Heterosis patterns and sources of self-compatibility, cross-compatibility and key nut traits within single and double hybrid crosses of kola [Cola nitida (Vent) Schott and Endl.]

Sexual incompatibility among kola genotypes accounted for over 50% yield loss. Compatible and high yielding varieties are in demand to develop commercial orchards. The objective of this study was to assess self-compatibility and cross-compatibility of kola (C. nitida) genotypes within self, single and double hybrid crosses and to determine heterosis pattern in the resulting hybrids for sexual compatibility and key nut yield and quality traits. Crosses among kola genotypes from three field gene banks (JX1, GX1, MX2) and one advanced germplasm (Bunso progeny) in Ghana were evaluated along their parents for sexual compatibility, nut yield and nut quality. Data were collected on pod set, pseudo-pod set, pod weight, number of nuts per pod, nut weight, brix, potential alcohol and nut firmness. Significant (P < 0.001) differential pod set was observed within Bunso progeny, JX1, GX1 and MX2 crosses; while pseudo-pod set differed only within JX1 and MX2 crosses (P < 0.001). Very large prevalence of mid-parent, heterobeltiosis, and economic heterosis was observed for sexual compatibility, outturn and brix for the single and double hybrid crosses. Heterosis was prominent among the double hybrid crosses as compared to the single hybrid crosses suggesting that recurrent selection of compatible varieties from advanced generations could result in genetic gain in kola improvement. The top five crosses with best heterosis for sexual compatibility and an appreciable positive heterosis for outturn and brix were B1/11 × B1/71 × B1/157 × B1/149, B1/11 × B1/71 × B1/296 × B1/177, GX1/46 × GX1/33 × B1/212 × B1/236, JX1/90 × JX1/51 and JX1/51 × JX1/36. These materials could serve as sources of beneficial alleles for improving Ghanaian kola hybrids and populations for yield and sexual compatibility.


Results
Variation in sexual compatibility among the BUNSO progeny crosses. There were significant differences (P < 0.001; df = 83) among the double hybrid crosses for pod set (Supplementary Table S1). Variability among the crosses for pseudo pod set was however not significant (P > 0.05; df = 83). Pod set ranged from 21.3% in DCS14 to 97.0% in DCC19 (Supplementary Table S1). Pseudo-pod set ranged from 0.0 in several crosses to 25.4% in DCC5 (Supplementary Table S1). Pod set was more than two-fold higher in the double hybrid crosses (DCC) as compared to the double hybrid selfs (DCS). By contrast, pseudo-pod set was significantly (P < 0.0001) higher in the DCS as compared with the DCC (Fig. 1). The higher pseudo pod sets among self-crosses than crosses between different genotypes suggests a linkage between pseudo-pod set and self-incompatibility in kola.
Variation in sexual compatibility among the AFOSU JX1 crosses. Variation among the single crosses of some JX1 accessions for the percentage pod set and percentage pseudo-pod set was significant (P < 0.001; df = 141) (Supplementary Table S2 Table S2). They are therefore good materials for involvement in kola breeding programmes to increase genetic gain for compatibility and yield in improved varieties.
Pod set of the single hybrid crosses (SCC) of JX1 was more than two-fold higher (P < 0.001) than pod set of single hybrid selfs of JX1 germplasm collection (Fig. 2). Pseudo pod set was observed to be significantly (P < 0.0001) higher for single hybrid selfs of JX1 as compared to single hybrid crosses (Fig. 2).
Grouping of crosses into compatible classes. Among the double hybrid crosses, none of the crosses had compatible classes 0, 1 and 2. Double hybrid cross had a compatibility score ranging from 3 to 5. In contrast, double hybrid self-crosses exhibited compatibility scores ranging from 2 to 3. The single hybrid crosses were distributed among the compatibility scores of 0 to 5 with compatibility class 4 mostly expressed (χ 2 = , df = 5, p < 0.001). On the other hand, single self-crosses had compatibility scores ranging from 0-4 and compatibility score 2 was predominant (χ 2 = , df = 5, P < 0.001) ( Supplementary Fig. S1). The cross-hybrids were more compatible than the self-hybrids for both the double hybrid crosses and single hybrid crosses.
Variation in yield components and nut quality traits of the BUNSO progeny crosses. There were significant differences (P < 0.001; df = 83) among the Bunso progeny crosses in number of pods, pod weight, number of nuts/pod, weight of unpeeled nuts, weight of peeled nuts and percentage outturn (Supplementary Table S1). Number of pods ranged from 14.4 in DCS9 and DCS16 to 83.6 in DCC19 (Supplementary Table S1). Number of pods for some crosses (DCC6, DCC19, DCC24, DCC 26) was about two-fold higher than number of pods for crosses (DCC11, DCC13, DCC17, DCC38) and more than three-fold higher than those of crosses DCC5, DCS12, DCS14 (Supplementary Table S1). The significant variation in pod weight ranged from 62.8 for DCS9 and DCS16 to 364.2 for DCC19 (Supplementary Table S1). The top five crosses with higher pod weight included DCC19, DCC26, DCC24, DCC6 and DCS3 (Supplementary Table S1). Number of nuts per pod of crosses (DCC19 and DCC26) was more than three-fold higher than that of other crosses (DCC5, DCS17, DCS23 and DCS25 (Supplementary Table S1). Weight of unpeeled nuts ranged from 41.0 for DCS9 and DCS16 to 238.1 for DCC19 (Supplementary Table S1). DCC19, DCC24, DCC26 were observed to express higher values for weight of peeled nuts. The highest outturn (%) of 92.5% was observed for the DCC19 while the lowest outturn (%) value of 28.7% was exhibited by DCS 14 (Supplementary Table S1). There was no significant variability www.nature.com/scientificreports/ ((P > 0.05) among the Bunso progeny crosses for nut quality traits including brix, potential alcohol and firmness of nuts (Supplementary Table S1).
Variation in yield components and nut quality traits of JX1 crosses. All  Variation in yield and nut quality traits among MX2 self-crosses. Differences among MX2 crosses for number of pods and pod weight were significant (P < 0.001; df = 26) ( Table 2). Number of pods for crosses A1 × A1 was about five-fold higher than that of ATTA 3 × ATTA 3 and two-fold higher than those of crosses A10 × A10, A22 × A22, JB 10 × JB 10, JB 3 × JB 3, JB 40 × JB 40, P2-1B × P2-1B and W25 × W25 (Table 2). Pod weight ranged from 59.3 for ATTA 3 × ATTA 3 to 297.0 for A1 × A1 (Table 2). ATTA 3 × ATTA 3, A10 × A10, A8 × A8, and JB 9 × JB 9 were the topmost crosses with the highest number of nuts/pod of 7.0 ( Table 2). All the crosses recorded above 50% outturn. However, outturn was significantly (P < 0.001) higher for A10 × A10, Club × Club, JB 15 × JB 15, JB 26 × JB 26, JB 3 × JB 3, JB 36 × JB 36, JB 9 × JB 9 and JB 34 × JB 34 (Table 2). Differences among the MX2 crosses for nut quality traits; brix, potential alcohol and firmness of nuts were also statistically significant (P < 0.001). For instance, brix of crosses ATTA Cluster and structuration of BUNSO progeny crosses based on pod set, yield and nut quality traits. Cluster analysis of double hybrid crosses of the Bunso progeny using pod set, pseudo-pod set and pod and nut traits grouped the crosses into 3 clusters (Fig. 3). Cluster 1 was composed of 30 individual crosses and cluster 2 had 35 individual crosses while Cluster 3 was made up of 18 crosses. Cluster 1 is characterized by crosses with significantly (P < 0.001) lower category means than overall mean of crosses for number of nuts/ pod, nut width, outturn, weight of unpeeled nuts, weight of peeled nuts, number of pods, pod weight, nut length and pod set (Supplementary Table S3). The category mean for weight of peeled nuts was two-fold lower than the overall mean of crosses for this trait (Supplementary Table S3). The category mean of cluster 2 for nut length, pod set, nut width, outturn and pod width were significantly (P < 0.001) higher than the overall mean of crosses for these traits (Supplementary Table S3). The standard deviation in category for cluster 2 ranged from 0.83 for nut length to 8.49 for pod set (Supplementary Table S3). Crosses grouped under cluster 3 expressed significantly higher category means for all traits than the overall mean of crosses (Supplementary Table S3 Cluster and structuration of JX1 crosses based on pod set, yield and nut quality traits. The pod set, pseudo pod set and pod and nut traits of the JX1 crosses grouped the single cross and single self-crosses into 3 clusters (Fig. 4). Crosses in cluster 1 are characterized by traits that had category means significantly (P < 0.001) lower than the overall mean of the crosses (Supplementary Table S4). The category standard deviation for cluster 1 ranged from 1.04 for nut width to 11.69 for outturn while the standard deviation for all the crosses ranged from 2.10 for nut width (NW) to 80.32 for pod weight (PW) (Supplementary Table S4). Structuration of JX1 crosses did not separate the single hybrid crosses and single hybrid self-crosses of JX1. The three groups of crosses were spatially distributed in all the four quadrants of the biplot. Single hybrid crosses of JX1 (SCC-JX1) were more associated with dimension 1 which contributed 68.4% of total variation. Also, the SCC-JX1 were more distributed on the positive quadrant of the biplot as compared to SCS-JX1 ( Supplementary  Fig. S3). The category means for crosses in cluster 2 for traits pseudo pod set, outturn, pod length, firmness of nuts was significantly (P < 0.001) higher than the overall mean of crosses for these traits. However, for traits such as nut length (NL), nut width (NW), number of nuts/pod (NN), weight of unpeeled nuts (WUN), weight of  www.nature.com/scientificreports/ peeled nuts (WPN), pod weight (PW), number of pods (NP) and pod set (PS%), the category mean of crosses in cluster 2 was significantly (P < 0.001) lower than the overall mean of crosses (Supplementary Table S4). The category means of crosses in cluster 3 were significantly (P < 0.001) higher than the overall mean of crosses for the traits except for pseudo-pod set where the category mean was significantly (P < 0.001) lower than the overall mean of crosses (Supplementary Table S4).
Cluster of GX1 single self-crosses based on pod set, yield components and nut quality traits. Pod set, pod and nut yield components and nut quality traits grouped the self-crosses of GX1 genotypes into three clusters (Fig. 5). Clusters 1, 2 and 3 encompassed of 7, 52 and 10 individual crosses, respectively. Cluster 1 was characterized by self-crosses that had significantly (P < 0.001) lower category mean for pod weight, number of pods, weight of unpeeled nuts, number of nuts per pod, pod set, weight of peeled nuts (Supplementary Table S5).
The self-crosses of GX1 genotypes grouped under cluster 3 exhibited category means that were significantly (P < 0.001) higher than the overall mean for weight of unpeeled nuts, number of nuts/pod, number of pod, pod weight, pod set and weight of peeled nut weight (Supplementary Table S5).
Cluster of MX2 crosses based on pod set, yield components and nut quality traits. The selfcrosses of MX2 were grouped into 5 clusters based on pod set, yield components and nut quality traits (Supplementary Fig. S4). Clusters 1, 2, 3, 4 and 5 had sizes of 2, 14, 18, 24, and 30 individual crosses, respectively. Cluster 1 was defined by self-crosses with significantly (P < 0.001) higher category mean than overall mean for nut length and nut width (Supplementary Table S6). Self-crosses in cluster 2 were characterized by significantly (P < 0.001) lower category mean for unpeeled nut weight, peeled nut weight, outturn and number of nuts/pod as compared to the overall mean of crosses for these traits (Supplementary Table S6).
Self-crosses in cluster 3 exhibited significantly (P < 0.001) higher category mean than overall mean of crosses for brix and pod width (Supplementary Table S6). Cluster 4 is defined by crosses with significantly (P < 0.001) higher category mean for number of pod, pod set and pod weight than the overall mean of crosses for these traits. However, the category mean of potential alcohol for crosses in cluster 4 was significantly (P < 0.001) lower than the overall mean indicating the crosses performed below average for this trait. The self-crosses of MX2 under cluster 5 exhibited significantly (P < 0.001) higher category mean than overall mean for number of nuts/pod, weight of unpeeled nuts, weight of peeled nuts, outturn and potential alcohol. The category mean for brix was however significantly (P < 0.001) lower than the overall mean (Supplementary Table S6).
Correlation among pod set, yield and nut quality traits of BUNSO progeny crosses. Highly significant (P < 0.001) and positive correlation was observed between number of pods and pod set, pod weight and pod set, number of nuts per pod and number of pods, number of nuts per pod and pod weight, outturn and pod set, outturn and number of pods, outturn and pod weight, outturn and weight of unpeeled nuts, outturn and weight of peeled nuts ( Supplementary Fig. S5).
Correlation among pod set, yield and nut quality traits of JX1 crosses. Correlation was significant (P < 0.001) and positive for the relationship between number of pods and pod set, pod weight and pod set, number of nuts per pod and pod set, number of nuts per pod and number of pods, number of nuts per pod and pod weight, nut length and pod width, nut width and nut length, weight of unpeeled nuts and pod set, weight of unpeeled nuts and number of nuts per pod, weight of peeled nuts and pod set, weight of peeled nuts and number   Heterosis for pod set, outturn and brix among the JX1 single cross hybrids. There was prevalence of mid-parent, better parent and economic heterosis for pod set, outturn and brix for single cross hybrids of JX1 (Supplementary Table S8 Table S8).
Prevalence of economic heterosis for pod set, outturn and brix was shown by the JX1 single hybrid crosses (Supplementary Table S8 In relation to standard variety 2, economic heterosis was observed for pod set, outturn and brix (Supplementary Table S8 Table S8).

Discussion
The results showed significant variation among Bunso progeny, JX1, GX1 and MX2 crosses for sexual compatibility suggesting there is a genetic variability within the gene banks to warrant selection and recovery of good performing lines for sexual compatibility. It is important to understand the amount of variation within a population in order to make a more informed selection decision 57,58 . This allows for planning crosses between compatible genotypes 59,60 . Nyadanu et al. 34  www.nature.com/scientificreports/ information accessible for breeding decisions [64][65][66] . Optimal cross-selection of these genotypes would increase genetic gain in breeding compatible varieties of kola 67,68 . The genetic variability in pseudo-pod set was significant for the JX1 and MX2 crosses. Bunso progeny and GX1 crosses expressed non-significant variation for pseudo pod set. These differences among the crosses of the various germplasm collections could be due to differences in the genetic materials. Pseudo-compatibility is the fertilization by pollen with which they would normally be incompatible; incomplete incompatibility in which gametes which would normally be incompatible form a viable embryo or fruit set 40 . Pseudo-pod set is inversely related to pod set and results in low yield in kola. There is a need to consider pseudo pod set as one of the target traits to select against in breeding for sexual compatibility in kola. Crosses such as B1/11 × B1/71 × GX1/4 6 × GX1/16, Club × JB32 × JX1/5 × JX1/9, JX1/9 × JX1/11 × GX1/46 × GX1/53, JX1/20 × JX1/9, JX1/51 × JX1/20, JX1/5 × JX1/23 and JX1/73 × JX1/23, A1 × A1, JB1 × JB1 which were very highly compatible and expressed no pseudo-pod set would be useful genetic materials to select against pseudo-pod set. Variation in the level of pseudo-compatibility among genotypes in Brassica oleracea has been shown to depend on the genetic background in which the S genes operate 69 . Johnson 70 reported pseudo-compatibility as one of the factors affecting the degree of self-incompatibility in inbred lines of Brussels sprouts.
The cross pollinations in general resulted to more pod set than the self-pollinations. Other authors reported that cross pollinations can enhance fertilization as shown by their higher fruit set reports 71,72 . Nevertheless, some genotypes were cross-incompatible and cannot fertilize each other. In some cases, this was expressed as low pod set. For example, among the single hybrids of JX1, 2% of the crosses made up of JX1/73 × JX1/36 and JX1/87 × JX1/118 were incompatible and 15 of the crosses (15% of total cross) exhibited very low compatibility. Selection against these crosses could help make progress in breeding for compatible varieties in kola. Griggs et al. 73 and Cuevas and Polito 72 reported similar findings in their work on olive.
The fairly good pod set of the double hybrid crosses as compared to the single hybrid crosses in this study could be due to an expression of genetic gains for pod set. The Bunso kola progenies were earlier selected for sexual compatibility or pod set in a hybrid breeding program 74 . However, the double hybrid self-pollinated crosses showed lower degree of compatibility reactions as observed in nearly all self -pollinations. This suggests that the phenomenon of self-incompatibility is expressed even in advanced generations of kola. This could have consequences of reduced yields in advanced generations of kola if recommended varieties are not inter-spaced with pollinizers and are solely established. The issue of self-incompatibility in advanced varieties of kola could be managed through inclusion of pollinizers in orchards to increase yield. In crops exhibiting SI, cultivars that serve as pollen donors ("pollenizers") are usually interspersed throughout orchards since fruit set depends largely on cross pollinations. Pollenizers are commonly used in canola (Brassica napus L.), sunflower, strawberry (Fragaria x anannasa (Weston), European pear (Pyrus communis), sweet cherry, Japanese plum (Prunus salicina Lindl) [75][76][77] . The use of pollenizers is also recommended in olive (Olea europaea L.) 63 . Two genotypes A1 and B1 were identified to be great pollinizers of Cola nitida in Ghana 34 .
The falling or distribution of double and single crosses into higher compatibility scores than the double and single self-hybrid crosses further confirmed existence of self-incompatibility in kola. Self-incompatibility is a genetically controlled mechanism that prevents self-fertilization in about half of angiosperm species 37,78 . Sporophytic self-incompatibility has been reported in Asteraceae, Betulaceae, Convolvulaceae, sterculiaceae and malvaceae 40 . Kola is of the family malvaceae.
Developing fruit crops with high qualities has been a goal in several breeding programs [82][83][84][85] . In this study, nuts of kola crosses were assessed for some important quality traits. The results confirmed that there were significant variations among the crosses for total soluble solids or brix for JX1 and MX2 kola gene banks. This suggests genetic diversity among the genotypes for brix and provides an opportunity to select crosses that are high in brix content.  86 . Considering the issue of astringency in kola, crosses that are high in brix could have an improved sensory attributes and good shelf-life during storage. The brix is very important because the higher the brix, the sweeter nut flavour and crop genotypes high in brix are preferred by consumers 87 . Astringency has been reported by Osei-Bonsu et al. 88 , Takrama et al. 35 and Lowor et al. 89 as a poor sensory attribute of kola consumption. Also, kola crosses with high brix content could store better during post-harvest storage. It is generally known that the higher the brix content of fruits, the better the shelf life during storage [90][91][92] .
Significant differences were observed among the crosses of JX1 and MX2 gene banks for potential alcohol. However, the variability in this trait for Bunso progeny and GX1 crosses was not statistically different. The difference in the significance of variation of this trait among different populations of kola could be explained by the differences in the genetic composition of genotypes and the environments where the accessions are established. Kola is used to develop various products including soft drinks and wine 17 . Varieties with higher potential alcohol content are therefore desirable. Potential alcohol does not only influence flavour and sensory perceptions ) and MX2 self-crosses (A10 × A10, A8 × A8, Atta1 × Atta 1, JB 9 × JB 9, and JB34 × JB 34) expressed high potential alcohol content. Though not significant, the following crosses of Bunso progeny ( DCS7, DCC1, DCS13, DCC33, DCC43, DCC51) and GX1 crosses (GX1/3 × GX1/3, GX1/30 × GX1/30, GX1/37 × GX1/37, GX1/60 × GX1/60, GX1/74 × GX1/74 and GX1/71 × GX1/71) had high potential alcohol content compared to the other crosses. These crosses would be important for developing industrial products like wine. Fruit or nut firmness is an important component of texture and influences sensory perception of consumers. Consumers regard texture as a positive quality attribute donating freshness and storability of products and contributing to the enjoyment of eating 94,95 . Texture properties are very important in foods for harvesting, processing, packaging, storage and presentation to the consumer/customer. For example, hardness/firmness, one of the texture properties, is one of the most substantial parameters generally used to determine freshness of fruits and vegetables 94 . All the four populations of kola used in this study did not exhibit significant differences among nuts of crosses for nut firmness except JX1 crosses. The top six crosses of the JX1 with higher nut firmness were JX1/30 × JX1/7, JX1/45 × JX1/45, JX1/62 × JX1/27, JX1/62 × JX1/7, JX1/63 × JX1/113 and JX1/J1 × JX1/33. These materials are great resources to improve nut firmness attribute of kola hybrids.
Cluster analysis was carried out to group crosses having similar performance in relation to pod set, pseudopod set, yield components and nut quality traits. Quantitative analysis regrouped the crosses into 3 clusters for Bunso progeny, JX1 and GX1 gene banks and 5 clusters for the MX2 gene bank which facilitates the selection of diverse crosses or parents for the kola breeding programme. Based on the results, the diversity panel was categorized into three clusters with cluster 2 and 3 containing the best performing crosses for Bunso progeny, JX1 and GX1. Crosses in cluster 2 and 3 for Bunso progeny, JX1 and GX1 and cluster 4 and 5 of MX2 were more distributed at the positive side of the quadrant and were more associated with dimension 1. The dimension 1 contributed 74.1%, 68.4%, 97.1% and 27.4% of the total variability for Bunso progeny, JX1, GX1 and MX2 crosses respectively. Given the information on the contribution of the traits to variation on the PC1 and PC2 axes, the biplots identified the genetic materials with high compatibility, yield and nut quality. The higher category means than overall mean for traits of the crosses that constitute cluster 2 and 3 and 4 and 5 of MX2 indicates that they performed above average of all the crosses. Such crosses would be rewarding if selected and used in kola breeding programmes to develop compatible and high yielding varieties.
The significant strong correlations of pod set, yield traits such as number of pods, pod weight, number of nuts/pod, weight of unpeeled nuts, weight of peeled nuts and outturn for Bunso progeny, JX1 and GX1 crosses suggested that these economic traits could be improved simultaneously. These results further suggest possible deployment of indirect selection for nut yield using pod set, pod weight, nut width and nut length as surrogate traits in kola breeding programmes. This agrees the findings of Nyadanu et al. 34 and Adebola et al. 81 who found significant and positive correlation between pod set and number of pods, number of nuts /pod, and nut weight. The negative and significant correlations between pod set and pseudo pod set indicates that an increase in pseudo-pod set would result in a decrease in pod set. This necessitates the need to select against pseudo-pod set in kola breeding programmes. Correlation between the nut quality traits and the yield traits was weak and not significant for Bunso progeny crosses. Similarly, correlation between the nut quality traits and the yield traits was weak and not significant for the MX2 crosses except correlation between brix and nut length (r = 0.46, P < 0.05), brix and nut width (r = 0.38, P < 0.05), firmness of nuts and nut length (r = 0.47, P < 0.05). However, in the case of JX1 and GX1 crosses, significantly strong correlations were observed between the nut quality traits and the agronomic/yield traits. For instance, the correlation between brix and pod length and between brix and outturn was 0.81 and 0.84 respectively for JX1 accessions. Similarly, the correlation between brix and outturn and brix and nut width was 0.6 and 0.56 respectively for GX1 accessions. The association of yield traits and nut quality traits for these genetic resources of kola allows the selection of promising genotypes that combines high yield with nut quality traits 96,97 . Due to the high demand and the search for new hybrids that meet the requirements of the consumer market, breeding strategies consist of exploring important agronomic traits and improvements in organoleptic properties to favour both higher quality and production. The contrasting results for the kola germplasm sets could be explained by genetic differences of their genotypes and the environments in which they were grown. Brix is affected by a lot of factors including genetics and environmental factors [98][99][100] . Selection and involvement of kola genotypes identified in this study to have appreciable higher contents of brix in breeding programmes could help develop improved varieties with enhanced nut quality.
The high prevalence of positive estimates of MPH, BPH and ECH for pod set and outturn among the double hybrid crosses of Bunso progeny suggest that the crosses performed better than their parents and standard varieties for this trait and further indicate absence of bidirectional dominance deviation. The prevalence positive heterosis observed among the single hybrid crosses of JX1 for pod set, outturn and brix indicates that effective progress can be made in the development of compatible and high yield kola varieties with quality nuts. The expression of high and positive heterosis among the double hybrids of Bunso progeny for pod set and outturn and pod set, outturn and brix in the case of JX1 crosses could be dependent on the degree of fit and genetic diversity 58,101 among the parental lines used which is worth testing to inform future decision making in the kola breeding pipeline. This could further be explained by the additive effects of several desired dominant alleles, or as 'overdominance' the combined effect of two different alleles at the same gene locus, or a combination of both [102][103][104][105] . Hence, heterosis helps a breeder to make more stringent selections. The parents of the double hybrid crosses were selected earlier for their high fruit set or compatibility potentials 74  www.nature.com/scientificreports/ could maximize progressive heterosis responses resulting in even higher compatibility in third generation of hybrid crosses. Lamkey and Edwards 106 and Alam et al. 107 suggested that positive heterosis is desired in the selection for yield and its components, whereas negative heterosis is desired for early cycling and short plant height. In our case however, a positive heterobeltiosis for compatibility, outturn and brix was desirable since it indicates that the crosses were more compatible, had higher nut outturn and brix than their parents.

Conclusion
Significant and large variations for sexual compatibility, nut yield and nut quality attributes were observed for crosses of Bunso progeny, JX1, GX1 and MX2 field genebanks of kola in Ghana. The cross pollinations in general resulted to more than two-fold pod set than the self-pollinations confirming the need for pollinizers to increase fruit set in kola. Self-compatible and cross compatible partners within single hybrid and double hybrid crosses were identified. Strong and significant correlation was observed between sexual compatibility and number of pods, pod weight, number of nuts/pod and outturn for the crosses of all the four field gene banks of kola. Significant and positive correlation between yield and nut quality traits observed for JX1 and GX1 crosses provides opportunity to simultaneously develop kola varieties that combine high yield with good nut quality. Very large prevalence of mid-parent heterosis, heterobeltiosis and economic heterosis was observed for the double hybrid crosses of Bunso progeny and single hybrid crosses. Mid-parent heterosis, heterobeltiosis and economic heterosis observed for the double hybrid crosses was above that of single hybrid crosses. This indicates that recurrent selection of compatible kola varieties from advanced generations of kola could be rewarding. The top five crosses with best heterosis for sexual compatibility and an appreciable positive heterosis for outturn and brix were B1

Materials and methods
Germplasm and experimental sites. The source of plant materials used for the study is the Cocoa Research Institute of Ghana (CRIG). The experiment was conducted on four different kola field gene banks held at CRIG namely, MX2, JX1, GX1 and Bunso progeny (advanced germplasm). All the materials were collected under the authority of CRIG in Ghana and all procedures were carried out in accordance with relevant guidelines for handling of plant genetic resources.
Assessment of self-compatibility and cross-compatibility of genotypes within the various gene banks was carried out from January 2019 to December 2020. MX2 kola field gene bank is located at CRIG, Tafo. CRIG is located at an altitude of 222 m above sea level with a minimum and maximum temperature of 21 to 37 °C, respectively. The weather conditions during 2019 and 2020 at CRIG are shown in Supplementary Fig. S9. JX1 and GX1 kola field gene banks are situated at Afosu. Afosu is a marginal area in terms of rainfall. Amount of rainfall received in the area is between 150 and 200 cm reaching its maximum during the two peak periods of May-June and September-October yearly with 25.2 °C and 27.9 °C as minimum and maximum temperatures, respectively. The Bekwai-Oda association is the predominant soil formation in the area. The weather conditions in 2019 and 2020 at Afosu is presented in Supplementary Fig. S10. Bunso progeny trial is located at Bunso. Bunso is located at altitude 145.00 m/475.72ft above sea level with minimum and maximum temperature of 23.2 °C and 34.9 °C, respectively. Supplementary Fig. S11 shows the weather conditions at Bunso during the experimental period in 2019 and 2020.
The physical and chemical properties of the soils at MX2 (Tafo); JX1 and GX1 (Afosu), and Bunso progeny trial (Bunso) are presented in Supplementary Table S9. Information on the year of establishment and the number of accessions of the four field gene banks of kola are presented in Supplementary Table S10.

Assessment of sexual compatibility within the germplasm collections. From May 2019 to
December 2020, crossing of some genotypes of MX2, JX1, GX1 and Bunso progeny trial was carried out. Partners used in the crossing depends on the availability of male and female flowers for the genotypes involved as flowering in the species is erratic across years. Twenty pollinations/crossings were targeted per each cross. The crossings/pollination was replicated thrice using three pollinators. Before pollination, female flowers about to open were covered or bagged with nylon nets and the male flowers were emasculated before pollen maturation. Freshly opened female flowers were pollinated with pollen grains from newly opened male flowers. Pollen grains from newly opened male flowers were collected with 15 cm thin sticks with sharp ends and smeared to each of the stigmatic lobes of freshly opened female flowers. Pollinated female flowers were re-bagged immediately with nylon nets for further 48 h. For each of the crosses carried out, pod sets, pseudo-pod sets, number of dropped flowers after pollination were recorded two weeks after pollination. Based on the records of pod sets, the crosses were categorized into incompatible, very low compatible, low compatible, moderate compatible, high compatible and very high compatible using a 0-5 scale-0 (no pod set), 1 (1-20% pod set), 2 (21-40% pod set), 3 ( www.nature.com/scientificreports/ pod set), 4 (61-80% pod set) and 5 (81-100% pod set). The proportion of fruit set to number of flowers pollinated was expressed in percentage and noted as percentage pod set as shown in Eq. (1). The proportion of number of pseudo fruit set to the number of flowers pollinated was expressed in percentage and noted as pseudo pod set as shown in Eq. (2).
Measurement of yield components of pod and nuts. The pods were harvested 130 days after pollination to ensure all the nuts of the kola accessions are of same maturity and data were collected on yield and yield components of pods and nuts. Number of pods harvested per each cross was counted and recorded. Pod length and width were measured using tape measure (Royal Dockyard Tape Measure-KA037) and digital caliper (NEIKO Stainless Digital Caliper 01407A) respectively. The weight of the pods was measured using weighing balance. The pods were broken and the number of nuts per pod was counted and recorded. Nut length and width were measured using digital caliper. Weight of nuts with the peels and weight of the nuts after removing the peels were measured using the weighing balance. The proportion of the total peeled weight of the nut of each cross to the weight of fresh unpeeled nuts was estimated as the outturn as shown in the formula below.
Determination of firmness, brix and potential alcohol content of the kola nuts. Firmness of the nuts from each cross within MX2, JX1, GX1 and Bunso progeny trial was determined using a handheld penetrometer (EFFEGI model ft327 (3-27Lbs, Italy). The nuts collected were stored in a nylon net and kept at room temperature for 24 h to ensure uniform temperature of nuts of the various crosses. A disc of about 2 cm in diameter was peeled of the skin of each nut using a stainless-steel vegetable peeler. A plunger with the tip size 5/16 was used. The plunger was forced into the nut through the peeled surface with the nuts against a stationary hard surface with a uniform speed of about 2 s. Depth of penetration was consistent to the inscribed line on the tip of the plunger. Readings were recorded in poundforce (lbf).
To measure the brix, the refractometer (HANNA instruments HI 968113, Romania) was calibrated using distilled water. The nuts of the crosses were crushed using mortar and pestle. The juice in the crushed nuts was then sieved through cheeth cloth into glass beaker. A disposable plastic pipette was then used to drop the juices into the window of the refractometer. Total soluble solids (TSS) was determined with a digital refractometer and was expressed in Brix 0 . Potential alcohol content of the juice from the nuts of the various crosses was also measured and recorded.
Statistical analyses. Data collected were analysed using the R environment (V. 3.6.2) 108 . To test the effect of the crossing type on pod and pseudo-pod set, a generalised linear model with a binomial distribution error to the data organized as success and failure events was fitted. The prominence of compatibility classes within each specific cross type was compared by using a multiple proportion comparison test from the prop.test () function of base R. An analysis of variance or a Kruskall-wallis test (where relevant) was used to assess the effect of the cross type on the pod and pseudo-pod yield as well as on the nut firmness, brix and potential alcohol content. When the analysis showed significance, mean comparisons were made according to the least significance difference test at P < 0.05 (LSD 0.05 ).To understand the grouping patterns of different type of crosses based on pod and pseudo-pod traits, a hierarchical clustering on principal component analysis was computed using the Fac-torMineR package 109 , and the graphical outputs were visualized using the factoextra package 110 . The correlation strength and significance among different pod and pseudo-pod traits was assessed using function corr_plot () of the Metan package 111 .
Heterosis for sexual compatibility (pod set %), outturn and brix were computed using the overall mean of each cross over replications. The heterosis of single cross hybrids and double cross hybrids were computed after Fehr and (1987) 112 . Relative heterosis or mid parent heterosis was estimated as percent deviation of hybrid cross value from its mid-parental self-cross value. The formulae used for estimating the relative heterosis was as follows: where di is the Heterosis over mid parental value (relative heterosis), F1 is the mean hybrid cross performance for pod set, outturn and brix. MP is the mid parental value, i.e., the arithmetic mean of self-cross pod set, outturn and brix of two parents involved in the respective cross combination.     www.nature.com/scientificreports/