The genus Abelmoschus
(Malvaceae) comprises several species with global or local importance as a vegetable crop. Okra (A. esculentus
(L) Moench) is an allopolyploid species with a variable chromosome number [1
] and unclear origin. It was suggested that okra originated in Africa [2
], while [3
] hypothesized Asian origin. As no truly wild A. esculentus
is known, okra is assumed to be a cultigen, whose cultivation has been documented since 1200 B.C. in Egypt and Arabia [5
]. Other Abelmoschus
vegetable crop species that can be hybridized with A. esculentus
are A. caillei
(A. Chev.) Stevels, A. moschatus
Medik, and A. manihot
(L.) Medik [7
]. A. caillei
is grown in West Africa for its edible pods [9
] and the leaves and unripe pods of A. moschatus
are consumed as a vegetable, while the roasted seeds with their sesame-like taste are used for flavoring foods and drinks [10
]. A manihot
is a popular leafy vegetable in Oceania and is grown in parts of Asia for its pharmacological value [11
The okra crop (A. esculentus
) is grown on about 2 million hectares producing almost 10 million tons of pods worldwide. The largest producer is India with about 6 million tons of fresh pods harvested, followed by Nigeria with about 2 million tons [12
]. The economic importance of okra is growing due to the rising local and international demand for fresh vegetables, but also for use as an alternative source of oil for human consumption, or as biofuel [13
] and for medicinal products [15
]. Breeding aims for okra are cultivars with improved pod quality and yield stability, resistance to abiotic and biotic stresses, and exploiting heterosis for hybrid development [16
Breeding requires access to crop biodiversity, which has been conserved in gene bank collections. The largest public sector Abelmoschus
germplasm collections are held by the Plant Genetic Resources Conservation Unit, Southern Regional Plant Introduction Station, University of Georgia, USDA-ARS (2971 accessions), the World Vegetable Center at its headquarters in Taiwan (WorldVeg, currently 1873 accessions), followed by the Agricultural Plant Genetic Resources Conservation and Research Centre, Sudan (http://seed.worldveg.org/search/passport
(accessed on 8 December 2020)). Assembly of a core collection that represents the diversity of okra conserved in gene banks in a smaller germplasm set would facilitate screening for traits of interest for breeding [17
Okra is a heat tolerant crop and therefore suitable for being cultivated during the hot season in tropical regions [18
]. In the humid tropics and in tropical monsoon regions, the hot season brings about episodes of heavy rain leading to flooding and waterlogging. Heat and heavy rain events may increase under climate change scenarios, causing additional stress for crops [19
]. Waterlogging inhibits aerobic respiration in roots, and plants respond to this stress with changes in metabolism, photosynthesis, growth, and development (reviewed by [20
]). Plants can adapt to waterlogging by developing adventitious roots or hypertrophied stem bases with lenticels and aerenchyma cells, which enhance root aeration [21
]. Few reports investigated the response of okra to water logging and came to contrasting results: [22
] reported that 30 days of flooding of two pot-grown okra accessions reduced net photosynthesis, but not shoot fresh or dry biomass in two okra accessions. In contrast, [23
] found that waterlogging reduced growth and yield of okra plants.
Measuring flooding stress responses of plants other than biomass or yield reduction at final harvest is laborious. Therefore, physiological investigations during waterlogging such as measuring net photosynthesis or water relations are generally conducted on a relatively small number of plants at a few time points, limiting the data availability for larger germplasm panels over time. Changes in plant growth over time would provide an integrated measurement for photosynthetic and metabolic processes in plants. Automatic field phenotyping devices can provide accurate plant growth data in a non-destructive manner also for large germplasm panels [24
]. Based on these data, germplasm with contrasting tolerance phenotypes can be identified for in-depth investigation of tolerance traits and selection of tolerant donor materials for breeding. The present study assessed the variation of the response to water logging in a biodiverse okra core collection using an automatized field phenotyping system and identified germplasm with maintained growth and health under flooding conditions.
2. Materials and Methods
2.1. Plant Material, Genotyping, Core Collection Selection
To ensure comparability of the fragment lengths of the microsatellite DNA fragments across all accessions, 61 accessions displaying in total 155 DNA fragments of 20 microsatellite markers were included as standards for genotyping the whole collection. The microsatellite bands were scored as present (1) and absent (0) in MS Excel and the 0/1 matrix was submitted to similarity analysis using Jaccard’s co-efficient in the Darwin package [26
]. Dendrograms were produced by the Unweighted Pair Group Method Average (UPGMA). An okra core collection was selected based on SSR genotypic data in Core Hunter software [27
] using the default Mixed Replica Algorithm optimizing the Modified Rogers’ distance (weight 0.7) and Shannon’s diversity index (weight 0.3) to define a core comprising about 20% of the entries of the entire collection.
Morphological descriptor data of 32 traits (Abe01-Abe32), including seedling, plant, inflorescence, fruit, and seed traits were collected according to http://seed.worldveg.org/download
(accessed on 8 December 2020) (Supplementary Materials Table S3
) on plants grown in experimental fields at WorldVeg, Shanhua, Tainan, Taiwan between 2012 and 2015. Diversity analysis by Principal Component Analysis (PCA) was performed with R package PCAmixdata [28
]. The percentage of principal component variance between the core collection and the whole collection was calculated by weighting the variance of each principal component by the proportion of eigenvalue. The representativeness of the core set for the morphological diversity of the whole collection was analyzed comparing results of the Principal Component Analyses between the whole and the core collection.
2.2. Phenotyping Flooding Stress Tolerance
A selection of A. esculentus genotypes of the core collection (75 accessions), 3 variants of core collection accessions, 15 breeder-preferred materials, and two commercial varieties (“Lucky Five” and “Ever Lucky”, Known You Seed, Taiwan) were submitted to flooding stress tolerance screening. Only A. esculentus accessions were included in this experiment, as A. caillei, A. manihot, and A. moschatus accessions showed different phenology, making comparison with the faster developing A. esculentus germplasm difficult. The experimental field (20 × 100 m) was flattened, laser-levelled, and evenly covered with black plastic mulch. One plot was 7 m long and 1 m large and accommodated 2 accessions with 4 plants each in a single row. The plants were spaced 60 cm from each other, the plots were separated by 1.6 m and rows by 1.3 m (center from center), except the two central rows were 2 m distant from each other. The soil pH was 8.2 and the organic matter proportion in the soil was 1.2%.
Seeds were sown on 13 July 2020 in seedling trays in a greenhouse, and seedlings were transplanted to the field on 14 August 2020. For each accession, 3 replicated blocks with 4 plants each were planted in complete randomized block design. Before transplanting, the soil in the field was mixed with 400 kg/ha complex fertilizer No.1 (N-P2O5-K2O- organic matter 20-5-10-60%) and 200 kg/ha complex fertilizer No.39 (N-P2O5-K2O- organic matter 12-18-12-50%). One day before flooding (3 September 2020), 200 kg/ha complex fertilizer No.1 was applied to the soil near the plant roots. Pesticide treatment (125 cm3/ha Abamectin 2%, 45 cm3/ha Calypso 40.4% SC, and 125 cm3/ha Alert 10% EC) was applied 14 days after transplanting on 28 August 2020. A second pesticide spray (125 cm3/ha chlorpyrifos 40.8% EC, 125 cm3/ha Pyriproxyfen 11% EC and 150 cm3 Bromopropylate 25% EC) was applied just before flooding, on 4 September 2020. During and after flooding, no pesticides or fertilizers were applied.
Cultivation before flooding was done under rain-fed conditions, as precipitation was sufficient to water the plot to field capacity. Three weeks after transplanting, the field was flooded for 9 days (4 September to 12 September) with ground water and the water level was kept at least 3 cm above the mulch surface. Subsequently, the field was allowed to drain and the plants were observed until final harvest on 28 September 2020, on day 77 after sowing/day 45 after transplanting.
From transplanting to final harvest, the plants were scanned at least two times per day with a Phenospex field scan device (Heerlen, The Netherlands) equipped with two sensor heads, each consisting of a PlantEye F500 high-resolution 3D laser dual scan unit (=2 scanners per unit) mounted on a gantry that moved automatically on rails over the field with a speed of 25 mm/s. The PlantEye sensors acquired 3D point clouds of the plants by simultaneously capturing the reflection of the near-infrared (NIR: 720–750 nm) laser line and from a multispectral flash light (RED: 620–645 nm, GREEN: 530–540 nm and BLUE 460–585 nm). Automated image analysis was performed using the PHENA analytics platform (Phenospex) and data were visualized and analyzed by HortControl 3.0 (Phenospex). Morphological plant parameters such as plant height, leaf area, digital biomass (DBM), and physiological indices such as normalized difference vegetation index (NDVI = (NIR − RED)/(NIR + RED)) and plant senescence reflectance index (PSRI = ((RED − GREEN) / NIR) were automatically obtained. DBM calculated by the Phenospex system as total leaf area * plant height was correlated to plant biomass at r2 = 0.96 (p < 0.01) and plant volume (r2 = 0.97, p < 0.01). Days to first flowering, percentage flowering, and days to harvest of the first fruits were determined visually, and the data were analyzed in R.
For the analysis of biomass and height increment per day and average NDVI and PSRI the weekly averages were taken as follows: Before flooding: Data from day −7 to 0 before flooding onset, during flooding: Data from day 3–9 after flooding onset, early recovery: Data from day 10–16 after flooding onset = from day 1 of drainage on, later recovery day 17–25 after flooding onset (from day 7 of drainage on).
The largest public sector Abelmoschus
germplasm collection comprising 2971 accessions of seven species (A. caillei, A. crinitus, A. esculentus, A. ficulneus, A. manihot, A. moschatus
, and A. tuberculatus
collected in 55 countries (https://www.genesys-pgr.org/
(accessed on 8 December 2020)) is held by USDA-ARS. The WorldVeg core collection used in this study was to be about a third of the size, and contained accessions of four species (A. caillei, A. esculentus, A. manihot
, and A. moschata
) collected from 40 countries. The USDA-ARS collection contains more accessions from Europe and Latin America than the WorldVeg collection. The representation of the species was maintained in the core collection, but the geographical diversity was reduced from 40 countries of the whole collection to 29 countries. In total, 30 accessions of the whole collection were derived from the 12 countries that were not represented in the core collection.
In spite of the reduction of the geographical diversity, the Abelmoschus
core collection developed in this study contained a large part of the variation of categorical and quantitative descriptors found in the WorldVeg okra collection. In previous studies, a core collection of 50 okra accessions drawn from 260 accessions retained 55% of the principal component variance of 10 descriptors [29
]. The present core collection maintains a similar percentage of the principal component variance from 32 descriptors. The PCA showed that some extreme phenotypes (located in the upper left quadrant in Figure 2
a) were not represented in the core collection. Nevertheless, the representation of morphological categories and quantitative descriptors in the core collection of 91 and 87%, respectively, suggested good representation of the diversity in the whole collection in the core.
Plants under flooding stress face inadequate oxygen supply of flooded plant parts and are affected by nutrient deficiencies and micronutrient toxicities [30
]. In susceptible plants, growth is stalled and shoots are wilting and leaves undergo premature senescence. Okra produces lenticels during flooding or water logging, therefore it is considered to be a waterlogging sensitive plant [23
]. However, it was also reported that it can survive water logging after ethylene priming for 15 weeks and produce fruits [23
], and suffers no reduction of growth under flooding for 30 days [22
]. Instructions for okra production, however, indicate to avoid flooding the plants during irrigation [31
]. Due to increasing production during the wet season and due to climate change, okra may become more exposed to flooding or waterlogging at many locations, therefore identifying variation in flooding sensitivity could support the development of more resilient cultivars. The biodiverse okra core collection reported here could be a good source of variation for flooding tolerance.
Tolerance to flooding implicates the ability to maintain photosynthesis and to avoid oxygen shortage in roots [32
]. Measuring photosynthetic efficiency on large germplasm panels over time is costly and practically challenging, and therefore difficult to implement for screening large germplasm panels for variation of flooding stress responses. Variation in photosynthetic efficiency translates into variation in growth rate and productivity [33
]. Biomass increment per unit of time integrates the effects of many plant processes and therefore is a representative measurement for the plant response to environmental factors. Automatized phenotyping through 3D laser scanning provides accurate measurements for DBM increment at relatively low cost and with minimum labor requirement [34
]. Multispectral data obtained in addition to the 3D scans from the Phenospex S500 plant eye can be used to determine physiological indices such as NDVI and PSRI. All these measurements were taken in a non-destructive manner over time and thus allowed the observation of the same plants during the whole experiment, before, during, and after flooding, yielding data for plant parameters over time.
NDVI is increasingly used in precision agriculture, as it correlates with plant growth, health, and often with yield [36
]. NDVI data can be obtained by various ways, ranging from remote sensing using satellites to data capture through hand-held sensors. NDVI during crop development generally follows a bell-shaped curve and increases during the vegetative phase and decreases during crop maturation [37
]. NDVI has been used to estimate the damage of flooding to corn [38
], for indirect selection for yield [39
] and to measure plant vigor and plant stress [40
], including at very early plant developmental stages [41
]. In the present study we tested variation of NDVI during okra development, before, during and after flooding. Besides a general decline of NDVI during flooding, we observed genotypic differences in the degree of this decline, leading to the hypothesis that plants with less decline are less stressed. Correlation between NDVI and the amount of biomass accumulation support this hypothesis. In parallel, PSRI was assessed. PSRI generally decreases during the growing phases of a crop, plateaus during the green phase at very low values, and rapidly increases during the senescent phase [37
]. A strong increase of PSRI during flooding compared to pre-flooding conditions indicates that the plants in average were affected by flooding. Differences in PSRI increase during flooding varied among the germplasm, suggesting variation of tolerance in the core collection. During recovery, PSRI ceased to increase in average, but remained elevated, probably because the duration of the experiment was not long enough to develop new canopy layers that completely mask the leaves that were affected by flooding stress.
Overall, screening of the core collection showed that gene bank accessions had in average higher DBM increment than the two commercial varieties included in the test set, while in variety “Ever Lucky” NDVI remained higher, and PSRI remained lower than in the average of the germplasm panel. The screening resulted in a set of genotypes with high average daily biomass increment, low decrease of NDVI, and low increase of PSRI under flooding stress, as well as genotypes that showed good recovery after the stress treatment. Interestingly, a large proportion of the candidate genotypes for flooding tolerance did not flower during the experiment. A longer vegetative phase may contribute to the maintained biomass increase in some of the genotypes that showed well maintained growth increment during flooding. Further investigations of the interactions between phenology responses to flooding are required to assess the interaction between apparent flooding tolerance on the level of DBM increment and delayed flowering. In future experiments, the candidate genotypes for flooding tolerance will be tested in a split plot design, comparing the performance of putatively tolerant and susceptible genotypes under flooded and control conditions to verify the flooding tolerance of the material and select the most tolerant genotypes for breeding. These materials will also be submitted to yield trials to assess the yield potential and pod quality of the tolerant materials.