Expression of starch-binding factor CBM20 in barley plastids controls the number of starch granules and the level of CO₂ fixation

Interference in protein–starch interactions in barley by the expression of an exogenous high-affinity starch-binding factor (Carbohydrate-Binding Motif 20, CBM20) reveals a link between starch granule biosynthesis and the level of photosynthesis.

Abstract

The biosynthesis of starch granules in plant plastids is coordinated by the orchestrated action of transferases, hydrolases, and dikinases. These enzymes either contain starch-binding domain(s) themselves, or are dependent on direct interactions with co-factors containing starch-binding domains. As a means to competitively interfere with existing starch–protein interactions, we expressed the protein module Carbohydrate-Binding Motif 20 (CBM20), which has a very high affinity for starch, ectopically in barley plastids. This interference resulted in an increase in the number of starch granules in chloroplasts and in formation of compound starch granules in grain amyloplasts, which is unusual for barley. More importantly, we observed a photosystem-independent inhibition of CO₂ fixation, with a subsequent reduced growth rate and lower accumulation of carbohydrates with effects throughout the metabolome, including lower accumulation of transient leaf starch. Our results demonstrate the importance of endogenous starch–protein interactions for controlling starch granule morphology and number, and plant growth, as substantiated by a metabolic link between starch–protein interactions and control of CO₂ fixation in chloroplasts.

Introduction

Biosynthesis of starch granules in plant plastids is carried out by the orchestrated action of enzymes directly involved in, or related to, α-glucan elongation and modification. Elongation is carried out by specialized glycosyltransferases termed starch synthases (SSs), which use ADP-glucose as donors of glucose units. Subsequently, α-1,6 branches are introduced by starch branching enzymes (SBEs). Maltooligosaccharides (MOSs) may act as primers for starch granule initiation and synthesis. MOSs are synthesized from units of glucose-1-phosphate by a combined action of α-glucan phosphorylase and α-1,6 starch branching enzymes (for reviews see Blennow et al., 2013; Tetlow and Emes, 2017). Starch granules are between 1–100 µm in size, with different morphologies depending on the species, such as round, ovoid, or polyhedral (Bertoft, 2017). In extreme cases, such as in starch granules that consist only of amylose, more amorphous granules are formed (Carciofi et al., 2012a). Regular starch granules have so-called growth rings with a thickness of 0.2–0.5 µm that alternate between semi-crystalline and amorphous blocklets (Goldstein et al., 2017).

Little is known about the organization of enzymes within and on the surface of the starch granules during their synthesis. Some starch-related enzymes bind through starch-binding domains (Pfister and Zeeman, 2016); however, others, such as granule-bound starch synthases (GBSSs), do not have such domains. Instead, these enzymes may interact with co-factors with starch-binding domains that guide them to starch granules. GBSSs have been shown to be transported to the granules by Protein Targeting to Starch 1 (PTST1) (Seung et al., 2015). The GBSS is then trapped within the granules during their synthesis and is clearly associated with the growth rings of the granules (Hebelstrup et al., 2017). In line with this, the proteome of starch granules to a large extent consists of GBSSs (Cao et al., 2015). Starch synthases operate by enzymatic mechanisms that are either processive or distributive. In a processive mechanism, elongation continues at the same acceptor α-glucan chain once it has initiated. In a distributive mechanism, the elongation is terminated after a limited number of reactions, and then it re-initiates at a new acceptor chain. Soluble starch synthases are generally not completely trapped within the starch granule, possibly because they elongate starch chains by a purely distributive mechanism (Cuesta-Seijo et al., 2015). In contrast, GBSSs operate by a processive mechanism, and hence they feature more largely in the proteome of starch granules than the soluble starch synthases (Cao et al., 2015). Previous studies have shown that loss of processive ability of GBSS results in it being trapped in small structures (≤1 µm) that are distributed separately from the starch granules (Hebelstrup et al., 2017). This indicates that the processive mechanism of GBSS is the reason that it is granule-bound. Plant starches mostly serve two different functions. Transient starch in chloroplasts is a way to temporarily store for later use carbohydrate that is formed during periods of high photosynthesis. In the leaf mesophyll this can usually be detected as a diurnal cycle of starch content, with transient reserves peaking at the end of the day and becoming minimal at the end of the night. In contrast, storage starches are deposited mainly for reproductive purposes in amyloplasts for a much longer time, often in seeds and storage organs such as tubers. Due to their different physiological functions, there are significant differences in biosynthesis and structure between transient and storage starches. In addition, granules of storage starches vary between organs and species, and between cultivars of individual crops. For example, rice and oat grain starch granules are of a compound type, where several small granules combine into larger ones, which often break apart into smaller polyhedral and sharp-edged granules during purification of granules in agricultural processing. It has been suggested that this characteristic shape is determined within amyloplasts by septum-like structures that separate individual granules (Yun and Kawagoe, 2010; Kawagoe, 2013). For transient leaf starch, there seems to be a link between chloroplast division and the number of granules per chloroplast (Crumpton-Taylor et al., 2012).

The mechanisms for granule initiation in both storage and transient starches are mostly unknown. There are several mutants with decreased numbers of storage granules per chloroplast, suggesting that the mutated proteins are involved in starch initiation. Mutation of the gene for soluble starch synthase IV (SSIV) in Arabidopsis results in the number of starch granules per chloroplast being limited to one. Although SSIV is a starch-interacting protein, it does not contain any known starch-binding domain. Instead, it binds though coiled-coil domains to the co-factor Protein Targeting to Starch 2 (PTST2), which has a starch-binding domain of the Carbohydrate-Binding Motif 48 (CBM48) family, and it has been suggested that PTST2 mediates the interaction between SSIV and MOSs for the initiation of transient leaf starch granules in Arabidopsis (Seung et al., 2017). PTST2 also binds to other plastidial proteins such as the thylakoid-associated protein MFP1, which is involved in granule initiation (Seung et al., 2018). Arabidopsis does not make storage starch, but a homolog of PTST2 in rice called FLO6 has been characterized that seems to be involved in the formation of grain storage granules, since their compound-like pattern is partly lost in flo6 loss-of-function mutants (Peng et al., 2014). A homolog of PTST2/FLO6 called FRA exists in barley, and fra loss-of-function mutants have fractured starch granules (Saito et al., 2018). The interactions of carbohydrate-binding domains with starch have also been shown to form a structural basis for control of the catalytic activity of enzymes in signal transduction. An example of this is the regulatory protein SnRK1 that is involved in responses to low-energy stress and in sensing of cellular carbon status. Subunits of SnRK1 contain carbohydrate-binding motifs that regulate its activity negatively when bound to starch (Ávila-Castañeda et al., 2014) and positively when bound to maltose (Ruiz‐Gayosso et al., 2018).

Despite of the obvious link between photosynthesis and starch biosynthesis, surprisingly few studies have examined starch metabolism in relation to control of photosynthesis. Photosynthesis is divided in two linked mechanisms: the light reactions that take place in chloroplast thylakoid membranes, where PSI and PSII harvest photons to produce NADPH and ATP; and the carbon reactions where NADPH and ATP are consumed by stromal enzymes of the Calvin–Benson cycle to fix CO₂. ssiv mutants and in particular ssiiissiv double-mutants show signs of inhibition of both the carbon reactions, as indicated by reduced CO₂ assimilation, and the light reactions, as indicated by reduced PSII efficiency. This inhibition of photosynthesis seems to be due to an accumulation of ADP-glucose, the content of which is more than 100-fold higher in these mutants than in wild-type plants (Ragel et al., 2013). These mutants also accumulate significantly higher levels of sucrose, which down-regulate photosynthesis.

There are different families of carbohydrate-binding motifs (CBMs). Some starch synthases and starch-modifying enzymes contain CBMs that allow them to target to starch granules (Pfister and Zeeman, 2016) whilst others depend on indirect interactions through other proteins that contain CBMs (Seung et al., 2015; Pfister and Zeeman, 2016). Glucoamylase from Aspergillus niger contains a domain of the CBM20 family that has a very high affinity for starch (Christiansen et al., 2009; Cuesta-Seijo et al., 2015; Tanackovic et al., 2016), and in this present study we examined whether this module (hereafter termed simply CBM20) may provide a means by which to explore starch–protein interactions in plastids. Our assumption was that since CBM20 has a very strong affinity for starch, it will competitively interfere with the large array of starch–protein interactions in plastids. We constructed seven independent transgenic lines of barley that ectopically express a synthetic protein, TP-CBM20-eGFP, in plastids. This protein consists of a transit peptide from barley Granule-Bound Starch Synthase Ia (GBSSIa), the CBM20 domain, and the fluorescent tag eGFP. The transit peptide guides the protein to plastids. It does not have any catalytic domains but despite this, the CBM20-expressing plants showed an increased number of disorganized starch granules in the chloroplasts. The plants also showed a transition to compound-type storage granules in grains, a significantly decreased C:N ratio in the leaves, and significantly lower photosynthesis, which seemed to be specifically regulated at the level of the Calvin–Benson cycle.

Materials and methods

Engineering of vectors and transformation of barley

Independent transgenic lines of barley (Hordeum vulgare cv Golden Promise) with expression of CBM20 from Aspergillus niger isolate G1 and vector control lines were generated by transformation with Agrobacterium tumefaciens (AGL0) (Hebelstrup et al., 2010; Carciofi et al., 2011, 2012b). The plasmid vectors pUCE-UBI:TP-eGFP:NOS and pUCE-UBI:TP-eGFP-CBM20:NOS were generated by In-Fusion® HD cloning (Takara Bio Europe) into the vector pUCE-UBI:TP:NOS (Hebelstrup et al., 2010), which has a unique PacI restriction site (TTAATTAA) between the transit peptide (TP) and the termination site (NOS). Primers used to clone the green fluorescent protein (eGFP) were: forward, 5´-CGCGCTGAGGT CTTAATCATGGT GAGCAAGGGC GAGGAG-3´ and reverse, 5´-AATGCTGAGGCATT AATTAAGCCCTTGTAC AGCTCGTCC ATGCCGAG-3´.

GFP was added in the plasmid as a marker to visualize the transgenic tissue in the embryonic callus. CBM20 was synthesized artificially by GenScript, and the codon was optimized for expression in H. vulgare. The primers used were: forward, 5´-CTGTACAAGGGCTTA ATTAATATGTGTAC GACCCCTACAGCCG-3´ and reverse, 5´-AATGCTGAGGCATTAA TTTACCTCCATGT GTCGGTAACGGT-3´. The primers used for the cloning procedure were designed following the In-Fusion protocols.

Plants were grown in a growth chambers with a 16/8 h light/dark photoperiod and temperature 12–14 °C.

Western blotting and measurements of Rubisco content

The 2nd leaves from the top of the plants (3.5 months old) were collected from the transgenic lines TP-eGFP-CBM20 and TP-eGFP and from the Golden Promise wild-type (WT). Leaf samples (100 mg FW) were ground in an automatic cell homogenizer (FastPrep™ FP120, Thermo Savant) for 20 s. Extraction buffer [1.0 ml 50 mmol l^–1 Tris, 5 mmol l^–1 β-mercaptoethanol, 12.5 % (v/v) glycerol] was added to ground samples and mixed by vortexing. The mixture was then centrifuged at 15 000 g for 15 min at 4 °C. The pellet was mixed with 100 µl 1× SDS-loading buffer [10 mmol l^–1 Tris, 1% (w/v) SDS, 10 % (v/v) glycerol, 0.4 mg ml^–1 bromophenol blue) and the samples were boiled for 10 min and centrifuged at 15 000 g for 2 min. The supernatant was used for SDS-PAGE with NuPAGE Novex (ThermoFisher Scientific).

Bis-Tris Mini Gels (4–12%, Invitrogen) were used in conjunction with Bio-Rad Mini-PROTEAN II Multiscreen Apparatus according to the manufacturer’s instructions (Bio-Rad bulletin 1721). Purified antibodies from rabbit serum raised against eGFP (Sigma-Aldrich, G1544) at a dilution of (1:1000) were used to detect eGFP-tagged TP or TP-CBM20. The secondary antibody (1:5000) was goat anti-rabbit purified IgG coupled with alkaline phosphatase (Sigma-Aldrich, A3687).

Quantification of Rubisco from SDS-PAGE gels was conducted using Image Lab software. During the quantification procedure, a standard protein BSA (0.1 mg ml^–1) was separately loaded in 5, 10, 15, and 20 µl volumes to four lanes on the same gel. The standard proteins were used to give a standard curve of known concentration. The content of Rubisco in each sample lane was then quantified.

Confocal laser-scanning microscopy

Granular localization in leaves was observed using an Olympus Fluoview FV1000 confocal microscope as described previously (Carciofi et al., 2012b; Hebelstrup et al., 2017). No fluorescence was detected in starch granules from non-transformed WT grains.

Endosperm was isolated from dry mature grains of transgenic plant TP-eGFP-CBM20 and WT plants. Transmission confocal laser-scanning microscopy was used to examine the morphology of starch granules.

Primary metabolite analysis

Relative levels of primary metabolites were measured as described in Hasler-Sheetal et al. (2016). In brief, 50-mg FW samples (2nd leaves from 4-month-old plants) from six independent T₀ CBM20-expressing plants and six vector control plants were extracted for 2 min in an ultrasound bath followed by 15 min on a thermo-shaker (both at 4°C) in 1 ml methanol/acetonitrile/water (4:4:2 v/v/v) chilled to –20°C and spiked with 0.4 mg l^{–1 13}C6 sorbitiol and reserpine as internal standards. After centrifugation at 16 000 g for 5 min, the relative metabolite levels were analysed by GC-quadrupole time-of-flight MS (7200 GC-QTOF MS, Agilent Technologies).

Staining of starch in leaves and purification of leaf starch

The second leaves from 3.5-month-old vector control plants and four independent T₀ TP-eGFP-CBM20 plants were collected after a 12-h light period or a 24-h dark period. Pigments were extracted in 80% (v/v) ethanol at 80 °C for 30 min and then stained for 10 min with Lugol’s iodine solution (stock: 250 mg I₂, 2.5 g KI, 125 ml ddH₂O; freshly diluted 1000× in 100 mM HCl). The Lugol solution was removed after ~10 min and the leaves were washed with 100 mM HCl to remove the excess iodine. White transmission light was used to observe starch granules using an Olympus Fluoview FV1000 confocal microscope.

Samples of 1–2 g FW of 2nd leaves from the top of TP-eGFP-CBM20 and vector control plants were collected and starch granules were purified as described by (Albi et al., 2014), but with Ficoll 400 instead of Percoll. Samples of 0.9 mg of the purified granules were analysed by SDS-PAGE.

Photosynthetic fluorescence induction curves

To evaluate differences in PSII efficiency between vector control and T₀ TP-eGFP-CBM20 plants, fluorescence induction curves were determined. Plants (three replicates) were dark-adapted for 40 min at room temperature. Measurements were taken with a MINI-PAM chlorophyll fluorescence meter (Waltz). Prior to dark adaptation, the leaf-clip holder 2030-B fitted with the standard MINI-PAM/F fiber optics (both Walz) was attached to the latest fully unfolded leaf. The MINI-PAM set points followed factory settings except for actinic light, which was set to 19 μmol m^–2 s^–1. F-zero was set once prior to measurements. The measuring light was turned on 30 s before starting a measurement. Measurements were carried out over a period of 5 h that was approximately centred around the middle of the photoperiod

CO₂ response curves and photosynthetic light-response curves

Mean CO₂ response (A/C_i) curves for three WT and three independent T₀ CBM20-expressing plants were obtained using a LI6400 (LI-COR) coupled to a 6400–02 red LED light source. The automated A/C_i program for the LI6400 was utilized with the following settings: leaf area, 3 cm²; bulk temperature, 20 °C; light, 1500 µmol m^–2 s^–1; and CO₂ concentrations of 400, 300, 250, 200, 150, 100, 50, 400, 700, 900, 1000, 1200, 1500, and 1800 ppm. One measurement was taken per CO₂ level on the 2nd leaf from the top of the plant. Plants were 4 months old. Prior to starting, the leaf was kept in the leaf chamber for 20 min. The intervals between readings were 10–12 min, except that an interval of 20 mins was used after increasing the CO₂ from 50 ppm up to 400 ppm. Measurements were done at room temperature on the latest fully unfolded leaf. The leaf vapor deficit varied between 0.7–2.0 kPa. Photosynthetic light-response curves (A/Q) were also determined. The basic settings were the same as for the A/C_i curves with the exception that the CO₂ level was constant at 400 ppm, and the values of photosynthetically active radiation (PAR) in the chamber were 400, 350, 250, 150, 120, 60, 40, 400, 700, 1000, 1500, and 1800 µmol m^–2 s^–1. Modelling of the A/C_i curves was done using ‘fitaci’ from the R package plantecophys (Duursma, 2015). Prior to modelling of the A/C_i curves, all start measurements at 400 ppm CO₂ were removed and the model parameters were kept as standard. Modelling focused on the maximum rate of Rubisco carboxylation (V_c,max), the maximum rate of electron transport used in the regeneration of ribulose 1,5-biphosphate (RuBP; J_max), and the respiration in the light (R_d,A/Ci). The triose phosphate use (TPU) part of the A/C_i curve was not added to the modeling.

Estimates for differences were calculated by one-way ANOVA and pairwise comparisons were done using glht() from the multcomp package (Hothorn et al., 2008). For multiple tests, P-values were adjusted to control for Type-I errors by using the multivariate T-distribution, which computed the maximum of the involved test statistics (Bretz et al., 2011).

A/Q curves were fitted as non-rectangular hyperbola using the nls() functions available in R.

$${\displaystyle \begin{array}{l}{\displaystyle {\mathrm{A}}_{\mathrm{n}}=\frac{\mathrm{\alpha }I{A}_{\mathrm{n}}-\sqrt{{\left(\mathrm{\alpha }I{A}_{\mathrm{m}}\right)}^2-4\mathrm{\alpha }I{A}_{\mathrm{m}}\mathrm{\Theta }}}{2\mathrm{\Theta }}-R_{\mathrm{d}A/Q}}\end{array}}$$

Where A_n is the net rate of photosynthesis in light (µmol CO₂ m^–2 s^–1), A_m is the maximum capacity of photosynthesis at 400 ppm, α is the apparent quantum yield (mol mol^–1), I is the irradiation in the PAR region (µmol m^–2 s^–1), Θ is the convexity (dimensionless), and R_d,A/Q is the respiration in the light (µmol CO₂ m^–2 s^–1).

First, the mathematic non-rectangular hyperbola used for the A/Q response curve was solved as the full model with four parameters (A_m, α, Θ, and R_d,A/Q). As Θ depends on several factors (Ögren and Evans, 1993) interpretation is limited, and hence the A/Q models were re-calculated with Θ fixed at 0.85. Akaike information criteria values (AIC) on the Θ parameter showed in general no restrictions for this reduction of the mathematical non-rectangular hyperbola.

C/N ratio

The 2nd leaf from the top of 4-month-old plants were sampled from WT, vector control, and independent T₀ CBM20-expressing lines (six plants each). All samples were freeze-dried for 48 h and ground through a 1-mm sieve. Nitrogen and carbon concentrations in the ground samples were measured using elemental analyser (vario EL-III, Germany).

Results

Expression of CBM20 in plastids

We generated plants with ectopic expression of the starch-binding domain (CBM20) from glucoamylase of A. niger G1 (1,4-α-D-glucan glucohydrolase, E.C. 3.2.1.3). This variant of CBM20 has a very strong affinity for starch and α-glucan (Tanackovic et al., 2016). We added an N-terminal transit peptide (TP) from barley granule-bound starch synthase (GBSSIa) as previous studies have shown that this TP effectively targets the transgenic proteins to plastids, and that it is effectively cleaved off during passage of the protein over the plastid envelope membranes (Carciofi et al., 2012b; Hebelstrup et al., 2017). To visualize the location of the transgenic CBM20 by fluorescence microscopy, we added the fluorescent tag eGFP to the construct, generating the TP-eGFP-CBM20 protein. For the vector control, we also generated plants with only the TP and the eGFP domains (TP-eGFP). The TP (67 aa) has a molecular weight of 7 kDa, eGFP (238 aa) has a molecular weight of 27 kDa, and CBM20 (108 aa) has a molecular weight of 12 kDa. Therefore, if the pre-protein of this construct is translocated into the plastids and the N-terminal TP is cleaved off, the final molecular weight of eGFP-CBM20 would be 39 kDa. Using western blotting with antiserum raised against eGFP, we confirmed the presence of a band at that size in extracts from leaves of transgenic (T0) plants expressing the TP-eGFP-CBM20 construct (Fig. 1A). There was no sign of the non-cleaved pre-protein (46 kDa). Similarly, for the vector control expressing TP-eGFP, a band with the expected size of eGFP (27 kDa) was observed, and there were no signs of a non-cleaved protein (34 kDa). Fluorescence microscopy confirmed that the transgenic TP-eGFP-CBM20 as well as the vector control TP-eGFP were correctly targeted to chloroplasts in leaves (Fig. 2).

Fig. 1.

(A) Western blotting of barley leaf proteins using antiserum raised against eGFP. No eGFP was detected in the wild-type (WT) plants. In transgenic vector control plants with expression of TP-eGFP (TP), a band of the expected size for eGFP at 27 kDa was observed whilst plants with expression of TP-eGFP-CBM20 showed a band of the expected size at 39 kDa. All lanes are from the same blot. (B) Images of representative plants at 4 months old. T₀ plants with strong expression of TP-eGFP-CBM20 showed a strong inhibition of growth.

Fig. 2.

Confocal laser-scanning microscopy of leaves from transgenic barley plants. (A–C) TP-eGFP-CBM20-expressing plants, (D–F) vector control plants expressing TP-eGFP, and (G–I) wild-type (WT) plants. (A, D) TP-eGFP-CBM20 and TP-eGFP were detected by their green fluorescence, and (B, E, H) chloroplasts were detected by their red fluorescence. (C) TP-eGFP-CBM20 was located as scattered dots inside chloroplasts while (F) TP-eGFP showed an even distribution throughout the chloroplasts. (I) No green fluorescence was observed in WT plants. Scale bars are 10 µm.

We produced seven independent lines of T₀ plants with expression of TP-eGFP-CBM20. Six of these were used in the subsequent analysis. They all showed very slow growth and after 4 months the TP-eGFP vector controls and the WT plants were much larger than those expressing CBM20 in plastids (Fig. 1B). CBM20-expressing plants continued to produce new tillers that could be propagated vegetatively; however, the stem apical meristem of the tillers either remained dormant or the stem grew only a few centimeters above the root (not shown). Only on one occasion did a single tiller eventually develop into a spike, which produced only one seed. One of the independent T₀ plants with only low expression of the transgenic CBM20 (Fig. 1A, lane 4) showed less retardation of growth than lines with higher levels of expression (lane 3). Several of the vector control lines showed levels of transgenic eGFP that were higher than those of the eGFP-CBM20 lines (Fig. 1A, lane 2), but none of them showed any sign of growth retardation, excluding a growth effect of eGFP itself.

Starch granule numbers in chloroplasts

We used confocal laser scanning microscopy to observe the localization of eGFP in the chloroplasts of the seven T₀ lines and the vector controls (Fig. 2A–F). Chloroplasts were identified by red autofluorescence from chlorophyll (Fig. 2B, E, H). Vector controls showed a distribution of eGFP throughout the chloroplast, in agreement with a stromal localization (Fig. 2F), whereas eGFP-CBM20 was localized in small granules, in agreement with its expected binding to starch (Fig. 2C, Supplementary Fig. S2 at JXB online). These granules disappeared when leaves were kept in darkness for 24 h (Supplementary Fig. S3). Surprisingly, there were many more and smaller starch granules in the eGFP-CBM20 plants than in the vector controls, and they often contained only one or two granules per chloroplast (Fig. 3, Supplementary Fig. S3). In addition, the granules in eGFP-CBM20 plants were randomly located, whereas those in the vector controls were well organized toward the center of the chloroplasts. This was evident both by fluorescence detection (Fig. 2, Supplementary Figs S2, S3) and by iodine staining (Fig. 3). To further examine the binding of eGFP-CBM20, we purified leaf starch granules to confirm their eGFP fluorescence (Fig. 3C). A similar eGFP signal was not detected in granules from the vector controls (Fig. 3E, F). The results of SDS-PAGE of proteins from the leaf starch granules were also in agreement with the presence of the 39-kD eGFP-CBM20 protein (Supplementary Fig. S4), and demonstrated a difference in the ratio of individual levels of endogenous starch-bound proteins between the transgenic and vector control plants.

Fig. 3.

Transmission light (laser) micrograph of iodine-stained starch granules from leaves of transgenic barley plants expressing (A) TP-eGFP-CBM20 and (B) vector control plants. Scale bars are 5 µm. There are many more and smaller starch granules randomly located within the chloroplasts in the eGFP-CBM20 expressing plants than in the controls, which often contain only one or two granules located at the centers of the chloroplasts. The insets show magnified images of the areas marked by the rectangles. (C–F) Confocal laser-scanning micrographs of purified starch granules from leaves of (C, D) TP-eGFP-CBM20 plants and (E, F) control plants; (C, E) show eGFP fluorescence channels and (D, F) show laser transmission channels. Scale bars are 2 µm.

Starch levels and metabolism

To determine how the expression of CBM20 affected metabolism, we measured the relative concentrations of primary metabolites in leaves from six independent T₀ plants at mid-day using GC-MS analysis. We focused on primary metabolites that are part of the Calvin–Benson cycle, glycolysis, and the citric acid cycle and on amino acid metabolism related to these pathways (Fig. 4). For the Calvin–Benson cycle, we identified relative concentrations of the intermediates 3-phosphoglyceric acid (PGA) and glyceraldehyde-3-phosphate (G3P). PGA is the product of the first step of the fixation in the cycle, where RuBP reacts with CO₂ to form two molecules of PGA in a reaction catalysed by Rubisco. G3P is the output metabolite of the cycle and in plastids it can be converted into ADP-glucose, which enters the starch biosynthesis pathway. Alternatively, when the leaves act as a source tissue, G3P is exported from the plastid to the cytosol where it is converted into sucrose, which is the major metabolite of the water-soluble carbohydrate pool in plant leaves. G3P exported from the plastid may also enter the glycolytic pathway in the cytosol. The concentrations of PGA, G3P, and sucrose as well as fructose and glucose were all significantly lower in eGFP-CBM20-expressing leaves than in vector control leaves, and the content of Rubisco was also lower Supplementary Fig. S1. This indicated lower photosynthetic productivity in plants with expression of eGFP-CBM20.

Fig. 4.

Relative concentrations of primary metabolites in leaves of barley plants expressing TP-eGFP-CBM20 and vector control plants. Red boxes indicate metabolites that have a significantly lower concentration in CBM20-expressing plants compared to the control, and blue boxes indicate metabolites that have a significantly higher concentration in CBM20 compared to the control. Black boxes indicate no significant difference. Relative concentrations are shown on a log₂-scale on the y-axis. 2-OG, 2-oxoglutarate; 3-PGA, 3-phosphoglyceric acid; G3P, glyceraldehyde-3-phosphate; PEP, phosphoenolpyruvate.

When glucose and other carbohydrates are metabolized in leaves they enter the glycolytic pathway. We identified and quantified the levels of the final products of cytosolic glycolysis, namely phosphoenolpyruvate (PEP) and pyruvate. In accordance with lower levels of the sugars that are the input for glycolysis, the levels of PEP and pyruvate were also lower. The relative concentrations of metabolites in the citric acid cycle were generally lower in plants expressing eGFP-CBM20 compared with the vector controls except for oxaloacetate, which is a substrate of citrate synthase that also uses actetyl-CoA produced from pyruvate. This step is the metabolic link between glycolysis and the citric acid cycle, and the accumulation of oxaloacetate may therefore have been due to lack of pyruvate. To examine whether the low levels of photoassimilates had an impact on starch biosynthesis, we produced a qualitative score of relative starch levels in leaves at the end of periods of light and darkness by means of iodine staining (Fig. 5). Starch levels were generally highest near the central leaf veins, and were higher in leaves from vector control plants than in those containing CBM20-bound starch, both at end of periods of light and darkness. Interestingly, the contents of glutamine and glutamate, both with a high N/C ratio, were higher in eGFP-CBM20 plants than in vector controls (Fig. 4). This was consistent with weaker carbon fixation relative to N assimilation in eGFP-CBM20 plants, which was confirmed by elemental analysis that showed a higher N content and lower carbon to nitrogen (C:N) ratio in these plants (Fig. 6).

Fig. 5.

Qualitative identification of relative starch levels in barley leaves at the end of periods of light and darkness as determined by iodine staining. (A, B) Leaves subjected to 24 h in darkness for (A) control plants and (B) plants expressing TP-eGFP-CBM20. (C, D) Leaves subjected to 12 h in the light for (C) control plants and (D) plants expressing TP-eGFP-CBM20.

Fig. 6.

Carbon and nitrogen contents (w/w) and C:N ratios of transgenic barley plants expressing TP-eGFP-CBM20, vector controls expressing TP-eGFP (control), and the wild-type (WT). Data are means (±SE) of six replicates. Different letters indicate significant differences as determined using one-way ANOVA followed by Tukey’s test (P<0.05).

Photosynthesis

The relatively lower level of the products of photosynthesis in the plants with expression of eGFP-CBM20 in the chloroplasts suggested that they might have a lower level of CO₂ fixation, and so we produced A/C_i and A/Q curves. The value of R_d,A/Ci did not vary significantly between vector control and CBM20-expressing plants (Table 1). Significant differences were observed for V_c,max and J_max according to the A/C_i curves (Fig. 7A–C), which suggested a lower level of CO₂ fixation in the eGFP-CBM20 plants. Stomatal densities and apertures were also found to be significantly different (Table 1). Using the data from the A/Q curves, stomatal conductance (g_s) was evaluated under ambient conditions (CO₂ 400 ppm; PAR 1500 μmol m^–2 s^–1). No differences in g_s were observed between vector control and eGFP-CBM20 plants, whilst C_i values were significant higher in the eGFP-CBM20 plants (Table 1). The difference in C_i seemed to originate from lower CO₂ accumulation as evidenced by the reduced photosynthetic rate (A_n) in the eGFP-CBM20 plants. That A_n was the limiting factor was further confirmed by the fact that the relationship between photosynthetic rate and stomatal conductance was more dependent on A_n than g_s (Fig. 7C). Differences between eGFP-CBM20 and vector control plants were found for both the slope (P≤0.001) and the intercept (P≤0.01).

Table 1.

Photosynthesis and stomatal parameters for barley plants expressing TP-eGFP-CBM20 and vector control plants

Parameter	Vector control	CBM20	t-test
A/Ci
V c,max (µmol CO2 m–2 s–1)	101.2±7.7	34.5±1.8	**
J max (µmol e– m–2 s–1)	122.8±4.0	58.9±8.1	**
R d,A/Ci (µmol CO2 m–2 s–1)	0.50±0.65	0.62±0.22	NS
A/Q
α (mol mol–1)	0.02±0.002	0.03±0.006	NS
A m (µmol CO2 m–2 s–1)	15.7±1.0	8.1±1.0	**
R dA/Q (µmol CO2 m–2 s–1)	0.25±0.08	0.03±0.37	NS
Photosynthesis at 400 µmol CO2 mol–1 and PAR 1500 μmol m–2 s–1
g s (mol H2O mol–2 s–1)	0.20±0.02	0.21±0.03	NS
A n (μmol CO2 m–2 s–1)	14.7±1.1	8.02±0.9	**
C i (μmol CO2 mol–1)	256.0±5.3	322.3±1.5	***
F v/Fm	0.79±0.004	0.77±0.004	*
Stomata
Length (μm)	31.33±0.71	29.34±0.78	*
Height (μm)	17.61±0.40	15.51±0.41	***
Aperture (μm)	6.94±0.33	3.84±0.44	***
Density (number mm–2)	91.7±1.1	143.4±3.1	***

A/C_i: PAR was set to 1500 μmol m^–2 s^–1. V_c,max, maximum velocity of Rubisco carboxylation; J_max, maximum rate of electron transport demand for RUBP regeneration; R_d,A/Ci, A/C_i curve estimation of respiration rate in the light.

A/Q: CO₂ level was set at 400 µmol mol^–1. α, apparent quantum yield; A_m, maximum capacity of photosynthesis; R_d,A/Q, A/Q curve estimation of respiration rate in the light.

*P≤0.05; **P≤0.01; ***P≤0.001; NS, not significant.

Fig. 7.

Photosynthesis in barley plants expressing TP-eGFP-CBM20 and in vector control plants (Control). (A, C) Measured values (n=3 plants) are shown as circles and fitted models are shown as lines. (A) Photosynthesis CO₂-response curves (A/C_i), showing the rate of photosynthesis (A_n) as a function of the internal CO₂ concentration (C_i). (B) Photosynthetic light-response curves (A/Q), showing the rate of photosynthesis as a function of the photosynthetically active radiation (PAR). (C) Relationships between stomatal conductance (g_s) and photosynthetic rate based on the light-response curves. eGFP-CBM20-expressing plants: A_n=37.18g_s– 0.24 (R²=0.89). vector control plants: A_n=91.58g_s–3.92 (R²=0.96). (D–F) Quenching curves based on the induction curve showing recovery for eGFP-CBM20 and WT plants. (D) PSII quantum yield, Y_II and (E) quantum yield of light-induced non-photochemical quenching, Y_NPQ. (F) Quantum yield of non-regulated heat dissipation and fluorescence emission, Y_NO.

The value of F_v/F_m was slightly but significantly, lower in eGFP-CBM20 plants than in vector controls (Table 1). However, as the contents of Rubisco and the RuBP product 3-PGA were lower in eGFP-CBM20, a photosynthetic protection mechanism might have been activated continuously and this could have been reflected in the difference in F_v/F_m. To investigate this, we evaluated the parameters quantum yield of PSII (Y_II), quantum yield of light-induced non-photochemical quenching (Y_NPQ), and quantum yield of non-regulated heat dissipation and fluorescence emission (Y_NO). Y_II was lower for CBM20-expressing plants compared to the vector controls (Fig. 7D). Excess energy was dispersed through Y_NPQ, a photo-protective mechanism, as pH- and/or xanthophyll-regulated thermal dissipation (Fig. 7E). In addition, excess energy was also dispersed through passive dissipation of energy in the form of heat and fluorescence, mainly due to the closed PSII reaction centers (Y_NO) (Fig. 7F). At the end of the dark adaptation, a steep increase in Y_NPQ indicated that CBM20 activated photo-protective mechanisms that removed energy by heat loss, which was seen as a peak in Y_NPQ within the first 80 s. Within this period, the Y_NO in CBM20-expressing plants decreased rapidly to a level comparable with the vector control, indicating opening of PSII reaction centers that had previously closed been closed by the first saturated pulse (F_m). After 145s, Y_NPQ in CBM20-expressing plants reached the level of the vector control whereas Y_NO decreased more slowly than in the vector control, being significantly higher between 167–294 s.

Starch granule formation in amyloplasts

We examined the morphology of starch grains from mature endosperm by transmission light confocal laser-scanning microscopy. Starch granules in barley endosperm are normally divided in two populations according to size, namely A-granules that have diameters in the range 8–15 µm and B-granules in the range 1–3 µm. Whilst the granules in the vector control plants had the normal appearance of large A- and small B-granules, CBM20-expressing plants had a compound structure, with 2–7 smaller granules combined into larger compound granules (Fig. 8A, B). There was no sign of fluorescence from these mature granules (data not shown) and, since only one grain was produced, we were unable to monitor granules in developing grains. As an alternative, we studied the localization of eGFP-CBM20 in transgenic embryonic callus containing starch granules, as described in our previous study (Carciofi et al., 2012b). Full-transgenic calli expressing eGFP-CBM20 generated a mix of single, round starch granules and compound granules made of two sub-granules. This callus is a step in the production of the transgenic plants (Carciofi et al., 2012b). The samples showed localized fluorescence on the starch surface (Fig. 8C), but with a particularly high intensity at the interface between the two granules of a larger compound granule (Fig. 8C, arrows). Such compound granules were not observed in transgenic calli of the vector control (Carciofi et al., 2012b), nor was any fluorescence associated with granules in the vector control embryonic callus (Fig. 8D).

Fig. 8.

Confocal laser-scanning microscopy images of barley starch granules. (A) Transmission light, showing compound starch granules in mature grain endosperm of plants with expression of TP-eGFP-CBM20. (B) Transmission light, showing normal A- and B-type granules in the mature endosperm of vector control plants. (C) Green fluorescence, showing that eGFP-CBM20 binds to the surface of starch granules from embryonic callus with expression of TP-eGFP-CBM20. Compound starch granules occasionally also developed in this embryonic callus, which was generated as described by Carciofi et al. (2012b). The concentration of eGFP-CBM20 in compound starch granules was high at the interface between sub-granules of the compound-type, as indicated by the arrows. (D) No fluorescence was detected in starch granules from control embryonic callus. Scale bars are 10 µm in (A, B) and 2 µm in (C, D).

Discussion

CBM20 binds to starch granules in transgenic plants

Starch synthases and other starch-related enzymes interact with starch during its synthesis. Some do this directly through carbohydrate-binding modules (CBMs) of the starch-binding domain type (SBDs) whereas others interact with CBM-containing co-factors. CBMs are classified into different families, as presented in the Carbohydrate-Active enZYmes database (www.cazy.org) (Lombard et al., 2014). At present, the database contains 85 different families, of which 13 are classified as SBDs. CBM20 is one of these SBDs, and entries with CBM20 modules are reported from organisms with broad phylogenetic diversity, including archaea, bacteria, and eukaryotes (Christiansen et al., 2009). The CBM20 used in the present study was the 12 kDa starch-binding domain (residues 509–616) of glycoamylase from Aspergillus niger, which was chosen based on a specific and high affinity for starch (Tanackovic et al., 2016). The affinity of CBM20 for starch is also significantly higher than that of the barley starch synthases GBSSI, SSI, SSIIa, SSIIIa, and SSIV (Cuesta-Seijo et al., 2015).The starch synthases GBSSI, SSI, SSII, and SSIV do not have any known CBMs; however, SSIII contains a CBM25 (Pfister and Zeeman, 2016). Other factors involved in starch biosynthesis also have CBMs. Starch-branching enzymes, starch-debranching enzymes, and members of the PTST family have CBM48 domains. Starch granules are located in plastids, and therefore in order to target the CBM20 transgenes to these, we fused it with the transit peptide from GBSSI, which we have previously shown to be effective for protein translocation across the barley plastid envelope membrane (Hebelstrup et al., 2010, 2017). Western blotting of plants expressing either TP-eGFP (vector controls) or TP-CBM20-eGFP confirmed that the transgenic pre-proteins were correctly processed by cleaving off the 67-bp N-terminal TP (Fig. 1A). In agreement with this, confocal microscopy showed that both types of proteins were translocated across the chloroplast envelope in leaves and both were localized within the chloroplast (Fig. 2, Supplementary Fig. S2). In the vector controls, eGFP was localized throughout the chloroplast (Fig. 2D, F), whereas eGFP-CBM20 was located in 5–10 randomly distributed granules within each chloroplast (Fig. 2A, C). By tagging GBSSIa with eGFP, we have previously shown that chloroplasts contain 1–2 starch granules, which are placed roughly in the middle of each chloroplast (Hebelstrup et al., 2017; see Supplementary Fig. S3 for an example). Here, we also found one large starch granule in the middle of each chloroplast in the vector control plants (Fig. 3B). In contrast, plants transformed with eGFP-CBM20- had 5–10 smaller starch granules in each chloroplast (Fig. 3A).

Binding of CBM20 to starch granules down-regulates the Calvin–Benson cycle

Plants with expression of eGFP-CBM20 bound to starch granules in chloroplasts had a significantly reduced CO₂ fixation rate (Table 1, Fig. 7). They also had a slow growth rate, which was not observed in vector control plants with overexpression of eGFP at a level that was similar (or higher) than that of the eGFP-CBM20 expression lines (Fig. 1B). This showed that the effect on growth and photosynthesis was not due to expression of eGFP, nor to its presence in the chloroplasts or its transport across the chloroplast envelope. By staining in whole leaves, we determined that plants with CBM20 bound to granules contained less starch than vector control lines both at end of periods of light and darkness (Fig. 5). It is possible that starch biosynthesis was reduced in plants with expression of CBM20 because, bound to granules, it inhibits starch synthases and starch-modifying enzymes. Inhibition of starch biosynthesis causes accumulation of its precursors, namely sugars and ADP-glucose, which then inhibit photosynthesis via reduced photosystem efficiency and CO₂ fixation rate (Ragel et al., 2013). However, sugars did not accumulate in the plants with starch granules covered with eGFP-CBM20-,. On the contrary, a metabolomic analysis (Fig. 4) that covered measurements of the levels of a number of metabolites in the Calvin–Benson cycle, in carbohydrate biosynthesis, glycolysis, the citric-acid cycle, and amino acid biosynthesis suggested that the pools of metabolites related to starch were all lower in eGFP-CBM20 plants. Sugar-sensing in plants is based on receptors that measure different, but related, carbohydrates such as sucrose and trehalose-6-phosphate (T6P), which have correlated concentrations (Lastdrager et al., 2014; Ruan, 2014). If the effect of CBM20 on metabolism had been due to direct inhibition of starch biosynthesis, this would have resulted in an accumulation of the pool of biosynthesis precursors, such as sucrose and monosaccharides, but that was not observed in the CBM20-expressing plants. Therefore, we suggest that the reduced starch levels in these plants was not a result of an inhibition of biosynthesis per se, but instead the result of limited metabolites due to decreased photosynthesis. Sucrose availability is sensed through T6P, which acts as an indicator of the metabolic status of plant cells (Figueroa et al., 2016). Low T6P and sucrose levels up-regulate photosynthesis (Oszvald et al., 2018). However, in the CBM20 plants, CO₂ assimilation by photosynthesis as well as the level of 3-PGA, which is the direct product of CO₂ fixation by Rubisco, were significantly lower despite low levels of sucrose and many other central metabolites. This indicated that the presence of the CBM20 bound to the starch granules in the chloroplast had an inhibiting effect on photosynthesis. Photosynthesis can also be regulated at the level of photosystem II, but the α-values for both TP-eGFP-CBM20 and the WT indicated that, under low-light conditions, the rate of electron transport was not limiting for CO₂ fixation. This suggests that the inhibition may have been acting directly at the level of CO₂ fixation in the Calvin–Benson cycle. Since CBM20 binds to the starch granules, it does not interact directly with Rubisco or other enzymes of the Calvin–Benson cycle and therefore its effect on photosynthesis must involve co-factors that link its inhibitory effect with the cycle. An alternative explanation is that the binding of CBM20 competitively releases unknown starch-binding factors with an inhibitory effect on CO₂ assimilation. However, we cannot exclude the possibility that there were small amounts of CBM20 in the stroma that were below the detection limits of our methods, which these may have bound with soluble glucans or other carbohydrates, thus causing the observed metabolic effects.

These conclusions about the metabolic effects of CBM20 expression are summarized in Fig. 9. The non-spiking phenotype of the CBM20 plants may be an effect of the general low metabolite status of these plants, since a lack of sucrose can prevent flowering via a number of different signaling pathways (Oszvald et al., 2018). Reduced photosynthesis in the CBM20 plants also resulted in a decreased C:N ratio (Fig. 6). In line with this, we observed a higher concentration of several amino acids in the CBM20 plants, most notably the high N-containing glutamate (Fig. 4).

Fig. 9.

Summary of the metabolic effects of TP- eGFP-CBM20 expression in barley leaves. TP-eGFP-CBM20 is transported across the chloroplast envelopes and the transit peptide (TP) is cleaved off. Inside the chloroplasts, eGFP-CBM20 binds to starch granules. Either this binding or the release of other starch-binding factors by competition with CBM20 inhibits CO₂ fixation by the Calvin–Benson cycle (the light reactions of photosynthesis located in the thylakoid membranes are more likely affected indirectly). In effect, less carbohydrate (including starch) is synthesized and the plants grow slowly and do not flower.

Binding of CBM20 to starch granules compromises their development

Expression of CBM20 increased the number of starch granules in the barley chloroplasts from 1–2 centrally localized granules into 5–10 smaller granules that were randomly distributed (Fig. 2). Similarly, expression of CBM20 turned some of the single, oval granules in embryonic callus into double-compound granule types, and the endosperm A-type granules into multiple compound granule types (Fig. 8). Expression of tandem starch-binding domains in potato tubers has been shown to have a similar effect of turning potato starch granules that would otherwise usually be ovoid or round into compound granules of a smaller, polyhedral shape (Nazarian-Firouzabadi et al., 2012). This was interpreted as an effect of the tandem SBDs, and a structural model was proposed where these would link smaller, initial granules into the compound type. We observed a similar effect in barley (Fig. 8), but the CBM20 SBD was expressed as a single protein rather than in tandem. Therefore, the effect was not likely to that of linking smaller granules into large, compound ones by bringing them together, unless CBM20 acts similarly to a tandem dimer because it contains two glucan binding sites (Sorimachi et al., 1997). Transient expression of CBM20 with a fluorescent tag has shown that it binds to the surface of starch granules and is absent from the granule interior (B. Svensson, Technical University of Denmark, pers. comm.). In our transgenic barley leaves, the starch granules were too small to conclude whether the CBM20 was bound throughout the granules or solely on the surface (Figs 2, 3). However, the larger starch granules that developed in the embryonic callus did not have much CBM20 trapped within them, and it was bound at the surface with particular high concentrations at the interface between the sub-granules of a compound granule (Fig. 8C). We therefore suggest that the actions of CBM20 that we observed in relation to metabolism and starch granule development were caused simply by it binding at the surface of the granules. CBM20 acts competitively towards amylolytic enzymes whether these have a SBD or not (Southall et al., 1999). CBM20 expression in the barley chloroplasts and amyloplasts is therefore likely to have an effect on numerous starch-interacting factors in these compartments. The effect may be directly competitive, or there may also be an effect of CBM20 on structural disruption of the starch, which in turn may have an indirect effect on the interaction of the starch granule surface with the endogenous starch-interacting factors. Barley endosperm starch granules are not of the compound type, whereas rice endosperm granules are. Formation of compound starch granules in rice is coordinated by formation of membrane septa inside plastids via a mechanism that shares similarities with that of plastid division (Yun and Kawagoe, 2009, 2010). In particular, these septa seem to derive from the plastid inner-envelope membrane (Kawagoe, 2013). This process must require proteins that mediate binding between the starch granules and the membrane. These factors are currently unknown, but it has been suggested that the protein PDV2, which is located in the plastid outer envelope, is involved in this interaction (Yun and Kawagoe, 2010). Another study has shown that a thylakoid-associated protein is involved in starch-granule interactions (Seung et al., 2018). Such a model involving starch-bound factors and their binding to plastid membranes where starch granules are coordinated within plastids is in agreement with several of our observations, namely the loss of coordination of starch granule positioning within the chloroplast; their increase in number; and the change in the endosperm towards compound-type granules. All of these deficiencies are likely to be effects of competitive binding of CBM20 to the starch. This in turn affects the interaction between starch granules and any factors that may mediate interactions between starch granules and plastid membranes. Loss of membrane organization in maize has been shown to have a similar effect on endosperm granule morphology, with simple, single starch granules being turned into compound types (Myers et al., 2011).

Perspectives on starch bioengineering by SBD-containing enzymes

SBDs are found in several amylolytic enzymes (Christiansen et al., 2009) and the activity of these enzymes may be further engineered by adding additional starch-binding CBMs, or CBMs with high affinity for starch (Valk et al., 2016). Similarly, glycosyltransferases may gain additional activity when fused with a CBM20 (Jia et al., 2017). Such an approach may provide useful enzymatic modifications of starches on an industrial scale. CBMs may also be fused with other potential starch-modifying enzymes to enhance their modification efficiency (Hebelstrup et al., 2015). Alternatively, such engineered enzymes could be expressed by organ-specific promoters directly in the cells of the storage organs of the crops that produce the starch, such as potatoes or cereal grains. Starch modification could thus be conducted at the crop cultivation stage, thereby avoiding the need for later costly and environmentally damaging industrial modification (Hebelstrup et al., 2015). Our study explored the expression of a CBM with a strong starch affinity. Since our transgenic module did not contain any enzymatic activity, we hypothesized that its presence in the plant would have only neutral effects unless it was fused with other enzymatic domains. However, our results showed several, mostly pleiotropic, effects of the expression of the high-affinity domain. This demonstrates that transgenic starch-binding in itself is a means to modify starch biosynthesis and starch granule morphology, and it may have pleiotropic effects on plant physiology and development. These effects therefore need to be carefully controlled when designing strategies for starch modification in plants that involve factors with starch-binding capacities.

Supplementary data

Supplementary data are available at JXB online.

Fig. S1. Rubisco content in TP-eGFP-CBM20 and vector control plants.

Fig. S2. Magnification of Fig. 2C, showing confocal laser-scanning micrograph of chloroplasts in eGFP-CBM20 plants.

Fig. S3. Starch granules in chloroplasts.

Fig. S4. Proteins from leaf starch granules in TP-eGFP-CBM20 and vector control plants.

Fig. S5. Images of leaf stomata of TP-eGFP-CBM20 and vector control plants.