Physiological adaptation to irradiance in duckweeds is species and accession specific and depends on light habitat niche

Abstract

Duckweeds span 36 species of free-floating aquatic organisms with body sizes ranging from 2 mm to 10 mm, where each plant body plan is reduced to a largely leaf-like structure. As an emerging crop, their fast growth rates offer potential for cultivation in closed systems. We describe a novel UK collection derived from low light (dLL) or high light (dHL) habitats, profiled for growth, photosynthesis, and photoprotection (non-photochemical quenching, NPQ) responses. Twenty-three accessions of three Lemna species and one Spirodela polyrhiza were grown under relatively low light (LL: 100 μmol m^–2 s^–1) and high light (HL: 350 μmol m^–2 s^–1) intensities. We observed broad within- and between-species level variation in photosynthesis acclimation. Duckweeds grown under HL exhibited a lower growth rate, biomass, chlorophyll, and quantum yield of photosynthesis. In HL compared with LL, carotenoid de-epoxidation state and NPQ were higher, whilst PSII efficiency (φPSII) and Chl a:b ratios were unchanged. The dLL plants showed relatively stronger acclimation to HL compared with dHL plants, especially Lemna japonica accessions. These achieved faster growth in HL with concurrent higher carotenoid levels and NPQ, and less degradation of chlorophyll. We conclude that these data support local adaptation to the light environment in duckweed affecting acclimation in controlled conditions.

Adaptation to high light levels in duckweed is characterized by growth and pigment alterations, with accessions from low light environments showing stronger acclimation than accessions from high light environments.

Introduction

Duckweeds are miniature, remarkably fast growing aquatic plants, with doubling times as low as 24 h and dry weight production of 106 t ha^–1 year^–1 (Cui and Cheng, 2015; Ziegler et al., 2015; Michael et al., 2021). Species are widely distributed across diverse waterbodies from swamps to flowing streams and canals. Consisting of simplified units of tiny, reduced leaf–stem structures (fronds) detaching to form independent clones, vegetative reproduction allows rapid colonization and full coverage of water surfaces. The 36 identified duckweed species have great phenotypic diversity within five genera, comprising larger and rooted Spirodela, Landoltia, Lemna, and the reduced and rootless Wolffia and Wolffiella. Duckweeds have worldwide market potential as human and animal feeds: growing rapidly without soil, they have global distributions and offer dietary protein including essential amino acids, with levels comparable with soybean (Cheng and Stomp, 2009). Duckweeds are also a source of high starch, fibre, and micro- and macronutrients (Appenroth et al., 2017; Yahaya et al., 2022). Historically used in Asian cooking, there is also growing interest in duckweeds for vertical farming and even as a live plant food for space travel (Smith et al., 2009, 2015; Appenroth et al., 2018; Stewart et al., 2020; Polutchko et al., 2022; Nguyen et al., 2023).

As miniature plants, duckweeds have little control over growing position, often lodged at water edges or passively motile, carried with the water flow or by water-dwelling organisms through an array of light environments. Duckweeds are tolerant of extremely low light conditions from grates and drains where light penetrates poorly, to vast duckweed carpets in expansive open lakes such as Titicaca in Peru and Bolivia, showing tolerance of growth in full light and high temperature. Although successful worldwide colonization strategies suggest natural variability of photosynthetic apparatus to challenging light climes, this has not been explored in a dedicated fashion (Ceschin et al., 2018; Paolacci et al., 2018; Stewart et al., 2020; Strzałek and Kufel, 2021).

Plants have many mechanisms that allow efficient acclimation to varied intensities and wavelengths of light, caused by, for example, season, cloud cover, and tree and shrub cover. Whilst light stimulates high photosynthesis under optimal growing conditions, high irradiance can cause damage to or inactivation of photosynthetic processes (photoinhibition) often in combination with other stresses such as low or high temperature. Plants can modify the area and width of leaves for shade and sun tolerance, which involves re-arrangements of the composition of light-harvesting complexes, photosystem stoichiometry, and Chl a:b ratios (Anderson et al., 1995; Maxwell et al., 1995; Walters, 2005; Poorter et al., 2006). Light acclimation is characterized by altered photosynthetic rates per unit leaf area and increased production of carotenoids and other counteracting antioxidants (Murchie and Horton, 1997; Foyer and Harbinson, 1999; García-Plazaola et al., 2004). Photoacclimation processes typically support high photosynthetic capacity in high light (HL) and more efficient light harvesting and quantum yield under low light (LL). This is commonly observed using light response curves for gas exchange and/or electron transport showing quantum yield and the saturation point of photosynthesis which can vary, for example 300–500 µmol m^–2 s^–1 for temperate annuals to ~1000 µmol m^–2 s^–1 in rice growing in the tropics (Murchie and Horton, 1997; Zhao et al., 2017). Acclimation to HL is also associated with photoprotective processes such as non-photochemical chlorophyll fluorescence (NPQ) which quenches excess excitation energy as heat and helps to prevent photoinhibition. NPQ consequentially down-regulates quantum yield of photosynthesis in the short term and can restrict plant growth in prolonged dynamic light exposure in natural conditions (Kromdijk et al., 2016; Ruban, 2016, 2017; Pniewski and Piasecka-Jędrzejak, 2020).

Extensive species-level variation exists in nature in terms of the ability to acclimate to light conditions, which can be linked to growth strategy (Murchie and Horton, 1997, 1998; Demmig-Adams et al., 2012; Burgess et al., 2023). Cultivar/accession-level variation in photosynthetic properties is seen in traditional food crops and their close relatives including wheat and rice (e.g. McAusland et al., 2020; Cowling et al., 2022), and indeed accession-level studies in duckweed have demonstrated broad variation in the accumulation of many elements (Smith et al., 2023, Preprint). It follows from this and the varied climes where duckweed is found that light adaptation studies in duckweed selection hold promise. Indications of variation in growth and changes in thylakoid pigments and photosynthesis in response to light have been observed in single Lemna clones of different species (Paolacci et al., 2018; Stewart et al., 2020), and Lemna species contain an enrichment in light-harvesting proteins compared with Spirodela species (An et al., 2018). However, there has not been a comprehensive analysis of variation in acclimation of photosynthesis to light across duckweeds, and its connection with adaptation to habitat irradiance has only begun to be explored (Strzałek and Kufel, 2021).

Duckweeds present an interesting challenge for photoacclimation due to their floating habit across diverse sites. They have unusual frond anatomy which includes large air spaces for floating, non-functional stomata, and minimal photoassimilate export to the vasculature (Shtein et al., 2017; Ziegler et al., 2023). Their rapid growth rate is usually assessed by clonal colony expansion rates rather than progressive (3D) canopy development and tropic responses toward light seen in other macrophytes. Growth rate in duckweed showed species and ecotypic variation in stable light conditions (Ziegler et al., 2015); however, we suggest that growth rate should be measured in the context of light acclimation to better understand how photosynthetic variations can help to achieve the high growth rates in these species.

Many duckweed species lack information regarding habitat of origin and accession-specific adaptations. Here, we report a new duckweed collection composed of different species and accessions in the UK. Attention was given to the collection environments which were differentiated by light maxima and all spectral composition (FR, R, G, B, UV ,R:FR). We hypothesized that ecotypes and species would show photosynthetic and growth adaptation footprints to local light environment when grown in artificial HL and LL environments. When ecotypes derived from high (dHL) or low light environments (dLL) were cultivated under controlled light conditions, we showed the unexpected outcome that only dLL ecotypes performed well under both LL and HL. Moreover, photosynthetic processes (F_v/F_m, φPSII, and NPQ) were modulated by growth light condition and species, but originating light habitat of ecotypes did not substantially influence photosynthetic parameters. We discuss these outcomes in terms of environmental light acclimation in duckweed and for seeking novel genetic resources for food production.

Materials and methods

Collection of duckweed accessions and measurements of environmental parameters

Twenty-four duckweed isolates were collected in May 2020 in the UK between latitudes of 49.9° and 53.9° and longitudes of –0.29° and –5.19° (Supplementary Table S1; Fig. 1A). Between 10 and 20 individuals from each site were collected into sealed tubes of local water. If mixed species were present across the site, individuals were taken of each species into separate tubes. Site KS06 was a special case where two Lemna minuta KS06A and KS06B varieties were sourced; however, KS06B came from a drain apparently excluding light, and was measured separately to test for potential extreme light-adaptive differences. All duckweed samples were stored at ambient temperature with local natural daylengths until return to the laboratory, where species were confirmed using genotyping by short read genome sequencing (see below).

Fig. 1.

Map of UK collection sites and high duckweed coverage found at high and low light sites. (A) Collection sites of duckweeds in this study, plotted by longitude and latitude coordinates (Supplementary Table S1). Coloured circles represent species groupings as shown in the key (as determined in Fig. 5). (B) High duckweed coverage in an open pond: KS12 in Bradford, UK, high light (dHL) site. (C) Close up of purple colouration in S. polyrhiza at KS12. (D) High duckweed coverage in a pond with steep banks and high tree cover: site KS25 in York, UK, low light (dLL) site.

Nineteen duckweed sites were visited to collect environmental data over a 2 d period in March 2021 and 2022 and July and October in 2020, 2021, and 2022 to monitor variation across the seasons of spring, summer, and autumn.

Duckweed coverage scores were estimated by analysing images of surface coverage. Three photos from above were taken per site with a Canon 650D camera suspended on a camera boom. A white reference and scale was provided for each photo, level with the water surface. Images were processed as follows: three representative areas of 5 × 4 cm rectangles were selected to determine duckweed coverage averages and variability per site. Images were split into red, green, and blue (RGB) stacks and the blue stack used with the threshold scale to best match the original photos. Percentage coverage was quantified using Fiji open-source software (Schindelin et al., 2012). Coverage data and locational coordinate data for duckweed sites are provided in Supplementary Table S1.

Light intensity (maximum PPFD, photosynthetic photon flux density intensity) was measured above 10 cm or up to 1 m (as close as possible to) from a water source using a 400–700 nm light meter (LICOR, 0Li-250A) with an attached quantum sensor head. Using a handheld spectrometer (LICOR, LI-180), total photon flux density (PFD), PPFD, and the ratio of light wavelengths making up the PFD (380–780 nm) were split and recorded as PFD-UV (380–400 nm), PFD-B (400–500 nm), PFD-G (500–600nm), PFD-R (600–700 nm), and PFD–FR (700–780 nm) in µmol m^–2 s^–1. Dominant (λd) and peak (λp) wavelengths (in nm) were recorded at each site. The dominant wavelength is defined as the colour perceived, and the peak wavelength is the highest intensity wavelength recorded per site. All measurements were taken three times over the 20 min period of each visit and the maximum was recorded. All light intensity and spectrum variables (total=50) were used together to group duckweed sites as dLL or dHL habitats using distance analysis and K-means clustering to split sites into two main groups by similarity (Supplementary Tables S2, S3). Distance matrices were computed using Manhattan, Euclidean, and Mikowski distances, and methods showed good consistency for two main groupings and site similarities. To measure variety in light environments, the proportion of spectral quality at each time point was calculated as the proportion of each individual spectral region/total light PFD×100 per site to give percentages, and R:FR was calculated as the ratio of R to FR from each site.

Time, weather (cloud cover), and atmospheric and water temperatures were noted across sites to account for variability across the 2 d periods and seasons. Climate data were collected for each longitude/latitude combination using bioclimatic variables extracted from worldclim.org using the R package Bioclim (Serrano-Notivoli et al., 2022), and altitudes were obtained relative to sea level using Google Earth Pro. The raw and grouped light datasets are summarized in (Supplementary Tables S2 and S3, respectively) and other environmental data are shown in Supplementary Table S4.

Maintenance of duckweed accessions

Wild duckweeds were treated with 0.5% sodium hypochlorite in well plates (Greiner bio-one, Cellstar) for 1–2 min to sterilize them. The duration of treatment was dependent on size of duckweed and visible inward bleaching rates, leaving a green meristematic pocket and then dipping in Milli-Q water 18 MΩ to recover. Multiple individuals from a site were designated A or B based on size (different species) and cultured separately. Sterile colonies were grown in individual flasks containing N medium. Duckweed stocks were grown in GEN-1000 cabinets (Conviron, Winnipeg, Canada) with light provision at 50 µmol m^–2 s^–1 PPFD using broad-spectrum white LED lights providing 16:8 h days with a ramp of light intensity at the start and end to represent sunset and sunrise. A temperature cycle of 22/18 °C day and night was used, and relative humidity was maintained at 60%. N medium was prepared with Milli-Q water 18 MΩ and consists of KH₂PO₄ (0.15 mM), Ca(NO₃)₂ (1 mM), KNO₃ (8 mM), MgSO₄ (1 mM), H₃BO₃ (5 µM), MnCl₂ (13 µM), Na₂MoO₄ (0.4 µM), and FeEDTA (25 µM) as described in Appenroth et al. (1996). Other trace elements were confirmed in N medium as Si 23 µM, Cu 0.27 µM, and Zn 0.15 µM. N medium was autoclaved before use at 121 °C for 20 min. Each week duckweeds were re-sterilized with 0.5% sodium hypochlorite followed by dipping into sterile Milli-Q water and placed into fresh flasks containing new medium to build up sterile stock populations.

DNA isolation and genome sequencing

Each of the sterile 24 duckweed accession stocks were harvested into small populations containing 5–20 individuals, and <50 mg of whole duckweed tissue was frozen in liquid nitrogen. Duckweeds were ground using a Tissuelyser II (Qiagen) and DNA extracted using a DNAeasy Plant kit (Qiagen). DNA quantification was performed on a Qubit 2.0 using the Qubit dsDNA HS assay (ThermoFisher Scientific). Individual library preparations and short read sequencing using Illumina HiSeq 2500 platform sequencing was performed by Novogene, Cambridge, UK.

Sequencing data preparation, alignment, and genotyping

Short read sequencing data for Lemna and Spirodela accessions published in the Sequence Read Archive (SRA) were also included (S. intermedia 9394, L. minuta 9260, L. japonica 7123, L. japonica 7868, L. minor 7210, L. minor 7194, L. minor 9441, L. minor 9541, L. minor 7016, and L. turionifera 6002). Reads were quality trimmed with Trimmomatic (version 0.39) (Bolger et al., 2014) and then aligned to the L. minor 7210 (SRR10958743) reference using BWA (version 0.7.17) (Li and Durbin, 2009) and further processed with Samtools (version 1.9) (Li et al., 2009). Duplicate reads were marked using Picard (version 1.134). Indels were realigned with GATK (version 3.5) (McKenna et al., 2010). The resulting variant call files (VCFs) were then filtered for biallelic sites and mapping quality (QD <2.0, FS >60.0, MQ <40.0, MQRankSum < –12.5, ReadPosRankSum < –8.0, HaplotypeScore<13.0). The VCF was then filtered by depth with a read depth cut-off of <650.

Genetic structure determination and species confirmation

For genetic structure analysis, putatively neutral 4-fold degenerate sites were extracted with DEGENOTATE (https://github.com/harvardinformatics/degenotate). These sites were then examined by a Neighbor–Joining tree using the VCFkit (Cook and Andersen, 2017) and plotted with ITOL (Letunic and Bork, 2021) using known individuals from the SRA to determine species clusters, alongside phenotypic observations (i.e. size of duckweed and presence or absence of seed formation). Species allocation was further explored using fastStructure v1 (Raj et al., 2014) utilizing a genetic admixture of 4-fold degenerate sites for each accession and known individuals to map ancestry. The partition with the lowest Bayesian information criterion (BIC) was chosen for population number, K=4, which was obtained using adegenet version 2.1.3. (Jombart, 2008).

Controlled growth conditions for light acclimation experiments

GEN-2000SH cabinets (Conviron, Winnipeg, Canada) installed with broad-spectrum white LED lights were used to provide growth light treatments. Six months after collection and cabinet acclimatization at 50 µmol m^–2 s^–1 light, UK accessions were subcultured to ~15–20 individuals from each accession into individual flasks and continued to be grown in long days 16/8 h at 22/18 °C day/night temperatures. Accessions were either placed for 2 weeks at LL (100 µmol m^–2 s^–1) (individual accessions n=24) or in 150 µmol m^–2 s^–1 for 1 week to acclimate to the intermediate light conditions and then transferred to HL (350 µmol m^–2 s^–1) for a further week (n=24), to acclimate to conditions (Stewart et al., 2020). The experiment with each light program then ran constantly for up to 6 weeks with a 1 h light and temperature linear ramp to simulate sunrise and sunset. N medium was changed weekly to maintain sterile populations and replenish nutrient dosage. Temperature and relative humidity were recorded using Datalogger (TinyTag Ultra 2, Gemini data loggers) in addition to the cabinet sensors. Light intensity and spectra were measured using a light sensor (LICOR, LI-180) (Fig. 2A).

Fig. 2.

Exploring duckweed photoacclimation using controlled light experiments. (A) Light spectra plotting available levels of light provided in two light treatments. The overall light input for the HL treatment (red) is PPFD 350 µmol m^–2 s^–1 and for LL (black) is PPFD 100 µmol µmol m^–2 s^–1. (B and C) False colour images applied to pixels corresponding to populations of duckweed accessions within well plates as measured by chlorophyll fluorescence by a Fluorcam (B) F_v/F_m and (C) NPQ. Individual accessions are labelled per well and grown at HL in these image examples. Scales represent ranges of measurement values 0.6–0.83 for F_v/F_m measured in the dark and 1–5 for NPQ at maximum light intensity.

Growth rate measurements

Single three-frond colonies from light level stock populations were added to individual conical flasks with 100 ml of N medium on day 0 (n=24) accessions for each HL and LL treatment. Growth rate was measured for each accession in each condition until ~95% surface coverage was achieved, or for up to 6 weeks in slower growing accessions. N medium was replaced each week to maintain high nutrient levels. Relative growth rate (RGR) was measured by colony gain, by counting the number of colonies in each flask, in each condition, every 7 d. Col RGRlog and RGRlog by area gain were calculated from raw data (shown in Supplementary Fig. S1). RGRlog by area gain was derived from total green area (mm²) measured using a digital Nikon D5100 camera and an imaging pipeline for quantification (Ware et al., 2023). The starting areas of the pioneering colonies were subtracted from total area gained each week to normalize differences due to size between accessions and species. RGRlog area was calculated using log(T21 area–T14 area)/7 and Col RGRlog by log(T21 colonies–T14 colonies)/7 (see Supplementary Fig. S1). The number of turions (seeds) formed between accession–light treatments was also assessed from photographs during each growth period. RGR difference by area and colony gain between treatments were obtained by mean growth in HL–mean growth in LL and proportion change calculated by RGR difference/mean growth in LL×100. Total fresh weight biomass (FW) per flask (normalized by weight of the starter colony) was measured after 6 weeks (T42). Fresh biomass per flask was harvested and snap-frozen in liquid nitrogen before being freeze-dried and weighed to obtain freeze-dried weight (FDW).

Chlorophyll fluorescence

Chlorophyll fluorescence imaging was carried out to measure photosynthetic parameters as in McAusland et al. (2019), with a protocol modified for duckweeds. Populations of each accession were grown for 4 or 6 weeks in each light level, and each accession–treatment combination was used to fill the surface of clear plastic 6-well plates containing 3 ml of N medium. Duckweed plates were imaged using a closed chlorophyll fluorescence imager (800C Fluorcam, Photon System Instruments, Brno, Czech Republic); the layout in 6-well plates is shown in Fig. 2B and C. After a 1 h dark adaption, white LEDs with actinic light provided a saturating pulse set at 4500 µmol m^–2 s^–1 for 0.8 s to measure F_v/F_m. Then a rapid stepwise light response curve was constructed with the following light intensities (0, 20, 130, 245, 365, 480, 600, 710, 830, 950, and 1050 µmol m^–2 s^–1 PPFD) with a 60 s illuminating pulse applied at the end of each step. Numeric averages were exported for each accession–treatment replicate using the in-built software (Fluorcam 7, Photon System Instruments). The following parameters were extracted from the protocol: maximum PSII efficiency or quantum yield (F_v/F_m), PSII operating efficiency (F_q'/F_m'), or φPSII and NPQ (F_m–F_mʹ)/F_mʹ) (non-photochemical quenching, i.e. the photoprotective dissipation as heat loss) at each light level (Murchie and Lawson, 2013). These photosynthetic responses are plotted as light response curves and boxplots in Supplementary Fig. S2.

Pigment extraction and analysis

For spectrophotometry, 5 mg of freeze-dried duckweed tissue was ground in 1.5 ml of 80% acetone using a TissueLyser II (Quigen) at 24 Hz s^–1 for 4 min and the cell debris was pelleted. Extracted supernatant was further diluted by 3.5 ml of 80% acetone to give a total volume of 5 ml. Absorbance was recorded for Chl a at 663 nm, Chl b at 646 nm, carotenoids at 470 nm, and absorbance at 750 nm as a correction turbidity factor using a UV/Visible spectrophotometer (Ultrospec 2100 pro, Amersham Biosciences). Total chlorophyll as mg g^–1 duckweed was calculated following Porra et al. (1989) and carotenoids as mg g^–1 calculated as in Lichtenhaler and Wellburn (1983).

Pigment extraction and analysis by HPLC

For carotenoid HPLC analysis, tissue was rapidly frozen in liquid nitrogen at mid-day. A 0.8 g aliquot was ground in 2 ml of 100% acetone (HPLC grade) in LL and centrifuged at 10 000 rpm at 4 °C for 2 min. The supernatant was filtered through a 13 mm diameter 0.2 µm polytetrafluoroethylene (PTFE) syringe filter (Whatman GmbH, Dassel, Germany) into a 1.5 ml amber Eppendorf and stored at –80 °C. Pigment separation was performed by reverse-phase HPLC as described in Färber et al. (1997) using a Dionex BioLC HPLC system (Sunnyvale, CA, USA) with a LiChrospher^® 100 RP-18 (5 µm) column (Merck, Darmstadt, Germany). The carotenoids, violaxanthin, zeaxanthin, antheraxanthin, lutein, neoxanthin, and β-carotene, and Chl a and Chl b were detected using 447 nm wavelength, shown in the chromatogram (Supplementary Fig. S3B). Each carotenoid was expressed as a percentage of the total carotenoid pool. The de-epoxidation state (DEPS) and total xanthophyll pool (XC) relate to ratios of violaxanthin, zeaxanthin, and antheraxanthin, calculated as described in Färber et al. (1997).

Experimental design and statistical analysis methods

Experiments were repeated on five separate occasions giving five independent sets of growth experiments for duckweed accessions growing in cabinet conditions. From each of these, three biological replicates were used for each accession–treatment combination for chlorophyll fluorescence measurements totalling 15 replicates. Pigment extraction by spectrophotometry was performed for four replicates of accession–treatment combinations from each independent experiment, maximum n=20. For HPLC analysis, 14 accessions grown in HL and LL were used, forming one data point for each accession–treatment combination. Significance of light treatment, accession, species, and light habitat (derived environmental light dHL or dLL) were determined on each growth rate parameter (RGRlog area, Col RGRlog, FW, and FDW), photosynthetic parameters, and pigment contents using all observation data with two-way ANOVA and Welch’s t-tests. Differences between light treatment responses amongst accession were also assessed (average HL minus LL) for growth, NPQ, φPSII, F_v/F_m, and pigment contents (total Chl a, Chl b, and carotenoids). HPLC carotenoid data were pooled for each accession to find differences between treatments. All data manipulation and analysis was performed in R (v3.6.3) using Rstudio (v1.2.5) with packages ggplot2 (Gómez-Rubio, 2017), corrplot (McKenna et al., 2016), FactoMineR (Lê et al., 2008), and factoextra (Kassambara and Mundt, 2020).

Results

Sites were classified as high light (dHL) or low light (dLL) intensity using distance analysis and K-means clustering utilizing all measured environmental light variables of light intensity PPFD and all spectral compositions from sites across time points (Fig. 3, and data summarized in Fig. 4 with statistics in Supplementary Table S3; see also ‘Collection of duckweed accessions and measurements of environmental parameters’ in the Materials and methods). Eleven sites were determined as LL and eight as HL (Fig. 3), such as ditches and locations under bridges and trees to full-scale open ponds and canals (Fig. 1B, D). Full duckweed coverage was evident in both sun and shade locations, with surface coverage variation between sites and seasons (Supplementary Table S1). Environmental light (Supplementary Table S2) was associated with surface coverage from 19 sites across the seasonal time points (Supplementary Table S1). In autumn, coverage was significantly negatively correlated with UV light (R= –0.53; P=0.02; Supplementary Fig. S4). Noteworthy also were no other correlations between light variables and coverage.

Fig. 3.

Native duckweed sites can be organized into derived from high light (dHL) or low light (dLL) using measured environmental light variables. Environmental light variables (n=50) measured in either µmol m^–2 s^–1 (PPFD, FR, R, G, B, UV) or nm (λp, λd) for 19 sites across three seasons for 2 years were used to group sites by relatedness using a dendrogram. The distance matrix was computed using Manhattan distances and the complete method. The rectangles represent site groupings after K-means clustering set at n=2, where a blue border indicates dLL sites and red a border indicates dHL sites. Coloured circles represent species groupings as shown in the key in Fig. 1.

Fig. 4.

dHL and dLL sites have stark contrasts in photosynthetically relevant light environments. Light measured between different seasons and years from 19 duckweed sites grouped as dHL or dLL showing averages with error bars as ±SD within site groupings. (A) Light intensity and spectral quality were measured for dLL and dHL sites in µmol m^–2 s^–1, for total light PFD made up of regions of light: FR, far-red; R, red; G, green; B, blue; UV, ultraviolet; and for the photosynthetic portion PFFD (B+G+R). (B) Wavelengths of light measured in nanometres were quantified for sites as λp, peak average wavelength; and λd, dominant average wavelength for dLL and dHL. (C) Proportions of spectral quality at each time point were calculated as the proportion of spectral region/total PFD light×100 per site and grouped by dLL and dHL. (D) R:FR was calculated as the ratio of R to FR light from raw values from each site and grouped as dLL or dHL. Bars are coloured by seasonal and yearly time point in chronological order, with dLL on the left and dHL on the right of each panel. Raw and grouped summary data and t-test results are presented in Supplementary Tables S2 and S3.

Seasonal light quantity and spectral quality were markedly habitat dependent

Light intensity (PPFD) and all spectral constituents (FR, R, G, B, UV, and R:FR) were higher in dHL sites than in dLL sites across all seasonal time points (Fig. 4; Supplementary Table S3). The maximum recorded light intensity overall in summer 2022 for dHL sites was 1116 µmol m^–2 s^–1 compared with an almost 10-fold reduction in dLL sites at 109 µmol m^–2 s^–1. dLL sites received higher levels of FR light (750 nm) compared with more G light (550 nm) in dHL, especially in summer as indicated by peak wavelength measurements (Fig. 4B). Light spectral proportions of % FR and % UV were higher in dLL sites whilst % R and % G were higher in dHL sites (Fig. 4C). This spectral shift contributes to substantially lower R:FR ratios in dLL sites consistent with year-round natural canopy shading, but with the greatest differences in summer (Fig. 4D). The % PPFD differed between sites, with dLL receiving less light in the photosynthetically active region (R+G+B) in summer and autumn. For other environmental variables, no significant differences were found between dHL and dLL sites for water and atmospheric temperature measured at each time point or for bioclimatic temperature and precipitation data (Supplementary Table S4). Such contrasting differences in light intensity and spectral quality indicate a marked difference in year-round habitats which present different challenges for light acclimation.

Lemna species were commonly identified in the UK duckweed panel

Species were determined using clustering of genome sequences employing known sequenced individuals as controls (Fig. 5). Five species were identified in the UK cohort: the majority were L. japonica (n=11), followed by L. minor (n=5), L. minuta (n=5), L. turionifera (n=2), and S. polyrhiza (n=1). Indeed, putative species clustered together with their respective relatives, namely L. turionifera (6002), S. intermedia (9394), L. minuta (9260), L. minor (7016, 7194, 7210, and 9441), and L. japonica (7123). In addition incongruence was noted with L. japonica 7868 and L. minor 9541, which are evidently switched between species clustering patterns, indicating that these samples have been misidentified previously as these two species are notoriously difficult to distinguish (Volkova et al., 2023). Individual accessions clustered into expected species groupings that corresponded with those from independent phenotypic assessments for the UK accessions. Lemna japonica individuals formed a cluster between L. minor and L. turionifera (Fig. 5A), and the same relationship was also apparent from fastStructure analysis, suggesting that L. japonica individuals arose from hybridization between L. minor and L. turionifera (see Fig. 5B). Divergence between KS12 from S. intermedia 9394, as well as high anthocyanin accumulation led to the classification of this accession as S. polyrhiza (Fig. 5; Supplementary Fig. S5; Fang et al., 2023, Preprint).

Fig. 5.

Duckweed species differentiated by genome alignment and allele frequencies used to classify the UK duckweed panel. UK duckweed short read sequences and known species sequences were aligned with L. minor 7210 and filtered down to 4-fold degenerate coding region single nucleotide polymorphisms (SNPs), consisting of >550 000 variants presumed to be under neutral selection. (A) Phylogenetic tree of duckweed accessions with n=5 clusters of duckweed species. Lemna minor accessions group with L. minor 7016, 9441, 7194, and 7210, but also with L. japonica 7868. Lemna japonica individuals cluster with L. japonica 7123 and L. minor (japonica?) 9541. Lemna minuta individuals (n=5) distinctly cluster with L. minuta 9260, L. turionifera (n=2) group with L. turionifera 6002, and KS12 clusters with S. intermedia 9394, but with some genetic differentiation, and is considered to be S. polyrhiza. Coloured circles represent dLL or dHL ecotypes relating to habitat light environment. (B) fastStructure analysis showing the ancestral genetic makeup of each duckweed individual used to classify species (n=24 UK) along the x-axis, with Bayesian probability of population allocation on the y-axis. fastStructure confirms duckweed species and presence of the hybrid L. japonica species composed from L. minor and L. turionifera parents. K=4 was used to determine the number of populations by which to colour the structure plot. Each species group is coloured as shown in the key in Fig. 1. Representative images of each species are given above the structure plot.

Lemna species were differentiated between both dHL and dLL environments, whilst a single Spirodela representative, S. polyrhiza, was found only at a dHL site (Figs 1B, 3, 5). There was high variation in species wild seasonal growth patterns. Lemna minor (e.g. KS13, KS27, and KS29) and L. japonica (KS17 and KS21) accessions from a mixture of dHL and dLL sites did not exceed 20% surface coverage across all time points, with low site coverage averages <5%, indicating they were often not present or only growing as sporadic colonies (Supplementary Table S1). In contrast, L. japonica KS03 and L. minuta KS06B from dLL had the highest average coverage of all sites overall, and maximum coverage in early spring and summer. The highest surface coverage in the wild in different seasons/year time points were all from dLL sites, including KS06B, KS25, KS18, and KS02, showing superior maintenance of L. minuta, L. japonica, and L. minor accessions in LL environments all year round. The exception was the KS12 Spirodela accession in summer, which had comparably high surface coverage as a dHL site, highlighting different species prevalence across extremes of seasonal light and temperature in the UK.

Duckweed growth rate in controlled conditions is dependent on light intensity, species, and original habitat light environment

Differences in RGR were derived from growth curves in each light treatment measured by area or colony counts (Supplementary Fig. S1). Duckweeds generally had higher RGRs during the log phase of growth in LL compared with HL (ANOVA, P≤0.001, Supplementary Table S5A, B; Fig. 6A, B, E, F). On average, RGRlog by area was 0.77, and 0.37 by colonies in LL compared with 0.65 and 0.28 in HL (P≤0.05, Welch’s t-test). For 18/24 accessions, HL negatively affected growth rate, three accessions had increased growth rate in HL, and three were unaltered (Supplementary Fig. S6A, B). Moreover, FW and FDW biomass positively correlated with RGRlog, especially in HL where FW/FDW and FW/RGRlog area R=0.88, P≤0.0001, and FDW/RGRlog area R=0.84, P≤0.001 were found (Fig. 9A; Supplementary Fig. S7).

Fig. 6.

Species show differences in growth rate in LL, but in HL the dLL accessions grow faster than the dHL accessions. Each pair of plots represent different growth rate parameters: RGRlog area, Col RGRlog, FW, and FDW (mg), and display LL treatment response on the left (grey) and HL on the right (red) of each panel. (A–D) Species differences coloured as shown in the key in Fig. 1. Lowercase letters indicate significant species differences within each light treatment by post-hoc Tukey test P≤0.05. (E–H) Light habitat effects by grouping accessions as dLL (blue) or dHL (red) for each boxplot. Welch’s t-test is indicated above, and significant differences are shown by *P≤0.05 and **P<0.01, with insignificant differences marked with n.s. The midlines on boxplots indicate the median and 25% and 75% quartile boxes; all observations are displayed as points.

Fig. 9.

(A) Fast growth in HL is associated with photosynthetic pigment contents. Principal component analysis (PCA) showing association of physiological responses from 24 accessions grown in HL and LL. The Cos2 scale shows how strongly the variables contribute to the dataset variability, and length and directions of arrows indicate the strength and relationships between variables. (B) Fast growth in HL is best achieved in dLL L. japonica individuals. Duckweeds broadly separate by HL response (top/bottom) and group into species (left/right) when using light physiological responses. PCA plot showing the relationship between accessions in the condensed landscape of 42 physiological variables under HL and LL treatment. Individual accession points are labelled and coloured by dHL (red) and dLL (blue). Centroids are marked with crosses for dHL and dLL accessions and represent averages for all physiological variables grouping by original habitat. Coloured ellipses represent the species types as shown in the key in Fig. 1. Ellipses overlap between L. minor and L. japonica species.

Species and light habitat were both significant for differences in RGRlog area and RGRlog colony gain by ANOVA (P≤0.0001 and P≤0.01 respectively, for RGRlog area), with an interaction for RGRlog area (Supplementary Table S5A, B). Growth responses in HL or LL were split, to consider the effects of species and derived environmental light (dHL or dLL) separately for light acclimation (shown in Fig. 6). In artificial LL, dHL and dLL accessions grew at the same rate. In HL, original light habitat had significant effects on growth rates by RGRlog area and RGRlog colony gain (Fig. 6E, F; Supplementary Fig. S6A, B), whereby the dLL accessions outperformed dHL accessions for growth rate (Supplementary Fig. S6A, B). There were no differences by colony gain, FW, and FDW between species in HL. Between treatments, L. japonica maintained the fastest growth by RGRlog area, and L. minuta growth was most severely affected by HL (Fig. 6A).

In LL, L. japonica and L. minor grew faster by area gain and produced more biomass than L. turionifera (Fig. 6A, C). Lemna japonica also produced colonies faster in LL relative to other Lemna species, with S. polyrhiza producing the least (Fig. 6B).

Photosynthetic processes are modulated by light intensity

Light response curves are presented for φPSII and NPQ for duckweed accessions grown in LL and HL treatments (Supplementary Fig. S2). At light levels equivalent to the LL growth condition (130 µmol m^–2 s^–1), φPSII had already declined by 50%, with a further 20–30% decrease at 365 µmol m^–2 s^–1 corresponding to HL levels. Linear regression models for φPSII between 130 µmol m^–2 s^–1 and 365 µmol m^–2 s^–1 showed a higher slope and lower intercept in HL-grown plants compared with LL-grown plants (Supplementary Fig. 2A). NPQ increased with increasing light levels, with a maximum at >1000 µmol m^–2 s^–1 (Fig. 7B). LL plants maintained a higher photosynthetic efficiency (φPSII) while a greater photoprotective capacity (NPQ) was observed in the HL plants, with an average NPQ of 2.98 in HL and 2.33 in LL at the maximum light intensity (P≤0.0001, Welch’s t-test).

Fig. 7.

Photosynthetic parameters φPSII, F_v/F_m, and NPQ display unique treatment–species effects at different light levels. (A) Boxplots showing φPSII and (B) NPQ measured in duckweed species at 130, 365, and 1050 µmol m^–2 s^–1. (C) F_v/F_m measured quantum yield in the dark for duckweed species grown in two light treatments. All panels are split by treatment LL (grey) and HL (red), and all boxplots show the median and 25% and 75% quartiles, with all individual points plotted. Significant differences between species within each light treatment are indicated by lowercase letters at the top or bottom of plots by Tukey post-hoc test <0.05, and differences between the same species grown in LL and HL are indicated with bars between them and marked with asterisks.

The maximum photosynthetic efficiency measured in the dark (F_v/F_m) and maximum NPQ were strongly affected by light treatment, accession, and species (ANOVA, P≤0.0001) and for F_v/F_m (ANOVA, P≤0.003; Supplementary Fig. S2C). For F_v/F_m, the interaction between factors was also important, showing both genetic and light treatment effects (Table 1; Fig. 7C). The original light habitat did not influence photosynthetic parameters in LL or HL, whilst dHL and dLL accessions had no clear differences in HL response (Table 1; Supplementary Fig. S6C, D).

Table 1.

Photosynthetic efficiency, NPQ responses, and quantum yield of photosynthesis at low and high light

	Treatment	Accession	Species	Light habitat	Treatment×Accession	Treatment×Species	Light habitat×Species
A. φPSII
φPSII 130 PPFD
F P	12.1 0.0005	1.6 0.04	4.4 0.001	1.1 0.3 ns	1.5 0.04	2.4 0.05	0.4 0.75 ns

φPSII 365 PPFD
F	11.0	1.8	5.1	2.5	1.4	1.5	0.4
P	0.0009	0.01	0.0005	0.11 ns	0.10 ns	0.22 ns	0.79 ns

φPSII 1000 PPFD
F	0.3	1.5	5.0	1.7	0.9	1.4	0.6
P	0.55 ns	0.06 ns	0.0007	0.18 ns	0.54 ns	0.24 ns	0.63 ns

B. NPQ
NPQ 130 PPFD
F P	26.6 <0.0001	5.1 <0.0001	19.3 <0.0001	0.03 0.8 ns	1.8 0.01	2.2 0.06 ns	1.0 0.4 ns
NPQ 365 PPFD
F	34.1	4.9	14.8	0.02	2.2	3.1	1.1
P	<0.0001	<0.0001	<0.0001	0.9 ns	0.001	0.02	0.3 ns

NPQ 1000 PPFD
F	121.0	4.3	11.5	0.5	2.5	6.4	3.1
P	<0.0001	<0.0001	<0.0001	0.5 ns	0.0002	<0.0001	0.03

C. Fv/Fm
F v /F m 0 PPFD
df	1	23	4	1	23	4	3
F	31.9	2	7.1	1.5	1.8	3.3	0.8
P	<0.0001	0.003	<0.0001	0.2 ns	0.01	0.01	0.5 ns

Photosynthetic efficiency (φPSII), NPQ responses, and quantum yield of photosynthesis (F_v/F_m) vary between accessions, species, and light treatments at corresponding low (130 μmol m^–2 s^–1) and high light (365 μmol m^–2 s^–1). φPSII measured at maximum light (1050 μmol m^–2 s^–1) was only sensitive to species differences, and light habitat does not appear to affect photosynthetic parameters at any light level. Parameters were derived from chlorophyll fluorescence.

ANOVA results for single factors and interactions of factors using significance as P≤0.05. Non-significant results are reported as ns.

Photosynthetic processes show species-specific differences

Photosynthetic responses differed between species, with sensitivity to both growth light level and measurement light level. Lemna minor had naturally higher F_v/F_m and φPSII than L. minuta in LL and HL, in fact L. minuta had the lowest F_v/F_m of all species and concurrent low φPSII, showing different achievable PSII quantum yields and capacity for light acclimation between species (Fig. 7A, C). The severity and direction of changes in F_v/F_m between light levels also differed between species. Lemna japonica had high F_v/F_m in LL, with a reduction in HL, whilst S. polyrhiza had comparable F_v/F_m with other species in HL, but lower photosynthetic efficiency in LL (Fig. 7C). Spirodela polyrhiza also showed an additional species-specific acclimation of photosynthesis with atypical improvement of φPSII at the maximum light intensity >1000 µmol m^–2 s^–1, in line with the observation that it is well adapted to high light (Fig. 7A).

Species differences for NPQ in LL-grown plants were also common, with L. japonica demonstrating higher inherent capacity for NPQ than L. minuta and S. polyrhiza (Fig. 7B). In HL-grown plants, when NPQ is normally higher, L. japonica accessions retain higher NPQ compared with all other species at specific light levels. Moving from low to high light, NPQ in L. minor rose whilst it declined in S. polyrhiza.

Higher F_v/F_m in HL was associated with faster growth in both light conditions; therefore, photoinhibition may be strongly associated with high rates of growth in HL (Fig. 9; Supplementary Fig. S7). Maintenance or increases in F_v/F_m in addition to high φPSII in HL relative to LL occurred in S. polyrhiza and L. turionifera, concordant with relatively better growth rates in HL relative to LL (Figs 6A, B, 7A, C; Supplementary Fig. S6A–C). The highest F_v/F_m values throughout were in fast growing L. minor accessions in LL, and the lowest F_v/F_m throughout were in L. minuta, which showed markedly reduced area gain in HL (Figs 6A, 7C). NPQ was not directly associated with high growth but, as L. japonica had high overall NPQ and high growth rates in LL and HL, it may be linked through species effects.

Changes in pigment concentrations occur in duckweed during light acclimation

Total chlorophyll and carotenoids measured spectrophotometrically were affected by light level, species, accessions, and habitat light environment (summarized in Supplementary Table 5C, D). Light treatment had significant effects (ANOVA, P≤0.0001), with total chlorophyll decline and carotenoid increase as shown in Supplementary Fig. S6E, F. T-tests showed that LL-grown accessions had higher Chl a and b, whilst HL-grown accessions had higher carotenoids and the carotenoid:chlorophyll (Car:Chl) ratio was also higher. Although there were clear pigment alterations in responding to the light treatments in terms of total chlorophyll and Car:Chl, duckweed accessions did not increase Chl a:b ratios to acclimate to HL (Fig. 8). The role of light habitat and species effects on pigment changes was further dissected for HL responses.

Fig. 8.

Chl a, Chl b, total chlorophyll, and carotenoid contents are higher in dLL duckweed accessions than dHL accessions when grown in HL. (A–F) Pairs of boxplots for Chl a, Chl b, Chl a+b, and carotenoid contents, and Chl a:b and Car:Chl ratios of accessions grown in HL (red) and LL (grey) treatment and coloured by dLL sites (blue) or dHL sites (red) on the x-axis. All P-values by Welch’s t-test are reported, and *P<0.05 and **P<0.01 indicate significance of environmental light, with insignificant differences marked with n.s. Pigment content is different between species at LL but not different in HL. (G–L) Paired boxplots show species effects on each pigment measurement in two light treatments, LL (grey) and HL (red). Lowercase letters at the top or bottom of plots represent significance by Tukey post-hoc test <0.05 between species within the treatment. Differences between the same species grown in LL and HL are indicated with bars between them and marked with asterisks. The midline on boxplots indicates median and 25% and 75% quartile boxes, and all are observations plotted as points. n=20 for each accession–treatment combination.

In a separate analysis, carotenoids were quantified using HPLC, pooling leaf samples so that statistical analysis was possible for light treatment effects only (Supplementary Fig. S3.). Here, xanthophyll cycle (XC) carotenoids increased under HL (ANOVA, P≤0.0001), especially zeaxanthin (ANOVA, P≤0.0001) and antheraxanthin (ANOVA, P≤0.05), with a reduction in violaxanthin (ANOVA, P≤0.05). The de-epoxidation state (DES) of the XC increased in HL, showing greater conversion to zeaxanthin in HL-grown plants generally, with high levels of DES from 43% in LL to 67% in HL, indicating relatively excessive light levels in both LL and HL in comparison with other higher plants under field conditions (e.g. Murchie et al., 1999). In this dataset, we note that Chl a:b was lower in LL-grown plants overall, indicating that acclimation of light-harvesting complex antenna size may have occurred.

Chlorophyll and carotenoid changes required for high light acclimation were better optimized in accessions from low light original habitats

Pigment content and the ratios normally associated with acclimation to HL differ between accessions from different light habitats. The dLL accessions acclimated to HL by increasing carotenoid and maintaining total chlorophyll, as shown by higher Chl a, Chl b, total Chl (a+b), and carotenoids in HL (Fig. 8A–C, E; Supplementary Fig. S6E, F). This response was less pronounced in dHL accessions. Interestingly, dHL accessions showed an increase in Chl b relative to dLL accessions only in LL. There were no significant differences in Chl a, total chlorophyll, and carotenoids between dHL and dLL accessions grown in the LL condition and there was no difference in overall Chl a:b or Car:Chl ratios between dHL or dLL in either treatment, which would typically be expected differences between sun- or shade-tolerant plants.

Pigment composition is naturally variable between species in controlled low light conditions

For LL-grown plants, pigment content varied among species, but this was weaker in HL (Fig. 8G–L). In LL, L. japonica had the highest Chl a:b and L. minuta the lowest, indicating sun- and shade-tolerant adaptations between species at low light. Spirodela polyrhiza and L. minuta had the highest Car:Chl in LL, with all four Lemna species showing significant increases due to treatment, whilst S. polyrhiza was unaffected. Spirodela polyrhiza was notably the only species with anthocyanin accumulation in response to HL treatment, contributing to its unique species adaptation to light (Supplementary Fig. S5A).

Low light-derived accessions acclimated to high light with higher pigment concentrations and strong growth performance

When accessions are plotted on the landscape of all physiological variables in light response, there is broad separation by environmental light groupings (Fig. 9B). The fastest growing accessions in HL were all L. japonica from dLL sites: KS03, LY02, LY03, and KS18, characterized by high chlorophyll and carotenoid contents in HL (Fig. 9). Lemna japonica were also fastest growing in LL, with high photoprotection via NPQ in both light intensities. Maximum quantum yield of photosynthesis in the dark (F_v/F_m) increases in S. polyrhiza in HL relative to LL and coincides with faster growth by colonies in HL (Fig. 6B). The centroids for dLL and dHL groupings show separation by performance in HL primarily driven by variation in growth rate and pigment composition (Figs 6, 8, 9).

Discussion

The relationship between light acclimation mechanisms and habitat has been relatively well studied in higher plants, but such information is less available for duckweeds, with their aquatic habitat complicating simple categorization. Typically, duckweed collections in stock centres are limited by unknown date of collection, specificity of collection locations, and environmental data, hampering extensive studies (Sree and Appenroth, 2020). Here we generated a novel duckweed collection from diverse sites in North and South UK alongside detailed habitat data. We then explored variation in light acclimation in controlled environments between accessions of five species from different light habitat types.

Colonization of duckweed in low light natural environments persisted all year round

The highest plant coverage was maintained all year round across dLL but not dHL sites. The differences between HL and LL habitats represented largely expected features. Diversity in light quantity by scattering and modified quality of green and FR light enhancement is caused by the presence of overhead vegetation (Lee et al., 1996; De Castro, 2000; Burgess et al., 2021). Previously, LL sites characterized by tree shading were proposed to also provide temperature protection for duckweeds and additionally contribute to higher nutrient injection into water from decaying biomass (Landolt, 1986; Landolt and Kandeler, 1987). Here, irradiance and spectral compositions were all significantly different between dHL and dLL sites all year round. Shade-enriched components FR and green are able to drive photosynthesis in well-acclimated plants in very low light conditions (Smith et al., 2017; Zhen and Bugbee, 2020). This may have contributed to dLL sites having the highest duckweed coverage of all seasons. The notable exception was S. polyrhiza KS12 from a dHL site which displayed high coverage in summer (Supplementary Table S1; Fig. 1B, C). UV was the only light spectral region to correlate negatively with duckweed coverage and, even in controlled conditions even at low levels, produces a stress response in L. minor (Farooq et al., 2000). No differences in temperature were noted between dHL and dLL sites, so we conclude that HL and species-specific acclimation were important drivers of adaptation in HL sites.

Low light was more supportive for fast growth than high light in controlled conditions

Growth rates were generally higher in controlled LL. Both low (<100 µmol m^–2 s^–1) and high light intensities have been cited as beneficial for biomass and growth rates in different duckweed species (Classen et al., 2000; Cheng et al., 2002; Paolacci et al., 2018; Stewart et al., 2020). At the same time, L. gibba grown in extremes of 50 µmol m^–2 s^–1 and 1000 µmol m^–2 s^–1 grew at the same rate, supporting better light use efficiency at LL (Stewart et al., 2021). Conclusions for species as a whole have been drawn from single clones of L. minuta, L. minor, and L. gibba grown in HL (Paolacci et al., 2018; Stewart et al., 2020). However, varied RGR measurement methods by area or biomass as well as different starting densities, experimental duration, and frequency of measurements challenge comparison of growth data between species across different studies. Optimal light intensities for growth are still under debate, and here we show that they are also affected by collection origin. In the LL treatment, L. japonica species had the natural growth advantage and L. turionifera and S. polyrhiza showed slower growth in LL. In the HL treatment, growth was faster in dLL accessions. We hypothesize that a survival or stress tolerance strategy in dHL accessions to local adaptation of HL may be at play. Similar established trade-off strategies between growth and survival can be seen for plants in nature (Pierce et al., 2017; Zhang et al., 2020). We suggest that this might be genetic in origin rather than epigenetic due to the multiple generations that took place between collection and experimentation (Huber et al., 2021; Antro et al., 2022).

Growth and acclimation are associated with species-specific responses

In addition to species differences in LL growth rate, we now show a dependence on habitat origin for HL growth rate. Lemna japonica is of note: it showed high growth in both LL and HL, higher Chl a and Chl b in LL, and highest Chl a:b and highest NPQ in LL and HL. Lemna japonica are hybrids between L. minor and L. turionifera reported here and recently in Braglia et al. (2021) and Volkova et al. (2023).

The turion-producing species L. turionifera (KS16 and KS22) and S. polyrhiza (KS12) were slow growing in LL but showed enhanced growth in HL. Additionally, effects on turionating capacity were evident between accessions and in response to light (Supplementary Fig. S5B). Lemna turionifera KS16 and S. polyrhiza KS12 also had higher F_v/F_m in HL than in LL. As the only Spirodela accession, KS12 appeared to have a distinct photosynthetic acclimation response profile. In its wild habitat, KS12 has high coverage in summer and is visibly red, and was the only accession to visibility accumulate anthocyanin in our controlled experiment (Fig. 1C; Supplementary Fig. S5A). Anthocyanin and flavonoid genes are more expansive in S. polyrhiza and less prevalent in Lemna species, suggesting alternative mechanisms for photo-oxidative stress tolerance (Landolt, 1986; Davies et al., 2022; Fang et al., 2023, Preprint).

Maximum quantum yield is closely linked to growth in high light

F _v/F_m is normally associated with photoinhibition and correlated positively with higher HL growth rate. The ability to limit photoinhibition is likely to contribute to the success of the dLL accessions in HL conditions. Thus it may be more important in this context to assess relative rates of damage to PSII and rate of repair, as well as the ability to cope with irreversible damage. Recovery from photoinhibition is also promoted by other mechanisms such as antioxidant production which deserve further attention in duckweeds. Related to this, accumulation of flavonoids such as anthocyanins is induced by abiotic stresses in S. polyrhiza (Landolt, 1986; Böttner et al., 2021).

Pigment responses aid light acclimation in fast growing accessions

Chlorosis in duckweed and increases in carotenoid levels can occur in response to light up to 1000 µmol m^–2 s^–1 (Paolacci et al., 2018; Stewart et al., 2020). Overall, plants from dHL lost more chlorophyll and carotenoids in HL than plants from dLL, suggesting that chlorosis was also part of the adaption between sites.

Acclimation to HL is often accompanied by an increase in Car:Chl, as the XC pool size increases, and an increase in Chl a:b as light-harvesting antenna size decreases. Here, growth in HL did induce a higher Car:Chl across all Lemna species and this also negatively correlated with growth rate, perhaps consistent with the importance of limiting PSII inactivation and the need for a higher F_v/F_m for high growth rates in HL. Unchanging Chl a:b ratios have been shown for Lemna clones already (Paolacci et al., 2018; Stewart et al., 2020) and also for other species such as barley (Murchie and Horton, 1997; Zivcak et al., 2014). Correlation of total Chl a and Chl b with HL growth rate indicates that total chlorophyll is an important attribute for light acclimation.

We note that Chl a:b was inconsistent in comparison with HPLC data which used pooled samples. Whilst spectrophotometry reports a Chl a:b ratio of ~3.5 and HPLC a ratio of 2.7 or 1.9 depending on the light condition, we expect that these differences in Chl a:b arose from differences in sample preparation. Whilst freeze-drying gives added accuracy when measuring pigments by mg g^–1 (Lichtenthaler and Buschmann, 2001), low temperature and vacuum treatment can degrade chlorophyll (King et al., 2001), especially Chl b, which could lead to an overestimation of Chl a:b from freeze-dried tissue (Lashbrooke et al., 2010). Relative species differences in Chl a:b ratios indicate that L. japonica was more typical of sun tolerance and L. minuta of shade tolerance. The HPLC data also showed high de-epoxidation rates in HL consistent with this being a highly light-saturating condition and further emphasizing the role of photoinhibition in determining growth.

Reduced light intensity, extremely different peak wavelengths of light and altered proportions of spectral quality, including higher FR, and reduced R and G, would be expected to affect light acclimation and chlorophyll content of accessions from LL habitats. Indeed, shade-tolerant plants grown in controlled HL had higher chlorophyll and carotenoid contents per dry weight which correlate with HL growth. dHL accessions may be less sensitive to controlled HL, previously experiencing >1000 µmol m^–2 s^–1 in their habitats, and may acclimate in other ways, linked to a survival strategy rather than a dominant or competitor strategy, characterized by increased growth. The exception is species-specific acclimation as seen in S. polyrhiza here. Related to this, we anticipate that the dHL conditions may have been hostile enough to induce a range of stress tolerance responses which we observe as F_v/F_m reduction and pigment composition alterations.

Applications to agriculture

The species and habitat of duckweeds are important in selection for commercial purposes as environmental light contrasts were substantial. Increased growth, chlorophyll retention, and carotenoid gain were greater in L. japonica dLL ecotypes in increased light conditions compared with dHL ecotypes. Further, we suggest that dHL ecotypes are less suitable for sustainable vertical farming systems, with the exception of S. polyrhiza, which had species-specific strategies for light adaptation, including anthocyanin accumulation.

Concluding remarks

We have focused here on accession-level and species-level variation in light adaptation, revealing widespread natural variation and broad local adaptation within species. Ecotypes from LL environments showed better acclimation of growth and pigment content to controlled HL. Pigment composition may be important in determining overall photosynthetic, growth and photoinhibitory traits because higher chlorophyll and carotenoid content in controlled conditions were related to LL habitats, suggesting biochemical or structural adaptation. How such adaptation occurs is intriguing, given the paucity of information on duckweed natural variation and adaptation to varied environmental factors. Our work provides a first step towards understanding environmental factors that are likely to select for genetic variation relevant also to subsequent breeding of duckweed accessions most useful by virtue of their distinct potentials.

Supplementary data

The following supplementary data are available at JXB online.

Fig. S1. Duckweed growth rates are affected by light intensity.

Fig. S2. Duckweed photosynthetic parameters are affected by light intensity.

Fig. S3. Carotenoid pigments show differential profiles in duckweeds in response to light treatment.

Fig. S4. Negative relationship between duckweed surface coverage and autumn UV light intensity.

Fig. S5. Species-specific responses to light treatment include anthocyanin production and altered turion formation.

Fig. S6. Ecotypic variation in light adaptive responses. Proportional changes in growth by area, colony gain, changes in F_v/F_m, NPQ, and pigment content between light treatments differ between accessions.

Fig. S7. Linear relationships between physiology of growth and photosynthetic responses in HL and LL treatments.

Table S1. Duckweed UK sites have five identified species, and site coverage varies across season and year time points.

Table S2. Seasonal light variables used to characterize dHL and dLL groupings.

Table S3. Grouped light variables into dHL and dLL sites show extreme light environments across seasons.

Table S4. Temperature and rainfall data across dHL and dLL sites indicate homogeneous conditions.

Table S5. Relative growth rate by area and colony gain and pigment contents between light treatments vary between accessions, original light habitats, and species.

Author contributions

KES: performing field collection, lab experiments, and data analysis; LC and BT: aiding with pigment extractions; LM: writing the code for chlorophyll fluorescence experiments; MH: writing the wrapper for short read data processing; LY: performing field collections; EHM: supporting the project; KES and EHM: conceptualization and writing the manuscript, with all authors contributing to the final manuscript.

Conflict of interest

The authors declare that there is no conflict of interest.

Funding

KES is supported by a Biotechnological and Biological Sciences Research Council (BBSRC) PhD scholarship (BB/M008770/1). The work was supported by the University of Nottingham Future Food Beacon of Excellence. EHM receives funding from the Biotechnological and Biological Sciences Research Council (BBSRC) through grant number (BB/S012834/1).

Data availability

The sequences for duckweed genomes in this panel are deposited under project PRJNA1026139 on the NCBI Sequence read archive (SRA).