Rapid transgenerational adaptation in response to intercropping reduces competition

  1. Laura Stefan
  2. Nadine Engbersen
  3. Christian Schöb

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This paper is of potential interest to people working at the interface between agronomy, ecology, and evolution. By growing experimental mixtures of crop species, i.e., intercrops, the study aims at testing whether positive interactions between species grown in association strengthen over generations of coexistence. The data are original and of high quality, and the statistical analysis are rigorous. The interpretation of the Results as well as the Discussion and Conclusions currently ignore an important discrepancy in the results for competition versus overall yield.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

By capitalising on positive biodiversity – productivity relationships, intercropping provides opportunities to improve agricultural sustainability. Intercropping is generally implemented using commercial seeds that were bred for maximal productivity in monocultures, thereby ignoring the ability of plants to adapt over generations to the surrounding neighbourhood, notably through increased complementarity, that is reduced competition or increased facilitation. This is why using monoculture-adapted seeds for intercropping might limit the benefits of crop diversity on yield. However, the adaptation potential of crops and the corresponding changes in complementarity have not been explored in annual crop systems. Here we show that plant – plant interactions among annual crops shifted towards reduced competition and/or increased facilitation when the plants were growing in the same community type as their parents did in the previous two generations. Total yield did not respond to this common coexistence history, but in fertilized conditions, we observed increased overyielding in mixtures with a common coexistence history. Surprisingly, we observed character convergence between species sharing the same coexistence history for two generations, in monocultures but also in mixtures: the six crop species tested converged towards taller phenotypes with lower leaf dry matter content. This study provides the first empirical evidence for the potential of parental diversity affecting plant – plant interactions, species complementarity and therefore potentially ecosystem functioning of the following generations in annual cropping systems. Although further studies are required to assess the context – dependence of these results, our findings may still have important implications for diversified agriculture as they illustrate the potential of targeted cultivars to increase complementarity of species in intercropping, which could be achieved through specific breeding for mixtures.

Article activity feed

  1. Version published to 10.7554/elife.77577 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    In this study, Stephan et al. use experimental data to test whether positive interactions between different crop species strengthen over time (generations) when these species are cultivated in association. Even if this has already been investigated in grassland species, we currently lack experimental data on such questions in crops, which makes the study original and with potentially important agronomic applications. To address the question, the authors designed two types of communities: monocultures and mixtures made from seeds collected on plants that evolved in the same community type in the previous two generations (i.e. mixtures from mixture seeds and monocultures from monoculture seeds), and monocultures and mixtures made from seeds collected on plants that evolved in a different community type in the previous two generations (i.e. mixtures from monoculture seeds and monocultures from mixture seeds). They then used multiple sets of indexes to characterize the magnitude and direction of plant-plant interactions in order to compare communities with different evolutionary histories. Interestingly, the experiment is also replicated across two fertilization treatments. At the individual plant level, the results suggest that facilitation increases and competition decreases for plants grown in the same community type as their progenitors compared to plants grown in a different community type as their progenitors. Community-level analysis, however, shows a different picture: both the total yield of the communities and the relative yield of the mixtures are not affected by the coexistence history of their parents, and the results do not provide evidence for increased complementarity between species that have evolved in mixtures in the previous generations. Finally, the authors show that several aboveground traits have lower variation in communities that evolved in the same community type in the previous generations compared to communities that evolved in a similar community type, which shows phenotypic convergence rather than the expected phenotypic divergence in mixtures. They also report differences in trait means, with for example lower mean leaf dry matter content in communities composed of offspring of plants that were grown in the same community type compared to communities composed of offspring of plants that were grown in a different community type.

    The study is based on original and high-quality experimental data. The number of species, communities, and replicates is relevant regarding the research question. It is also very nice to have contrasted environments (fertilized vs unfertilized). All these combinations of factors have been replicated over three consecutive years, which also needs to be acknowledged as an impressive experimental effort.

    The statistical analysis is rigorous and successfully accounts for design features when testing the effects of interest. My main criticisms concern, by order of importance, the disconnection between the results and the claims of the paper, some weaknesses in the experimental design, and the clarity of the Materials and Methods.

    The main hypothesis of the study, which is that higher facilitation/complementarity should occur in mixtures made from plants that evolved in mixtures compared to mixtures made from plants that evolved in monoculture, is not supported by the results. However, the results are presented in such a way that it seems that this hypothesis is verified. For example, the first index which is used by the author (Relative Intensity Index, RII) shows a significant effect on the treatment of interest (coexistence history). However, this index is not the most relevant to capture facilitation or complementarity effects in multi-species communities, notably because it is computed at the single plant level, and only with three plants per species. The Loreau & Hector partitioning (Net Biodiversity >Effect, NBE, partitioned into a Selection Effect, SE, and a Complementarity Effect, CE), which is used a second time, is the gold standard in the field. This is acknowledged by the authors given that they check the validity of RII by measuring its correlation with CE in Fig. S13). Unfortunately, RII and NBE give very different results: the coexistence history of the species has no effect on NBE, and most notably no effect on the CE component. Yet, the authors claim that the coexistence history has an effect on NBE in the fertilized treatment, but we do not have any information supporting the statistical significance of this result (the p-value used to support this claim l. 103 is > 0.05). More generally, several non-significant results are discussed (e.g. l. 100 to 105, l. 136 to 144). The Figures in the main text are also misleading. They show the means ({plus minus} standard error) in the different treatment but do not report statistical significance. Very often, the related boxplots in the Supplementary Information show much fewer differences between the treatments (e.g. Figure 2 vs Fig. S1, Figure 3 vs Figure S3), and the ANOVA tables confirm that these differences are not statistically significant. Overall, the fact that coexistence history has no effect on the total yield, on NBE, and most notably on the CE component, together with the fact that species' traits converge, notably towards taller plants, do not support reduce competition nor higher facilitation in mixtures with a mixture history compared to mixtures with a monoculture history.

    The p-values are indicated in the main text along the figures in the Result section. Significance stars were now added and we hope it will be clearer. Furthermore, the discrepancy between RII results and NE/CE was also discussed more extensively in the discussion and we refer to our responses of the general recommendations (1) and (3) by the review editor above. In summary, with our interpretations we tried to stick to the results provided by the data and now provide explanations to reconcile the different results. This does hopefully help to demonstrate that the different results we obtained are not contradictive.

    The amount of phenotypic and genetic variation within each species at the beginning of the experiment has not been controlled and reported in the study. It seems that inbred lines were chosen for some species (e.g., wheat, oat, or lentil) which means that there was no genetic variation for these species, whereas landraces or open-pollinated varieties were chosen for others (e.g., coriander or camelina). It thus means that the evolutionary potential of the different species was not the same. It would have been more rigorous to choose either only fixed genotypes for all species (inbred lines or hybrids), which would then have evolved under the sole effect of epigenetic changes, or only mixtures of genotypes for all species (either varietal mixtures or open-pollinated varieties), which would then have evolved under natural selection and changes in gene frequency.

    We recognize that the underlying mechanisms remain unclear, as indeed we did not measure the initial amount of standing variation and we could not find seed or populations with the exact genetic variation. This was not done as investigating the potential genetic mechanism was beyond the scope of the study. Therefore, we can only speculate regarding the potential mechanisms, and we have now added an extensive paragraph discussing the possibilities in the discussion (L305-323).

    Several aspects of the Materials and Methods could be clarified. It is not clear how the different plots and community types were re-allocated each year. This is important to interpret the results, as soil legacy effects could also affect the outcomes of plant-plant interaction. For example, were mixture plots with a "pure" mixture history grown in the same plot from one year to the other, or were plots reshuffled each year?

    Plots were reshuffled each year precisely to avoid soil legacy effects. This was clarified in the methods.

    Also, the sowing pattern of the 4-species mixtures is not explained. Was it also alternate rows, as the 2 species mixtures? Was the pattern the same across the different replicates and treatments for a given 4-species mixture?

    Since plots were sown with four lines a 50 cm length, in the case of 4-species mixtures, it was for each crop species 1 line. We did not mix species within sowing lines. The pattern across the different replicates and treatments was randomized and is not necessarily the same for a given 4-species mixture. This was now added to the methods.

    We do not have information on the sowing densities in mixtures plots (was it simply their monoculture densities divided by the number of species?).

    Yes, indeed. We kept the monoculture sowing densities in the mixture plots (i.e. if we planted 10 lentils per line in the monoculture, we also planted 10 lentils per line in the mixture). This was added in the method description.

    An important aspect of index computation is also not explained. For example, monoculture yield is used as a reference to compute Relative Yield (RY), and single plant yield is used as a reference to compute RII. There are several ways to compute these reference values, given that there are multiple replicates of monocultures and single plants for a given species. It can be either the value of the closest replicate in the experiment in the same treatment, the average value of all replicates in the same treatment, or a model-derived prediction which accounts for design effects (BLUE or BLUP). In this experiment, monocultures and single plants are also replicated across different evolutionary histories. So, we need to know which type of monoculture plots or single plant plots were used to compute RII and NBE.

    This was specified in the Methods and discussed in the response to the general comments. In short, we always used the average value over the replicates with the same treatment combination, e.g. a single plant with a monoculture history in unfertilised plots as a reference for monoculture history treatments in unfertilised plots.

    Reviewer #2 (Public Review):

    The paper offers a very novel experimental framework for assessing how coexistence history could influence intercropping success in agricultural systems. The authors do a very nice job combining the science from multiple fields into a coherent and useful framework. Based on this framework we should conclude that growing crops in polyculture fields for multiple generations will increase the benefits of intercropping for growing food.

    However, on the ecological side, there are some weaknesses that need to be addressed:

    1. The introduction and discussion need more context for how co-occurrence can lead to more facilitation. I see how co-occurrence could lead to trait displacement and less niche overlap, so less competition. But what is the facilitation part of this? The introduction doesn’t introduce any potential mechanisms for this despite many indications that facilitation could also change as a result of coexistence history.

    We included a paragraph covering the evolution of facilitation in the introduction (L49-57).

    1. The authors should think carefully about their use of net effects, RII_facilitation, and RII_competition. It appears to me as though all three are measuring net effects but in some cases facilitation > competition and in other cases competition > facilitation. Even though that’s true, it doesn’t mean that the indices aren’t still measuring net effects. Given that, the authors should temper that language and consider reinterpreting some of their data.

    Indeed, as our RII measures net effects, for more clarity, we decided to skip this distinction between RII facilitation and RII competition – thereby also acknowledging concerns raised by reviewers 1 and 3. Furthermore, we rephrased our statements to make clear that what we measure is the outcome of plant-plant interactions and that we could not always distinguish between increased facilitation and/or reduced competition.

    1. The authors should also give careful consideration to the relative balance of inter vs. intraspecific competition. Many (if not all) of these trends could be indicative of stronger intraspecific competition than interspecific competition. This will need to be considered very carefully.

    We agree in that increased complementarity with increasing diversity is per se due to stronger intraspecific competition than interspecific competition, and consequently higher benefits of alleviation of the most important source of competition. This is the underlying hypothesis of all BEF studies and actually the reason why we tackled BEF in this study by means of plant-plant interaction metrics. We clarified this in the introduction (L96-99).

    1. Have the authors considered separating their data into plots with and without legumes? The strong selection effects with co-occurrence history would also support this. Nitrogen enrichment is one of the most heavily studied facilitation mechanisms and thus this separation might help give insight into the mechanisms operating here.

    We tried to separate effects with and without legumes, but since all our 4-species mixtures necessarily included a legume, the plots without legumes were only monocultures and 2-species mixtures, and therefore it did not give a representative picture of the experimental design.

    1. Overall, a lack of clarity of underlying mechanisms is the greatest weakness of the paper.

    We recognize that the underlying mechanisms remain unclear, as indeed we did not measure the initial amount of standing variation and we could not find seed or populations with the exact same genetic variation. This was not done as investigating the potential genetic mechanism was not the goal of the study. Therefore, we can only speculate regarding the potential mechanisms, and we have now added an extensive paragraph discussing the possibilities in the discussion (L305-323).

    Reviewer #3 (Public Review):

    This work investigates the effects of growing annual crops for two generations in the same or a different social environment (coexistence history being single plant, monoculture, mixture of species) on measures of competition and yield. This is a very interesting and timely topic; diversification in agriculture is a promising means to help reduce the global decline of biodiversity. The experimental setup appears to be sound and the experiment is carefully executed (though this is not my area of expertise). The authors conclude that growing plants in the same community as their parents did reduces competition.

    However, I am not convinced by the interpretation of the results. Particularly the results for competition versus overall yield are in conflict. This discrepancy is not properly discussed and is largely ignored in the conclusions. Hence, I doubt whether the results support the conclusions.

    We would like to thank the reviewer for the constructive feedback and the appreciation of our work. Regarding the potentially conflicting findings mentioned by the reviewer here, we would like to refer to our previous statement in response to the review editor, in particular response (1) and (3), where we show that these results are not necessarily conflicting.

    My most important comment relates to the discrepancy between results for total yield (Figure 3b) versus those for competition (Figure 2a) and for net biodiversity effect (Figure 3a). Results for all those measures are based on yield records. Figure 2a and 3b (panel fertilizer) show clearly that plants that have the same coexistence history as the tested plants outperform those having a different co-existence history. Figure 3b, however, shows no effect of coexistence history on yield; total yield for Same and Different do not differ. How to reconcile these results? Remarkably, this discrepancy is not discussed at all; the discussion largely ignores the absence of an effect on total yield.

    This discrepancy is now discussed in the discussion (L230-245), which was largely rephrased to take into account the comments. The discrepancy between the response of plant-plant interactions and the response of net biodiversity effects to coexistence history can stem from various reasons. First, net biodiversity effects are driven both by complementarity and selection effects (6); therefore, a reduction in competition does not necessarily lead to an increase in net biodiversity effects, as this can be compensated by concurrent changes in selection effects. Changes in RII should however correlate with complementarity effects, which they do in our study (Fig. S11, p-value = 0.033), indicating that reduced competition and/or increased facilitation correlates with higher complementarity effects. As mentioned in the response to the general comments, the reference for RII (single plant) is not the same as for community level measures such as NE (monoculture). This can also explain why coexistence history affects RII but not NE (see how we calculate this extra RII_monoculture in the main response). Finally, our RII calculations and net biodiversity effects also take into account the baseline effect of coexistence history on the reference plant or community (i.e. single plant for RII, monoculture for net effects). This allows to explicitly distinguish the effects of coexistence history on the interactions, independently of the baseline effect on plant performance overall (35), and can explain why the effect of coexistence history on relative metrics (such as RII) does not appear in absolute metrics (such as total yield). We also suggest that the limited timeframe of this study – two generations – might be the reason for the lack of more significant changes in total yield.

    Related to the previous comment, the title includes the phrase “reduces competition”. In the manuscript, competition is derived from effects on yield. Still, there is no benefit of the same coexistence history for total yield. This is somewhat misleading.

    See response to main comments.

    A second important comment relates to the absence of results from the second year. The Methods section explicitly states that the comparisons made in year 3 (as shown in Figure 2) were also made in the second year (2018; L350-358). However, no results are presented. Why are those results excluded?

    We collected only partial data after 1 year, as this was considered as an intermediate stage, where adaptation was less likely to give significant results. Notably, we put fewer efforts into collecting data at the individual-level and reduced the number of traits measured, which prevented us from having a full picture of the response of plant-plant interactions as well as of the trait space. Therefore, we decided to not present these partial pieces of data in the study.

    A third comment relates to the distinction between competition and facilitation (Equations 3 and 4, and corresponding results), which is artificial and not very meaningful in my opinion. Since RII will never be precisely equal to zero (i.e., the RII=0 category is empty), an increase (decrease) in facilitation must go together with a decrease (increase) in competition, and vice versa. This must be the case since the total of both categories must add up to the number of comparisons made. In other words, if we have a total of N objects, being either apples or pears, then, if we have fewer apples, we must have more pears. (hence, L93-94 is a tautology). I suggest dropping this distinction from the manuscript.

    The decomposition between competition and facilitation was dropped (as already mentioned above).

    The Discussion seems to ignore some of the results that don’t seem to match the “desired” outcome. For example, L178 speaks about niche differentiation as if this was found, but it was not. Same for L200. Similarly, L181 speaks about “the yield benefit”, which was not there.

    We tried to reformulate the discussion and notably to emphasize that we find some clues for niche differentiation (light use, RII) but this did not consistently match other measures (traits, CE).

    While the manuscript is well written with respect to the language, it is not always easy to follow and absorb. This is partly because the number of traits is large. A table with the traits could help. Also, the writing could be improved to help the reader get the message. For example, when showing results in Figure 2, it could be mentioned from the start that these are relative to single plants, whereas those in Figure 3a are relative to monoculture. This can be found in the methods but should be clear from the Results as well.

    This was clarified and specified in both the main text of the results and in the figure legends.

  3. eLife

    Evaluation Summary:

    This paper is of potential interest to people working at the interface between agronomy, ecology, and evolution. By growing experimental mixtures of crop species, i.e., intercrops, the study aims at testing whether positive interactions between species grown in association strengthen over generations of coexistence. The data are original and of high quality, and the statistical analysis are rigorous. The interpretation of the Results as well as the Discussion and Conclusions currently ignore an important discrepancy in the results for competition versus overall yield.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    In this study, Stephan et al. use experimental data to test whether positive interactions between different crop species strengthen over time (generations) when these species are cultivated in association. Even if this has already been investigated in grassland species, we currently lack experimental data on such questions in crops, which makes the study original and with potentially important agronomic applications. To address the question, the authors designed two types of communities: monocultures and mixtures made from seeds collected on plants that evolved in the same community type in the previous two generations (i.e. mixtures from mixture seeds and monocultures from monoculture seeds), and monocultures and mixtures made from seeds collected on plants that evolved in a different community type in the previous two generations (i.e. mixtures from monoculture seeds and monocultures from mixture seeds). They then used multiple sets of indexes to characterize the magnitude and direction of plant-plant interactions in order to compare communities with different evolutionary histories. Interestingly, the experiment is also replicated across two fertilization treatments. At the individual plant level, the results suggest that facilitation increases and competition decreases for plants grown in the same community type as their progenitors compared to plants grown in a different community type as their progenitors. Community-level analysis, however, shows a different picture: both the total yield of the communities and the relative yield of the mixtures are not affected by the coexistence history of their parents, and the results do not provide evidence for increased complementarity between species that have evolved in mixtures in the previous generations. Finally, the authors show that several aboveground traits have lower variation in communities that evolved in the same community type in the previous generations compared to communities that evolved in a similar community type, which shows phenotypic convergence rather than the expected phenotypic divergence in mixtures. They also report differences in trait means, with for example lower mean leaf dry matter content in communities composed of offspring of plants that were grown in the same community type compared to communities composed of offspring of plants that were grown in a different community type.

    The study is based on original and high-quality experimental data. The number of species, communities, and replicates is relevant regarding the research question. It is also very nice to have contrasted environments (fertilized vs unfertilized). All these combinations of factors have been replicated over three consecutive years, which also needs to be acknowledged as an impressive experimental effort.

    The statistical analysis is rigorous and successfully accounts for design features when testing the effects of interest. My main criticisms concern, by order of importance, the disconnection between the results and the claims of the paper, some weaknesses in the experimental design, and the clarity of the Materials and Methods.

    The main hypothesis of the study, which is that higher facilitation/complementarity should occur in mixtures made from plants that evolved in mixtures compared to mixtures made from plants that evolved in monoculture, is not supported by the results. However, the results are presented in such a way that it seems that this hypothesis is verified. For example, the first index which is used by the author (Relative Intensity Index, RII) shows a significant effect on the treatment of interest (coexistence history). However, this index is not the most relevant to capture facilitation or complementarity effects in multi-species communities, notably because it is computed at the single plant level, and only with three plants per species. The Loreau & Hector partitioning (Net Biodiversity Effect, NBE, partitioned into a Selection Effect, SE, and a Complementarity Effect, CE), which is used a second time, is the gold standard in the field. This is acknowledged by the authors given that they check the validity of RII by measuring its correlation with CE in Fig. S13). Unfortunately, RII and NBE give very different results: the coexistence history of the species has no effect on NBE, and most notably no effect on the CE component. Yet, the authors claim that the coexistence history has an effect on NBE in the fertilized treatment, but we do not have any information supporting the statistical significance of this result (the p-value used to support this claim l. 103 is > 0.05). More generally, several non-significant results are discussed (e.g. l. 100 to 105, l. 136 to 144). The Figures in the main text are also misleading. They show the means ({plus minus} standard error) in the different treatment but do not report statistical significance. Very often, the related boxplots in the Supplementary Information show much fewer differences between the treatments (e.g. Figure 2 vs Fig. S1, Figure 3 vs Figure S3), and the ANOVA tables confirm that these differences are not statistically significant. Overall, the fact that coexistence history has no effect on the total yield, on NBE, and most notably on the CE component, together with the fact that species' traits converge, notably towards taller plants, do not support reduce competition nor higher facilitation in mixtures with a mixture history compared to mixtures with a monoculture history.

    The amount of phenotypic and genetic variation within each species at the beginning of the experiment has not been controlled and reported in the study. It seems that inbred lines were chosen for some species (e.g., wheat, oat, or lentil) which means that there was no genetic variation for these species, whereas landraces or open-pollinated varieties were chosen for others (e.g., coriander or camelina). It thus means that the evolutionary potential of the different species was not the same. It would have been more rigorous to choose either only fixed genotypes for all species (inbred lines or hybrids), which would then have evolved under the sole effect of epigenetic changes, or only mixtures of genotypes for all species (either varietal mixtures or open-pollinated varieties), which would then have evolved under natural selection and changes in gene frequency.

    Several aspects of the Materials and Methods could be clarified. It is not clear how the different plots and community types were re-allocated each year. This is important to interpret the results, as soil legacy effects could also affect the outcomes of plant-plant interaction. For example, were mixture plots with a "pure" mixture history grown in the same plot from one year to the other, or were plots reshuffled each year? Also, the sowing pattern of the 4-species mixtures is not explained. Was it also alternate rows, as the 2 species mixtures? Was the pattern the same across the different replicates and treatments for a given 4-species mixture? We do not have information on the sowing densities in mixtures plots (was it simply their monoculture densities divided by the number of species?). An important aspect of index computation is also not explained. For example, monoculture yield is used as a reference to compute Relative Yield (RY), and single plant yield is used as a reference to compute RII. There are several ways to compute these reference values, given that there are multiple replicates of monocultures and single plants for a given species. It can be either the value of the closest replicate in the experiment in the same treatment, the average value of all replicates in the same treatment, or a model-derived prediction which accounts for design effects (BLUE or BLUP). In this experiment, monocultures and single plants are also replicated across different evolutionary histories. So, we need to know which type of monoculture plots or single plant plots were used to compute RII and NBE.

  5. eLife

    Reviewer #2 (Public Review):

    The paper offers a very novel experimental framework for assessing how coexistence history could influence intercropping success in agricultural systems. The authors do a very nice job combining the science from multiple fields into a coherent and useful framework. Based on this framework we should conclude that growing crops in polyculture fields for multiple generations will increase the benefits of intercropping for growing food.

    However, on the ecological side, there are some weaknesses that need to be addressed:
    1. The introduction and discussion need more context for how co-occurrence can lead to more facilitation. I see how co-occurrence could lead to trait displacement and less niche overlap, so less competition. But what is the facilitation part of this? The introduction doesn't introduce any potential mechanisms for this despite many indications that facilitation could also change as a result of coexistence history.
    2. The authors should think carefully about their use of net effects, RII_facilitation, and RII_competition. It appears to me as though all three are measuring net effects but in some cases facilitation > competition and in other cases competition > facilitation. Even though that's true, it doesn't mean that the indices aren't still measuring net effects. Given that, the authors should temper that language and consider reinterpreting some of their data.
    3. The authors should also give careful consideration to the relative balance of inter vs. intraspecific competition. Many (if not all) of these trends could be indicative of stronger intraspecific competition than interspecific competition. This will need to be considered very carefully.
    4. Have the authors considered separating their data into plots with and without legumes? The strong selection effects with co-occurrence history would also support this. Nitrogen enrichment is one of the most heavily studied facilitation mechanisms and thus this separation might help give insight into the mechanisms operating here.
    5. Overall, a lack of clarity of underlying mechanisms is the greatest weakness of the paper.

  6. eLife

    Reviewer #3 (Public Review):

    This work investigates the effects of growing annual crops for two generations in the same or a different social environment (coexistence history being single plant, monoculture, mixture of species) on measures of competition and yield. This is a very interesting and timely topic; diversification in agriculture is a promising means to help reduce the global decline of biodiversity. The experimental setup appears to be sound and the experiment is carefully executed (though this is not my area of expertise). The authors conclude that growing plants in the same community as their parents did reduces competition.

    However, I am not convinced by the interpretation of the results. Particularly the results for competition versus overall yield are in conflict. This discrepancy is not properly discussed and is largely ignored in the conclusions. Hence, I doubt whether the results support the conclusions.

    My most important comment relates to the discrepancy between results for total yield (Figure 3b) versus those for competition (Figure 2a) and for net biodiversity effect (Figure 3a). Results for all those measures are based on yield records. Figure 2a and 3b (panel fertilizer) show clearly that plants that have the same coexistence history as the tested plants outperform those having a different co-existence history. Figure 3b, however, shows no effect of coexistence history on yield; total yield for Same and Different do not differ. How to reconcile these results? Remarkably, this discrepancy is not discussed at all; the discussion largely ignores the absence of an effect on total yield.

    Related to the previous comment, the title includes the phrase "reduces competition". In the manuscript, competition is derived from effects on yield. Still, there is no benefit of the same coexistence history for total yield. This is somewhat misleading.

    A second important comment relates to the absence of results from the second year. The Methods section explicitly states that the comparisons made in year 3 (as shown in Figure 2) were also made in the second year (2018; L350-358). However, no results are presented. Why are those results excluded?

    A third comment relates to the distinction between competition and facilitation (Equations 3 and 4, and corresponding results), which is artificial and not very meaningful in my opinion. Since RII will never be precisely equal to zero (i.e., the RII=0 category is empty), an increase (decrease) in facilitation must go together with a decrease (increase) in competition, and vice versa. This must be the case since the total of both categories must add up to the number of comparisons made. In other words, if we have a total of N objects, being either apples or pears, then, if we have fewer apples, we must have more pears. (hence, L93-94 is a tautology). I suggest dropping this distinction from the manuscript.

    The Discussion seems to ignore some of the results that don't seem to match the "desired" outcome. For example, L178 speaks about niche differentiation as if this was found, but it was not. Same for L200. Similarly, L181 speaks about "the yield benefit", which was not there.

    While the manuscript is well written with respect to the language, it is not always easy to follow and absorb. This is partly because the number of traits is large. A table with the traits could help. Also, the writing could be improved to help the reader get the message. For example, when showing results in Figure 2, it could be mentioned from the start that these are relative to single plants, whereas those in Figure 3a are relative to monoculture. This can be found in the methods but should be clear from the Results as well.

  7. eLife

    Reviewer #4 (Public Review):

    In this paper, the authors conducted a multi-year study in which they grew six crop species either in monoculture or mixture to find out whether species could adapt to growing in mixture. They found that there was a shift within species to different strategies for growing in mixture vs. monoculture, which increased the yield. Contrary to expectations, they found convergent trait selection and a reduction in overall trait variation. They discuss this surprising finding in depth and connect it well to previous and ongoing research. The authors conclude that crops that so far are mainly cultivated and bred in monocultures could instead be bred for higher yield in intercropping set-ups (thus mixtures of different crops at the same time in the same place).

    This research is original and novel. Previous research had shown similar effects, but for non-crop plant species. Extending those previous findings to agriculture is highly relevant and timely. The experiment was conducted carefully. Not only was total yield measured, but also a series of traits were assessed to find out how the trait space changed over the three years the experiment was conducted. The manuscript is very well written. The abstract summarizes the results and implications of the study very well. The introduction set up the study very well and the discussion is appropriate in length and depth.

  8. Version published to 10.1101/2022.01.14.476288v1 on bioRxiv