The four populations of D. mojavensis feed and breed on four different species of fermenting cacti, so we began our study of the role of olfaction in host plant shift by measuring the attraction of each population to their respective host plants across a range of fermentation stages. The purpose of these experiments was to determine at what stage(s) the flies are attracted to their own host cactus necroses, an essential step in identifying the volatile(s) underlying host specific behavioral attraction. Thus, we conducted two choice experiments using a behavioral trap assay system. For each fly population the following comparisons were performed using their own respective host plants: uninoculated (NI) vs. one week (W1) fermented cactus, one week (W1) vs. five week (W5) fermented cactus, five week vs. nine week (W9) fermented cactus and lastly one week vs. nine week fermented cactus. For the mainland Sonoran population, comparisons between uninoculated and five week fermented cacti are shown as no significant difference between uninoculated and one week fermented samples was found (data not shown).

In these behavioral choice tests, all four populations overall showed greater attraction to fermented rather than fresh (uninoculated) cactus tissue (Figure 1B–E). Additionally, preferences in fermentation stage varied among the four populations. The Mojave Desert and S. Catalina populations had clear preferences for earlier fermentation stages (Figure 1B, C; Table S1A, B). Both one week and five week stages were equally attractive to flies, and attraction to nine week old tissue was reduced for both cacti. The mainland Sonoran and Baja populations, on the other hand, were attracted similarly to all of the stages of fermentation (Figure 1D, E; Table S1C, D). Moreover, increased attraction to fermented cactus in the mainland Sonoran population was only observed after five weeks of organ pipe fermentation, despite there being no difference in attractiveness between one and five week fermented samples (Figure 1D). There is also an indication of sex specific responses between some rot stage comparisons and overall females tended to have stronger responses. These results are expected given previous findings in Drosophila that show differences in olfactory responses between the sexes and increased behavioral responses in D. mojavensis females relative to males [12], [16], [17]. Finally, repetition of these behavioral tests using flies from a second fly line for each population lead to the same conclusion (data not shown), suggesting that changes in attraction with changes in headspace volatiles over time reflect a general population specific result.

These differences in preference for different fermentation stages arise because volatile composition of fermenting cactus tissues is dynamic [18]. To determine changes in volatiles over time, we sampled the headspace of all four host cacti at one week intervals for nine weeks. Headspace volatiles varied between host plants and varied in their relative amounts across time (Figure 2A–D; Figure S1A–D). Seventy seven compounds were identified, with six unique to barrel, eight to prickly pear, two to agria and none to organ pipe (Table S2). More specifically, 1-dodecene, 2-methoxy-4-propyl phenol, durenol, isopropyl acetate, isopropyl propionate, and N,N'-diethyl-1,3 benzenediamine were unique to barrel cactus. The compounds 2-methyl-3-nonanol, 2-octanol acetate, ethyl propionate, isobutyl tiglate, isopropyl isopentanoate, isopropyl pentanoate, n-propyl 3-mercapto-propanoate, and pentanoic acid 1-methylpropyl ester were unique to prickly pear, and 6-methyl-2-heptanone and acetic acid were unique to agria. In general, the majority of identified compounds were esters (38%) and aromatics (30%). The volatile blends of prickly pear and agria were primarily equal in number of esters and aromatics, but organ pipe and barrel cacti were enriched for esters and aromatics, respectively. These host specific differences are illustrated by a principal component analysis (PCA) based on the volatile composition, in which the four host plants are clearly segregated into separate groups (Figure 2E; Table S3).

Given that the volatile compositions of these host cacti differ substantially, we asked whether there was evidence of host specific adaptations in the olfactory systems of the fly populations. Intraspecific variation in odor-guided behavior has been observed previously in the tephritid fly Rhagoletis pomonella and these differences were accompanied by subtle changes in the peripheral odor detection machinery [19], [20]. We examined whether there were alterations in the electrophysiological response properties of the antennae and maxillary palps using electroantennograms (EAG) and electropalpograms (EPG). We measured responses to 110 compounds for the antennae and 32 compounds for the maxillary palps among D. mojavensis populations and its sister species, D. arizonae. The odorants included diverse chemical groups as well as compounds present in fermenting host cacti ([13], [14] and this study). Both EAG and EPG measurements indicated significant differences in odor detection, especially for the Mojave population (Figure 3A, B), as illustrated by PC analyses based on both EAG and EPG response characteristics. In both cases, the PCA grouped the D. mojavensis mainland Sonoran, Baja, and S. Catalina populations together with D. arizonae, and separate from the D. mojavensis Mojave population (EAG: ANOSIM based on Bray-Curts similarity, R = 0.72, P<0.0001, Figure 3C; EPG: ANOSIM based on Bray-Curts similarity, R = 0.625, P<0.0001, Figure 3D). Specifically, the antennae of the Mojave population differed primarily in having an overall reduced response to straight-chain esters, and the palps differed in having a strong response to 4-ethylguaiacol, a compound that elicited minimal responses from the other populations and D. arizonae. Moreover, as with rot preference behavior, electrophysiological responses did not differ significantly within populations (Table S1E). Therefore, these distinct odor sensitivities of the Mojave population presumably constitute host-specific adaptations. The headspace of barrel cacti, the sole host of the Mojave population, had a lower number of esters and those esters identified typically constitute only minor components of the volatile blend (see section on the identification and comparison of volatile composition over time, above). However, aromatics are a dominant component of barrel cactus headspace, with compounds such as 4-ethylguaiacol, present across all fermentation stages but present only in trace amounts in prickly pear and organ pipe and not detected in agria headspace (Table S2).

Given the shifts in peripheral detection, we measured behavioral responses of the four D. mojavensis populations to a mixture of thirteen compounds for which the majority showed population differences in electrophysiological responses and/or were consistently present in barrel cactus. Since responses (electrophysiological and behavioral) did not differ significantly within populations, we focused our behavioral analyses of synthetic compounds on a single line per population. As expected, the Mojave population showed greater attraction to the mixture than the other D. mojavensis populations (Figure 4; Table S1F). Females in particular, showed attraction to the mixture in a dose dependent manner, which is consistent with previous studies [16], [17] showing increased behavioral responses in females relative to males. Moreover, females of the Mojave population continued to show attraction to the mixture at a 10⁻² dilution, while responses of females from the other three populations ranged from decreased attraction to repulsion. In the case of the Baja population, for example, dose responses were shifted to lower concentrations. We also tested responses to the thirteen compounds that were components of the mixture individually, at several concentrations (Figure 5; Table S1G). Most of the single compounds elicited minimal attraction or repulsion across all populations. Furthermore, those single compounds that elicited population specific differences did not recapitulate the host specific responses of the mixture, indicating that a combination of volatiles is essential for appropriate host plant identification and preference.

We observed differences in olfactory preferences in D. mojavensis for different stages of host plant fermentation. These fermentation stages varied in the composition and abundance of volatiles produced. Early studies of the Baja population have shown that flies prefer fermenting cactus to fresh tissue and have an attraction to initial fermentation stages [13]. Our results are in accordance with these studies in that the Mojave and Catalina populations also exhibit a preference for early stages of cactus fermentation. No difference in preference among fermentation stages, however, was observed for the mainland Sonoran and Baja populations for their respective host plants. This lack of differential attraction was consistent with comparatively little change in the volatile compositions of these cacti over the test period and most likely reflect methodological differences between our study and Downing, 1985 [13]. Moreover, while rotting in nature likely occurs more rapidly and additional studies on microbe colonization of damaged cactus tissue in nature are needed [21], these laboratory experiments can identify volatiles and changes in volatiles that underlie shifts in host plant preference behavior.

Previous studies of single compounds have been instrumental in the development of our understanding of how olfactory cues are processed [22]. In nature, however, organisms encounter a vast array of volatiles and the importance of single compounds in an ecological context remains less clear. Our results show that individual compounds may elicit behavioral responses in D. mojavensis, but that host-specific attraction to a mixture of these compounds could not be explained by responses to a single compound alone. The importance of odor mixtures in mediating appropriate behavioral responses has been observed in other systems (e.g., grapevine moth (Lobesia botrana), [23]; oriental fruit moth (Cydia molesta), [24]; hawkmoth (Manduca sexta) [25]) and our results support a model in which synergistic or antagonistic effects among mixture compounds result in host specific behavioral responses to olfactory cues.

Adaptation to different ecological environments can result in divergence of olfactory preference [26]. We examined whether alterations at the sensory periphery were found among populations of D. mojavensis that differ in their host plant use. Notably we observed that the Mojave Desert population, specializing on barrel cactus, has diverged in its olfactory sensitivities with an overall decreased response to esters and increased response to aromatics. This divergence from the other three populations coincides with the fact that the volatiles released by fermenting barrel cactus are heavy in aromatics compared to the other three host cacti whose primary volatiles were enriched for esters or balanced equally with aromatics ([13], [14] and this study). These electrophysiological differences most likely reflect alterations in ligand binding or odorant clearance at the sensory periphery, either through changes in gene expression or protein structure-function. The latter, caused by amino acid substitutions in chemosensory receptors, has been shown to confer differences in odorant sensitivity [27], [28], [29]. On the other hand, changes in the number of olfactory sensory neurons can also tune olfactory sensitivity towards host-specific volatiles, as in the case of D. sechellia, a specialist on Morinda fruit [30], [31]. Moreover, this preference by D. sechellia for Morinda fruit volatiles has been shown to be mediated by odorant binding proteins [32] and host-driven sensory augmentation has been shown for other insects, such as Culex mosquitoes [33].

In short, we have begun to understand the evolution of olfactory preference in response to host plant shift in D. mojavensis, a model of incipient speciation. Our results suggest rapid adaptation to changes in host plant utilization in this system. Estimates of divergence between.

D. mojavensis and D. arizonae range between 1.91 and 2.97 million years ago [34], [35]. Moreover, Smith et al., 2012 [36] estimates that the Baja population diverged from an ancestral mainland Sonoran/Mojave Desert group 230,000 to 270,000 years ago. Separation of the mainland Sonoran and Mojave Desert populations are then estimated to have occurred 117,000 to 135,000 years ago. Such rapid adaptation of the olfactory system has also been observed in Rhagoletis, with shifts in olfactory preferences from hawthorn to apple within 150 years [37]. In the aforementioned studies of shift of host preference in D. sechellia, the shift was proposed to be in response to competition with D. simulans [38], with divergence between species less than half a million years ago [39]. Our findings in this system will help unravel mechanisms underlying the process of species formation and the evolution host-plant specialization.

An analysis of the volatile compositions of Stenocereus gummosus, S. thurberi, Ferocactus cylindraceus, and Opuntia littoralis, was obtained through headspace solid phase microextraction (SPME, Polydimethylsiloxane/Divinylbenzene, Sigma-Aldrich, St. Louis, MO). For each cactus, volatile compounds emitted from uninoculated and inoculated cactus were identified. The tissue was kept frozen and for the experiments it was thawed, placed in a glass jar with a polyethylene lined cap, sterilized and subsequently inoculated. More specifically, a 70 g piece of cactus tissue was inoculated with 1.0 ml of seven yeast species (Pichia cactophila, P. mexicana, Starmera amethionina, Candida valida, C. sonorensis, Diapodascus starmeri and Sporopachyderma cereana) mixture and 0.5 ml of one pectolytic bacterium Erwinia cacticida ([40], [41] and Etges pers. comm.). Both yeast and bacteria cultures were freshly grown on Yeast Complete media or Glucose Yeast Calcium carbonate media plates respectively. After 48 hours, the microorganisms were harvested and suspended in sterile water. To facilitate the even distribution of the microorganisms, the cactus tissue was subsequently inoculated at multiple spots using a syringe [40], [41]. These yeast species have been documented on necrotic cacti [18] and used previously in D. mojavensis rearing experiments [41]. Cactus tissue was then incubated at 30°C and the volatiles present in the headspace were determined at weekly intervals over nine weeks. The volatile compositions of two to three replicate rots per cactus were examined over the nine week period. The identification of volatiles emitted from uninoculated cactus tissue was determined after one day.

The SPME fiber was exposed for one hour to the sample headspace, and the fiber assembly was then placed into the GC-MS injector port. Volatiles were analyzed using an Agilent 7890A GC with 5975C MSD apparatus (Santa Clara, CA) in a pulsed splitless mode. The GC-MS was equipped with a polyethylene glycol column (Nukol, Supelco Co.). GC conditions were optimized with standards and subsequent analyses done at injector and detector (FID) temperatures of 250°C and 280°C, respectively. Helium was used as the carrier gas at 25 ml min⁻¹, and at a split ratio of 2∶1. The oven temperature was initially set at 40°C for 1 min and then ramped to 210°C at a rate of 7° min⁻¹. Mass spectra were recorded from 35 to 700 amu, with electronic impact ionization at 70eV. Compounds were identified using the NIST Mass Spectral Library, by comparison to their retention times, and by mass spectra analyses of select standards. Compounds with more than a 90% match with the NIST library were labeled. Raw data was subjected to principal component analysis (PCA) using JMP 9.0 (SAS Institute, Cary, NC).

Flies were obtained from the Drosophila Species Stock Center or kindly provided by Dr. Bill Etges and are as follows: Baja California population, [Punta Prieta (stock number 15081–1351.30) and San Quintin (SQ59a)]; the mainland Sonoran population [Organ Pipe National Monument, Arizona, stock number 15081–1352.32 and OPNM9]; the Mojave population [Grand Canyon, Arizona (stock number 15081–1352.10) and Providence Mountain, CA (A997b)]; Santa Catalina Island [stock numbers 15081–1352.30 and 15081–1352.22]; D. arizonae [Sinaloa, Mexico (stock number 15081–1271.33)]. All flies were reared on cactus-banana-agar medium and were maintained at 25 ^oC, under a 12 h L/D cycle.

Free walking behavioral assays consisted of twenty flies placed into a polystyrene arena (6 cm (H) ×15 cm (Ø)) containing two traps. Each trap was constructed using a 10 ml glass beaker, fitted with a polypropylene plastic funnel. Traps were then symmetrically placed within the testing arena. To prevent dehydration of the flies, a cotton ball saturated with 20 ml of water was placed into the arena. Flies were tested at 10–12 days post-eclosion and flies were starved overnight prior to the experiment. Assays were performed in the dark and the number of flies trapped was recorded after 48 hours. For tests of single odorants or synthetic mixtures, traps contained 2 ml of the vehicle control and 0.1% Triton X with or without odorant(s). All odorants were obtained from Sigma-Aldrich (St. Louis, MO) at the highest purity available. Behavioral responses were measured to a synthetic mixture of 13 individual compounds at proportions reflective of the headspace of a three week fermented barrel cactus (Table S2). These mixture included hexyl acetate, acetone, phenethyl acetate, guaiacol, 1-hexanol, 2-nonanol, 4-ethyl guaiacol, phenol, isopropyl benzoate, phenethyl propionate, isoamyl propionate, 4-ethyl phenol, and 4-vinyl guaiacol. Because 4-vinyl guaiacol was present in the majority of early stage rots, with the exception of week 3, it was included at its average relative amount across weeks one through five. All mixture components were also tested singly. Response indices were calculated by subtracting the number of flies present in the control traps from the number of flies present in the trap containing odor and dividing by the total number of flies. Ten replicate measurements per sex, population and odorant concentration were conducted. Statistical analyses were conducted using ANOVA, followed by a Tukey-Kramer post-hoc test. For behavioral tests using fermenting cactus, all four cacti were inoculated as described in the above identification of host plant volatile section. Two grams of uninoculated or fermented cactus tissue was used per trap. Five replicate measurements per sex and per choice test were conducted. Statistical analyses were conducted within a given sex and two choice test using ANOVA. All analyses were done using JMP 9.0 software (SAS Institute, Cary, NC).