Characterization of Various Subunit Combinations of ADP-Glucose Pyrophosphorylase in Duckweed (Landoltia punctata)

Background Landoltia punctata can be used as renewable and sustainable biofuel feedstock because it can quickly accumulate high starch levels. ADP-glucose pyrophosphorylase (AGPase) catalyzes the first committed step during starch biosynthesis in higher plants. The heterotetrameric structure of plant AGPases comprises pairs of large subunits (LSs) and small subunits (SSs). Although several studies have reported on the high starch accumulation capacity of duckweed, no study has explored the underlying molecular accumulation mechanisms and their linkage with AGPase. Therefore, this study focused on characterizing the roles of different L. punctate AGPases. Methodology. Expression patterns of LpAGPs were determined through comparative transcriptome analyses, followed by coexpressing their coding sequences in Escherichia coli, Saccharomyces cerevisiae, Arabidopsis thaliana, and Nicotiana tabacum. Results Comparative transcriptome analyses showed that there are five AGPase subunits encoding cDNAs in L. punctata (LpAGPS1, LpAGPS2, LpAGPL1, LpAGPL2, and LpAGPL3). Nutrient starvation (distilled water treatment) significantly upregulated the expression of LpAGPS1, LpAGPL2, and LpAGPL3. Coexpression of LpAGPSs and LpAGPLs in Escherichia coli generated six heterotetramers, but only four (LpAGPS1/LpAGPL3, LpAGPS2/LpAGPL1, LpAGPS2/LpAGPL2, and LpAGPS2/LpAGPL3) exhibited AGPase activities and displayed a brownish coloration upon exposure to iodine staining. Yeast two-hybrid and bimolecular fluorescence complementation (BiFC) assays validated the interactions between LpAGPS1/LpAGPL2, LpAGPS1/LpAGPL3, LpAGPS2/LpAGPL1, LpAGPS2/LpAGPL2, and LpAGPS2/LpAGPL3. All the five LpAGPs were fusion-expressed with hGFP in Arabidopsis protoplasts, and their green fluorescence signals were uniformly localized in the chloroplast, indicating that they are plastid proteins. Conclusions This study uncovered the cDNA sequences, structures, subunit interactions, expression patterns, and subcellular localization of AGPase. Collectively, these findings provide new insights into the molecular mechanism of fast starch accumulation in L. punctata.


Introduction
Duckweed is a general term for plants belonging to the Lemnaceae family, comprising five genera (Spirodela, Landoltia, Lemna, Wolffiella, and Wolffia) and more than 40 species [1]. The plants in this family do not differentiate into stems and leaves but exhibit leaflike fronds bearing none or several roots on the underside. In warm climates, duckweed thrives throughout the year, and their biomass increases significantly due to exponential growth rate in nutrient water.
Indeed, duckweed can achieve between 13 to 38 metric tons dry weight (DW)/hectare/year [2,3] depending on the environmental conditions. Duckweed can be used in bioethanol production because of its comparatively high starch and low lignin percentage [4,5]. Through fermentation, duckweed can be used for the industrial production of biofuel and sewage treatment simultaneously [6,7]. Also, duckweed positively affects eutrophic water because it can eliminate up to 98% of ammonia, 85% of total nitrogen, and 78% of total phosphorous [8]. Duckweed can be applied to purify foul water and generate renewable forms of energy. It is also an excellent source of animal feed [9]. Therefore, in a world of increasingly scarce resources, there is a need to constantly source for new applications of natural products.
L. punctata is the most efficient species in accumulating starch among over 800 Lemnaceae specimens collected from various parts of the world. L. punctata OT cultivated in wastewater increased its starch percentage to 52.9% (DW) during a field experiment [10]. In our previous study, the starch percentage of L. punctata 0202 plants increased from 3% to 45.5% (DW) when shifted from standard Hoagland nutrient solution to distilled water [11]. Additionally, the enzymatic activity of ADP-glucose pyrophosphorylase (AGPase, EC: 2.7.7.27) increased by 53.1%.
As one of the essential enzymes in starch biosynthesis, AGPase is the first and most crucial rate-limiting enzyme and thus has been a genetic engineering target for crop improvement [12]. Regulating AGPase enzyme activity can change the carbon metabolism of plants and influence their starch content [13]. In higher plants, AGPase functions as a heterotetramer comprising two large subunits (LSs) and two small subunits (SSs) [14,15], whereas, in cyanobacteria and prokaryotes, it functions as a homotetramer [16]. Generally, the primary function of the large AGPase subunit is to modulate the regulatory properties of the enzyme in higher plants, while the small subunit mainly functions in catalysis [17]. However, the small subunit exhibited partial regulatory activity, and the large subunit had a little catalytic activity [18][19][20].
In specific instances, starch-derived and reservoir organs of plants harbor different AGPase large subunits. The expression patterns of AGPases vary in different organs and can also differ in the same organ at different developmental stages [21]. This spatiotemporal specific expression enables various AGPase to regulate starch synthesis in different organs at different plant developmental stages. AGPase is located in the plastids of most plant tissues, while most of the AGPase in rice [22], maize [23], barley endosperm [24], and early-stage wheat endosperm [25] are located in the cytoplasm. Thus, plants have both cytosolic AGPase and plastidial AGPase. To understand the role of different AGPases in plant starch synthesis, the enzymatic activity and spatiotemporal specificity of various subunit combinations need to be studied. Among the five Lemnaceae genera, Spirodela and Lemna are the most studied. However, the value of Landoltia has not been comprehensively explored, and its genome has not been decoded. Additionally, the relationship between various AGPases and starch accumulation has not been elucidated in Landoltia.
BioMed Research International MA, USA) using oligo (dT) as primers. PCR amplification was performed using KOD-Plus-Neo DNA polymerase (ToYoBo, Tokyo, Japan). The PCR products were separated by agarose gel electrophoresis and purified using EasyPure® Quick Gel Extraction Kit (TransGen Biotech, Beijing, China).

Yeast Two-Hybrid Analysis.
Primers were designed based on the LpAGP sequences without the regions encoding the plastid transit peptides (Table S1). The purified PCR products of LpAGPSs and pGBKT7 yeast expression vectors (Clontech) were double-digested with Nco I/BamH I (TaKaRa, Shiga, Japan) and separated by agarose gel electrophoresis. Next, the desired bands were extracted from the gel and purified using EasyPure® Quick Gel Extraction Kit (TransGen Biotech, Beijing, China). The purified products were then ligated by T4 DNA ligase (TaKaRa, Shiga, Japan) to obtain recombinant plasmids pGBKT7-LpAGPS1 and pGBKT7-LpAGPS2. The purified PCR products of LpAGPLs were digested with Pci I/EcoR I (TaKaRa, Shiga, Japan), while pGADT7-Rec was doubledigested with Nco I/EcoR I (TaKaRa, Shiga, Japan). After purification, the digestion fragments of LpAGPLs and pGADT7-Rec were ligated by T4 DNA ligase (TaKaRa, Shiga, Japan) after purification and transformed into E. coli DH5α competent cells via electroporation. Positive clones were identified by colony PCR, and the recombinant plasmids pGADT7-Rec-LpAGPL1, pGADT7-Rec-LpAGPL2, and pGADT7-Rec-LpAGPL3 were obtained. The LpAGPS and LpAGPL recombinant plasmids were cotransformed pairwise into Saccharomyces cerevisiae strain AH109 and used for a two-hybrid analysis as per the Clontech protocol. Meanwhile, pGADT7 and pGBKT7-LpAGPS1 were cotransformed into S. cerevisiae strain AH109 as the negative control, while the same strain was used to cotransform pGADT7-T and pGBKT7-53 (Clontech, Tokyo, Japan) as a positive control. AH109 transformants were selected using a synthetic growth medium lacking His, Leu, and Trp. The selection of subunit interactions was confirmed on SD/-Leu-Trp-His-Ade and X-α-gal medium.

Coexpression of AGPase LSs and SSs in E. coli.
Primers were designed based on LpAGP sequences without plastid transit peptide coding region (Table S1). The PCR products of LpAGPSs and pACYC-Duet-1 were purified and double-digested using Nco I/BamH I (TaKaRa, Shiga, Japan), then ligated. Meanwhile, the PCR products of LpAGPLs and pRSF-Duet-1 were purified and double digested using Hind III/EcoR I (TaKaRa, Shiga, Japan), then ligated. The products were transformed into E. coli DH5α competent cells, and positive clones were identified by colony PCR. Finally, the plasmids were extracted and sequenced by the Sanger sequencing method (ABI 3730 DNA sequencer, Life Technologies, USA).
The identified recombinant plasmids of LpAGPSs (pACYC-LpAGPS1 and pACYC-LpAGPS2) and LpAGPLs (pRSF-LpAGPL1, pRSF-LpAGPL2, and pRSF-LpAGPL3) were cotransformed pairwise into E. coli BL21 (DE3). Meanwhile, plasmids LpAGPS1-pMD18-T, LpAGPS2-pMD18-T, LpAGPL1-pMD18-T, LpAGPL2-pMD18-T, LpAGPL3-pMD18-T, pACYCDuet-1, and pRSFDuet-1 were separately transformed into E. coli BL21 (DE3) as a negative control. Bacterial colonies containing one or two expression plasmids were selected using antibiotics. Positive clones were identified by colony PCR, and the positive E. coli strains were grown in Luria-Bertani medium supplemented with corresponding antibiotics at 37°C and 220 rpm overnight until an OD600 of 0.9 was achieved. After that, 1 mM isopropyl β-D-thiogalactoside (IPTG) was added, and the bacterial culture was further incubated at 37°C and 220 rpm for 30 h, to induce the expression of LpAGPs. The cultures were then centrifuged, washed, and resuspended in distilled water to OD600 of 0.8-1.0, suitable for the glycogen accumulation assay [16]. The experimental procedure was slightly modified: after adding 300 μL iodine solution (0.01 M I 2 and 0.03 M KI) to 500 μL of cell suspensions, the solution was mixed and incubated at room temperature for 2-3 min. Glycogen accumulation in the cells was indicated by brownish coloration, and the different shades of the color indicated different amounts of accumulated glycogen.
Leaf blades were cut into filaments (0.5-1 mm wide) and placed in 0.4 M mannitol. The filaments were completely immersed in the enzyme solution and incubated on a rotary shaker (40 rpm) in the dark for 2-3 h. Next, W5 solution was added to terminate enzymatic hydrolysis. The hydrolysate was sucked up with a 50 mL syringe (without needle), filtered into a round-bottom centrifuge tube through a 450 μm nylon mesh, and then centrifuged at 4°C, 100 g for 8 min with acceleration and deceleration set at 3-4 gear. Afterward, the supernatant was discarded, and the protoplasts were suspended in precooled W5 solution and then centrifuged at 4°C, 100 g, for 8 min. Next, the supernatant was removed, and the protoplasts were resuspended and incubated on ice for 30 min. Subsequently, the supernatant was removed, and the protoplasts were resuspended in

BioMed Research International
MMG solution (the final number of protoplasts was about 2 × 10 5 mL -1 ). Exactly 10 μL plasmid was added to 200 μL of protoplast in 2 mL tube and incubated for 5 min. Then, 210 μL PEG solution was added to the mixture, mixed gently, and incubated at 23°C for 5-30 min. Subsequently, 800 μL W5 solution was added to the mixture, mixed gently, and then centrifuged at 23°C, 100 g for 8 min. The supernatant was then gently discarded, and the pellet was resuspended in 1 mL WI solution, incubated for 12 h at 20-23°C under weak light, and then centrifuged at 23°C, 100 g for 8 min. After that, the supernatant was eliminated, 500 μL WI solution was added, and the mixture was incubated for 20 min in the dark, then centrifuged at 23°C, 100 g for 8 min. The supernatant was again discarded, and 1 mL WI solution was added; then, the mixture was centrifuged at 23°C, 100 g, for 8 min. Thereafter, 500 μL WI solution supplemented with 1 μL 4′,6-diamidino-2-phenylindole (DAPI) was added after discarding the supernatant, and the mixture was agitated, followed by incubation for 5 min in the dark, and centrifugation at 23°C, 100 g, for 8 min. Finally, the supernatant was discarded, and 1 mL WI solution was added; then, the mixture was agitated, incubated for 5 min in the dark, and centrifuged at 23°C, 100 g, for 8 min. Fluorescence images were taken under excitation light of 405 nm and 488 nm using a confocal microscope.

Bimolecular Fluorescence Complementation (BiFC).
Primers were designed based on the LpAGP sequences without the regions encoding the plastid transit peptide (Table S1). About 5 mL YEB culture medium supplemented with appropriate selection antibiotics (rifampicin + kanamycin) was inoculated with Agrobacterium strain GV3101 that had previously been transformed with fusion protein binary vector. The culture was left to grow overnight at 28°C and 200 rpm. About six weeks old and healthy Nicotiana tabacum plants were prepared for BiFC analysis. First, the regions of the leaves were prepared for infiltration by puncturing them with a thin needle. Next, the syringe tip (without needle) was placed against the upper side of the leaf and pressed down slowly on the plunger, while the upper side of the leaf was supported with a finger. The infiltrations were repeated in the two different midrib regions of three leaves per plant. The infiltrated regions were marked in black for further identification.
The infiltrated plants were placed in a growth cabinet under normal growth conditions for 72 h. Afterward, 1-2 cm 2 of the leaf segments within the infiltrated zone was excised and mounted on glass microscope slides containing water droplets. The slides were coved with coverslips and examined for expression using a confocal or epifluorescence microscope.

3.1.
Cloning and Analysis of L. punctata AGPase Subunit cDNAs. Comparative transcriptome analysis was performed in a previous study to investigate the high starch accumulation of L. punctata under nutrient starvation (distilled water treatment) [11]. It was shown that AGPase large subunits (LSs) are encoded by three genes (LpAGPL1, LpAGPL2, and LpAGPL3), while small subunits (SSs) are encoded by two genes (LpAGPS1, LpAGPS2) [11]. The expression levels analysis of the five LpAGPs showed that LpAGPS1 (comp27906_c1_seq1, comp27906_c1_seq2), LpAGPL2 (comp43464_c0_seq1), and LpAGPL3 (comp43482_c0_ seq2) were significantly upregulated under nutrient starvation (distilled water treatment) during the first 24 h (Figure 1). The expression patterns of the five LpAGPs have been verified by qRT-PCR in previous study [27]. Comp27906_c1_seq1 and comp27906_c1_seq2 possess an identical open reading frame (ORF) (LpAGPS1), whereas comp43482_c0_seq3, comp43482_c0_seq23, and comp43482_c0_seq26 possess an identical LpAGPL1 ORF.
Five LpAGPs were cloned according to their ORFs. The length of these LpAGPs genes ranges from 1254 bp to 1611 bp, corresponding to 417 to 536 amino acids. Sequence comparison of the five LpAGP revealed that LpAGPL2 and LpAGPL3 share the highest sequence identity (67%), while LpAGPS2 shares the lowest sequence identity with the other LpAGPs. These results show that the large subunit of L. punctata is relatively conserved, and the similarity between small subunits is lower than between large subunits. The sequences of LpAGPs were also aligned using BLASTN against the GenBank database. The gene encoding mannose-1-phosphate guanyltransferase in Hordeum vulgare shares significant sequence homology with LpAGPS1 (XM_045126568.1, sequence identity of 77%). Moreover, sequence identity between LpAGPS2 and the gene encoding the small subunit of Colocasia esculenta AGPase (MT445784.1) is 82%. The LpAGPLs show remarkable sequence homology (sequence identity ranges from 85% to 89%) to S. polyrhiza (JN180634.1, JN180635.1, JN180636.1).
A phylogenetic tree was constructed using MEGA 6.0 ( Figure 2) based on the deduced amino acid sequences of the five LpAGPs and AGPase subunits from other plant species. Sequences alignment was conducted using ClustalW2, and the phylogenetic parameters were estimated via maximum likelihood analyses. The phylogenetic tree was created using MEGA 6.0 ( Figure 2). Data were obtained from the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/) database. The results revealed that LpAGPS2 is less related to the other four LpAGPs but is more closely related to AGP of Oryza sativa, P. trichocarpa, and T. cacao. LpAGPS1 is more related to O. sativa, A. thaliana, N. tabacum, and Solanum tuberosum. The three AGPase large subunits of L. punctata were further aligned with those of S. polyrhiza. The sequence identity between LpAGPL1 and SpAGPL1 (AEV40471) is 94.39%, LpAGPL2 and SpAGPL2 (AEV40472) is 95.52%, and LpAGPL3 and SpAGPL3 (AEV40474) is 83.30% (Table 1). These results suggest that AGPase LSs of L. punctata and S. polyrhiza may have similar functions.

Interactions between the Large and Small Subunits (LSs and SSs) of L. punctate AGPase (LpAGP)
3.2.1. Yeast Two-Hybrid and X-α-gal Assay. Yeast twohybrid assay (Y2H) was performed to determine the interaction between the small and large LpAGP subunits. Five LpAGPs were cloned and inserted into pGBKT7 (LpAGPS1, LpAGPS2) or pGADT7-Rec (LpAGPL1, LpAGPL2, LpAGPL3) vectors to construct recombinant plasmids. LpAGPSs recombinant plasmids were free combined with those of LpAGPLs and cotransformed into S. cerevisiae strain AH109 for Y2H assay. The yeast cells harboring the S1-L1 combination and the other five SS-LS combinations grew on both the selective-interaction medium (a synthetic growth medium lacking Trp, Leu, Ade, and His) and X-αgal medium (Figure 3). These results indicate that LpAGPS1 does not interact with LpAGPL1.

Bimolecular Fluorescence Complementation. Five
LpAGPs were subcloned into the pSPYNE-35S BiFC (bimolecular fluorescence complementation) expression vector and then transferred into A. tumefaciens GV3101 competent cells via electroporation. Positive control (pSPYNE-ABI2 and pSPYCE-RCAR1) exhibited relatively high transient expression of YFP (Figure 4(a)), while no fluorescence was observed in the negative control (pSPYNE-LpAGPS2, Figure 4(b)). Besides, no fluorescent signal was observed under simultaneous expression of LpAGPS1 and LpAGPL1. Simultaneous transient expressions of LpAGPS1 and LpAGPL3, LpAGPS2 and LpAGPL1, LpAGPS2 and LpAGPL2, and LpAGPS2 and LpAGPL3 in Arabidopsis protoplasts resulted in stronger BiFC fluorescence than LpAGPS1 and LpAGPL2 expression. These results confirm that the LpAGPS1 does not interact with LpAGPL1.
3.2.3. Coexpression of L. punctata AGPase LS and SS cDNA in E. coli. Previous studies have shown that coexpressing of cDNA sequences of plant AGPase LSs and SSs in E. coli yields functional heterotetrameric enzymes, as detected by iodine staining of glycogen for the product of AGPase [28]. In this study, the coding regions of LpAGPS and LpAGPL were cloned into the expression vectors pACYC-Duet-1 and pRSFDuet-1, respectively. Subsequently, the recombinant plasmids were transformed alone or pairwise into E. coli strain BL21 (DE3) and analyzed by iodine staining. The E. coli BL21 (DE3) expressing individual subunits showed no visible brownish color, indicating that single subunits cannot produce glycogen. Meanwhile, simultaneous expression of LpAGPS1/LpAGPL3, LpAGPS2/LpAGPL1, LpAGPS2/LpAGPL2, and LpAGPS2/LpAGPL3 formed functional enzyme with cells displaying a brownish color when exposed to iodine staining ( Figure 5). Because the color depth of the stained cells was determined by glycogen accumulation, indicating AGPase activities, it can be deduced that LpAGPS2 may have higher regulative activity than LpAGPS1.

Subcellular Localization.
Analysis of LpAGPs subcellular localization was performed to determine if they are plastid or cytoplasmic proteins. The five LpAGP coding regions were cloned into the vector 16318-hGFP, which was then transformed into A. thaliana protoplasts. Fluorescence images of the protoplasts showing the GFP expression were taken using a confocal microscope. The results demon-strated that the five AGPase subunits of L. punctata are uniformly located in the chloroplast, indicating that they are plastid proteins (Figure 6, Figure S1-S4). Notably, the AGPase LSs are possibly located on the cell membrane ( Figure S3, S4).

Discussion
Duckweed is a regenerable bioenergy crop with enormous potential for boosting energy supplies. Its bioethanol yield is about 50% higher than maize, demonstrating a promising starch source for bioethanol production [29]. The genomic characteristics of two Spirodela and Lemna species have been reported [30,31]. Although less studied, L. punctata can quickly accumulate high starch levels under nutrient starvation or uniconazole [11,32]. However, the mechanism of its starch accumulation remains unknown. Therefore, studying how L. punctata differs from other species is essential in utilizing its full potential.
Wang et al. [31] reported the first genome of Lemnaceae: S. polyrhiza in 2014. Based on their results, S. polyrhiza genome has 158 Mb and encodes 19,623 proteins, including four AGPase, one GBSS, four soluble starch synthase (SSS), and three starch branching enzyme (SBE) coding genes. In comparison to S. polyrhiza, L. punctata has five AGPase, two GBSS, two SSS, and five SBE coding genes. The three genes encoding AGPase large subunits of S. polyrhiza have been cloned; however, the small subunit coding gene is yet to be cloned. A comparison between the three AGPase LSs of L. punctata and the AGPase of S. polyrhiza shows that the similarity between LpAGPL1 and SpAGPL1 (Genbank AEV40471), LpAGPL2 and SpAGPL2 (AEV40472), and LpAGPL3 and SpAGPL3 (AEV40474) is 94.39%, 95.52%, and 83.30%, respectively. These results suggest that the LpAGPSs and SpAGPSs coding genes may play similar functions. Unlike S. polyrhiza, which develops a seed-like turion for storing starch [33], L. punctata has a different storage mechanism. Metabolic pathway analysis showed that nutrient starvation and uniconazole treatment enhance starch anabolism but weaken starch catabolism in L. punctata, resulting in high starch accumulation by modulating the global carbon metabolism flux [34].
Full-length LpAGPS1 and LpAGPL1 cDNA were previously cloned from L. punctata strains 0202 [35] and 5632 [36]. The lengths of LpAGPS1 (1,578 bp) and LpAGPL1 (1,554 bp) cDNA were the same in the two L. punctata strains. In contrast, the corresponding sequences cloned in 2015 (LeAPS: KJ603243.1; LeAPL1: KJ603244.1) by Taq DNA polymerase are different from the sequences cloned in this study [35]. Here, we cloned all the five LpAGP cDNAs. Phylogenetic analyses showed that three LpAGPLs (LpAGPL1, LpAGPL2, and LpAGPL3) are more conserved than two LpAGPSs (LpAGPS1, LpAGPS2). Notably, LpAGPS2 has a rather distant phylogenetic relationship with the other four LpAGPs. The sequence similarity between LpAGPS2 and the AGPase small subunit of Theobroma cacao was 85.13%. Furthermore, we found that LpAGPS2 has a high degree of sequence homology to many GDPmannose pyrophosphorylase (GMPPA, EC: 2.7.7.13),  [39], uniconazole [40,41], abscisic acid (ABA) [42], nutrient starvation [11], and pectinase [43] can stimulate high starch accumulation in duckweed. For example, the starch levels of S. oligorrhiza increased to 75% DW [44] when cultivated in a glucose medium without phosphate. This is the highest starch accumulation in duckweed ever recorded. Regarding L. punctata, 52.9% is the highest ever recorded DW starch content [10]. Studies have shown that enzyme activities related to starch biosynthesis increase during the starch accumulation process, especially the activity of AGPase [41]. However, not all AGPase subunit combinations can form active AGPase in plants. AGPase SSs or LSs solely have no catalytical and allosteric properties without subunit synergy [45]. However, there are exceptions; for example, the tomato L3 subunit has catalytic activity as a monomer [46].
In this study, yeast two-hybrid and bacterial coexpression assays were performed to detect the interaction of different subunits. The results showed that LpAGPS1 failed to interact with LpAGPL1 but was weakly associated with LpAGPL2. Additionally, LpAGPS1 interacts with LpAGPL3 to form an active AGPase. LpAGPS2 interacts with all the LpAGPLs; its coexpression in E. coli caused a color change when exposed to iodine staining, indicating that it can interact with LpAGPLs despite being different from most AGPase SSs. Meanwhile, BiFC assay showed that LpAGPS1 interacts with LpAGPL2. This could be because LpAGPS1 and LpAGPL2 can form a functional AGPase with weak activity.
AGPase is typically located in the plastids of most plant tissues. However, most AGPases in mid-developmental stages of rice, maize, barley endosperm, and early-stage wheat endosperm are localized in the cytoplasm [21,26,[47][48][49][50][51]. Two AGPase classes have been reported in plants, including cytosolic AGPase and plastidial AGPase. Correspondingly, AGPase subunits are categorized into small cytosolic, small plastidial, large cytosolic, and large plastidial subunits. In this study, five L. punctata AGPase subunits were fusion-expressed with hGFP in Arabidopsis protoplasts. The results revealed that the green fluorescence of the five fusion-expressed AGPase subunits are all located in chloroplast, suggesting that the five AGPase subunits are all plastidial subunits. In cereals, a significant portion of AGPase activity occurs in the cytosol of endosperms and a minor portion in amyloplasts [47]. In contrast, AGPases of most dicotyledonous plants are plastid proteins. Spatiotemporal expression of AGPases results in different AGPase compositions in various tissues, adapting to specific metabolic requirements of specific tissues.
During starch biosynthesis, various enzymes jointly determine the quality and quantity of starch. Therefore, plant breeders hope to improve starch synthesis and yield by increasing the expression of starch synthesis pathwayrelated enzymes, in which AGPase is considered important [52]. As a fast-growing aquatic plant, duckweed can adapt to the various pH and temperature ranges [53]. Duckweed is used for industrial production of ethanol and enzymes [54] and used as a bioreactor for producing antibodies, vaccines, etc. due to its relatively lower production cost [55,56]. However, most have studies focused on the industrial application of Spirodela, Lemna, and Wolffia belonging to Lemnaaceae, while studies on Landoltia are rarely reported. Thus, this study lays a theoretical foundation on the application of Landoltia in industries, medicine, and other fields.

Conclusion
In this study, five LpAGPs of L. punctata (duckweed) were cloned and characterized. Nutrient starvation significantly upregulated the expression of LpAGPS1, LpAGPL2, and LpAGPL3. The yeast two-hybrid and BiFC assays revealed the interactions between LpAGPS1/LpAGPL2, LpAGPS1/ LpAGPL3, LpAGPS2/LpAGPL1, LpAGPS2/LpAGPL2, and LpAGPS2/LpAGPL3. All the five LpAGPs are localized in the chloroplast. These results pave the way for future functional tests of LpAGPs in vivo.

Conflicts of Interest
The authors declare no competing interests.