A source for leaf architecture of different plant species/genus?

I'm currently working on extracting some features of the leaf architecture based on their images (like L:W ratio, laminar shape,… ). I use the Manual of Leaf Architecture as the reference. It is really great in mathematically describing the features and how to calculate them.

But, the problem is I am not able to find a source (a book, or a data set) to have this architecture of different leaves (plants). For example, whether the Acer genus has leaves with a certain (or a range of) L:W ratio or the laminar shape for this species is mainly ovate,…

Does anybody know any source of this kind?

Quantitative descriptions of leaf shape used as diagnostics are hard to come by. There are numerous qualitative descriptions (lyrate, cordate, acicular, etc.), and I think this fits within the example you give that "the laminar shape for this species is mainly ovate." But actual quantitative ranges as you mention (e.g., that the L:W ratio of Acer lies within a certain range) are much more difficult to come by. Also, there are ways of quantifying leaves using all the shape variance present in their outlines, and I describe some of these methods below. By using quantitative methods such as these, you can avoid the trap of a subjective criteria that arbitrarily distinguishes leaf morphs by a particular feature. Plus, you won't be biased by using a particular shape attribute, as you will be measuring the entirety of shape. Use of these various methods though depends on the species you are studying.

Oreotrephes I think is on track saying that, if possible, you should build your own diagnostic criteria for whatever system you are working in. You mentioned certain "ranges" that can diagnose a species… this is a critical question. Measuring intra-genotype/individual variance in shape for your system is critical. Which leaf to measure for a species?… i.e., shoot position, developmental age of the leaf, if working with compound leaves, the placement of the leaflet along the proximal distal axis of the leaf, what about heteroblasty (changes in leaf form at different nodes), leaves from lateral vs. main branches, environmental effects on leaves (e.g. sun vs. shade leaves)?

For tomato, I measure shape quantitatively using Elliptical Fourier Descriptors, which reduces the shape of an object to a series approximation and explains variance with a Principal Component Analysis. I use the resulting Principal Component values of leaves as a trait and explain shape quantitatively that way. A very useful program to perform Elliptical Fourier Descriptors is described in Iwata & Ukai 2002 and can be downloaded here. In Chitwood et al 2012, I describe how to perform shape analysis on tomato leaves, separating components of shape based on genotype (e.g., by species), by position in the leaf series (heteroblasty), the ontogentic age of the leaf, and placement of leaflets along the proximal-distal axis.

Other easy ways to measure shape: Circularity, solidity, L:W ratio (or aspect ratio) as you mentioned. All very easy to measure in ImageJ!

You mentioned Acer leaves, which piqued my interest because I study differences in the shape of grape leaves. But like grape, many Acer leaves are palmately lobed. If every one of your samples has certain features (e.g. 5 lobes and 5 sinuses as in Acer, including petiolar sinus) then you can perform a landmark analysis using a Generalized Procrustes Analysis (The "Shapes" package in R is useful for this analysis).

Using global methods to describe shape, that is, methods to quantify the totality of shape variance, are critical to distinguish leaves. It may very well be that two leaves are so similar from different species that they are not statistically separable by shape! Be creative in the ways you measure shape, and if need be, create your own system for doing so!

• Chitwood DHD, Headland LRL, Kumar RR, Peng JJ, Maloof JNJ, Sinha NRN. 2012. The developmental trajectory of leaflet morphology in wild tomato species. Annu Rev Plant Physiol 158: 1230-1240.

• Iwata H, Ukai Y. 2002. SHAPE: a computer program package for quantitative evaluation of biological shapes based on elliptic Fourier descriptors. Journal of Heredity 93: 384-385.

For qualitative features, you could try essentially any flora. Most of these will have descriptions like this one in Flora of China

1. Acer… Leaves mostly simple and palmately lobed or at least palmately veined, in a few species pinnately veined and entire or toothed, or pinnately or palmately 3-5-foliolate.

The problem with using florae is that a) they're limited in geographical scope and b) they're usually just blocks of text (not databased or usually even XML tagged). However, a big flora like FoC will have plenty of genera for you to work with, and it should be easy to manually look up the information you want for a few genera, and possible for you to automate the information retrieval with online florae.

Florae can be a pain to use for non-botanists, but I think the FoC site is pretty explanatory - you can navigate from the family list to any family, and then down at the bottom will be a link called "list of lower taxa" (or just a list, if it's a small family) where you'll seelinks to each of the genera in the family. Then you'll see the genus description page, for instance, for Acer as quoted above.

Genetical genomics of Populus leaf shape variation

Leaf morphology varies extensively among plant species and is under strong genetic control. Mutagenic screens in model systems have identified genes and established molecular mechanisms regulating leaf initiation, development, and shape. However, it is not known whether this diversity across plant species is related to naturally occurring variation at these genes. Quantitative trait locus (QTL) analysis has revealed a polygenic control for leaf shape variation in different species suggesting that loci discovered by mutagenesis may only explain part of the naturally occurring variation in leaf shape. Here we undertook a genetical genomics study in a poplar intersectional pseudo-backcross pedigree to identify genetic factors controlling leaf shape. The approach combined QTL discovery in a genetic linkage map anchored to the Populus trichocarpa reference genome sequence and transcriptome analysis.

Results

A major QTL for leaf lamina width and length:width ratio was identified in multiple experiments that confirmed its stability. A transcriptome analysis of expanding leaf tissue contrasted gene expression between individuals with alternative QTL alleles, and identified an ADP-ribosylation factor (ARF) GTPase (PtARF1) as a candidate gene for regulating leaf morphology in this pedigree. ARF GTPases are critical elements in the vesicular trafficking machinery. Disruption of the vesicular trafficking function of ARF by the pharmacological agent Brefeldin A (BFA) altered leaf lateral growth in the narrow-leaf P. trichocarpa suggesting a molecular mechanism of leaf shape determination. Inhibition of the vesicular trafficking processes by BFA interferes with cycling of PIN proteins and causes their accumulation in intercellular compartments abolishing polar localization and disrupting normal auxin flux with potential effects on leaf expansion.

Conclusions

In other model systems, ARF proteins have been shown to control the localization of auxin efflux carriers, which function to establish auxin gradients and apical-basal cell polarity in developing plant organs. Our results support a model where PtARF1 transcript abundance changes the dynamics of endocytosis-mediated PIN localization in leaf cells, thus affecting lateral auxin flux and subsequently lamina leaf expansion. This suggests that evolution of differential cellular polarity plays a significant role in leaf morphological variation observed in subgenera of genus Populus.

Author summary

Plant identification is not exclusively the job of botanists and plant ecologists. It is required or useful for large parts of society, from professionals (such as landscape architects, foresters, farmers, conservationists, and biologists) to the general public (like ecotourists, hikers, and nature lovers). But the identification of plants by conventional means is difficult, time consuming, and (due to the use of specific botanical terms) frustrating for novices. This creates a hard-to-overcome hurdle for novices interested in acquiring species knowledge. In recent years, computer science research, especially image processing and pattern recognition techniques, have been introduced into plant taxonomy to eventually make up for the deficiency in people's identification abilities. We review the technical status quo on computer vision approaches for plant species identification, highlight the main research challenges to overcome in providing applicable tools, and conclude with a discussion of open and future research thrusts.

Citation: Wäldchen J, Rzanny M, Seeland M, Mäder P (2018) Automated plant species identification—Trends and future directions. PLoS Comput Biol 14(4): e1005993. https://doi.org/10.1371/journal.pcbi.1005993

Editor: Alexander Bucksch, University of Georgia Warnell School of Forestry and Natural Resources, UNITED STATES

Published: April 5, 2018

Copyright: © 2018 Wäldchen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: We are funded by the German Ministry of Education and Research (BMBF) grants: 01LC1319A and 01LC1319B (https://www.bmbf.de/) the German Federal Ministry for the Environment, Nature Conservation, Building and Nuclear Safety (BMUB) grant: 3514 685C19 (https://www.bmub.bund.de/) and the Stiftung Naturschutz Thüringen (SNT) grant: SNT-082-248-03/2014 (http://www.stiftung-naturschutz-thueringen.de/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

23 Species of Acacia Trees and Shrubs

Acacia trees and shrubs come from the Acacia genus, Fabaceae (legume) family, and Mimosoideae subfamily. The majority of the species are found in Australia, but some acacia species are found in Africa, Europe, Asia, and North and South America. They are generally long-lived and fast-growing plants, often with deep roots that enable them to thrive under dry, drought conditions. Acacias have a variety of landscape uses. Most have clusters of flowers that are yellow or cream in color..

What appears to be leaves on some acacia trees are not leaves at all—they are modified petioles, the parts of the stem that attach the leaves to the branch. When the petioles form in this manner, they are called phyllodes. The plant may start out with real leaves that change to phyllodes as it matures. Other species have a modified stem called a cladode. On the species that do have true leaves, the leaves are pinnately compound—consisting of rows of leaflets around a central stem.

Warning

Some species of Acacia include a psychoactive alkaloid in the leaves, seed pods, flowers, or stems. The psychoactive agent, known as DMT (dimethyltryptamine), is a powerful but short-lived hallucinogen that has been used for spiritual purposes by indigenous peoples. Acacia acinacea, Acacia acuminata ssp. acuminata, Acacia burkittii, and Acacia adunca are all species known to contain this psychoactive substance, though none of these are common landscape plants. Accidental ingestion to a degree that produces psychoactive effects is very rare, but it has been suggested that you should use caution not to breathe the smoke when burning brush that that contains acacia plants. Some species also bear sharply modified stems or thorns which can be useful for preventing access in certain locations. If you choose a thorny variety be sure to place it away from high traffic areas.

Discussion

Stomatal study is one of the important research areas in current rice science. Changes in stomatal features can alter the carbon and water flux of this crop and may contribute significantly to making climate resilient rice. Because stomata are the major organ controlling water and gaseous exchange in rice plants, it can be a potential target for modification in order to increase and sustain rice yield with minimal water input (Müller et al. 2018 Franks et al. 2015 Flexas 2016 Zhang et al. 2012). Being a positive regulator of photosynthesis (Kusumi et al. 2012), manipulation of stomatal traits, especially the number and size, could pave new ways to break the rice yield barrier especially in areas with limited water availability (Oshumi et al. 2007 Xu et al. 2009).

Extensive Stomatal Diversity Is Present in Oryza Species

A combination of domesticated and non-domesticated rice species (Table  1 ) allowed us to explore stomatal diversity in a whole range of primary (close to cultivated rice), secondary (relatively distantly related) and tertiary (fairly distantly related) genetic pools of the genus Oryza (Vaughan et al. 2003 Sweeney and McCouch 2007). A significant amount of natural diversity in stomatal traits is documented in different Oryza species (Fig.  5 , ​ ,6, 6 , ​ ,7) 7 ) and among the different Oryza complexes (Tables S1, S2, S3, S4, S5, S6). Substantial diversity in the stomatal structure has been noticed in Oryza family from the very small stomata of O. nivara, O. meridionalis, O. granulata and O. meyeriana to the relatively very big stomata of O. grandiglumis, O. ridleyi, O longiglumis and O. coarctata (Fig.  2 , ​ ,3, 3 , ​ ,4). 4 ). Notably, genetically close species do not necessarily have a similar kind of stomatal structure, which shows that the stomatal variation in rice family does not follow a robust phylogenetic trend and might be driven by domestication or environmental factors. This is further supported by very low phylogenetic signals in the stomatal trait values (Table S9).

It is interesting to note that the evolutionary dynamics of the number and size of stomata in the Oryza family vary differently on different sides of the leaf and in different genetic groups or complexes. In general, there is a greater variation in stomatal number on the adaxial side but, greater variation in size on the abaxial side (Table S2). The Oryza species that belong to the Sativa complex or the AA genome, are more diverse in their stomatal number (SDab =�.42%, SDad =�.96%), whereas, the species of the Officinalis complex are more diverse in their stomatal size i.e., length, width and area (example, SCAab =�.8%, SCAad =�.6%). Less stomatal diversity estimated in Meyeriana and Ridleyi complexes is likely due to the limited number of members in these two complexes, although they contain unique stomatal features as evidenced from Fig.  4 . It was quite expected to see the differences in the stomatal structure in the species of distant rice genomes. But interestingly, there is extensive structural diversity present even among the species of most recently evolved Sativa complex (Table S1, Table S2), where the two cultivated rice species O. sativa and O. glaberrima belong to. This clearly suggests a possibility to re-introduce favorable stomatal traits into the cultivated varieties to improve its physiology.

Rice Has Increased gmax through an Increased Stomatal Number

Superimposing stomatal features and phylogeny of Oryza species, suggests no doubt that speciation has led to an increased stomatal number per unit area and a smaller stomata (Fig. S1). This is accompanied by a steady increase in both the abaxial and adaxial conductance (gmax) (Fig. S2). These two observations can easily relate that the increased stomatal conductance of rice is actually due to its increasing stomatal number over time, instead of increasing the stomatal size. This is in accordance with some of the earlier reports that showed that stomata in domesticated rice are generally numerous and smaller compared to other crops (Chen et al. 1990 Teare et al. 1971) with indica species having more stomata than japonica (Maruyama and Tajima 1990).

Structure-Function Correlations Reveal Opportunities for Adjusting the Stomatal Size to Reduce Stomatal Water Loss in Rice

Stomatal development in rice follows a typical one cell spacing pattern as in other plants (Sachs 1991 Geisler et al. 2000). Our correlation results display a negative relation between stomatal density and size (Fig.  8 ), associated with the reduction of the neighboring epidermal cell width. It is known that gas exchange through these stomata is highly affected by its architecture, the density and size, specially the pore depth (Fanourakis et al. 2015b), which limits the length of the diffusion pathway of gases, exchanged through these micropores (Franks and Farquhar 2001 Franks and Farquhar 2007). Our results give clear evidence that both the increased stomatal conductance and ∆ 13 C in rice is caused by increasing stomatal density as these are positively related to each other. Previous reports argue that smaller stomata would respond faster and more efficiently in the uptake of CO2 inside the leaf, and thus will increase the photosynthesis and WUE (Lawson et al. 2014 Giday et al. 2013). But, in contrast, in our results, ∆ 13 C consistently shows a significant negative relationship with stomatal and guard cell size at P <𠂐.05 (Fig.  8 , Fig.  10 b, Fig. S3). This suggests that the larger stomata will have less discrimination and thus theoretically will improve the WUEi in rice. Ideally, either an increase in stomatal number or size should lead to an increase in conductance (Parkhurst 1994), which is indeed reflected by the positive relation of stomatal number and size with gmax (Fig.  8 ). However, considering the negative relation of GC size and ∆ 13 C, we propose, increasing gmax through increased stomatal size would be a better approach to retain minimum water loss while maintaining substantial CO2 intake in rice, under well-watered conditions. Possibly, higher gmax through increased stomatal size would increase photosynthesis in a much better way without losing more water through transpiration. Some recent reports using genome editing also supports that reduced stomatal density actually helps in increasing the water use efficiency in rice (Caine et al. 2018). The mechanism is of course complicated and the hypothesis needs to be tested by actual field experiments. In addition, the non-significant relationship of SC area with Δ 13 C in non-corrected correlation r, suggests that there are other drivers as well controlling this physiological trait in rice. However, our result still suggests that there is scope for improving the carbon-water balance in elite rice varieties by introducing larger stomata. Several species in the Officinalis and Ridleyi complex possess larger stomata although, it is difficult to use these species in a breeding program due to their reproductive barrier to cultivated rice. Therefore, Oryza barthii, that belongs to the Sativa complex, with a larger GC size of 21.8 ×𠂖.3 μm 2 , might be the most appropriate for this purpose, as this species can be used in crossing program with more success of hybridization. Unfortunately, there is no gene directly reported for controlling stomatal size that can be edited immediately. Epidermal Patterning Factor (EPF) is the reported candidate gene, which controls the formation of different epidermal cells and thus also the stomata. EPF controls stomatal number negatively and has pleotropic effect on stomatal size. Plants overexpressed with this gene have reduced stomatal number, increased stomatal size and less adversely affected by reduced water availability (Doheny-Adams et al. 2012). Another aspect of stomatal domestication, the stomatal adaxialization (Milla et al. 2013), is not very prominent in species of Sativa complex. Although collectively the Sativa complex shows increased ad/ab conductance (Table S5), altogether the 23 species show a steady increase in total stomatal conductance through maintaining almost a consistent improvement in both abaxial and adaxial gmax in Oryza (Fig. S2). Hence, improving gas exchange properties through adjustment of adaxial stomatal features in rice is another viable prospect in breeding programs.

Discussion

Acclimation to high light

Variation in leaf mesophyll anatomy with growth irradiance was most pronounced in A. rufinerve, a light-demanding species (Figs 1 & 2). The notable mesophyll thickness, Smes and Sc in A. rufinerve in full sun were mainly due to thick palisade tissue (Fig. 1). Sun-type leaves typically had thick palisade tissue ( Lambers, Chapin, & Pons 1998 ) and high Sc ( Syvertsen et al. 1995 ). Notable palisade thickness along with Sc may enhance the capture of photons on an area-basis ( Evans 1999 Nobel 1999 ).

The high leaf assimilation rate in A. rufinerve was expected from high Sc, because the latter enhances the diffusion of CO2 from the surface of mesophyll cells to chloroplasts, resulting in the positive correlation between Sc and Amax/area ( Araus et al. 1986 Evans et al. 1994 Syvertsen et al. 1995 Evans 1998 ). The higher Rubisco content in A. rufinerve than in A. palmatum for full-sun-grown plants (Fig. 3b) suggests a high photosynthetic potential in A. rufinerve compared with A. palmatum. In fact, when compared among the light levels, Amax/area was positively correlated to Sc and Rubisco in A. rufinerve and A. mono (Fig. 6a & b). However, the Amax/area in A. rufinerve was similar to those in other species in full sun (Fig. 4a).

Experimental conditions in the present study (e.g. sizes of the pots) did not cause the reduction in Amax/area in A. rufinerve. The value of Amax/area in A. rufinerve grown at full sun was comparable with those of 10-year-old A. rufinerve grown at an open site (7·6 µmol m −2 s −1 A. Ishida, personal comm.) and seedling grown at a light regime of gap centre ( Lei & Lechowicz 1998 ). In addition, Amax/area values in A. mono and A. palmatum grown at full sun were consistent with those obtained in previous studies ( Koike 1988 Kitao et al. 2000 ).

The possible factors that limit Amax in A. rufinerve relative to the other species are CO2 diffusion processes from the atmosphere to carboxylation sites. Despite high photosynthetic potential, the value of gs in A. rufinerve was no greater than that of the other species (Fig. 4b). This may be partly because stomatal density did not increase at full sun in A. rufinerve (Table 1), although stomatal density generally increases in high light compared to low light ( Boardman 1977 ). In addition, CO2 conductance inside the leaf (gi) in A. rufinerve was also no greater than in the other leaf species at full sun (Fig. 4c). As a result, CO2 partial pressure at the carboxylation site (Cc) was lower in A. rufinerve (130 µmol mol −1 ) than in A. palmatum (150 µmol mol −1 ) and A. mono (190 µmol mol −1 ), when measured at high irradiance (Fig. 4d). This low Cc in A. rufinerve minimizes the increase of Amax/area in A. rufinerve grown at full sun. Our present result supports the result by Evans (1999 ) that lower assimilation rate in sclerophyllous leaves was accompanied by lower Cc compared to mesophyllous leaves when measured under high irradiance.

Low gs in A. rufinerve at full sun (Fig. 4b) causes a marked increase in intrinsic WUE at full sun (Fig. 5a), suggesting that A. rufinerve is capable of accumulating carbon with less water loss. Maximum daily temperature in an open field in the Japanese temperate zone is high in summer (29·3–31·5°C from July to September 1999 at the experimental field). Thus, leaves may be water-stressed by increasing VPD, although all plants were well-watered. Many previous studies showed that increasing WUE is an important aspect of plant acclimation to water limitation ( Ehleringer 1993 Nilsen & Orcutt 1996 ). The increase in WUE in midday was observed for a perennial herb in summer ( Muraoka et al. 1998 ). Tree species in xeric habitat showed higher WUE than in mesic habitat ( Wuenscher & Kozlowski 1971 Garten & Taylor 1992 ) and the difference in WUE was highly genetic ( Anderson et al. 1996 Lauteri et al. 1997 ). The high WUE in A. rufinerve is a favourable characteristic at full sun, especially when summer drought occurs.

When grown in high light, light-demanding species are expected to have a higher Amax/area than shade-tolerant species ( Walter, Kruger, & Reich 1993 Kubiske, Abrams, & Mostoller 1996 Niinemets, Kull, & Tenhunen 1998 Kitao et al. 2000 ), which was not the case in the present study (Fig. 4a). In addition, light-demanding species were found to have higher gs than the shade-tolerant species under high light ( Kubiske et al. 1996 ), which is in contrast to our present result (Fig. 4b). These discrepancies may be caused by the fact that we are comparing species of the same genus (Acer), which is relatively shade-tolerant. In general, shade-tolerant species are drought sensitive ( Kubiske et al. 1996 Niinemets, Kull, & Tenhunen 1998 ). Therefore, among the Acer species, acclimation to high light can be characterized by traits that maximize WUE, rather than maximizing only leaf net carbon gain.

Low light acclimation

There were few interspecific differences in mesophyll anatomy at shade conditions (Figs 1 & 2). This result indicates that anatomical acclimation to shade was not correlated to the difference in light demand of the species. On the other hand, LMA appeared to be highest for the shade-tolerant species A. palmatum in 7% light (Table 2), although the difference was not statistically significant. The higher LMA in shade-tolerant species in shade has been widely observed in the growth experiments of tropical evergreen and deciduous tree seedlings ( Kitajima 1994 Walters & Reich 1999, 2000 Kitao et al. 2000 ) and during field observations of the tall trees of neotropical species ( Rijkers et al. 2000 ). Low mesophyll porosity (Fig. 1 Table 1) and the increase in the density of leaf trichome in 7% light (data not shown) partly contribute to the apparent LMA increase in A. palmatum. These characteristics increase leaf toughness, which may enhance their defence against herbivores and pathogens, thus improving their survival rate in shade ( Kitajima 1994 ).

Concerning the carbon-assimilating capacity, there was no evidence that shade-tolerant species A. palmatum had some advantage compared with the light-demanding species. In 7% light, Amax/area and Rd/area in A. palmatum were similar to those of the other species (Fig. 4a Table 2). Quantum efficiency was also similar among the three species in 7% light (21–33 µmol CO2 mol −1 PPFD, data not shown). These findings support the results obtained in the seedlings of evergreen and deciduous tree species having different shade tolerances ( Walters & Reich 1999 ). In addition, similar chlorophyll concentration among the species in 7% light (Table 2) suggests that light-harvesting capacity would also be similar among the species. The resource allocation in light harvesting and in carbon-assimilating capacity was also similar among the species in shade conditions (Fig. 5c).

Correlation between mesophyll anatomy and gi

An increase in CO2 conductance inside the leaf (gi) for full-sun grown A. mono was accompanied by the increase in Sc (Figs 2c & 4c). There was also an apparent increase in Sc and a significant increase in gi for full-sun-grown A. palmatum. In these species therefore Sc would be one of the factors determining gi. Positive correlations between gi and Sc were obtained by some previous workers ( Evans et al. 1994 Syvertsen et al. 1995 Evans 1998 ), whereas the correlations between gi and Sc were poor in the present study (Fig. 6c).

In A. rufinerve, gi increased only slightly in full sun, irrespective of the remarkable increase in Sc. The thick cell walls of spongy cells (Table 1) may partly contribute in reducing gi in A. rufinerve grown at full sun, because CO2 conductance through mesophyll cell walls is low and it limits liquid phase CO2 diffusion ( Evans et al. 1994 ). Kogami et al. (2001 ) showed that highland Polygonum cuspidatum had a lower gi and thicker mesophyll cell walls than lowland plants, which suggests the importance of cell wall in determining gi. Low mesophyll porosity at full sun (Table 1) might also decrease the gas phase conductance inside the leaves of A. rufinerve. It is possible that both thick mesophyll cell wall and low mesophyll porosity are responses to water limitation in a high light environment. Other factors such as the structural and chemical composition of cell walls, conductance across plasma membrane ( Terashima & Ono 2002 ) and chloroplast envelopes may contribute to the low gi in A. rufinerve grown at full sun. Further study is needed to examine the effects of leaf anatomical characteristics on gi.

Discussion and conclusions

Many factors modulate foliar stomata including abiotic factors such as temperature, moisture, radiation, carbon dioxide in the atmosphere and humidity and nutrient in the soil, as well as plants’ morphological and anatomical factors such as foliar architecture and position, plants’ photosynthetic and transpiration traits [23,24,25,26,27]. In our study, different A. montana provenances were planted in the same site to reduce environmental variation. Also, leaves from generally the same position of the plants were sampled and stomata distributing in consistent areas of adaxial sides were examined to reduce sampling errors. It’s true that the relative positions might still slightly differ because of the individual morphological differences in plants and leaves, however, the difference was minor.

Significant differences existed in stomatal characteristic parameters from different A. montana provenances, consistent with conclusions drawn by Shanna Wen et al. regarding Manglietia conifer Dandy leaf micromorphological characteristics (stomatal density, area, width, and length) [28]. Yanhua Zhu et al. found that growth environment significantly affected micromorphological features of plant leaves [29], suggesting that the stomatal characteristic parameter differences between different A. montana provenances might be affected by both genetic variation and environmental factors. A study by Rongbo Jiang showed that the wider a tree species was distributed, the bigger their genetic variation was, as well as their phenotypic and physiological differences. Furthermore, genetic variation was closely correlated with the complexity of their growth environment [30]. In this study, the variation coefficients of the leaf stomata of different A. montana provenances differed significantly, with a comparison of the average variation coefficient of each characteristic showing that the Putian provenance had the highest average stomatal characteristic variation coefficient of 20.46% and the Zhenghe provenance the lowest coefficient of 11.21%. These indicated that different provenances of the same plant species had different stomatal adaptability under the same growth environment. In terms of environmental changes, low variability means the plant has stronger self-regulation and environmental adaptability and better leaf stomata density stability. The Zhenghe, Fuding, Youxi, Shaxian, and Jianyang provenances, therefore, possessed better stomatal density stabilities that might be associated with genetic factors.

Casson [31] showed that stomatal density and size directly affected the transpiration and photosynthesis rates of plant leaves and that the key factors associated with plant environmental adaption were two contradictory characteristics the adjustment of stomatal movement and optimization of stomatal density and size. Thus, correlation analysis of stomatal characteristics was of great importance. According to several studies, stomatal area, perimeter, long axis length, and short axis length are closely related to stomatal density. Observation of wheat leaf stomata by Weiyue Chen[32] showed that leaf stomatal density was significantly negatively correlated with the stomatal long axis and short axis length. In this study, consistent with other reports, stomatal density was significantly negatively correlated with stomatal perimeter, area, long axis length, and short axis length.

Comparison of five stomatal characteristics (stomatal density, area, perimeter, long axis length, and short axis length) in leaves of nine different A. montana provenances and cluster analysis based on those characteristic parameters showed that, when the Euclidean distance was set to 5, these nine A. montana provenances could be classified into four groups. Most had high stomatal densities that were generally accompanied by smaller stomatal sizes. In terms of environmental changes, smaller stomata had shorter reaction times and faster open/close times compared with larger ones, meaning that the combination of high density and small size improved the stomatal conductance under benign environments and optimized the CO2 diffusion rate to decrease the stomatal conductance under adverse environments [33,34]. This indicated that most A. montana provenances had strong environmental adaptabilities, in particular, the Jianyang, Zhenghe, and Shaxian provenances.

To summarize, analysis of stomata showed the A. montana, Jianyang, Zhenghe, Fuding, and Shaxian provenances were better transplant choices when considering their stomatal stabilities and stomatal environmental adaptabilities, thereby facilitating the selection of better A. montana provenances. Currently, there is no information about the effect of stomatal features on some other foliar functional traits such as photosynthesis, water balance, and construction cost. This might be a future challenge.

Results

Contrasting water‐use patterns identified in wild and cultivated lettuce

When grown under well-watered conditions, the wild (L. serriola) and cultivated (L. sativa) parents of the recombinant inbred lines (RILs) showed significant variation in their diurnal pattern of transpiration (repeated measures ANOVA F1,9 = 24.76, P < 0.001, Fig. 1). For cultivated lettuce, transpiration rose from 05:00 until 13:00 h when it declined until measurements ceased at 23:00 (Fig. 1). Transpiration continued to rise until 15:00 h for wild lettuce, which demonstrated a significantly higher transpiration rate than its cultivated relative consistently throughout the course of the day under well-watered conditions (F1,9 = 24.76, P < 0.001, Fig. 1) until 23:00 h. This effect was observed in several experiments (data not shown).

Diurnal transpiration of cultivated (L. sativa cv. Salinas) and wild (L. serriola) lettuce. Transpiration pattern (mmol m − 2 s − 1 ) (a), with example thermal images of cultivated and wild lettuce (b)

Transpiration rate was also higher in wild lettuce under drought (t10 = -2.35, p < 0.05, Fig. 2a) as was stomatal conductance (t10 = -2.90, p < 0.05, Fig. 2b). Although leaf temperature did not vary significantly between the two parents, there was a trend for lower leaf temperatures in wild lettuce when compared to the cultivated parent under drought (Fig. 2c), confirming the data from stomatal conductance. Leaf temperature was significantly higher in wild lettuce under well-watered conditions (t10 = -3.83, p < 0.01, Fig. 2c). Though differences between wild and cultivated lettuce were observed, the gas exchange response of both genotypes was negligible when the well-watered and drought 1 experiments were compared for each individual (Fig. 2a–c), however leaf temperature was significantly reduced by imposing water stress for wild lettuce (t10 = 3.73, p < 0.01, Fig. 2c). Carbon isotope discrimination (Δ 13 C) was consistently higher for wild lettuce compared to cultivated lettuce (Fig. 2d). Oxygen isotope discrimination was higher in cultivated lettuce (31.31 ± 0.75) than wild (29.08 ± 0.03), although differences were not significant.

Drought response of cultivated (L. sativa cv. Salinas) and wild (L. serriola) lettuce. Transpiration (a), stomatal conductance (b), leaf temperature (c), carbon isotope (d) and oxygen isotope discrimination (e). * indicate significant differences (see text for details)

Phenotypic variation for water‐use traits in the RIL population

Phenotypes for water-use traits segregated under well-watered, mild and moderate drought conditions within the RIL population and bidirectional transgressive segregation was evident for transpiration, stomatal conductance, leaf temperature, fresh and dry weight (Figure S1). The RILs demonstrated transgressive segregation below either parent for carbon isotope discrimination under drought, indicating this population may have an improved water-use efficiency under these conditions.

Infrared thermal measurements of leaf temperature correlated well with porometry measurements under well-watered (r 2 = 0.62, p < 0.001, Fig. 3a) and drought conditions (r 2 = 0.81, p < 0.001, Fig. 3b). Transpiration (E) was strongly positively correlated with stomatal conductance (gs) under well-watered (r 2 = 89, p < 0.001, Fig. 3a) and drought conditions (r 2 = 0.75, p < 0.001, Fig. 3b). Both E and gs were significantly negatively correlated with fresh (r 2 =-0.19 and r 2 =-0.20, respectively, P < 0.01) and dried whole plant biomass (r 2 =-0.35 and r 2 =-0.39, respectively, P < 0.001), but positively correlated with fresh:dry weight ratio (r 2 = 0.36 and r 2 = 0.40, respectively, P < 0.001) in the Dr1 trial, although significant variation was observed in the data. Application of drought led to a reduction in gs measured using a porometer (r 2 =-0.18, p < 0.01, Fig. 3b) and by thermal imagery (r 2 =-0.27, p < 0.001, Fig. 3b), though the opposite effect was seen under well-watered conditions when temperature was measured using the porometer (r 2 = 27, p < 0.01, Fig. 3a). As expected, carbon isotope discrimination was found to be significantly negatively correlated with above ground fresh weight biomass under both drought treatments (r 2 =-0.33, P < 0.05, Fig. 3b and r 2 =-0.69, P < 0.001, Fig. 3c, for Dr1 and Dr2 treatments, respectively). Carbon isotope discrimination was positively correlated with E and gs under drought stress (r 2 = 0.36 and r 2 = 0.35, respectively, P < 0.01, Fig. 3b).

Correlations between water-use traits. Observed under well-watered conditions (a) and under drought 1 (b) and 2 (c) trials. Estimated using Spearman’s correlation, with scatterplot (bottom left) and significant r 2 correlation values (top right) shown. * indicates significance at P > 0.001 (***), P < 0.01 (**) and P < 0.05 (*). Transpiration (e), stomatal conductance (gs), temperature measured by porometry (Temp), temperature measured by thermal imaging (IR), carbon isotope discrimination (Δ 13 C), whole fresh weigh (FW), dry weight (DW) and their ratio (FW:DW)

QTL for water‐use traits in lettuce

A genetic linkage map with 1,099 markers spanning a total of 1,414.7 cM across 10 linkage groups was generated using regression mapping in Joinmap. Collinearity of marker ordering was validated using the physical map. Due to a region of high segregation distortion on chromosome 3, which has been previously noted by Truco et al. [7], this linkage group did not coalesce and was split into two segments labelled as 3a and 3b, which were 62 and 33 cM in length, respectively. Maximum marker interval was 16.9 cM with an average spacing of 1.3 cM (approximately 2.2 Mb) across all LGs. Utilising this molecular marker map, 30 significant QTL were identified for nine of the ten traits investigated, with no QTL identified for whole plant dry weight. These QTL accounted for 4.8–23.6 % of the phenotypic variation (PV), with 22 small effect QTL (< 10 % PV) and eight moderate effect QTL (10–25 % PV) and spanned eight of the ten linkage groups, with no QTL identified on LG5 or LG3b (Table 1 Fig. 4).

QTL identification for water-use-associated traits in the RIL population. Bars represent each LG with position in centiMorgan on the left, LG number at the top of each bar and horizontal lines indicating marker positions. QTL are shown as filled boxes to the right of each LG representing the 1-LOD interval, with error bars showing the 2-LOD interval for QTL detected in the well-watered (blue), Dr1 (red) and Dr2 (black) trials. 13 C, Δ 13 C, FW, fresh weight FW_ Lf56, fresh weight of fifth and sixth true leaves, FW:DW, fresh:dry weight ratio gs, stomatal conductance, E, transpiration, E:DW, transpiration:(dry weight) ratio Temp, leaf temperature measured via porometry, IR, leaf temperature measured via IR thermography

Two QTL for E were identified on LGs 2 and 9, under the Dr1 and well-watered treatments and accounting for 8.4 and 7.5 % of the PV respectively, with L. sativa allele inheritance increasing the trait value. A QTL for leaf temperature measured via steady state porometry mapped to the same position as E on LG 9 accounting to 9.1 % of the PV. Four QTL for gs were identified, two under the well-watered treatment on LGs 7 and 8, cumulatively accounting for 17.2 % of the PV and two under Dr1 treatment, located 53 cM apart and accounting for 10.6 % of the PV. Three moderate and one small-effect QTL for Δ 13 C measured in the Dr2 trial were identified on LGs 6, 8 and 9, together accounting for 63.4 % of the PV. QTL for Δ 13 C co-located to those for whole plant fresh weight (FW) on LGs 6 and 8 in the Dr1 trial and a second QTL for Δ 13 C on LG 8 located to the same position as FW, gs and the ratio between E and whole plant dry weight measured in the Dr1 trial. Two QTL for leaf temperature, measured by porometry and IR thermal imaging in the Dr1 trial co-located on LG 4, accounting for 5.6 and 9.5 % of the PV. Mapping has identified interesting candidate regions for further functional investigations.

Candidate genes for WUE in lettuce

Nine locations with large-effect or multiple overlapping QTL were selected for candidate gene analyses (Table S2). The 2-LOD QTL intervals were mined for genomic features, identifying > 1,400 putative genes from 73.8 Mbp of genome sequence and 87 % of these genes retrieved a BLASTp hit against 15 plant protein databases (Table S3).

Four regions of interest were located on LG8 (Fig. 5). QTL_8–51, comprising two QTL for gs and FW:DW, and these harboured a cluster of six xyloglucan endotransglucosylase/hydrolases (XTH) which have roles in modifying the extensibility of the cell wall and have been linked to drought tolerance through influencing stomatal pore size [42]. Other candidates in this region included a subtilisin-like serine protease, which modulate cell differentiation during stomatal development [43], a glutaredoxin family protein, associated with drought stress tolerance through ROS detoxification [44] WRKY and BHLH transcription factors. Significantly enriched GO terms within QTL_8–51 included those for cell wall (GO:0005618), cellular polysaccharide metabolic process (GO:0044264) and xyloglucan:xyloglucosyl transferase activity (GO:0016762 Table S4). A subtilisin-like protease and glutaredoxin family protein were identified within QTL_8–89, a QTL for Δ 13 C accounting for 9 % of the PV. A QTL for gs accounting for 10 % of the PV on LG 8, QTL_8–65, mapped to the same position as two aquaporin-like proteins involved in water transport and an ABA-responsive element binding protein, involved in ABA-induced stomatal closure following water deficit [45]. Another aquaporin protein was identified within QTL_8-100 a hotspot on LG8 in which QTL for Δ 13 C, gs, FW and the ratio between E and DW co-located. Other notable candidates in this QTL hotspot included a subtilisin-like serine protease, a dehydration-associated protein, three BZIP and one BHLH transcription factors (Fig. 5). Significantly enriched GO terms within QTL_8-100 included defence response (GO:0006952), response to stress (GO:0006950), stimulus (GO:0050896), oxidative stress (GO:0006979) and antioxidant activity (GO:0016209 Table S4).

Candidate gene mining of LG8 QTL. Illustration of LG 8, with the QTL investigated for candidate genes highlighted and gene information provided

Under QTL_6–6, a region in which large-effect QTL for Δ 13 C and FW traits measured from the Dr2 trial co-located, a late embryogenesis abundant (LEA) protein along with several transcription factors reported to influence response to drought were identified, including three MYB-like domain containing proteins, a NAC, APETALA2 (AP2)-like ethylene-responsive factor and WRKY transcription factor (Hadiarto & Tran, [46] Nuruzzaman et al., [47] Table S3). The same region contained ten glutathione S-transferases and a glutathione peroxidase, involved in reactive oxygen species (ROS) detoxification in response to drought [48]. A region in which a QTL for E and leaf temperature co-localised, designated QTL_9–27, contained a LEA protein, an aquaporin-like protein, a glutathione S-transferase and several transcription factors (AP2, MYB, NAC, WRKYand BZIP Table S3).

Cleome gynandra

If you can supply pictures for this datasheet please contact:

Compendia
CAB International
Wallingford
Oxfordshire
OX10 8DE
UK
[email protected]

Don't need the entire report?

Generate a print friendly version containing only the sections you need.

Pictures

Title Flowers Gynandropsis gynandra Flowers. Krishnagiri, Tamil Nadu, India. November 2015. ©siddarthmachado/via inaturalist - CC BY-NC 4.0
Title Flowering plant Gynandropsis gynandra Flowering plant. AVRDC (The World Vegetable Center), Taiwan. September 2013. ©Michael Hermann/via Wikimedia Commons - CC BY-SA 3.0
Title Flowers Gynandropsis gynandra Flowers. September 2012. ©Pau Pámies Grácia/via Wikimedia Commons - CC BY-SA 4.0
Title Foliage Gynandropsis gynandra Foliage. Kouhu, Yunlin County, Taiwan. August 2018. ©Kuan-Chieh (Chuck) Hung/via inaturalist - CC BY-NC 4.0
Title Leaves Gynandropsis gynandra Leaves. Kouhu, Yunlin County, Taiwan. August 2018. ©Kuan-Chieh (Chuck) Hung/via inaturalist - CC BY-NC 4.0
Title Leaves Gynandropsis gynandra Leaves. September 2012. ©Pau Pámies Grácia/via Wikimedia Commons - CC BY-SA 4.0
Title Habit Gynandropsis gynandra Habit. Kouhu, Yunlin County, Taiwan. August 2018. ©Kuan-Chieh (Chuck) Hung/via inaturalist - CC BY-NC 4.0
Title Fruit Gynandropsis gynandra Fruit. Krishnagiri, Tamil Nadu, India. November 2015. ©siddarthmachado/via inaturalist - CC BY-NC 4.0
Title Fruit and seeds Gynandropsis gynandra Fruit and seeds. February 2019. ©Sosicles Ennius/via Wikimedia Commons - CC BY-SA 4.0

Taxonomic Tree

• Domain: Eukaryota
• Kingdom: Plantae
• Phylum: Spermatophyta
• Subphylum: Angiospermae
• Class: Dicotyledonae
• Order: Capparidales
• Family: Capparaceae
• Genus: Cleome
• Species: Cleome gynandra

Notes on Taxonomy and Nomenclature

Cleome is one of 10–17 genera of the small family Cleomaceae that were formerly included in the Capparaceae, but raised to comprise a distinct family based on chloroplast and nuclear DNA evidence. This evidence suggested that the genera are more closely related to the Brassicaceae than they are to the Capparaceae (Hall et al., 2002). According to The Plant List (2013) there are multiple synonyms of C. gynandra L. but the most commonly encountered are Gynandropsis gynandra (L.) Briq. and G. pentaphylla (L.) DC. On the basis of a numerical analysis of 100 morphological, anatomical and seed protein characters, Mohamed (2009) supports the subsuming of G. gynandra under C. gynandra as does El-Ghani et al. (2007) on the basis of leaf architecture.

An extensive list of vernacular names is tabulated by Chweya and Mnzava (1997) and these include cat's whiskers, African spider flower, bastard mustard (English) kurhur, karaila (India) phak sian (Thai) babowan (Indonesia) caya blanc, brede caya, mouzambé (French) musambe (Angola) mgagani (Swahili) and boanga, mugole (West Africa).

Description

Erect, herbaceous annual, 0.5–1(–1.5) m tall, branched and with a long tap root with few secondary roots. Stems and petioles thickly covered with glandular hairs, rarely glabrous, varying in colour from green to pink, or violet to purple. Leaves alternate, digitately palmate, (3–)5(–7) leaflets, sessile, pinnately dissected, sparsely hairy, obovate to elliptic, 2–10 cm long, 2–4 cm wide, finely toothed margin or rounded ends,  petioles 3–23 cm long. Inflorescences showy, up to 30 cm long, terminal and axillary determinate racemes flowers arise singly in axils of small sessile and trifoliate to simple bracts which are smaller than the leaflets flowers 1–2.5 cm in diameter pedicels long 4 sepals, free, ovate to lanceolate, up to 8 mm long, glandular 4 narrow clawed petals, 6 stamens with long purple filaments arising from an elongated gynophore style short extending to a purple capitate stigma depressed at the apex ovary bicarpellary syncarpous, unilocular with numerous ovules on parietal placentation, a false septum develops during fruiting petals white, pale pink or lilac floral formula K4C4A6G(2). Fruit long stalked silique, spindle shaped, 12 cm long, 8–10 mm wide green in colour and yellow when ripe easily dehiscent when dry releasing seeds. Seeds numerous, 1.0–1.5 mm in diameter, suborbicular, sharply tuberculate with many concentric ribs and irregular cross ribs grey-black in colour seed cleft narrow. Seedlings have oblong cotyledonary leaves, hairy petioles and petiolate trifoliate to elliptical leaflets. Terminal leaflet generally larger than lateral leaflets (Chweya and Mnzava, 1997 Raju and Rani, 2016).

Distribution

Widely distributed in drier areas of the tropics and subtropics, C. gynandra is likely to have originated in tropical Africa and South East Asia (Chweya and Mnzava, 1997). Although species of the genus Cleome mostly occur in Africa, Zhang and Tucker (2008) state that the centre of diversity of Cleome is in South West Asia. Semicultivated and considered native across sub-Saharan Africa and Asia, and in parts of northeast Thailand it is grown on a large scale commercially (JIRCAS, 2010). C. gynandra has been introduced and regarded as a weed in many areas, including the Caribbean islands (Bermuda, Bahamas, Cuba), southern USA, Central and South America, Central and Northern Europe, Russia, China, Japan, Korea, Australia, New Zealand and Pacific Islands (Chweya and Mnzava, 1997). Information on distribution in relevant countries, locations and whether invasive or cultivated is detailed on the Pacific Island Ecosystems at Risk website (PIER, 2018).

Distribution Table

The distribution in this summary table is based on all the information available. When several references are cited, they may give conflicting information on the status. Further details may be available for individual references in the Distribution Table Details section which can be selected by going to Generate Report.

Biology and Ecology

Growth stages and physiology

After germination, vegetative growth is rapid and, if conditions are favourable, a plant height of 60–90 cm may be achieved before flowering. Flowering is early often within 4–6 weeks of growth, and the growth rate is highest between the 5th and 6th weeks of growth (Chweya and Mnzava, 1997). Leaves follow the direction of the sun throughout the day. Sailaja et al. (1997) noted diurnal constancy in laser induced chlorophyll fluorescence of leaves. This heliotropic behaviour maximises light use efficiency throughout the day avoiding a midday dip in photosynthesis.

Compared to the ancestral C3 state, C4 photosynthesis is a highly efficient process of carbon fixation originating from a complex phenotype of tightly regulated biochemical and anatomical modifications of leaf architecture. It has evolved at least 66 times across several lineages and the evolutionary route from C3 to C4 is not necessarily identical but is likely conserved (Covshoff et al., 2014). A clear understanding of the post-transcriptional mechanisms that regulate gene expression central to the onset of C4 photosynthesis is required before aiming to transform C3 crops into C4 crops (Fankhauser and Aubry, 2017). As it is related to Arabidopsis, a model C3 system, there is great interest in developing C. gynandra as a model C4 species (Marshall et al., 2007). Koteyeva et al. (2011) surveyed several Cleome species for structural and functional forms of C4 and found that C. gynandra has an atriplicoid-type Kranz anatomy with individual veins surrounded by multiple simple Kranz units. Analysis of leaf tissue found high levels of expression of markers for key photosynthetic enzymes of the C4 cycle. Marshall et al. (2007) notes that C. gynandra belongs to the NAD-dependent malic enzyme subtypes of C4 photosynthesis.

Reproduction

Over a period of 2–3 months, 200–400 flowers develop acropetally on the terminal inflorescence in a tight spiral with several flowers open simultaneously. Axillary branches develop and flowering continues on terminal inflorescences for 6–12 months. Taking 2 weeks for flowers to develop, sepals first develop and elongate fully, followed by petals, stamens, stigmas and gynophores. Pedicels elongate throughout floral development and gynophores elongate to around 7 cm before seed development commences and reach 8 cm during seed maturation (Chweya and Mnzava, 1997). Raju and Rani (2016) demonstrated that C. gynandra is polygamodioecious with flowers are of two types: staminate (with residual ovary without ovules) or bisexual (functional ovary and fertile stamens). Bisexual flowers are classified into four different floral morphs based on length of gynoecium: medium gynoecium flowers (MGF), long gynoecium flowers (LGF), medium gynoecium short stamen flowers (MGSSF) and medium gynoecium sessile shortest stamen flowers (MGSeSF). The morphs SGF, MGF and LGF occur on the same individual plant while the other two flower morphs (MGSSF and MGSeSF) are produced singly on different individual plants. Plants producing MGSSF and MGSeSF floral morphs are less common. Flowers are protogynous. In all floral morphs pollen grains are monads, spherical and yellow, 19.92 μm in diameter, prolate, tricolpate and reticulate. C. gynandra is facultatively self-pollinated but outcrossing is thought to occur facilitated by insects and the wind (Chweya and Mnzava, 1997). Omondi et al. (2017) conducted hand pollination experiments between 30 lines and showed that they were self- and cross-compatible. According to Raju and Rani (2016), the number of seeds per fruit varies with floral morph ranging from 82.8 (MGF) to 148 (LGF). Seed set also varies with floral morph ranging from 71% (MGF) to 97% (MGSSF). The light seeds are dispersed by wind during the dry season and by rain in the wet season.

Environmental requirements

Occurring in semi-arid, subhumid and humid climates it is grows in cultivated or fallow fields and alongside roads. Although found mainly in semi-arid conditions it is adapted to moist soils alongside rivers, irrigation canals and ditches. C. gynandra grows from sea level to 2400 m in Africa in the temperature range 18–25°C. Tolerating many soil types providing they are deep, well drained and with a pH range of 5.5 to 7, plants grow well on rubbish dumps and in soils amended with manure (Chweya and Mnzava, 1997). C. gynandra is tolerant of salinity with only a slight reduction in plant height resulting from application of a saline solution (75 mM NaCl) for 15 days. Proline contents of stem, roots and leaves were unaffected (Kulya et al., 2011). Mwai et al. (2004) noted that it can grow and reproduce (although somewhat retarded) under salt stress of -0.9 MPa (0.2 mol/kg NaCl). They suggest that although C. gynandra is salt sensitive it has a capacity for osmotic adjustment. Flooding is not tolerated.

Young leaves and stems of C. gynandra are cooked and eaten as a vegetable either alone or in stews. Various methods are used to ameliorate bitterness, particularly in eastern African countries. In Kenya leaves are steeped overnight in milk. In Botswana, initial blanching of leaves is performed, the water discarded and cooking continued in fresh water (Sogbohossou et al., 2018). Leaves are also cooked with other leafy vegetables such as nightshades (Solanum spp.), Amaranthus and cowpea (Vigna unguiculata) leaves. In addition, leaves are dried, ground and used as a powder or blanched, rolled into balls and then dried. In Thailand the leaves, young shoots and flowers are fermented in salt water prior to serving as a side dish with nam phrik (JIRCAS, 2010). Seeds are used for mustard and contain edible polyunsaturated oil (Mnzava and Chigumira, 2004).

Leaves, stems and roots are widely used in traditional medicine. Boiled leaves are used to boost immunity in women and children, to treat blood loss in new mothers and after injury, and as a general medicinal meal. Leaf infusions are used to treat diarrhoea and anaemia, while leaves and seed are used externally and internally to treat rheumatism. Root infusions are used to treat chest pains. C. gynandra has also been used to treat various digestive disorders, inflammation, epilepsy and malaria (Mnzava and Chigumira, 2004 Iwu, 2014 Sogbohossou et al., 2018). In Zambia C. gynandra is used to treat chancroid, a venereal infection causing ulceration of the groin lymph nodes (Chinsembu, 2016).

Pharmacological analyses of the leaves have shown high concentrations of various compounds including flavonoids, tannins, glucosinolates, and iridoids. These contribute to antibacterial, antifungal, antiviral, analgesic, anticarcinogenic and anti-inflammatory properties (Yang et al., 2008 Ghogare et al., 2009 Moyo et al. 2013 Bala et al., 2014).

In comparison with currently used chemicals to control tick infestations in livestock, C. gynandra is one of several species with acaricidal and larvicidal effects with 90–100% efficacy (Adenubi et al., 2016).

Discussion

The purpose of this study was not to partition the influence of the different components of palatability, as has often been attempted in previous works, for example by the use of reconstituted plant material in agar gels ( Cronin 1998 Hay et al. 1994 Pennings & Paul 1992 ). Because it is also not realistic to screen large numbers of species in this way, and because there is evidence that these components are interactive or synergistic in their effects ( Duffy & Paul 1992 Hay et al. 1994 ), we regard the underlying covariation between the different traits involved in plant palatability as a global feature of natural populations. We are therefore concerned with an evaluation of DMC as a convenient and potentially integrative expression of palatability.

Cruz-Rivera & Hay (2001 ) previously observed an inverse correlation between the ash-free DMC of nine marine macrophyte species and their consumption by an amphipod in a non-choice experiment. As the most consumed species in the non-choice assay were little consumed in a multiple-choice experiment, they attributed this pattern to compensatory feeding on food of low organic matter content when alternative foods were not available. A negative correlation was also observed in a multiple-choice assay between the feeding rate of a salt-marsh crab and the DMC of seven vascular-plant species, which was probably explained by the link between plant DMC and toughness ( Pennings et al. 1998 ). The present results indicate that there is a highly significant negative correlation between DMC and palatability, reproduced between the 20 species studied, within Elodea species, and within P. lucens over time and between sites. Consequently, it is unlikely that the relationship observed among the complete set of species is an artefact of phylogeny, although we cannot test this definitively due to the lack of an established phylogeny for the Alismatales. Moreover, in a comparison of the feeding rate of L. stagnalis in multiple-choice and non-choice experiments on eight freshwater macrophyte species (including five of the species investigated in the present paper), the ranked palatability of the species studied was not significantly affected by the possibility of choice. This indicates that, in this instance, the results of non-choice tests tend to order the species correctly in terms of palatability and are little influenced by compensatory feeding ( Elger et al. 2002 ).

For the complete set of species, the effects of amphibiousness and leaf shape can be explained by the smaller values of DMC of potentially amphibious species and/or those with linear leaves, as the differences in snail consumption were not significant when these factors were tested after log (DMC) in the ancova . The difference of palatability between the two Elodea species was also partly related to the smaller DMC of E. nuttallii. At an intraspecific level, the effect of date is also fully explained by the temporal variability in DMC, for the Elodea species and for P. lucens.

We attempted to minimize the influence of mechanical defence by deliberately using exclusively submerged leaves, devoid of trichomes or thickening. Potential macro-scale indicators of mechanical defence, in the form of leaf shape and denticulate leaf margins, failed to explain significant variation in palatability between species with similar DMC. Moreover, the use of plant fragments was assumed to reduce the effect of plant architecture. Nevertheless, differences in plant architecture make it impossible to completely homogenize the nature of the fragments used in the feeding experiments, and this parameter had a significant effect on herbivore consumption rate when added after log (DMC) in the ancova . This effect is probably caused by the fact that large, flat leaves, independently of their shape, offer a more favourable surface to snails that must physically occupy the substratum they graze, than do small-leaved shoots ( Steneck & Watling 1982 ). However, the nature of fragments was of secondary importance in snail feeding rate, as its effect was hidden by variations in plant DMC.

We hypothesized that DMC could be a good indicator of physical and chemical defences of freshwater macrophytes. These traits are the major determinants of plant palatability to invertebrates in freshwater systems, nutrient content generally being of secondary importance ( Kolodziejczyk & Martynuska 1980 Lodge et al. 1998 Newman, Kerfoot & Hanscom 1996 ). However, nutrient content, which is assumed to be negatively correlated with DMC ( Mattson 1980 ), may be elevated in importance once interspecific variation in physical and chemical defence is subdued through the choice of affiliated species. Similarly, leaf constituents such as lignin, fibre and silica contents, which contribute to leaf toughness and reduce palatability or leaf decomposition rate ( Cornelissen & Thompson 1997 Grime et al. 1996 ), have a transparent link to DMC. In contrast, some chemical defences [mobile defences sensu Coley et al. (1985 ) such as alkaloids or cyanogenic glycosides] may be relatively cheap to produce, and are thus independent of factors such as relative growth rate ( Almeida-Cortez, Shipley & Arnason 1999 Fenner et al. 1999 ). In the present study, such chemicals could help explain the residual variation in palatability at an interspecific level or between sites at an intraspecific level.

The ability of intraspecific variability of DMC to explain seasonal variations in macrophyte palatability suggests that the balance between growth and defence can switch over time. In spring, freshwater macrophytes take advantage of high light and nutrient availability to complete their growth phase in a short period ( Best 1977 Sand-Jensen et al. 1989 ). Defence is therefore constrained, most of the resources being allocated to growth, which explains why the palatability and water content of the macrophytes studied are both high in spring ( Feeny 1970 Mattson 1980 ). Later in the season, as growth slows, allocation to defence processes presumably increases ( Herms & Mattson 1992 ), which is consistent with our results. Because we used consistently young shoots, the DMC trends that we observed cannot be due simply to leaf maturation. It is unclear to what extent they are an adaptive response to reduce palatability, or whether they primarily reflect increased structural support for light harvesting tissue at the shoot tip within maturing plant stands, with any defensive role being secondary.

The exact roles of nutrient content and internal leaf defences, as the functional relations between these traits and DMC, remain to be tested. Also, the relation between leaf palatability and DMC remains to be confirmed outside the aquatic system. However, DMC is an effective indicator of macrophyte palatability, despite the probable variation in defensive investment for a given DMC. Palatability as a global trait depends on many plant features. Quantifying these and assessing their functional relevance is likely to be challenging. Thus, from an ecological perspective, we regard the ability of DMC to explain a significant part of the differences in palatability observed between species and over time as a potentially useful means of predicting palatability which integrates some of these features.