Introduction

Coffee-trees (family Rubiaceae) differ greatly in morphology, size and ecological adaptation, thereby leading to the description of a large number of species. The subgenus Coffea (genus Coffea L.) encompasses more than 80 taxa so far identified, including the two species of economic importance: C. arabica L. and C. canephora Pierre (Charrier and Berthaud, 1985). Higher quality is associated with C. arabica, which accounts for 70% of world coffee production. Coffea arabica is the only tetraploid species (2n = 4x = 44) in the genus while other species are diploid (2n = 2x = 22). Recent investigations established that C. arabica is an amphidiploid formed by natural hybridisation between two closely related diploid species, C. canephora and C. eugenioides (Lashermes et al, 1999). In spite of the low divergence between the two constitutive genomes, C. arabica displays a diploid-like meiotic behaviour (Krug and Mendes, 1940; Lashermes et al, 2000a,Lashermes et al, 2000b). Furthermore, C. arabica is characterised by low genetic diversity and the transfer of desired traits from diploid relative species has been a continuous priority in coffee breeding (Carvalho, 1988; Van der Vossen, 2001).

Plant interspecific hybridisation is a common means of extending the range of variation beyond that displayed by the parental species. However, inherent problems of interspecific introgression such as hybrid instability, infertility, non-Mendelian segregations, and low levels of intergenomic crossing-over can constitute important limitations (Stebbins, 1958). Moreover, features associated with polyploidy or ploidy dissimilarity between species may result in supplementary constraints for interspecific gene flow (Rieseberg et al, 2000).

Occurrence of spontaneous hybrids between C. arabica and diploid relative species such as C. canephora and C. liberica is common, especially when these species grow in direct proximity (Cramer, 1957). In artificial conditions, the success in producing viable interspecific hybrids appears to depend on the direction of the cross. When C. arabica is used as maternal parent, successful crosses have been obtained with a large number of diploid species. In contrast, attempts to produce hybrids using C. arabica as pollen donor have not yet been successful (Carvalho and Monaco, 1968; Le Pierès, 1995). The resulting interspecific hybrids are usually triploid and rather vigorous. While hexaploid hybrids (ie, obtained by duplication of triploid hybrids) and tetraploid hybrids (ie, resulting from the hybridisation between C. arabica and auto-tetraploidised diploid parents) appear reasonably fertile (Berthaud, 1978), the triploid hybrids are highly sterile. Meiosis in these triploid hybrids is especially disturbed (Krug and Mendes, 1940; Chinnappa, 1968; Kammacher and Capot, 1972). For instance, a high frequency of meiotic irregularities (eg, 7.8 to 14.4 univalents depending on the study) in chromosome associations at metaphase I has been observed. Similarly, anaphase distributions appear very irregular. Nevertheless, viable gametes occur occasionally and these hybrids have been successfully backcrossed to C. arabica (Orozco, 1976). The agronomic evaluation of advanced lines (ie, F4) derived from such triploid hybrids reveal an important potential of production often combined with an enhanced resistance to leaf rust caused by the fungus Hemileia vastatrix (Alvarado and Cortina, 1997). The resistance to this serious disease of coffee is assumed to result from the introgression of one or several resistance genes that have been identified in C. canephora (Bettencourt and Rodrigues, 1988).

Despite a considerable interest for coffee breeding, genetic study in relation to interspecific triploid hybrids has been very limited. In particular, development of efficient strategies for selection of triploid-derived progenies would require a better knowledge of genome interactions and factors affecting genetic exchange in triploid hybrids. The purpose of the study presented here was to gain insights into the type of viable gametes that are produced in triploid C. arabica × C. canephora hybrids, and to evaluate the gene flow occurring from C. canephora to C. arabica in this context. Six progenies resulting from the backcross to C. arabica of several interspecific triploid hybrids C. arabica × C. canephora were produced. The nuclear DNA content of both triploid F1 and BC1 individuals was assessed by flow cytometry which is a simple and efficient tool for estimating ploidy level in coffee trees (Barre et al, 1996). Molecular markers combined with evaluations of morphological characteristics and resistance to leaf rust were applied to verify the occurrence of gene transfer from C. canephora to C. arabica, and to estimate the amount of introgression present in BC1 individuals.

Materials and methods

Plant material

Triploid F1 hybrids were generated by crossing tetraploid C. arabica cv caturra individuals, used as female parents, with three different diploid accessions (ie, EC 40, EC 103 and EC 137) of C. canephora. The accessions of C. canephora were selected to represent contrasted genotypes and different collecting sites (ie, Angola and Uganda) in the natural distribution area of this species in Africa. A total of six triploid hybrids, two from each C. canephora accession, were produced as part of the experiment Meg0635 from the breeding programme of CENICAFE. These plants were named H 880, H 881, H 861, H 866, H 846 and H 855, respectively. Then, the different hybrids were backcrossed as the male parent to the elite line ASIII of C. arabica. Six BC1 progenies were produced (Table 1) and 60 BC1 plants (10 plants/progeny) were retained for further analysis. The different crosses were performed by means of artificial pollination using standard hybridisation techniques as described by Carvalho (1988). Briefly, flowers of selected healthy branches are emasculed 2 or 3 days before anthesis and protected from pollen contamination. Pollen is collected and stored under dry-condition at 5°C in a hermetically sealed box up, and when required directly put on the style with the help of a small brush.

Table 1 Production of interspecific backcross progenies by pollination of Coffea arabica (line ASIII) with pollen originated from different interspecific triploid hybrids (C. arabica × C. canephora)
Full size table

Determination of nuclear DNA content

The total DNA amount in nuclei of parental species, hybrid and BC1 individuals was assessed by flow cytometry (Dolezel et al, 1989). Segments (5 × 5 cm) from fresh leaves were excised, and chopped with a razor blade in a plastic Petri dish containing 1 ml of nuclei extraction buffer. The cell suspension was then filtered through 48 μm-mesh nylon and the nuclei suspension was incubated for 5 min in propidium iodide (Sigma P4170) at a final concentration of 0.33 mg/ml as described by Barre et al (1996). Two independent replicates (ie leaf samples) per individual were prepared, and for each run a minimum of 5000 nuclei per sample were analysed using a FACScan (Becton Dickinson) cytometer equipped with an argon laser (15 mW). Each measurement was repeated at least twice. Coefficients of variation of the 2C fluorescence peaks ranged from 4.3% to 6.2% depending on the individuals. To ascertain the ploidy level of hybrid and BC1 individuals, the 2C peak values were compared with the values determined (four replicates) for the diploid (C. canephora accession EC 37) or tetraploid (C. arabica var. Caturra) parental species. Comparison of nuclear DNA content between the mean values of BC1 progenies was made using the Newman and Keuls test while the BC1 progenies and the recurrent C. arabica parent were compared by the rank-sum test developed by Mann and Whitney (1947).

Molecular marker assay

Nineteen microsatellite loci, previously identified as polymorphic between C. arabica and C. canephora, were analysed using PCR. Some of these microsatellite loci (Table 2) have been mapped in C. canephora (Lashermes et al, 2001). The specific primer pairs, amplification conditions, radioactive labelling and polyacrylamide gel electrophoresis were as reported elsewhere (Combes et al, 2000). For each hybrid, the microsatellite loci revealing C. canephora specific bands (ie alleles) were determined and surveyed in the corresponding BC1 progeny. The amplified fragment length polymorphism (AFLP) procedure was performed as previously reported (Lashermes et al, 2000). Briefly, 500 ng of genomic DNA was digested with the restriction enzymes EcoRI and MseI. Restriction fragments were then ligated with double-strand EcoRI and MseI adapters. A selective preamplification was performed using the appropriate primers (named E and M, respectively) with one selective nucleotide at the 3′ end (ie E+A/M+C). The reaction mixture was diluted 1/30 and 10tμL was used for the final amplification with two primers, each containing three selective nucleotides (Table 3).

Table 2 Microsatellite markers showing Coffea canephora specific bands in the different triploid interspecific hybrids (C. arabica × C. canephora)
Full size table
Table 3 Number of Coffea canephora specific AFLP markers identified in three interspecific triploid hybrids (C. arabica × C. canephora). Primer combinations named as E+3 and M+3, represents the 3′ end selective nucleotides of the primers complementary to the EcoRI and MseI adapters, respectively
Full size table

Field characterisation of BC1 progenies

BC1 individuals were further evaluated under field conditions at the Naranjal field experimental station of CENICAFE (Chinchina, Colombia). Each individual tree was recorded for three different morphological traits (ie plant height, leaf shape, and branching pattern) that are known (Stoffelen, 1998) to discriminate the two parental species, namely C. arabica and C. canephora. Subsequently, three morphological types of reference were defined and scored based on a 1–3 scale where 1 designated the ‘arabica type’, 2 the ‘intermediate type’, and 3 the ‘canephora type’, respectively. Furthermore, the resistance to leaf rust (Hemileia vastatrix) was scored on a 0–9 field scale as proposed by Eskes and Toma-Braghini (1981), where 0 represents a complete resistance and 9 indicates a high rust susceptibility. Surveys were performed twice a year (March and October) during the most favourable periods for rust development. The final tree score was the average of the separate observations.

Results

Crossability of interspecific triploid hybrids

Results on BC1 progeny production from 6 different interspecific triploid C. arabica × C. canephora hybrids are summarised in Table 1. BC1 progenies were produced for all attempted backcrosses, the triploid hybrids being used as male parent. For all crosses, the number of initiated fruits 1 month after pollination was rather high. The fruit set varied from 69.1 to 92.3% depending the cross considered. However, a large proportion of fruits fell during the subsequent months. Thus, the number of fruits was markedly reduced seven months after pollination, the fruit sets varying from 6.0 to 11.9%. Finally, the frequency of recovered seeds per pollinated flower ranged from 5.5 to 13.2%, which was rather low and constant whatever the interspecific triploid hybrids used.

DNA content variation

2C nuclear DNA contents of parental species, hybrid and BC1 individuals were successfully assessed (Figure 1). As expected, all interspecific triploid hybrids exhibited a DNA content intermediate between the two parental species, thus confirming their ploidy level (ie 3×) and the possibility of comparing genome sizes by flow cytometry. The DNA content of BC1 plants varied appreciably around the value of the recurrent tetraploid C. arabica parent. The mean value of BC1 plants (ie 775.3) was not significantly different (P = 0.398) from the estimate for C. arabica (ie 795.2). On the other hand, the mean value for BC1 plants was significantly higher (P > 0.01) than a hypothetical value (ie 706.7) corresponding to the mean between the two parents involved in the backcrosses, namely the interspecific triploid hybrids and the recurrent tetraploid C. arabica.

Figure 1
figure 1

Distribution of interspecific triploid hybrids and BC1 individuals according to their DNA content as determined by flow cytometry. For comparison, mean values estimated for the diploid (Coffea canephora) and recurrent tetraploid (C. arabica) parent species are indicated.

Full size image

Analyses of variations in DNA content within and between the different BC1 progenies are given in Table 4. Within progenies, important differences in DNA content were observed. Nevertheless, with the exception of one offspring (ie BC 4530), all analysed BC1 progenies were not significantly different (P > 0.05) from C. arabica. Slight differences were observed among progenies. For instance, the mean value varied significantly between BC1 progenies. It is notable that the two BC1 progenies (BC 4535 and 4530) exhibiting a lower mean DNA content (P > 0.05) were derived from two hybrids involving the same accession (ie EC 037) of C. canephora as parent.

Table 4 Variation in DNA content among the BC1 individuals, and between the different BC1 progenies and the recurrent Coffea arabica parent
Full size table

Molecular analysis of BC1 progenies

Molecular analysis was performed to detect C. canephora-specific markers in the interspecific triploid hybrids and the BC1 individuals. A total of 19 microsatellite loci that are distributed on at least six of the 11 linkage groups of the C. canephora genetic map were investigated. For two interspecific triploid hybrids (ie H880 and H881), only subsets of 16 and five microsatellite loci, respectively, were used. The number of identified microsatellite loci exhibiting a C. canephora-specific allele varied from five to 15 depending on the interspecific triploid hybrids considered (Table 2). Subsequently, all pertinent microsatellite loci were surveyed in the six corresponding BC1 progenies (Table 5). Coffea canephora-specific markers were detected in only eight out of the 58 BC1 individuals analysed. These introgressed plants (ie showing at least one C. canephora-specific marker) belong to four different BC1 progenies and exhibited between 1 and 10 C. canephora-specific markers, the mean number being 3.8 markers. The frequency of C. canephora-specific markers was determined for each locus as well as at the plant and backcross level. Coffea canephora-specific markers were observed for 16 of the 19 microsatellite loci studied. These loci belong to at least six different linkage groups of the C. canephora map. The loci M 20, M 171 and M 189 for which no C. canephora-specific markers was observed, are those analysed for only a restricted number of backcrosses. Furthermore, frequency of C. canephora-specific markers was 4.1% (range 0 to 13%) depending on the BC1 progeny considered. The mean percentage of C. canephora-specific markers in the introgressed individuals was about 30%.

Table 5 DNA marker analysis of Coffea canephora introgression in different BC1 progenies derived from triploid interspecific hybrids (C. arabica × C. canephora)
Full size table

In complement to microsatellites, AFLP analysis using 15 distinct primer combinations was performed on three BC1 progenies (Table 3). The number of C. canephora-specific bands in the different triploid hybrids varied between 52 and 62. Out of the 30 BC1 plants (10 per progeny) scored, only two displayed C. canephora-specific bands (Table 5). One of these two plants also presented C. canephora-specific microsatellite-allele. The mean frequency of C. canephora AFLP specific bands (ie, 2.3, ranging between 1 to 8) in BC1 individuals was similar to the value estimated for the microsatellite loci.

Field observation of BC1 progenies

Sixty BC1 individuals were evaluated for tree morphology and resistance to leaf rust (Figure 2). In agreement with the observed absence of C. canephora-specific molecular markers, a large majority of BC1 individuals showed an arabica-like phenotype (type 1). No more than six BC1 individuals were classified as intermediate or canephora morphological types (types 2 and 3, respectively). Unsurprisingly, these plants displayed C. canephora-specific markers.

Figure 2
figure 2

Relationship between phenotype, rust resistance and introgression level among 60 BC1 individuals derived from different triploid interspecific hybrids (Coffea arabica × C. canephora). Plant morphological types were designated as 1 ‘arabica’, 2 ‘intermediate’, or 3 ‘canephora’ according to the morphological similarities with parental species. Leaf rust resistance was scored according to the field scale (0–9) proposed by Eskes and Toma-Braghini (1981), where 0 represents complete resistance and 9 indicates high rust susceptibility.

Full size image

Regarding leaf rust, most BC1 individuals appeared either susceptible or moderately resistant. Only three trees exhibited a complete resistance to leaf rust. These trees belonged to the canephora morphological type and appeared introgressed. It is notable that, for two BC1 individuals, the complete resistance to leaf rust is combined with a low frequency of C. canephora-specific markers.

Discussion

In the present study, all crosses performed between interspecific (C. arabica × C. canephora) triploid hybrids used as male parent and C. arabica successfully yielded progeny plants. The mean production of 10 seeds per 100 pollinated flowers clearly reflected the low fertility of triploid hybrids. Fruit setting in coffee tree is affected by many factors including the vegetative state of the mother plant, the pollination conditions and the climatic environment during fruit development (Carvalho, 1988). However, in typical artificial conditions (ie intraspecific), fruit setting varied between 30% and 60% in C. arabica (Le Pierres, 1995; Alvarado, unpublished data).

Whatever the origin of the C. canephora accessions used to produce the interspecific triploid hybrids, an overall homogeneous behaviour was observed in the BC1 progenies. The flow cytometric analysis revealed that the generated BC1 individuals had a DNA content close to that of C. arabica. Although allowing only a crude estimation (Barre et al, 1996), this observation indicates that most BC1 plants were nearly tetraploids with about 44 chromosomes. Thus, among the male gametes produced by the interspecific triploid hybrids, only those presenting a high number (ie diploid) of chromosomes could successfully generate viable hybrid seeds. Pollen with about 22 chromosomes would have a selective advantage over gametes presenting a lower number of chromosomes and level of ploidy.

In angiosperms, the successful formation of an embryo is not only dependent on fertilization but also on normal development of the endosperm. Endosperm is a major food and feed source of embryos, playing a capital role in seed formation or failure in interploidy or interspecific crosses (Brink and Cooper, 1947). Analyses in several genera have led to the development of the endosperm balance number (EBN) hypothesis (Johnston et al, 1980). The balance of qualitative genetic factors in endosperm (EBN) has been proposed as a determinant mechanism in endosperm development, and thus of the embryo after intra- and inter-ploid interspecific crosses (Ehlenfeldt and Ortiz, 1995; Carputo et al, 1999). Although this concept has not yet been fully investigated in coffee, related mechanisms have been invoked to explain the contrasted results obtained in reciprocal crosses between C. arabica and several diploid species at different ploidy levels (Carvalho and Monaco, 1968; Berthaud, 1978). It has been suggested (Le Pierres, 1995) that the normal development of the endosperm, and indirectly of the coffee embryo, would be conditioned by the presence of a minimal dosage (ie ploidy) of one parental genome species whatever the endosperm/embryo balance.

In the present study, the reduced fruit set observed in backcrosses of interspecific triploid hybrids to C. arabica appeared clearly associated with a severe drop of fruits during the development phase. It is probable that only pollen mother cells having nearly 22 chromosomes generated effective ploidy, the others leading to deficient endosperm and fruit dropping. In addition to this strong post-zygotic barrier, fertilisation with aneuploid gametes which are likely to be produced by the interspecific triploid hybrids may also result in embryo abortion and fruit or seed development failure as commonly reported in coffee material derived from interspecific crosses (Carvalho, 1988).

Interspecific hybrids are often characterised by distorted marker-segregations (Lytte, 1991). In the present investigation, the combined use of microsatellite-PCR and AFLP methods were appropriate to mark and ensure a good coverage of the C. canephora genome. However, among the different BC1 progenies, the number of individuals exhibiting C. canephora-specific markers appeared particularly small (ie mean frequency of 13.8%). In the same way, an overall low frequency of C. canephora-specific markers was observed, even when considering only the identified introgressed plants (ie. observed frequency of 5.3%). Assuming that each of the chromosomes in the sets of three in the triploid hybrid have an equal chance to be included in the diploid gametes formed, the expected proportion of C. canephora alleles in a tetraploid backcross would be 0.66. Analysing on average twelve microsatellite loci in a tetraploid backcross, the expected proportion of individuals without any C. canephora alleles would be 1:531441 which is much less than the observed value of 1:1.2. Hence, our data revealed a strong deficiency in C. canephora alleles indicating a severe counter-selection against the introgression of genetic material from C. canephora into C. arabica by way of triploid hybrids (C. arabica × C. canephora). In contrast, the segregation of C. canephora alleles transmitted by tetraploid interspecific (C. arabica × C. canephora 4×) hybrids has been reported to conform with the expected ratio assuming random chromosome segregation and the absence of selection (Herrera et al, 2002). The presence of C. canephora alleles in the functional gametes of tetraploid hybrids was neither significantly beneficial nor detrimental. This suggests that the deficiency in C. canephora alleles is more likely to be due to mechanisms specific to triploid hybrids, associated with the formation of diploid gametes, rather than to the consequence of a counter-selection related to the expression of lethal or sub-lethal C. canephora genes. For instance, detailed studies carried out in triploid hybrids of Lolium species showed that the orientation and respective position of chromosomes within the multivalents could have a significant effect on the incorporation of specific alleles in the balanced gametes (Thomas et al, 1988).

Genetically modified coffee plants including C. canephora are approaching commercialisation (Leroy et al, 2000). Among the concerns related to genetically engineered crops, the risk of gene escape toward wild relatives is one of the most discussed (Rissler and Mellon, 1996). Regarding coffee species, spontaneous hybridisation between coffee species are known to occur, and the results presented here reveal a potential gene flow, between C. canephora and C. arabica, which should be studied further.

The use of triploid (C. arabica × C. canephora) hybrids provides an opportunity to transfer desirable traits for crop improvement. For example, introgressants displaying a high resistance to leaf rust were obtained that constitute a promising starting point for breeding programmes (Alvarado and Cortina, 1997). In comparison with the data reported for tetraploid interspecific (C. arabica × C. canephora 4×) hybrids (Herrera et al, 2002), the level of introgression observed in C. arabica when using triploid hybrids appears noticeably low. This could be an advantage by reducing the possibility of introgression of undesirable traits and facilitating the recovery of the recurrent parent. Hence, the triploid hybrid route could recommended for single trait introgressions. On the other hand, this could be a limitation when attempting the transfer of complex traits or several simply inherited traits.