2.1. Terpene chemistry and chemotyping of wild G. hirsutum populations A correlation analysis (Fig. 2) performed on the relative abundance of mono- and sesquiterpenes in the leaf solvent extracts revealed two distinct highly supported (approximately unbiased (AU) p < 0.05) groups of monoterpenes. One group comprised the monoterpenes γ- terpinene, limonene, α- thujene, α- terpinene, terpinolene, and p-cymene (the “γ- terpinene group”, designated chemotype class A), while the other consisted of the pinene-type compounds α- and β- pinene (the “α- pinene group”, designated chemotype class B). The sesquiterpene compounds α- humulene and β- caryophyllene were also strongly intercorrelated. These strong correlations suggest a shared or highly linked biosynthesis. A theorised pathway of the synthesis of these compounds highlights their shared biochemical precursors (Degenhardt et al., 2009, Fig. 3). Taking the relative abundance of only monoterpenes, a negative relationship was observed between the proportions of γ- terpinene and α- pinene groups with two clusters of plants identified (Fig. 4a). Representative chromatograms from plants belonging to chemotype classes A and B respectively can be seen in Fig. 4b. Chemotype class A was composed of plants where the γ- terpinene group compounds made up 4% or more of the total monoterpene concentration; chemotype class B comprised plants where the γ- terpinene group compounds made up less than 4% of total monoterpenes (Fig. 4c). The major difference is the strong reduction or absence of the γ- terpinene cassette of monoterpenes in chemotype class B. In a principal coordinate analysis (PCoA) on relative abundances of monoterpenes, with samples separating based on their compositional similarity (using a Euclidean dissimilarity matrix), the chemotype classes separated along the first axis (Fig. 5), which explained ca. 65% of the variation. Plants belonging to chemotype class B group together, while plants in class A are spread along the first axis. Around 42% of overall variance in the monoterpene profiles of the plants was explained by chemotype class (adonis: R2 = 0.419, p = 0.001). Most modern varieties of G. hirsutum produce and store a range of mono- and sesquiterpenes, including α- and β- pinene, limonene, camphene, β- myrcene, α- humulene, and β- caryophyllene (Elzen et al., 1985; Minyard et al., 1965; Yang et al., 2013), and plants producing significant amounts of γ- terpinene are rare. Only one naturalised variety has been reported to emit γ- terpinene in large quantities (Loughrin et al., 1995). Although modern cultivated varieties came from diverse landraces, this could indicate that the main sources for domestication events of G. hirsutum were obtained from B chemotype wild plants, or that perhaps the artificial selection throughout the domestication process selected against chemotype A plants due to pleiotropic interactions or linkage with traits of interest. The large diversity of terpenoid compounds found in plants is due to terpene synthases (TPSs), which are a diverse family of enzymes that catalyse terpenoid compounds from single substrates (Karunanithi and Zerbe, 2019). The genes GhTPS1 and GhTPS2, identified and characterised in Gossypium hirsutum (cultivar CCRI12) (Huang et al., 2013), encode for active TPSs that are expressed in young developing leaves, and may be involved in the production of constitutive terpenoids that are stored in glands. It has been found that GhTPS1 mainly produces the sesquiterpenes β- caryophyllene and humulene at a ratio of ~4:1. GhTPS2 was found to mainly produce the monoterpenes α- pinene and β- pinene, at the ratio of ~6:1. These sets of compounds were highly correlated in our dataset, with β- caryophyllene and humulene being present at a ratio of 3.8:1 (SE =0.013), and α- pinene and β- pinene at a ratio of 5.3:1 (SE = 0.027). The highly similar product ratios between our dataset and those described in Huang et al. (2013), suggests that GhTPS1 and GhTPS2 contribute to the production of these compounds in wild G. hirsutum. Nonetheless, other TPSs may also be involved, as multiple G. hirsutum TPSs have been shown to produce several terpenes at varying levels (including the compounds that comprise the “γ -terpinene group” (Huang et al., 2018)). Eighty-five TPS genes have been putatively identified in G. hirsutum to date (Zhang et al., 2022). Our results suggest that one or more enzymes present or overexpressed in plants of chemotype A are multifunctional TPSs that synthesise the monoterpenes of the γ- terpinene group. 2.2. Association between chemotype and geographic location The correlation between geographic location of the sites and chemical distance was not very strong but significantly different from zero (Mantel test, r = 0.165, p = 0.001), indicating that as geographic distance increased, the chemical profiles of the plants became more dissimilar. An association between site location and γ- terpinene content was observed, with more plants in western sites containing low amounts of compounds belonging to the γ- terpinene group, and plants in eastern sites containing higher proportions of γ- terpinene group compounds (Fig. 6). Future studies could examine the genetic structuring of these populations to determine if these chemotypic differences are related to genetic differentiation (i.e., isolation among populations) and/or adaptation to their local environment. 2.3. Chemotypes of mature plants In addition to the plants grown from seed, we also assessed the chemical profiles of 165 mature plants from which a subset (34) were also maternal plants. Thirty-seven compounds were detected in the mature plants, with 20 compounds identified as MTs, 10 SQTs, 4 GLVs, 1 hydrocarbon, and 2 nitrogen-containing volatiles (see Table S3). Mature plants were classified using the same chemotyping process as described above. PcoA ordination of all mature plants harvested in situ shows grouping by chemotype class, separating along axis 1 (Fig. S1; variation explained by first two ordinations: 82.6% and 15%). More than twothirds of the variation was explained by chemotype group (adonis: R 2 =0.698, p =0.001). Terpenoid production is known to be affected by a variety of environmental stresses, including exposure to extreme temperatures, water stress (flooding, drought), and salinity, as well as herbivory (Gouinguene´and Turlings, 2002; Isah, 2019). A differential regulation in terpenoid production because of environmental pressures could explain why chemotype explained a higher percentage of total variation in the chemical profiles of the mature plants compared to plantlets; abiotic and biotic stress are known to affect the levels of VOCs in plants (Holopainen and Gershenzon, 2010). Plant ontogeny can also influence the terpenoid profile. Although the age of the mature plants was unknown, they had been surveyed for several years and were well established flowering plants. Meanwhile, offspring plants were seedlings with only four developed leaves (around one month old), and it is possible that mature G. hirsutum plants exhibit a more defined chemotype. For example, in Melaleuca alternifolia, six terpene chemotypes can be clearly differentiated in mature trees, however young plants only have four identifiable chemotypes, and two chemotypes with high terpinolene concentration are indistinguishable in young plants (Bustos-Segura et al., 2015). 2.4. Heritability of chemotypes Most of the progeny had the same chemotype as their parent plant. Offspring of plants with chemotype class A matched their maternal plant 80% of the time (63/79 plants), and plants in chemotype class B matched their maternal plant almost 93% of the time (64/69). In wild G. hirsutum populations, both self-pollination and cross-pollination have been described by Wegier et al. (2011). Vel´azquez-L´opez et al. (2018) reported an outcrossing rate of 0.72 for a metapopulation of wild G. hirsutum plants from the Yucatan Peninsula, indicating that although the level of self-pollination is important, there is also a large contribution from cross-pollination to the next generation. Thus, as we only knew the chemotype of the maternal plant and not the paternal plant, this could explain why there was some disparity between maternal and offspring chemotypes. Broad sense heritability estimated with the animal model was 0.85 (CI = 0.75, 0.95). In concordance with what we observed, chemotype is generally considered to be heritable in plants (Hare, 2011). In the tree species Eucalyptus globulus, O’ Reilly-Wapstra et al. (2011) found moderate to high broad-sense heritability of foliar terpenes, and Karban et al. (2014) found that in sagebrush (Artemisia tridentata) chemotypes are highly heritable between parent and offspring. 2.5. Concentration of leaf terpenoids To determine the degree to which biological and geographic factors can explain the variation in the concentration of compound classes, we used generalised linear mixed models. Approximately a quarter of the total variance (from 21 to 29%) in concentrations of total volatiles, total monoterpenes, total sesquiterpenes, and total green leaf volatiles was significantly explained by ‘genotype’ (Table 2), whereas ‘population’ explained only a small fraction of the variance and was not statistically significant. Concentrations of total volatiles and total monoterpenes were significantly higher in plants belonging to chemotype Class A (VOCs: χ 2 (1) = 4.9101, p = 0.027, MTs: χ 2 (1) = 8.298, p = 0.004). No significant difference was observed in total concentrations of sesquiterpenes or green leaf volatiles between chemotypes (SQTs: χ 2 (1) = 1.5618, p = 0.2114, GLVs: χ 2 (1) =1.646, p = 0.1995). At the time of harvesting, all leaves were at a similar developmental stage (4th leaf fully unfurled). However, leaf area showed considerable variation with a bimodal distribution (excess mass test, p = 0.004, Fig. S2); therefore, all analyses on concentration included leaf area as a covariate. Plants with smaller leaves contained significantly higher summed concentrations of all compounds found (summed GLVs, MTs and SQTs; χ 2 (1) = 184.9882, p < 0.001). Total monoterpene and sesquiterpene concentrations were significantly higher in smaller leaves (MTs: χ 2 (1) = 585.716, p < 0.001, SQTs: χ 2 (1) = 339.2854, p < 0.001), whereas total concentrations of green leaf volatiles were higher in larger leaves (GLVs: χ 2 (1) =51.733, p < 0.001). Opitz and colleagues had found that accumulation of terpenoids in glands in cotton was strongly influenced by leaf area and developmental stage, with the two youngest/ smallest leaves having the highest total terpenoid concentration (Opitz et al., 2008). They also showed that younger leaves have a much higher gland density than older leaves, but contain less total terpenoids per gland. Eisenring et al. (2017) made the observation that smaller leaves have a higher density of glands than larger leaves of the same age. We found that smaller leaves contained higher concentrations of MTs and SQTs and lower concentrations of GLVs in relation to larger leaves. 2.6. Ecological implications of chemical diversity in G. hirsutum The chemotype of a plant plays an important role in the outcome of environmental adaptation, and volatile-mediated plant-insect and plant-plant interactions. For instance, in Artemisia tridentata (sagebrush), plants with the same chemotype respond more strongly to volatiles emitted by each other than to volatiles from individuals with a different chemotype (Karban et al., 2014). Moreover, when Tanacetum vulgare (tansy) plants were challenged by insect herbivores from two feeding guilds, distinct chemotypes responded differently. Following caterpillar feeding, some chemotypes exhibit a stronger volatile response when also pre-treated with aphids, while the opposite was observed for other chemotypes (Clancy et al., 2020). In Thymus vulgaris (thyme), phenolic chemotypes show reduced tolerance to freezing compared to non-phenolic chemotypes (Thompson et al., 2013) and occur more frequently in a region that has undergone regional warming (where occurrence of extreme winter freezing events has dwindled). Accordingly, the observed geographic gradient in the frequency of the two identified chemotypes across wild G. hirsutum populations calls for further work testing whether the terpene chemotypes are associated with a pattern of local adaptation to biotic (herbivores and/or pathogens) or abiotic factors. There have been numerous investigations concerned with describing the volatile terpenoids found in G. hirsutum. While chemotypes have not, to our knowledge, been described in wild cotton to date, differences in the amounts of herbivory-induced plant volatiles (HIPVs) released by different cultivars have been recorded (Hagenbucher et al., 2016). Magalh˜aes et al. (2020) investigated VOC differences between seven G. hirsutum genotypes and found few qualitative and quantitative differences. R¨ose and colleagues (R¨ ose and Tumlinson, 2005) showed that cotton plants respond in specific ways to herbivory by systemically releasing distinct blends of volatile compounds, and suggest that plants that emit higher amounts of induced defence chemicals are better at defending themselves. Loughrin et al. (1995) showed that upon feeding damage by beet armyworm larvae (Spodoptera exigua), the naturalised variety TX2259 was found to emit much greater quantities of HIPVs than five other cultivated lines (almost sevenfold higher). Moreover, this variety released higher amounts of γ- terpinene. Thus, it would be interesting to explore if wild (and naturalised) cotton emits higher amounts of HIPVs than cultivated plants of G. hirsutum, and whether the constitutive terpene chemotype is also linked to quantitative differences in the volatile emissions. HIPVs in G. hirsutum have also been implicated in interactions with the third trophic level. Campoletis sonorensis and Microplitis croceipes (two parasitoid species) were found to respond to volatiles from undamaged cotton plants (Elzen et al., 1987), yet volatiles emitted by caterpillar-damaged plants seem to provoke a considerably stronger response (De Moraes et al., 1998; Turlings et al., 1995). Importantly, cotton volatiles are also involved in plant-plant signalling; studies have shown that cotton plants exposed to the volatiles of a nearby damaged plant are more resistant to herbivory (Llandres et al., 2018; Renou et al., 2011). Therefore, chemotypic variation could be a factor that not only affects the attraction of natural enemies of herbivores to damaged or undamaged plants but also be a key determinant of the effectiveness of airborne signalling between plants. 3. Conclusions In brief, we demonstrate that wild G. hirsutum plants from 16 naturally occurring populations along the Yucatan peninsula could be grouped into two chemotypes based on differences in their stored monoterpene profiles. The identification of chemotypes in wild G. hirsutum could have implications for cotton direct defences against herbivory, and indirect defences through the attraction of natural enemies or plant-plant signalling. We recommend performing experiments exploring the consequences of the reported terpene variation on the different levels of plant defence and its effects on the outcome of associated interactions. In combination with functional characterisation of terpene synthases in wild cotton chemotypes such research would provide valuable information into the chemical ecology of cotton and may reveal opportunities to enhance the control of cotton pests without the current excessive use of pesticides. 4. Experimental 4.1. Plant material Gossypium hirsutum Cav. (Malvaceae) is a perennial shrub that is native to Mexico and is distributed throughout the Caribbean Basin and Central America (d’ Eeckenbrugge and Lacape, 2014). It grows up to 2 m tall under natural conditions. Wild populations of the species can be found growing in coastal shrubland along the Mexican Yucatan Peninsula coast. Seeds from 16 populations of wild G. hirsutum plants located around the Yucatan Peninsula (Fig. 1) were collected in 2019 and 2020. From each population, approximately 100 seeds from six maternal plants were harvested at each site. Offspring from the same mother are either full or half-siblings and referred to here as a genotype. The seeds were shipped to the University of Neuchˆatel, Switzerland, where the experiment was performed between July–November 2020., Published as part of Clancy, Mary V., Mamin, Marine, Flückiger, Galien, Quijano-Medina, Teresa, P, Biiniza, erez-Nino ˜, Abdala-Roberts, Luis, Turlings, Ted C. J. & Bustos-Segura, Carlos, 2023, Terpene chemotypes in Gossypium hirsutum (wild cotton) from the Yucatan Peninsula, Mexico, pp. 113454 in Phytochemistry (113454) (113454) 205 on pages 2-7, DOI: 10.1016/j.phytochem.2022.113454, http://zenodo.org/record/8160599, {"references":["Elzen, G. W., Williams, H. J., Bell, A. A., Stipanovic, R. D., Vinson, S. B., 1985. Quantification of volatile terpenes of glanded and glandless Gossypium hirsutum L. cultivars and lines by gas chromatography. J. Agric. Food Chem. 33, 1079 - 1082. https: // doi. org / 10.1021 / jf 00066 a 015.","Minyard, J. P., Tumlinson, J. H., Hedin, P. A., Thompson, A. C., 1965. Isolation and identification, constituents of cotton bud terpene hydrocarbons. J. Agric. Food Chem. 13, 599 - 602.","Yang, C. - Q., Wu, X. - M., Ruan, J. - X., Hu, W. - L., Mao, Y. - B., Chen, X. - Y., Wang, L. - J., 2013. Isolation and characterization of terpene synthases in cotton (Gossypium hirsutum). Phytochemistry 96, 46 - 56. https: // doi. org / 10.1016 / j. phytochem. 2013.09.009.","Loughrin, J. H., Manukian, A., Heath, R. R., Tumlinson, J. H., 1995. Volatiles emitted by different cotton varieties damaged by feeding beet armyworm larvae. J. Chem. Ecol. 21, 1217 - 1227.","Karunanithi, P. S., Zerbe, P., 2019. Terpene synthases as metabolic gatekeepers in the evolution of plant terpenoid chemical diversity. Front. Plant Sci. 10.","Huang, X., Xiao, Y., Kollner ¨, T. G., Zhang, W., Wu, Junxiang, Wu, Juan, Guo, Y., Zhang, Y., 2013. Identification and characterization of (E) - β- caryophyllene synthase and α / β- pinene synthase potentially involved in constitutive and herbivore-induced terpene formation in cotton. Plant Physiol. Biochem. 73, 302 - 308. https: // doi. org / 10.1016 / j. plaphy. 2013.10.017.","Huang, X. - Z., Xiao, Y. - T., K ¨ ollner, T. G., Jing, W. - X., Kou, J. - F., Chen, J. - Y., Liu, D. - F., Gu, S. - H., Wu, J. - X., Zhang, Y. - J., Guo, Y. - Y., 2018. The terpene synthase gene family in Gossypium hirsutum harbors a linalool synthase GhTPS 12 implicated in direct defence responses against herbivores. Plant Cell Environ. 41, 261 - 274. https: // doi. org / 10.1111 / pce. 13088.","Zhang, C. - P., Zhang, J. - L., Sun, Z. - R., Liu, X. - Y., Shu, L. - Z., Wu, H., Song, Y., He, D. - H., 2022. Genome-wide identification and characterization of terpene synthase genes in Gossypium hirsutum. Gene 828, 146462. https: // doi. org / 10.1016 / j. gene. 2022.146462.","Gouinguen´e, S. P., Turlings, T. C., 2002. The effects of abiotic factors on induced volatile emissions in corn plants. Plant Physiol. 129, 1296 - 1307.","Isah, T., 2019. Stress and defense responses in plant secondary metabolites production. Biol. Res. 52, 39. https: // doi. org / 10.1186 / s 40659 - 019 - 0246 - 3.","Holopainen, J. K., Gershenzon, J., 2010. Multiple stress factors and the emission of plant VOCs. Spec. Issue Induc. Biog. Volatile Org. Compd. Plants 15, 176 - 184. https: // doi. org / 10.1016 / j. tplants. 2010.01.006.","Bustos-Segura, C., Kulheim, C., Foley, W., 2015. Effects of terpene chemotypes of Melaleuca alternifolia on two specialist leaf beetles and susceptibility to myrtle rust. J. Chem. Ecol. 41, 937 - 947. https: // doi. org / 10.1007 / s 10886 - 015 - 0628 - 0.","Wegier, A., Pi´neyro- Nelson, A., Alar´on, J., G´alvez-Mariscal, A., Alvarez-Buylla´, E. R., Pinero´, D., 2011. Recent long-distance transgene flow into wild populations conforms to historical patterns of gene flow in cotton (Gossypium hirsutum) at its centre of origin. Mol. Ecol. 20, 4182 - 4194. https: // doi. org / 10.1111 / j. 1365 - 294 X. 2011.05258. x.","Hare, J. D., 2011. Ecological role of volatiles produced by plants in response to damage by herbivorous insects. Annu. Rev. Entomol. 56, 161 - 180. https: // doi. org / 10.1146 / annurev-ento- 120709 - 144753.","O' Reilly-Wapstra, J. M., Freeman, J. S., Davies, N. W., Vaillancourt, R. E., Fitzgerald, H., Potts, B. M., 2011. Quantitative trait loci for foliar terpenes in a global eucalypt species. Tree Genet. Genomes 7, 485 - 498. https: // doi. org / 10.1007 / s 11295 - 010 - 0350 - 6.","Karban, R., Wetzel, W. C., Shiojiri, K., Ishizaki, S., Ramirez, S. R., Blande, J. D., 2014. Deciphering the language of plant communication: volatile chemotypes of sagebrush. New Phytol. 204, 380 - 385.","Opitz, S., Kunert, G., Gershenzon, J., 2008. Increased terpenoid accumulation in cotton (Gossypium hirsutum) foliage is a general wound response. J. Chem. Ecol. 34, 508 - 522.","Eisenring, M., Meissle, M., Hagenbucher, S., Naranjo, S. E., Wettstein, F., Romeis, J., 2017. Cotton defense induction patterns under spatially, temporally and quantitatively varying herbivory levels. Front. Plant Sci. 8.","Clancy, M. V., Haberer, G., Jud, W., Niederbacher, B., Niederbacher, S., Senft, M., Zytynska, S. E., Weisser, W. W., Schnitzler, J. - P., 2020. Under fire-simultaneous volatilome and transcriptome analysis unravels fine-scale responses of tansy chemotypes to dual herbivore attack. BMC Plant Biol. 20, 1 - 18.","Thompson, J., Charpentier, A., Bouguet, G., Charmasson, F., Roset, S., Buatois, B., Vernet, P., Gouyon, P., 2013. Evolution of a genetic polymorphism with climate change in a Mediterranean landscape. Proc. Natl. Acad. Sci. 110, 2893 - 2897. https: // doi. org / 10.1073 / pnas. 1215833110.","Hagenbucher, S., Birgersson, G., Chattington, S., Anderson, P., 2016. Domestication influences choice behavior and performance of a generalist herbivore. Perspect. Plant Ecol. Evol. Systemat. 23, 63 - 72. https: // doi. org / 10.1016 / j. ppees. 2016.09.001.","Magalhtaes, D. M., Borges, M., Laumann, R. A., Caulfield, J. C., Birkett, M. A., Blassioli- Moraes, M. C., 2020. Inefficient weapon - the role of plant secondary metabolites in cotton defence against the boll weevil. Planta 252, 94. https: // doi. org / 10.1007 / s 00425 - 020 - 03497 - w.","R ¨ ose, U. S., Tumlinson, J. H., 2005. Systemic induction of volatile release in cotton: how specific is the signal to herbivory? Planta 222, 327 - 335.","Elzen, G. W., Williams, H. J., Vinson, S. B., Powell, J. E., 1987. Comparative flight behavior of parasitoids Campoletis sonorensis and Microplitis croceipes. Entomol. Exp. Appl. 45, 175 - 180. https: // doi. org / 10.1111 / j. 1570 - 7458.1987. tb 01078. x.","De Moraes, C. M., Lewis, W. J., Par´e, P. W., Alborn, H. T., Tumlinson, J. H., 1998. Herbivore-infested plants selectively attract parasitoids. Nature 393, 570 - 573. https: // doi. org / 10.1038 / 31219.","Turlings, T. C., Loughrin, J. H., McCall, P. J., R ¨ ose, U. S., Lewis, W. J., Tumlinson, J. H., 1995. How caterpillar-damaged plants protect themselves by attracting parasitic wasps. Proc. Natl. Acad. Sci. 92, 4169 - 4174. https: // doi. org / 10.1073 / pnas. 92.10.4169.","Llandres, A. L., Almohamad, R., Br´evault, T., Renou, A., T´er´eta, I., Jean, J., Goebel, F. - R., 2018. Plant training for induced defense against insect pests: a promising tool for integrated pest management in cotton. Pest Manag. Sci. 74, 2004 - 2012.","Renou, A., T´er´eta, I., Togola, M., 2011. Manual topping decreases bollworm infestations in cotton cultivation in Mali. Crop Protect. 30, 1370 - 1375. https: // doi. org / 10.1016 / j. cropro. 2011.05.020.","d' Eeckenbrugge, G. C., Lacape, J. - M., 2014. Distribution and differentiation of wild, feral, and cultivated populations of perennial upland cotton (Gossypium hirsutum L.) in Mesoamerica and the Caribbean. PLoS One 9."]}