In spite of an increasing number of successes, transgenic plant production is still difficult and limited to a small number of species and genotypes. This is particularly true of woody species, as these are often recalcitrant to genetic transformation. Chevreau  reported that only three fruit woody species were distinguished by a large number of successfully transformed genotypes: more than 10 species of Citrus , 40 genotypes of apple tree , and 35 of grapevine . For the majority of species, the efficiency of the methods used remains variable and weak. However, efficient methods have been described in a number of woody crops. One category of methods consists in directly transforming zygotic embryos or young seedlings (epicotyls, hypocotyls) to rapidly obtain transformed organogenic calli and derived plantlets; this is the case in Allocasuarina , citrange  and poplar . Another category consisting in regenerating transformants from established embryogenic cultures has been described in several species including grapevine [21, 37], Citrus , rubber tree , Prunus subhirtella [18, 19] and American chestnut . These procedures were shown to be efficient not only for basic research in functional genomics aimed at characterizing T-DNA insertion [21, 39, 40], the long-term stability of transgene expression  or transgene silencing , but also for the production of transgenic varieties. In the last ten years, embryogenic callus directly produced from leaf explants has been the most widely used tissue for Agrobacterium-mediated transformation in coffee . Nevertheless low transformation frequencies were reported and, like for other perennial crops, the availability of embryogenic tissues suitable for genetic transformation remains one of the main bottlenecks for developing genetic transformation strategies. In this work, we demonstrate that established embryogenic cultures can be used to improve the efficiency of coffee tree transformation. Another advantage of such cultures is that, unlike other target tissues used for transformation whose availability is seasonal, embryogenic cultures are available all year round. This means transformation experiments can be planned at any time, which is a prerequisite for setting up a more efficient and faster transformation pipeline.
The reliability of GFP fluorescence to monitor transformation efficiency in coffee was confirmed by a large-scale trial aimed at regenerating fully-transformed plantlets using the hygromycin marker gene. In this trial, high frequencies of hygromycin resistant calli (82.5%) were evaluated that were similar to those of GFP expressing calli (90-93%) obtained in the same transformation conditions. In a previous work on coffee, transformation efficiency ranged from less than 1% for the recalcitrant C. arabica  to 33% for C. canephora . The fluorescent marker GFP has a major advantage compared to other reporter genes in that it enables non-invasive detection of transformed cells without the introduction of co-factors or the destruction of the biological sample . Similarly, GFP visual selection was recently used as the only marker to detect transgenic calli lines of Hevea brasiliensis without antibiotic pressure . As previously reported [7, 45], highly efficient selection of transformed coffee tissues was achieved and escapes were prevented when the antibiotic hygromycin was used in the four-month selection step.
It is noteworthy that in our study most progress in improving transformation efficiency was achieved by optimizing the production conditions of the embryogenic cultures used as target tissues for transformation rather than by optimizing the physical co-cultivation conditions. The positive effect on genetic transformation of reducing the strength of mineral salts in the proliferation medium of coffee embryogenic cultures (MS/2 or MS/4) is consistent with reports in the literature. Although it has been widely demonstrated that ions are involved in bacterial attachment to plants , transformation has been successfully achieved in numerous species by using half-strength or even more diluted salt solution media or solutions that specifically lack certain salts such as CaCl2 in the pre-culture medium  or co-cultivation medium [48, 49]. It has been shown that during long-term culture, plant cells may lose the need for auxin and/or cytokinin to maintain active growth. This process known as 'habituation', which is common in callus cultures in some plant species such as sugarbeet , was described as a shift from auxo- to autotrophic state for growth regulator requirements . Coffee embryogenic cultures do not require auxin or cytokinin to proliferate, although it was possible to stimulate proliferation by optimum exogenous concentrations of both growth regulators. Similarly, genetic transformation of long-term cultures was possible without auxin or cytokinin. Our results specifically revealed a very negative effect of cytokinin and a positive effect of auxin in optimizing transformation ability. Blanc et al.  showed that simultaneously increasing the auxin and cytokinin supply in the pre-culture medium prior to transformation of rubber tree using embryogenic cultures stimulated the development of active and fast growing cells, hence improving transformation efficiency.
Coffee embryogenic callus cultures were successfully established by sub-culturing the recently formed active embryogenic tissues every four weeks. It is widely acknowledged that a lack of regular subcultures leads embryogenic calli to lose their embryogenic potential due to cell ageing. Applying short subcultures (14-21 days) on a maintenance (i.e. proliferation) medium is an essential step in establishing long-term cultures in many perennial crops [52–56]. The process of establishing coffee embryogenic cultures leads to the development and maintenance of three morphologically different callus phenotypes with highly contrasted potential for genetic transformation. Several morphological variants with contrasted transformation potential have also been observed during the proliferation and maintenance of embryogenic calli in cotton . Andrade et al.  highlighted great variability in the way in which embryogenic cultures proliferate and that this variability should be taken into account as it strongly affects further genetic transformation. These authors described embryogenic culture proliferation as a continuum along a developmental gradient from undifferentiated embryogenic callus through slightly more differentiated proembryogenic masses or PEMs  to repetitive embryogenesis at the globular embryo stage or even later . Our histological studies in coffee showed that, depending the callus phenotype or the age of the embryogenic culture, either distinct cell types can co-exist, resulting in a heterogeneous proliferating callus, or only one cell type is present, forming a very homogeneous tissue.
One priority of any team involved in developing a transformation procedure is to identify the suitable target cell type. At the histological level, the yellow coffee callus with high transformation ability (transformation efficiency >90%) proved to consist of PEMs similar to those observed in the Daucus carota model [58, 60, 61] that were described as proliferating compact cell masses able to produce somatic embryos. In our study, improved transformation efficiency was also systematically associated with increased quantities of PEMs when different auxin/cytokinin balances (data not shown) and embryogenic culture ages were tested. Taken together, these results indicate that PEMs are probably the competent target tissue for coffee genetic transformation in C. arabica. PEMs have already been identified as suitable target cells for the transformation of several woody species for which transformation methods using embryogenic cultures have been established like Vitis vinifera , V. rotundifolia , avocado  and American chestnut . The particular structure of PEMs could favorably influence the Agrobacterium-mediated transformation process. For instance in Arabidopsis thaliana, Sangwan et al.  showed that, irrespective of their origin, the competent cells were small, isodiametric with thin primary cell walls, small vacuoles, prominent nuclei and dense cytoplasm. Most of these characteristics correspond to those of coffee PEMs. The small size of PEMs associated with their looser organization increases bacterial accessibility. In addition, the high regenerative potential of this cell type is an indispensable quality for regenerating transgenic plants.
Although embryogenic callus directly regenerated on leaf explants is the most widely used tissue for coffee transformation, low transformation efficiencies (< 1%) were obtained . However, the histological heterogeneity of coffee embryogenic calli has already been reported . Our work confirmed the strong heterogeneity of this tissue but also revealed its low transformation-competent PEM contents. Consequently, we recommend avoiding direct use of embryogenic callus for transformation experiments and instead completing a number of preliminary proliferation cycles to increase the quantity of PEMs.
Many authors have studied the influence of ageing on transgenic lines. Indeed, as embryogenic cultures of some species can be maintained for several years without loss of embryogenic capacity, they provide a model to study the long-term expression of transgenes during in vitro culture and in regenerated plants [19, 66]. Surprisingly, little attention has been paid to the effect of the age of embryogenic cultures used as target tissues on subsequent transformation efficiency. In our study, transformation efficiency increased markedly with the age of the coffee embryogenic culture to reach maximum at seven months of proliferation, and remained at over 90% for several months. To our knowledge, this is the first time that a strong positive effect of ageing of embryogenic cultures on their transformation competence has been demonstrated. As the regeneration capacity of coffee embryogenic cultures remained stable over two years (data not shown), this means that highly efficient and reliable regeneration of transgenic plants is possible. It has been reported that some Vitis vinifera genotypes produced transgenic embryo lines irrespective of 4, 8, or 12 month embryogenic culture age, whereas others produced embryo lines only from the youngest 4-month cultures . However, this response was attributed to differences in genotype to maintain their regeneration potential under long-term conditions and not directly to a loss of transformation competency. The same was recently observed in Citrus by Dutt and Grosser , who compared the transformation competence of six-year-old and one-year-old embryogenic cultures. In that study, although some EGFP expression in the callus phase of the old cell line was observed, it was not possible to regenerate transgenic embryos and plants from such cells. Among the parameters we studied, the age of the culture was the most important factor in improving the efficiency of coffee transformation. The optimization of this parameter should thus be taken into account in other species for which transformation procedures using embryogenic cultures as target tissues have already been established.
Agrobacterium-mediated transformation often allows transgenic plants containing single copy insertions to be obtained . Low transgene copy numbers have recently been reported in transformants derived from embryogenic cultures in several trees [21, 23, 38]. Similarly, our results in coffee showed that all the selected plants had T-DNA integrated in their genome and that almost half contained single transgene insertion sites. These results are consistent with previous observations in coffee plants transformed via A. tumefaciens with from one to five inserted T-DNA copies, of which 69% harbored one T-DNA copy . It has been shown that transgene expression is influenced by the number of T-DNA copies and/or the integration site . Even if inactivation of transgene expression can occur in plants with a single copy , this phenomenon is less frequent than with multiple transgene copies [69, 70]. For this reason, the regeneration of a high proportion of coffee plants with a low copy number of the inserted T-DNA is important for the application of this transformation technology for both genomic purposes and breeding programs.