Two Plastidial Coiled-Coil Proteins Are Essential for Normal Starch Granule Initiation in Arabidopsis

MFP1 and MRC are two plastidial coiled-coil proteins that play an important role in initiating starch granules in Arabidopsis chloroplasts.

Abstract

The mechanism of starch granule initiation in chloroplasts is not fully understood. Here, we aimed to build on our recent discovery that PROTEIN TARGETING TO STARCH (PTST) family members, PTST2 and PTST3, are key players in starch granule initiation, by identifying and characterizing additional proteins involved in the process in Arabidopsis thaliana chloroplasts. Using immunoprecipitation and mass spectrometry, we demonstrate that PTST2 interacts with two plastidial coiled-coil proteins. Surprisingly, one of the proteins is the thylakoid-associated MAR BINDING FILAMENT-LIKE PROTEIN1 (MFP1), which was proposed to bind plastid nucleoids. The other protein, MYOSIN-RESEMBLING CHLOROPLAST PROTEIN (MRC), contains long coiled coils and no known domains. Whereas wild-type chloroplasts contained multiple starch granules, only one large granule was observed in most chloroplasts of the mfp1 and mrc mutants. The mfp1 mrc double mutant had a higher proportion of chloroplasts containing no visible granule than either single mutant and accumulated ADP-glucose, the substrate for starch synthesis. PTST2 was partially associated with the thylakoid membranes in wild-type plants, and fluorescently tagged PTST2 was located in numerous discrete patches within the chloroplast in which MFP1 was also located. In the mfp1 mutant, PTST2 was not associated with the thylakoids and formed discrete puncta, suggesting that MFP1 is necessary for normal PTST2 localization. Overall, we reveal that proper granule initiation requires the presence of MFP1 and MRC, and the correct location of PTST2.

INTRODUCTION

Starch is the primary storage carbohydrate in plants, and is synthesized and stored as insoluble semicrystalline granules in the plastids of leaves and storage organs. Many plants use photoassimilates to produce a starch reserve in leaf chloroplasts during the day and then degrade this reserve during the subsequent night to provide metabolic energy required for growth (Stitt and Zeeman, 2012; Scialdone et al., 2013). In Arabidopsis thaliana leaves, each chloroplast typically contains five to seven starch granules (Crumpton-Taylor et al., 2012). The biochemical mechanism that initiates the synthesis of each starch granule, and the factors that control the number and size of granules within each chloroplast, are relatively poorly understood. The identification of key proteins involved in this process would suggest biotechnological strategies to manipulate granule size, which is of significant interest for the many food and industrial applications of starch (Jobling, 2004; Lindeboom et al., 2004).

Starch is composed of the glucose polymers (glucans) amylopectin and amylose. Amylopectin constitutes 70 to 90% of the starch granule and is a highly branched polymer, consisting of α-1,4-linked glucan chains and α-1,6-linked branch points. Amylose consists of long α-1,4-linked glucan chains and is mostly unbranched. Plants have multiple starch synthase (SS) isoforms that elongate α-1,4-linked glucan chains using ADP-glucose as the glucosyl donor. The SS1, SS2, and SS3 isoforms are involved primarily in amylopectin synthesis, and Arabidopsis mutants lacking any of these isoforms have altered amylopectin structure (Delvallé et al., 2005; Zhang et al., 2005, 2008). These SS isoforms act together with branching enzymes and debranching enzymes (ISOAMYLASE), which are required for the formation and correct arrangement of the branch points (Delatte et al., 2005; Dumez et al., 2006; Streb et al., 2008; Pfister et al., 2014, 2016). The GRANULE-BOUND STARCH SYNTHASE (GBSS) elongates amylose chains and is tightly associated with the starch granule (Smith et al., 2004; Seung et al., 2015). In Arabidopsis, this location of GBSS on starch depends on PROTEIN TARGETING TO STARCH1 (PTST1), a protein containing coiled coils and a glucan binding domain belonging to the Carbohydrate Binding Module 48 (CBM48) family (Lohmeier-Vogel et al., 2008; Seung et al., 2015). GBSS interacts directly with PTST1 via its own coiled-coil motifs. PTST1 then interacts with starch via its CBM48 domain and presumably aids GBSS localization. The amount of starch-bound GBSS, as well as amylose content, is greatly reduced in the Arabidopsis ptst1 mutant.

In Arabidopsis, the process of starch granule initiation that determines the final number of granules per chloroplast appears to involve a specialized set of proteins, distinct from those that synthesize the bulk of the starch polymers. Chloroplasts of the Arabidopsis mutant lacking SS4 have a dramatic reduction in the number of starch granules (typically containing zero or one large starch granule), but the starch has only minor alterations in amylopectin structure and amylose content (Roldán et al., 2007; Crumpton-Taylor et al., 2013). The mutant also accumulates ADP-glucose, indicating that the other starch synthases cannot effectively utilize this substrate without SS4 (Crumpton-Taylor et al., 2013; Ragel et al., 2013). These findings suggest that SS4 is not a major contributor to amylopectin structure like the other SS isoforms but has a specialized function upstream in granule initiation.

Recently, we identified two homologs of the PTST1 protein, PTST2 and PTST3, that are involved in granule initiation (Seung et al., 2017). Like PTST1, both PTST2 and PTST3 contain coiled coils and a CBM48 domain. PTST2 interacts specifically with SS4, and PTST3 interacts with PTST2. The ptst2 mutant typically has zero or one large granule per chloroplast and accumulates ADP-glucose. The ptst3 mutant has a minor reduction in granule number per chloroplast, but the ptst2 ptst3 double mutant has more starchless chloroplasts than the ptst2 single mutant. Since the CBM48 domain of PTST2 can bind to longer maltooligosaccharides (DP 10), but not shorter ones (DP 7), we proposed that PTST2 and PTST3 act in granule initiation by delivering maltooligosaccharide primers to SS4 that are suitable for elaboration (i.e., into a starch granule initial). The rice (Oryza sativa) ortholog of PTST2 (also referred to as FLOURY ENDOSPERM6 [FLO6]) and a putative ortholog in barley (Hordeum vulgare), affect starch synthesis in the seed endosperm (Peng et al., 2014; Saito et al., 2018). In the case of the rice ortholog, evidence was presented that it interacts with ISOAMYLASE1 (ISA1). However, we found no biochemical or phenotypic evidence that PTST2 interacts with ISA1 in Arabidopsis leaves (Seung et al., 2017).

In addition to its interaction with the PTSTs, Arabidopsis SS4 is proposed to interact with the plastidial fibrillins (Gámez-Arjona et al., 2014; Raynaud et al., 2016). This interaction, which is mediated by the nonenzymatic N-terminal domain of SS4, was proposed to localize SS4 to the thylakoid membrane and plastoglobules, and defines specific regions in the chloroplast where granule biogenesis can originate (Gámez-Arjona et al., 2014). So far, no phenotypic evidence for the importance of the fibrillins in granule initiation has been presented. However, it was shown that the N-terminal domain of SS4 is important for the formation of granules in their characteristic lenticular shape (Lu et al., 2018).

Here, we aimed to discover more proteins that are involved in starch granule initiation, using immunoprecipitation and mass spectrometry to look for interaction partners of PTST2 and PTST3. We found that PTST2 not only copurifies with SS4 and PTST3, but also with two plastidial coiled-coil proteins, MAR BINDING FILAMENT-LIKE PROTEIN1 (MFP1) and MYOSIN-RESEMBLING CHLOROPLAST PROTEIN (MRC). We present evidence that both proteins are required for normal starch granule initiation in Arabidopsis chloroplasts.

RESULTS

Identification of Two Coiled-Coil Proteins That Interact with PTST2

Having identified PTST2 and PTST3 as proteins required for proper granule initiation in Arabidopsis chloroplasts (Seung et al., 2017), we looked for proteins that interact with either protein in planta. We conducted immunoprecipitation experiments using Arabidopsis lines stably expressing PTST2 or PTST3 with a C-terminal YFP tag (Seung et al., 2017). Proteins extracted from rosette leaves were incubated with beads conjugated to a GFP antibody. Proteins recovered in the immunoprecipitate (IP) were digested with trypsin in solution, and the resulting peptides were analyzed using shotgun mass spectrometry. The identified peptides were searched against a database of Arabidopsis proteins from the TAIR10 genome annotation. As controls, we also conducted immunoprecipitations from lines expressing YFP-tagged PTST1, which is not involved in the granule initiation process, but rather in amylose synthesis (Seung et al., 2015), and from lines expressing CFP targeted to the plastid using the chloroplast transit peptide of the Rubisco small subunit. CFP and YFP, both variants of GFP, share 96% amino acid identity. An average of 151 ± 43 peptides matching GFP were detected in three replicates of the plastidial CFP IP, and peptides from a total of 638 proteins were identified in at least one replicate. These proteins were assumed to be contaminants resulting from unspecific binding to the tag or beads and were excluded from the final list of PTST-interacting proteins.

Peptides matching the PTST-YFP proteins were most abundant within each IP, showing that the fusion proteins were purified effectively, and several copurifying proteins were identified for each PTST isoform (Table 1). First, many PTST3 peptides were identified in the PTST2-YFP IP, and vice versa, confirming our previous finding that PTST2 and PTST3 interact with each other (Seung et al., 2017). Our data also confirm that PTST2 interacts with SS4, although a relatively low number of SS4 peptides were identified (Table 1). This was not unexpected as the interaction between PTST2 and SS4 is likely to be transient (Seung et al., 2017), although it should be noted that peptide counts are not strictly correlated with protein abundance. In the PTST2-YFP IP, we identified peptides corresponding to a myosin heavy chain-related protein encoded at locus At4g32190. We refer to this protein as MYOSIN-RESEMBLING CHLOROPLAST PROTEIN (MRC), since the TargetP/ChloroP programs for predicting the presence of organelle targeting peptides (Emanuelsson et al., 2007) suggested that this protein is located in the chloroplast. MRC is also referred to as PROTEIN INVOLVED IN STARCH INITIATION1 (PII1) by Vandromme et al. (2018), who recently identified it as an SS4-interacting protein. We also identified many peptides matching the MAR BINDING FILAMENT-LIKE PROTEIN1 (MFP1). MFP1 was previously described as a nucleoid binding coiled-coil protein that is bound to the chloroplast thylakoid membranes (Jeong et al., 2003), but its exact physiological function is unknown. Additionally, we found a small number of peptides matching VESICLE INDUCING PROTEIN IN PLASTIDS1, which is thought to play roles in maintaining the integrity of thylakoid and envelope membranes (Vothknecht et al., 2012; Zhang et al., 2012).

Table 1. Interaction Partners of PTST1, PTST2, and PTST3.

Accession	Protein		Loc.	Total Spectral Counts
				Rep. 1	Rep. 2	Rep. 3	∑ All
PTST1-YFP IP
–	GFP		–	82	52	48	182
AT5G39790	PTST1		M*	26	23	23	72
AT2G19750	Ribosomal protein S30 family protein		M	1	2	1	4
PTST2-YFP IP
AT1G27070	PTST2		C	219	295	752	1266
AT3G16000	MFP1, MAR binding filament-like protein		C	206	204	233	643
–	GFP		–	72	91	238	401
AT5G03420	PTST3		C	11	21	9	41
AT4G32190	MRC, myosin-resembling chloroplast protein		C	5	15	12	32
AT4G18240	SS4, starch synthase 4		C	1	4	3	8
AT1G65260	VIPP1, vesicle-inducing protein in plastids		C	4	3	1	8
PTST3-YFP IP
AT5G03420	PTST3		C	364	374	143	881
–	GFP		–	160	181	82	423
AT1G27070	PTST2		C	63	10	1	74
AT1G16610	SR45, serine/arginine-rich family protein		C**	1	3	5	9
AT1G50840	PolIA, DNA polymerase		C	2	5	2	9

YFP-tagged PTST proteins were expressed in Arabidopsis rosettes and purified by immunoprecipitation. Copurifying proteins were identified by LC-MS/MS. Values represent the total spectral count for all peptides matching each protein. Analyses were carried out on three independent purifications (Rep. 1, 2, and 3), and the sum of all spectral counts over the three replicates is shown (∑ all). The listed proteins had at least one matching peptide identified in all three replicates and were not identified in the control IP with the plastidial CFP. Proteins hits that were not identified in all three replicates were also removed. The location (Loc.) of each protein in the chloroplast (C) or mitochondria (M) as predicted by TargetP is indicated.

^*, Despite the predicted mitochondrial location for PTST1, the protein has been experimentally verified to localize to plastids (Seung et al., 2015).

^**, Despite the predicted chloroplast location of SR45, the protein has been experimentally verified to localize to the nucleus (Ali et al., 2003).

No peptides matching any of these proteins were detected in the PTST3-YFP IP. Instead, we detected relatively low numbers of peptides matching an SR45 serine/arginine-rich protein involved in mRNA processing (Day et al., 2012) and the chloroplast DNA polymerase PolIA (Parent et al., 2011; Baruch-Torres and Brieba, 2017). Although TargetP/ChloroP programs predicted that SR45 is located in the chloroplast, this protein is known to be located in the nucleus (Ali et al., 2003). Given that the major functions of these proteins are unrelated to starch metabolism, these interactions may not be biologically significant. Peptides matching ISA1 were also absent from both PTST2 and PTST3 IPs. This supports our previous evidence that PTST2 does not interact with ISA1 in Arabidopsis leaves (Seung et al., 2017), although its ortholog in rice, FLO6, is proposed to interact with ISA1 in the endosperm (Peng et al., 2014). None of the interaction partners identified for PTST2-YFP and PTST3-YFP were detected in the PTST1-YFP IP. We also note that GBSS was not detected in the PTST1-YFP IP. This was expected as most GBSS is bound to starch and therefore not present in our extracts, and the GBSS-PTST1 interaction in the stroma is likely to be transient (Seung et al., 2015).

Since MFP1 and MRC appear to be major interaction partners of PTST2 in the chloroplast, we further analyzed them using bioinformatics. First, since PTST2 and MFP1 are coiled coil-containing proteins, we used the COILS program (Lupas, 1996) to predict whether MRC also has coiled coils. The program predicted long coiled coils throughout the sequence of the protein with high probability scores (Figure 1A). This is consistent with the protein being annotated to be like myosin, which is also composed of long coiled coils (Schwaiger et al., 2002). The chloroplast transit peptides of MFP1 and MRC were predicted to be 41 and 27 amino acids in length, respectively, using ChloroP (Emanuelsson et al., 2007). Neither MFP1 nor MRC contain known domains, but both show weak similarity to the STRUCTURAL MAINATINANCE OF CHROMOSOMES (SMC) superfamily. SMC proteins are involved in chromosome organization and dynamics, and the weak similarity is likely due again to the presence of coiled coils in these proteins.

Figure 1.

MFP1 and MRC Are Highly Conserved Plant Proteins.

(A) Schematic illustration of domains in the Arabidopsis PTST2, MFP1, and MRC proteins. The chloroplast transit peptide (TP) predicted using ChloroP is shown in green, and the coiled-coil domain (CC) and CBM of PTST2 are shown in blue and purple, respectively. The locations of coiled coils, predicted using the COILS program with a 14-amino acid prediction window, are shown with an indication of probability score (from 0 to 1, where 1 is the highest probability). a.a., amino acids.

(B) and (C) Maximum likelihood phylogenetic tree of MFP1 (B) and MRC (C). Branch lengths represent the number of substitutions per site, indicated by the scale bars. One thousand bootstrap replicates were performed for each tree, and bootstrap values greater than 50 are shown above or underneath the nodes. The amino acid alignments used to generate these trees are provided in Supplemental Data Set 1.

We then conducted a phylogenetic analysis for both MFP1 and MRC proteins. We could not accurately align MFP1 and MRC amino acid sequences with each other as the two proteins were mostly dissimilar in sequence, suggesting that they are only distantly related, if at all. We therefore created separate alignments (Supplemental Data Set 1) and phylogenetic trees for each protein. We found that both MFP1 and MRC are highly conserved among higher plant genomes (Figures 1B and 1C), suggesting that they play an important role in this group. Interestingly, several species had multiple copies of MFP1, most notably grasses (maize [Zea mays], rice and Brachypodium distachyon), where a duplication early during their divergence resulted in the formation of two separate clades of MFP1-like sequences. An MFP1-like sequence was identified in the green algae Physcomitrella patens, but not in Selaginella moellendorffii, suggesting it may have been lost in the latter. Interestingly, an MRC-like sequence was found in S. moellendorffii, but not in P. patens.

MFP1 and MRC Play an Important Role in Granule Initiation

Since PTST2 participates in granule initiation, we investigated whether MFP1 and MRC also play roles in the same process. We obtained Arabidopsis mutants carrying T-DNA insertions in either gene, with two independent insertion alleles for each gene. The exact insertion site was determined for each allele using PCR amplification of genomic DNA and sequencing (Figure 2A; Supplemental Table 1). For MFP1, we obtained mfp1-1 and mfp1-2, both carrying insertions in exon 4. For MRC, we obtained mrc-1, with the insertion in the 5′ untranslated region (UTR; 3 to 5 bp upstream of the start codon), and mrc-3 with the insertion in exon 2. To determine the effect of the insertions on protein accumulation, we assessed MFP1 and MRC protein abundance in the mutants using immunoblots. We generated specific antibodies for each protein by immunizing rabbits with either synthetic MFP1-derived peptides or MRC recombinant protein expressed in and purified from Escherichia coli. Both MFP1 and MRC were readily detectable in total protein extracts of wild-type plants (Figure 2B). The MFP1 protein was undetectable in extracts from either mfp1 mutant. MRC protein was undetectable in the mrc-3 mutant, but a small amount was observed in the mrc-1 mutant. The T-DNA insertion in mrc-1 is in the 5′ UTR and thus could allow some translation of full-length protein. Levels of MFP1 were not altered in mrc mutants, and levels of MRC were not altered in mfp1 mutants. We also generated an mfp1 mrc double mutant by crossing the mfp1-1 mutant with mrc-3 to address possible redundancy as both proteins have long coiled coils. Levels of PTST2 and SS4 proteins were not altered in either mfp1 or mrc single mutant or in the double mutant.

Figure 2.

Arabidopsis Knockout Mutants for MFP1 and MRC.

(A) Schematic illustration of the gene model at the MFP1 and MRC loci. Exons are depicted by the dark blue boxes, while the 5′ and 3′ UTRs are shown in light blue. The position of the translation start (ATG) and stop (TGA/TAA) codons are indicated. Red arrowheads represent the insertion sites of the T-DNA.

(B) Immunoblot detection of MFP1, MRC, PTST2, and SS4 in total protein extracts of leaves from mfp1 and mrc single and double mutants. An immunoblot of actin was performed as a loading control. Lanes were loaded on an equal fresh weight basis (equivalent to 1.8 mg). The migration of molecular weight markers (in kD) is indicated on the left of each panel.

The rosettes of the mfp1 and mrc single and double mutants were visually indistinguishable from the wild type (Figure 3A). To check for alterations in the pattern of starch accumulation in these mutants, we harvested the rosettes at the end of the day and stained starch with iodine. The ss4 mutant, and to a lesser extent the ptst2 mutant, did not stain strongly for starch at the center of the rosette (Figure 3B), as previously reported (Crumpton-Taylor et al., 2013; Seung et al., 2017). However, this phenotype was not observed in the mfp1 and mrc single and double mutants, which all stained like the wild type. As iodine staining is not quantitative and is also influenced by starch composition, we quantified starch in the rosettes at the end of the day. The mfp1-2, mrc-3, and both mfp1 mrc double mutants had slightly more starch than the wild type. Therefore, the loss of MFP1 or MRC does not reduce the overall amount of starch synthesized during the day. However, at the end of the night, all mfp1 and mrc single and double mutants had slightly more starch than wild-type plants (Table 2). This effect was also seen in ptst2 and ss4, as reported previously (Crumpton-Taylor et al., 2013; Seung et al., 2017). Although previous studies reported a significant decrease in end-of-day starch content in the ss4 mutant relative to the wild type (Roldán et al., 2007), we failed to find a significant difference in our current and previous study (Seung et al., 2017), suggesting some variation in the severity of this aspect of the ss4 phenotype. Together with the results of the iodine staining, this suggests an overall change in the distribution of starch in the rosette, where there is less starch in the younger leaves and more in the older leaves.

Figure 3.

Phenotype of the mfp1 and mrc Single and Double Mutants.

(A) Photographs of 4-week-old rosettes.

(B) Iodine-stained rosettes. Plants were harvested at the end of the day, cleared of chlorophyll, and then stained in an iodine solution. Black arrows indicate typical leaves taken for analysis in (C) and (D), from genotypes with wild-type-like plant morphology and in the smaller rosette of ss4.

(C) Starch granules in mesophyll cell chloroplasts. Segments were excised approximately two-thirds along the length of young leaves toward the distal end (examples of young leaves marked with arrows in [B]) at the end of the day and were fixed and embedded in resin. Semithin sections were stained with toluidine blue and observed with light microscopy. Similar phenotypes were observed in at least three independent experiments performed on different batches of plants, and representative images are shown. Bars = 10 µm.

(D) Quantification of granule sections in mesophyll chloroplast sections. Light micrographs were acquired as in (C), and the number of starch granule sections observed within each chloroplast section was counted. Histograms show the relative (rel.) frequency of chloroplast sections containing a given number of granule sections, relative to the total number of chloroplast sections analyzed. Equal numbers of chloroplast sections were analyzed in leaf sections prepared from three different plants for each genotype (resulting in a total of n = 459 chloroplast sections per genotype). Arrows indicate the bin containing the median value.

Table 2. Starch Content of mfp1 and mrc Mutants.

Genotype	Starch Content (mg/g FW)
	End of Day	End of Night
Wild type	6.28 ± 0.31	0.22 ± 0.05
mfp1-1	6.99 ± 0.33	0.77 ± 0.26
mfp1-2	7.71 ± 0.45*	0.52 ± 0.06*
mrc-1	7.10 ± 0.39	0.76 ± 0.11*
mrc-3	7.68 ± 0.26*	1.08 ± 0.09*
mfp1-1 mrc-1	7.86 ± 0.23*	1.08 ± 0.14*
mfp1-1 mrc-3	8.15 ± 0.28*	1.41 ± 0.12*
ptst2-7	6.75 ± 0.43	0.99 ± 0.05*
ss4	5.71 ± 0.31	1.25 ± 0.11*

Values are the mean ± se from n = 5 to 6 individual plants. Values that are significantly different to the wild-type value (at each time point) under a two-tailed t test at P < 0.05 are indicated with an asterisk. FW, fresh weight.

We then investigated whether chloroplasts of mfp1 and mrc mutants contained fewer starch granules than wild-type chloroplasts. Segments were excised from a young leaf of a 4-week-old rosette at the end of the day (three individuals per genotype), and fixed and embedded in resin. Semithin sections were stained with toluidine blue to observe starch granules using light microscopy (Figure 3C). There were far fewer granules per chloroplast in both mfp1 and mrc mutants than in wild-type plants, as most chloroplast sections in the mutants contained only one granule section. To quantify this phenotype, we counted the number of granule sections in a total of 459 chloroplast sections per genotype (Figure 3D). Note that the number of granule sections in a chloroplast section is representative of the number of granules in each chloroplast, but is not the actual number of granules, as this information cannot be obtained from two-dimensional images (Crumpton-Taylor et al., 2012). Most chloroplast sections in the wild type contained between two and four granules. By contrast, ∼50% of chloroplast sections in both mfp1 mutants and the mrc-1 mutant contained one granule. Fewer granules were present in mrc-3 chloroplast sections, with only one granule observed in more than 75% of the cases. This difference in phenotype severity between mrc-1 and mrc-3 is likely due to the presence of residual MRC protein in mrc-1 (Figure 2B). Chloroplast sections in the mfp1-1 mrc-3 double mutant contained fewer granules than in either mfp1-1 or mrc-3 single mutant, as ∼25% had no visible starch granule and very few had more than one granule (Figure 3D). This was comparable to the ptst2 mutant. The greatest reduction in granule number relative to wild-type plants was observed in the ss4 mutant, which had no visible granule in over 50% of chloroplast sections.

For a more detailed view, we used transmission electron microscopy (TEM) to observe the chloroplasts of the mutants. Apart from the reduced frequency of starch granules, there were no obvious changes in chloroplast structure, such as in the architecture of thylakoids or grana stacks (Figure 4). Starch granules in the mutants were larger than those of the wild type, which could explain why total starch content was not altered in the mutants, despite the reduced frequency of granules per chloroplast. The granules in mfp1, mrc, and ptst2 appeared to have the typical flattened shape of wild-type granules. Granules in the mfp1 mrc double mutant either had the same flattened shape as granules from the wild type and single mutants or occasionally appeared slightly rounder. However, they were not as rounded as ss4 granules (Roldán et al., 2007). Closer examination of granule morphology by scanning electron microscopy on purified starch confirmed that the mfp1 and mrc mutants, as well as ptst2 and ss4, have much larger starch granules than the wild type (Figure 5). The change in the shape of mfp1 mrc granules observed using TEM was not obvious from the scanning electron microscopy analysis, and only granules from the ss4 mutant had the clear round morphology.

Figure 4.

Chloroplasts in the mfp1 and mrc Single and Double Mutants as Observed via TEM.

Sections were prepared from leaf segments harvested at the end of the day, and mesophyll chloroplasts were imaged. Bars = 2 µm.

Figure 5.

Starch Granules from the mfp1 and mrc Single and Double Mutants Observed with Scanning Electron Microscopy.

Granules were purified from 4-week-old rosettes. The upper panels were imaged under low magnification and are shown at the same scale. The lower panels were imaged under high magnification and are shown at the same scale. Bars = 4 µm.

We then tested whether the reduction in granule number in mfp1 and mrc mutants was accompanied by an accumulation of ADP-glucose, as previously observed in the ptst2 and ss4 mutants (Crumpton-Taylor et al., 2013; Ragel et al., 2013; Seung et al., 2017). Metabolites were extracted from entire rosettes harvested at the end of the day and analyzed by ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS). The wild-type and mfp1 plants had similar amounts of ADP-glucose, while the mrc plants had 2- to 3-fold more ADP-glucose than the wild type (Table 3). The mfp1 mrc double mutant accumulated 20-fold more ADP-glucose than the wild type. This was approximately one-third of the ADP-glucose measured in the ptst2-7 mutant. Interestingly, ADP-glucose only accumulated in the mutants in which we observed a substantial proportion of chloroplasts containing no starch granules (Figure 3).

Table 3. ADP-Glucose Content in mfp1, mrc1, and ptst2 Mutants.

Genotype	ADP-Glucose Content (nmol/g FW)
	Experiment 1	Experiment 2
Wild type	1.4 ± 0.1a	3.2 ± 0.4a
mfp1-1	1.5 ± 0.1a	–
mrc-3	4.0 ± 0.3b	–
mfp1-1 mrc-3	29.1 ± 1.7c	49.7 ± 1.7b
ptst2-7	78.3 ± 3.6d	138.6 ± 14.6c
ptst2-7 mfp1-1	–	150.7 ± 4.6c
ptst2-7 mrc-3	–	299.6 ± 12.1d
ptst2-7 mfp1-1 mrc-3	–	297.8 ± 3.7d
ss4	–	364.8 ± 33.4d

Metabolites were extracted from single 4-week-old rosettes harvested at the end of the day using the chloroform/methanol procedure. ADP-glucose was quantified using UHPLC-MS/MS. Values represent mean ± se of n = 4 rosettes. Within each experiment, values that are annotated with different letters are significantly different under a two-tailed t test at P < 0.05.

Our results show that both MFP1 and MRC are involved in starch synthesis in Arabidopsis and are required for the proper initiation of starch granules. The loss of MRC has a slightly greater impact on granule initiation than the loss of MFP1. However, granule initiation is more severely affected when both proteins are absent. The results also revealed a physiological role for MFP1 that is perhaps unrelated to its previously assumed role in nucleoid maintenance. Additionally, we investigated whether MFP1 and MRC also play a role in amylopectin and amylose biosynthesis. We found no evidence of alterations in amylopectin chain length distribution in the mfp1 and mrc single and double mutants (Supplemental Figure 1A), and there were only minor increases in the amylose content of starch granules (Supplemental Figure 1B). Therefore, both proteins appear to have specific roles in granule initiation.

The mfp1 Mutation Is Epistatic to ptst2

To further understand the role of MFP1 and MRC in granule initiation, we generated a series of double and triple mutants by crossing ptst2 with mfp1 mrc. We then assessed the severity of the granule initiation phenotype indirectly by measuring ADP-glucose. ADP-glucose levels in mfp1 ptst2 and ptst2 were similar, indicating that mfp1 is epistatic to ptst2 (Table 3). However, the mrc ptst2 double mutant accumulated twice the ADP-glucose content of ptst2, approaching the levels measured in the ss4 mutant. This suggests that MRC acts in granule initiation not only through its interaction with PTST2, but also through other mechanisms. As expected from the epistasis of mfp1 to ptst2, the mfp1 mrc ptst2 triple mutant had the same ADP-glucose levels as mrc ptst2.

PTST2 Associates with the Thylakoid Membrane via MFP1

MFP1 was previously shown to associate tightly with the thylakoid membranes on the stromal side (Jeong et al., 2003). Thus, if PTST2 interacts with MFP1, then the interaction must take place on the thylakoid membrane. To confirm the previous localization of MFP1 and investigate whether PTST2 was also thylakoid-associated, we sequentially extracted soluble and membrane-bound (insoluble) proteins from wild-type and mfp1 leaves. Gels were loaded on a fresh weight basis, and MFP1 and PTST2 were detected and quantified by immunoblotting. As reported, essentially all MFP1 was in the insoluble fraction (97% ± 1%, mean ±se, 3 extracts from different plants), although it should be noted that unspecific bands were present in the soluble fractions of both the wild type and mfp1, resulting in the overestimation of MFP1 abundance in the soluble fraction (Figure 6A). About 20.2% ± 0.2% of PTST2 was in the insoluble fraction in the wild type, but only 4.0% ± 0.2% was in the insoluble fraction in the mfp1 mutant. Surprisingly, we found that SS4 was primarily soluble in the wild type (97% ± 2%) and also in the mfp1 mutant (97% ± 3%). This is in contrast to previous reports that SS4 is almost exclusively associated with the thylakoids (Gámez-Arjona et al., 2014). MRC was primarily soluble in both the wild type (92.6% ± 0.2%) and the mfp1 mutant (95.1% ± 0.6%).

Figure 6.

PTST2 Associates with the Thylakoid Membranes.

(A) Immunoblot detection of MFP1, PTST2, SS4, and MRC in soluble and membrane-bound protein fractions. Young leaves were homogenized in a medium without detergent, and soluble proteins were recovered in the supernatant. After washing the pellet, the membrane-bound proteins were solubilized using 1% Triton X-100. Lanes were loaded on an equal fresh weight basis (equivalent to 1.35 mg). Three replicate extractions from leaves of different rosettes are shown for both Col and mfp1. The migration of molecular weight markers (in kD) is indicated on the left of each panel.

(B) Immunoblot detection of PTST2 in soluble and membrane-bound proteins in Col and mfp1, mrc, and mfp1 mrc mutants. Proteins were extracted as described for (A), and lanes were loaded on an equal fresh weight basis (equivalent to 1.35 mg). Three replicates prepared from different rosettes are shown for each genotype.

(C) Immunoblot detection of MFP1, PTST2, and SS4 in total (T), stromal (St.), and thylakoid (Th.) protein fractions of chloroplasts. The AGPase small subunit (APS1; stromal marker) and the photosystem II PsbA protein (thylakoid marker) were also detected. Fractions were prepared from chloroplasts isolated from 4-week-old rosettes. Equal amounts of protein (5 µg) were loaded on each lane. The migration of molecular weight markers (in kD) is indicated on the left of each panel.

We then investigated whether MRC also influences the amount of membrane-bound PTST2. The amount of PTST2 in the soluble and insoluble fractions was similar between the wild type and the mrc mutant (Figure 6B). In the mfp1 mrc double mutant, the PTST2 was primarily in the soluble fraction, as in the mfp1 single mutant. Therefore, MRC does not appear to influence the interaction between PTST2 and the membranes. To verify that PTST2 was interacting specifically with the thylakoid membranes, we lysed isolated Arabidopsis chloroplasts and fractionated the stroma and thylakoids by centrifugation. Protein preparations from the fractions were used for immunoblotting. We detected the known stromal protein, AGPase small subunit (APS1), and the known thylakoid-bound photosystem II protein, PsbA, exclusively in the appropriate fractions, showing that the fractionation was effective (Figure 6C). MFP1 was only detected in the thylakoid-bound fraction. PTST2 was detected both in the stromal and the thylakoid fractions in wild-type chloroplasts, but exclusively in the stromal fraction in the mfp1 mrc double mutant. SS4 was not detected in the thylakoid fraction, but was highly enriched in the stromal fraction. These results are consistent with the fractionation between soluble and membrane-bound proteins in Figures 6A and 6B, confirming that PTST2 is partially associated with the thylakoid membrane.

MFP1 Controls the Location of PTST2 within the Chloroplast

Since PTST2 associates with the thylakoid membranes through MFP1, we examined whether this interaction affects the localization of PTST2 within chloroplasts. We previously reported that the Arabidopsis PTST2-YFP protein, when expressed transiently in Nicotiana sylvestris leaves, was located in small distinct patches that were distributed throughout the chloroplast (Seung et al., 2017). These patches were not formed by the protein binding to starch granules, as similar patches were observed when PTST2-YFP was expressed in a starchless tobacco mutant. To localize PTST2 in Arabidopsis leaves, we created a series of transgenic lines expressing mCitrine-tagged PTST2 under its native promoter. The PTST2-mCitrine protein was functional, as its expression in the ptst2 mutant could restore normal granule numbers in the chloroplast (Supplemental Figure 2). Chloroplasts in the epidermal cell layer were imaged with confocal microscopy. PTST2-mCitrine was located in patches within the chloroplast (Figure 7A), similar to the localization we observed in tobacco leaves (Seung et al., 2017). We could also confirm in Arabidopsis that the patches were not likely to be starch granules, as we observed similar patches when we expressed the W475A mutant of PTST2-mCitrine. This mutant has the conserved glucan binding Trp-475 residue of PTST2 mutated to an alanine, which results in the loss of glucan binding in vitro (Supplemental Figure 3). The W475A mutant of PTST2 could not complement the ptst2 phenotype, demonstrating the importance of the CBM48 domain for PTST2 function (Supplemental Figure 2).

Figure 7.

MFP1 Is Required for the Proper Localization of PTST2 in Chloroplasts.

All images of mCitrine (mCit.)-tagged proteins were acquired from the lower epidermis of young Arabidopsis leaves using confocal laser-scanning microscopy. The mCitrine fluorescence (shown in yellow) was overlayed onto the chlorophyll autofluorescence (shown in red). Bars = 5 µm.

(A) PTST2-mCitrine or the W475A variant (which cannot bind to starch in vitro), expressed in the ptst2-3 mutant background under its native promoter (PTST2_pro). Each image was taken from an independent transformant in the T1 or T2 generation.

(B) PTST2-mCitrine expressed in the Col, mfp1, mrc, and mfp1 mrc backgrounds. Each image was taken from an independent transformant in the T1 or T2 generation.

(C) MRC-mCitrine expressed under its native promoter (MRC_pro) in the Col and mfp1 backgrounds. Each image was taken from an independent transformant.

We then examined the localization of PTST2 in the absence of MFP1 and/or MRC. The PTST2-mCitrine construct was transformed into mfp1, mrc, and mfp1 mrc. The location pattern of PTST2-mCitrine in the mfp1 background was very different to that in the wild-type background, typically appearing in one or two puncta per chloroplast, and occasionally in ring-like structures (Figure 7B). The altered location in mfp1 was not due to differences in PTST2-mCitrine abundance, as similar levels of expression were achieved in the wild-type and mfp1 backgrounds (Supplemental Figure 4). MFP1 is therefore required for the proper localization of PTST2 in the chloroplast. However, MRC does not appear to influence the location of PTST2, as the PTST2-mCitrine location pattern was similar in the wild-type and mrc backgrounds, and also similar in the mfp1 and mfp1 mrc backgrounds.

We also examined the location of MRC and MFP1. When MRC-mCitrine was expressed under its native promoter in Arabidopsis, it was located in distinct puncta within the chloroplast, which were more discrete and less abundant than the patches observed for PTST2 (Figure 7C). The location of MRC-mCitrine was similar in wild-type and mfp1 backgrounds. This suggests that although MFP1 is necessary for the proper localization of PTST2, it is not required for correct localization of MRC. Fluorescence from MFP1-CFP was not detectable when the protein was expressed on the MFP1 promoter, so the constitutive Ubiquitin10 promoter was used instead. MFP1-CFP also located to small patches that were distributed throughout the chloroplast, consistent with a previous study that examined MFP1 localization using immunofluorescence (Samaniego et al., 2006). The location pattern of MFP1-CFP was very similar to the pattern observed for PTST2 (Figure 8). We therefore coexpressed MFP1-CFP with PTST2-mCitrine in Arabidopsis leaves and observed a close overlap in the signals from the two proteins. This observation, combined with the altered location of PTST2-mCitrine in mfp1, strongly suggests that the location of PTST2 is determined by its interaction with MFP1.

Figure 8.

Colocalization of PTST2 and MFP1 in Chloroplasts.

PTST2-mCitrine (mCit.) was expressed under its native promoter (PTST2_pro), and MFP1-CFP was expressed under the constitutive Ubiquitin10 promoter (UBQ10_pro) in wild-type Arabidopsis plants. The lower epidermal cells of young leaves were imaged using confocal laser-scanning microscopy. Colors represent mCitrine (yellow), CFP (cyan), and chlorophyll (red). All images are shown at the same scale, and bar = 5 µm.

To observe the location of PTST2-mCitrine patches relative to the location of starch granules in Arabidopsis leaves, we coexpressed PTST2-mCitrine with a construct encoding a CFP-tagged GBSS, as GBSS associates tightly with starch granules in vivo. This was expressed under the constitutive Ubiquitin10 promoter as strong expression could ensure reliable visualization of the CFP tag, which is relatively dim compared with other fluorescent protein tags. To minimize the impact of this protein on starch synthesis, we abolished its starch synthase activity by mutating the Glu-486 residue to an alanine (GBSS E486A-CFP) (Seung et al., 2015). As expected, numerous GBSS E486A-CFP spots per chloroplast were observed in the wild-type background, consistent with the presence of multiple starch granules (Figure 9). We often observed PTST2 patches adjacent to GBSS spots, but a strict pattern was difficult to discern; there were many PTST2-mCitrine patches within each chloroplast, and each granule had a high probability of being in close proximity to a PTST2-mCitrine patch. There were fewer, larger GBSS E486A-CFP spots in mfp1 chloroplasts than in wild-type chloroplasts, consistent with the fewer, larger starch granules observed in this mutant. We observed again that PTST2-mCitrine formed fewer puncta or rings in mfp1 than in the wild type, but the puncta always located adjacent to the starch granules in mfp1. In cases where PTST2-mCitrine puncta were observed, GBSS E486A-CFP signal was seen between the puncta, whereas when PTST2-mCitrine rings were observed, the GBSS E486A-CFP signal was enclosed within the ring. In summary, the location of PTST2 is restricted to a few puncta in mfp1 chloroplasts, and starch granules are present in their vicinity.

Figure 9.

Localization of PTST2 and GBSS in Chloroplasts.

PTST2-mCitrine (mCit.) was expressed under its native promoter (PTST2_pro), and GBSS E486A-CFP was expressed under the constitutive Ubiquitin10 promoter (UBQ10_pro) in wild-type Arabidopsis plants. The lower epidermal cells of young leaves were imaged using confocal laser-scanning microscopy. Colors represent mCitrine (yellow), CFP (cyan), and chlorophyll (red). All images are shown at the same scale, and bar = 5 µm.

DISCUSSION

Plastidial Coiled-Coil Proteins Play an Important Role in Granule Initiation

In this study, we aimed to identify more proteins involved in starch granule initiation in Arabidopsis leaves and found two new proteins, MFP1 and MRC, by characterizing the central interaction partners of PTST2. MFP1 and MRC are long coiled-coil proteins that locate to the plastid (Figures 1, 7, and 8). Long coiled coils have been implicated in a variety of biological processes in eukaryotic cells, providing scaffolding, acting in the cytoskeleton, regulating transcription, organizing membrane networks (such as the Golgi apparatus), and also exerting physical force by acting as springs or levers (Kohn et al., 1997; Mason and Arndt, 2004; Rose and Meier, 2004). To our knowledge, there have been no reports of long coiled-coil proteins participating in starch metabolism, but shorter coiled-coil regions have been identified in the sequences of SS4 (Leterrier et al., 2008; Crumpton-Taylor et al., 2013), all PTST proteins (Seung et al., 2015, 2017), and other enzymes involved in starch synthesis and degradation (Lohmeier-Vogel et al., 2008).

It now appears that proper starch granule initiation in Arabidopsis leaves requires at least five proteins (SS4, PTST2, PTST3, MFP1, and MRC), which coordinate the process through protein-protein and protein-glucan interactions. Chloroplasts in the mfp1 and mrc mutants, like those in the ptst2 mutant, contained fewer starch granules than wild-type chloroplasts (Figure 3). Similar findings for the mrc (pii1) mutants were reported by Vandromme et al. (2018). Our studies suggest that both MFP1 and MRC are likely to act in granule initiation by interacting with PTST2, either directly or indirectly through common interaction partners like SS4. The mrc knockout mutant (mrc-3) had fewer granules than the mfp1 mutants, and the mfp1 mrc double mutant had fewer granules than either single mutant, with a higher proportion of chloroplast sections in which no granules were observed (Figure 3). We previously proposed that the elongation of suitable maltooligosaccharide primers by SS4 and PTST2 is an important step in granule initiation (Seung et al., 2015, 2017). The degree to which this process is hindered in the various mutants may be reflected in the proportion of starchless chloroplasts observed. A stronger inhibition of the overall process makes it less likely that a successful granule initiation can occur at the single chloroplast level, resulting in more starchless chloroplasts.

The level of ADP-glucose accumulated in the mutants appeared to correlate with the proportion of starchless chloroplasts. The mfp1 mutant did not accumulate ADP-glucose, and the mrc mutant only accumulated a small amount, despite the substantial reduction in the number of granules per chloroplast observed in both mutants. However, a much greater accumulation was observed in the mfp1 mrc double mutant (Table 3). These observations are consistent with our previous suggestion that ADP-glucose accumulation is not caused by reduced starch granule numbers per se, but by the complete failure of granule initiation in a significant proportion of chloroplasts (Seung et al., 2017). High levels of ADP-glucose might only occur in starchless chloroplasts, which are more common in ptst2 and mfp1 mrc than in the other mutants examined.

However, it should be noted that the amount of glucose sequestered in this ADP-glucose pool was relatively low compared with the total amount of starch synthesized. This could explain why total starch content was unaffected in the mutants, despite the accumulation of ADP-glucose.

The fact that granule initiation is more strongly affected in the mfp1 mrc double mutant than either single mutant could suggest that the two proteins play partly redundant roles in granule initiation. However, we suspect that they play distinct roles and that their loss has a cumulative impact on the initiation process. Our reasoning is as follows; MFP1 is a thylakoid-associated protein, while MRC is primarily soluble (Figure 6), and the two proteins have very different location patterns within the chloroplast (Figures 7 and 8). MFP1 plays a major role in localizing PTST2 within the chloroplast, as in its absence, PTST2 no longer associates with the thylakoid membrane and does not form its typical patchy location pattern within the chloroplast. By contrast, the loss of MRC does not have a noticeable effect on the location of PTST2. Also, MFP1 probably acts in granule initiation through ptst2, as the mfp1 ptst2 double mutant had the same amount of ADP-glucose accumulation as the ptst2 single mutant. By contrast, the mrc ptst2 double mutant accumulated much more ADP-glucose than either single mutant, strongly suggesting that MRC has functions in granule initiation that are independent of PTST2 (Table 3). The exact role of MRC is unknown, but it may have a direct influence on the function of other granule initiation proteins, such as SS4, aside from acting in the process by interacting with PTST2. Indeed, MRC (PII1) was found to interact with SS4 in a yeast two-hybrid assay, suggesting that the two proteins are likely to interact directly in the absence of PTST2 (Vandromme et al., 2018).

We previously demonstrated that PTST2 interacts with PTST3 and suggested that the two proteins play partially overlapping roles in granule initiation, since the ptst2 ptst3 double mutant initiated fewer granules than the ptst2 and ptst3 single mutants (Seung et al., 2017). The exact role of PTST3 in granule initiation is unclear. We confirmed here that PTST2 and PTST3 interact, as PTST3 was detected in the PTST2-YFP IP, and vice versa (Table 1). However, we did not detect SS4, MFP1, or MRC in the IP with PTST3-YFP. This could indicate that PTST3 only weakly interacts with these proteins in comparison to PTST2; thus, the interactions were lost during the IP procedure. This would be consistent with the mild reduction in granule frequency in the ptst3 single mutants relative to the ptst2 mutants (Seung et al., 2017). It is also possible that PTST3 does not interact with these proteins directly, but only indirectly via PTST2. This would imply that in the ptst2 mutant, PTST3 acts in granule initiation via a mechanism that is independent of the other granule-initiating proteins. Further biochemical analyses determining the stoichiometry and stability of these protein-protein interactions, as well as more genetic analyses, will shed light on the function of PTST3 in granule initiation.

A Novel Role for MFP1 in Starch Synthesis

MFP1 is highly conserved in plants (Figure 1) and is distinct from the characterized classes of long coiled-coil proteins from animals and fungi (Rose and Meier, 2004). Our discovery that it is involved in starch biosynthesis is unexpected. It was originally discovered as a nuclear-located DNA binding protein in tomato (Solanum lycopersicum; Meier et al., 1996) but was later shown to locate to chloroplast thylakoid membranes, where it could be anchoring nucleoids (Jeong et al., 2003). MFP1 could be detected in both chloroplasts and nuclei using immunofluorescence and immunogold microscopy, although it cannot be seen in the nuclei using GFP fusions, possibly because the GFP may hinder nuclear import (Samaniego et al., 2006). The exact role of the DNA binding activity of MFP1 is not known, and to our knowledge, there is no report of altered plastid nucleoid distribution or plastome transcription in the mfp1 mutant. Recent studies demonstrated that MFP1 is relatively loosely bound to nucleoids, compared with core components of nucleoid transcription (Melonek et al., 2010; Powikrowska et al., 2014). It is possible that MFP1 is a versatile protein that plays multiple roles, by controlling the localization of starch granule-initiating proteins and other bodies (e.g., nucleoids) within the chloroplast. Multiple copies of the MFP1 gene were present in many plants, most notably in grasses where we observed two distinct clades of MFP1, but also in Brassica rapa, Gossypium raimondii, and Musa acuminata (Figure 1B; Supplemental Data Set 1). We speculate that these duplications could allow specialization of multiple functions.

In wild-type Arabidopsis leaves, starch granules are distributed throughout the chloroplast and form within “pockets” between the thylakoids (Figure 4). We rarely observe chloroplasts where all the starch granules aggregated in a single region, although sometimes there is more than one in a single pocket. The pattern of starch granule location within each chloroplast may therefore involve coordination between granule initiation and plastid membranes. SS4 was previously reported to associate almost exclusively to the thylakoid membranes, more specifically to the plastoglobules by interacting with plastidial fibrillins (Gámez-Arjona et al., 2014; Raynaud et al., 2016). Although our study could not confirm the thylakoid association of SS4, the authors of the previous studies also proposed that starch granules may be initiated in distinct regions of the chloroplast.

The interaction between PTST2 and MFP1 provides a direct link between a component involved in granule initiation and a thylakoid-bound protein. PTST2 probably associates with the membrane primarily through its interaction with MFP1, as it is not predicted to contain any transmembrane domains or signal peptides. This is consistent with the fact that there was almost no interaction between PTST2 and the thylakoids in the mfp1 mutant (Figure 6). The exact proportion of PTST2 that is associated to the membrane is not yet clear. Using semiquantitative immunoblots, we estimated that around 20% of total PTST2 was associated with membranes in wild-type leaves (Figure 6). However, the location of PTST2 observed with microscopy closely overlapped that of MFP1, suggesting that the majority of PTST2 is associated with MFP1 on the thylakoid membrane in intact chloroplasts (Figure 8). On the one hand, it is possible that a substantial amount of thylakoid-bound PTST2 dissociates into the soluble fraction during protein extraction, leading to an underestimation of PTST2 abundance in the membrane-bound fraction using the immunoblot. On the other hand, it is also possible that stromal PTST2 is difficult to detect using microscopy, in comparison to the signal from the MFP1-associated PTST2, where fluorescence is concentrated in discrete patches. Since the interaction between SS4 and PTST2 is also likely to be transient (Seung et al., 2017), it is not surprising that only a fraction of SS4 was membrane-associated (around 3%). However, our findings are in contrast to those of Gámez-Arjona et al. (2014), who reported that almost all SS4 is associated with the thylakoid membrane. We do not yet know the reason for this difference.

We propose that MFP1 determines specific areas of the chloroplast at which starch granules are initiated, by restricting the location of PTST2. The patchy location of fluorescently tagged MFP1 implies that it is not uniformly distributed on the thylakoid membranes, since other fluorescently tagged thylakoid-bound proteins appear homogenously distributed in the chloroplast under confocal microscopy (Gerdes et al., 2006; Vainonen et al., 2008). The proper location of PTST2 with MFP1 appears to be important for initiating the correct number of granules. In the mfp1 mutant, PTST2 is mislocated into a few large puncta, limiting its distribution within the chloroplast (Figure 7). This may restrict the formation of granules, thus reducing the final number of starch granules. Indeed, we only observed the presence of starch granules adjacent to the few PTST2 puncta in chloroplasts of the mfp1 mutant (Figure 9). We do not know why PTST2 forms puncta in the mfp1 mutant, rather than forming a diffuse location pattern throughout the stroma. However, the PTST2 protein itself appears to be soluble (Figure 6), and the puncta are not likely due to the PTST2-mCitrine protein interacting with existing starch granules, as both the PTST2 WT- and W475A-mCitrine variants had similar location patterns in the mfp1 background (Supplemental Figure 5). Although the few puncta resemble the MRC location pattern, it is not certain that PTST2 aggregates with MRC in the absence of MFP1, since PTST2-mCitrine formed similar few puncta in the mfp1 mrc double mutant. However, there could be other, as yet undetected interaction partners that are responsible for the formation of these puncta in the mfp1 background.

The localization of fluorescently tagged PTST2, MFP1, and MRC, as well as the previously studied localization of PTST3 and SS4 (Szydlowski et al., 2009; Seung et al., 2017; Lu et al., 2018) suggests that all proteins known to be important in granule initiation locate to distinct subdomains within the plastid. The location of SS4 depends on its N-terminal domain containing the coiled coils (Lu et al., 2018). The correct location of SS4 is also important for starch granule morphology, as the expression of truncated SS4 lacking the N-terminal domain in the ss4 mutant appears to restore the initiation of multiple granules, but those granules are spherical rather than lenticular (Lu et al., 2018). Neither the correct location of PTST2, nor the presence of PTST2 itself, is required to establish correct granule morphology, as both mfp1 and ptst2 mutants have lenticular granules (Figure 5). The subdomains or patches observed for each granule initiation protein may therefore play distinct roles in determining granule number and/or granule morphology. To further understand these distinct roles, we are currently determining the extent of colocation between all the granule-initiation proteins and localizing each protein-protein interaction.

Potential Targets for the Biotechnological Modification of Starch

Overall, our discovery of more proteins involved in the granule initiation process opens up gene targets for the biotechnological modification of starch, particularly to modify the size of starch granules. There is significant industrial interest in modifying granule size in starch crops, as granule size affects both the ease of commercial starch preparation, and the final quality of various food and non-food products (Chen et al., 2003; Lindeboom et al., 2004). MFP1 and MRC may be useful targets for increasing granule size, as the mfp1 and mrc mutants produce starch granules that are substantially larger than wild-type granules (Figure 5) and do not have major effects on total starch content or the structure of amylopectin or amylose content or display the deleterious ADP-glucose accumulation seen in ss4 and ptst2, the two other mutants with single, large granules (Table 3). Both proteins are conserved in major starch crops, including the cereals, potato (Solanum tuberosum), and cassava (Manihot esculenta) (Figure 1). The fact that mutations in PTST2 in rice and barley (also referred to as FLO6) have altered starch synthesis in the endosperm suggests that PTST2 participates in starch synthesis in that tissue (Peng et al., 2014; Saito et al., 2018). Further investigations will now be required to determine the role of MFP1 and MRC in storage organs of different starch crops. This will be particularly interesting for MFP1, given that amyloplasts lack true thylakoid membranes. However, there is evidence that granule initiation in the maize endosperm is affected by alterations in the amyloplast membrane. The maize opaque5 mutant, defective in monogalactosyldiacylglycerol synthase activity, has altered galactolipid composition in the amyloplast membranes and initiates starch granules in compound form (several granules that initiate in close proximity and later fuse) rather than as single simple granules (Myers et al., 2011). It is possible that MFP1 associates with amyloplast membranes in storage organs.

METHODS

Plant Materials and Growth Conditions

Arabidopsis thaliana plants were grown on soil in controlled environment chambers: either AR-95L (Percival Scientific) fitted with fluorescent lamps and supplemented with red LED panels or Climatic Cabinet (Kälte 3000) fitted with fluorescent tubes. The chambers were set to 20°C with relative humidity of 65% and a 12-h-light/12-h-dark regime with light intensity of 150 μmol photons m⁻² s⁻¹.

All T-DNA insertion mutants are in the Columbia background and were obtained from the Nottingham Arabidopsis Stock Centre (accession numbers provided below and in Supplemental Table 1). All mutants were homozygous for the T-DNA allele, as determined by PCR-based genotyping (primer sequences are provided in Supplemental Table 2). The exact T-DNA insertion site for each mutant can be found in Supplemental Table 1. The mfp1, mrc, and ptst2 double and triple mutants were generated by crossing plants carrying the mfp1-1, mrc-3, and ptst2-7 alleles.

Immunoprecipitation experiments were performed on Arabidopsis lines stably expressing PTST2-YFP and PTST3-YFP, which were generated in a previous study (Seung et al., 2017). The PTST1-YFP line was generated in this study, as described below.

Cloning and Plant Transformation

The sequences of primers used for all the cloning steps described below are provided in Supplemental Table 2.

For PTST2-mCitrine expression under the native PTST2 promoter: The PTST2 promoter and 5′ UTR (1728 bp immediately preceding the start codon) was amplified from Arabidopsis genomic DNA, flanked with attB4 and attB1R recombination sites. The amplicon was recombined into the Multisite Gateway-compatible entry vector, pDONR P4-P1R, using BP clonase II (Thermo Fisher Scientific). The PTST2 coding sequence in Gateway entry vector pDONR 221 was previously cloned by Seung et al. (2017). The coding sequence of the mCitrine tag was synthesized as a double-stranded gBlocks DNA fragment (Integrated DNA Technologies), with codon optimization for Arabidopsis expression using the online tool provided by the manufacturer, and flanked by attB2R and attB3 recombination sites. This sequence was recombined into the multisite Gateway-compatible entry vector, pDONR P2R-P3, using BP clonase II (Thermo Fisher Scientific). The entry vectors containing the promoter, the coding sequence, and the tag were used in an LR clonase II plus (Thermo Fisher Scientific) recombination reaction, assembling the 5′-PTST2 promoter:PTST2 coding sequence:mCitrine tag-3′ expression cassette in the binary vector, pH7_m3,4GW,0 (University of Ghent/VIB). Site-directed mutagenesis to introduce the W475A mutation into the PTST2 coding sequence was performed with the Quikchange Lightning kit (Agilent) according to the manufacturer’s instructions.

For MRC-mCitrine expression under the native MRC promoter, the genomic fragment consisting of the promoter and 5′ UTR (1674 bp upstream of the start codon), and all exons and introns, was amplified from Arabidopsis genomic DNA, flanked at the 5′ end by the sequence CACC. The amplicon was then cloned into the Gateway entry vector, pENTR, between the attL1 and attL2 sites using the D-TOPO/pENTR cloning kit (Thermo Fisher Scientific). The MRC:pENTR and the mCitrine:pDONR P2R-P3 vectors were used in an LR clonase II plus (Thermo Fisher Scientific) reaction to assemble the MRC-mCitrine expression cassette into the binary vector pH7_m3,4GW,0. As this was a three-way Gateway recombination vector, an additional entry vector with 93 bp encoding 3X myc epitopes between attL4 and attL1R sites was included in the LR clonase reaction. This sequence is not translated as it recombines before the MRC promoter.

To clone the MRC-cTP:pET21a+ vector for MRC expression with a C-terminal His-tag in Escherichia coli, the coding sequence of the protein (excluding the region encoding the predicted transit peptide, the first 29 amino acids) was amplified from an Arabidopsis cDNA preparation and cloned into the pET21a+ vector using the NdeI and NotI sites, in frame with the C-terminal His-tag.

For MFP1-CFP expression in plants, the MFP1 coding sequence was amplified from an Arabidopsis cDNA preparation, flanked with attB1 and attB2 sites. The amplicon was recombined into the Gateway entry vector, pDONR 221, using BP clonase II (Thermo Fisher Scientific). The coding sequence was then recombined into the pUBC-CFP vector (Grefen et al., 2010) using LR clonase II, in frame with the C-terminal CFP tag and downstream of the Ubiquitin 10 promoter. For GBSS E486A-CFP expression, site-directed mutagenesis (with the Quikchange Lightning kit) was used to introduce the E486A amino acid substitution into the coding sequence of GBSS in the GBSS:pCR8 Gateway entry vector (Seung et al., 2015). The mutated coding sequence was recombined into the pUBC-CFP vector using LR clonase II (Thermo Fisher Scientific). For PTST1-YFP expression, the PTST1 coding sequence within the PTST1:pCR8 vector (Seung et al., 2015) was recombined into the pB7YWG2 vector (Karimi et al., 2002) using LR clonase II, in frame with the C-terminal YFP tag and downstream of the 35S promoter.

For the stable transformation of Arabidopsis plants, we used the floral dip method with Agrobacterium tumefaciens cultures carrying the appropriate transformation vector, as described in Zhang et al. (2006).

Immunoprecipitation and Mass Spectrometry

Proteins were extracted from leaves harvested from Arabidopsis rosettes stably expressing PTST1-YFP, PTST2-YFP, or PTST3-YFP by homogenizing in ice-cold immunoprecipitation medium (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% [v/v] Triton X-100, 1 mM DTT, and Complete Protease Inhibitor cocktail [Roche]) at a ratio of 1 mL medium per 100 mg fresh weight. Insoluble material was removed by centrifugation at 20,000g for 5 min at 4°C, and the supernatant was incubated for 1 h at 4°C with µMACS magnetic beads conjugated to anti-YFP (Miltenyi Biotec). The beads were captured with a µColumn on a magnetic stand (Miltenyi Biotec) and washed five times with immunoprecipitation medium before eluting the bound proteins with elution medium (50 mM Tris-HCl, pH 6.8, and 2% [w/v] SDS). Proteins were precipitated with 10% (w/v) TCA and pelleted. After two washes with cold acetone, proteins were redissolved in 10 mM Tris-HCl, pH 8.2, and 2 mM CaCl₂ and digested with Trypsin for 30 min at 60°C. The resulting peptides were dried, dissolved in 0.1% (v/v) formic acid, and analyzed by LC-MS/MS on a nanoAcquity UPLC (Waters) coupled to a Digital PicoView source for electrospray ionization (New Objective) and a Q Exactive mass spectrometer (Thermo Fisher Scientific). Peptides were trapped on a Symmetry C18 trap column (5 µm, 180 µm × 20 mm; Waters) and separated on a BEH300 C18 column (1.7 µm, 75 µm × 150 m; Waters) at a flow rate of 250 nL/min using a gradient from 1% solvent B (0.1% formic acid in acetonitrile)/99% solvent A (0.1% formic acid in water) to 40% solvent B/60% solvent A over 90 min. Mass spectrometry settings were as follows: precursor scan range, 350 to 1500 m/z; resolution, 70,000; maximum injection time, 100 ms; threshold, 3e6; fragment ion scan range, 200 to 2000 m/z; resolution, 35,000; maximum injection time, 120 ms; and threshold, 1e5. Peptides were searched against the TAIR10 Arabidopsis proteome database with the Mascot search engine (Matrix Science, version 2.4.1), with fragment ion mass tolerance of 0.030 D, parent ion tolerance of 10.0 ppm, and oxidation of methionine as a variable modification. Scaffold (Proteome Software) was used to validate MS/MS-based peptide and protein identifications, with protein identification threshold set to 95%, and peptide identification threshold set to 90%. Each IP was conducted in triplicate, and proteins that were identified in the IP with the plastidial CFP were removed from the final list of interactors. Proteins hits that were not identified in all three replicates were also removed.

Bioinformatics Analysis

Coiled coils were predicted in amino acid sequences using the COILS/PCOILS program (Lupas, 1996), with a 14-amino acid prediction window.

For phylogenetic analysis, sequences of MFP1 and MRC homologs were identified using pBLAST against the Phytozome database, searching in selected species that represent the major evolutionary branches of land plants. Sequences were aligned using Muscle (Edgar, 2004) and the alignment was edited to remove poorly aligning regions. A tree was constructed using the maximum likelihood method based on the JTT matrix-based model in MEGA v7 using 1000 bootstrap replicates (Kumar et al., 2016). The alignments used to build the trees are available as Supplemental Data Set 1. Note that the alignment was edited to remove poorly aligned regions, and the sequences in Supplemental Data Set 1 may not be full amino acid sequences of the proteins.

Starch Content and Iodine Staining

Starch content was quantified as described by Smith and Zeeman (2006). Briefly, the entire rosette was homogenized in ice-cold 0.7 M perchloric acid, and the insoluble fraction was collected by centrifugation at 5000g for 5 min at 4°C. The pellet was washed three times with cold 80% ethanol and resuspended in water. Starch in the pellet was gelatinized at 95°C and then digested to glucose using α-amylase and amyloglucosidase (Roche). Glucose was quantified using the hexokinase/glucose-6-phosphate dehydrogenase-based assay (Roche), and starch was calculated in glucose equivalents.

To stain starch in the rosette with iodine, the harvested plant was decolorized in 80% ethanol and then stained with Lugol solution (I₂/KI solution; Sigma-Aldrich).

Visualization of Starch in Chloroplasts Using Light and Electron Microscopy

Leaf segments (∼2 mm²) were excised from approximately two-thirds along the length (toward the distal end) of a young leaf on a 4-week-old rosette (examples of leaves taken are indicated in Figure 3B). Segments were fixed in glutaraldehyde, osmocated, and dehydrated as described in detail by Eicke et al. (2017) and then embedded in Epon (Sigma-Aldrich) or Spurr resin (Polysciences) according to the manufacturer’s instructions. For light microscopy, semithin sections were produced, stained in toluidine blue, and visualized with an AxioImager Z2 microscope fitted with a 100× oil immersion lens (1.4 numerical aperture) and AxioCam monochrome camera (Zeiss). For TEM, ultrathin sections were imaged on a Morgagni 268 microscope (FEI) as described by Eicke et al. (2017).

Quantification of ADP-Glucose Content

ADP-glucose was quantified using the method described by Seung et al. (2016). Briefly, a whole rosette was harvested at the end of the day and was immediately frozen by plunging in liquid N₂ while still illuminated. The frozen rosette was ground into a fine powder, and metabolites were extracted using chloroform/methanol, as described by Arrivault et al. (2009). Metabolites were analyzed by UHPLC-MS using a 1290 Infinity UHPLC (Agilent) with an Aquity T3 end-capped reverse phase column (Waters) and a QTRAP 5500 triple quadrupole MS (AB Sciex). ADP-glucose was quantified against a standard curve produced using commercial ADP-glucose (Sigma-Aldrich).

Amylose Content and Scanning Electron Microscopy

Starch was purified from Arabidopsis rosettes using the method described by Seung et al. (2015). Amylose content was determined in purified starch granules using the iodine colorimetry method, described by Hostettler et al. (2011). The purified granules were also observed with an SU5000 scanning electron microscope (Hitachi).

Protein Extraction and Immunoblotting

For extraction of total protein, leaves were homogenized directly in SDS-PAGE loading medium (50 mM Tris-HCl, pH 6.8, 2% [w/v] SDS, 100 mM DTT, 3% [v/v] glycerol, and 0.005% [w/v] bromophenol blue) at a ratio of 1 mL medium per 100 mg fresh weight. Gels were loaded on an equal fresh weight basis.

For the sequential extraction of soluble and membrane-bound proteins, leaves were homogenized in ice-cold protein extraction medium (100 mM Tricine-KOH, pH 7.9, 5 mM MgCl₂, and Complete Protease inhibitor cocktail [Roche]), at a ratio of 100 mg tissue per 1 mL medium. Soluble proteins were recovered in the supernatant after spinning at 20,000g at 4°C for 5 min. The pellet was then washed twice in cold protein extraction medium and then resuspended in the same medium supplemented with 1% Triton X-100. The volume used for each wash and resuspension was equivalent to the original volume of medium used for the initial homogenization. Membrane-bound proteins were then collected in the supernatant after spinning at 20,000g at 4°C for 5 min. Samples were mixed with SDS-PAGE loading medium, and gels were loaded on an equal fresh weight basis.

For the analysis of stromal and thylakoid-associated proteins, proteins were fractionated from isolated chloroplasts exactly as described by Bals and Schünemann (2011). Proteins in each fraction were precipitated using TCA and resuspended in total protein medium (40 mM Tris-HCl, pH 6.8, 5 mM MgCl₂, 1× Protease inhibitor cocktail, and 2% SDS). Proteins were quantified with the Pierce BCA Protein Assay (Thermo Fisher Scientific) before the addition of SDS-PAGE loading medium, and gels were loaded on an equal protein basis.

To produce antibodies against MFP1, two synthetic peptides from the MFP1 amino acid sequence (peptide 1, C-VNSTDNKEKSDNTVT; and peptide 2, C-ALADERGNEIKTSKV; both with a cysteine residue added at the C-terminal to facilitate conjugation) were conjugated to Keyhole limpet hemocyanin (KLH), and both peptides were used together to immunize rabbits (Double X program; Eurogentec). Antibodies were purified from the antiserum over Protein A-agarose (Roche), and antibodies specific for KLH were removed by running the purified antibodies over a HiTrap NHS-Activated HP column (GE Healthcare) conjugated with commercial KLH (Sigma-Aldrich) according to the manufacturer’s instructions.

To produce antibodies for MRC, the MRC-cTP:pET21a+ vector was used to express recombinant His-tagged MRC protein in E. coli and was purified as described previously (Seung et al., 2015). The purified protein was used to immunize rabbits (Eurogentec). Antibodies specific for MRC were affinity purified from the antisera using a HiTrap NHS-Activated HP column (GE Healthcare) conjugated with MRC recombinant protein according to the manufacturer’s instructions.

For immunoblotting, the following primary antibody dilutions were used: anti-MFP1, 1:1000; anti-MRC, 1:200; anti-PTST2 (Seung et al., 2017), 1:200; anti-SS4 (Roldán et al., 2007), 1:5000; anti-actin (Sigma-Aldrich; A0480, lot 056M4819V), 1:10,000; anti-GFP/YFP (Abcam; ab290, lot GR278073-1), 1:10,000; anti-APS1 (Agrisera; AS11 1739, lot 1406), 1:5000; and anti-PsbA (Agrisera; AS05 084, lot 1207), 1:10,000. Proteins were detected using chemiluminescence from horseradish peroxidase-coupled secondary antibodies (Bio-Rad) or using near-infrared fluorescence from IRDye 800CW-conjugated secondary antibodies with an Odyssey CLx detection system (Li-Cor).

Confocal Microscopy

Fluorescence of YFP, mCitrine, CFP, and chlorophyll was imaged in leaf tissue using confocal laser scanning microscopy as described by Seung et al. (2015). YFP/mCitrine was excited with an argon laser at 514 nm, and emission was captured between 518 and 557 nm. CFP was excited with an argon laser at 458 nm, and emission was captured between 462 and 500 nm. Chlorophyll autofluorescence was excited with an argon laser at 514 nm, and emission was captured between 662 and 721 nm.

Starch Binding Assay

Recombinant His-PTST2 protein was expressed in and purified from E. coli as described by Seung et al. (2017). Starch binding assays were performed exactly as described by Seung et al. (2017).

Accession Numbers

Sequence data of the genes studied in this article can be found in TAIR (www.arabidopsis.org) under the following accession numbers: MFP1 (At3g16000), MRC (At4g32190), PTST1 (At5g39790), PTST2 (At1g27070), PTST3 (At5g03420), SS4 (At4g18240), and GBSS (At1g32900). The accession numbers of T-DNA mutants used in this study are as follows: mfp1-1 (SALK_124298), mfp1-2 (SALK_037017), mrc-1 (SALK_122445), mrc-3 (SAIL_1151_E06), ptst2-3 (SAIL_1148_C07), ptst2-7 (SALK_73591), and ss4 (GABI_290D11).

Supplemental Data

• Supplemental Figure 1. Minor alterations in amylopectin chain length distribution and amylose content in mfp1 and mrc single and double mutants.

• Supplemental Figure 2. PTST2-mCitrine and MRC-mCitrine are functional proteins in vivo.

• Supplemental Figure 3. The PTST2 W475A mutant cannot interact with starch in vitro.

• Supplemental Figure 4. PTST2-mCitrine expression levels in the Col and mfp1 backgrounds.

• Supplemental Figure 5. PTST2 W475A-mCitrine does not form a diffuse location pattern in the stroma of mfp1 chloroplasts.

• Supplemental Table 1. Arabidopsis mutants used in this study.

• Supplemental Table 2. Oligonucleotide primers used in this study.

• Supplemental Data Set 1. Amino acid sequence alignments used for phylogenetic analysis of MFP1 and MRC.

Dive Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• APS1 Gramene: AT5G48300

• APS1 Araport: AT5G48300

• UBQ10 Gramene: AT4G05320

• UBQ10 Araport: AT4G05320

• amylopectin CHEBI: CHEBI:28057

• amylose CHEBI: CHEBI:28102

• amylose content Planteome: TO:0000196

• branch AmiGo: PO:0025073

• chlorophyll a CHEBI: CHEBI:18230

• glucose CHEBI: CHEBI:17234

• leaf AmiGo: PO:0025034

• mfp1 Gramene: AT3G16000

• mfp1 Araport: AT3G16000

• oligonucleotide CHEBI: CHEBI:7754

• peptides CHEBI: CHEBI:16670

• proteins CHEBI: CHEBI:36080

• starch CHEBI: CHEBI:28017

• starch content Planteome: TO:0000696