Analyses of phylogenetics, starch granule morphology and consumer preference of Canna indica L. grown in Sri Lanka

Canna indica is a tuber crop which has many medicinal values. In Sri Lanka, C. indica tubers are consumed in rural areas and mainly available in street-markets of NuwaraEliya and Kandy Districts. In the present study, we assessed the phylogenetics of C. indica, starch granule morphology and consumer preference of C. indica tubers in comparison to the popular tuber crops. The phylogenetic analysis was conducted based on the sequence polymorphism at rbcL, atpB gene, trnLtrnF and trnH-psbA marker-loci with respect to the ornamental Canna spp. in Sri Lanka and the previously published sequences of Canna spp. The starch granules were isolated and observed under optical and scanning electron microscopes. The diameter and the surface area of the starch granules were measured under the optical microscope and subjected to analysis of variance. As C. indica tubers are consumed as boiled tuber pieces in Sri Lanka, the consumer preference analysis was conducted using the boiled tuber pieces C. indica, Xanthosoma sagittifolium, Manihot esculenta, Solanum tuberosum, and Ipomoea batatas. The phylogenetic tree based on rbcL marker revealed that C. indica in Sri Lanka is slightly divergent from the other Canna spp. Only the polymorphism of the atpB gene can be used to differentiate C. indica from the ornamental Canna sp. in Sri Lanka. The morphological analysis of starch granules revealed that C. indica has the biggest scallop-seashell shaped starch granules compared to other tuber species. The boiled C. indica tubers were accepted better than that of X. sagittifolium, rated equally to the tubers of S. tuberosum and M. esculenta, and rated less than I. batatas. The hardy and fibrous nature of C. indica tubers must be the major limiting factors for achieving the highest consumer preference highlighting the need of breeding for better texture in tubers.

The tubers of C. indica are a common commodity in ayurvedic herbal medical preparations (Kankanamalage et al., 2014). Although C. indica possesses significant medicinal and food values, it is currently an underutilized tuber crop in Sri Lanka (Malkanthi, 2017). The tubers of C. indica rarely appear in the country-wide market. The younger generation who generally appreciate fast food and limited array of choices, even do not know that C. indica tuber is a human food. However, C. indica is fast growing (Zhang et al., 2008) with an excellent ornamental value (Zhang et al., 2008) and recently being identified as one of the best species for the phytoremediation of the polluted water released from reverse osmosis process of the water purification units in Sri Lanka (Gunarathna et al., 2016). It is believed that C. indica contains the largest starch granules ever known (Hermann et al., 1997) thus, can be used as a candidate species to broaden our knowledge on starch science and technology.
In Sri Lanka, C. indica is grown in lowland plains (30 -200 m above sea level) of Sri Lanka without much attention. However, C. indica plants thrive well in the central highlands of Sri Lanka (500 m above mean sea level, 2000 mm annual rainfall and 23-26 0 C of mean temperature) and grow as large dense bushes where suckers keep coming from the mother plants. The tubers can be harvested for human consumption after six months of planting. There are slight variations in leaf size, leaf margin and plant height exist among the C. indica populations found in different habitats. However, when suckers from one location is planted in a different habitat, these variations get changed based on the soil fertility and moisture content. However, the genetic diversity or the exact species delimits of the C. indica have not been studied in Sri Lanka. In addition to the edible C. indica, there are ornamental Canna plants which can produce attractive flowers and foliage grown in many places of the country. The species differences between edible C. indica (i.e. Arrowroot) and the ornamental Canna spp. have not been studied to date. The extensive multidisciplinary research attempts are needed to uplift C. indica from its current underutilized status. Also, the taste and other sensory attributes of C. indica in comparison to the popular tuber crops must be studied. Therefore, the present study was conducted to assess the species delimits and phylogenetic relationships, starch granule morphology, and consumer preference of C. indica in Sri Lanka.

Phylogenetic analysis
The initial and end noises of the sequencing data obtained from four DNA markers (Table 1) were observed independently using MEGA 7 (Kumar et al., 2016). Then the identities of the trimmed sequences were confirmed independently by subjecting to a search in Basic Local Alignment Search Tool (BLAST). We created separate alignments for each marker in MEGA 7. In order to relax the family Cannaceae topology, we did not use phylogenetically distant outgroup, or a too closely related one. Thus, we used two species from family Commelinaceae as outgroups which can also be found in the Commelinids clade where family Cannaceae is located. To compare the Canna spp. Sri Lanka, the sequences generated were aligned for the marker, rbcL, with the dataset generated in Prince, (2010) ( Table 2). A model selection analysis in Jmodel test (Posada, 2008) was implemented to infer the best nucleotide substitution process of the dataset. A phylogeny was constructed in Maximum Likelihood framework in RAxML (Stamatakis, 2006). We implemented rapid+boostrap algorithm with 1000 iterations and GTRGAMMA as the nucleotide substitution model. The analysis was implemented in the CIPRES supercomputer (Miller et al., 2010). We also carried out a tree search in the Bayesian framework in MrBays (Huelsenbeck and Ronquist, 2001) in CIPRES science gateway. Two hot and cold chains of Markov Chain Monte Carlo (MCMC) running for 30 million generations were used to probe trees in the tree-space. The 10% of the initial trees were removed as burn-in, and the rest of the probed trees were used to draw the final 50% majority rule consensus tree. The Effective Sample Size (ESS) of all the priors used were checked in TRACER v1. 4 (Rambaut and Drummond, 2007). To check the intra species variability of the Sri Lankan Canna spp., UPGMA trees were constructed for the markers, using uncorrected pairwise genetic distance matrices. The gaps (INDELS) were considered as pairwise deletions and considered uniformed rates across all the states. The trees were constructed in MEGA7. Finally, all the drawn trees were modified using FigTree v1.4.3 (Rambaut, 2014).

Assessment of the starch granules
The starch granule morphology of C. indica was assessed in comparison to the starch granules of well-known tuber crops in Sri Lanka; Xanthosoma sagittifolium, Manihot esculenta, Solanum tuberosum, Ipomoea batatas purpletuber, and I. batatas yellow-tuber. The healthy tubers at the right maturity stage for harvesting were collected from all six species.

Isolation of starch
The starch was isolated from each species using the modified protocols explained in Alves et al., (1999) andDe Pater et al., (2006). The tubers were thoroughly washed, peeled and chopped into small pieces. The chopped pieces were homogenized using a solution containing 0.075% sodium bisulfate to avoid the browning. The homogenateslurry was then filtered to eliminate the impurities and fiber. Then the filtrate was allowed to settle for three hours. The supernatant was discarded and the sediment was resuspended in about five volumes of distilled water for further purification. The starch was allowed to settle for 30 minutes. The procedure of decanting the supernatant and re-suspending the sediment in water was repeated for three to four more times to make sure that the extracted starch was completely free of impurities. Finally the starch was allowed to settle for one hour followed by treating with a solution containing 0.1% NaOH. This step was carried out to reduce the effects of non-starch polysaccharides which may greatly interfere with starch isolation procedure by retaining the smaller granules in the remaining fibrous matrix. A magnetic stirrer was used to stir the viscous solution. The extracted starch was then washed with distilled water to remove the excess NaOH and allowed to dry overnight at room temperature. The resulting starch was stored at -20 °C.

Starch granule observation under optical microscopy (OM)
A few drops of starch granule suspension of each type of tuber were treated with iodine and observed under the optical microscope at ×1000 magnification. As a control, the same observations were made by using starch granule suspensions without treating with iodine. An optical microscope (Carl Zeiss Microscopy GmbH, SN 3150000610) equipped with the camera Zeiss AxioCam ICc 5 was used for the observations, and the captured photos were analyzed using the image analysis software, Zen lite 2.1. The diameter and the cross-sectional area of 20 starch granules were measured for each type of tuber assessed. The data were subjected to ANOVA procedure using the statistical package SAS 9.4 (SAS Institute, NC, Cary, USA).

Starch granule observation under scanning electron microscopy (SEM)
The freshly harvested tubers from each of the six species were used to obtain tissue samples. The tissues were subjected to the vacuum for about 15 min. The vacuumdried tissue samples were mounted on stubs with carbon tabs. The tissue samples were exposed to gold particles for staining purpose. The observations and the analysis of tissues were done by using the Zeiss scanning electron microscope (SEM) (Jeol SEM 6400, Tokyo, Japan). The images were taken using same magnification, ×1000, as in optical microscope to illustrate the arrangement of starch granules inside the cells. The same procedure was carried out to capture the images of extracted starch granules.
Here, instead of tissues, extracted starch granules of all of the tuber types were used to capture the images and, the magnification was adjusted to ×2000 for better resolution.

Assessment of the consumer preference on boiled tubers
The fresh and healthy tubers at the right maturity stage were collected from all six species. The tubers were peeled, washed thoroughly, cut into approximately 3 cm x 3 cm pieces, and boiled in water for 20 mins. The water was filtered out, and the salt was added to ascertain the generally preferred taste. A taste panel of 40 human subjects was employed representing randomly selected equal proportions of males and females in the age range of 18 -65 years. The required instructions were provided to the panelists before the assessment. Each panelist was given 40 g of boiled tuber pieces of each tuber type for tasting. The panelists were requested to taste and rank the tubers for seven sensory attributes; color, aroma, texture, bitterness, fibrous nature, hardness, and overall taste according to a three-tier scoring system. Score "1" was assigned for the least preferred level. Score "2" was assigned for medium preferred level and Score "3" was assigned for the highest preferred level. The care was taken to refresh the palates of each panelist by providing water in between the tasting each tuber dish (Leighton et al., 2010). The ranked data generated by the taste panel were subjected to association analysis using FREQ procedure in SAS.

Morphology of C. indica
Canna indica is a perennial herb that grows up to 1-3 m in height at flowering stage ( Figure 1). The newly emerged rhizome ( Figure 1A) is elliptical in shape and covered with whitish scale leaves. The interior of the newly emerged rhizome is cream-white to ivory in color (Figuers1B  and 1C). The fully-grown rhizome is irregular shaped, monopodial or sympodial, stoloniferous or tuberous, branched and also covered with scale leaves where the distal part of the scale leaves are blackish-brown, and proximal parts are whitish-yellow ( Figure 1D). However, at maturity, the external color of the rhizome is brown.
The root system of C. indica consists of an excessively branched network of fibrous roots. The roots do not go deep into the soil; however, forms a mat within 30 cm of soil depth ( Figure 1E). The roots are cylindrical and thick with numerous hairs. The color of the roots is creamy white ( Figure 1E). As C. indica has a rhizome, the roots are adventitious in nature. When rhizomes are separated from the mother plant or in case where mother plant is dead, sprouting of rhizomes happen immediately upon touching with moist soils ( Figure 1F). After two weeks, R.W.K.M. Senevirathna et al.
the shoot gets elongated up to 5 -15 cm. At this stage, no roots are apparent. In the next three to four weeks, the shoot is extending further, and roots start to grow ( Figure 1G). Within another week or so, the first leaf is unfolding and continuing to produce new leaves sequentially ( Figure 1H). The plant remains in the vegetative stage for four months and then starts flowering. A potted C. indica plant at its fully grown vegetative stage is shown in Figure 1I. The stem is a pseudo stem covered with sheathing leaves. They are erect, herbaceous and cylindrical ( Figure 1I). Appearance of the leaves is somewhat similar to the banana leaves; however, the size is less as C. indica leaves are 15 -20 cm wide and 20 -50 cm long. The leaf shape is more or less ovate. The leaf lamina is dark green with purplish brown margins. Moreover, the leaf lamina is parallel veined and has smooth and wavy margins. The leaf apex is acute ( Figures  1J and 1K). Red, solitary flowers are normally 1 cm to 2 cm long and sepals are also having the same length range.
In flowers, the reddish-brown coral tube is a prominent character. The flowers are hermaphrodite ( Figure 1L). The fruits are spiny, 2 cm to 3 cm long structures consisting of green capsules ( Figure 1L). The seeds are 5 -6 mm long, 2 -3 mm wide and white color at the initial stage, and black color at the maturity stage (Rao and Donde, 1955). A cultivated C. indica which appears as a bush at five months of age is shown in Figure 1M. Figures 1N and 1O show two C. indica plants grown in a natural habitat and an infertile land, respectively.

Phylogenetics of C. indica
The best substitution model for combined datasets of Prince, (2010) and the Canna spp. in Sri Lanka is TPM3uf+I+G (A/ C=2.100; A/G=8.260; A/T=1.000; C/G=2.100; C/T=8.260; G/T=1.000). The maxmum convergence for the MCMC tree is at 20 millon chain runs and the ESS value for each prior is above 200. Almost similer branching pattern is obtained for the trees constructed in the baysin and ML frameworks. However, the ML tree resolves the best, thus presented in this study ( Figure 2). The Canna spp. in Sri Lanka are separately clustered in the Cannaceae clade of the phylogram (Figure 2A). However, the phylogenetic relationships of the Cannaceae clade in this phylogeny are poorly resolved with few polytomic relationships. This has also been described in the Prince, (2010), where they concluded that plastid markers failed to resolve the correct phylogenetic relationships within the Cannaceae family.
The nucleotide variation is recognized in the selected loci of the chloroplast genome of C. indica and ornamental Canna spp. (Figures 2B, 2C and 2D). Out of the two coding and two non-coding markers employed in the present study, only rbcL, atpB gene and trnH-psbA can be effectively employed to identify the nucleotide polymorphism and thereby to designate the species in respective clusters within a UPGMA tree. The marker trnL-trnF is monomorphic for all the samples sequenced. The informative substitutions in the atpB gene are detected; thus an enhanced separation is obtained in the phylogeny ( Figure 2B). Accordingly, five edible Canna spp. are cladded in the same cluster. Remarkably, the point mutation (A/C) at the 2nd position in atpB gene is sufficient to ascertain the variance between the ornamental Canna spp. from the edible Canna spp. with a genetic divergence of 0.0008. However, the ornamental Canna sp. (red flower bearing) is diverged out from the clade which contains the other two ornamental Canna spp. (yellow and orange flower bearing). This might be attributed to the synonymous mutations at 1417th and 1418th positions in the atpB gene.  Figure 2C). Based on the results, it can be suggested that the usefulness of the two loci; trnH-psbA and rbcL, in discriminating the ornamental and edible Canna species is less, compared with the informativeness of atpB gene. The ornamental Canna spp. are also identified as C. iridiflora and C. jaegeriana in the published literature (Prince, 2010;Tanaka et al., 2009). However, in the present study for rbcL and trnH-psbA, the ornamental Canna spp. do not show any clear separation from C. indica (edible Arrow-root) indicating that ornamental Canna spp. could also come under C. indica. conflicting with the nomenclature provided in (Prince, 2010;Tanaka et al., 2009) (Figures 2C and 2D). The ornamental Canna spp. and C. indica are clearly separated by the polymorphism at the atpB gene ( Figure  2B) highlighting the possible speciation or sub speciation which requires further studies.

Morphological variability of starch granules
The variability of the size of the starch granules are given in the Table 3. The results revealed that the granule size and the shape are highly variable among the tuber crops studied. C. indica starch granules are significantly larger than the granules of the other popular starch crops studied for the comparison purpose. The mean diameter and the mean cross-sectional area of the starch granules are 53.72 µm of 2180.46 µm2 respectively for C. indica and they are the largest reported values for the six species studied (P<0.05). The microscopic analysis of the starch granules in the present study further verifies the past observations reporting that C. indica has large starch granules (Hermann et al., 1997). The smallest size of the starch granules is observed in I. batatas (purple) (mean diameter: 10.56 µm, cross sectional area: 91.44 µm2) (P<0.05). The appearance of starch granules under the OM and SEM are presented in Figure 3. A pronounced difference in the shape of the starch granules is observed in C. indica. The shape is like a shell of a bivalve with curve-like patterns (scallop-seashell shaped) throughout the surface when viewed under OM ( Figure  3A1; 3A2). The granules displayed a disk shape with a smooth surface under SEM ( Figure 3A3) and the similar results have previously been reported by Jane et al., (1994). However, within the tissue, the cells contain much smaller number of starch granules in C. indica in comparison to the other tuber crops studied ( Figure 3A4). According to the SEM images and the surface area measurements under OM, the highest surface area is recorded for C. indica. Several other studies also have reported the same observation (Wickramasinghe et al., 2009, Piyachomkwan et al., 2002, Hung and Morita, 2005, Cisneros et al., 2009. Moreover, the high viscosity and the ability to make a clear paste of C. indica starch make it a good candidate as a thickening agent (Andrade-Mahecha et al., 2012).    In contrast, X. sagittifolium is found to possess a significantly high number of starch granules within the tissue with irregular shapes ( (Figures 3E4; 3F4). However, M. esculenta granules possess both spherical and irregular shapes ( Figure 3C3). As shown by SEM image, the number of granules within the tissues of M. esculenta is higher than that of C. indica, however, less than that of X. sagittifolium ( Figure 3C4). Moreover, S. tuberosum displays relatively a smaller number of starch granules within the tissue, nevertheless a high number of granules compared to C. indica.

Consumer preference
The representative samples of the boiled tubers prepared for the taste panel are shown in Figure 4. The boiled tubers of C. indica are brown in color ( Figure 4A) compared to other tubers boiled. All the taste parameters assessed except bitterness are significantly associated with the type of tuber (i.e., tuber species) ( Figure 5, P<0.05). The strongest association is detected between the type of tuber and the preferred color ( Figure 5A). The highest preferred color is recorded for I. batatas (yellow) (88%) and S. tuberosum (68%). The color of C. indica, M. esculenta and I. batatas (purple) are equally preferred (15-18%). It is interesting to note that the color of C. indica is preferred more than that of X. sagittifolium which is a more popular tuber crop than C. indica. The highest preferred aroma is observed for I. batatas (yellow) (28%) followed by S. tuberosum (15%). The respondents equally rank C. indica and M. esculenta for the preferred level of aroma ( Figure 5B). The highest preferred texture is reported for the two types of I. batatas; however, the texture of C. indica is preferred more than X. sagittifolium ( Figure 5C). The bitterness felt is low for all the tuber types assessed; however, 10 % of the respondents rank the highest felt bitterness for C. indica tubers ( Figure  5D). The highest felt fibrousness is observed for C. indica (73%) followed by I. Batatas (purple) (21%) ( Figure 5E). The highest felt hardness is also reported for C. indica. The felt hardness is found be the least for S. tuberosum and I. batatas (yellow). The tubers of X. sagittifolium and M. esculenta are ranked as the hardest by 25% and 20% respondents respectively ( Figure 5F). In the ranking of tubers for overall taste, 70% and 50% of the respondents said they highly prefer the taste of I. batatas (yellow) and I. batatas (purple) respectively. Interestingly, C. indica and M. esculenta get equal ranks for the overall taste ( Figure  5G).
The association analysis based on the taste panel data provided important insights on the potentials of C. indica tubers and the avenues for the crop improvement. It is evident from the analysis that consumers prefer yellow color in boiled tubes but the color of C. indica is not entirely rejected. The color of C. indica accepted better than that of X. sagittifolium. The highest fibrousness and the hardness could be the reasons for the less preference for the tubers of C. indica. However, when considering the overall taste, C. indica gets equally ranked to the highest levels as the two of the world famous and established tuber crops, S. tuberosum and M. esculenta ( Figure 5G). Thus, improving the tuber properties; softness and less fibrousness, through breeding and selection will undoubtedly uplift C. indica from its current underutilized status.

CONCLUSIONS
The phylogenetic tree constructed based on rbcL marker revealed that C. indica is slightly divergent from the other Canna species in the world. The polymorphism of the atpB gene can be used to differentiate C. indica from ornamental Canna spp. The polymorphism of rbcL and trnH-psbA cannot be used to differentiate C. indica from the rest. The starch granules morphological analysis revealed that C. indica has the significantly large (scallop-seashell shaped) starch granules compared to Xanthosoma sagittifolium, Manihot esculenta, Solanum tuberosum, and Ipomoea batatas. The consumer preference analysis indicates that boiled C. indica tubers are accepted better than that of X. sagittifolium. The C. indica tubers rate equally to the tubers of Solanum tuberosum, and M. esculenta. The relatively hardy fibrous nature of the tubers of C. indica is the major limiting factor for achieving the highest consumer preference. The breeding programs on C. indica must be planned to achieve softer and less fibrous tubers for higher appetite.