Quantitative modelling of biological response dynamics reveals novel patterns in plant volatile signalling

Reviewer #1 (Public review):

Summary:

The manuscript by Waterman et al. describes the development of a mathematical model that quantifies plant volatile emissions dynamics in response to mechanical/biotic stress. Model outputs were based on volatile emission measurements from maize plants using PTR-MS. Modeling revealed differences in emission patterns dependent on the intensity of wounding damage, application of herbivore oral secretions, age of leaf, circadian clock, and genotype. Differences were also observed between different types of volatiles, and the response curves somewhat correlated with expression patterns of biosynthetic genes. Moreover, the model showed priming effects from overlapping response curves upon multiple wounding events.

Strengths:

As a non-expert in modeling, this reviewer assesses the work from a broader point of view. Overall, I consider this model to be useful for other researchers to quantify volatile emission dynamics for their plant system. Generating the models does not seem to be overly complicated as long as emissions can be measured with a real-time system such as PTR-MS, which is costly and not available to every lab. The advantage of this approach is that it does not rely on parameters of underlying enzymatic pathways or transport processes. The authors claim that it can be easily applied to other biological responses.

Weaknesses:

The manuscript lacks a deeper discussion of how the model can help make predictions of volatile emission dynamics from plants in the greenhouse or field. Can the model be trained and validated with volatile measurements from plants under different environmental conditions? How realistic is this approach given the complexity of a field environment? It would be helpful to provide a better outlook of the application of the model for scientists in the field of plant volatile biology and beyond.

The authors state that "emissions can be regulated independently of each other" (Line 359). I would assume that regulatory mechanisms in different genotypes are similar but show genotype-specific variation.

Reviewer #2 (Public review):

This is a study of the dynamics of plant volatile emissions, using a curve-fitting approach to describe salient properties of the dynamics of plant volatile chemicals. The study is interesting and unique in taking this approach. Some of the dynamics uncovered (e.g. lagged emission of many sesquiterpenes) are already well known using less sophisticated approaches, while other properties (diurnal cycles in emission dynamics) are newly uncovered. The approach in general is new for the topic of plant volatile emissions, but curve-fitting is widely used to describe the dynamics or function-valued responses of plants and other organisms. The study thus reads as rather methods-focused, giving tidbits of interesting properties of the dynamics of plant VOCs rather than being structured strongly around clear biological hypotheses. The method seems like a logical and robust way to analyze the dynamics of plant VOCs. I believe the impact of the work will largely depend on whether there are substantial and meaningful outcomes (for herbivores, downstream processes of induction, etc) due to the differences in VOC dynamics described via these methods that would be hard to observe in other ways. If so, there will be a need to adopt robust methods such as this to describe the salient features of those dynamics. At present, I do not believe there is evidence one way or another as to whether the subtle differences in VOC dynamics have large consequences.

The paper sells itself as describing a new technique for describing response curves generally across biological systems, but it only uses this technique to look at the dynamics of induced plant volatiles. I believe to show general utility of this approach, a wider range of examples of plastic responses to stimuli across organismal groups would be needed. I am, however, convinced that this approach is both novel and useful within the scope in which the examples are shown (i.e. in describing the dynamics of induced plant responses). Some of the text purporting novelty in uncovering shared and divergent responses across the tree of life seems pretty overstated.

Much of the introductory and discussion text is quite broad, and I wonder if the technique is really meant to be applicable to the specific case that is described (repeated measures of an induced volatile response). Likewise, there has been considerable work in such realms as behavioral science, function-valued traits (e.g. Stinchcombe et al 2012), performance curves (Kingsolver various papers), etc to describe dynamic or variable responses phenomenologically, and there are approaches including GAMs, parametric curve fitting, and other techniques that probably report the same salient features as the approach here. Indeed, there are already statistical techniques to assess the macroevolution of response curves (e.g. Goolsby 2015) and wide discussions as to how to compare function-based responses among organisms (The Functional Phylogenies Group 2012). So in the broad scheme of biology, I am not sure I'm convinced of the novelty of the approach. However, I believe it is novel within the context in which it is used here. The salient part of the methods is that it uses predefined attributes of dynamics (onset, duration, etc) based on a gamma distribution that the researchers (with good reason) believe to be biologically meaningful. This is in contrast to multivariate approaches (e.g. Izem et al 2005) that attempt to find salient dynamic features in a less constrained way.

I would have liked to see a clear description of model fits (e.g., how much of the variation in the real data is described by the fitted model). This seems important because there are quite a number of constraints placed on model fitting - so presumably when a model blind to those constraints picks unrealistic parameters, that would suggest that the constrained model probably does not fit the data all that well.

I am curious about the normalization process in the 'normalized emission' that is analyzed throughout the study. Normalization to leaf size makes sense, though I was less clear about L459: "Additionally, values were normalized to the maximum response observed in each experiment, yielding a range of positive values < 1." Why was this needed? Is the 'maximum response observed in each experiment' across all plants/compounds/treatments or within a single plant? In general, is there a way of reporting VOC emission rates in absolute values (e.g. umol / Liter air)? Normalization would presumably not impact most curve properties very much, but it could have effects on 'integral', and the need for within-experiment normalization would suggest a lack of transferability or comparability among datasets from different experiments (at least as regards 'integral'), which is suggested as a major advantage of this approach in the discussion.