Availability and distribution of Crotalaria species in Kenya
The present study determined that there are 21 morphologically distinct Crotalaria species distributed in 15 counties in Kenya (Supplementary table 1) with three species already domesticated by two communities in the country. These species were predominantly found in Kakamega, Vihiga, Busia, Bungoma, Kisumu, Migori, Siaya and Homa Bay counties. Only 9 of 133 accessions (6.77%) provided by the Genetic Resources Research Institute of Kenya are domesticated by local communities in Kenya, and belonged to the species C. brevidens Benth and C. trichotoma Bojer.
Edible Crotalaria species in Kenya
Three Crotalaria species were found to be cultivated by farmers in Kenya (Figure 3). These are C. brevidens Benth. var. brevidens, C. brevidens Benth. var. intermedia, C. ochroleuca G.Don. and C. trichotoma Bojer./C. zanzibarica Benth.. Morphologically these three species closely resemble one another but with a few differences. C. brevidens Benth. var. brevidens and C. brevidens Benth var. intermedia both have slender pods at maturity compared to C. ochroleuca and C. trichotoma which have relatively broad pods. The calyx of C. brevidens Benth are relatively puberulous with bright yellow petals while those of C. ochroleuca, and C. trichotoma are glabrous with pale yellow petals. The difference between the two varieties of C. brevidens is the colour of leaflets which are dull green in C. brevidens Benth. var. brevidens and bright green in Crotalaria brevidens Benth. var. intermedia. C. trichotoma, sometimes referred to as C. zanzibarica is an erect herb which grows taller than the other three (up to 2.70m) with a woody base. It is morphologically close to C. ochroleuca which is relatively shorter (up to 2.50m) but highly branching. C. ochroleuca, and C. trichotoma have a mild taste while C. brevidens have a bitter taste although the degree of bitterness depends on the place and time of propagation. Out of the 83 accessions used in this study, 28 (33.7%) were C. ochroleuca, 4 (4.8%) were C. trichotoma and 11 (13.3%) were C. brevidens.
Cultivation aspects of Crotalaria vegetables.
From a sampled population of 37 farmers, 86.5% knew of two types of edible Crotalaria species, 10.8% knew of three types while 2.7% did not know of the different types. However, this was based on the taste of cooked vegetables, which they categorized as either mild, bitter or very bitter. Edible Crotalaria species in Kenya were found to be cultivated by small holder farmers, with land sizes ranging from 0.20 to 8.0 hectares. The crop was mainly cultivated in kitchen gardens with average land size under the crop observed of 0.058 Ha. Of the farmers interviewed, 89.2% planted the crop in less than 0.05 Ha (Supplementary material 3). A correlation analysis of the household size, farm size and the total land area under Crotalaria vegetables revealed that there was a negative correlation between household size and farm size (r2=-0.21) although this was not significantly different (p=0.901). There was a positive correlation between household size and the portion under which Crotalaria vegetables were grown (r2=0.17) which was also not significant (p=0.314). Farm size and the portion of land under Crotalaria vegetables were positively correlated (r2=0.665, p=0.001) (Supplementary table 4). Analysis of variance revealed that there was a significant difference in the portions of land under the crop among all the interviewed farmers (p=0.001), with the largest being 0.8 Ha and the least being less than 0.05 Ha. Household sizes for the interviewed farmers ranged from two to eight people and this was found to have an effect on how much piece of land will be cultivated to plant Crotalaria vegetables, with large households cultivating more land than the small households.
It was observed that 89.2% of the farmers who cultivated crotalaria vegetables were aged between 31-65 years with none of the respondents aged between 0-30 years. Only 6 (16.2%) of the 37 respondent farmers who cultivated Crotalaria vegetables had formal education beyond the secondary level. All the farmers with tertiary education planted the vegetable for commercial and domestic purposes. However,15 (40.5%) farmers with primary and secondary education planted the vegetables solely for domestic consumption. Majority (70.3%) of the farmers admitted to eating the vegetable weekly, while 4 farmers (1.2%) ate the vegetable daily and fortnightly. Of all the interviewed farmers, only 5 (13.5%) were male, while 32 were female, representing 86.5%.
Variation in Quantitative characters in Kenyan Crotalaria species
Significant variations (p<0.0001 for all parameters) between accessions were recorded for all the quantitative variables; height, number of leaflets, leaf length, leaf width, number of branches, pod length, pod diameter, pods per plant and leaf area (Supplementary Table 2). Mean height ranges were between 2.65 m in sample GBK-005664 (C. trichotoma) and 0.34 m in sample GBK-005489 (C. greenwayi). From the 83 samples used in this study, 76 had compound leaves with three leaflets except for C. grahamiana which had seven leaflets. The other seven samples had simple leaves, representing the five species; C. juncea, C. spectabillis, C. retusa, C. paulina and C. pancira. Leaf lengths ranged from 19.0 cm in sample GBK-005209 (C. paulina) to 1.13 cm in sample GBK-005664 (C. greenwayi). The leaf area followed the same pattern, with sample GBK-005209 (C. paulina) having the largest leaf area (122.51 cm2) and GBK-005664 (C. greenwayi) with the smallest leaf area (0.606 cm2). The most profuse branching was observed in samples GBK-005685(C. spectabilis) and GBK-005201 (C. pancira) which had a mean of 34.33 branches while the least number of branches observed in sample Fmg 0298 (C. pallida) with a mean of 3.00 branches. The mean pod length ranged from 6.7 cm in sample Fvh 0119 to 0.4 cm in sample GBK-047548 (C. cephalotes) while mean pod diameter ranged from 5.83 cm in sample Fvh 0119 to 0.90 in sample GBK-005664 (C. greenwayi). Sample GBK-005189 (C. laburnifolia) had the highest mean number of pods while Fki 0123 (C. ochroleuca) had the least mean number of pods.
Variation in qualitative characters
The qualitative characters that were studied include; leaf shape, pod shape, leaf type, flower colour, stem colour, hairiness of the stem, dry pod colour, seed colour and growth habit (Figure 4). There were 6 leaf shapes observed within the genus, of which three were observed in 74 accessions, accounting for 89.1% of the shape diversity observed. Oblanceolate leaves/leaflets were the most common and were observed in 27 (32.5%) samples, followed by the elliptical 25 (30.1%) and linear 22 (26.5%). The obovate, orbicular and ovate leaves/leaflets constituted 5 (6%), 2 (2.4%) and 2 (2.4%) accessions respectively. The long broad smooth pod was the most frequent and was observed in 49 (59%) accessions, followed by the long slender smooth, observed in 22 (26.5%) accessions. The long broad hairy, short broad hairy and short slender smooth pods were the lest frequent, with each constituting 1.2%. The short broad smooth and long slender hairy pods constituted 7.2% and 3.6% respectively. The compound leaf type was observed in 76 (91.6%) accessions while 7 (8.4%) accessions had simple leaves. The pale yellow corolla with purple stripes were observed in 29 (34.9%) samples while the deep yellow with purple stripes were observed in 15 (18.1%) samples. The lemon yellow corolla with purple stripes and the chartreuse yellow were the least, representing only 1.2% of the samples. The brown-green stems were the most abundant in the population (51.8%), while green and pale green were 14.5% and 13.3% respectively. Purple-green and shiny-green stems each represented 6% of the samples. Dark-green and yellow-green stems were the least abundant in the samples representing 4.8% and 3.6% respectively. The stems were either smooth (75.9%) or pubescent (24.1%). Black dry pods constituted 63.9% while brown dry pods represented 33.7%. Brown with black striped pods and green with brown striped pods were 1.2% each. Pale-yellow seeds were the majority (32.5%), while the mixture of pale-yellow and red-orange, and the purely red-orange seeds constituted 16.9%. Other seed colours observed were black (6%), brown (7.2%), grey (6%), pale grey (2.4%), black and grey mixture (2.4%), grey and pale-yellow mixture (1.2%) striped grey (3.6%), maroon (2.4%) and purple (2.4%). Most of the cultivated accessions had mixed seed colours in a pod whereas the wild species had distinct uniform seed colours.
Correlation between quantitative variables
A weak positive correlation was observed between height and the following traits: number of leaflets, number of branches, pod length, pod diameter and the number of pods per plant (Table 1). Height and leaf length had no correlation. However, height only correlated significantly with the number of leaflets, (r2=0.158, p=0.0128), number of branches (r2=0.160, p=0.0111), pod length (r2=0.191, p=0.0024) and pod diameter (r2=0.125, p=0.0481). Both leaf width (r2=-0.170, p=0.007) and leaf area (r2=-0.149, p=0.018) had a significant inverse correlation with height. All parameters correlated negatively with the number of leaflets except pod length (r2=0.192, p=0.002) and pods per plant (r2=0.074, p=0.240). Among the negatively correlating parameters, a moderate negative correlation was observed only between the number of leaflets and leaf area (r2=-0.561), which was significantly different (p<0.0001). Although leaf length and leaf width were moderately negatively correlating with number of leaflets, this correlation was significantly different with p<0.0001 for both parameters. Leaf length correlated positively with all the other parameters. However, only leaf area (r2=0.74; p<0.0001) and had strong positive correlations with leaf length, with leaf width (r2=0.531; p=<0.0001), number of branches (r2=0.297, p=<0.0001), pod length (r2=0.241, p=0.0001), pod diameter (r2=0.227, p=0.0003) and pods per plant (r2=0.201, p=0.0015) all showing moderate to weak positive correlations. Leaf width had no correlation with pod length (r2=-0.052, p=0.4057), but had positive correlation with all the other parameters. However, only leaf area strongly correlated with leaf width, (r2=0.886, p<0.0001). The number of branches had a positive weak correlation which was significantly different from zero with the following parameters: pod length (r2=0.157, p=0.013), pod diameter (r2=0.214, p=0.0007), pods per plant (r2=0.220, p=0.0005) and leaf area (r2=0.153, p=0.015). Pod length had a significantly moderate positive correlation with pod diameter (r2=0.654, p<0.0001) but had no correlation with pods per plant and leaf area. Pod diameter and pods per plant were weakly negatively correlated, which was significantly different from zero (r2=-0.257, p<0.0001). A significant weak positive correlation was observed between pods per plant and leaf area (r2=0.197, p=0.0018).
Agglomerative hierarchical clustering of qualitative and quantitative traits
Agglomerative hierarchical clustering of both qualitative and quantitative traits of the 83 Crotalaria accessions assembled them into eight major clusters at 7.5 Euclidean distance (Figure 5). The first cluster (A) had only one wild accession, C. spectabilis. The distinct feature of this accession was the simple leaf type with the highest leaf area index. The second cluster (B) consisted of four wild accessions of different species with two sub-clusters. The first sub-group consisted of the species C. incana and C. endecaphylla while the second consisted of C. cephalotes and C. greenwayi species. All these species have trifoliate compound leaves and other similarities in qualitative traits such as brown dry pods and hairy stems. The third cluster (C) comprised three wild accessions representing two species; C. juncea and C. greenwayi. These species are herbaceous with brown dry pods. Cluster four (D) consisted of four wild accessions with four different species; C. spectabilis, C. retusa, C. pancira and C. paulina. These accessions are perennial shrubs with simple leaves, their dry pods are brown with black or grey seeds and have shiny green stems. Cluster five (E) was made up of 14 accessions representing 10 species. All the accessions in this cluster were wild perennial shrubs with compound leaves. The first sub-group consisted of five accessions of the species C. pallida, C. laburnifolia, C. intermedia and C. anagyroides, all with three leaflets and brown dry pods. The other sub-group in this cluster consisted of a single accession of the species C. grahamiana which has seven leaflets. The other sub-cluster in the group consisted of two accessions namely; C. scassellatii and accession Fkj 0294 (collected from Kajiado county). The last sub-cluster in this group consisted of six accessions, representing only two species; C. recta, and C. endecaphylla, characterised by compound trifoliate leaves and black dry pods. The sixth cluster (F) consisted of 36 cultivated accessions. A common feature among these accessions is that their dry pods are black, and their flowers have purple stripes. The cluster consisted of the two species, C. brevidens and C. ochroleuca. This cluster had three sub-clusters, two entirely made of C. ochroleuca samples while the third consisted of a mixture of C. brevidens and C. ochroleuca accessions. The Seventh cluster (G) consisted of the wild type perennial herb C. intermedia. The last cluster (H) was made up of 20 accessions, comprising of both cultivated and wild species. This cluster contained three sub-clusters, representing six species; C. brevidens, C. intermedia, C. incana, C. deserticola, C. lanceolata, and C. trichotoma.
Table 1: Correlation of nine quantitative traits of 83 Crotalaria accessions from Kenya
Variable
|
Height (ft)
|
No of leaflets
|
Leaf length (cm)
|
Leaf width (cm)
|
Number of Branches
|
Pod length (cm)
|
Pod Diameter (cm)
|
Pods per Plant
|
Leal area (cm2)
|
Height (ft)
|
1.00000
|
|
|
|
|
|
|
|
|
No of leaflets
|
0.15752
|
1.00000
|
|
|
|
|
|
|
|
Leaf length (cm)
|
0.04413
|
-0.42131
|
1.00000
|
|
|
|
|
|
|
Leaf width (cm)
|
-0.17014
|
-0.43052
|
0.5311
|
1.00000
|
|
|
|
|
|
Number of Branches
|
0.16074
|
-0.09821
|
0.29659
|
0.00957
|
1.00000
|
|
|
|
|
Pod length (cm)
|
0.19154
|
0.19164
|
0.24146
|
-0.05292
|
0.15774
|
1.00000
|
|
|
|
Pod Diameter (cm)
|
0.12537
|
-0.01905
|
0.22668
|
0.07796
|
0.21449
|
0.65484
|
1.00000
|
|
|
Pods per Plant
|
0.11343
|
0.07467
|
0.2005
|
0.179
|
0.22013
|
-0.00918
|
-0.25797
|
1.00000
|
|
Leaf area (cm2)
|
-0.14985
|
-0.56131
|
0.74386
|
0.88564
|
0.15342
|
-0.00792
|
0.10351
|
0.1971
|
1.0000
|
Values in bold are different from 0 at a significance level of p<0.05.
Table 2: The contribution of variables to the first five Principal Components of 83 crotalaria accessions in Kenya.
Traits
|
PC 1
|
PC 2
|
PC 3
|
PC 4
|
PC 5
|
Eigenvalue
|
2.961
|
1.730
|
1.398
|
0.838
|
0.764
|
Variance %
|
0.329
|
0.192
|
0.155
|
0.093
|
0.085
|
Cumulative %
|
0.329
|
0.521
|
0.676
|
0.770
|
0.855
|
Factor loading
|
|
|
|
|
|
Height (ft)
|
0.092
|
-0.346
|
0.394
|
-0.467
|
0.6803
|
No of leaflets
|
0.379
|
-0.212
|
0.232
|
0.458
|
-0.1205
|
Leaf length
|
-0.524
|
-0.117
|
0.034
|
-0.050
|
-0.0489
|
Leaf width
|
-0.421
|
0.158
|
-0.101
|
0.342
|
0.4087
|
Number of branches
|
-0.234
|
-0.399
|
0.294
|
-0.344
|
-0.5601
|
Pod length
|
-0.001
|
-0.616
|
-0.149
|
0.346
|
0.1838
|
Pod diameter
|
-0.045
|
-0.507
|
-0.488
|
0.072
|
-0.0548
|
Pods per plant
|
-0.167
|
-0.026
|
0.657
|
0.459
|
-0.0482
|
Leaf area
|
-0.558
|
0.006
|
-0.013
|
0.023
|
0.0214
|
variable-reduction by Principal Components
The principal component analysis (PCA) grouped the nine quantitative traits into nine principal components, which accounted for all the genetic variation observed in the 83 Crotalaria accessions. The first five Principal Components (PCs) explain 85.5% of the observed variation in the accessions, out of which 52.1% of the variability was accounted for by the first two PCs (Table 2). The traits that contributed to the most variability among accessions both in the negative and positive loadings in the first PC1 were leaf area, leaf length, leaf width and number of leaflets. With an Eigen value of 1.73, the second PC accounted for 19.2% of the phenomic variation. The phenotypic attributes that mainly contributed to PC2 were pod length, pod diameter number of branches and plant height. Pods per plant, pod diameter, and height were the traits that contributed to the variation accounted for by PC3, whose Eigen value was 1.39.
The correlation between variables and PCs revealed that the variables height, leaf length, and leaf area are well explained by the first two PCs, since they appear at the periphery of the correlation circle. These two PCs however cannot properly explain the variables pod length and pod diameter as they appear at the center of the correlation circle (Figure 6a). Based on this bi-plot, PCs 1 and 2 have almost no information on pods per plant but provide adequate information on diversity aspects based on leaf area and leaf length. A bi-plot on the correlation between accessions and PCs revealed a similar grouping as that of the cluster dendrogram, with the furthest accession in the bottom left quartile clustering as a distant relative of the other accessions (Figure 6b). The bi-plots also revealed that the variables leaf area and pod length and also leaf length and height are orthogonal.
Most of the observed variation was contributed by the quantitative traits, with the qualitative traits contributing little to the clustering. The magnitude of contribution of quantitative and qualitative traits to variation as depicted by the heat maps generated based on the counties of collection (Figure 7) and agro-ecological regions (supplementary figure 1) resonates with the PCA output. The most important quantitative traits were the number of pods per plant, the leaf area, number of branches, leaf length, and pod length. Seed colour followed by pod shape were the most contributing qualitative traits to diversity.
NMDS clustering
According to the Non-metric multidimensional scaling (NMDS) analysis, the close affinities among the accessions in each sample type are supported by their proximity to one another though there is still a little morphological overlap with other samples. The NMDS ordination with a stress value of 0.165 indicated an acceptable ordination of the data. Characteristic of the sampling counties, fifteen NMDS groups were formed (Figure 8). Similarly, six groups were formed representing the six agro-ecological sampling regions (Figures 9). The two most significant variables that generated the clusters in the two-dimension model at P<0.05 were number of pods per plant and leaf area index (with a score of 12% and 0.8% respectively). Based on the administrative clusters (Figure 8), accessions collected from Kakamega county and the neighbouring counties exhibited similar morphological traits with exception of Nairobi county. A unique overlap between accessions collected in Migori and Kajiado counties was observed. Samples from Makueni and Embu counties as well as GBK accessions exhibited peculiar characteristics and thus showed no significant relatedness to the other accessions. Based on the agro-ecological regions, accessions collected from Western region and the neighbouring Nyanza and Rift valley regions exhibited similar morphological traits. Accessions collected from Nyanza region shared a noticeable proportion of similarities with accessions collected from the Rift valley region. Accessions collected from the Eastern region and gene bank exhibited peculiar characteristics thus showed no significant relatedness to the other study accessions. The ANOSIM value of (R=0.4807, P=0.001) supported the even distribution of high and low ranks within and between groups of Crotalaria species in the different counties and agro-ecological regions. Although distantly separated, the accessions from Nairobi clustered together with those from Kakamega and exhibited similar morphological appearance.