Temporal Regulation of the Metabolome and Proteome in Photosynthetic and Photorespiratory Pathways Contributes to Maize Heterosis

We have received reviews of your manuscript entitled "Temporal regulation of metabolome and proteome in photosynthetic and photorespiratory pathways contributes to maize heterosis." Thank you for submitting your best work to The Plant Cell. The editorial board agrees that the work you describe is substantive, falls within the scope of the journal, and may become acceptable for publication, pending revision and potential re-review.

after a short vacation in China due to travel ban. During the revision, Dr. Li has creatively collaborated with Dr. Qingxin Song, a former postdoc in my lab and now a professor at Nanjing Agricultural University, to perform the enzyme activity assays. Dr. Li is still in China, and we hope he will return to the U.S. after the travel ban is lifted. Regardless, this incidence has definitely shown some positive aspects of international collaboration to advance science and to build a peaceful world.

Reviewer #1:
Heterosis has been exploited by breeders for thousands of years and was rigorously described by Darwin who connected heterosis to inbreeding depression. Since then progress in understanding these twin phenomena has been slow. The dramatically greater size and health of hybrids are belied by transcriptome and proteome profiles that are mostly the average of their parents, and by inheritance patterns that suggest heterozygosity per se is more important than loci for hybrid vigor. The authors have previously reported that the timing of expression of the circadian clock gene, ZmCCA1, is shifted to the early morning in hybrids which causes early expression of transcripts for photosynthesis. The current manuscript follows up with profiles of the proteome and metabolome to reveal whether there are rhythmic differences between hybrids and their inbred parents, as predicted by the previous study, which may account for elevated photosynthetic rates.
Author response: Thank you for the positive assessment of our work by this expert reviewer.
For metabolomics, the authors measured two replicates of the hybrid BM, which is the commercial form, and one replicate of the reciprocal hybrid, MB. Which values were used to calculate MPV? Only one value is graphed in Fig 5  E. Is it the mean MPV from 2 or 3 "reps"? In "Results" the authors state that 3 biological replicates were used. This is potentially misleading and should be clarified in the Discussion or earlier in the manuscript. For proteomics, the authors measured two replicates each of BM and MB. In "Results" the authors state that 3 replicates of each sample were used. This is in contrast to the section on "Statistical analysis" which states only two replicates were used: "The difference of the average number of expressed proteins for inbreds and hybrids at each time point were statistically tested by calculating the standardized normal deviate parameter z, which is given by calculating the difference in the two numbers and divided by the square root of their sum." This is potentially misleading and should be clarified in the Discussion or earlier in the manuscript.
Author response: Again, we are sorry for the ambiguity of the replicates. As for the metabolome data, we also used three biological replicates for the proteome analysis of each genotype at each time point. "The difference of the average number of expressed proteins between inbreds and hybrids at each time point was tested by calculating the standardized normal deviate parameter z (Pocock, 2006). Here, z = (inbred -hybrid) / sqrt (inbred + hybrid), which estimates the difference of the two values (hybrid and hybrid) and divided by the square root (sqrt) of their sum. The "inbred" and "hybrid" in the formula indicate the numbers of expressed proteins for inbred parents (the average between B73 and Mo17) and hybrids (the average between BM and MB), respectively, at each time point." We have revised the description in "Statistical analysis" section to avoid the potential confusion.
The strength of this manuscript is provided by an integrated analysis of metabolites and proteins. The technical work is good although it would have been best to have three biological replicates of each sample type.
Author response: Thanks for the supportive comment. Indeed, the analyses and results present in this work were all based on three biological replicates for each genotype in a given time point. We have clarified some confusing descriptions in the revision.
In key aspects these orthogonal data coincide to provide a convincing description of quantitative differences associated with hybrids. Previous studies have found indications that transcripts for photosynthetic proteins are elevated in hybrids and this manuscript shows direct evidence that at least some carbon reaction enzymes are up and this is concomitant with differences in some metabolites. Perhaps more novelty is provided by the observation that enzymes and metabolites of photorespiration are substantially lower in hybrids, although the enzyme data are mixed and it isn't clear why the observed differences in protein levels would be expected to cause the lower abundances of Gln, Gly, and Ser. The authors should comment on the relative importance of photorespiration in C4 plants.
Author response: We further measured the enzyme activities of GOX and CAT, which were consistent with their protein levels and showed nonadditive increases during the mid-day and at night, respectively (Supplemental Figure 9), supporting their respective roles in H2O2 generation and scavenging. As a result, these enzymes may accelerate rapid turnover of metabolites such as Gln, Gly and Ser during mid-day to early afternoon, leading to lower abundance at night. However, additional experiments will be needed to test causal relationships. We have discussed this aspect of photorespiration in C4 plants in the revision. "In C4 plants such as maize, photorespiration is suppressed due to the increased CO2 concentration in BS chloroplasts by the carbon shuttle, which results in higher photosynthetic efficiency compared to C3 plants. However, photorespiration is not only essential for C4 plants, but also plays an important role the evolution of C4 photosynthesis (Sage, 2004). In photorespiration, GOX is responsible for the catalyzation of the oxidation of glycolate to glyoxylate in higher plants ( Figure 7B)…." In Fig 3 C-D the authors accept as meaningful metabolite level differences from MPV of 2% or more. They should discuss how this level of difference was chosen as the threshold, how it compares to the error rate, and how it may account for the growth rate or photosynthetic rate differences observed.
Author response: This is a good suggestion. We observed the median MPH for metabolites in day and night times is around ~2.5%. Instead of using this as a threshold, we used 2% and 5% as indicative values to show the number of metabolites that have "small" absolute mean MPH (2.0%-5.0%) and that have relatively "large" (greater than 5%) absolute mean MPH. We revised this in the text.
The authors focus on metabolites and proteins with diel expression patterns. Suppl 5 C-D shows that only 30% of non-additive proteins are also rhythmic. Maybe these are the most important for heterosis but the authors don't explain why the other 70% of the non-additive proteins are unlikely to contribute to heterosis; this point should be addressed.
Author response: In this analysis, we mainly focus on nonadditive proteins that are also rhythmic and found these proteins with upregulation are exclusively enriched in photosynthesis related GO terms. We agree that the other 70% nonadditive proteins can also be very important for heterosis, perhaps through other mechanisms. Their central hypothesis is that regulation is provided by the circadian clock in a manner that causes enzyme expression earlier in the day which gives the hybrids a photosynthetic head-start. It's clear that some metabolites and proteins are rhythmic and that hybrids have more extreme amplitudes. It isn't clear that the phase or period is different in hybrids. It's also not clear whether diel regulation is driven by the clock or the environment. The authors should explain why the data cannot arise from increased amplitude with unchanged rhythmic phase and period, which would also cause elevated expression earlier in the day.
Author response: This is a good comment. Maintaining proper circadian phase and period is essential for optimizing growth and development, while altering (either increase or decrease) phase or period can lead to altered growth patterns. We used JTX CYCLE to test the phase distributions for rhythmic proteins in the hybrids and inbreds. The data showed similar phases between the hybrids and inbreds, but hybrids exhibited a larger fraction of rhythmic proteins in early morning (ZT3) and a smaller fraction at night (ZT21) (Supplemental Figure 6). This data is supportive to our previous findings for the temporal shift of ZmCCA1-binding targets to the early morning in the hybrids as a head-start for photosynthetic  Author response: Thanks for the suggestion, we added boundary lines between the clusters. After carefully examining the dendrogram, we revised it as three main clusters, instead of four, as in the prior manuscript: the prior cluster 3 and 4 were combined into one cluster (the new cluster 3 in the revised manuscript). The boundaries between clusters were defined according to the distances between clusters calculated by R software. This is included in the Methods.
The authors make claims about different numbers of proteins expressed between inbreds and hybrids at various time points. These numbers are in the 2300-3000 range which is substantially fewer than the number of expressed proteins so the authors should refer to what is detectable by their method not what is expressed.
Author response: Thanks for the suggestion. We clarified this this in the revised manuscript. "A total of 4,792 proteins that were detectable by the two criteria noted above in at least one genotype or time point were classified as expressed proteins." The authors only analyzed the hybrids of two inbred parents. While this is perfectly acceptable it leaves unanswered an important question. Are the observed differences associated with heterosis, as suggested, or could they be associated with heterozygosity per se, independent of heterosis. The authors should comment on this possibility.

Author response: This is an interesting comment, which could invoke a debate about the genetic models for heterosis, namely, dominance and overdominance. Heterozygosity including nonadditive components could be related to heterosis; alternatively, other factors such as additive factors could be related to heterosis. Although our data show most changes in the hybrids (heterozygous) that are associated with heterosis, we cannot rule out other possibilities.
Overall, this manuscript provides a valuable addition to the description of differences between hybrids and their inbred parents. It presents specific hypotheses for heterosis arising from differences in metabolism and these can guide future research.
Author response: We appreciate this positive summary of our work along with a future perspective.

Reviewer #2:
Previous studies have shown that diurnal regulation of gene expression in maize can be extensive and is one of the factors that impact heterosis in maize and other plant species. It was further demonstrated that higher levels of carbon fixation and starch accumulation in maize hybrids as compare to their respective inbred are associated with altered temporal gene expression, which contributes to biomass heterosis. Nevertheless, the connection between the metabolic and proteomic responses and heterosis is not well understood. In the present manuscript, the authors addressed this issue by performing integrative approach of time series of metabolome and proteome analyses from maize seedlings of hybrids and their inbred. They characterized the metabolites and the proteins expression during one diurnal cycle and found that many of them are diurnally regulated and have nonadditive abundance in the hybrids. They also found that although metabolic heterosis was detectable, it remained relatively mild compared to other known heterotic traits. Interestingly, the authors observed different trends for categories such as sugars and amino acids, a finding that aligned with previous observations regarding the role of carbon fixation and photosynthetic efficiency in hybrid vigor.
A more in depth analysis further demonstrated that the metabolites associated with photosynthesis pathway showed a positive MPH, while those in the photorespiratory pathway showed a negative MPH. Selective analysis on the key enzymes that are associated with these processes showed that during specific time points, several key enzyme protein abundance corresponded with the trends of their pathway-related metabolites during the day or night (those however were not analyzed per time point as in the proteome). Based on these observations and previous studies, the authors proposed that hybrids optimize the abundance of relevant metabolites to improve carbon efficiency. They further suggested that hybrids are efficient in removing toxic metabolites generated during photorespiration and therefore have higher photosynthetic efficiency. However, the observed metabolic heterosis was mild and only protein abundance studies were performed, and not the enzyme activity measurements. H2O2 measurement was only significant during limited time points. To suggest that the interaction between photosynthesis and photorespiration drives more pronounced heterosis phenotypes, I would think that a more causal evidence is needed.
Author response: We appreciate the positive summary of our work. We agree that measuring the enzyme activity of key enzymes should substantiate the evidence. As suggested, we tested enzyme activities of four key enzymes: GOX, CAT, TK and RCA during the revision. The results are generally consistent with protein data with a few exceptions. The enzymes in the Calvin cycle such as RCA and TK have higher activities in the hybrids than the MPV, which may promote carbon assimilation in hybrids (Supplemental Figure 9). In photorespiration, GOX had higher activity from mid-day to early night in one hybrid (BM), while CAT activity was lower in one hybrid (BM) and higher at night in another hybrid (MB). This may suggest that either H2O2 or glyoxylate pathway can help reduce toxic metabolites in the hybrids. Testing additional enzymes in each pathway would clarify this notion.
Author response: As suggested, we added axis annotation in Supplemental Figure 1B. 2. Can the authors address why less than 35% was detected in their PCA and how does it influence their inferences.
Author response: Metabolic data are inherently noisy. The two PCs with the largest variations have captured the two major treatments (components): genotype and time point. Other factors such as tissue type, number and type of metabolites may also influence the explained variation but cannot be teased apart by this method. We believe that two main PCs have separated two major treatments and should represent the overall trend of these results. Figure 1 is not clear, it seems like there are 3 main clusters and not 4, please review. It would also be helpful to show the average trend of each cluster next to it for clarity.

The separation of clusters in
Author response: Thanks for the insightful comment. After carefully checking the dendrogram, we indeed found three main clusters, instead of four; the prior clusters 3 and 4 should be combined into one cluster. We added boundary lines between clusters to make it clear and clarified this in the Methods. "To define cluster borders, we used R package 'heatmap.2' {R Core Team, 2018 #5393}, which calculates the Euclidean distance between measurements to obtain distance matrix and employs complete agglomeration method for clustering." 4. I think it would be useful to show whether the shard nonadditive metabolites are showing the same trends in the two hybrids and whether they align with the authors' hypothesis.
Author response: For the 175 overlapped nonadditive metabolites between reciprocal hybrids, the majority (133, 76%) showed changes in the same direction, either higher or lower than the MPV in both BM and MB. We clarified this in the revision. Figure 3C and D, it is not possible to see how specific metabolites behaved similarly/differently during the day and night. Maybe it would be more useful to show the metabolites by categories followed by day and night comparison.

In
Author response: This is a good suggestion. Some specific metabolites were grouped under day and night and shown in Figure 3A, 3B. For example, amines and amino acids tend to show negative MPH, while lipids and nucleosides tend to show positive MPH under both day and night (Figure 3A, 3B). It would be difficult to show the day and night comparison for all of them. We also showed specific amino acids that display positive MPH during the day and at night ( Figure 4F, 4G). Furthermore, by focusing on a group of key metabolites related to photosynthesis, we found that some of these metabolites behave differently under day and night (Figure 5C and 5E), such as PEP and pyruvate.
6. Since the study shows specific time points for key proteins it is worthwhile to do the same for key metabolites to see if there was any correlation between the metabolome and the proteome.
Author response: As suggested, we examined some key metabolites involved in photosynthesis (PEP, pyruvate), photorespiration (glycine, glutamine), and Calvin cycle (sucrose, fructose) (Supplemental Fig. 9 and also attached below). For the hybrids, PEP and pyruvate showed higher abundance in the morning or early afternoon, and glycine and glutamine showed lower abundance in the afternoon. Carbon assimilations such as sucrose and fructose showed higher abundance in the afternoon and night for some time points. These results are consistent with protein abundance (Figures 6 and 7) and enzyme activities (Supplemental Fig. 8) of some key enzymes examined in these pathways. We added the data and clarified these views in the Results.
genotypes can be expected to function highly similar in principal, I wonder how the authors explain such small overlap and the high genotype specificity of basic metabolite diel profiles.
Author response: Thanks for the insightful comments. By examining the overlapping separately for inbred parents and the hybrids, we found over half (60% ~ 65%, Figure 4C) of the rhythmic metabolites were overlapped between the two inbred parents and between the reciprocal hybrids respectively, which is much higher than the overlapping among all genotypes. We believe that the low overlap may also indicate the metabolome divergent between the two inbreds and between the inbreds and hybrids. In addition, there is a relatively low overlap of nonadditive proteins between reciprocal hybrids (20%) and nonadditive metabolites between reciprocal hybrids (30%, Figure 2B). This low overlap of the nonadditive proteins (and large variation) was also reported in maize seedling seminal roots of MB and BM hybrids (~9%) (Marcon et al., 2011, J. Proteomics). There is also low overlap (10%) of the nonadditively expressed genes between BM and MB maize hybrids (Stupar et al., 2007, Plant Physiol.). We speculate that this low overlap observed may be related to the genetic and physiological diversity among genotypes as well as technical factors such as the limitation of GC-MS and LC-MS methods. We included these views in the Results and Discussion.
In the introduction (line 109) the authors delimitate their statement on diurnal regulation of the metabolome and proteome, which is not taken further in the manuscript. The JTK_CYCLE method was used to define circadian regulation, however, an explanation on the validity or shortcomings of this method for the available dataset is missing.
Author response: JTK_CYCLE method has been widely used to identify rhythmic mRNAs and metabolites over the years for its fast speed and high accuracy (Hughes et al., 2010, JOURNAL OF BIOLOGICAL RHYTHMS). Indeed, our datasets are shown to be effectively supported by JTK_CYCLE (JTK_CYCLE User's Guide). Although the available datasets of metabolome and proteome in one diel cycle with limited time points may hinder JTK_CYCLE to capture certain subtle rhythmicity resides, our analysis has identified several general trends, providing some directions for future studies. We added this view in the Discussion.

TPC2020-00320-RAR1 2 nd Editorial decision -acceptance pending
Sept. 11, 2020 We are pleased to inform you that your paper entitled "Temporal regulation of metabolome and proteome in photosynthetic and photorespiratory pathways contributes to maize heterosis" has been accepted for publication in The Plant Cell, pending a final minor editorial review by journal staff.
Final acceptance from Science Editor Sept. 29, 2020