Further insight into the involvement of PII1 in starch granule initiation in Arabidopsis leaf chloroplasts

Further insight into the involvement of PII1 in starch granule initiation in Arabidopsis leaf chloroplasts


Introduction
Starch is a storage polymer that is synthesized in the plastids of photosynthetic organisms.It is composed of two structurally distinct a-linked glucans : amylose and amylopectin that are intricately combined in the form of dense, cold-water insoluble granules.Starch granule morphology and size are variable between species (Streb & Zeeman, 2012).In wild-type Arabidopsis leaves, starch granules, that accumulate between thylakoids, have a smooth, more, or less regular, lenticular form.The average number of granules per plastid also fluctuates.In photosynthetic plastids of a source organ, several granules are generally synthesized.For instance, Arabidopsis leaf chloroplasts generally contain 5-7 starch granules (Crumpton-Taylor et al., 2012).The number of granules in nonphotosynthetic plastids is lower, being, in some species, reduced to only one, like in potato tubers (Kram et al., 1993).
Currently, the process underlying the initiation of a new starch granule and the control of the number of granules per plastid has gained a lot of attention since it impacts the granule size and shape, parameters that strongly influence starch transformation and the quality of starch-based end products (Jobling, 2004;Singh et al., 2007;Guo et al., 2014;Tao et al., 2016).Over the past few years, numerous proteins have been described being involved in the priming and hence the control of starch granule formation.Most of these proteins are noncatalytic such as Protein Involved in starch Initiation 1 (PII1, also called MRC for Myosin-Resembling Chloroplast protein), Mar-binding Filament-like Protein 1 (MFP1), Protein Targeting to Starch 2 and 3 (PTST2 and PTST3) and Starch synthase 5 (SS5; Seung et al., 2017;Seung et al., 2018;Vandromme et al., 2019;Abt et al., 2020).Other proteins are catalytically active such as SS4.The inactivation of SS4 activity in Arabidopsis leaves leads to a grossly modified pattern of starch granule morphology (rounded rather than lenticular) and number (0-1 granule per chloroplast rather than 5-7) suggesting a preeminent function of this enzyme in the control of starch granule formation and initiation (Rold an et al., 2007;Crumpton-Taylor et al., 2013;B€ urgy et al., 2021).Interestingly, starch granule initiation is tightly controlled and occurs in specific regions of the chloroplast located between thylakoid membranes (B€ urgy et al., 2021).How this happens remains unclear but is likely related to the control of SS4 distribution within the plastid.This distribution could be driven by plastidial proteins directly or indirectly interacting with SS4.This is for instance the case of the thylakoid-localized fibrillins and the noncatalytic MFP1 protein that were both reported to physically interact with SS4 (G amez-Arjona et al., 2014;Seung et al., 2018).In addition, it was recently demonstrated that PTST2 and PTST3 are plastidial proteins that can bind glucans and that PTST2 interact with SS4 (Seung et al., 2017).It was then suggested that both PTST2 and PTST3 could provide the priming substrate required by SS4 to initiate the synthesis of starch granules in specific plastidial regions (Seung et al., 2017).
Starch synthase 4 is composed of two domains.The Cterminal moiety of the protein contains the catalytic Glycosyl transferase 5 (GT5) domain conserved among the starch synthases and was shown to control the starch granule number in Arabidopsis chloroplasts (Lu et al., 2018).The N-terminal moiety of SS4, which contains several coiled-coil regions, controls both the proper localization of the protein within the stroma and the shape of the granule (Raynaud et al., 2016;Lu et al., 2018).
In a previous study, using double-hybrid and BiFC methods, we identified PII1 as a protein that physically interacts with SS4.Protein Involved in starch Initiation 1 has no known activity, and no specific function can be determined through its primary structure.In pii1 KO mutants, we found a reduction in the starch granule number per chloroplast (one granule per plastid on average).However, contrary to what happens in ss4 mutants, these granules remain lenticular such as in the wild-type.Thus, we hypothesized that PII1 is required for the proper catalytic activity of SS4 but not for the control of the lenticular growth of starch granules by SS4 (Vandromme et al., 2019).
Moreover, other works revealed that in the absence of SS4, the synthesis of the residual starch granules in the Arabidopsis chloroplasts depends on the presence of SS3, another isoform of starch synthase.Indeed, starch content drops to almost zero in the ss3 ss4 double mutant (Szydlowski et al., 2009;Crumpton-Taylor et al., 2013).
From these results, several issues remain to be elucidated.The first was to determine which starch synthase (SS3 or SS4) is responsible for starch granule initiation in pii1.Indeed, it is still possible that SS4 activity is altered and not completely inactivated in pii1 leading to the formation of one granule per chloroplast.Alternatively, it is also possible that in pii1, the priming activity of SS4 is totally inactivated while that of SS3 is stimulated because of the lack of PII1 (the N-terminal part of SS4 being still able to drive the lenticular growth of the granule).Both hypotheses could explain why the phenotype of pii1 is less severe than that of ss4.If SS3 primes starch granules synthesis in the pii1 mutant background, the ss3 pii1 double mutant, such as the ss3 ss4 double mutant, should be starchfree.If not, ss3 pii1 should resemble the phenotype of pii1.The second issue was to determine whether PII1 is also involved in the mechanism of starch initiation fulfilled by SS3 in the absence of SS4.Thus, removing PII1 in addition to that of SS4 would give different phenotypes whether PII1 assists or not SS3 for its priming activity.To answer these two questions, we generated and analyzed both ss3 pii1 and ss4 pii1 mutants, respectively.The results presented here show that the ss3 pii1 mutant is not free of starch but, on the contrary, contains granules with distorted shapes, while the ss4 pii1 mutant accumulates less starch than the ss4 single mutant.The results are discussed and provide further insight into the function of PII1 in the control of starch granule synthesis.

Plant material and growth conditions
Arabidopsis thaliana Wassilewskija (Ws) and Columbia (Col-0) lines can be ordered from SiGnAL (Alonso et al., 2003).All mutant alleles used in this study have been described previously (Zhang et al., 2005;Rold an et al., 2007;Szydlowski et al., 2009Szydlowski et al., , 2011;;Vandromme et al., 2019).Germplasm references and position of the different mutations are described in Fig. 1.Crosses and selection of the double mutant lines were performed from plants grown in the glasshouse.Regulation system was set to obtain 16 h : 8 h, light (150 lmol photon m À2 s À1 ) : dark photoperiod with c. 23°C : 20°C, day : night.For plant growth analysis or starch extraction, plants were grown in a growth chamber with 16 h : 8 h, light (120 lmol photon m À2 s À1 ) : dark photoperiod, 23°C : 20°C.In all cases, 3 d before being sown, seeds were incubated at 4°C in 0.1% phytagel solution.

Chloroplast observation and determination of starch priming efficiency
Leaves of 2-wk-old plants were harvested at the end of the light phase.Sections of proximal or distal part of the leaves were transferred in 1 ml of fixating solution (4% (w/v) paraformaldehyde, 4% (w/v) sucrose, and 19 PBS at pH 7.3).Samples were placed under vacuum during 6 h and stored at 4°C.Leaf samples were placed between microscope slide and coverslip and observed under A1 Nikon confocal microscope (Nikon Instruments Europe B.V., Amstelveen, the Netherlands) with a Plan Apo 960 Oil (NA = 1.4) objective.Autofluorescence was acquired with k ex = 488 nm-k em = 500-550 nm and k ex = 561 nm-k em = 580-620 nm.The observation and determination of starch granule number per chloroplast were carried out through live Z scanning of the whole chloroplasts to avoid bias due to out-of-focus starch granules.Granules were counted during the acquisition by two researchers, and a representative image was acquired.After acquisition, we used FIJI (Schindelin et al., 2012) to apply: (1) a median filter with a 'radius' of 0.5; (2) a high accuracy Contrast Limited Adaptive Histogram Equalization (CLAHE (Zuiderveld, 1994)) with a maximum slope of 2.5; and (3) a brightness equalization between images.

Starch extraction, purification, and quantification
The leaves of 3-wk-old plants were harvested at the end of the light or dark phases and immediately frozen in liquid nitrogen.Depending on the purpose, quantification, or purification, two different protocols were applied.
For starch quantification, leaf samples were weighted and 0.3-0.5 g were disrupted using a polytron blender in 5 ml of 1 M perchloric acid.Samples were centrifuged at 4500 g for 10 min at 4°C.The pellet containing starch was rinsed three times with sterile deionized water and stored at 4°C in 1 ml of 20% ethanol.
The starch amount was determined using Enzytec TM starch kit.30 ll of different sample dilutions was boiled during 10 min; New Phytologist (2023) www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation 60 ll of amyloglucosidase (14 U ml À1 ) was added.The medium was incubated for 45 min at 58°C and centrifuged for 10 min at 10 000 g and 4°C.30 ll of digested starch was transferred in a 96-well plate.The following steps were carried out according to the manufacturer's instruction with the provided buffer and enzymes.100 ll of buffer 1 and 100 ll of deionized water were added.The initial absorbance was recorded at 340 nm. 2 ll of enzyme mix were used and the final absorbance at 340 nm was recorded to complete the assay.Four independent cultures were grown, and for each culture, three biological samples were harvested per genotype.We then normalized the values by calculating, for each culture, the average obtained for the wild-type reference.We established this value as 100% and expressed the amount of starch accumulated by the different genotype as a percentage of the wild-type amount.We then calculated a mean and a standard error from the 12 values obtained for each of the genotypes.A two-tailed t-test with unequal variance was applied using an EXCEL function.Statistically meaningful differences were indicated by either one asterisk (P < 0.05) or two asterisks (P < 0.001).
A different protocol was used when purified starch was needed for microscopy analysis, starch fractionation, or determination of chain length distribution (CLD).Several grams of leaves were disrupted using a Polytron blender in a buffer composed of 100 mM MOPS, pH 7.2, 5 mM EDTA, and 10% (v/v) ethylene glycol.The homogenate was filtered through two layers of Miracloth (Calbiochem; EMD Millipore Corp., Billerica, MA, USA).The filtrate was centrifuged at 4000 g during 10 min at 4°C.The supernatant was discarded, and the pellet was resuspended in 2 ml of 90% (v/v) Percoll.The sample was centrifuged for 45 min at 10 000 g and 4°C.The starch pellet was washed several times with 2 ml of sterile deionized water and stored in 1 ml of 20% ethanol.

Scanning electron microscopy and starch granule size distribution
The procedure was identical to the one described in Vandromme et al. (2019).Aqueous suspensions of purified starch granules were dropped on a piece of glow-discharged copper tape.When dried, samples were coated with Au/Pd and secondary electron images were recorded with a FEI Quanta 250 scanning electron microscope equipped with a field emission gun and operating at 2 kV.The starch granule size was determined from the scanning electron microscopy (SEM) images using FIJI software.The major axis of 300 granules was measured, and the average value was calculated.

Starch fractionation and determination of the amylose content
To evaluate the proportion of amylose in starch granules, two different methods were applied.First, we fractionated starch by size exclusion chromatography to separate amylopectin and amylose.1.8 mg of starch was dispersed by boiling for 10 min in 200 ll of 100% dimethylsulfoxide (DMSO).The glucans were then precipitated by the addition of 800 ll of 100% ethanol and stored during several hours at À20°C.The samples were centrifuged at 10 000 g during 10 min at 4°C.The pellet was air-dried and solubilized in 300 ll of 10 mM NaOH.The sample was loaded at the top of a column (0.5 i.d. 9 65 cm) filled with Sepharose CL-2B matrix and equilibrated with 10 mM NaOH.The flow rate was set at 12 ml h À1 .Fractions of 300 ll were collected.80 ll of each fraction was mixed with 20 ll of a solution made of I 2 (0.1% w/v) and KI (1% w/v).The presence of glucans was detected by spectrophotometry analysis.Fractions containing amylopectin or amylose were pooled and the amount of polysaccharides was determined by enzymatic assay (as previously described).
The second method that was applied is described in Helle et al. (2019) and used the Megazyme (Wicklow, Ireland) kit amylose/amylopectin assay (K-Amyl).1.5 mg of starch was gelatinized in 200 ll of 100% DMSO at 100°C during 15 min.Polysaccharides were precipitated by the addition of 800 ll of 100% ethanol.The sample was centrifuged for 5 min at 2000 g.The pellet was suspended in 120 ll of 100% DMSO and heated at 100°C during 15 min.Amylopectin was precipitated following the manufacturer's instructions using concanavalin A (ConA).The sample containing the polysaccharides was gently mixed with 1.5 ml of ConA solvent (this 1.5 ml containing amylose and amylopectin is designed as Solution A). 300 ll of this solution was mixed with 150 ll of ConA in order to precipitate the amylopectin fraction during 1 h of incubation at room temperature.A 10 min centrifugation at 14 000 g allowed the separation of amylose (in the supernatant) and amylopectin (in the pellet).300 ll of the supernatant was mixed with 300 ll of 100 mM sodium acetate buffer (pH 4.5) and heated at 100°C during 5 min.This allowed ConA to be denatured.Then, 800 ll of sodium acetate buffer was added to 100 ll of solution A. 20 ll of the amylose fraction and 20 ll of diluted solution A were digested with 20 ll of solution 2 (provided by manufacturer and containing amyloglucosidase and a-amylase) during 30 min at 40°C.40 ll of the digested samples was transferred into 96-well plates, and 160 ll of GOPOD reagent was added to each well.After 20 min of incubation at 40°C, the absorbance was measured at 510 nm using a spectrophotometer.The amylose percentage was calculated according to the manufacturer's instructions.

Polysaccharide chain length distribution
The amylopectin CLD was determined using fluorescenceassisted capillary electrophoresis as described in Helle et al. (2019).1.8 mg of starch was boiled in 250 ll of sterile ultrapure water during 10 min.250 ll of sodium acetate 110 mM pH 3.5; 20 units of isoamylase (Pseudomonas sp.from Megazyme); and 1 unit of pullulanase (from Sigma-Aldrich E2412) were added.Samples were incubated overnight at 42°C.Proteins were eliminated by boiling the samples during 10 min before centrifugation (15 000 g for 10 min).The glucans-containing supernatants were desalted using a Carbograph column (Grace TM Alltech TM , Extract-Clean TM ) and lyophilized.The glucans were solubilized in 2 ll of sodium cyanoborohydride in tetrahydrofuran and 2 ll of 0.2 M 8-aminopyrene-1,3,6-trisulfonic acid trisodium salt (APTS) in 15% acetic acid (v/v).The derivatization reaction occurred during overnight incubation at 42°C.46 ll of ultrapure water was added at the end of the reaction.1 ll of the dilution was mixed with 49 ll of water before injection in a Beckman Coulter PA800-plus Pharmaceutical Analysis System (Beckman Coulter, Brea, CA, USA).A silicon capillary column (inner diameter: 50 lm, outer diameter: 360 lm, and length: 60 cm) was filled with Beckman Coulter N-linked carbohydrate separation gel buffer diluted to the third with ultrapure water.The sample was injected during 15 s at 10 kV.Migration was performed at 10 kV during 1 h at 20°C.A 488 nm solid-state laser module and a laser-induced fluorescence detector were used to detect the glucans labeled with APTS.

X-ray diffraction
The sample preparation protocol was adapted to take into account the very small amount of starch granules available.To prevent any loss of material, the specimens were manipulated in the form of aqueous suspensions that were first concentrated by centrifugation.A small volume of the wet pellets (c. 4 ll) was pipetted and poured into 0.7-mm (outer diameter) glass capillaries.The capillaries were centrifuged again to concentrate the granules at the bottom, then flame-sealed.The pellets (in excess water) were X-rayed in transmission in a vacuum chamber using a Philips PW3830 generator operating at 30 kV and 20 mA (Nifiltered CuKa radiation, k = 0.1542 nm).Two-dimensional Xray diffraction (XRD) patterns were recorded on Fujifilm imaging plates read with a Fujifilm BAS 1800-II bioimaging analyzer.Diffraction profiles were calculated as radial averages of the 2D patterns.

Additional materials and methods
Supporting Information Methods S1 describes the methods used for the selection of ss3pii1 and ss4pii1 lines, for the SDSacrylamide gel electrophoresis and measurement of ADP-glucose concentration.

Plant growth phenotype
We used already characterized ss3, ss4, and pii1 mutants to combine mutations in either Ws or Col-0 ecotypes (Figs 1, S1; Table S1).Since the pii1-1 mutation isolated in a Col-0 is leaky (Seung et al., 2018), all the experiments described in this work were carried out in Ws genetic background.However, the main results obtained in the Ws ecotype were confirmed in the Col-0 background and are presented in Figs S2-S6.The plants were grown under 16 h : 8 h, light : dark cycles.The aboveground organs were collected at different time points and weighed.As already reported under these conditions, ss3 and pii1 single mutant behave like the wild-type (Zhang et al., 2005(Zhang et al., , 2008; Vandromme et al., 2019), while the ss4 single mutant displayed a severe growth retardation (Rold an et al., 2007).Combining ss3 and pii1 mutations led to reduced growth kinetics compared with the respective single mutants (Fig. 2a).Moreover, this double mutant displayed a slight delay (of c. stems with flowers while ss4, ss3 pii1, and ss4 pii1 plants have not (Fig. 3).Note that the growth kinetics of the ss4 pii1 were similar to that of the ss4 single mutant and that all plants bearing ss4 mutation also have the pale green leaf phenotype described for this mutation (Rold an et al., 2007; Fig. 2b).

Starch distribution in leaves
Leaf starch distribution was investigated by iodine-staining of 3wk-old plants collected at the end of the light phase.In accordance with what has already been described, starch is homogeneously distributed in the leaves of wild-type, ss3, and pii1 plants, while it is heterogeneously distributed in ss4 (Fig. 3) being mostly present in the apex of the leaves.We show here that leaf starch distribution in the ss3 pii1 line is similar to that of the wild-type, ss3, and pii1 lines, while that of the ss4 pii1 double mutant resembles that of the ss4 plants.
The number of starch granules in plastids was determined from confocal microscopy images of all lines analyzed in this work (Figs 4, S7; Table S2).Both proximal (close to the stem) and distal (apart of the stem) regions of the leaves were investigated and, depending on the genotypes, 90-250 chloroplasts were analyzed.As expected, most chloroplasts from wild-type and ss3 plants contain 5-7 starch granules (rarely < 5 and > 6) with an average number between 5.32 (AE0.08) and 5.72 (AE0.11)whatever the considered leaf region (proximal or distal).In the pii1 and ss3 pii1 mutants, only one large starch granule was generally observed in the chloroplasts of both proximal and distal regions, with the average number of granules varying between 1.05 (AE0.04) and 1.3 (AE0.05).The ss4 pii1 line behaves similarly  to the ss4 mutant although its phenotype is exacerbated compared with the single mutant.Most plastids of the proximal region are starch-free in both ss4 and ss4 pii1, the average number of granules being 0.55 (AE0.06) in the ss4 mutant and 0.22 (AE0.03) in the ss4 pii1 double mutant.In both ss4 and ss4 pii1 genotypes, the number of granules per plastid is increased in the distal region of the leaf in comparison with the proximal region (1.03 AE 0.05 and 0.47 AE 0.03 respectively).Regardless of the region observed, there is, in average less starch granules per chloroplast in ss4 pii1 than in ss4 (Table S2).The same approach was conducted for ss4 and ss4 pii1 mutants obtained in the Col-O genetic background and similar results were obtained (Fig. S4).

Shape and size of starch granules
The confocal microscopy observations described above suggested that the starch granules of ss3 pii1 and ss4 pii1 are modified both in size and shape compared with the wild-type.Indeed, the counting of the granules in the ss3 pii1 double mutant was particularly challenging, even by varying the focus along the z-axis, because of the apparent grossly modified shape.Therefore, the starch granules were extracted and purified from these lines and observed by SEM (Fig. 5a).The ss4 pii1 starch granules are spheroidal and very large compared with the smaller lenticular wild-type granules.This phenotype is quite similar to that already described for the ss4 single mutants.
In the ss3 pii1 mutant, starch granules are larger than those of the wild-type.Instead of the lenticular form (thin and round) usually observed for wild-type granules, the ss3 pii1 granules have a considerably modified shape.Although they remain thin like the wild-type granules, they are much larger, like those of the pii1 single mutant, but contrary to these latter, most of them do not remain flat.In the most extreme cases, the granules are significantly buckled (Fig. 5a,b for an enlarged view) giving them an irregular shape.This buckled phenotype is specific to the ss3 pii1 mutant and was not described in the corresponding single mutants (Szydlowski et al., 2011;Vandromme et al., 2019).
The average size of the starch granules was estimated from the SEM images using the FIJI software.Three hundred starch granules were analyzed for each genotype and the major axis of each granule was measured (Fig. 5c).The average size of the major axis was 2.4 lm (AE0.8) for the wild-type, while it reached 5.5 lm (AE1.7) for ss3 pii1 and 4.1 lm (AE1.0) for ss4 pii1.Due to the distorted shape of the ss3 pii1 starch granules, the identification and measurement of the major axis were difficult.Therefore, the value should be taken as indicative, confirming that this double mutant and the ss4 pii1 double mutant accumulate starch granules which size is approximately twice larger than that of the wild-type ones.

Leaf starch content and ultrastructure
We determined the leaf starch content in the different genotypes at the end of the light and the dark periods.Four independent cultures were prepared, and for each of them, three independent biological samples were collected.Within an independent culture, all genotypes were cultivated in the same growth chamber under the same conditions and with the same treatment.For each culture, the leaf starch content of the different genotypes was normalized to that of the corresponding wild-type, the average of the three wild-type biological samples being set to 100%.Finally, the mean and the standard error were calculated for each genotype from 12 samples (four independent cultures 9 three biological samples; Figs 6, S8).
ss3 behaves as the wild-type since no significant difference in leaf starch content at the end of the day and at the end of the dark period was noticed.The ss4 mutant accumulated significantly less starch at the end of the day compared with the wild-type but more starch at the end of the night.The pii1 mutant contained more starch than the wild-type at the end of both periods.No statistically significant difference was detected when ss3 pii1 was compared with the wild-type or the ss3 single mutant.The ss4 pii1 double mutant accumulates significantly less starch than the wild-type and the ss4 lines at the end of the day (Fig. 6a).However, at the end of the night, the ss4 pii1 contained in average more starch than the wild-type while the difference between the ss4 and the ss4 pii1 was not considered as statistically significant (Fig. 6b).
The amylose content was determined for each genotype after the fractionation of 1.8 mg of purified starch by size exclusion chromatography on Sepharose CL-2B resin.The presence of glucans in each collected fraction was assessed by iodinespectrophotometry analysis (Fig. 7a).This analysis suggests that ss3 pii1 starch contains more amylose than the wild-type.Amylopectin-and amylose-containing fractions were pooled independently and the amount of each glucan was determined by enzymatic assay.The starch of the ss3 pii1 mutant contains 22% amylose while that of the wild-type contains only 15% amylose (Fig. 7b).The amylose content was also estimated using another method that requires the specific precipitation of amylopectin with Concanavalin A. We applied this method on three starch samples prepared from three independent cultures.Here again, the amylose content was increased in the ss3 pii1 compared with the wild-type being 19.4% AE 1.08 and 6.51% AE 0.64, respectively (Fig. 7c).This increase in amylose content is not associated with a higher content of GBSS1 in starch nor to an increase in ADP-glucose concentration, as these parameters were similar in both ss3 pii1 and the wild-type lines (Figs S9, S10).
The starch ultrastructure was also characterized by the determination of the amylopectin CLD.1.8 mg of starch was enzymatically debranched using a mix of isoamylase and pullulanase.The linear glucans were then derivatized by APTS and separated by capillary electrophoresis.After migration, the detection of fluorescence allowed the quantitative determination of the relative proportion of each glucan length.Overall, the profiles of the wild-type and mutant starches were similar (Fig. 8).However, we observed a slight increase in DP 5-10 glucans in the ss3 mutant, associated with a slight reduction in DP 13-20 glucans, which is in agreement to previously reported results (Zhang et al., 2008).
The profile of the ss3 pii1 was similar to that of the ss3, indicating that PII1 has no function in amylopectin chain elongation.Similarly, the profile of the ss4 pii1 mutant resembles that of the ss4 mutant, which is further evidence that PII1 is not involved in determining the final structure of amylopectin.
The crystalline organization of amylopectin molecules was analyzed by XRD.The starch granules from Ws and the double mutants yield similar intensity profiles (Fig. 9).By comparison with the profile of standard potato starch, all specimens correspond to the B-type allomorph, with main characteristic diffraction peaks located at 2h = 5.6, 17.1, 22.2, and 24.1°.In particular, the peak at 5.6°is typical of B-type.It has the highest intensity since the starch granule pellets were X-rayed in excess water, hence in a fully hydrated state.The profile of the ss4 pii1 starch granules is nearly identical to that of the wild-type, while the peaks of the ss3 pii1 specimens are somewhat less defined and with a lower intensity, suggesting a slightly lower degree of crystallinity, in line with the higher amylose content of the starch granules.Similar results were observed in the Col-0 genetic background (Fig. S6).

Discussion
Several proteins have already been described as being involved in the control of starch granule initiation in Arabidopsis leaves.However, SS4 is to date the only one with an identified enzymatic activity.ss4 KO mutants synthesize less than one starch granule per plastid, whereas 5-7 are typically observed in wildtype individuals (Rold an et al., 2007).Moreover, SS4 was shown to control the anisotropic growth of Arabidopsis starch granules by canalizing the deposition of newly synthesized glucans on the equatorial regions, thus explaining their flat and lenticular shape in the wild-type and their spherical shape in the ss4 mutant (B€ urgy et al., 2021).The analysis of the ss3 ss4 double KO mutant, which is nearly starch-free, highlighted the partial redundancy between SS3 and SS4 in starch initiation suggesting that SS3 is responsible for the priming of the residual starch granules in the ss4 mutant (Szydlowski et al., 2009).No other enzyme seems capable to compensate for the loss of both SS3 and SS4 to maintain a correct initiation of starch synthesis in Arabidopsis leaves, the very few residual granules being likely the result of stochastic initiation events.
More recently, the PII1 protein was described as mandatory for the correct activity of SS4 (Vandromme et al., 2019).Indeed, PII1 physically interacts with SS4 and in the absence of the former, the number of starch granules produced per plastid significantly drops compared with the wild-type, in a way that resembles that of the ss4 KO mutant.Therefore, the objective of the present work was to analyze Arabidopsis lines combining mutations of PII1 with those of SS3 or SS4 to better understand the role of these three proteins in the initiation of starch synthesis.
SS4 consistently initiates one starch granule per plastid in the absence of PII1 whatever the status of SS3 As mentioned above, the pii1 mutant has a reduced number of starch granules in chloroplasts.In most cases, only one starch

Research
New Phytologist granule can be found in the plastid.However, these granules are larger than those of the wild-type but still have the same flatlenticular shape.This phenotype suggests that PII1 has a function in controlling the number of starch granules that is distinct from that of SS4.The fact that the shape of the pii1 mutant granules is identical to that of the wild-type (i.e.lenticular and not round like in the ss4 mutant) implies that the loss of PII1 does not alter the anisotropic granule growth driven by SS4.
The phenotype of the pii1 mutant can be explained in two ways.The first explanation is that SS4 is still responsible for the synthesis of, on average, one starch granule per plastid in the absence of PII1, but cannot support the production of the 5-7 starch granules per chloroplast seen in wild-type plants.The second explanation is that, in the absence of PII1, SS4 is no longer able to initiate the starch granule synthesis, this function being then ensured by SS3 only (through a similar or different mechanism and/or partners than that driven by SS4).For that second explanation, one should assume that the N-terminal part of SS4 is still able to control the anisotropic growth of the granules, since these remain flat and lenticular in the pii1 mutant background.
The analysis of the ss3 pii1 double mutant allowed testing these two hypotheses.The results revealed that ss3 pii1 still produces nearly wild-type amount of starch in the form of one large starch granule per plastid.Since SS3 cannot be responsible for starch initiation in this mutant, this function must still be provided by SS4 in this line.The formation of at least one starch granule per chloroplast, driven by SS4, is very robust since there are almost no starch-free chloroplasts in pii1 or ss3 pii1 mutants.
The reason why ss3 pii1 starch is enriched in amylose is unclear.This could be caused by an increased amount of GBSS responsible for amylose synthesis and/or an increased concentration of ADP-glucose that would stimulate GBSS activity.However, this does not seem to be the case since these parameters are similar in both ss3 pii1 and wild-type lines (Figs S9, S10).We could only speculate that the increase in the amylose content results from the ability of GBSS to partially compensate for the lack of SS3.It has been demonstrated that SS3 synthesizes the intermediate to long chains of amylopectin (Fontaine et al., 1993;Ral et al., 2006;Fujita et al., 2007;Bor en et al., 2008).This capacity can be to a certain extent fulfilled by GBSS1 (Ral et al., 2006), leading to a moderate impact of the SS3 deficiency on amylopectin CLD.Moreover, studies conducted in Chlamydomonas reinhardtii revealed that GBSS produces amylose by the extension of and the subsequent cleavage from amylopectin (van de Wal et al., 1998).Consequently, in the absence of SS3, GBSS activity could be partially diverted to synthesizes the very long chains of amylopectin and hence could promote the production of amylose or amylose-like polymers.In fact, the lack of SS3 often induces higher amylose starch as well as modified amylopectin structure and granule shape (Inouchi et al., 1983;Fontaine et al., 1993;Maddelein et al., 1994;Abel et al., 1996;Fujita et al., 2007;Bor en et al., 2008).Amylose content is also slightly increased in the Arabidopsis ss3 mutant (Fig. 7) and (Szydlowski et al., 2011).This effect is exacerbated in ss3 pii1 probably because of the increased size of the starch granule in that mutant.Indeed, this can amplify the anomaly of the granule shape formation engendered by higher amylose content and altered amylopectin structure.It is therefore possible that the combination of ss3 with another mutation that greatly reduces the starch granule number per plastid (mfp1, ptst2, or ss5 for instance) would lead to the same effect.

SS3 can initiate starch synthesis, a process facilitated by PII1
It was demonstrated that in the absence of SS4, SS3 can prime the synthesis of starch in Arabidopsis leaves (Szydlowski et al., 2009).However, this ability is limited to one starch granule per plastid and not all plastids contain a starch granule.The chloroplasts of cells in the distal region of the leaf (away from the stem) are more likely to produce a starch granule than those in the proximal region (close to the stem) (Vandromme et al., 2019).We investigated whether the SS3-priming aptitude is, mechanically speaking, similar to that of SS4 or if different.To this end and to determine whether SS3-priming ability is affected by PII1, we combined KO mutations for both SS4 and PII1 and we analyzed the phenotype of this double mutant, notably the number of granules per plastid in both the distal and proximal regions of the leaf.We found that the ss4 pii1 double mutant initiates half as many granules as the single ss4 mutant whatever the region of the leaf is considered (Fig. 4).Identical result was obtained in both Ws and Col-O genetic backgrounds (Fig. S9).In addition, we also found that ss4 pii1 accumulates significantly less starch at the end of the day than ss4.The fact that the lack of PII1 further reduces the number of granules in an ss4 mutant background suggests that PII1 facilitates the action of SS3 to sustain the synthesis of the starch granule when SS4 is inactive.It is not known whether this functional interaction arises from a physical interaction between SS3 and PII1 but it has no impact on the interpretation of the results.The study of the physical interaction between SS3 and PII1 as well as of SS3 with the other proteins involved in the process of starch granule initiation (MFP1, SS5, and PTST2) can be the subject of a dedicated complementary work.
The fact that the mechanism of starch granule initiation is similar for both SS3 and SS4 and that it is facilitated by PII1 is not completely surprising if one considers that both starch synthases share common features facilitating interactions with the same protein partners.Indeed, contrary to other canonical starch synthases, SS3 and SS4 share a common chlamydial phylogenetic origin (Ball et al., 2013), and are the only ones that possess two distinct domains of equivalent size with a catalytic GT5-like Cterminal part, and an N-terminal part comprising coiled-coil domains involved in protein-protein interactions (Abt et al., 2020).This functional overlap is however limited, and in the absence of SS4, a single starch granule is synthesized per chloroplast with a nonnegligible part of starch-free chloroplasts.This overlap of SS3 and SS4 function is not restricted to Arabidopsis since in rice the combined knockdown of SS3a and SS4b leads to a strong alteration of the shape of the starch granules while the corresponding single mutants display wild-type phenotypes (Toyosawa et al., 2016).Arabidopsis genome also codes for a fifth starch synthase (SS5) that can physically interact with PII1 (Abt et al., 2020).Although this protein is closely related to SS4, it is unlikely that it can compensate for the loss of SS4 since it lacks the GT1 domain and is supposed to be catalytically inactive.
Interestingly, the impact of the ss4 mutation is not homogeneously distributed along the Arabidopsis leaf (Vandromme et al., 2019).Indeed, the average number of starch granules per chloroplast is divided by two in the proximal part of the leaf (0.55 AE 0.06) compared with the distal region (1.03 AE 0.05).We observed the same phenomenon with the ss4 pii1 double mutant in which the number of starch granules per chloroplast is also divided by approximately two in the proximal region (0.22 AE 0.03) compared with the distal region (0.47 AE 0.03).Since cells and chloroplasts division is greatly reduced in the distal region of the leaf (Gonzalez et al., 2012), proximal and distal regions could be regarded as immature (high cell and plastid division activity) and mature (low cell and plastid division activity) leaves, respectively.Immature leaves of the ss4 mutant were described as low starch-containing leaves suggesting that starch granule formation is controlled by the status of leaf expansion in a way that is still not understood (Crumpton-Taylor et al., 2013).Our results suggest that PII1 is likely not involved in that control even in the absence of SS4 because the removal of PII1 activity in the ss4 mutant background does not modify the difference in the chloroplast granule content between distal and proximal regions.
During this study, and by creating and analyzing combinations of mutations, we have demonstrated that, in the absence of PII1, SS4 indeed initiates, almost systematically, the synthesis of one starch granule per chloroplast a number that is similar in pii1 and ss3 pii1 (Figs 4,10).The number of starch granules per plastid is also similar in the wild-type and in ss3 indicating that when SS4 is present, SS3 does not contribute, or has a minor effect, on starch granule initiation (only a slight difference can be seen in the distal part of ss3 leaves compared with wild-type).In addition, SS3 can initiate the synthesis of new starch granules, but only if SS4 is absent (i.e. in ss4 and ss4 pii1) and preferentially in cells where the chloroplasts no longer divide (distal part of leaves).Since the phenotype of ss4 pii1 is more pronounced than that of ss4, we could conclude that this mechanism of starch initiation by SS3 is facilitated by PII1 (Fig. 10).
The fact that pii1 and ss3 pii1 mutants accumulate one large starch granule per chloroplast and that there are apparently no starch-free chloroplasts in these plants, can be explained in two different ways.First, in the absence of PII1, SS4 can only initiate the synthesis of one starch granule per plastid.It is not clear yet how PII1 could control SS4 activity.The region of SS4 involved in the interaction with PII1 is not clearly defined, but it is likely that, in this hypothesis, PII1 modulates the C-terminal catalytic core region of SS4, which is known to control the initiation of starch.Second, we can speculate that in the absence of PII1, the initiation of starch granules still occurs as in the wild-type plant, but then most, if not all, starch granules merge at the very early stage of their development to finally form a unique large granule in the plastid.This hypothesis is reinforced by the discovery that some starch granule initials coalesce during their growth in Arabidopsis leaves (B€ urgy et al., 2021).Thus, in the absence of PII1, coalescence of granule initials would occur at a larger scale compared with the wild-type, leading to the production of a reduced number of larger granules without modification of their morphology since SS4 could still control their anisotropic growth.However, further investigations will be required to discriminate between these two hypotheses and to precisely determine the role of PII1 in the determination of starch granule number per plastid.

Fig. 1
Fig. 1 Identification of the different alleles.(a) Structure of Arabidopsis Starch synthase 3 (SS3), SS4, and Protein Involved in starch Initiation 1 (PII1) genes.3 0 and 5 0 UTR are indicated by gray boxes while introns and exons are represented by black lines and boxes respectively.The locations of T-DNA insertions are indicated by black triangles (insertion in Col-O genetic background) or gray triangles (insertion in Ws genetic background).(b) Summary of the different alleles used in this work, germplasm identification, genetic background, and references where the mutant lines are described.

Fig. 2
Fig. 2 Plant growth phenotypes.Arabidopsis plants were grown in a growth chamber under 16 h : 8 h, light : dark cycles.(a) The weight of aboveground organs is plotted against the number of days after germination.Two independent experiments were performed.For each experiment, all six genotypes are grown in the same chamber.Each value corresponds to the mean of three samplings (AESD).Wild-type (Ws) is depicted in blue lines, ss3 in orange; ss4 in gray; pii1 in yellow; ss3 pii1 in black; and ss4 pii1 in green.Note that, on these graphs, ss3 (orange) and wild-type (blue) profiles overlap and cannot be differentiated.ss4 (in gray) and ss4 pii1 (in green) are also almost identical.(b) Pictures of 3-wk-old plants.All pictures are at the same scale.Bar, 5 cm.

Fig. 3
Fig. 3 Starch distribution in leaves.Arabidopsis plants were collected at the end of the light phase, 24 d after germination.Pigments were removed by incubation of plants in 80% ethanol at 80°C for 10 min.Plants were stained with iodine.Bar, 5 cm.

Fig. 4
Fig. 4 Starch granule number in plastids.Leaves of 2-wk-old Arabidopsis plants were collected at the end of the day.Pieces of proximal (orange frame) and distal part (blue frame) of leaves were fixed in paraformaldehyde and analyzed by confocal microscopy.At the bottom of each picture, the average number of starch granules per chloroplast is indicated (AESE).Number of chloroplasts analyzed is indicated in brackets.Chlorophyll fluorescence in purple delimits the chloroplast volume; starch granules are black.Bar, 10 lm.

Fig. 5
Fig. 5 Starch granule size and morphology.Starch granules were extracted at the end of the illuminated period from 3-wk-old Arabidopsis plants cultivated in 16 h : 8 h, light : dark photoperiods.(a) The purified starch granules were observed by scanning electron microscopy (SEM), and their major axis was measured using FIJI software.The average values (AESD), calculated from the measurement of 300 granules, are indicated.Bars, 10 lm.(b) Enlargement of ss3pii1 starch granules selected from the image in (a), with marked alteration in their shapes.Bar, 5 lm.(c) For each genotype, the major axis of 300 starch granules was measured from SEM images using FIJI software.The solid line in the box corresponds to the median.The bottom and top of the boxes correspond respectively to the 25 th and 75 th percentiles.The dotted lines delimit the interval between the 5 th and the 95 th percentiles.The average value is indicated by a black dot.

Fig. 6
Fig.6Starch content.Arabidopsis plants were grown for 3 wk in a growth chamber under 16 h : 8 h, light : dark cycles.Leaves were collected at the end of light (a) or dark phase (b).Four independent cultures were made and each time, three biological samples were harvested.For each culture, the average of wild-type amount was calculated and set at 100%.We then calculate the percentage of wild-type starch amount for the three biological samples of each genotype within the culture.Mean and SE are calculated from the 12 biological samples.Thin vertical bars represent SE.Value obtained for mutant line was compared with the wild-type (and between ss4 and ss4 pii1) using a two-tailed t-test (unequal variance).Asterisks indicate statistically significant differences: *, P < 0.05; **, P < 0.001.

Fig. 8
Fig.8Amylopectin chain length distribution.An amount of starch (1.8 mg), extracted from Arabidopsis leaves, was debranched using isoamylase and pullulanase.Linear glucans are derivatized with APTS and separated using capillary electrophoresis.Peaks corresponding to DP 4-80 were integrated.Bars represent the mean of relative percentage of each DP obtained from three independent analyses with starch prepared from three independent cultures.Thin black lines correspond to the differential plot between mutant and wild-type profiles.

Fig. 9
Fig.9X-ray diffraction profiles recorded from starch granule pellets in excess water.The contributions of the glass container and the excess water (amount not known with precision) to the scattering have not been subtracted.The profiles recorded from the Arabidopsis Ws specimens have been compared with that of standard potato starch granules prepared in the same conditions.All profiles correspond to a pure B-type allomorph.The intensity has been normalized by calculating the area under each profile between 4 and 28°.au, arbitrary units.