Genetic diversity in Colocasia esculenta and Xanthosoma mafaffa in Togo, West Africa

Taro and new cocoyam are root and leaf crops commonly grown in tropical to warm temperate regions. In Togo, they are neglected and underuti­ lized. Here we report the genetic diversity of 26 accessions of taro and 101 accessions of new cocoyam. Analysis of simple sequence repeats revealed low polymorphic information content of 0.43 and 0.25 in taro and new cocoyam, respectively. PCA scatterplots and Neighbour Joining dendrograms based on the SSR data clustered accessions into groups that more­or­less correspond to morphological diversity in both species. AMOVA within and between morpho­ logical groups revealed greater variances within groups than between. This indicates weak genetic differentiation between morphological groups, particu­ larly for taro. Genetic diversity was greater among taro cultivars. Taro has a longer history of introduction and dispersal in Africa, and has had more oppor­ tunity for multiple introduction and local cultivar development. Different strategies are suggested for future development of these crops in Togo and Africa. For taro, further studies of existing diversity and recent experimental introductions to Africa are likely to be rewarding. New cocoyam, a modern his­ torical introduction, has spread widely in Africa with little genetic diversity. For this crop, international collaboration is needed to clarify taxonomy, and to introduce further cultivars for evaluation under local conditions in Africa.


Introduction
Root and tuber crops are important sources of food and income for household in rural areas of Africa. In subSaharan Africa, they provide about 20% of calories (PinstrupAndersen et al., 1999). Taro, Colocasia esculenta (L.) Schott and new cocoyam, Xanthosoma spp., are grown for food and income generation at the household level. Both crops are grown in tropical regions of Africa, Asia, Oceania and America, and taro is also common in temperate regions of Africa, Asia, and Oceania (Gonçalves, 2011;Matthews, 2014;Onyeka, 2014;Grimaldi, 2016;Matthews and Ghanem, 2021). Taro is considered an ancient crop in Africa, with multiple likely routes of introduction from Asia (Matthews, 2006;Fuller et al., 2011;Chaïr et al., 2016;Grimaldi, 2016). New cocoy am is known to have been introduced to Ghana in 1843 (Karikari, 1971), though earlier introduction fol lowing European contact with America has been sug gested (Bown, 2000).
Most parts of these plants (corms, sidecorms, stolons, petioles, leaf blades, and floral spathes) are edible, the leaves and corms are also commonly used as animal fodder in Asia (Coursey, 1984;Matthews, 2010;Mwenye et al., 2010;Masuno et al., 2012;Matthews, 2014;Wada et al., 2017), and medicinal uses are also known (Plowman, 1969;Ribeiro Pereira et al., 2021). The specific parts eaten vary according to cultivar attributes, local food knowledge, and cul tural or personal preferences. Both crops have great potential for development in Africa and globally (Okereke, 2020).
Under cultivation, clonal propagation is universal for both taro and new cocoyam. In a global survey of taro, Chaïr et al. (2016) found the greatest genetic diversity and the largest number of private alleles in Asian cultivars, especially in India. Low genetic diver sity was found in western Africa, among diploid and triploid cultivars, and also in southern Africa, where triploid cultivars were dominant. Their observations correspond broadly to what is known about the breeding of taro: flowering, fruiting and seed produc tion by wild and cultivated taros are common in trop ical regions of Asia and Oceania (Matthews, 2014), but have not been observed in Africa despite occa sional reports of flowering (e.g. Traore, 2013). Natural fruiting and seed production by new cocoy am has not been reported outside South America, but sterile inflorescences are often produced (obser vation by authors), and induced flowering and experi mental crosses have been reported in Cameroon (Onokpise et al., 1992).
Globally, taro has undergone genetic erosion due to changes in cropping patterns, the spread of improved varieties, and replacement by other crops (Lakhanpaul et al., 2003;Matthews and Ghanem, 2021), including new cocoyam (Coursey, 1984). The acridity of taro (and resulting special care needed for cooking) (Matthews, 2010), spread of taro leaf blight (TLB) (Singh et al., 2012), and shortages of planting materials are contributing factors. In order to identify and preserve cultivars of economic value, maintain living germplasm collections efficiently (without excessive duplication), and provide baseline data for future breeding programs, genetic diversity and mor phology must be assessed in cultivars of both species.
The taxonomy of cultivated Xanthosoma species and relationships with wild species are uncertain, and historically there has been a tendency to use the name X. sagittifolium for all cultivated Xanthosoma (Giacometti and Leon, 1994;Castro, 2006;Quero Garcia et al., 2010;Doungous et al., 2015). Although X. sagittifolium is the name used in many previous studies in Africa, the plant is most likely to be X. mafaffa (Gonçalves, 2011). The taxonomy of cultivat ed Xanthosoma spp. in tropical America has been revised by Croat and Delannay (2017). Various culti vated species of Xanthosoma are also circulating internationally and may have reached Africa in the modern historical period. These include X. atrovirens C. Koch & Bouché (blackish green blades, and "blue" wax on dark green petioles creating a dark purple or black appearance), X. robustum (which can reach 4 m in height, with tall aboveground stems) and X. violaceum (with violet petiole tissue below a waxy sur face) (Gonçalves, 2011). In addition to the 1843 intro duction of Xanthosoma (Karikari, 1971), an unsuc cessful attempt was made to introduce X. brasiliensis from Puerto Rico in 1974 (Karikari, 1979). In Central America, X. mafaffa cultivars vary in corm parenchy ma color (red or white), and those with red corms are also tinged with red in the petiole, leaf sheath and spathe (Gonçalves, 2011).
In Togo, a collection of taro and new cocoyam cul tivars from throughout the country was assembled, and morphological groups were described in both species (Bammite, 2018;Bammite et al., 2018 a, b). However, data related to genetic diversity among these crops, based on molecular tools such as SSR, are lacking to enhance effective usage and conserva tion of these neglected species and develop a breed ing program to improve the quality of their germplasm. This study aimed to assess simple sequence repeat (SSR) diversity in the same collec tion of taro and new cocoyam. Polymorphic SSR loci have codominant alleles (repeat sequences of vary ing length, detected by PCR amplification), and been used in many studies of taro (Devi, 2012;You et al., 2015) and other edible aroids (Suppl . Table S1), and in genetic linkage mapping of the taro nuclear genome (QueroGarcía et al., 2010;Soulard et al., 2017).

Plant materials
In 2016, cultivars of taro and new cocoyam were collected from 42 localities randomly selected across the five ecological zones of Togo ( Fig. 1). At each vil lage, farmers were invited to bring corms of different cultivars grown in their village. For each distinct culti var recongnized in group discussions, the local name was recorded and collected corm samples were planted at the Centre de Recherche Agronomique du Litoral (CRAL), an experimental farm of the Institut Togolais de Recherche Agronomique (ITRA) located at Davié, at latitude N 6°23' and longitude E 1°12' and at 88 m above sea level (Table S2). Accessions of both species were classified in a binary manner based on morphological characters that are easily observed in the field: taro accessions were identified as either dasheen (with large mother corms, and either stolons or sidecorms; Pop1), or eddoe (with small mother corms, and few to many sidecorms; Pop2); new cocoyam accessions were identified as either green (leaves entirely green; Pop1) or purple (petioles purple or pink to some extent; Pop2). Morpological diversity in the same collection was previously analysed with reference to a wide range of agronomic and morphological characters (Bammite et al., 2018 b;Figs. S1S3). Young leaf tissue from one plant from each of 26 accessions of taro and 101 accessions of new cocoyam was dried on silica gel and taken to the International Institute of Tropical Agriculture (IITA) Bioscience Centre, Ibadan, Nigeria for DNA extraction and genotyping.

DNA isolation and quantification
DNA was extracted using an optimized SDS proto col recommended by IITA Bioscience Centre (2017). About 100 mg of dry leaf tissue was put in a tube with two steel balls and reduced to powder using a SPEX Genogrinder2000. Preheated extraction buffer (450 µl of 1M TrisHCl, 0,5M EDTA, 5M NaCl, 20% SDS and 1% PVP) was added. Tubes were incubated at 65°C for 20 mins and inverted occasionally to homog enize each sample. Tubes were removed from bath, allowed to cool for two mins, then 200 µl of icecold 5M potassium acetate was added and the mixture incubated on ice for 20 mins to precipitate proteins. Tubes were then centrifuged at 3500 rpm for 10 mins, and each supernatant was transferred to a new labeled tube. A volume of 200 µl of 4% polyvinylpyrrolidone (PVP) was added to the super natant and gently mixed. To precipitate and remove proteins and lipids, 45 µl of chloroform isoamylal chohol (24:1) was added, mixed gently and tubes were centrifuged at 3500 rpm for 15 mins. Each supernatant was transferred to a new tube, and a 2/3 volume of icecold isopropanol was added, mixed, and incubated in 80°C for 15 mins to precipi tate the DNA. After centrifuging at 3500 rpm for 15 mins, the DNA pellet was washed by adding 400 µl of 70% ethanol, centrifugation at 3500 rpm for 15 mins, and decanting the supernatant until the last drop. The DNA pellet was airdried then resuspended in 100 µl low salt TE buffer (10 mM TrisHCl, 1 mM EDTA). A volume of 2 µl of RNase A (10 µg/ ml) was added and incubated at 37°C for 40 mins. The quanti ty and quality of extracted DNA was checked using electrophoresis with 1% agarose gel, and a Nanodrop 8000 spectrophotometer, and the extracts were stored at 20°C until use.

PCR amplification
Initial testing was carried out with 47 primer pairs designed in previous SSR studies: 19 for for Amorphophallus paeoniifolius (Santosa et al., 2007), 11 for C. esculenta (Hu et al., 2009.) and 17 for X. sagittifolium (Cathebras et al., 2014) (Table S3). The 47 primer pairs were tested first with five samples from each target species to determine which pairs could amplify scorable DNA products in each species.
PCR amplification was performed in PCR mixture (25 µl) containing 2.5 µl of template DNA (20ng/µl), 2.5 µl of 10x NH4 PCR reaction buffer, 1 µl of 50mM MgCl 2 ), 1 µl of 5 µM forward primer, 1 µl of 5 µM reverse primer, 0.2 µl of 5mM each dNTP, 0.1 µl of BIOTAQ DNA polymerase and 16.1 µl of water. The PCR program consisted of initial denaturation (94°C, 5 mins), 42 cycles each consisting of 20 s denatura tion (93°C), 1 min annealing at temperatures ranging from 47 to 59°C (as recommended by the authors above; Table S3), and 2 mins elongation (72°C). Finally, an extension period of 10 mins was included. After PCR completion, the products were stored at 4°C until gel electrophoresis. Ten µl of each PCR product was electrophoresed alongside a 50 bp DNA ladder (New England Biolab) in polyacrylamide gel (10% InstaPAGE gel) at 110 V for one hour, and bands were visualized by silver staining (1L TBE 0,5X buffer + 500 µL of SafeView) for 3 mins. Amplified fragment sizes were determined by comparison to the 50 bp ladder, and bands were examined and recorded using the ENDURO TM Gel Documentation System.

Data analysis
Recorded gel images with PCR products were analysed the the Image Studio Lite Ver 5.2 software, generating binary matrix data for all accessions based on the band patterns observed at each SSR locus: presence of an amplified band was scored as "1"; absence was scored as "0". Summary statistics for each locus were estimated using PowerMarker 3.25 software.
For statistical comparisons within each species, the binary classifications of morphotypes were used: taro dasheen = Pop 1, eddoe = Pop2; new cocoyam green = Pop1, purple = Pop2. GenAlex 6.5 software (Peakall and Smouse, 2006) was used to calculate the number of polymorphic loci (no. PL), the percentage of poly morphic loci (% PL), the observed number of alleles (Na), the effective number of alleles (Ne), average expected heterozygosity (He) (also known as Nei's gene diversity, Nei, 1973), and Shannon's information index (I) (a measure of genetic diversity suitable for codominant data). Analysis of molecular variance (AMOVA) was performed to evaluate the genetic varia tion within and among morphotype populations by using GenAlEx version 6.503 (Peakall and Smouse, 2006), and PhiPT (the proportion of total genetic vari ance derived from variance between individuals among populations, i.e. an estimate of population genetic dif ferentiation). Here, our H 0 = no genetic difference among populations, H 1 = there is genetic difference among populations, and p= probability of an observed PHiPT value =/> than that observed by chance, if the null hypothesis (no genetic difference) is true.
Principal Coordinate Analysis (PCA) was carried out using the GenAlEx version 6.503 (Peakall and Smouse, 2006). The genetic distance matrix was con structed by calculating the shared allele distance for each pair of individuals in PowerMarker version 3.25. From this matrix, a neighbour joining (NJ) tree was constructed using Nei's genetic distances (Nei et al., 1983) in the same software.

Results
In the initial test with five samples of each target species, only 27 of 47 SSR primer pairs, amplified and gave polymorphic, scorable bands ( Fig. 2 Tables 1 and 4, and are sum marised below. Among 26 accessions of taro, 33 alleles were observed at 13 loci, an average of 3.15 alleles per locus (loci are hereafter identified by the primer code names). The frequency of major alleles ranged from 0.38 for locus HK7 to 0.88 for HK5, with an average of 0.62. Nei's gene diversity ranged from 0.21 for HK7 to 0.71 for HK5, with an average of 0.49. Polymorphic information content (PIC) values ranged from 0.20 to 0.65 with an average of 0.43. The primers HK35, HK26, HK38, Ampa9, HK7 with PIC values >/= 0.5 were most discriminating (Table 1). The percentages of polymorphic loci for each morphotype of taro were 82% (dasheen, Pop1) and 58% (eddoe, Pop2), with an average of 70% (rounded figures). The num bers of different (Na) and effective (Ne) alleles, Shanons Information Index (I), and Nei's gene diversi ty (He) were all higher in dasheen and lower in eddoe ( Table 2). df = degrees of freedom, SS = sum of squares, MS = mean sum of squares, Est. var. = estimated variance, % var = percentage of variation, PhiPT = proportion of total genetic variance derived from variance between individuals among populations, p = probability value for PhiPT. Among the 101 accessions of new cocoyam, 48 alleles were observed at 16 loci, an average of 3.0 alleles per locus. The frequency of major alleles ranged from 0.47 for Ampa9 to 0.97 for Ampa15, with an average of 0.83. Nei's gene diversity (He) ranged from 0.06 for Ampa9 to 0.67 for Ampa15, with an average of 0.28. PIC values ranged from 0.06 to 0.62, with an average of 0.25. Only the Ampa9 locus gave a PIC value greater than 0.5 (Table 3).
The percentages of polymorphic loci for each morphotype of new cocoyam were 74% (green, Pop1) and 94% (purple, Pop2) and with an average of 84% (rounded figures). A lower number of different alleles was recorded in the green population (Na = 1.74, n = 23) and a higher number in the purple pop ulation (Na = 1.87, n = 78), but the number of effec tive alleles (Ne) and other measures of diversity (I, He) were higher in the green population (Table 4).

Cluster analysis and structuring of genetic diversity
Analysis of molecular variance (AMOVA) gave per centages of molecular variance of 83% within and 17% between the dasheen and eddoe populations of taro, indicating weak differentiation overall (Table 2). For new cocoyam, the percentages of molecular vari ance were 64% within and 36% between the green and purple population ( Table 4). The probability value (p) for PhiPT is higher in taro (0.014) than in new cocoyam (0.001) providing for taro a weaker rejection of the H 0 of no genetic difference between populations .
The first and second coordinates of the PCA scat ter plot (Fig. 3) represent, respectively, 48% and 16% (in total 64%; rounded figures) of the detected vari ability among taro accessions. For cocoyam acces sions, the coordinates represented 41% and 13% (in total 54%) (Fig. 4). Some dasheen taros formed a dis tinct group along the first coordinate, but apart from this, there is no clear separation of dasheen and eddoe accessions overall in the SSR data. In contrast, the green and purple new cocoyam formed very dis tinct groups along the first PCA coordinate. These relationships between SSR diversity in PCA scatter plot and simple morphotype classification are mir rored in the NJ dendrograms. The NJ dendrogram of SSR diversity in taro (Fig. 5) revealed one larger cluster (C1) that includes a mix of dasheen (7 accessions) and eddoe types (10 acces sions), and a two clusters (C2, C3) that include dasheen types only (9 accessions). The accessions in C3 were all dasheen types, from wet, flooded envi ronments, with purple petiole and petiole junction, and generally producing many stolons (see Discussion and Conclusions) (Fig. S1A, Fig. S2).
The NJ dendrogram of SSR diversity in new cocoy am revealed two large clusters (C1 and C2) (Fig. 6) that largely correspond to the green and purple mor photypes of this species (13 and 88 accessions in each category; and many identical haplotypes).
With regard to the more specific morphological groups previously reported, C1 includes mostly G1, and C2 includes a mix of mostly G2 and G3 (purple morphotypes; Fig. S3).

Discussion and Conclusions
The present SSR results indicate few duplicate accessions in the smaller collection of taro (Fig. 5) and many apparent duplicates in the larger collection of X. mafaffa (Fig. 6). Qualitative and quantitative traits for morphological and agronomic characters of taro and new cocoyam were previously recorded and analysed by Bammite et al. (2018 b). Thirtyeight characters were selected from the descriptor list of IPGRI (1999) for C. esculenta and 28 from the descriptor list of IBPGR (1989) for X. saggitifolium. Based on these detailed observations, morphological groups within each species were identified using UPGMA analysis (Fig. S1). Although these groups (G1   G4 in taro, G1G3 in new cocoyam) do not always correspond as expected to the singlecharacter cate gories used in the present study (corm size and shape in taro, plant colour in new cocoyam) (Figs. S2 and S3), congruences are apparent between SSR genetic diversity and morphological diversity, however the latter is defined. Complete correspondence between singlecharacter and multicharacter classifications is not expected, but future studies of morphological diversity can be improved by ensuring greater unifor mity in the planting materials used. Ahmed et al. (2020) found that the dasheen/ eddoe classification corresponds to two distinct evo lutionary lineages within C. esculenta, and suggested that the existence of many intermediate or mixed morphotypes may reflect hybridisation between these lineages. In Togo, SSR cluster C1 includes both dasheen and eddoe forms, consistent with the sug gestion of mixing (hybridisation) between eddoe and dasheen lineages (see also Lakhanpaul et al., 2003). Clusters C2 and C3 corresponded entirely to dasheen forms with large mother corms and either stolons or sidecorms. The three accessions in C2 had morpho logical aboveground traits similar to those typical of eddoe in the Togo collection (large pendant or drooping leaves) and underground morphological traits of dasheen type (G2, with large central corm with small sidecorms). These accessions were col lected in home gardens of farmers from Kabyè ethnic group in the ecologic zone IV (southern Togo Mountains) (Fig. 1). They reported having introduced them from ecological zone II (northern Togo Mountains), which is consistent with the report by Ern (1979) of an expansion of banana, cassava and taro production on steep slopes in Zone IV by Kabyè settlers from the north. The observation of two major SSR clusters in new cocoyam (C1C2) suggests that more than one species of Xanthosoma is present in Togo, not just X. mafaffa (Bammite et al., 2018 a, b). A survey of Amplified Fragment Length Polymorphism (AFLP) in new cocoyam in Ethiopia also revealed two major clusters (Wada et al., 2018). The existence of such distinct lineages and the general uncertainty of iden tification of Xanthosoma species suggest an urgent need for direct and detailed comparisons between cultivated Xanthosoma spp. in Africa, and the wild and cultivated species of Xanthosoma in tropical America.
For both species, the Neighbour Joining analysis of SSR allelic diversity at a small number of loci pro vides tree diagrams in which terminal branching (near tree tips) is not a reliable indicator of phyloge ny. Much of the "within population" variation in both species may reflect somatic mutation within clonal cultivar lineages. The small numbers of loci analysed here (13 in taro, 14 in new cocoyam) make it inher ently difficult to distinguish clones with certainty, as there are 14 chromosomes in the haploid comple ment of C. esculenta (Coates et al., 1988;Cusimano et al., 2012), and 13 in Xanthosoma spp. (Cusimano et al., 2012;Wada et al., 2018) giving rounded aver ages of just 0.9 loci (taro) and 1.1 loci (new cocoyam) sampled per chromosome. For taro, actual coverage is less than 0.9 loci/chromosome, as two HK loci have been mapped to one linkage group and chromosome (Table S4). Although Chaïr et al. (2016) introduced a method to estimate clonality based on just 11 loci, the same research group also employed a much more robust method using Diversity Array Technology (DArT TM ) to screen polymorphic loci (pos sibly thousands) across the entire taro genome (Vandenbrouke et al., 2016). By combining the latter method with a detailed survey of morphological diversity in an assemblage of Vanuatu cultivars, Vandenbrouke et al. (2016) could unequivocally iden tify clonal lineages within which somatic mutation has produced distinct phenotypes that are recog nised, selected, and maintained by farmers. Most recently, Soulard et al. (2017) have mapped polymor phic SNP and SSR loci across the entire taro nuclear genome, while Yin et al. (2021) have published nearly complete sequences for all 14 chromosomes in taro. Somatic mutation may explain some of the mor phological and genetic diversity found in taro in Togo and Africa generally, but spontaneous breeding among diploid cultivars in Africa may also be involved. The chromosome numbers of Togo cultivars have not been studied, but triploid taros are widespread and common in Africa  and the neighbouring country of Burkina Faso (Traore, 2013), and can be assumed to be inherently sterile because they are triploids. The eddoetype taros (C. esculenta var. antiquorum) in Togo are likely to be triploids, as this morphotype is generally triploid in neighbouring Burkina Faso (Traore, 2013) and eastern Asia (Plucknett, 1983;Matthews, 2014;Wang et al., 2020), but this cannot be assumed if the diploid (fertile) progenitors of triploid eddoe cul tivars still exist, some might share the eddoe mor photype. It also cannot be assumed that the dasheen types are diploid. There are multiple triploid lineages in taro, and some dasheen and intermediate mor photypes are also triploid (Kreike et al., 2004).
The results of our initial primer screening corrobo rate those of Traore (2013), who found that primers designed for C. esculenta are not transferable to Xanthosoma spp. accessions. We also found that, conversely, the primers designed for Xanthosoma do not amplify C. esculenta accessions. In their original report of the HK primers, Hu et al. (2009) surveyed 30 plants from several provinces of China. Chaïr et al. (2016) screened 64 primer pairs developed from C. esculenta and A. paeoniifolius, and selected 11 from C. esculenta, of which three were from the HK primer series. This study (the largest survey of SSR diversity in taro) included 321 cultivars from 19 countries in Asia, Africa, America and the Pacific. Several HK primers were also used by Hunt et al. (2013). Including the present Togo survey, results for HK7, HK22 and HK26 can now be compared across four studies ( Table 5). The largest number of alleles was found in the largest sample set representing many countries , which is not surprising. The surprise here is that Togo, a relatively small country far from Asia, displayed only slightly fewer alleles than a similar number of plants from across China (Hu et al., 2009), a much larger country that is also a candidate region for the origin of triploid taros (Matthews, 2014;Wang et al., 2020;Zhu et al., 2000). The number of alleles in two wild breeding populations in Papua New Guinea and an adjacent region of northern Australia was larger than in the China and Togo cultivars, but also much less than in the large survey by Chaïr et al. (2016).
The relatively low number of alleles detected in Togo presumably reflects the small number of plants  3   18  3  HK26  5  8  28  3  HK31  3  4   2  HK34  3  9   3  HK35  3  11   4 tested, and the relatively low genetic diversity of taro in Africa generally . Nevertheless, the overall diversity of taro in Togo, and in the neigh bouring countries of Ghana and Burkina Faso (Traore, 2013) does suggest a complex history of the crop in the region, and in Africa. Among published studies of SSR diversity in taro (Table S1), no two studies have used the same meth ods to collect, maintain and test plants, and no stan dard set of primer pairs and target loci has emerged. Crucially, different sample sets differ in whether they represent initial collections created to assess diversi ty in possiblyidentical cultivars from different loca tions (as in the present study), or laterstage collec tions in which apparent duplicates have been removed. Observed diversity depends on how plants are collected, how many are collected, and how each collection is maintained over time. Taro and new cocoyam collections are constructed "populations" of clones, not random samples from freely breeding populations. For all of these reasons, we do not com pare our statistical estimates (calculated data) with those of other smallscale studies. In the near future, new techniques for largescale and lowcost DNA sequencing may allow more accurate, comprehen sive and direct comparison of genotypes in different cultivar assemblages. Already for taro, public databases contain records of thousands of SSR and SNP (single nucleotide polymorphism) loci revealed by wholegenome and transcriptome studies You et al., 2015;Helmkampf et al., 2017;Soulard et al., 2017;Wang et al., 2020), and a draft sequence for all 14 chromosomes has been published (Yin et al., 2021).
In the Togo collection, flowering occurred among accessions of both species, but fruiting and seed pro duction were not observed. Togo has a tropical savannah climate, with distinct wet and dry seasons, and annual rainfall ranging from around 800 mm to 1,600 mm (Djaman et al., 2017). In Burkina Faso, in the same general climate zone, but further Northwest, most taro is mostly grown in provinces with annual rainfall ranging from around 700 mm to 1,100 mm, near Togo (Traore, 2013). Although condi tions during the wet season in Togo (April to October) might be suitable for breeding by taro, dry and windy conditions during the winter harmattan (Ern, 1979) would be fatal for unprotected seedlings. These are very different circumstances from those in the natural range of taro, in the tropical rainforest zone of Asia and the western Pacific, where wild breeding populations are found (Matthews, 1991;Hunt et al., 2013;Matthews, 2014). Togo itself lies in a dry savannah corridor (the Dahomey Gap) flanked by tropical rainforest (the Upper and Lower Guinean Forests) where spontaneous breeding by taro may be possible. Breeding and selection of taro cultivars in these nearby forest regions might have contributed to some of the diversity found in Togo.
Different strategies are suggested here for future development of taro and new cocoyam in Togo. For taro, it will be rewarding to study the existing range of eddoe and dasheen cultivars further, and to make experimental introductions of new cultivars from outside Africa, following the example of Ouedraogo et al. (2018). Efforts will be needed to produce dis easefree stocks of existing cultivars so that fair com parisons can be made with newlyintroduced plants that are disease free. New cocoyam, a relatively modern historical introduction, has spread widely in Africa, and a lack of diversity is clear among the accessions collected in Togo. International collabora tion is needed to identify and introduce new cultivars for evaluation under local conditions. This will be dif ficult, as there are no international breeding pro grammes for the crop, and little is known about the origins and diversity of cultivated Xanthosoma species in Central and South America. Xanthosoma spp. are even more neglected as orphan crops than taro (Matthews and Ghanem, 2021).
Together, taro and new cocoyam offer a range of cultivars suitable for cultivation in wetland to dryland environments. Agriculture in Togo is predominantly rainfed and often experiences both flooding and drought (Djaman et al., 2017). In seasons and loca tions when water is abundant, the flooding tolerance of dasheen taro (Onwueme, 1999) is a positive attribute that can enhance the food security of farm ers working in riverine flood plains. When irrigation is provided, very good yields of taro can be expected in otherwise dry environments: this is shown by the success of taro as an irrigated summer crop in the eastern Mediterranean (Matthews, 2006), a region with long dry summers and relatively little annual rainfall (approx. 400 mm per year in the main agricul tural districts of Cyprus). Over the last 50 years, cli mate change has been very obvious throughout Togo, with the wet season becoming 12 months shorter (Djaman et al., 2017;GadédjissoTossou, 2018). During the period 19612001, for example, annual precipitation decreased at 80% of weather stations across the country (Djaman et al., 2017).
Under these circumstances, maintaining or expand ing the cultivation of taro and new cocoyam may come to depend on the success or otherwise of efforts to improve methods for water storage, con servation and irrigation in Togo.