Coffee is one of the most important commodities cultivated worldwide and has a great economic impact in many countries, especially in South America (FAOSTAT, 2014). Although more than 130 different species belonging to the Coffea genus have been described, only two of them are mainly commercially exploited: Coffea arabica and C. canephora. There exist other species such as C. iberica, C. dewevrei, and C. racemosa that are thus far only cultivated to satisfy local markets (Kumar et al., 2006). The most cultivated variety for C. canephora is robusta. C. arabica has many different important cultivars and is responsible for 60% of the world production. It is considered to have superior beverage qualities compared to robusta. However, the production costs for C. arabica are much higher due to crop management practices, higher susceptibility toward diseases and the need for more stringent environmental conditions (less adapted to temperature changes, requirement of more rain or irrigation and prefers higher altitudes due to more mild temperatures) (van der Vossen et al., 2015). While C. canephora is a diploid species (2n = 2x = 22) and presents a higher diversity, C. arabica is tetraploid (2n = 4x = 44) and shows a very narrow genetic diversity attributed to its evolution and reproductive biology (self-pollination) (Lashermes et al., 1999, 2000). Moreover, coffee plants take 2 years to complete their life cycle. All these characteristics make classic genetic improvement a big challenge, taking at least 20 years to have a new genotype in the market (Etienne, 2005; Santana-Buzzy et al., 2007; Tonietto et al., 2012).

One powerful biotechnological tool used in crop improvement is somatic embryogenesis (SE). By SE we understand the production of an embryo from somatic tissues without fecundation (de Feria et al., 2003). Embryogenic cells show two important characteristics; they are able to multiply or to proliferate, which makes SE suitable for mass production of elite cultivars (de Feria et al., 2003), and the fact that plants can be regenerated from one single cell. The latter characteristic is essential for genetic engineering and somatic hybridization. Moreover, SE can also be used to conserve interesting genotypes and/or the ones that are threatened with extinction (Yang and Zhang, 2010). Because SE formation is based on cellular totipotency it has also been used as a model to investigate morphological, physiological, molecular and biochemical events that occur during the onset and development of embryogenesis in higher plants (Quiroz-Figueroa et al., 2006). In coffee, the most direct application of SE is the rapid multiplication of elite genotypes, specially hybrid heterozygous ones (Etienne et al., 2016).

The SE technology has been studied in coffee since 1970 (Staritsky, 1970). The methodologies to induce SE and the further development of embryos are genotype-dependent, which leads to an almost empirical development of specific protocols for each species (Santana-Buzzy et al., 2007). In the recent years, information about the molecular mechanisms of SE induction has been gained. In this review, we provide an overview about the general concepts of somatic embryogenesis and important molecular markers found in model plants, how this knowledge has been applied on coffee SE and how future knowledge could be applied to improve protocols. Additionally we introduce a new hypothesis about the acquisition of embryogenic capacity. We want to instigate researchers to see the classical concept from a different point of view and reflect about the type of explant that is used for SE induction.

In general, somatic embryogenesis for all species is initiated by exposing plant tissues to the right stimulus, most often to plant hormones (Yang and Zhang, 2010). A right balance between the applied hormones and internal factors can induce the reprogramming of a differentiated somatic cell, but it could also promote the proliferation of totipotent undifferentiated cells that are dormant present in some tissues, being the plant stem cells. The “classical” theory about the formation of SE is that differentiated somatic cells can regain their embryogenesis capacity and be reprogrammed to differentiate into new viable embryos (Marsoni et al., 2008; Yang and Zhang, 2010). Irrespective of the initiation of the totipotency theory, somatic embryos can be obtained in two different ways, directly or indirectly. Following the direct way, the embryos are formed without intermediate callus formation (proliferation of cells) directly on the explant (leaves, roots or other part of plant). In the indirect way, first an embryogenic callus is formed and then embryos arise from this callus (Yang and Zhang, 2010). In the indirect way, two distinct phases are involved, called induction and expression. The induction stage is marked by changes in the metabolism and gene expression, leading to the differentiation into embryos in the expression phase (Fehér et al., 2003; Jiménez, 2005).

The use of growth regulators is practically essential to obtain somatic embryos in both processes, direct and indirect way. The optimal concentration, time of treatment and type of growth regulator varies according the species or even according the cultivars within the same species. Although essential to induction of SE process, auxins are negatively effecting embryo development. To solve this problem, most of protocols suggest dropping the concentration or even omitting auxins after the multiplication of totipotent cells to allow differentiation and consequent protoderm development. Associated periclinal cell divisions result in tissue invagination and establishment of the embryonic axis, being the beginning of embryo development (Toonen et al., 1994; van Boxtel and Berthouly, 1996; Simões-costa et al., 2009; Silva et al., 2015).

Embryogenic calli can be distinguished from non-embryogenic calli based on their morphological characteristics (Yang and Zhang, 2010; Padua et al., 2014; Silva et al., 2014). In general, embryogenic calli are yellow and friable and their cells are small, isodiametric, arranged in clusters, with a dense cytoplasm, a nucleus with salient nucleoli and rich in small amyloplasts. Non-embryogenic calli are spongy and translucent showing cells that are more elongated, with the vacuole occupying a big volume of the cytoplasm, higher number of vesicles and absence of cytoplasmic organelles (Figure 1) (Gatica-Arias et al., 2008; Padua et al., 2014; Silva et al., 2014). Zygotic and somatic embryo development processes for all plant species are quite similar. That is the reason why somatic embryogenesis is often used to study zygotic embryogenesis. Generally, morphogenetic and metabolic phases are distinguished. In the morphogenetic stage the structure of embryos are established and the second stage is marked by a pronounced biochemical activity. In dicots, the morphogenetic stage is divided in four phases; globular, heart, torpedo and cotyledonary stage (Etienne, 2005; Yang and Zhang, 2010; Etienne et al., 2013; Padua et al., 2014). Despite that both zygotic and somatic embryogenesis are similar, there are some differences in morphological and histochemical aspects during embryo development. The shape of somatic embryos are normally more irregular than zygotic embryos due to a different cell elongation and different storage components between both process (Bertrand et al., 2012; Etienne et al., 2013).

The “classical” concept for initiation of somatic embryogenesis affirm that molecular reprograming of somatic cells is essential. In this process, already differentiated cells regain the totipotency capacity (Fehér et al., 2003; Yang and Zhang, 2010) and under the right stimulus they can differentiate into new embryos. Although this is the most accepted theory about totipotency in SE, we would like to discuss an alternative hypothesis. We hypothesize that the cells that are able to differentiate into embryos don’t pass through the process of dedifferentiation. They are pools of meristematic cells that keep the totipotency throughout plant development and under the right stimulus they multiply and differentiate to form a new viable embryo. These cells would act then, as a plant stem cell. The fact that only a limited amount of cells are responsive to SE induction through the right stimulation is a point to strengthen our hypothesis (Quiroz-Figueroa et al., 2002; Yang and Zhang, 2010). Not matter the stimulation provided, some cells will never become an embryo. The responsive cells in our point of view are the ones that keep the embryogenic capacity and could be considered as plant stem cells. Explants currently used for direct SE induction are microspores, ovules, immature zygotic embryos, seedlings and young leaves (Yang and Zhang, 2010), all young tissues. Such young tissues relatively contain more stem cells. With aging, the ratio totipotent stem cells/differentiated cell decreases. Auxins are considered essential to induce the process of SE in the view of the general theory about totipotency capacity and it would be the induction of the redifferentiation and proliferation of the embryogenic cells. In our theory it is the breakage factor of ‘dormancy’ and the proliferation, being also essential to complete the SE process. This theory arose from our personal discussion during years of embryogenesis study. It is an invitation to see the classical concept of embryogenic acquisition from a different point of view and stimulate new researches in this topic.

The first trails on coffee SE were executed using orthotropic shoots in different coffee species, with obtaining of embryos only in the robusta explants (Staritsky, 1970). Since then, many protocols using different kinds of explants were developed for different coffee species, including C. Arabica (Loyola-vargas et al., 2016). In coffee, the SE direct way is often described as a low frequency method and the indirect way as high frequency method (Etienne, 2005). In the low frequency process, only one medium is used and the embryos are obtained faster (approximately within 70 days) but in a smaller number, maximum 10 per explant. In the high frequency process, multiple media are used: (i) the callus induction medium, to provide the stimulus necessary for the cells to start dividing, (ii) the embryogenic callus production medium, were the callus will grow exponentially and (iii) the callus maintenance medium to further multiply the cells in exponential phase, select and keep the callus in the undifferentiated state and the embryo maturation and regeneration medium (van Boxtel and Berthouly, 1996) (Table 1 and Figure 2). The firsts media to induce and produce embryogenic callus normally use a higher concentration of auxin, followed by a decrease or even complete removal of auxins, for embryo development (van Boxtel and Berthouly, 1996; Padua et al., 2014). In this methodology, the embryo production takes more time to develop (9–10 months), but hundreds of them can be obtained per gram of callus. It is possible to use a liquid medium for multiplication as well as for maturation. This is the preferred method for mass propagation since bioreactors can be used improving the embryo formation rates (Etienne, 2005). The use of semi-solid medium is less efficient, probably due to lower and less homogeneous diffusion of growth regulators on this type of medium which can lead to a lower concentration of auxins with increasing distance to the culture medium (van Boxtel and Berthouly, 1996; Etienne, 2005; Simões-costa et al., 2009).

In the past, studies about SE were mainly empirical and focused on establishing and optimizing somatic embryo production in different species, much more than to understand the mechanisms behind this event. Even though different studies investigated the cellular and molecular changes during SE in many different plant species, the molecular basis of the factors involved in initiation of SE process are not completely understood (Yang and Zhang, 2010; Mukul-lópez et al., 2012; Guzmán-García et al., 2013). The Brazilian Coffee Genome Project sequenced 214.964 ESTs (Expressed Sequence Tags) belonging to 33.000 different genes from 37 cDNA libraries of leaves, roots, flowers, fruits, embryogenic calli and zygotic embryos of C. arabica, C. canephora, and C. racemosa. Those tissues were collected under different physiological states as different stages of fruit development and flowering, young and mature leaves and under influence of biotic and abiotic stress (Esteves Vieira et al., 2006). The large number of data generated by this consortium is extremely valuable to help to clarify molecular events like SE (Silva et al., 2013). Recently the genome of C. canephora was published, opening new possibilities to explore and understand the coffee diversity (Denoeud et al., 2014). We also published the proteome profile of embryogenic calli, a first step toward insight into totipotency (Campos et al., 2016). The main challenge now is to integrate all the generated knowledge about coffee physiology to elucidate processes such as somatic embryogenesis. It is already known that the expression of some genes are linked to different stages of SE: BABY BOOM1 (BBM), LEAFY COTYLEDON 1 and 2 (LEC1/LEC2), WUSCHEL-RELATED HOMEOBOX (WUS) and SOMATIC EMBRYOGENESIS RECEPTOR KINASE (SERK) (Hecht et al., 2001; Arroyo-Herrera et al., 2008; Yang and Zhang, 2010; Yang and Karlson, 2011; Ma et al., 2012; Nic-Can et al., 2013). Since these four genes are relatively well known in other species, especially in model organisms, their role and specifically presence or similarity in coffee are further discussed. Silva et al. (2013) identified 15 EST-contigs related to the embryogenesis process in coffee. Among those 15 ESTs, 1 sequence was annotated to abiotic stress. 9 EST-contigs could not be identified in the EST database generated by the Brazilian Coffee Genome Project, but were detected only in their own embryogenic material. The other 5 ESTs could be readily associated with coffee embryogenesis showing similarities with proteins as polygalacturonase, cysteine-proteinase, expansine, allergenine, and WUS.

Studies in plant biology trough proteomics have increased considerably in the recent years. The main cause of this growing interest is that proteomics provides an overview of the metabolism, complementary to the genomics results (Palma et al., 2011). It is known that the correlation between mRNA and protein at the same moment of extraction is often low (Jansen et al., 2002; Carpentier et al., 2008). High throughput proteomics for non-model plants has been used to solve this problem and generate more applicable results (Carpentier and America, 2014). Compared to transcriptomics, the proteomics approach has a great potential to study non-model plants since DNA sequences are less conserved as amino acid sequences being a promise alternative to understand plant metabolism (Zivy et al., 2015). This is specifically valid for C. arabica of which the genome is not sequenced yet but from which RNA-seq data are available. The sequence of C. canephora genome is also a really helpful tool that can be used to extrapolate for C. arabica and provide good results for plant understanding and coffee breeding. Proteomics studies on SE have already been executed in some species such as V. vinifera (Marsoni et al., 2008; Croce et al., 2009), cassava (Baba et al., 2008; Li et al., 2010), cacao (Noah et al., 2013), avocado (Guzmán-García et al., 2013), maize (Varhaníková et al., 2014), sugarcane (Heringer et al., 2015), and coffee (Tonietto et al., 2012; Campos et al., 2016). We identified 1052 non-redundant proteins, being 5 annotated to embryogenic capacity (Campos et al., 2016). Many proteins are without any annotation and so are still uncharacterized, showing the poor molecular knowledge about SE in coffee. Our publication of the proteome of somatic calli in coffee can be considered as the initial step to understand the embryogenic process in coffee and the annotation of proteins can guide further investigations in this topic. Most of these studies focus on differences between embryogenic and non-embryogenic calli. Some proteins specifically appear in embryogenic and not in non-embryogenic cell lines and proposed some molecular markers for different stages of SE as enolase and globulin S11 for torpedo stage of embryo development (Lippert et al., 2005; Tonietto et al., 2012). Tonietto et al. (2012) identified 14 proteins in coffee linked to the different phases of embryo development during SE. The most abundant were proteins related to energy production. The need for more energy is compatible with the major cell division that follows these stages of development (Baba et al., 2008). Correlation to specific proteins could be found for each embryo development stage (globular, torpedo, and cotyledonar). Proteins related to stress response were also identified, such as HEAT SHOCK PROTEIN (HSP 70) and cytoplasmic aldolase being more abundant in the cotyledonary stage. Stress is known to be an important factor to induce somatic embryogenesis and the presence of these proteins corroborate to this fact (Fehér et al., 2003; Tonietto et al., 2012; Fehér, 2015).