin response factors CRFs), closely related to BBM, are transcriptionally upregulated by cytokinin signaling 45 . CRFs mutant combinations have demonstrated to be lethal to embryos 46 . LEC1, a nuclear transcription factor Y subunit B-9, was upregulated in SE-M9 (Figs. 4, 5). LEC1 is a transcriptional activator required for both embryo maturation and cellular differentiation. FUS3, a regulator of gene expression during late embryogenesis, was upregulated in SE-M9 (Figs. 4, 5, and Table S2).

Figure 5.
A model of SEs maturation regulatory network in two different media. Differences in shoot and root apical meristems observed in the SEs are related to expression changes of key genes involved in cytokinin signaling, stem cell maintenance and embryogenesis. Initial samples were SEs from PEM at globular and early torpedo steges. The second sample were secondary SEs at cotyedonary stage in the second month of maturation. SEs were cleary developed with specific shape in osmotic stress medium contrasting with non-osmotis stress medium.  Table S2). AGL15 is preferentially expressed during embryogenesis and gibberellic acid catabolism.
The above-mentioned differential expression of key genes is similar to that detected during the successive stages of seed maturation in C. arabica. Dussert et al. 48 , reported that one coding gene for AHK3 was upregulated in subsequent stages of seed maturation of C. arabica. In this transcriptional analysis it was found other upregulated genes related with cytokinin signaling: one gene for AHK4, three for AHP1, four type-B response regulators and five type-A response regulators. In the case of transcription factors, BBM2 and ARR2, a redundant gene of ARR1, were differentially upregulated in the last four stages of maturation, a similar behavior was observed with FUS3 which was differentially upregulated in two stages of maturation (4 and 5 out of 7) 48 . This genetic network represents a preliminary evidence in the understanding of SEs maturation process of C. arabica var. Typica. It is conceivable that epigenetic reprogramming could have also occurred during the maturation developmental process of SEs. Changes or modifications in DNA methylation patterns are linked with regulatory mechanisms of master regulators genes of SE such as LEC1, LEC2, FUS3 and AGL15 49 . Further analysis of each component is required to clearly elucidate this developmental phenomenon.
In this work we were able to reach up to 95.9% of germination and complete plant regeneration with SEs under osmotic stress treatment in contrast with 39.34% in untreated SEs. The effect of different water potentials in SE conversion in C. arabica was reported. Etienne et al. 3 analyzed the developmental process of zygotic and SEs of coffee C. arabica grown in temporary immersion bioreactors, a system with changing water potential depending on immersion cycles. They found that a higher conversion rate of germination of zygotic embryos (96%) versus SEs (55%). Conversion of SE was increase up to 91% using a bigger bioreactor with differences in water characteristics (water potential). It is important to notice that shoot and root apical meristems were developed compared to untreated SE. Failure in forming proper SAM and RAM in SEs was published elsewhere 1,3,7,9,13,50 . The importance of the apical meristems in plant growth and development has long been recognized 51 , but in woody plants, little is known about meristems formation during SE 11 . Most of the studies on the molecular network regulating embryo development have been made in Arabidopsis 52 .
In C. arabica, a mutation called laurina occurred in highland Ethiopia, alters shoot apical meristem (SAM) dimensions, resulting in smaller leaf primordia and incipient internodes 53 . In our work, under osmotic stress and cytokinin presence we showed the activation of SAM and RAM dimensions and SEs conversion to plantlet (Fig. 3B,D).
If whether or not our results are related with laurina mutation, analogies with the observations reported in Arabidopsis in early timing within plastochron reducing SAM dimensions for some genes such as AXR1, ARGOS, ANT, related to auxin regulation pathway may explain the negative role in SEs maturation of auxins [54][55][56] .
Regenerated plants derived from osmotic stress showed an increased fitness including vigorous roots, increased leaf area, and stem length, leading to an increased anchorage, nutrition and productivity. In economic terms, a reduction of costs (up to 80%) in plant production derived from somatic embryos subjected with osmotic stress with improved agronomic performance is expected.
Additionally, development of SAM and RAM in mature SEs derived from osmotic treatment can be considered a morphological marker for quality control in SEs development in C. arabica var. Typica.

Conclusions
Desiccation using high content of gelrite in the presence of cytokinins in SEs plays an important role for maturation and conversion to plantlet of C. arabica var. Typica.
Reprogramming of genes involved in shoot and root apical meristems occurred during maturation of SEs under osmotic stress, enhancing robustness of converted plants.
Development of SAM and RAM in mature SEs derived from osmotic treatment can be considered a morphological marker for quality control in SEs development in C. arabica var. Typica.
Bioinformatic tools such as STRING achieved improvement in the understanding of molecular mechanisms involved in SEs maturation.
The application of strategies derived from this knowledge can be applied in tissue culture media composition and morphological and physiological changes in the expected phenotype.

Methods
Induction of SE in C. arabica var. Typica. SE in C. arabica var. Typica (INIFAP-Tapachula, Chiapas), was induced from leaf explants, derived from eight-months old trees. Explants were disinfected by gasification in a vacuum chamber with a mixture of 50 mL of sodium hypochlorite from a commercial bleach (1.2% active chlorine) and 50 mL HCl 6 N for 15 min and washed four times with sterile distilled water 57 . Disinfected leaves were cut into 1 cm 2 pieces and cultured on the callus induction CIM medium as described by Van Boxtel et al. 14 : half-strength Murashige and Skoog medium 58 , 30 g/L sucrose, 100 mg/L casein hydrolysate, 400 mg/L malt extract, 10 mg/L thiamine, 1 mg/L nicotinic acid, 1 mg/L pyridoxine, 1 mg/L glycine, 100 mg/L myo-inositol, 0.5 mg/L 2.4-D, 1 mg/L indole-3-butyric acid (IBA), 2 mg/L 2-isopentenyladenine (2-iP), and solidified with 2.4 g/L gelrite, pH was adjusted to 5.8 before autoclaving. After two months in dark, PEM were sub-cultured into SE-P medium, as described Van Boxtel et al. 14  The protocol of SEs maturation included two stages: In the first one, PEM (with soft consistency, 0.7 to 10 mm 2 size), derived from CIM medium, were used as source of SEs. 100 PEM (each one containing about 200 SEs at globular and torpedo stage) yielding a total of 20,000 SEs. PEM were individually transferred to respective cultured medium (SE-M3 and SE-M9), using forceps and slightly smash to allow SEs to get contact with the medium. 10 PEM were cultured in each plastic petri dish (90X20 mm, Phoenix biomedical product). PEM were incubated at 25 ± 2 °C, under a 12/12 h photoperiod at 50 µmol/m 2 -s irradiance provided by fluorescent lamps T8 Phillips P32T8/TL850 combined with natural sunlight, producing a unique light spectrum ( Figure S1). In the second stage, we selected and subcultured only SEs derived from secondary SE developed from SEs from the first stage. SEs at early torpedo stage, symmetric and having a suspensor-like structure were separated, and individually subcultured to the corresponding fresh medium (SE-M3 or SE-M9), 25 SEs in each plate, for one month until cotyledonary stage developed according to Valencia-Lozano et al. 8 .
Conversion of SEs to plantlets of C. arabica var. Typica. Secondary SEs subjected to osmotic stress (SE-M9) (19,000 total) and non-osmotic stress (SE-M3) (6192 total) in cotyledonary stage, green color, symmetric and having a suspensor-like structure were subcultured to plantlet conversion in SE-G medium: MS medium 58 30 g/L sucrose, 1 g/L activated charcoal, 3 g/L gelrite and smoke water 0.5% (karrikins) 59 . Incubation was under the same conditions mentioned in SEs maturation. Regenerated plantlets were subcultured in flasks with SE-G medium until plants reached 10 leaves and 27 cm root length. Regenerated plants were potted in vermiculite and soil (1:1 v/v), incubated in a plant growth chamber at 50 µmol/m 2 -s irradiance provided by T8 Phillips P32T8/TL850 fluorescent lamps with a photoperiod of 16/8 h and temperature of 19 to 23ºC, as described by Valencia-Lozano et al. 8 . Efficiency of SEs maturation was calculated according to the number of SEs converted to plantlets derived from SEs at cotyledonary stage produced at the end of the second stage of maturation.

Isolation of RNA and gene expression analysis.
Total RNA from SEs cultured in osmotic-medium (SE-M9) and non-osmotic medium (SE-M3) was isolated using Trizol (Invitrogen, Carlsbad, CA, USA). RNA concentration was measured by its absorbance at 260 nm, ratio 260 nm/280 nm was assessed, and its integrity confirmed by electrophoresis in agarose 2% (w/v) gels. Samples of cDNA were amplified by PCR using SYBR Green qPCR (BioRad) in Real-Time PCR Systems (CFX96 BioRad). The reference genes in this work were ACT, 24S and RPL39, according to Freitas et al. (2017) applied for qPCR analysis of embryogenic calli, somatic embryos and callus of coffee. Retfinder, NormFinder, Bestkeeper and Delta-Ct were used in this analysis. Three replicates were done with these reference genes, calculated the relative expression, and weighted ct, and next a delta ct in each gene analyzed and relative amount of target gene expression using the 2 −ΔΔCT method 60 . qPCR analysis was based on at least three biological replicates for each sample with three technical replicates and control treatment SE-P. To select genes involved in SE development process, a gene network with a high confidence (0.700) was performed with STRING (v11.0, http:// string-db. org), based on C. arabica homologous genes present in A. thaliana genome. The selected genes were: Two component system and cytokinin signaling genes, AHK1, AHK3, AHP4, ARR1, ARR7, ARR15, auxin signaling; ARF5 (monopteros), homedomains; WUSCHEL, WOX5; master regulator of embryogenesis, LEC1, BBM, FUS3, and AGL15 (Table S1, Figure S2). Gene identifier was made according to UNIPROT (http:// www. unipr ot. org), NCBI (http:// www. ncbi. nlm. nih. gov) database. Sequences of all genes were analyzed from A. thaliana using blastN, blastP in the coffee genome homologous sequences. Homologous sequences in C. arabica genome greater than 40% in protein sequence with A. thaliana were considered. Proteins of C. arabica homologous with A. thaliana aligned by BlastP were identified as: (XP_027075540, XP_027097259, XP_027125278, XP_027115945, XP_027098055.1, XP_027065231.1, XP_027100463.1, XP_027109768.1, XP_027090708.1, XP_027085797, XP_027062561, XP_027102113.1, XP_027113896.1) (Table S1). Oligonucleotides were designed to qPCR (2 −ΔΔCT method analysis) gene expression or transcriptional analysis (Table S1).