Flowering and Seed Production across the Lemnaceae

Plants in the family Lemnaceae are aquatic monocots and the smallest, simplest, and fastest growing angiosperms. Their small size, the smallest family member is 0.5 mm and the largest is 2.0 cm, as well as their diverse morphologies make these plants ideal for laboratory studies. Their rapid growth rate is partially due to the family’s neotenous lifestyle, where instead of maturing and producing flowers, the plants remain in a juvenile state and continuously bud asexually. Maturation and flowering in the wild are rare in most family members. To promote further research on these unique plants, we have optimized laboratory flowering protocols for 3 of the 5 genera: Spirodela; Lemna; and Wolffia in the Lemnaceae. Duckweeds were widely used in the past for research on flowering, hormone and amino acid biosynthesis, the photosynthetic apparatus, and phytoremediation due to their aqueous lifestyle and ease of aseptic culture. There is a recent renaissance in interest in growing these plants as non-lignified biomass sources for fuel production, and as a resource-efficient complete protein source. The genome sequences of several Lemnaceae family members have become available, providing a foundation for genetic improvement of these plants as crops. The protocols for maximizing flowering described herein are based on screens testing daylength, a variety of media, supplementation with salicylic acid or ethylenediamine-N,N′-bis(2-hydroxyphenylacetic acid) (EDDHA), as well as various culture vessels for effects on flowering of verified Lemnaceae strains available from the Rutgers Duckweed Stock Cooperative.


Introduction
The five genera and 36 species comprising the Lemnaceae family, commonly known as duckweeds, are the smallest, fastest growing, most morphologically reduced, and widely distributed family of angiosperms [1][2][3]. These aquatic monocots are found floating and rapidly clonally dividing on still, nutrient rich, waters worldwide. Duckweeds are rare among plants for being a complete protein source, with an amino acid content similar to eggs [4,5]. Wild or greenhouse grown Wolffia species are an especially resourceefficient food source. Protein concentrates from Lemna species are being explored as a scalable and economical way to meet the rapidly increasing demand for plant-based proteins [6,7]. Members of the Lemnaceae are promising industrial crops that can grow at rates of 13-38 dry tons/hectare per year, compared to 6-9 and 2.5-3.5 dry tons/hectare per year for maize and soybean, respectively, reported by the FAO [8,9]. Duckweeds can grow on non-arable land in agricultural or industrial wastewater, recapturing nutrients instead of needing energy intensive fertilizers, without pesticides, while cleaning the water and producing a biomass that can be used as animal feed or biofuel [6][7][8]. Research into this family's unique biology will result in improving the ways we use these distinctive plants to sustainably provide clean water, food, and energy in the future.
The Lemnaceae plant body generally consists of a small (0.5 mm-2 cm), flat leaf-like structure called a frond. Plants in the Spirodela, Landoltia, and Lemna genera have rhizoid Members of the Lemnaceae grow in a juvenile and asexual manner indefinitely, and then produce flowers in a matter of days if exposed to the correct stimuli. This allows precise study of individual stimuli or genetic pathways which initiate floral development, rather than the earlier or later flowering phenotypes observed in Arabidopsis and other commonly studied annual plants, where regulation of flowering is complicated by multiple active pathways [43,44]. The Lemnaceae were used to search for the mythical "florigen", first hypothesized in 1937 [45]. Landolt and Kandeler's multivolume monograph, The Family of Lemnaceae-a monographic study Volumes 1 and 2 [1,3] and a later review by Pieterse [46] describe 112 studies to investigate factors affecting floral regulation using duckweed. Day-length was paramount in every study, while salicylic acid (SA), a plant hormone now understood as important for abiotic stress response and pathogen resistance, was the most studied chemical inducer. Chelating agents such as ethylenediamine-N,Nbis(2-hydroxyphenylacetic acid) (EDDHA), were also frequent players in those studies.
While many studies of the duckweeds describe ways to induce flowering in the laboratory, there are only two reports of seed production in the laboratory. This is important as the production of seed underlies establishment of breeding protocols to improve agricultural traits. In the 1950s, Maheshwari described the anatomy of W. microscopica and L. paucicostata (now L. aequinoctialis) seed collected from the wild, showing the presence of a cellular endosperm [47]. A study of IAA accumulation in a tissue culture induced large mutant line of L. gibba G3 [16] described that gentle shaking of flowering plants growing on liquid "E" medium resulted in production of seed, which fell to the bottom of the culture flask and germinated when placed in fresh media. Recently, Fu et al., 2017 [48] were the first to produce hybrid duckweed seed by manually crossing under a dissecting microscope, two strains of L. gibba G3 flowering on modified Hoagland's (MH) medium containing 20 µM SA. In that study, only one of the strains used in the cross, 7741, was able to produce viable pollen. Strain 7741 was manually crossed to a male sterile strain to produce the first hybrid strain created in a laboratory.
The long history of research using duckweeds and the rapidly increasing omics resources available make these plants attractive for use in exploring basic biological questions related to the neotenous, aqueous lifestyle, while simultaneously accelerating the domestication of the fastest growing plants into a sustainable crop species. Yet, one could argue that no duckweed species would ever be a "model organism" unless there are robust flowering protocols and the ability to produce seed. We describe here protocols for reliably obtaining flowering cultures of three genera, that will enable development and floral regulation studies, and facilitate development of protocols for production of improved varieties, whether by classical breeding or gene editing. Such protocols are fundamental to performing the mutant screens, genetic experiments and genome wide association studies that have facilitated research in other model plants, and the commercial breeding of other crops.

Spirodela Polyrhiza
Initial screening experiments involved three media, various SA or EDDHA concentrations and combinations, and several light regimes. Guided by previous work on other Spirodela strains [49,50], we initially screened 3 media: E; Hoagland's (Hg); and Schenk Hildebrandt (SH) (Table S1) with and without SA and EDDHA over 28 days in 16 h light:8 h dark long days (LD), or continuous light (CL) for induction of flowering (Table S2). Flowering was very sparse or completely absent for both genomically characterized strains 7498 and 9509 on any medium in LD. Neither strain flowered on E (pH 4.6), or E at pH 5.8, although a senescent culture of 9509 flowered on E supplemented with 25 µM EDDHA. In these screens, the highest flowering rate, 4%, was seen with strain 7498 on SH and Hg, supplemented with 1.5 or 2.0 µM SA under CL. By day 14, strain 9509 was senescing on SH, so screening was continued with Hg only, and all later experiments were performed using Hg. A second screen was performed to test whether using different culture vessels affected flowering. Strain 7498 exhibited higher flowering rates than 9509 in any culture vessel. Cultures in Petri dishes and larger flasks marginally outperformed those in small flasks. In this set of screening conditions, the optimal SA concentrations were 1.0 µM and 1.5 µM (Table S2).
Although a low red/far-red light ratio accelerates the flowering response in Arabidopsis, [51], screens with cultures growing with supplemental far-red light failed to produce flowers (Table S2). Shifting cultures of strain 7498 from LD or CL to Petri dishes under 12 h light:12 h dark led to 9% of fronds flowering with some dehiscent anthers (Table S2). Replicated testing of strain 7498 with these conditions showed that visible flowering started at day 11, and peaked at day 14 with a maximum rate of 6% independent of whether the inoculant was grown in LD or CL ( Figure S1).
Using 4 week old cultures as a source of inoculant improved flowering rate. In Hg medium, Petri dishes of strain 7498 inoculated with three 3-5 frond colonies from one month old cultures growing in Hg medium, began flowering after 10 days in CL ( Figure 1). By day 24, strain 7498 had a significantly higher flowering rate than strain 9509 the 1.5 µM SA condition (p < 0.01). By day 28, when the culture filled the dish, 11 ± 1% of the 7498 fronds growing on media containing 1.5 µM SA were flowering, significantly different from 2.0 µM SA (p < 0.01). Under the same conditions, only 1% of strain 9509 fronds were flowering ( Figure 1). Both strains produced flowers with pistils secreting stigmatic fluid as well as flowers with dehiscent anthers, although 7498 did so at a higher rate ( Figure 2). at pH 5.8, although a senescent culture of 9509 flowered on E supplemented with 25 μM EDDHA. In these screens, the highest flowering rate, 4%, was seen with strain 7498 on SH and Hg, supplemented with 1.5 or 2.0 μM SA under CL. By day 14, strain 9509 was senescing on SH, so screening was continued with Hg only, and all later experiments were performed using Hg. A second screen was performed to test whether using different culture vessels affected flowering. Strain 7498 exhibited higher flowering rates than 9509 in any culture vessel. Cultures in Petri dishes and larger flasks marginally outperformed those in small flasks. In this set of screening conditions, the optimal SA concentrations were 1.0 μM and 1.5 μM (Table S2).
Although a low red/far-red light ratio accelerates the flowering response in Arabidopsis, [51], screens with cultures growing with supplemental far-red light failed to produce flowers (Table S2). Shifting cultures of strain 7498 from LD or CL to Petri dishes under 12 hr light:12 h dark led to 9% of fronds flowering with some dehiscent anthers (Table S2). Replicated testing of strain 7498 with these conditions showed that visible flowering started at day 11, and peaked at day 14 with a maximum rate of 6% independent of whether the inoculant was grown in LD or CL ( Figure S1).
Using 4 week old cultures as a source of inoculant improved flowering rate. In Hg medium, Petri dishes of strain 7498 inoculated with three 3-5 frond colonies from one month old cultures growing in Hg medium, began flowering after 10 days in CL ( Figure  1). By day 24, strain 7498 had a significantly higher flowering rate than strain 9509 the 1.5 μM SA condition (p < 0.01). By day 28, when the culture filled the dish, 11 ± 1% of the 7498 fronds growing on media containing 1.5 μM SA were flowering, significantly different from 2.0 μM SA (p < 0.01). Under the same conditions, only 1% of strain 9509 fronds were flowering ( Figure 1). Both strains produced flowers with pistils secreting stigmatic fluid as well as flowers with dehiscent anthers, although 7498 did so at a higher rate ( Figure 2).  Occasionally, abnormal flowers were seen in Petri dish cultures. Unusual kidney shaped fronds of strain 7498 produced flowers that did not appear to emerge from the normal location on the frond ( Figure 3A). Strain 9509 occasionally produced flowers with what appeared to be the purple ventral side of the frond overgrowing and covering the anthers ( Figure 3B). On this frond, a rhizoid, which would normally grow from the ventral surface of the frond, is growing vertically up out of the medium. Completely detached, aborted developing pistils and anthers enclosed in the spathe were found floating in strain 9509 cultures ( Figure 3C). Occasionally, abnormal flowers were seen in Petri dish cultures. Unusual kidney shaped fronds of strain 7498 produced flowers that did not appear to emerge from the normal location on the frond ( Figure 3A). Strain 9509 occasionally produced flowers with what appeared to be the purple ventral side of the frond overgrowing and covering the anthers ( Figure 3B). On this frond, a rhizoid, which would normally grow from the ventral surface of the frond, is growing vertically up out of the medium. Completely detached, aborted developing pistils and anthers enclosed in the spathe were found floating in strain 9509 cultures ( Figure 3C).  Both strains of Spirodela produced some viable pollen according to the Alexander's stain [52], that was capable of germinating in vitro ( Figure 4). In strain 7498, 69% or 161 of 234 total pollen grains from dehiscing anthers stained as viable, and 48% (47 of 97 total) germinated and produced pollen tubes. Only 26% (207 of 803) of the pollen from strain 9509 dehiscing anthers stained as viable, and only 9.7% (10 of 103) produced pollen tubes. Both strains of Spirodela produced some viable pollen according to the Alexander's stain [52], that was capable of germinating in vitro ( Figure 4). In strain 7498, 69% or 161 of 234 total pollen grains from dehiscing anthers stained as viable, and 48% (47 of 97 total) germinated and produced pollen tubes. Only 26% (207 of 803) of the pollen from strain 9509 dehiscing anthers stained as viable, and only 9.7% (10 of 103) produced pollen tubes. Both strains of Spirodela produced some viable pollen according to the Alexander's stain [52], that was capable of germinating in vitro ( Figure 4). In strain 7498, 69% or 161 of 234 total pollen grains from dehiscing anthers stained as viable, and 48% (47 of 97 total) germinated and produced pollen tubes. Only 26% (207 of 803) of the pollen from strain 9509 dehiscing anthers stained as viable, and only 9.7% (10 of 103) produced pollen tubes. On average, strain 7498 formed 518 ± 98 turions per flask over a 2 month culture period, and an average of 309 ± 66 turions were found per flask of strain 9509 over the same time period. After 2 months, no fruit or seed were found at the bottom of flasks or On average, strain 7498 formed 518 ± 98 turions per flask over a 2 month culture period, and an average of 309 ± 66 turions were found per flask of strain 9509 over the same time period. After 2 months, no fruit or seed were found at the bottom of flasks or plates of either strain. Gentle rotary shaking of both strains, and manually self-pollinating 6 fluid secreting pistils of 7498 with dehiscent anthers failed to result in seed formation.

Wolffia Microscopica
A screen in which plants were cultured on Hg, with daylength, chelating agent ED-DHA, and sucrose addition varied, clearly showed that LD were better for flowering than CL, provided that sucrose and EDDHA were present (Table S3). EDDHA enhanced frond multiplication when sucrose is present, although plants were capable of multiplying without either sucrose or EDDHA. However, cultures without sucrose and EDDHA senesced earlier than those with sucrose but without EDDHA. Overall, the screen indicated that in LD, Hg supplemented with sucrose and either 25 or 75 µM of EDDHA resulted in the highest flowering rates (Table S3).
Further screens with E and Hg media in flasks and Petri dishes exhibited very high variability, but overall greater flowering was seen with E, and on this medium no influence of EDDHA on flowering rate was observed (Table S4). This screen indicated that when growing on E, W. microscopica was greatly affected by the type of culture vessel.
No flowering was observed if Wolffia was grown in cotton stoppered flasks containing 100 mL of E, whereas if the plants were grown in 10 mL of E in a 6-well plate, 38 ± 2% of the fronds were flowering over the same time period ( Figure 5). After day 7, flowering was significantly greater in parafilm sealed 6-well plates (red) than in parafilm sealed Petri dishes (green) (p < 0.05) and no significant differences were found between 6-well plates sealed or not sealed with parafilm ( Figure 5). Under these conditions, the abundant pistils were easily observed due to a drop of stigmatic fluid, which was almost the size of the frond, and dehiscent anthers were also abundant ( Figure 6). Pollen from these anthers stained as 56% (94 of 169) viable, and 28% of the pollen grains formed germination tubes (65 of 235) ( Figure 6). Despite plentiful flowering, none of the cultures, either gently shaken or stationary, were observed to produce seed. earlier than those with sucrose but without EDDHA. Overall, the screen indicated that in LD, Hg supplemented with sucrose and either 25 or 75 μM of EDDHA resulted in the highest flowering rates (Table S3).
Further screens with E and Hg media in flasks and Petri dishes exhibited very high variability, but overall greater flowering was seen with E, and on this medium no influence of EDDHA on flowering rate was observed (Table S4). This screen indicated that when growing on E, W. microscopica was greatly affected by the type of culture vessel.
No flowering was observed if Wolffia was grown in cotton stoppered flasks containing 100 mL of E, whereas if the plants were grown in 10 mL of E in a 6-well plate, 38 ± 2% of the fronds were flowering over the same time period ( Figure 5). After day 7, flowering was significantly greater in parafilm sealed 6-well plates (red) than in parafilm sealed Petri dishes (green) (p < 0.05) and no significant differences were found between 6-well plates sealed or not sealed with parafilm ( Figure 5). Under these conditions, the abundant pistils were easily observed due to a drop of stigmatic fluid, which was almost the size of the frond, and dehiscent anthers were also abundant ( Figure 6). Pollen from these anthers stained as 56% (94 of 169) viable, and 28% of the pollen grains formed germination tubes (65 of 235) ( Figure 6). Despite plentiful flowering, none of the cultures, either gently shaken or stationary, were observed to produce seed.

Lemna
No flowering was observed in our initial experiments with DWC130, in contrast to what was reported for this strain in Slovin and Cohen 1988 [16]. As of 2017, there were 5 strains of Lemna gibba G3 listed in the RDSC inventory: DWC114 (7741), collected in Sicily, which is the pollen producing strain described by Fu et al. [48], DWC130 and DWC131, labeled as a self-fertile parental line and a larger auxin mutant jsR1 regenerated from tissue culture [16], DWC132, which was contributed by Biolex Corporation, and DWC128 from the Waksman Collection, which was not included in this study. Contrary to expectations, DWC114 was, in fact, the largest of the 4 strains we grew. DNA barcoding using chloroplast atpH-F primers [28] unequivocally showed that strain DWC114 is L. gibba G3, and revealed that the other 3 strains are, in fact, L. minor ( Table 1). The DWC114 barcode sequence is an exact match to both L. gibba 5504 and L. gibba 7741 described by Fu et al. [48]. DWC132 was further barcoded using matK primers, which clarified its identity as L. minor. Barcoding revealed a high degree of similarity among the three L. minor strains, DWC130, 131, and 132, with all closely matching L. minor strain 7210. The auxin mutant jsR1 appears to have been lost in one of the collection's multiple transfers, while DWC114 appears to be the L. gibba G3 originally described in Hillman 1961 [53]. Based on the results, our colleagues at the RDSC updated the database as part of their curation of the global collection.

Lemna
No flowering was observed in our initial experiments with DWC130, in contrast to what was reported for this strain in Slovin and Cohen 1988 [16]. As of 2017, there were 5 strains of Lemna gibba G3 listed in the RDSC inventory: DWC114 (7741), collected in Sicily, which is the pollen producing strain described by Fu et al. [48], DWC130 and DWC131, labeled as a self-fertile parental line and a larger auxin mutant jsR1 regenerated from tissue culture [16], DWC132, which was contributed by Biolex Corporation, and DWC128 from the Waksman Collection, which was not included in this study. Contrary to expectations, DWC114 was, in fact, the largest of the 4 strains we grew. DNA barcoding using chloroplast atpH-F primers [28] unequivocally showed that strain DWC114 is L. gibba G3, and revealed that the other 3 strains are, in fact, L. minor ( Table 1). The DWC114 barcode sequence is an exact match to both L. gibba 5504 and L. gibba 7741 described by Fu et al. [48]. DWC132 was further barcoded using matK primers, which clarified its identity as L. minor. Barcoding revealed a high degree of similarity among the three L. minor strains, DWC130, 131, and 132, with all closely matching L. minor strain 7210. The auxin mutant jsR1 appears to have been lost in one of the collection's multiple transfers, while DWC114 appears to be the L. gibba G3 originally described in Hillman 1961 [53]. Based on the results, our colleagues at the RDSC updated the database as part of their curation of the global collection.  As preliminary screening for flowering rate, all 4 strains were grown in flasks of E with 0-100 µM SA in CL, and also in E with either 20 µM SA or 20 µM SA + 75 µM EDDHA in either LD or CL ( Figure S2). Overall, this screen showed that DWC114 had the highest flowering rate of the 4 Lemna, that 24 h of light promoted flowering, that the optimum SA concentration for flowering was 20 µM, and that EDDHA had no obvious additive effect on flowering.
Under these conditions, L. gibba DWC114 produced perfect flowers, with a translucent pistil often topped with a drop of stigmatic fluid and two dehiscent anthers ( Figure 7A). L. minor DWC130 ( Figure 7B) produced small pistils observable at 16× magnification with no stigmatic fluid, but no anthers. L. minor strains DWC131 ( Figure 7C) and DWC132 ( Figure 7D) produced large pistils and large drops of stigmatic fluid, with two white but non-dehiscent anthers which were prone to detaching and sinking to the bottom of the flask.
As preliminary screening for flowering rate, all 4 strains were grown in flasks of E with 0-100 μM SA in CL, and also in E with either 20 μM SA or 20 μM SA + 75 μM EDDHA in either LD or CL ( Figure S2). Overall, this screen showed that DWC114 had the highest flowering rate of the 4 Lemna, that 24 h of light promoted flowering, that the optimum SA concentration for flowering was 20 μM, and that EDDHA had no obvious additive effect on flowering.
Under these conditions, L. gibba DWC114 produced perfect flowers, with a translucent pistil often topped with a drop of stigmatic fluid and two dehiscent anthers ( Figure  7A). L. minor DWC130 ( Figure 7B) produced small pistils observable at 16× magnification with no stigmatic fluid, but no anthers. L. minor strains DWC131 ( Figure 7C) and DWC132 ( Figure 7D) produced large pistils and large drops of stigmatic fluid, with two white but non-dehiscent anthers which were prone to detaching and sinking to the bottom of the flask.  To determine if flowering of L. gibba G3, DWC114, was affected by growth in constant light, plants were cultured, in triplicate, in cotton stoppered flasks containing E supplemented with no SA or 20 µM SA, in LD or CL ( Figure 7E). The results show that the presence of SA was required to produce flowering fronds until very late in the culture cycle.
In the presence of SA, flowering under CL was significantly greater than in LD at day 14 (p < 0.01) but the appearance of flowers was essentially equal thereafter under either light condition until the end of the culture cycle.
In the presence of SA, flowering under CL was significantly greater than in LD at day 14 (p < 0.01) but the appearance of flowers was essentially equal thereafter under either light condition until the end of the culture cycle.
The production of viable pollen by DWC114, DWC131, and DWC132 was assessed with modified Alexander's stain [52] (Figure 8A,B) and by pollen germination assays (Figure 8C). Pollen grains from dehisced anthers of L. gibba G3 DWC114 were largely viable as assessed by staining (83% viability) ( Figure 8A1,A2), with 43% capable of forming a pollen tube (59 of 131) ( Figure 8A3). The non-dehiscent anthers of L. minor strains DWC131 and DWC132 contained mostly non-viable pollen, with 1% (3 of 227), and 2% (11 of 454), respectively, of pollen grains staining as viable ( Figure 8B1,B2 and Figure 8C1,C2). None of the pollen from these anthers produced germination tubes ( Figure 8B3,C3).  The effect of SA on seed production by L. gibba G3, DWC114 (Figure 9) was determined under two light regimes and at four concentrations of SA (10-30 µM SA) over a 40 day culture period. Although similar rates of flowering were observed in LD if SA was present, after 25 days in culture flowering percent dropped to zero by 40 days if SA was not present ( Figure 9A). In contrast, the number of seed produced remained low until 25 day of culture, and increased thereafter until 40 days, independent of whether SA was present or not ( Figure 9B). In LD, plants growing on E with 30 µM SA produced the fewest seed, an average of 25 ± 0.3 seed per flask. Under CL conditions, flowering remained high throughout the 40 day culture period with small increases attributable to the presence of SA ( Figure 9C). In comparison to cultures growing in LD, seed production by plants in CL was substantially higher, with cultures containing 30 µM SA producing almost 150 seed per flask by day 35 ( Figure 9D). day of culture, and increased thereafter until 40 days, independent of whether SA was present or not ( Figure 9B). In LD, plants growing on E with 30 μM SA produced the fewest seed, an average of 25 ± 0.3 seed per flask. Under CL conditions, flowering remained high throughout the 40 day culture period with small increases attributable to the presence of SA ( Figure 9C). In comparison to cultures growing in LD, seed production by plants in CL was substantially higher, with cultures containing 30 μM SA producing almost 150 seed per flask by day 35 ( Figure 9D). Additional media supplements, such as yeast extract and bactopeptone (Eye) with 75 μM EDDHA, were tested for their reported ability to enhance seed production [16]. Addition of 20 μM SA slightly improved flowering ( Figure S3A), however, seed production was decreased substantially with SA in the enriched medium ( Figure S3B). In comparison, plants growing on Eye + 75 μM EDDHA produced fewer seed than plants growing on E supplemented with 30 μM SA. Additional media supplements, such as yeast extract and bactopeptone (Eye) with 75 µM EDDHA, were tested for their reported ability to enhance seed production [16]. Addition of 20 µM SA slightly improved flowering ( Figure S3A), however, seed production was decreased substantially with SA in the enriched medium ( Figure S3B). In comparison, plants growing on Eye + 75 µM EDDHA produced fewer seed than plants growing on E supplemented with 30 µM SA.
Developing seed of L. gibba G3 DWC114 were visible within the frond due to the highly pigmented operculum at one end ( Figure 10A). Occasionally, developing seed within the ovary ( Figure 10B) were seen detached from the mother frond, although most seed became apparent when they became free of the ovary and dropped to the bottom of the flask ( Figure 10B). Seeds could then be harvested using a Pasteur pipette and stored dry, or in the cold in diluted E. Seeds germinated if transferred to fresh E (Figure 10C), achieving a germination rate by day 3 of 65% or higher, independent of LD or CL light conditions ( Figure 10).
highly pigmented operculum at one end ( Figure 10A). Occasionally, developing seed within the ovary ( Figure 10B) were seen detached from the mother frond, although most seed became apparent when they became free of the ovary and dropped to the bottom of the flask ( Figure 10B). Seeds could then be harvested using a Pasteur pipette and stored dry, or in the cold in diluted E. Seeds germinated if transferred to fresh E (Figure 10C), achieving a germination rate by day 3 of 65% or higher, independent of LD or CL light conditions ( Figure 10).

Discussion
The majority of the vast body of flowering research using members of the Lemnaceae was conducted before the sequencing of the human and Arabidopsis genomes [54][55][56] and before the first miRNA was discovered in Caenorhabditis [57] or miRNAs were found in Arabidopsis [58]. Today's technology, which now allows for determination of the genetic mechanisms involved in flowering in these neotenic aquatic monocots, is still dependent on being able to have a convenient and reliable means of inducing flowering in the laboratory. While previous publications on flowering almost always focused on a single genus of the Lemnaceae, we developed protocols for obtaining flowering dependably in 3 different genera, and applied lessons between them, as a means of enabling the further study of the genetic mechanisms regulating floral induction and progressing toward breeding protocols in this plant family.
The first step in developing reliable flowering protocols was the standardization of the number and the birth order of the inoculating fronds. In Lemna, first daughter fronds

Discussion
The majority of the vast body of flowering research using members of the Lemnaceae was conducted before the sequencing of the human and Arabidopsis genomes [54][55][56] and before the first miRNA was discovered in Caenorhabditis [57] or miRNAs were found in Arabidopsis [58]. Today's technology, which now allows for determination of the genetic mechanisms involved in flowering in these neotenic aquatic monocots, is still dependent on being able to have a convenient and reliable means of inducing flowering in the laboratory. While previous publications on flowering almost always focused on a single genus of the Lemnaceae, we developed protocols for obtaining flowering dependably in 3 different genera, and applied lessons between them, as a means of enabling the further study of the genetic mechanisms regulating floral induction and progressing toward breeding protocols in this plant family.
The first step in developing reliable flowering protocols was the standardization of the number and the birth order of the inoculating fronds. In Lemna, first daughter fronds are more robust that those produced subsequently [59][60][61]. Birth order also influences production of turions by S. polyrhiza [62]. We consistently used 4-frond colonies of Lemna and 3-5 frond colonies of Spirodela as inoculant to help standardize growth and flowering rate. In this way, a colony would have a grandmother frond, with an attached mother frond, and a developing daughter frond, which would be the first daughter. Depending on environmental conditions, a first daughter frond of Lemna gibba is capable of producing 12-14 daughters of its own before senescing, although the production of a flower significantly reduces the number of possible daughter fronds [63].
It is, therefore, worth considering that while the number of daughter fronds produced by a particular meristem before it can produce a flower is likely to be genetically determined, flowering may also be influenced epigenetically through the frond's mother. A first daughter will already have its first daughter developing in the meristematic pouch, with at least one generation of internally developing fronds in L. minor and S. polyrhiza [64], and W. microscopica containing up to three unseen internal generations of daughter fronds nested within each other [10]. Evolutionarily conserved epigenomic stress responses such as DNA methylation and histone modification that influence flowering time in rice, sugar beets, Arabidopsis, and other plants [65,66], are likely to be active through these internal generations and perhaps future generations, and may help explain the long term effects of previous culture conditions on flowering rates. Comparing the results in early Wolffia screens (Table S4) to experiments where inoculants came from fully acclimated cultures ( Figure 5) underscores the importance of acclimating cultures for 4 weeks to fully eliminate variation before running experiments, as suggested in Duckweed Forum issue 8 [67].

Spirodela Polyrhiza
We focused on flowering of S. polyrhiza strains 7498 and 9509 because of the number of recent genomic and transcriptomic studies on them [11,29,30,34,35,37,38,68]. An abundance of molecular evidence illustrates the neotenous nature of Spirodela that must be overcome to induce flowering. Genome sequencing revealed a strong enrichment of genes for floral repressors, and a reduction in the number of members of the floral promoter gene families in both Spirodela strains compared to Arabidopsis [29,30]. About 9 members of the juvenile marker miR156 gene family were found, compared to 5 members of the adult marker miR172 gene family in the 7498 and 9509 genomes [29,30,68]. Sequencing of small RNAs in strain 7498 growing in 8 different stress conditions or hormonal applications, found that the ratio of miR156 to miR172 ranged from 71-408 [34]. These two highly conserved microRNAs regulate the abundance of the key floral promoters which integrate flowering responses to day length, stress, and accumulated sugar states [43,[69][70][71].
Most strains of S. polyrhiza were classified as day neutral plants [1,72]. We found flowering in 12 h days, (Figure S1), although growth in CL produced the highest flowering rates (Figure 1). If inoculant was taken from 4-week-old cultures growing in CL in flasks containing 100 mL of Hg media, and used to inoculate Petri dishes containing 50 mL of Hg + 1.5 µM SA, we obtained optimal flowering of 11% for 7498 and 1% for 9509 ( Figure 1). While both 7498 and 9509 produced pistils with stigmatic fluid and pollen capable of producing germination tubes (Figures 2 and 4), gentle shaking, which was found to promote seed production in L. gibba G3 [16] failed to produce seed in S. polyrhiza. Our attempts to manually self-pollinate 7498 by touching dehiscent anthers to pistils, as done with L. gibba by Fu et al. [48] were also unsuccessful.
Diverse types of stress can influence flowering in plants through SA signaling [73,74]. In Spirodela, a more common stress response than flowering and seed formation may be dormancy through the production of the starch-rich turions. In our optimal conditions, a 4 weeks old culture could have up to 39 mature and developing flowers, while at the same time having an average of 240 turions for strain 7498 and 134 turions for strain 9509. After two months, up to 500 turions could be found per flask of strain 7498, suggesting that genetic or environmental manipulation of turion production could dramatically influence the production of flowers, and vice versa.

Wolffia Microscopica
W. microscopica flowers abundantly in the wild. These diminutive plants were used for many early studies on induction of flowering [75,76], and the morphology and anatomy of seed obtained from the wild have been studied [47]. In addition to being a promising crop plant which provides a complete protein [4], Wolffia is a prime candidate for molecular studies into how a tiny monocot manages to maintain both a vegetative and sexual reproductive cycle. The consistent laboratory protocol for maximum flowering induction we describe supports efforts to maximize research on, and use of, these plants.
Our initial screen validated the reported abundant flowering of these plants, although there was a high degree of variability of replicates within a trial, and variability between replicated experiments ( Figure S4). As with Spirodela and Lemna, variability in flowering was decreased when inoculants came from acclimated cultures ( Figure 5). Culture inoculant size and age, EDDHA, and hormones like SA, are known to play roles in regulation of flowering of the Lemnaceae. We found that under our conditions, EDDHA had no effect on flowering rate (Table S4), unlike what was reported earlier for a different strain of W. microscopica [77], or S. polyrhiza SP 20 [78] and L. gibba G3 [79]. In addition to its chelating effects, EDDHA breaks down in a matter of weeks to form SA and SA-like molecules in sunlight [80], yet this transition from EDDHA to SA may not have occurred in our experiments where EDDHA was added to the medium after autoclaving.
The differences we observed in flowering when plants were grown under the same conditions but in different culture vessels (Table S4) suggests that a physical factor, such as rate of evaporation or accumulation of volatiles, surface to volume ratio, or even quorum sensing influences flowering in Wolffia and perhaps other duckweed genera. To test whether evaporation, or exchange of gases such as ethylene may be responsible, we tested flowering rate in Petri dishes and 6-well plates with and without parafilm. We found no significant difference due to parafilm, suggesting that surface to volume ratio or crowding may play a large role in flowering ( Figure 5), although the mechanism for increased flower induction remains ambiguous, it may be SA independent.

Lemna
While gene editing and transformation present avenues for production of new duckweed varieties for basic and applied research, L. gibba G3 (strain DWC114) is currently the ideal strain for attempts to breed new varieties through sexual reproduction. This strain flowers readily, produces abundant dehiscent anthers, and self-pollinates to readily produce seed [16,48], (Figures 7 and 10). Two different protocols to produce flowers and seed in L. gibba G3 have been described [16,48]. In the first, L. gibba G3 was grown on liquid E media in flasks in CL to produce flowers. Gentle shaking on a rotary shaker for 1 h twice a day was required to produce seed [16]. In the second protocol, the plants are grown in Petri dishes containing semi-solidified Modified Hoagland's media supplemented with 20 µM SA in LD [48]. Flowers were either self or cross-pollinated manually under a stereoscope to produce seed [48]. It is possible that flowering under CL is influenced by a decrease in Circadian Clock Associated 1, which oscillates every 24 h under LD or SD, but is transcriptionally silenced when L. gibba G3 is grown under CL for 48 h or more [81]. Our best seed production protocol incorporated aspects of both previous protocols to both strongly induce flowering, and easily pollinate flowers with a gentle stir twice per week.
The need for supplementary SA to induce flowering if plants are grown in LD as found by Fu et al. [48] as compared to CL without SA [16] may also be explained if CL is acting as an abiotic stressor and thus increases endogenous levels of SA. Many plants respond to abiotic and biotic stress by the production of SA [73]. Intriguingly, L. gibba grown under CL exhibited increased auxin turnover [82]. Additionally, tight control of auxin levels is mediated through conjugation by GH3 acyl acid amido synthases, one of which, atGH3.5, has been shown to have a role in both auxin and SA homeostasis [83]. SA interacts with the flowering developmental pathway through Flowering Locus D. Flowering Locus D demethylates the histones associated with Flowering Locus C and thus inhibits transcription of floral promoters [84][85][86][87][88].
Fortunately, for a mechanistic understanding of floral regulation, we are no longer reliant on comparisons to other plant species, but can now analyze the transcriptome of flowering L. gibba G3 recently published by Fu et al., 2020 [89]. This study revealed differential expression of 1501 genes throughout the flowering process, and highlighted the role of photoperiod and SA. In LD L. gibba had very low levels of transcripts of two genes, LgCO and LgGI, which in Arabidopsis are responsible for integrating circadian rhythms. Fu et al. [89] found a lack of endogenous SA production, suggesting that flowering is being activated through exogenous SA from the medium. Surprisingly, this transcriptome suggested a flowering pathway distinctly different from other plant models, with a lack of expression of the floral regulators LgCO, LgGI, LgSOC1, and LgFD, and a co-expression of FT with LgTEM1 and LgSVP, which act as floral inhibitors in Arabidopsis [89]. Clearly there is a need for flowering transcriptome data from Spirodela, Wolffia, and the improperly developing Lemna minor strains, with which to build a larger picture of floral regulation strategies across this family.
Under optimal conditions for floral induction, L gibba DWC114 produced pistils with stigmatic fluid and both dehiscent and apparently non-dehiscent anthers, L. minor DWC130 produced pistils but no stigmatic fluid, L. minor DWC131 and L. minor DWC132 pistils produced stigmatic fluid but anthers remained non-dehiscent ( Figure 8). Under our conditions, none of the L. minor strains produce seed, whereas L. gibba readily does. Over 70 attempts to manually fertilize L. minor DWC131 and DWC132 with L. gibba pollen were unsuccessful, indicating that the two species may be too diverged to form an interspecific hybrid (data not shown). However, the marked differences in floral development and fertility of these genetically related strains of L. minor and L. gibba provides an opportunity to identify critical genes in floral development.
L. gibba seed can be left in a Petri dish to dry overnight, wrapped in parafilm, and stored dry at 4 • C for long-term storage. Seed can also be stored in diluted E medium, although light must be blocked to prevent germination. Of 234 seed stored in diluted E medium for 15 months, 36% germinated when placed in fresh E medium in the light. L. gibba seeds collected in the wild germinated at 70% efficiency after being stored in water for 2 years at room temperature, but germination dropped to 1% after 3 years of wet storage [90]. Only 5.6% of seed germinated after dry storage [90]. To eliminate possible fungal or bacterial contamination, seeds were surface sterilized with 5% bleach for 3 min or 10 min, then washed 3 times with sterile water, or more conveniently, rinsed with 70% ethanol and air dried. Ninety seven percent of one month old seed treated with any of the three methods germinated with no contamination. Single seed germinated within 4 days in as little as 5 mL of E medium in a 6-well plate, making it easy to rapidly obtain progeny from a cross.
Similar to many plant species that rely on protogyny (the sequential development of the pistil followed by the anther) to promote cross-fertilization while ensuring fruit production [91], like most family members the three genera of Lemnaceae we studied are protogynous [3,[92][93][94]. Due to the clonal nature of Lemnaceae populations, protogyny would be substantially less effective at promoting cross-pollination and reducing deleterious in breeding than other self-incompatibility (SI) strategies [3]. Mechanisms involving the determinants at the S locus are diverse, and have evolved at least 35 times in the angiosperms, yet very little is known about them in monocots [95], and certain monocot orchids appear to have a novel SI mechanism [96]. Observations of flowering and seed production in the wild led to the suggestion that 22 species of duckweed are likely self-compatible [3]. Historical reports stated that W. welwitschii was able to selfpollinate [97], and that two L. minor populations were self-incompatible but capable of cross fertilization [98]. This work is summarized in Landolt and Kandeler (1987) [1]. Although Maheshwari (1956) [47] reported seed production by W. microscopica in the wild, it is not clear whether the wild populations were clonal or heterogeneous. We were unable to obtain seed from our W. microscopica and S. polyrhiza cultures even though pollen was capable of forming germination tubes, suggesting self-incompatibility in these species. It is therefore likely that SI is variable across the family, and co-cultivation of genotypes within a species may be required to develop breeding protocols. It will be interesting to investigate the relationship of SI mechanisms to genetic diversity in native habitats and the ability of a species to produce seed or develop turions. Both of these organs are capable of becoming dormant and surviving desiccation, and thus their formation might represent alternative stress-response strategies.

Summary
The Lemnaceae can be seen as new model plants ideally suited to probing important questions in plant biology, and for developing a new crop that can be a part of sustainably providing humanity with clean water, food, and fuel. To make future floral and breeding experiments easier we developed culturing protocols for obtaining flowering plants in the laboratory.
In CL or LD, L. gibba G3 flowers and sets seed in E medium supplemented with 20 µM SA. Occasional movement of the culture increases seed production.
W. microscopica flowers conveniently under LD in E medium in 6-well plates. S. polyrhiza flowers in Petri dishes with 50 mL of Hg under CL when 1.5 µM SA is added.
Neither Wolffia or Spirodela produced seeds under these conditions, suggesting that selfincompatibility may be at play, and requiring further testing. With these culture conditions and modern molecular tools, it is now possible to determine the genetic mechanisms leading to flowering in W. microscopica, and determine if flowering in S. polyrhiza can reveal new insights in transposon mobility and DNA methylation. The Lemnaceae family is both exciting in its own right and an excellent basal monocot model with which to study the genetic mechanisms of floral regulation, or the mechanisms behind self-incompatibility in clonal populations.
The media and culture vessels used for this study are described in Tables 2 and 3. The chemical compositions of each of the media tested are provided in Supplementary  Table S1, while the instructions for making stock solutions for, and preparing, E medium are in Document S1. All media were supplemented with 29 mM sucrose unless specified otherwise with a minus (−) sign after the medium abbreviation. Additions of yeast extract and bactopeptone to E (Eye media) were made before autoclaving. SA was prepared as a 100 mM stock solution in ethanol. EDDHA was prepared as a 100 mM stock solution in water, and filter sterilized. SA and EDDHA were added to cooled media at the concentrations indicated. Flasks were fitted with cotton stoppers and loose aluminum foil covers. Petri dishes and 6-well plates were wrapped with strips of parafilm to maintain sterility and reduce evaporation.   New cultures of W. microscopica were inoculated with a single 1.5 × 1.5 cm mesh loopful of floating fronds from a 3-5 week old mature culture growing in a 250 mL Erlenmeyer flask containing 100 mL E medium. Spirodela experiments were started with three 3-5 frond colonies from a 1, 2, or 4 week old culture grown in a flask of Hg medium. For L. gibba and L. minor, all experiments were started with 3 four-frond colonies from a 7 or 14 day old population growing in E medium in flasks.

Flowering, Seed Production and Seed Storage
Flowers were counted if they were developing or mature. An observable pistil, anthers, or a perfect flower (pistil with 2 anthers) were counted as one flower. Flowering rate is expressed as the percent of developing and mature flowers per 100 fronds. The number of fronds scored was either every frond in the culture vessel, or ≥100 fronds from any one culture. Flower or seed production data are presented as the average ± SEM (standard error of the mean, (n ≥ 3). Statistically significant difference between variables were calculated by Student's t-test. Seed production is expressed as the total number per culture vessel. Seeds were stored dry, or in a flask of depleted media diluted 1:2 with sterile water at 4 • C. Unless otherwise specified, flowers were counted on days 7, 10, 14, 17, 21, 24, 28 after subculture.

Pollen Viability and Fertility
Pollen from all strains that produced anthers was tested for viability with a modified Alexander's stain [52] and for the ability to form pollen tubes by gently spreading anthers on E solidified with agar. For staining, pollen from 5-10 anthers was stained, and at least 169 grains per anther were scored as viable or dead. Pollen tube formation was observed at 20× using a Leica DM550B microscope 1 h later. Fertility is described as the percent of pollen grains with germination tubes. At least 97 grains from 4-7 anthers were observed.
Supplementary Materials: The following are available online at https://www.mdpi.com/1422-006 7/22/5/2733/s1, Figure S1: Flowering rates of Spirodela polyrhiza in LD and CL. Figure S2: Screen for the effects of 0-100 µM SA and 75 µM EDDHA in E on flowering rate of L. gibba G3 and 2 strains of L. minor in LD or CL. Figure S3: Flowering rate and seed production in strain DWC114. Table S1: Media compositions. Table S2: Summary of Spirodela screens and experiments. Table S3: Screening for optimal flowering of W. microscopica 2005. Table S4: Summary of W. microscopica flowering rates in different containers and media, with and without EDDHA. Document S1: E medium protocol.
Author Contributions: Conceived, guided, and supervised the research, J.M.; designed and conducted experiments, P.F.; contributed expertise to the project, J.S.; wrote and edited the manuscript, P.F. and J.S. All authors have read and agreed to the published version of the manuscript.