Peruvian Journal of Agronomy 4(3): 93–103 (2020) ISSN: 2616-4477 (Versión electrónica) DOI: http://dx.doi.org/10.21704/pja.v4i3.1341 http://revistas.lamolina.edu.pe/index.php/jpagronomy/index © The authors. Published by Universidad Nacional Agraria La Molina Received for publication: 14 October 2020 Accepted for publication: 20 November 2020 Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú Colecta y caracterización morfológica de 149 accesiones de achiote (Bixa orellana L.) provenientes de siete departamentos del Perú Juan Nolasco-Chumpitaz1,2*; Paul Ccoyllo-Llacsa1; Gabriela Koc-Sanchez1; Percy Medina-Morales1 *Corresponding author: juanluisnolasco@outlook.com https://orcid.org/0000-0001-6358-5497 Abstract The aim of this study was to characterize and identify groups of achiote (Bixa orellana L.) with characteristics related to increased yield and bixin content. A total of 149 achiote accessions were collected from the departments of Loreto, San Martin, Junín, Pasco, Huánuco, Ucayali, and Cusco, in Perú. These were then evaluated using ten quantitative and three qualitative capsule and seed descriptors. Undesirable characteristics such as high spinosity and dehiscence predominated, while none of the quantitative descriptors correlated significantly with bixin content. Principal component analysis indicates that the quantitative descriptors (except for bixin content) are statistically significant, forming four clusters. Notably, one of the clusters included accessions characterized by heavy seeds, and another cluster included accessions with high number of seeds and bixin content. Keywords: achiote, annatto, bixin, morphological characterization, Perú Resumen El objetivo de este estudio fue caracterizar e identificar grupos de achiote (Bixa orellana L.) con características relacionadas con mayor rendimiento y contenido de bixina. Se recolectaron un total de 149 accesiones de achiote de los departamentos de Loreto, San Martín, Junín, Pasco, Huánuco, Ucayali y Cusco, en Perú. Luego se evaluaron utilizando diez descriptores de cápsulas y semillas cuantitativos y tres cualitativos. Predominaron características indeseables como alta espinosidad y dehiscencia, mientras que ninguno de los descriptores cuantitativos se correlacionó significativamente con el contenido de bixina. El análisis de componentes principales indica que los descriptores cuantitativos (excepto el contenido de bixina) son estadísticamente significativos, formando cuatro grupos. Notablemente, uno de los grupos incluye accesiones caracterizadas por semillas pesadas, y otro grupo incluye accesiones con alto número de semillas y contenido de bixina. Palabras clave: achiote, annatto, bixina, descriptores morfológicos, Perú. 1 Instituto Nacional de Innovación Agraria. Sub-Dirección de Productos Agrarios, Dirección de Desarrollo Tecnológico Agrario, Av. La Molina 1981 La Molina, Lima, Perú. 2 Universidad Nacional Agraria La Molina, Facultad de Agronomía. Lima, Perú. Cite this article: Nolasco-Chumpitaz, J., Ccoyllo-Llacsa, P., Koc-Sanchez, G., & Medina-Morales, P. (2020). Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú. Peruvian Journal of Agronomy, 4(3), 93–103. http://dx.doi.org/10.21704/pja.v4i3.1341 Introduction Achiote (Bixa orellana L.), a shrubby plant standing from 3 to 10 m tall, is native to the American tropics, and bears globose to ovoid fruits with colors such as red, green, yellow, or shades of these. It has historically been used as a food, medicine, and colorant (Camacaro et al., 2018; Raddatz-Mota et al., 2017; Valério et al., 2015). The use of achiote as a colorant in the food, textile, pharmaceutical, and cosmetic industries is due to the carotenoid compounds present in the seeds, especially the apocarotenoid bixin (Alcázar et al., 2017; Habibi-Najafi et al., 2018; Viuda-Martos et al., 2012). Bixin is the second most economically important natural colorant in the world and 60% of the world’s supply comes from Latin America, mainly Perú, Brazil, and Mexico (Raddatz-Mota Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú September - December 2020 94 et al., 2017; Rajagopal et al., 2016; Yolmeh et al., 2015). Moreover, Perú is recognized as one of the main achiote seed exporters in the world (Stringheta et al., 2018). According to the Peruvian Ministerio de agricultura y riego [MINAGRI] (2019), in 2018, achiote was cultivated in 11,635 ha of land. The departments with the largest areas of achiote cultivation were Cusco (6,205 ha), Pasco (4,390 ha), Huánuco (414 ha), Junín (203 ha), and Ucayali (185 ha). The national average yield was 601 kg ha−1, producing 6,988 t of grain. Other countries such as Mexico and Colombia had annual yields of 1.2 and 1.13 t ha−1, respectively, for the year 2017 (Servicio de Información Agroalimentaria y Pesquera [SIAP], 2019; Minagricultura, 2020). Currently, the bixin content of achiote seeds varies from 1% to 6%, with Brazil reporting a national average of 3.5% and a maximum of 5% in the state of Sao Paulo (Albuquerque & Meireles, 2011). Brazilian varieties such as Embrapa 37 have bixin contents that exceed 5%, which is relevant because the market requires contents greater than 4% for export (Dias et al., 2017). In Perú, promising accessions with bixin contents of 3.55% to 4.55% have been reported in the national germplasm bank (Instituto Nacional de Investigación Agraria [INIA], 2009). Aside from this report, there is no official report or census on the bixin contents of Peruvian achiote, although several companies own accessions with bixin contents lower than the minimum percentage required for export in Brazil. Despite being an important agro-industrial crop in Perú, local achiote production is characterized by low yields, with lower bixin contents compared to those of other countries. This may be partly attributed to the absence of an established variety or ecotype. Little is known about the characteristics of local achiote plants and there are also very few published studies on the characterization of Peruvian material. Such information is necessary to improve this crop genetically. The objectives of this research are to analyze and classify, using morphological descriptors, achiote accessions from seven departments in Perú in order to identify accessions with great potential to increase yield and bixin content. Methodology Germplasm collection Achiote capsules (fruits) were collected between May and November 2019. A total of 149 achiote accessions were collected from the departments of Loreto, San Martín, Junín, Pasco, Huánuco, Ucayali, and Cusco (25, 24, 25, 23, 4, 24, and 24 accessions, respectively). Each accession was provided by farmers, and was derived from plants that grew from seeds. The majority of plants were older than 3 years and less than 2.5 m tall. The location of each collection site was mapped by GPS (Fig. 1 and mean GPS data on Table 1). Prior to sampling, we verified, together with the farmer that the plant was a different type. Samples consisted of two panicles collected at the stage of commercial maturity (capsules and seeds in good quality). The panicles were transported to the laboratory of the La Molina Experimental Center of the Instituto Nacional de Innovación Agraria (INIA), where the fruits were characterized. The passport data for each sample were compiled in accordance with directive No. 001-2005-INIEA-DGIA- SUDIRGEB of INIA-Perú, specifically, within the “Norms that define the standardized use of formats for the documentation of the Data of Passport in the SUDIRGEB Ex Situ Germplasm Bank.” (Instituto Nacional de Investigación y Extensión Agraria [INIEA], 2006). Each Figure 1. Sites (red dots) where the 149 achiote accessions were collected. Department Number of accessions UTM* latitude UTM* length Elevation (m.a.s.l.) Loreto (L) 25 683192 18M 9563896 102 San Martín (SM) 24 275200 18M 9326970 708 Junín (J) 25 522614 18L 8780152 659 Pasco (P) 23 502876 18L 8864447 282 Huánuco (H) 4 474506 18L 8932754 288 Ucayali (U) 24 515248 18L 9039663 222 Cusco (C) 24 745212 18L 8609063 793 Table 1. Average locations of collection points of achiote accessions by department. *Universal Transverse Mercator J. Nolasco-Chumpitaz; P. Ccoyllo-Llacsa; G. Koc-Sanchez; P. Medina-Morales Peruvian Journal of Agronomy 4(3): 93–103 (2020) 95 of the samples was admitted into the National Achiote Germplasm Bank of the El Porvenir-San Martin Agrarian Experimental Station-INIA as new accessions. The conditions of the plants from which samples were collected varied greatly. For example, they differed in management, age (from 3 years to older than 15 years), soil, and other characteristics. The data of yield per plant (which is linked to descriptors such as number of capsules per panicle, number of panicles per plant, among others) was not considered in our analysis. The accessions were coded with the initials of the department from where they were collected followed by a correlative number (for example, code L-020 corresponds to accession number 20 collected in Loreto). Characterization of the fruits and seeds The samples were cleaned, and then dried in an oven at 40℃ for 2 days (Arce, 1999) prior to evaluating the morphological descriptors of the fruits. The seeds were extracted from each capsule and they were stored under controlled conditions (seeds were packed in trilaminate aluminum envelopes and placed in a sealed container without light, with a temperature of 2°C–8°C and relative humidity of 50%–60%) for a maximum of 3 days before determining the bixin content. Quantitative (Table 2) and qualitative (Table 3) descriptors reported by Arce (1999) were used to characterize achiote fruit and seeds. These descriptors have been used previously in similar studies (Valdez-Ojeda et al., 2008). In addition, we also used other descriptors (López et al., 2018) such as capsule thickness, weight of seeds per capsule, seed length, and seed diameter. The lengths, widths, and thicknesses of 20 randomly selected capsules were measured from each accession using a digital vernier according to the methodology of Arce (1999). Seed weight was determined with an analytical balance and the number of seeds per capsule was counted manually. The lengths and diameters of the seeds were measured using Dinocapture 2.0 (2019) software after photographing 20 randomly selected seeds from each accession using a digital microscope (Dino-Lite, New Taipei City, Taiwan). The qualitative characteristics (degree of spinosity, length of spines, and dehiscence), were scored by visual observation and compared to the graphs of Arce (1999). Determination of bixin content The bixin content of achiote seeds was determined using standard methods described by the Food and Agriculture Organization [FAO] (1982); and seeds were extracted using Peruvian standard methods for the extraction of bixin and norbixin, as reported in NTP 209.256: 1991 (Instituto de Investigación Tecnológica Industrial y de Normas Técnicas [INITEC], 1991). The bixin extraction step was performed exhaustively on the seeds using chloroform until complete discoloration. The resulting extracts were further diluted with chloroform and the absorbance’s of the diluted samples were measured at 470 nm using a UV-Vis spectrophotometer (Thermo Scientific, Genesys 10S). Absorbance values were converted to the equivalent g of bixin 100 g−1 sample, using a standard curve of known bixin (chemical purity 99.3%) concentrations (i.e., from 0.25 to 3 mg L−1). Bixin concentrations were calculated using the following equation: Y = 0.406x − 0.005 (r2 = 0.9998), where Y = bixin content in mg L−1; and X = absorbance. Organ Descriptor Unit Corresponding abbreviation used in text Capsule Length cm LC Width cm AC Thickness cm GC Length-width ratio - RLA Number of seeds per capsule - NS Seed weight per capsule g PSC Seed Weight of 100 seeds g PP100S Seed length mm LS Seed diameter mm DS Bixin content g 100g−1 BIX Table 2. Quantitative descriptors used to characterize achiote (Bixa orellana L.) accessions. Descriptor Score Description Corresponding abbreviation used in text Capsule spinosity 0 No spines EC 1 Very low number 3 Low number 7 High number 9 Very high number Spine length 1 Very short LE 3 Short 7 Long 9 Very long Capsule dehiscence 0 Non-dehiscent Deh 1 Dehiscent Table 3. Qualitative descriptors used to characterize achiote (Bixa orellana L.) fruits. Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú September - December 2020 96 Statistical analysis Data were analyzed using SAS (2020) software. Pearson coefficients and the descriptive statistics were generated using the corr function. Principal component analysis (PCA) (only quantitative descriptors) and cluster analysis (quantitative and qualitative descriptors) were performed using RStudio (2020) software. Specifically, we implemented the prcomp function within the package factoextra (Kassambara & Mundt, 2020), and the functions daisy and hclust within the cluster package (Maechler et al., 2019), to perform PCA and cluster analysis, respectively. We determined the number of clusters using the function kgs within the package maptree (White & Gramacy, 2012). PCA was performed by first standardizing the data using the scale function within R. For the analysis, we employed the criteria of Cliff and Kaiser as described by Franco & Hidalgo (2003). According to these criteria, the analysis should only consider the components that, when added together, contribute 70% or more of the variance and whose eigenvalues are greater than or equal to one. In the cluster analysis, a dendrogram based on both quantitative and qualitative descriptors was constructed. The distance matrix was based on Gower distances, which is recommended for mixed data (Franco & Hidalgo, 2003). Groupings are based on Ward’s hierarchical method, and the number of clusters was estimated using the Kelley- Gardner-Sutcliffe penalty function as described by Grum & Atieno (2007). Results Morphological variability The means of the ten quantitative characteristics measured in the 149 achiote accessions, collected in seven departments of Perú, are shown in Table 4. The results indicate a moderate level of variation as evidenced by high coefficients of variability (7.87% to 30.54%), with the highest value observed for bixin content. This is reflected by the wide range of bixin contents among the accessions, with the highest and lowest values observed for accessions L-021(3.997 g 100 gr−1) and J-061 (0.895 g 100 gr−1). Table 5 shows the frequencies of the grades of each qualitative descriptor. High and very high grades (quantity of spines) of capsule spinosity (EC) were observed in 71.2% of all observations. Furthermore, long and very long spines (LE) were observed in 67.8% of the accessions. Three accessions (P-081, P-082, and P-083) have no spines. The distribution of dehiscent and non-dehiscent fruits among the accessions is roughly equal (54.4% and 45.6%, respectively). Correlation between the descriptors The pairwise Pearson correlation coefficients between the ten quantitative descriptors are shown in Table 6. Most of the correlations (36) between descriptors were significant (p < 0.05), and only nine were not significant. Among the morphological descriptors of the capsule, significant correlations were found between capsule length (LC) and length-width ratio (RLA) (0.726), capsule width (AC) and capsule thickness (GC) (0.637), and capsule width (AC) and RLA (−0.655). This means that capsule length is directly correlated with the length/width ratio, LC* AC GC RLA NS PSC PP100S LS DS BIX Population (N = 149) Average 4.72 2.97 2.15 1.64 39.55 1.15 2.93 4.87 3.97 2.27 s.d. 0.91 0.53 0.48 0.42 8.86 0.34 0.73 0.43 0.31 0.69 C.V. 19.29 17.78 22.38 25.76 22.39 29.86 24.98 8.73 7.87 30.54 Minimum Value 2.54 1.75 1.15 0.86 18.60 0.50 1.16 3.43 3.23 0.895 Accession J-055 SM-047 J-051 U-107 U-125 U-123 U-123 U-123 L-024 J-061 Maximum Value 7.17 4.81 3.48 2.74 69.90 2.42 5.14 6.09 4.67 3.997 Accession J-072 J-061 L-002 J-060 L-008 J-072 C-128 J-072 C-129 L-021 Table 4. Statistical values of ten quantitative descriptors used to describe 149 achiote accessions and the identities of the accessions having the minimum and maximum values. s.d.: Standard deviation; C.V.: Variability coefficient (%). *LC: Capsule length (cm); AC: Capsule width (cm); GC: Capsule thickness (cm); RLA: Capsule Length-width ratio; NS: Number of seeds per capsule; PSC: Seed weight per capsule (g); PP100S: weight of 100 seeds (g); LS: Seed length (mm); DS: Seed diameter (mm); BIX: Seed bixin content (g 100g−1) Descriptor Grade Frequency Number Percentage EC* 0 3 2.0 1 11 7.4 3 29 19.5 7 60 40.3 9 46 30.9 LE 1 15 10.1 3 33 22.1 7 48 32.2 9 53 35.6 Deh 0 68 45.6 1 81 54.4 Table 5. Frequencies of the grades of the qualitative descriptors used to characterize 149 achiote accessions. EC: Capsule spinosity; LE: Spine length; Deh: Capsule dehiscence J. Nolasco-Chumpitaz; P. Ccoyllo-Llacsa; G. Koc-Sanchez; P. Medina-Morales Peruvian Journal of Agronomy 4(3): 93–103 (2020) 97 while width is negatively correlated to capsule length. RLA and GC are also negatively correlated (−0.482). Among the seed-related descriptors, the number of seeds per capsule (NS) is correlated (0.459 and 0.506) to GC and seed weight per capsule (PSC), respectively. Moreover, the weight of the seeds (PSC) is significantly correlated with LC (0.474) and weight of 100 seeds (PP100S) (0.702). Lastly, PP100S is positively correlated to LC (0.501). Thus, accessions with high seed weight per capsule tend to have longer capsules and heavier individual seeds. Capsule length is correlated with seed length (LS) (0.450), RLA (0.509), PSC (0.480), and PP100S (0.620), suggesting that accession with elongated seeds tend to have elongated capsules and heavier seeds. Furthermore, seed diameter (DS) is also positively correlated with PP100S (0.644) and LS (0.634). Bixin content (BIX) is significantly positively correlated (albeit at low coefficients) to AC, GC, and NS and negative correlated to RLA, PP100S, LS, and DS. These results indicate that, among accessions, bixin content tends to be higher with the more oval-shaped fruits and with fruits with high number of seeds (NS) per capsule. Principal component analysis The first three principal components (PC) contribute a total of 72.3% of the variance; and their corresponding eigenvalues are all greater than 1, as shown in Table 7. Along the axis of Component 1, (which contributes 39.1% to the variance), the descriptors RLA, PP100S, LC, LS, DS stand out, followed by PSC, AC, and GC. NS and BIX are less relevant. Component 2 contributes 22.3% of the morphological variance, and along this axis, the descriptors NS, PSC, AC, and GC stand out. Component 3 contributes 10.9% of the variance, and along this axis, the descriptors RLA, NS, and AC stand out. The rest of the descriptors have low discriminatory power. A graph of the first two PCs (Fig. 2) shows a somewhat positive correlation between NS, GC, and AC; these descriptors are correlated to a lesser extent with BIX. Conversely, the fruit descriptors RLA, LC, PSC, PP100S are more or less correlated to seed descriptors LS and DS due to the narrow angle between them. GC and AC are most closely related, as indicated by the smallest angle between the descriptors in this graphic. Whereas, BIX and RLA are furthest related, as indicated by the near horizontally opposite angle. Overall, the results show that LC* AC GC RLA NS PSC PP100S LS DS AC 0.012 - GC −0.063 0.637*** - RLA 0.726*** −0.655*** −0.482*** - NS 0.059 0.329*** 0.459*** −0.168* - PSC 0.474*** 0.089 0.152 0.270*** 0.506*** - PP100S 0.501*** −0.153 −0.191* 0.441*** −0.233** 0.702*** - LS 0.450*** −0.280*** −0.291*** 0.509*** −0.135 0.480*** 0.620*** - DS 0.407*** −0.213** −0.254** 0.422*** −0.323*** 0.350*** 0.644*** 0.634*** - BIX −0.076 0.166* 0.193* −0.186* 0.174* −0.037 −0.202* −0.178* −0.206* *, **, *** indicates statistical significance at p-value of 0.05, 0.01, and 0.001, respectively. Table 6. Pearson correlation coefficient matrix of the quantitative descriptors used to characterize 149 achiote accessions. *LC: Capsule length (cm); AC: Capsule width (cm); GC: Capsule thickness (cm); RLA: Capsule Length-width ratio; NS: Number of seeds per capsule; PSC: Seed weight per capsule (g); PP100S: weight of 100 seeds (g); LS: Seed length (mm); DS: Seed diameter (mm); BIX: Seed bixin content (g 100g−1) Component 1 2 3 Eigen value 3.9099 2.2292 1.0889 Variance percent 39.099 22.292 10.889 Cumulative variance percent 39.099 61.392 72.281 Descriptors Correlation coefficients LC* −0.667 0.383 −0.259 AC 0.498 0.624 0.395 GC 0.496 0.666 0.167 RLA −0.816 −0.150 −0.478 NS 0.283 0.712 −0.459 PSC −0.515 0.761 −0.070 PP100S −0.804 0.286 0.298 LS −0.799 0.139 0.119 DS −0.760 0.049 0.391 BIX 0.313 0.191 −0.372 Table 7. PCA output of the quantitative descriptors used to characterize achiote accessions. *LC: Capsule length (cm); AC: Capsule width (cm); GC: Capsule thickness (cm); RLA: Capsule Length-width ratio; NS: Number of seeds per capsule; PSC: Seed weight per capsule (g); PP100S: weight of 100 seeds (g); LS: Seed length (mm); DS: Seed diameter (mm); BIX: Seed bixin content (g 100g−1) Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú September - December 2020 98 Figure 2. Contributions of the achiote descriptors* as determined by the two principal component axes. *LC: Capsule length (cm); AC: Capsule width (cm); GC: Capsule thickness (cm); RLA: Capsule Length-width ratio; NS: Number of seeds per capsule; PSC: Seed weight per capsule (g); PP100S: weight of 100 seeds (g); LS: Seed length (mm); DS: Seed diameter (mm); BIX: Seed bixin content (g 100g−1) Figure 3. Distribution of the 149 achiote accessions along the two principal components. The different clusters (in graphic named groups) are indicated by a color and symbol (Cluster 1: red dot, Cluster 2: green triangle, Cluster 3: blue square, and Cluster 4: purple plus sign). J. Nolasco-Chumpitaz; P. Ccoyllo-Llacsa; G. Koc-Sanchez; P. Medina-Morales Peruvian Journal of Agronomy 4(3): 93–103 (2020) 99 Figure 4. Dendrogram of 149 achiote accessions generated using cluster analysis by processing 13 quantitative and qualitative morphological descriptors. Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú September - December 2020 100 accessions with higher bixin contents tend to have wide or rounded capsules and high numbers of seeds. The PCA plot of all the accessions (Fig. 3) was constructed based on the first two components. While the plot shows no clear clusters, we nevertheless identify the clusters that were generated by cluster analysis. Cluster analysis Four clusters were identified at a Gower distance of 0.75 (Fig. 4). The features of each cluster are summarized below. Cluster 1 contains 26 accessions from Loreto, San Martín, Junín, Pasco, and Ucayali departments. This group is characterized by the absence of capsule dehiscence and high spinosity (EC) and long spines (LE). The values for all descriptors (including BIX) were generally average, except for NS and LC, which were less than average. Cluster 2 contains 42 accessions from Junín, Pasco, Huánuco, Ucayali, and Cusco departments. This group is characterized by non-dehiscent capsules, low degree of capsule spinosity, and short spines. In addition, the accessions in this cluster have the lowest AC, GC, and BIX values, while having the highest LC, RLA, PSC, PP100S, LS, and DS values. Furthermore, the highest seed weights were observed in this cluster, which suggests that it has good yield potential, however, such yields will have low bixin contents. Cluster 3 is made up of 37 accessions from Loreto, San Martín, and Ucayali departments. It is characterized by dehiscent capsules with a very high degree of spinosity and very long spines. On average, this cluster has highest BIX, NS, AC, and GC values, while its RLA, PSC, PP100S, LS, and DS values are the lowest. Moreover, accessions in this cluster tend to have higher BIX values. Cluster 4 is made up of 44 accessions from Loreto, San Martin, Junín, Pasco, Huánuco, Ucayali, and Cusco. It is characterized by dehiscent capsules with high degree of spinosity and long spines. Most of the descriptors (including BIX) tend to have average values, except for PSC values, which tend to be lower than the general average. Figure 4 shows 2 major groups, which differ mainly by the non-dehiscent capsules of one group (which include clusters 1 and 2) and the dehiscent capsules of the other group (which contain clusters 3 and 4). Discussion The ranges of values observed for LC (7.17 to 2.54 cm), AC (4.81 to 1.75 cm) and GC (3.48 to 1.15 cm) (Table 4) agree to some extent with those obtained from India, Perú, and Venezuela by Akshatha et al. (2011), López et al. (2018), and Mazzani et al. (2000), respectively. These authors reported capsule dimensions ranging from 5.7 to 3.0 cm long and from 4.8 to 2.0 cm wide. We found RLA values ranging from 0.86 to 2.74 (average 1.64), which are higher than those (0.75 to 1.5) reported by Valdez-Ojeda et al. (2008) for Mexican accessions. RLA values can also be used to classify capsules as oval, elongated, or spherical (Akshatha et al., 2011). Our findings indicate that our accessions are predominantly elongated. The NS values of our accessions range from 18.6 to 69.9 seeds per fruit, which are higher than those (30 to 60 seeds per fruit) reported by other authors (Akshatha et al., 2011; Moreira et al., 2015; Rivera-Madrid et al., 2006). The average PSC of our accessions (1.15 g) is lower than that reported by López et al. (2018) (1.55 ± 0.29 g). The PP100S values of our accessions (5.14 to 1.16 g) are generally higher than the reference values of 3.8 to 1.10 g reported previously (Akshatha et al., 2011; Mantovani et al., 2013). The LS (6.09 to 3.43 mm) and DS (4.67 to 3.23 mm) values of our accessions are slightly higher than the previously reported length (0.5 ± 0.02 cm) and width (0.4 ± 0.02 cm) of seeds in Perú (López et al., 2018). Comparing our results to previous studies, the extreme values we found for various descriptors suggest that the achiote populations are diverging in terms of capsule and seed characteristics. Bixin contents in our accessions range from 0.895 (J-061) to 3.997 g 100 g−1 (L-021), which is consistent with previously reported values (0.49 to 2.65 g 100 g−1) generated by similar extraction methods (Akshatha et al., 2011; Rivera-Madrid et al., 2006; Valdez-Ojeda et al., 2008; Viuda-Martos et al., 2012). Other authors, using exhaustive methods such as supercritical fluid extraction, have obtained higher BIX values, i.e., from 4.90 to 7.58 g 100 g−1 (Albuquerque & Meireles, 2012; Rodrigues et al., 2014). Meanwhile, extractions using alkaline solvents result in BIX contents ranging from 1.66 to 5.05 g 100 g−1 (Dias et al., 2017; Mantovani et al., 2013). In both the cases of exhaustive methods (super critical fluid extraction and alkaline solvents), Brazilian achiote seeds were used. We suggest that commonly used standard methods of bixin extraction often underestimate the bixin content of the accessions. Therefore, in the future, we intend to use exhaustive methods to better extract pure bixin. In our study, only two accessions have bixin contents that are close to the 4% bixin requirement for export; L-021 and L-024 produce achiote seeds with 3.997% and 3.934% bixin, respectively. Bixin is degraded by light, high temperatures (Arce, 1999), and high environmental humidity (Biego et al., 2013). Dehiscent fruits tend to expose seeds to such unfavorable environmental conditions, generating low- quality seed (Medina et al., 2001). Furthermore, capsules with little or no spines are preferable because they make the fruit harvesting easy. The frequencies of these desirable characteristics (low spinosity and no dehiscence) in our J. Nolasco-Chumpitaz; P. Ccoyllo-Llacsa; G. Koc-Sanchez; P. Medina-Morales Peruvian Journal of Agronomy 4(3): 93–103 (2020) 101 accessions were low to medium (Table 5), indicating that selection may be an effective genetic improvement tool. The low Pearson coefficients (Table 6) indicate little correlation between the descriptors, especially for BIX. This may be related to the differences in location, climate, age, and management among the accessions. Most likely, however, this is because the accessions are biologically independent from each other. Therefore, future characterization studies of achiote should focus on accessions of the same age, planted in the same plot, and managed under controlled conditions. The PCA results (Table 7) indicate that most of the descriptors are relevant, that is, they are indeed important in describing the accessions. The exception is BIX, which is of little relevance in discriminating between accessions. This agrees with the Pearson correlation analysis and can be explained by the fact that BIX is unrelated to other traits studied. Analyzing achiote accessions using PCA and morphological descriptors such as those evaluated in the present study has been done previously (Mazzani et al., 2000). These authors obtained three main components that explained 83.8% of the variance, however, they did not evaluate bixin contents. The results of correlation analysis and PCA are somewhat concordant with the clusters generated by cluster analysis (both graphically represented in Fig. 3). For example, Cluster 3 includes accessions with high values of NS, AC, and BIX. in Clusters 2 and 3 include accessions with high values in descriptors related to yield (PP100S and PSC) and quality (NS and BIX), respectively. Therefore, these clusters may identify potential candidate accessions for subsequent genetic improvement work. The results of cluster analysis show that the departmental origin of the accessions is not related to their clustering, which indicates that the achiote plants in each department are highly diverse, based on the descriptors used in this study. Future work should include establishing the progeny of the accessions used in this study in a germplasm bank to monitor their morphological descriptors in order to confirm the stabilities of these characteristics and the clusters obtained in this study through time. We have also identified accessions with potential for genetically improving this crop; these accessions with the highest bixin contents (3.997 to 3.761 g 100 g−1 bixin, top 5th percentile) are L-021, L-024, C-130, L-27, and L-22. Conclusion We characterized 149 achiote (Bixa orellana L.) accessions from the departments of Loreto, San Martin, Junín, Pasco, Huánuco, Ucayali, and Cusco. Various fruit and seed traits are correlated. A few accessions have high values for characteristics related to crop yield and bixin content such accessions may have potentially important roles to play both commercially and as materials for future achiote breeding programs. Acknowledgements To the staff of the INIA Agrarian Experimental Stations, for their invaluable support in the collections, as well as to the Programa Nacional de Innovación Agraria (PNIA) for financing through the PNIA 187_PI Project. References Akshatha, V., Giridhar, P., & Ravishankar, G.A. (2011). Morphological diversity in Bixa orellana L. and variations in annatto pigment yield. Journal of Horticultural Science and Biotechnology, 86(4), 319–324. https://doi.org/10.1080/14620316.2011.1 1512767 Albuquerque, C.L.C., & Meireles, M.A.A. (2011). Trends in annato agroindustry: bixin processing technologies and market. Recent Patents on Engineering, 5(2), 94–102. https://doi. org/10.2174/187221211796320738 Albuquerque, C.L.C., & Meireles, M.A.A. (2012). Defatting of annatto seeds using supercritical carbon dioxide as a pretreatment for the production of bixin: experimental, modeling and economic evaluation of the process. Journal of Supercritical Fluids, 66, 86–95. DOI: https://doi.org/10.1016/j. supflu.2012.01.004 Alcázar-Alay, S.C., Osorio-Tobón, J.F., Forster-Carneiro, T., & Meireles, M.A.A. (2017). Obtaining bixin from semi-defatted annatto seeds by a mechanical method and solvent extraction: process integration and economic evaluation. Food Research International, 99(1), 393–402. https://doi. org/10.1016/j.foodres.2017.05.032 Arce, J. (1999). El achiote (Bixa orellana L.): Cultivo Promisorio Para El Trópico. Costa Rica. Editorial Escuela de Agricultura del Trópico Húmedo (EARTH). Biego, G., Yao, K., Koffi, K., Ezoua, P. & Kouadio, L. (2013). Influence of the storage conditions on moisture and bixin levels in the seeds of Bixa orellana L. African Journal of Agricultural Research, 8, 5342–5347. https://academicjournals. org/journal/AJAR/article-stat/055A8E241683 Camacaro, J., Goméz, J., Jiménez, M., Vega, C., & Manganiello, L. (2018). Un colorante liposoluble de semillas de Onoto (Bixa Orellana L.) como insumo para la industria alimentaria. Revista Collection and morphological characterization of 149 accessions of achiote (Bixa orellana L.) from seven departments in Perú September - December 2020 102 INGENERÍA UC, 25, 2. https://www.redalyc.org/ jatsRepo/707/70757669017/html/index.html Dias, N.O., Rebouças, T.N., São José, A.R., & Amaral, C.L. (2017). Morpho-agronomic characterization and estimates of genetic parameters in annatto plant. Horticultura Brasileira, 35(2), 242–246. https://doi.org/10.1590/s0102-053620170214 Dinocapture (2019). DinoCapture 2.0 ver. 1.4.5. Dino-Lite Digital microscopes. Food and Agriculture Organization. (1982). Food and nutrition paper 25: JECFA specifications for identity and purity of buffering agents, salts; emulsifiers, thickening agents, stabilizers; flavouring agents, food colours, sweetening agents and miscellaneous food additives. Rome: FAO Franco, T., & Hidalgo, R. (2003). Análisis Estadístico de Datos de Caracterización Morfológica de Recursos Fitogenéticos. Boletín técnico no. 8. Colombia. Instituto Internacional de Recursos Fitogenéticos (IPGRI). https://www.bioversityinternational.org/ fileadmin/_migrated/uploads/tx_news/Análisis_ e s t a d í s t i c o _ d e _ d a t o s _ d e _ c a r a c t e r i z a c i ó n _ morfológica_de_recursos_fitogenéticos_894.pdf Grum, M., & Atieno, F. (2007). Statistical analysis for plant genetic resources: clustering and indices in R made simple. Italy. Bioversity International. https://www. bioversityinternational.org/fileadmin/_migrated/ uploads/tx_news/Statistical_analysis_for_plant_ genetic_resources_1283.pdf Habibi-Najafi, M.B., Fatemizadeh, S.S., Boroojerdi, S.R., Hosseini, F., & Karazhyan, R. (2018). In vitro evaluation of antimold activity of annatto natural dye and its effects on microbial, physicochemical, and sensory properties of bread. Journal of Food Protection, 81(10),1598–1604. https://doi. org/10.4315/0362-028X.JFP-17-533 Instituto de Investigación Tecnológica Industrial y de Normas Técnicas. (1991). Norma técnica peruana 209.256:1991: Achiote y sus derivados. Determinación de Bixina y norbixina. Perú. Instituto Nacional de Investigación Agraria. (2009). Accesiones promisorias Banco de germoplasma de la SUDIRGEB (Sub dirección de recursos genéticos y biotecnología)–INIA. Lima Perú. https:// repositorio.inia.gob.pe/ Instituto Nacional de Investigación y Extensión Agraria. (2006). Directiva 03-05: Normas de procedimiento para la transferencia de germoplasma conservado por la SUDIRGEB en las estaciones experimentales agrarias. Lima Perú. Kassambara, A., & Mundt, F. (2020). factoextra: extract and visualize the results of multivariate data analyses. R package version 1.0.7. https://CRAN.R-project.org/ package=factoextra López, S.E., Caicedo, M., Gil, A., López, A. & Pazos, A.E. (2018). Morphometry of fruit and seed of Bixa orellana L. “achiote.” SCIÉNDO, 21(2), 213–216. http://dx.doi.org/10.17268/sciendo.2018.022 Maechler, M., Rousseeuw, P., Struyf, A., Hubert, M., & Hornik, K. (2019). Cluster: cluster analysis basics and extensions. R package version 2.1.0. Mantovani, N., Grando, M., Xavier, A., & Otoni, W. (2013). Avaliação de genótipos de urucum (Bixa orellana L.) por meio da caracterização morfológica de frutos, produtividade de sementes e teor de bixina. Ciência Florestal, 23(2), 344–362. https://doi. org/10.5902/198050989281 Mazzani, E., Marín, C., & Segovia, V. (2000). Estudio de la variabilidad existente en la colección de onoto (Bixa Orellana L.) del CENIAP; FONAIAP; Venezuela. Revista Facultad Nacional De Agronomia Medellin, 17, 492–504. https://produccioncientificaluz.org/ index.php/agronomia/article/view/26377 Medina, A., Michelangeli, C., Ramis, C., & Díaz, A. (2001). Caracterización morfológica de frutos de onoto (Bixa orellana L.) y su correspondencia con patrones de proteinas e isoenzimas. Acta Científica Venezolana, 52, 14–23. Minagricultura. (2020). Agronet: Área, Producción y Rendimiento Nacional por Cultivo. Gobierno de Colombia. https://www.agronet.gov.co/estadistica/ Paginas/home.aspx?cod=1 Ministerio de Agricultura y Riego. (2019). Anuario de Produccion Agricola 2018. Perú. http://siea. minagri.gob.pe/portal/ Moreira, P.A., Lins, J., Dequigiovanni, G., Veasey, E.A., & Clement, C.R. (2015). The domestication of annatto (Bixa orellana) from bixa urucurana in Amazonia. Economic Botany, 69(2), 127–135 (2015). https:// doi.org/10.1007/s12231-015-9304-0 Raddatz-Mota, D., Pérez-Flores, L.J., Carrari, F., Mendoza-Espinoza, J., Díaz, F., Pinzón-López, L. Godoy-Hernández, G., & Rivera-Cabrera, F. (2017). Achiote (Bixa orellana L.): a natural source of pigment and vitamin E. Journal of Food Science and Technology, 54(6), 1729–1741. https://doi. org/10.1007/s13197-017-2579-7 Rajagopal, P.L., Sreejith, K.R., & Premaletha, K. (2016). Natural colorants as safe additives: a review. World wide Journal of Multidisciplinary Research J. Nolasco-Chumpitaz; P. Ccoyllo-Llacsa; G. Koc-Sanchez; P. Medina-Morales Peruvian Journal of Agronomy 4(3): 93–103 (2020) 103 and Development, 2, 28. http://wwjmrd.com/ upload/1507790072.pdf Rivera-Madrid, R., Escobedo-GM, R.M., Balam-Galera, E., Vera-Ku, M., & Harries, H. (2006). Preliminary studies toward genetic improvement of annatto (Bixa orellana L.). Scientia Horticulturae, 109(2), 165– 172. https://doi.org/10.1016/j.scienta.2006.03.011 Rodrigues, L.M., Alcázar-Alay, S.C., Petenate, A.J., & Meireles, M.A.A. (2014). Bixin extraction from defatted annatto seeds. Comptes Rendus Chimie, 17(3), 268–283. https://doi.org/10.1016/j. crci.2013.10.010 RStudio Team (2020). RStudio: integrated development environment for R. RStudio, PBC, Boston, MA, USA. http://www.rstudio.com SAS Institute (2020). SAS University Edition. SAS Institute Inc. North Carolina, United States of America. Servicio de Información Agroalimentaria y Pesquera. (2019). Anuario Estadístico de la Producción Agrícola. Mexico. https://nube.siap.gob.mx/ cierreagricola/ Stringheta, P.C., Silva, P.I., & Costa, A.G.V. (2018). Annatto/Urucum—Bixa orellana. Exotic Fruits, 23–30. https://doi.org/10.1016/B978-0-12-803138- 4.00006-X Valdez-Ojeda, R., Hernández-Stefanoni, J.L., Aguilar- Espinosa, M., Rivera-Madrid, R., Ortiz, R., & Quiros, C. F. (2008). Assessing morphological and genetic variation in Annatto (Bixa orellana L.) by sequence-related amplified polymorphism and cluster analysis. HortScience, 43(7), 2013-2017. https://doi.org/10.21273/HORTSCI.43.7.2013 Valério, M.A., Ramos, M.I.L., Braga Neto, J.A., & Macedo, M.L.R. (2015). Annatto seed residue (Bixa orellana L.): nutritional quality. Food Science and Technology, 35(2), 326–330. https://doi. org/10.1590/1678-457X.6539 Viuda-Martos, M., Ciro-Gómez, G.L., Ruiz-Navajas, Y., Zapata-Montoya, J.E., Sendra, E., Pérez-Álvarez, J.A., & Fernández-López, J. (2012). Antioxidant activity of annatto extracts. Journal of Food Safety, 32(4), 399–406. https://doi.org/10.1111/j.1745- 4565.2012.00393.x White, D., & Gramacy, R. (2012). Maptree: mapping, pruning, and graphing tree models. R package version 1.4-7. https://CRAN.R-project.org/ package=maptree Yolmeh, M., Habibi-Najafi, M.B., Shakouri, S., & Hosseini, F. (2015). Comparing antibacterial and antioxidant activity of annatto dye extracted by conventional and ultrasound-assisted methods. Zahedan Journal of Research in Medical Sciences, 17(7). https://doi. org/10.17795/zjrms1020