Exploring natural odour landscapes: A case study with implications for human-biting insects

Reviewer #1 (Public Review):

This study assesses the volatile profiles from the hair and bodies of 64 vertebrate species to compare odor constituents across taxa. Compared to a similar data set for floral volatiles, the study suggests that vertebrate odors are significantly less diverse and show little phylogenetic relationship regarding profile similarity. Human odors were particularly unique from other species. It is concluded that this may influence the odor coding of organisms (like vectors) who respond to these odors compared to plant-feeding organisms like most other insects. While the study is compelling, several methodological issues leave the conclusions less convincing. It is suggested that the paper be tempered accordingly with these issues mentioned.
The study makes several assumptions about the methodology to be considered when interpreting the results:

Major Concerns:

1. Body hair as a proxy for animals. "Hair odour is likely a reasonable proxy for mammalian body odour, but may lose some volatile compounds during storage. Live-animal odour, on the other hand, can be contaminated with compounds from faeces or urine occasionally excreted during sampling." The study has addressed this by testing hair against the bodies of 4 humans, two rats, and one guinea pig (Figure S1). However, the results show that there are both quantitative and qualitative differences among all the samples. While the presence of waste accounts for some of this variation, this, too, is a natural response of the animal and could be present in natural settings. Also would not body heat in mammals have an impact on odors? The authors should support this. While this does not require reanalysis, the authors should address these differences, particularly when qualitative and quantitative differences are discussed heavily in the results.

2. Sampling medium: Tenax TA was used to sample the vertebrate odors. Please note that any sorbent will exhibit specificity regarding selectivity and sensitivity to VOCs. See https://www.eva.mpg.de/documents/Elsevier/Marcillo_Comparison_JChromA_2017_2452774.pdf for one comparison. For example, it is not surprising that "Aldehydes, ketones, alcohols, aromatics, terpenes, and hydrocarbons dominated" the samples given that these types of compounds are well retained by the Tenax polymer:
   https://www.sisweb.com/index/referenc/tenaxta.htm
   Many chemical ecology studies will employ multiple polar and non-polar polymers to retain different VOCs for better profile comparison.
   By itself, this limitation must be noted. However, it becomes even more relevant when compared to the floral volatile study, which used a different sorbent (Poropak) which is less hydrophobic and may retain more polar compounds than Tenax: https://hero.epa.gov/hero/index.cfm/reference/details/reference_id/2859526
   Such differences must be considered when comparing these two datasets, particularly when the study makes conclusions about their differences. Alternatively, a small set of experiments with poropak and a few species (like for the hair vs. body control experiments) could clarify the effect of sorbent type on VOC retention.

3. Sampling quantification: The methods note that " All extractions were run for 5-80 minutes depending on the expected odour concentration of the sample." What does this mean? Such differences in sampling timing in our lab have shown profound differences in the type and amount of volatiles collected. Generally, it is best to sample for as long as possible to ensure that the most volatiles are collected (up to 24 hours if possible). The compounds will eventually come into equilibrium with the sorbent. However, for quantification, the timing must be calibrated carefully, usually by using a representative set of likely compounds with different functional groups to determine the optimal length of sampling time. Was this done in this case? If not, how can one account for the significant variation in sampling time regarding quantification?
   A second issue with quantification is the need for an internal standard. Even with robotic assistance, slight variations in processing can significantly affect the quantity of volatile retained through detection at the MS. This is generally avoided by using an internal standard in the sampling arena. See this example with multiple sampling techniques (also employing TD-GCMS): https://www.frontiersin.org/articles/10.3389/fevo.2021.607555/full
   Without these methodological controls, it is unclear how effective quantification can be performed. It might be more prudent to confine the results to qualitative discussions.

Reviewer #2 (Public Review):

Summary:
Zung et al. use a comparative approach to examine the volatile headspace of diverse mammals and host species to understand the differences in chemical profiles that may provide mosquitoes with signatures of appropriate hosts. The authors collect the volatiles from hair samples and conduct qualitative analyses of the headspace composition. The authors' results suggest that mammals share overlapping volatile signatures, although the sampling method and statistical approaches reduce the veracity of the authors' findings. Additional comparisons between mammalian and floral odours were conducted, although the datasets were limited.

The inter-species comparisons will be helpful in the field, although the data pipeline and approaches may underestimate the headspace chemical diversity, and sampling artifacts and contaminants occur in the datasets, which further weakens the study's findings.

Strengths:
The comparative approach is a strength of the manuscript. The authors identify an important gap in mosquito natural history by attempting to characterize the odours from various mammalian, bird, and reptile species that mosquitoes may use as blood hosts. Although others have compared the skin volatiles of humans, apes, and ungulates (Verhulst et al., 2018, not cited in the current manuscript), Zung and coworkers expand this sampling by using hair samples from collections and zoos. Unfortunately, the sampling approach leads to potential artifacts associated with the collected volatiles and statistical analyses.

Weaknesses:
There are three major points of weakness associated with the manuscript: (1) sampling approach and analysis pipeline; (2) statistical analyses; and (3) premise and prior work.

1. Sampling approach and pipeline
A. The authors have described their sampling and analysis as quantitative, but they use a qualitative approach by not quantifying their samples and using a low-res MS. I outline several approaches that would allow the authors to quantitate their samples. The authors must run synthetic standards for peak verification (the mass spectra alone are insufficient for compound identification). The authors are also encouraged to run the standards in a concentration curve to allow quantification of the compounds. The authors have only tentatively identified 120 compounds. Using an autosampler and standard analyses in the software, the authors could easily quantify their samples which would take less than a week's time (this is not impossible, as the authors state in the methods). Based on the volatile fragmentation and the MS detector, the compounds will differ in their relative abundances - running calibration curves, co-injection of authentic standards, and using multiple column types are necessary for the resulting statistical analyses to prevent mischaracterization of the abundances in the hair samples. Using an internal standard, by spiking the Tenax before collection, would also allow determination if column conditions change over the course of the experiment. These measurements would provide some quantitative measures to explore the differences in host odors. Details on these approaches can be found in Methods in Chemical Ecology, Techniques in Pheromone Research, and article reviews that describe more recent approaches and analyses (Tholl and Rose, 2006; Stashenko and Martínez, 2008; Spicer et al., 2017; Tholl et al., 2020; Eisen et al., 2021; Schulz and Mollerke, 2022).

B. Abundant contaminants in the samples. In the supplemental table of partially identified compounds, many contaminants are associated with the headspace collection method and environmental contaminants. Under thermal deadsorption, Tenax degradation produces many compounds, including quinolones and benzenoid compounds. Phenyl-substituted carbonyl compounds (benzaldehyde, acetophenone, benzene acetaldehyde) are formed as artifacts from the oxidation of Tenax with environmental contaminants. Other compounds, like phenol or -ethyl and methylated benzene compounds, are known to be released from the Tenax traps. The authors' pipeline and blank subtraction should have identified these compounds.

C. Hair and live headspace volatiles. I appreciate the authors' experiments comparing the composition and abundance of volatiles from live collections and hair samples. However, the results demonstrate that the hair does not always match the volatiles from the live animal. Humans 1, 3, and 4 differ significantly in their aldehyde abundances, especially nonanal. The hamster and mice samples also differ significantly. The matrix of the hair will adsorb and modify the emissions and ratios of compounds, which makes the inter-species comparisons difficult if not impossible if the headspace collection approaches differ. The authors need to change their phrasing of the host odours to "hair odours", and soften their statements associated with the complete host odour profile, and use hair samples as a standard matrix for the headspace collections. The comparison of human odour collections relative to hair samples is like the comparison of apples and oranges.

D. The authors need to use another column type to characterize their peaks further. Some of the compounds are enantiomers or closely elute from the column. Although the authors suggest their methods may separate these compounds, they may be misidentified without a different GC temperature ramp or column.

E. The authors should replace their retention indices with KRI values to further identify their compounds. The methods section does not describe whether the alkane standards were run parallel to the hair samples, and the manuscript's retention indices do not match published KRI values.

F. The number of compounds across species (including flower compounds) is very low (approximately 120 compounds) and surprising. This suggests that the analysis pipeline and thresholding may miss many compounds in the headspace. I would encourage the authors to lower their threshold to 10^-5 AU, or to perform a sensitivity analysis on their ability to identify the peaks. Running authentic standards would also allow the identification of compounds missed in the analysis.

G. I understand the difficulty in obtaining these samples across the different species. However, additional information is needed for those species that are limited in the number of replicates (individuals). Sampling the individual multiple times may indicate the variability in the hair volatiles. Although the authors and many others have shown the reproducibility of human skin volatiles through time, additional sampling would indicate this also occurs for other mammals while strengthening the authors' approach.

H. An important measure of natural odour statistics is the odor emission rates, and normalizing across samples by the sample mass. More information on the methods would have clarified these aspects. It needs to be clarified why the samples were collected for different time periods (5 to 80 minutes). The sample mass for each specimen should also be included as this would allow normalization by time and mass, and should be described in the methods. This would allow quantitative measurements of the samples.

I. A critical missing component in the headspace is the acids. Tenax does not perform well at collecting these compounds. However, Gerstel Twisters and other collection matrices can capture those compounds. The authors must use these other collection methods to sample the hair specimens and identify those compounds to include in their table and analyses. Without this information, the manuscript lacks a critical dimension in the human odour landscape that is critical for mosquito attraction.

2. Statistical Analyses
A. Sampling effort and the replicate numbers used in the analyses is an important consideration that the authors do not address, but should be discussed in more detail. In many subfields of chemical ecology, a minimum of ten replicates per species has been suggested to accurately identify the composition of compounds, and even with ten samples, this may not be enough to characterize the volatile profile (Raguso and Pellmyr, 1998; Campbell et al 2019). The authors could perform a power analysis, or an accumulation curve to represent the needed sample number to identify the number of compounds in the hair headspace accurately.

B. It would be worthwhile for the authors to provide more detail on their supervised and unsupervised approaches, and how their data fits the assumptions of the analyses. The PCA parametric method may require log or square root transformation of the data to make residuals fit the normality assumption, but it's unclear if this was the case with the authors' datasets.

C. PCA is also not appropriate when many samples have zero values in the data matrix, which occurs in the authors' data. In such a case, the approaches of NMDS or canonical analysis of principal coordinates would be more appropriate, and allow distance measures (the Bray-Curtis distance) to define dissimilarity of different groups. An analysis of similarity (ANOSIM) could be used to determine if the data clustered significantly by species or by mosquito host.

D. The authors are encouraged to use alternate approaches, such as random forest (ML) approach, to determine if the odor classification is based on host or non-host. This method has been used for the last fifteen years in chemical ecology and human odor analysis (Cutler et al, 2007, Kwak et al 2008).

E. The authors use a phylogenetic framework for their analyses. Multivariate methods are now available to test evolutionary hypotheses about scent composition in a phylogenetic framework (Goolsby, 2017), and the authors are encouraged to use these approaches.

F. Comparison to floral odour space section. I would encourage the authors to examine other datasets of plant headspace samples, including plants used by mosquitoes. There are many datasets out there that the authors could use (El-Sayed 2021, Farré-Armengol et al 2020). Expanding the authors' dataset would provide more statistical power, and provide control of differences in plant visitor and plant phylogenetic relatedness.

G. Adding context related to mosquito olfaction. The authors describe how their work could provide insight into the coding of olfactory information by the mosquito. I would encourage the authors to analyze their data further by collapsing the host volatiles into groups based on biochemical pathways, or knowledge of the detection of the volatiles by the mosquitoes (such as using electroantennogram responses) to filter and identify only those responsive volatiles to keep in their dataset.

Premise and Background Knowledge
A. Analyses of odour headspace have been known for the last three decades, e.g. (Methods in Chemical Ecology, Techniques in Pheromone Research, George Petri's work, Tholl and Rose, 2006; Stashenko and Martínez, 2008; Spicer et al., 2017; Tholl et al., 2020; Eisen et al., 2021; Schulz and Mollerke, 2022). But in many places, the paper conveys the impression that these are new discoveries and analyses. For example,
-"Yet we remain remarkably ignorant of the composition of the chemical world."
-"Our work provides one of the first quantitative descriptions of a natural odour space"
-"Progress in understanding natural odours has also been hindered by the technical challenges of capturing and analyzing odour, especially the complex blends that constitute most natural odours"
The Introduction and Discussion are rife with these overblown statements. I found this frustrating as the authors were not giving due credit to prior work on that topic while (maybe unintentionally) giving an impression that this specific idea was a new contribution. More care is needed to delineate which aspects of the study are 1) based on prior understanding, or 2) totally new). The authors are adding to an already extensive field of chemical ecology and olfactory processing of mixtures, and are contributing to this knowledge by adding datasets related to mammalian odor. I plead that the authors clearly describe these gaps, and place their results into proper context.

B. Similarly to the above statements relating to chemical ecology, the authors have numerous statements about gaps in odour processing. Mixture processing has been an important topic of study for the last forty years (Shorey, 1973, Caprio, 1988, Riffell et al 2009, Su et al 2009, Rokni et al 2014, Mathis et al 2016), which is based on encoding the temporal and concentration-dependent statistics of the odour.
-"Yet compared to visual and auditory scenes, we know very little about the statistics of natural olfactory scenes"
As described above, this is surprising and frustrating because of the rich literature on these topics (searching for "odour mixtures" provides 32,000 articles). In their manuscript, the authors are providing a strawman argument for their analyses by focusing on single odorant signatures, when the literature has repeatedly demonstrated the importance of odour mixtures for behavior and combinatorial processing.

C. There are increasing studies examining the mosquito behavioral and electrophysiological responses to hosts and other odours. However, this literature is not cited or included in the authors' analyses. The chemical ecology of mosquito attractants and natural odours has been studied in the Carde, Leal, Ignell, Carlson, Kline, Riffell, Takken, Torto, Verlhurst, Vosshall labs, and many others. The authors could use this information in their analyses and cite the literature.

Reviewer #3 (Public Review):

This study focused on collecting and analyzing odour samples from a wide range of vertebrate species to understand the composition and characteristics of vertebrate body odours. The researchers used dynamic headspace sampling to collect odour samples from 120 individual animals representing 64 vertebrate species. They collected odour from both live animals and hair samples, with hair being a reasonable proxy for mammalian body odour.

The odour samples were analyzed using thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS) to identify compounds and estimate their abundance. They identified a total of 116 compounds in the vertebrate odour extracts, including aldehydes, ketones, alcohols, aromatics, terpenes, and hydrocarbons. The compounds varied in prevalence across species, but a large number of compounds were found in at least 15 samples, indicating a broad overlap in odour composition among vertebrates.

The study compared the vertebrate odour space to floral odour space and found that vertebrate odours shared more compounds compared to floral odours. Floral odours tended to be less complex and more likely to contain unique compounds found only in a single species. The analysis also revealed that odour profiles did not show strong phylogenetic signals, indicating that closely related species did not necessarily have similar odour profiles. However, within-species clustering was observed, suggesting that body odour composition may be species-specific.

The researchers also investigated specific compounds that could serve as host-seeking cues for animals. They compared the odour of live vertebrate hosts to non-host stimuli and identified straight-chain aldehydes as abundant compounds in vertebrate odours. These aldehydes were found at substantially lower levels in non-host stimuli. Additionally, when comparing human odour to other vertebrate species and non-host stimuli, several compounds, including decanal, sulcatone, geranylacetone, and undecanal, emerged as strong predictors of human hosts.

Three shortcomings of the study can be highlighted:
1. Undersampling of certain compound classes: The study acknowledged that they undersampled carboxylic acids, which are generally too polar or non-volatile to be analyzed without a special derivatization step. This limitation could have resulted in an incomplete understanding of the full range of compounds present in vertebrate odours.
2. Missing highly volatile compounds: The study mentioned the difficulty of capturing and quantifying highly volatile compounds reliably. This limitation suggests that certain compounds with high volatility may not have been adequately represented in the analysis, potentially impacting the comprehensiveness of the odour space.
3. Lack of controlled experiment for species replicates: Although the study observed strong within-species clustering for some species in their dataset, they cautioned that many of the species replicates came from the same farm or zoo, which could confound the results with sample origin. The lack of a well-controlled experiment limits the generalizability of the findings regarding consistent and characteristic odour profiles across animals.

These shortcomings should be considered when interpreting the results of the study and could be addressed in future research to further advance our understanding of vertebrate body odours.

The manuscript highlights three open questions. First, the authors discuss the implications of the differences between vertebrate and floral odors for olfactory coding in blood feeders and floral visitors. Specialist mosquitoes require odor blends to detect hosts, while honeybees can generalize from attractive mixtures to individual components. The authors suggest that these differences may be influenced by the different odor spaces mosquitoes and bees inhabit.

Second, the authors note that although compounds in vertebrate odor are shared broadly across species, they are also common in other natural odors. This poses a challenge for generalist blood feeders, but the study suggests that straight-chain, saturated aldehydes, which are highly abundant in vertebrate odors, may still serve as useful indicators. These aldehydes have been shown to enhance host-seeking in mosquitoes and are even used by malaria parasites and orchids to attract mosquitoes. However, the study did not capture highly volatile or polar compounds that may also indicate the presence of a vertebrate host.

Third, the manuscript discusses the lack of phylogenetic signal in the odors of mammals, which make up the majority of the sampled species. This may explain why few mosquitoes exhibit preferences for taxonomic groups at the family or order level. The study suggests that within a species, there is high consistency in odor-blend composition, which may mediate species-specific host preference through olfactory cues.

The authors also focus on odor features that may serve as valuable cues for human specialists. They find that certain components of human odor, such as sulcatone, geranylacetone, decanal, and undecanal, are distinctive and enriched in human odor. Undecanal, despite being less common across non-human animals and in nature overall, is a more reliable indicator of human odor than decanal. The two ketones are even more reliable indicators. The authors speculate that the reliance on aldehydes by human-specialist mosquitoes may be due to the evolutionary history of these mosquitoes, which arose from an ancestral generalist subspecies.

In conclusion, this manuscript presents a quantitative study of vertebrate animal odors, highlighting the differences between vertebrate and floral odors. It raises questions about olfactory coding in blood feeders and floral visitors, the challenges faced by generalist blood feeders, and the lack of phylogenetic signal in mammalian odors. The study also explores odor features that may be valuable cues for human specialists and discusses the evolutionary implications of these findings.

Author Response

We thank the editors for their care in handling our manuscript. We also thank the reviewers, especially reviewer 2, for their thorough comments. We will work to address their concerns in a revised version and provide some initial comments below.

A major concern of two reviewers was that odour profiles were not quantified rigorously. We acknowledge that our study does not achieve the level of quantitative rigour standard in most chemical ecology work. We plan to conduct a few additional analyses to help address this shortcoming. We will also adjust the text to clarify the semi-quantitative nature of the data.

Reviewers also suggested using several different analytical approaches (e.g., different column, different sorbent) to broaden the type and number of detectable compounds. The reviewers rightly point out that such choices strongly affect which compounds we are likely to sample. No single approach is comprehensive, and ours is no exception. We will work to ensure that the appropriate caveats are included prominently in the text.

However, we believe this concern in fact underscores a special strength of our study: analysing the odour of a large number of species in a single study using the same analytical approach, so that the inherent biases of different approaches do not complicate cross-species comparisons. We are aware of very few such large-scale studies in any system and welcome suggestions from reviewers or readers of any we might have overlooked.

In general, we believe many of the reviewers’ methodological concerns reflect standards in the field of chemical ecology established for studies that aim to describe the odour of one or a few species as comprehensively as possible with a high level of quantitative rigour. This was not our goal, and we will temper our language in the revised paper to make that clear. Instead, we aimed to sample as broadly as possible across species to gain insight into the general statistics of a large 'odour landscape' or 'odour space' — an endeavour that, to our knowledge, is less common in the chemical ecology literature. In doing so, we prioritized breadth over depth. We believe the resulting dataset provides solid evidence for our major conclusions, though we will revisit our analyses and conduct a small number of additional experiments to further substantiate our claims.