Domestication in Real Time: The Curious Case of a Trigenomic Sunflower Population

Round 1

Reviewer 1 Report

The manuscript addresses ongoing efforts to domesticate perennial herbaceous species, which provide both an economically viable product and ecological services such as conservation of topsoil. In this study, authors crossed annual, diploid Helianthus annuus common sunflower, with perennial, hexaploid H. tuberosus, a tuber-producing wild species. Helianthus tuberosus is the product of a complex hybridization history involving H. divaricatus and H. grosseserratus. Following crosses, plants underwent six generations of selection in an attempt to generate an “ideotype” which displays large head size, seed size, apical dominance, and tuber production. Genome size was also selected for with the goal of achieving genomic stability and fertility.

This manuscript is an important empirical contribution to the growing body of literature describing ongoing attempts to domesticate perennial, herbaceous species. The approach taken here involves phenotyping of an intermated interspecific hybridization derived from perennial and annual parents. Authors present data on genome size, pollen size and viability, seed dimensions, seed number, head number, head size, and tuber presence (scored on a scale of 0-5). The authors report a positive correlation between seed size and number, but overall low fertility in hybrids. Selection resulted in increases in seed size and apical dominance over generations; however, male fertility declined across generations, as did perenniality (tuber production).

The manuscript is generally well-written but could be strengthened with a few, relatively minor tweaks. First, key details are lacking with regards to the number and generation of individuals sampled for each trait. More detailed descriptions are required in the methods (e.g., lines 91 – 96). Second, more detailed introduction and discussion of what is apparently the primary perennating structure of Helianthus spp. Is needed. The authors equate tuber production with perenniality, but there are many perennial, herbaceous species that do not produce tubers that are also perennial. For readers less familiar with this system, clarification regarding what structures constitutes perenniality in this taxon is needed. Do hybrids that lack perennial structures have the capacity to regrow? Has this been tested? Please address this. Finally, the authors allude to the presence of extensive phenotypic variation in hybrid individual, but do not spend much time in the manuscript presenting or discussing this. I would like to see this variation highlighted in figures, and also would love to see a focus on the outliers. For example, It is interesting that tuber production decreases in some individuals over generations but increases in others (Figure 3C).

Specific comments:

Line 57: H. tuberosus is misspelled (shown as H. tuberosis). This happens throughout the manuscript.

Line 90: define IM (presumably intermated) the first time it is used

Lines 91 – 96: how many and what plants were used in this study? What generation were they? More details are needed.

Line 167: What is the range of variation in fertility exhibited by hybrids? Is it possibly to show this in a figure?

Line 183: description of the plants actually used in this study is needed. How many of each population were used? What age were the plants at the time of phenotyping? What was the rationale for the selection? I recommend adding a separate subheading, beyond the population development, that describes in detail individuals and quantitites of individuals surveyed here.

Figure 1: Pedigree information apparently summarizes work in previously published papers; having said that, the figure as presented here leaves out key information for readers who may not be familiar with previously work. What does RM mean? What happened when one line split into two? How many plants currently exist in each line, and how long have these plants been living? I recommend either adding much more detail to the explanation -  which could go in the caption – or simplifying or removing the figure.

Line 112: Which plants, and how many, and from which generation, were plants phenotyped? A table describing traits measured, and numbers of individuals screened for each trait (if different) would be helpful.

Line 129: I am curious about equating perenniality with tuber presence. In this population, do all tubers sprout in the subsequent year? Are there any other perennating structures in this system (e.g., persistent root systems)? Also curious about the scale of perenniality (0 – 5). Does this number reflect the number or size of tubers? Please explain the quantitative nature of this measure. Its not clear what attributes “resemble H. tuberosus”.

Line 131: Are there any other perennial organs besides tubers in H. tuberosus? Is there any evidence that non-tuber bearing progeny are perennial? Has this been quantified? More information about the basic biology, morphology, and lifespan of the parents and progeny are needed.

Line 132: The paper would be strengthened with the addition of mixed models that explain observed patterns of variation. Also, it would be interesting to consider patterns of covariation among traits.

Lines 148 – 156: Description of correlations/covariation analyses should go in methods.

Lines 179 – 192: Relationships among morphological features and fertility are an important component of this paper. The paper would be much stronger if these results, including range of variation in traits, regressions among traits, and correlation analyses, were shown in graphs, either in the main manuscript or supplement information. I recommend de-emphasizing pollen staining, given the statement “that this imprecise measure of pollen viability attenuated this signal” (line 177) and “limited reliability of pollen staining as a measure of pollen viability” , and focusing instead on variation in female fertility and tuber production, and the relationships between these features.

Line 190: “No other morphological traits had significant relationships with male or female fertility.” It would be helpful to remind the reader here what those traits are. Also I recommend adding this table to the main body of the manuscript, as the lack of significant relationships between morphology and fertility is important for understanding how selection for increased yield might impact other parts of the plant.  

Line 194: Replace “this program” with the specific name of the population. E.g., “The H. annuus x H. tuberosus breeding population has been used to select for classical domestication syndrome traits such as….”

Lines 205 – 208: Previous comments by the authors and pollen staining data in earlier sections (and see above for comments on this) indicate that pollen staining may not be reliable. Given this, some explanation and/or justification for the use of pollen staining, and guidance for interpreting the results, is warranted here and in other sections of the manuscript.

Line 209 – 213: To what extent are trends in “perenniality” shaped by the way in which perenniality is measured? As raised previously, are tuber the only mechanism of perenniality in these species, or are non-storage roots perennial too? Anotherr potentially valuable metric might be the comparison of the number of non-tuber bearing individuals over time. Also curious – are there multiple trends apparent in Figure 3C? It looks like some individuals may be increasing in their perenniality measure, whereas others may be moving towards a lack of tubers all together. For breeding purposes, the outliers (here, the individuals that are increasing in perenniality over time) may matter more than the overall population trend. Would be interesting and important to dissect this further and discuss implications for selection of perennial, herbaceous crops.

Figure 4: It is difficult to see the numbers in 1A. Also, it is not clear from these pictures what specific traits are being scored as perennial. What is a perennatingAs raised in a previous comment, it is not clear what constitutes a perennating structures are in this species. Have they been tested?

Line 244-245: “though deviations from tetraploidy show an increase across generations…which is to be expected given the genomic background of our population.” Figure 5  - what is going on in Generation 7? It looks like all but two of the tetraploids disappear and instead reduct to triploids. Is this correct? What causes this, and  why in generation 7 after relative stability in generations 4, 5, and 6? Finally, additional discussion describing the relationship between fertility and tuber production would strengthen this manuscript. The authors note abundant phenotypic variation (line 314) but this is not explored fully. Additional interrogation and interpretation of the observed variation, with a special focus on the outliers, would be interesting to see in this manuscript. Another area of potential expansion is the apparent bimodal trend of tuber production increase/decrease in later generations (Fig. 3C). Additional comments are listed below.

Additional comments.

Lines 250 – 257: Add a figure describing this result.

Lin 312 – 315: “Though the apparent antagonism between desired above- and below-ground morphologies has been an obstacle to achieving the ideoype, a closer look at the data reveal abundant phenotypic variation. This means that we can likely continue to make progress, for example, by choosing plants with larger seeds and high seed set…while making only modest concessions in other traits like perenniality.” This statement seems to contradict the data presented here: if high seed set is negatively correlated with tuber production, then it seems selection for increased seed set would lead to concomitant decrease in seed set.

Author Response

Perenniality

Comments:  Second, more detailed introduction and discussion of what is apparently the primary perennating structure of Helianthus spp. Is needed. The authors equate tuber production with perenniality, but there are many perennial, herbaceous species that do not produce tubers that are also perennial. For readers less familiar with this system, clarification regarding what structures constitutes perenniality in this taxon is needed. Do hybrids that lack perennial structures have the capacity to regrow? Has this been tested? Please address this.

Line 129: I am curious about equating perenniality with tuber presence. In this population, do all tubers sprout in the subsequent year? Are there any other perennating structures in this system (e.g., persistent root systems)? Also curious about the scale of perenniality (0 – 5). Does this number reflect the number or size of tubers? Please explain the quantitative nature of this measure. It’s not clear what attributes “resemble H. tuberosus”.

Line 131: Are there any other perennial organs besides tubers in H. tuberosus? Is there any evidence that non-tuber bearing progeny are perennial? Has this been quantified? More information about the basic biology, morphology, and lifespan of the parents and progeny are needed.

Line 209 – 213: To what extent are trends in “perenniality” shaped by the way in which perenniality is measured? As raised previously, are tuber the only mechanism of perenniality in these species, or are non-storage roots perennial too? Anotherr potentially valuable metric might be the comparison of the number of non-tuber bearing individuals over time. Also curious – are there multiple trends apparent in Figure 3C? It looks like some individuals may be increasing in their perenniality measure, whereas others may be moving towards a lack of tubers all together. For breeding purposes, the outliers (here, the individuals that are increasing in perenniality over time) may matter more than the overall population trend. Would be interesting and important to dissect this further and discuss implications for selection of perennial, herbaceous crops.

Figure 4: It is difficult to see the numbers in 1A. Also, it is not clear from these pictures what specific traits are being scored as perennial. What is a perennatingAs raised in a previous comment, it is not clear what constitutes a perennating structures are in this species. Have they been tested?

Response: Within Helinathus, plants are perennial due to both the formation of rhizomes and tubers. Within Helinahthus tuberosus, the key perennial organ is the production of tubers. Root mass is much denser in H. tuberosus and interspecific progeny which present both tubers and rhizomes. In interspecific hybrids between H. annuus and H. tuberosus, tuber formation is necessary for perenniality, but not sufficient to make the plant perennial—the plant needs to both form a perennial organ and be cold tolerant (Kantar et al. 2014; Kantar et al., 2018). Previous studies have explored the ability to regenerate different tubers and tuber parts (Kantar et al., 2012; Bock et al., 2018) and there have been extensive efforts to phenotype the underground biomass of these populations (Kantar et al., 2014,2018). However, over the 17 years these populations have been grown there is no evidence of perenniality in non-tuber bearing plants. Tuber production is an important perenniality trait that segregates in the intermated populations. Tuber-producing plants are generally harvested at physiological maturity (e.g., R-9 for sunflower, usually recognized based on the production of brittle and brown phyllaries).

We have added more information about the perennial index to Table 1, and we have added more information about the importance of tuber tolerance and cold-hardiness to lines 59-62.

Depending on the population of H. tuberosus (invasive or domestic) there is large variation in tuber number and size (Bock et al., 2018), and we agree that outliers are very important, so we have added details in the figure 3 caption about potential invasiveness and the potential utility of outlier individuals.

Since the background in Figure 4A makes it very difficult to see any colors in the foreground, we have moved the index labels just outside the frame of each picture. To orient the reader and help keep the images clean, we have also added a more helpful description of tubers in the legend.

Bock, DG, Kantar MB, Caseys, C, Matthey-Doret, R, Rieseberg, LH. 2018. Evolution of invasiveness by genetic accommodation. Nature Ecology & Evolution. doi:10.1038/s41559-018-0553-z

Sample sizes

Comments:

First, key details are lacking with regards to the number and generation of individuals sampled for each trait. More detailed descriptions are required in the methods (e.g., lines 91 – 96).

Lines 91 – 96: how many and what plants were used in this study? What generation were they? More details are needed.

Line 183: description of the plants actually used in this study is needed. How many of each population were used? What age were the plants at the time of phenotyping? What was the rationale for the selection? I recommend adding a separate subheading, beyond the population development, that describes in detail individuals and quantities of individuals surveyed here.

Line 112: Which plants, and how many, and from which generation, were plants phenotyped? A table describing traits measured, and numbers of individuals screened for each trait (if different) would be helpful.

Line 90: define IM (presumably intermated) the first time it is used

Response: We have added the sample sizes for each pooled ID and a more descriptive legend to Figure 1, and we have added a Table 1 which displays the sample sizes for each trait for each generation.

We have clarified which plants were saved and which plants were selected for breeding in lines 100-104. We define intermated in line 96 address phenotyping timing information in lines 137-143.

Variation

Comment: Finally, the authors allude to the presence of extensive phenotypic variation in hybrid individuals, but do not spend much time in the manuscript presenting or discussing this. I would like to see this variation highlighted in figures, and also would love to see a focus on the outliers. For example, it is interesting that tuber production decreases in some individuals over generations but increases in others (Figure 3C).

Line 167: What is the range of variation in fertility exhibited by hybrids? Is it possibly to show this in a figure?

Lines 179 – 192: Relationships among morphological features and fertility are an important component of this paper. The paper would be much stronger if these results, including range of variation in traits, regressions among traits, and correlation analyses, were shown in graphs, either in the main manuscript or supplement information.

Response: We have rewritten the figure 3 caption to talk about outlier individuals and potential that these are transgressive segregates, and directed the reader to the scatter along ideotype regressions in line 360.

The seed set range is listed, and we have added the ranges to pollen fertility metrics (Line 192), as well as the ranges of morphological traits (lines 237-239, 260-261). The individual values in Fig 2B-D are intended to show the variation in each fertility trait using transformed data.

Other comments

Comment: The manuscript is generally well-written but could be strengthened with a few, relatively minor tweaks.

Response: Thank you.

Comment: Line 57: H. tuberosus is misspelled (shown as H. tuberosis). This happens throughout the manuscript.

Response: These have been changed to tuberosus.

Comment: Figure 1: Pedigree information apparently summarizes work in previously published papers; having said that, the figure as presented here leaves out key information for readers who may not be familiar with previously work. What does RM mean? What happened when one line split into two? How many plants currently exist in each line, and how long have these plants been living? I recommend either adding much more detail to the explanation -  which could go in the caption – or simplifying or removing the figure.

Response: We have added sample size information to the figure and changed the caption to be more descriptive “Pedigree of maternal lines continued in the intermated (IM) generations. Parents and F₁ populations were previously phenotyped for perenniality and domestication syndrome traits including seed size, branching, pollen fertility, head diameter, and number of heads [40]. Initial populations were selected based on a domestication syndrome trait index developed to move the population toward the perennial sunflower ideotype [37]. For the first three generations this index was used to select the top twenty percent of the maternal half sib families for inclusion in the next generation of plant material. After the IM₃F₁ generation, selection was based on theoretical fertility, as seed production was the limiting factor in the ability to move toward commercialization. The individuals in the IM₃F₁ generation were designated random mated (RM) as all previous selected lines were grown in the same field and pollen control was not used. Downstream progeny are given pooled identifiers here based on the year they were produced, the breeder’s ID, and their maternal line. Each pooled ID is connected to its maternal ID (individual or pooled) by yellow lines and situated under the corresponding pollen donor cohort. Sample sizes are provided for each pooled ID. We note the four individuals with unknown maternity (i.e. seeds that escaped their individual envelope during collection).”

Comment: Line 132: The paper would be strengthened with the addition of mixed models that explain observed patterns of variation. Also, it would be interesting to consider patterns of covariation among traits.

Response: We agree that a mixed model could be more appropriate, but recognize that this introduces a cost of complexity in analysis, and interpretation, and therefore present the simplified (but imperfect) linear model. See lines 182-185

Comment: Lines 148 – 156: Description of correlations/covariation analyses should go in methods.

Response: We moved this to the methods (lines 169-171).

Comment: I recommend de-emphasizing pollen staining, given the statement “that this imprecise measure of pollen viability attenuated this signal” (line 177) and “limited reliability of pollen staining as a measure of pollen viability”, and focusing instead on variation in female fertility and tuber production, and the relationships between these features.

Lines 205 – 208: Previous comments by the authors and pollen staining data in earlier sections (and see above for comments on this) indicate that pollen staining may not be reliable. Given this, some explanation and/or justification for the use of pollen staining, and guidance for interpreting the results, is warranted here and in other sections of the manuscript.

Response: As the reviewer noticed, we attempt to de-emphasize pollen staining in our analyses, but wish to honestly present out results, as anomalous as they may seem. Pollen staining is a reliable way to characterize pollen viability in other (less complicated) systems, and provides a link to previous work on interspecific Helianthus populations as well as a way to understand how this interspecific population compares to its progenitor species. We note this in lines 152-153.

Comment: Line 190: “No other morphological traits had significant relationships with male or female fertility.” It would be helpful to remind the reader here what those traits are. Also I recommend adding this table to the main body of the manuscript, as the lack of significant relationships between morphology and fertility is important for understanding how selection for increased yield might impact other parts of the plant. 

Response: We have added a phenotyping Table 1 to explain the traits explored.

Comment: Line 194: Replace “this program” with the specific name of the population. E.g., “The H. annuus x H. tuberosus breeding population has been used to select for classical domestication syndrome traits such as….”

Response: This was changed.

Comment: Line 244-245: “though deviations from tetraploidy show an increase across generations…which is to be expected given the genomic background of our population.” Figure 5  - what is going on in Generation 7? It looks like all but two of the tetraploids disappear and instead reduct to triploids. Is this correct? What causes this, and why in generation 7 after relative stability in generations 4, 5, and 6? Finally, additional discussion describing the relationship between fertility and tuber production would strengthen this manuscript. The authors note abundant phenotypic variation (line 314) but this is not explored fully. Additional interrogation and interpretation of the observed variation, with a special focus on the outliers, would be interesting to see in this manuscript. Another area of potential expansion is the apparent bimodal trend of tuber production increase/decrease in later generations (Fig. 3C).

Response: Pollen contamination is a possibility, and we have indicated this in line 189. We attempt to show the observed variation by plotting individual data points in Figures 2-4. We now direct the viewer to this aspect of the figure (line 360). We dedicate a paragraph to the relationship between fertility and tuber production (lines 211-221) and include the statistics in a supplemental table. We believe this is a sufficient amount of attention for a relationship that is weak at best. In lines 356-357, we note that plants with lower perennial indices are not necessarily doomed to winter-kill. We further explain in the new Table 1 that any plant with a perennial index of one or more has demonstrated winter-hardiness.

Comment:  Lines 250 – 257: Add a figure describing this result.

Response: Figure five is intended to illustrate this information.

Comment: Line 312 – 315: “Though the apparent antagonism between desired above- and below-ground morphologies has been an obstacle to achieving the ideoype, a closer look at the data reveal abundant phenotypic variation. This means that we can likely continue to make progress, for example, by choosing plants with larger seeds and high seed set…while making only modest concessions in other traits like perenniality.” This statement seems to contradict the data presented here: if high seed set is negatively correlated with tuber production, then it seems selection for increased seed set would lead to concomitant decrease in seed set.

Response:  The negative relationship between fertility and tuber production is weak, and outliers abound. Thus, since we can find outlier individuals that clonally propagate we may be able to leverage population variability. We hope that the aforementioned increased attention to variation will help the reader see how progress can be made by selecting on the variation, and the outliers in particular.

Reviewer 2 Report

This paper investigates genome stability and trait correlations in later hybrid generations of intermated annual x perennial sunflower as part of a plant breeding program to combine the most desirable traits of both parents. The study is largely descriptive but it identifies interesting limitations and possibilities in the breeding material and interprets these in terms of inherent meiotic incompatibilities associated with this wide cross. Some future approaches are recommended in the light of the presented results.
In the introduction, I found some terms "ecosystem services", "natural resources", "conservation practice" to be used in a vague sense without linking them enough to the particular factors associated with this study of breeding for novel combinations of crop traits. The lettered genomes should be more clearly linked to their progenitor species as this also forms an important part of the discussion later. The aims of the study are clearly presented in the final paragraph of the introduction. The selection regime is less clearly explained. At different points in the paper, there is mention of maintaining trait variation and at other points, selection for desired traits. The methods only mention selection for perenniality explicitly and one could also assume implicit selection for increased fertility. If other traits were not directly selected for, the results and discussion need to acknowledge that these are spontaneous trends associated with genome stabilization or linkage to the selected traits. The discussion provides good interpretation of the results in terms of genomic interactions and references the wider literature in other crops, raising its general interest. I recommend including some short discussion of fertility issues related to the endosperm as this a second tissue requiring fertilization and development for successful seed production. The supplementary data and tables are commendably comprehensive.
This paper would make an interesting contribution to the plant breeding literature involving changes in ploidy following these general revisions and the additional minor specific changes listed below.

Specific comments
L17 and L34 "ecosystem services" is too vague. What benefits in particular are possible?
L35-36 "natural resources" is vague. What particular resources could be preserved?
L36 and L41 "conservation agricultural practices" What is being conserved? I'm more used to the term "sustainable" in this context.
L67 "e.g." should probably be replaced with "previously called" or similar if the intention is to notify the reader of an alternative name for this species.
L70 Specify which genome letter stands for which species. This could be done by listing the letter with species names in the preceding sentences.
L88 Wild collected from what region?
Figure 1 explains well the complex intermatings that were performed each year as well as the generation stage of each tested line. Would it be possible to include numbers of progeny per line to this figure also as this information is currently missing from the methods?
L126-127 Mention if flower heads were all manually crossed or allowed to open pollinate.
L138-139 The preceding methods sections did not discuss selection other than for perenniality. Make this point again here. If not directly selected, I would therefore expect other traits to trend towards H. tuberosis or no change.
L141-142 Not clear how this correction was performed as part of a regression analysis. Did you use residuals here? Later explained in the results at L172-173 and L183-184. Move this explanation to the methods.
L186-188 You could also mention the issue of multiple testing increasing the possibility of some false positive results.
L194-195 These selection activities need to be described in the methods. What were the selection thresholds?
L197 What "other traits" are you referring to here?
L231 How could backcrosses to H. annuus have occurred if controlled crosses were being performed?
Figure 4A Add arrows to the figure to highlight relevant features.
L276 Again it would be useful for the letters of these genomes to be clearly listed in the introduction as a reference for the reader.
L314-316 Perhaps some short discussion of the observation of differences between family lines would be useful as an additional selective approach here.
P324-325 I disagree with synthetic polyploidy inevitably causing a bottleneck. It can be avoided if multiple genetically diverse individuals are made octoploid and later intermated.
L327-329 Rather vague. What genomic signals would indicate that proper meiotic pairing has taken place?
L329-330 Summarize again the subset of most promising targets for selection that you identify in this study.

Author Response

Comments: I found some terms "ecosystem services", "natural resources", "conservation practice" to be used in a vague sense without linking them enough to the particular factors associated with this study of breeding for novel combinations of crop traits.

L17 and L34 "ecosystem services" is too vague. What benefits in particular are possible?

L35-36 "natural resources" is vague. What particular resources could be preserved?

L36 and L41 "conservation agricultural practices" What is being conserved? I'm more used to the term "sustainable" in this context.

Response: We have substituted the term “continuous ground-cover” (Line 17) for ecosystem services in the abstract, and use “soil and water” (lines 35, 37) instead of “natural resources”—this is more precise. And use “sustainable” instead of “conservation”—the reviewer suggested a better phrasing that is less “buzzy.”

Comments: The lettered genomes should be more clearly linked to their progenitor species as this also forms an important part of the discussion later.

L70 Specify which genome letter stands for which species. This could be done by listing the letter with species names in the preceding sentences.

L276 Again it would be useful for the letters of these genomes to be clearly listed in the introduction as a reference for the reader.

Response: Thank you for pointing this out. This has been addressed in line 72-74, and we reiterate the labels to assist the reader later in lines 317-322.

Comment: At different points in the paper, there is mention of maintaining trait variation and at other points, selection for desired traits. The methods only mention selection for perenniality explicitly and one could also assume implicit selection for increased fertility. If other traits were not directly selected for, the results and discussion need to acknowledge that these are spontaneous trends associated with genome stabilization or linkage to the selected traits.

Response: We have added a sentence about selecting for the ideotype in lines 102-104.

Comment: The discussion provides good interpretation of the results in terms of genomic interactions and references the wider literature in other crops, raising its general interest. I recommend including some short discussion of fertility issues related to the endosperm as this a second tissue requiring fertilization and development for successful seed production.

Response: Since the endosperm number should be balanced (theoretically, at least, using parents of equal ploidies), we posit that the primary barrier occurs before endosperm development. We base this on observed meiotic anomalies that have been reported by several previous studies [e.g. refs 31,35,36], and the highly irregular pollen observed in this study.

Comment: The supplementary data and tables are commendably comprehensive. This paper would make an interesting contribution to the plant breeding literature involving changes in ploidy following these general revisions and the additional minor specific changes listed below.

Response: Thank you.

Comment: L67 "e.g." should probably be replaced with "previously called" or similar if the intention is to notify the reader of an alternative name for this species.

Response: We removed mention of H. hirsutus, as this detail may be too distracting.

Comment: L88 Wild collected from what region?

Response: The collected H. tuberosus were collected within 5 miles of the research field, now clarified in line 94.

Comment: Figure 1 explains well the complex intermatings that were performed each year as well as the generation stage of each tested line. Would it be possible to include numbers of progeny per line to this figure also as this information is currently missing from the methods?

Response: We’ve added the sample sizes to the figure.

Comment: L126-127 Mention if flower heads were all manually crossed or allowed to open pollinate.

Response: Lines 96-97 explain that crosses largely involve the mixed pollen of verified tetraploid plants with desirable phenotypes.

Comments: L138-139 The preceding methods sections did not discuss selection other than for perenniality. Make this point again here. If not directly selected, I would therefore expect other traits to trend towards H. tuberosus or no change.

L194-195 These selection activities need to be described in the methods. What were the selection thresholds?

Response: We have added a sentence about selection in lines 102-104.

Comment: L141-142 Not clear how this correction was performed as part of a regression analysis. Did you use residuals here? Later explained in the results at L172-173 and L183-184. Move this explanation to the methods.

Response: We moved this to the methods (lines 170-171).

Comment: L186-188 You could also mention the issue of multiple testing increasing the possibility of some false positive results.

Response: We added this caveat (lines 182-185).

Comment: L197 What "other traits" are you referring to here?

Response: We added a table 1 with all the morphological traits. These are also shown in the supplemental analyses referenced.

Comment: L231 How could backcrosses to H. annuus have occurred if controlled crosses were being performed?

Response: This is just a possible explanation for the presence of diploid annual- behaving plants we obtained (and eliminated from our dataset). Although we make every effort to exclude pollinators, we can only be so certain that a tiny pollinator didn’t infiltrate our bags.

Comment: Figure 4A Add arrows to the figure to highlight relevant features.

Response: We have added a better legend to keep the figure as tidy as possible.

Comment: L314-316 Perhaps some short discussion of the observation of differences between family lines would be useful as an additional selective approach here.

Response: We have added the sentence, “There is variance in individuals and across maternal lines, this implies that stronger pollen control may lead to better response to selection.” Line361.

Comment: P324-325 I disagree with synthetic polyploidy inevitably causing a bottleneck. It can be avoided if multiple genetically diverse individuals are made octoploid and later intermated.

Response: We agree with this point in principle, but due to the small number of resources, it would not be possible to create enough successful individuals to have a large effective population size in the synthetic polyploidy population. Nonetheless, we have softened this statement in the text (line 374).

Comment: L327-329 Rather vague. What genomic signals would indicate that proper meiotic pairing has taken place?

Response: We would like to know of any chromosome loss, structural variation, or if recombination occurs at all. We suspect that “normal” recombination is anomalous in these plants, but would like to know how this differs among lines and individuals. We have added an example of the questions we could answer in lines 348-349.

Comment: L329-330 Summarize again the subset of most promising targets for selection that you identify in this study.

Response: We have added in the specific ideotype traits to this section (line 358).

Reviewer 3 Report

This paper is quite interesting, and mostly well written and explained, despite the complications of the system and the topic. 

My main concerns are with the way the authors deal with some of the results and discussion. To me it looks like the effort to develop a perrenial seed crop using this approach worked quite well until generations 7-8. There was good progress towards combining the seed traits with perenniality, and little change in genome size / ploidy from generations 3-6. However, in generation 7 there was a sudden and unexplained devation from that n=4 ploidy, becoming for some reason much more variable. Perhaps related to this, there was a complete loss of perenniality in generation 8. Neither of these issues is explained, and it is not clear why these sudden changes happened only in these last 2 generations. 

Possible explanations for these changes should be explored, and also addressed in terms of how this system can be developed further. Was there some mistake made in one of these generations? A mix-up of seed or pollen, or accidental interbeeding wth something else? Or what happened to cause these last 2 generations to be so different?

Whatever the explanation, it seems to me that once perenniality is lost as a trait in generation 8, selection will not be able to easily recover it using that plant material. Since that trait was the entire purpose of doing this cross, it seems strange to say things like the last statement before the conclusions,  (lines 314-316)

"we can likely continue to make progress, for  example, by choosing plants with large seeds and high seed set (which have a positive relationship)  while making only modest concessions in other traits like perenniality"

It makes it sound like the complete loss of perenniality in generation 8 is just a "modest concession", though the authors likely don't mean that.

How can this actually be reversed? The authors need to address that. Perhaps by going back to genration 6 or 7? Or making crosses among generations, maybe? I'm not sure, but something should be proposed rather than ignoring the problem. 

Author Response

Comment: My main concerns are with the way the authors deal with some of the results and discussion. To me it looks like the effort to develop a perennial seed crop using this approach worked quite well until generations 7-8. There was good progress towards combining the seed traits with perenniality, and little change in genome size / ploidy from generations 3-6. However, in generation 7 there was a sudden and unexplained devation from that n=4 ploidy, becoming for some reason much more variable. Perhaps related to this, there was a complete loss of perenniality in generation 8. Neither of these issues is explained, and it is not clear why these sudden changes happened only in these last 2 generations. Possible explanations for these changes should be explored, and also addressed in terms of how this system can be developed further. Was there some mistake made in one of these generations? A mix-up of seed or pollen, or accidental interbeeding wth something else? Or what happened to cause these last 2 generations to be so different?

Response: This is a very good point, it is quite possible that there was pollen contamination in IM7F1, and we have added lines 189-290 to acknowledge this possibility.

Comment: Whatever the explanation, it seems to me that once perenniality is lost as a trait in generation 8, selection will not be able to easily recover it using that plant material. Since that trait was the entire purpose of doing this cross, it seems strange to say things like the last statement before the conclusions,  (lines 314-316) "we can likely continue to make progress, for  example, by choosing plants with large seeds and high seed set (which have a positive relationship)  while making only modest concessions in other traits like perenniality" It makes it sound like the complete loss of perenniality in generation 8 is just a "modest concession", though the authors likely don't mean that.

Response: You are correct, once perenniality is lost, plant material needs to be crossed back to a perennial parent. We did not intend for this to be a minor concession, we intended this to illustrate that outliers provide variation for selection to proceed. Plants which lost perenniality completely were removed from the population within the first year (Methods line 141). We have also changed the language of the confounding sentence so it will more clearly communicate our meaning, “This means that we can likely continue to make progress, for example, by choosing plants with large seeds and high seed set (which have a positive relationship) that also show sufficient perenniality.”

Comment: How can this actually be reversed? The authors need to address that. Perhaps by going back to genration 6 or 7? Or making crosses among generations, maybe? I'm not sure, but something should be proposed rather than ignoring the problem.

Response: We have added a sentence about this as a potential way around the problem, “Since plants are perennial, when subsequent generations do not produce the desired results (IM7F1 and IM8F1), we can cross them back to previous generations to reintroduce beneficial characteristics.” (Line 370-371.)