Coalescence and directed anisotropic growth of starch granule initials in subdomains of Arabidopsis thaliana chloroplasts

Living cells orchestrate enzyme activities to produce myriads of biopolymers but cell-biological understanding of such processes is scarce. Starch, a plant biopolymer forming discrete, semi-crystalline granules within plastids, plays a central role in glucose storage, which is fundamental to life. Combining complementary imaging techniques and Arabidopsis genetics we reveal that, in chloroplasts, multiple starch granules initiate in stromal pockets between thylakoid membranes. These initials coalesce, then grow anisotropically to form lenticular granules. The major starch polymer, amylopectin, is synthesized at the granule surface, while the minor amylose component is deposited internally. The non-enzymatic domain of STARCH SYNTHASE 4, which controls the protein’s localization, is required for anisotropic growth. These results present us with a conceptual framework for understanding the biosynthesis of this key nutrient.

They also show that the formation of a starch granule occurs after the coalescence of what they call "granule initials" in WT I do have some comments or suggestions for the authors: 1) I think that it would be interesting for the reader to see the same diagram ( Fig. 4C) applied for the ss4-mutant and the GS and NterSS4-GS expressing lines (Fig. 5).
2) I would appreciate a clear definition of what a "pocket" is in the stroma, where the starch granules are synthesized. To me, the stroma is a continuum inside chloroplasts. In this continuum, thylakoids and grana float without the formation of subsets (pockets?) that would be isolated from others. Reading the manuscript suggests that the stroma is organized in pockets isolated from each other, which is not, I think, real.
3) It is not clear how the starch granules were unquestionably identified during SBF-SEM analysis. Was it done manually or after computational analysis? 4) There are strong differences in signal intensity in some NanoSIMS imaging (Fig 3 for instance). Indeed, the signal observed in Fig. 3B or fig. 3D would have been included in the background signal in Fig. 3A or Fig. 3C. Is there any explanation for that? In addition, could it be possible to produce better quality superimposed images (center of Fig. 3, Fig. 4, Fig. 5) by deleting most of the background of NanoSIMS picture that darkens the picture? This would give a better view of the localization of the observed signal. Fig. 6: a gbss-mutant has been used that shows a reduced incorporation of 13C in the core of the starch granule (Fig. 6B) compared to the wild-type (Fig. 6A). Which kind of gbss-mutant is it? Is the GBSS protein still present within the granule but inactive, or a mutant without GBSS in starch? If the second option is correct, could it be that the reduced signal in the gbss-mutant is related to the lack of the protein within the granule? I do believe that 13C incorporation from 13CO2 is not limited to starch but could also be found in proteins or any other organic compounds that are synthesized during the light phase. 6) Lines 327-329: the authors suggest that, since granule growth is anisotropic and preferentially observed at the equatorial regions of the granule, synthetic enzymes must be concentrated at the same region. This makes sense; however, it could also be that precursor (ADP-glucose) synthesis and/or concentration occurs preferably at the same region through a still not uncovered mechanism, thus privileging granule growth in specific direction. 7) Lines 338: I understand the reference to fibrillins (FBN1a and FBN1b) that were shown to interact with SS4. However, none of the corresponding mutant display a phenotype that is related to that of the ss4-mutant. Thus, the actual implication of FBNs in the process of starch granule initiation is questioned. 8) A comparative analysis was performed between Arabidopsis wild-type and mutant lines for SS4, which is known to affect the initiation process of starch synthesis in vivo by significantly reducing the number of granules per chloroplast and modifying their morphology. I think it would have been worth it to include a line with a phenotype opposite to that of the ss4-mutant. Indeed, the isa1-mutant would have been interesting. In that mutant, starch synthesis is altered but not stopped. Numerous small granules are synthesized in the isa1-mutant that could, somehow, correspond to the "granule initials" suggested by the authors. 9) Lines 310-319. I don't exactly understand the demonstration made here. What would be "opposing orientations of lamellae"? Why the lamellae would be unlikely radially oriented in granules initials? Do the authors have any evidence for that? By the way, if granule formation results from the merging of several "granule initials", starch granules would then consist of several hila. I don't remember that Arabidopsis starch granules have been described containing several hila. Moreover, how does the growth ring would be organized in that case? Starch growth rings are organized around the hilum of the granule as already described elsewhere. This raises the question of how growth rings will organize if several "granule initials" merge to form one bigger granule. One last thing, which is about semantic: to my point of view, the 9-10 nm lamellae are not "radially arranged in starch". This is the molecules inside the 9-10 nm lamellae that are arranged radially (the glucans are more or less oriented parallel to the radius of the starch granule). The lamellae are arranged tangentially to the surface of the granule (i.e. they are placed concentrically from the center (hilum) of the granule). Christophe D'HULST.

5)
Reviewer #2: Remarks to the Author: The manuscript is well structured and presented, and I have a few minor suggestions: more data is needed for explanation of growing conditions, eg. how the authors have provided that test plants were not exposed to water/nutrient stress? Also, more information is needed for plant material sampling (eg. leaf from which position) and handling prior analyses line 425, reference stile is incorrect.
Reviewer #3: Remarks to the Author: The current study is firmly based on innovative imaging techniques to provide cell-biological insights into starch granule formation and growth.
By applying SBF-SEM and nanoSIMS to Arabidopsis wt and mutants, it is convincingly shown that multiple starch granules are initiated in stromal cavities, and subsequently these initial structures fuse and grow anisotropically and expand equatorially until reaching their final size and shape. The use of selected starch synthesis mutants, in particular ss4, indicates that the non-enzymatic domain of STARCH SYNTHASE 4 is vital for anisotropic growth.
The imaging is conducted in an excellent manner and the results are of very high quality and lend unequivocal support to the conclusions that are put forward. One aspect that could perhaps be made a bit clearer is the time frame of starch granule formation. In figure 3D for example, only one starch grain is labeled, while the others are not. In this context the chase of 4 hours could perhaps be set in proportion to the duration of granule formation. Figure 1F would perhaps benefit from an inset with higher magnification.

RESPONSE TO REVIEWER COMMENTS
Reviewer #1:

RESPONSE: We very much appreciate the positive and detailed comments from Reviewer 1 and address the specific comments below:
1) I think that it would be interesting for the reader to see the same diagram (Fig. 4C) applied for the ss4-mutant and the GS and NterSS4-GS expressing lines (Fig. 5).

RESPONSE: We agree and provide these plots as a new Supplemental Fig. S6. The plots reinforce the conclusions of our manuscript -that there is a regular pattern of anisotropic growth, which is lost or becomes irregular in the absence of STARCH SYNTHASE 4, but regained in the NSS4-GS lines.
2) I would appreciate a clear definition of what a "pocket" is in the stroma, where the starch granules are synthesized. To me, the stroma is a continuum inside chloroplasts. In this continuum, thylakoids and grana float without the formation of subsets (pockets?) that would be isolated from others. Reading the manuscript suggests that the stroma is organized in pockets isolated from each other, which is not, I think, real.

RESPONSE: we understand the point of the reviewer and have tried to define better in the text what we mean by a 'stromal pocket', i.e. a defined volume of the stroma in which we observed numerous starch granule initials forming in close proximity (Lines 99-101). We also altered the discussion (Lines 305-311) to note that if our speculation about phase separation events is correct, this could result in the formation of real membrane-less subdomains within the stroma (Line 368).
3) It is not clear how the starch granules were unquestionably identified during SBF-SEM analysis. Was it done manually or after computational analysis?

RESPONSE: all granule identification was done manually. This is noted in the revised text (Lines 89-90).
4) There are strong differences in signal intensity in some NanoSIMS imaging (Fig 3 for instance). Indeed, the signal observed in Fig. 3B or fig. 3D would have been included in the background signal in Fig. 3A or Fig. 3C. Is there any explanation for that? In addition, could it be possible to produce better quality superimposed images (center of Fig. 3, Fig. 4, Fig. 5) by deleting most of the background of NanoSIMS picture that darkens the picture? This would give a better view of the localization of the observed signal. Lines 457-461). However, we prefer not to delete the background from the NanoSIMS images and hope that the presentation of the two images and the merge is sufficient for readers to see the enrichment pattern and its overlap with the cellular ultrastructure. Fig. 6: a gbss-mutant has been used that shows a reduced incorporation of 13 C in the core of the starch granule (Fig. 6B) compared to the wild-type (Fig. 6A). Which kind of gbss-mutant is it? Is the GBSS protein still present within the granule but inactive, or a mutant without GBSS in starch? If the second option is correct, could it be that the reduced signal in the gbss-mutant is related to the lack of the protein within the granule? I do believe that 13 C incorporation from 13 CO2 is not limited to starch but could also be found in proteins or any other organic compounds that are synthesized during the light phase.

RESPONSE: The GBSS mutant used in this study is that described in Seung et al. (2015), as stated in the methods (Line 389) which lacks the GBSS protein (now stated (Line 273). The reviewer is right that 13 C will eventually be found in proteins and other cellular structures synthesized during the light phase. However, we believe that the labelling experiment is too short for this to occur to any significant extent. The 13 CO2 supplied would need to be fixed via photosynthesis, make its way through metabolism to significantly enrich free amino acids in the cytosol. These would need to be incorporated into GBSS pre-proteins, re-imported into chloroplasts, processed, folded, and be targeted to -and enriched on -the starch granule. We consider it unlikely that this would produce a detectable NanoSIMS signal during the 60-minute experiment. Furthermore, even if detectable amounts of 13 C-labbeled GBSS protein were produced and targeted to the granule during the experiment, it would presumably bind to the granule surface and contribute to the surface signal rather than appearing in the granule core, since only small molecules like ADPGlc can diffuse into the semi-crystalline amylopectin matrix. GBSS present in the cores of wild-type granules would have been made and incorporated into the granule prior to labelling. Hence, we do not believe that the label detected inside wild-type granules (but not in the gbss mutant) can be attributed to the GBSS protein itself.
6) Lines 327-329: the authors suggest that, since granule growth is anisotropic and preferentially observed at the equatorial regions of the granule, synthetic enzymes must be concentrated at the same region. This makes sense; however, it could also be that precursor (ADP-glucose) synthesis and/or concentration occurs preferably at the same region through a still not uncovered mechanism, thus privileging granule growth in specific direction.

RESPONSE: we appreciate this interesting and valid suggestion, which complementary and not mutually exclusive to our suggestions. We mention this option in the revised manuscript (Lines 341-342).
7) Lines 338: I understand the reference to fibrillins (FBN1a and FBN1b) that were shown to interact with SS4. However, none of the corresponding mutant display a phenotype that is related to that of the ss4-mutant. Thus, the actual implication of FBNs in the process of starch granule initiation is questioned.

RESPONSE: we agree with the reviewer on this point. While the previously published proteinprotein interaction data seem valid, the lack of phenotypic impact upon mutating both FBN1a and
FBN1b fibrillins calls the biological significance into question. We also mention in a prior publication that we could not reproduce this interaction. We have consequently de-emphasized this point by removing one reference to the fibrillins, without removing the concept altogether. 8) A comparative analysis was performed between Arabidopsis wild-type and mutant lines for SS4, which is known to affect the initiation process of starch synthesis in vivo by significantly reducing the number of granules per chloroplast and modifying their morphology. I think it would have been worth it to include a line with a phenotype opposite to that of the ss4-mutant. Indeed, the isa1mutant would have been interesting. In that mutant, starch synthesis is altered but not stopped. Numerous small granules are synthesized in the isa1-mutant that could, somehow, correspond to the "granule initials" suggested by the authors.

RESPONSE: The reviewer suggests an interesting experiment. Indeed, we are pursuing further research in this direction, including mutants which show alterations in starch granule initiation frequency (e.g. ptst2 and the corresponding PTST2 overexpressing lines). The isa1 mutant would make an interesting addition. However, we respectfully suggest that this would better contribute to a second manuscript where we explore a number of recently described genetic factors in parallel.
9) Lines 310-319. I don't exactly understand the demonstration made here. What would be "opposing orientations of lamellae"? Why the lamellae would be unlikely radially oriented in granules initials? Do the authors have any evidence for that? By the way, if granule formation results from the merging of several "granule initials", starch granules would then consist of several hila. I don't remember that Arabidopsis starch granules have been described containing several hila. Moreover, how does the growth ring would be organized in that case? Starch growth rings are organized around the hilum of the granule as already described elsewhere. This raises the question of how growth rings will organize if several "granule initials" merge to form one bigger granule. One last thing, which is about semantic: to my point of view, the 9-10 nm lamellae are not "radially arranged in starch". This is the molecules inside the 9-10 nm lamellae that are arranged radially (the glucans are more or less oriented parallel to the radius of the starch granule). The lamellae are arranged tangentially to the surface of the granule (i.e. they are placed concentrically from the center (hilum) of the granule). Lines 322-337). We also appreciate the correction about it being the amylopectin molecules -not the lamellae -that are radially arranged, which was incorrectly phrased at one point.