An organism's sensory systems are typically well-tuned to the requirements of its habitat and lifestyle. Major lifestyle transitions are therefore often accompanied by dramatic changes to sensory systems, such as the secondary loss of eyes in cave-dwelling fish or the evolution of electroreception in Guiana dolphins (Darwin, 1859; Czech-Damal et al., 2011; Cartwright et al., 2017). Similarly, it is conceivable that both the conquest of land and taking flight by arthropods have created opportunities for evolutionary innovation and at the same time changed the demands on early hexapod and insect chemosensory systems.

Insects detect odorants and tastants in their environment using members of three gene families, ionotropic receptors (Irs), gustatory receptors (Grs) and odorant receptors (Ors) (Clyne et al., 1999, 2000; Gao and Chess, 1999; Vosshall et al., 1999; Scott et al., 2001; Benton et al., 2009). Ionotropic receptors are believed to have their origin early in the protostome lineage and Grs are probably even more ancient with gene family members present in animals as distantly related to insects as sea urchins (Croset et al., 2010; Robertson, 2015, 2019; Saina et al., 2015; Eyun et al., 2017). In comparison, Ors appear to be an insect-specific expansion from within the Gr gene family, with the functionally essential and highly conserved odorant receptor co-receptor (Orco) thought to be the ancestral Or (Robertson et al., 2003; Robertson, 2019).

At the molecular level, the functional Or signal transduction complex is a heteromultimer (most likely a heterotetramer) of a specific Or (Orx), which defines the ligand-specificity of the complex, and the ubiquitously expressed Orco (Dobritsa et al., 2003; Hallem et al., 2004; Hopf et al., 2015; Butterwick et al., 2018). Orco functions as a chaperone in intracellular Or trafficking and as an essential component of a non-selective cation channel that opens upon ligand-binding to the specific Orx leading to the depolarization and hence excitation of the sensory neuron expressing the Orx/Orco complex (Larsson et al., 2004; Sato et al., 2008; Wicher et al., 2008). In accordance with its central function in Or trafficking and signal transduction, exactly one Orco ortholog has been identified in most studies concerned with insect chemosensation, and insects lacking a functional Orco typically display severely impaired olfactory function (e.g., Krieger et al., 2003; Larsson et al., 2004; Robertson and Wanner, 2006; DeGennaro et al., 2013; Terrapon et al., 2014; Koutroumpa et al., 2016; Li et al., 2016; Yan et al., 2017).

Two hypotheses have been proposed to explain the origin and expansion of the Or multigene family in insect genomes: (1) Ors may represent an adaptation to a terrestrial lifestyle (Robertson et al., 2003; Brand et al., 2018), which is supported by the tendency of Ors to detect hydrophobic compounds, whereas the more ancient Irs and Grs mostly detect hydrophilic substances. Moreover, Ors display, on average, a wider ligand spectrum than Irs, presumably allowing for the detection of a greater number of volatiles (Hallem and Carlson, 2006; Ai et al., 2010; Abuin et al., 2011; Silbering et al., 2011; Min et al., 2013); (2) Ors may represent an adaptation to insect flight (Missbach et al., 2014), which is supported by the observation that Ors are typically more sensitive to their ligands than Irs, that Or-expressing sensory neurons respond more readily and reliably to short odor pulses typical for in-flight odor detection than their Ir-expressing counterparts, and that sensitivity of the Or/Orco complex can be adjusted by a variety of modulatory mechanisms (Sargsyan et al., 2011; Getahun et al., 2012, 2016; Mukunda et al., 2014; Guo et al., 2017).

An apparent lack of Or gene family members in antennal transcriptomes of the bristletail Lepismachilis y-signata and the detection of three Orco-like sequences but no Orxs in the antennal transcriptome of the firebrat Thermobia domestica supported the latter hypothesis, placing the origin of the derived Orx/Orco system alongside the advent of insect flight (Missbach et al., 2014). However, this interpretation was recently challenged by the observation that one of the Orco-like sequences in the firebrat associates phylogenetically with damselfly Ors (Ioannidis et al., 2017), and even more so by the discovery of many more Or gene family members in the firebrat genome and some Or genes but no Orco ortholog in the genome of the bristletail, Machilis hrabei (Brand et al., 2018). Based on these findings the authors of the latter study argue that the Orx/Orco olfactory system was already fully established in the ancestor of silverfish/firebrats (i.e., the order Zygentoma) and winged insects and hence represents an adaptation to a terrestrial lifestyle.

In summary, there is growing consensus that the evolutionary origin of the Or multigene family (in the broad sense) lies in the common ancestor of insects based on transcriptomic and genomic information. However, both bristletail and Zygentoma Or repertoires display peculiarities hitherto unobserved in winged insects which appear to be specific to these wingless insect orders, with a lack of Orco in the examined bristletails and an expansion of Orco-like sequences in the firebrat. More data are clearly needed to examine whether these observations in single species generalize in the respective insect orders. Here we report results of our efforts to identify Or gene family members in the antennal transcriptomes of further examples from the Zygentoma, the common silverfish Lepisma saccharina and the forest silverfish Tricholepidion gertschi, to improve the resolution of early evolutionary events in the chemosensory systems at the base of the insect phylogeny.

We performed transcriptomic surveys of several tissues in two additional zygentome species. The monophyletic order Zygentoma is the sister group to all winged insects and comprises four families (Koch, 2003; Blanke et al., 2014; Misof et al., 2014). We obtained transcriptomes from the only extant representative of the most basal family, Lepidotrichidae, the forest silverfish Tricholepidion gertschi, and from the common silverfish Lepisma saccharina from the same family as the previously examined firebrat Thermobia domestica, from the family Lepismatidae. Including a representative of the most basally branching zygentome family allows us to estimate whether a larger Or repertoire than that observed in bristletails is a specific character of the Lepismatidae, or a general feature of zygentome chemosensory systems. Unless specified otherwise in the following sections, we will use the term Or in its broader sense (i.e., including Orco and without distinguishing between Orco and specific Or or Orx).

Despite occupying a key position in the insect phylogeny of being the most derived wingless insect order, members of Zygentoma have historically received little attention regarding their chemosensory systems and chemical ecology. Firebrats and silverfish are known to possess a variety of cuticular sensory hairs, sensilla, on their appendages, some of which have been shown to have a chemosensory function by physiological methods (Adel, 1984; Berg and Schmidt, 1997; Hansen-Delkeskamp, 2001; Missbach et al., 2014; Hädicke et al., 2016). Behaviourally, firebrats and silverfish tend to aggregate in response to some chemical stimuli and to avoid others (Wang et al., 2006; Woodbury and Gries, 2007, 2008, 2013). In contrast, little has been known concerning the molecular makeup of their Or-based chemosensory system until a publication reported three Orco-like transcripts expressed in the antenna of the firebrat, Thermobia domestica (Missbach et al., 2014). The presence of three co-receptors without a single Orx seemed strange at the time, but a recent study examining the genome of the same species identified 44 members of the Or multigene family and proposed that Zygentoma already possess a full Orco/Orx-based olfactory system (Brand et al., 2018). The discrepancy between the number of Or transcripts identified from chemosensory tissue and the number of Or genes present in the genome may in part be attributable to an insufficient depth of the RNAseq experiments or point to an expression profile that involves other tissues or other life stages.

The present study provides several lines of evidence that Ors have a chemosensory, if not necessarily olfactory, function in the order Zygentoma. We find eight and 15 Or transcripts in transcriptomes from the silverfish T. gertschi and L. saccharina, respectively, numbers which are higher than those previously reported from the firebrat antenna. The 15 Ors from L. saccharina and eight from T. gertschi that we report here are likely to be underestimates of the full repertoire of these genes found within the respective genomes given that we are sampling just a small number of tissues and that there are often low expression levels of these genes per tissue due to their expression being limited to certain sensory neuron types. Consistent with an olfactory function and unlike identified Grs, the Or transcripts were predominantly detected in chemosensory tissues such as antennae and maxillary palps, and some of them were also expressed in other tissues. Expression of chemoreceptors (including Ors) outside the classical chemosensory tissues, e.g., in reproductive tissues and legs, has also been reported in other insect species (Bray and Amrein, 2003; Leal et al., 2013; Pitts et al., 2014; Haverkamp et al., 2016; Klinner et al., 2016; Raad et al., 2016), although the tissue-specific function of these genes respect to the physiology and behavior of the respective insects is not always clear. Given the generally low expression levels of Ors and that many of the firebrat Ors identified from the genome form clades with silverfish Ors identified from our RNAseq experiments in our phylogenetic analysis, it is likely that many of the firebrat genes described from the genome by Brand et al. (2018) are indeed also expressed in the firebrat antennae and that they were missed in previous RNAseq experiments because of insufficient sequencing depth.

Both insect Ors and the phylogenetically unrelated vertebrate Ors are assumed to evolve rapidly following a birth-and-death evolutionary process with new genes being born by gene duplication and others being lost by pseudogenisation (Nei and Rooney, 2005; Nei et al., 2008; Sánchez-Gracia et al., 2009; Ramdya and Benton, 2010; Hansson and Stensmyr, 2011; Andersson et al., 2015; Benton, 2015; Hughes et al., 2018). In combination with the adaptation of the Or repertoire to the species' ecological niche, this evolutionary process leads to Or gene phylogenies mostly lacking clear orthologs between species, and with species-specific expansions and reductions, which do not necessarily follow the underlying species phylogeny. Such a branching pattern with species-specific expansions and a lack of clear orthologs between species is clearly apparent in our phylogenetic reconstruction of neopteran ligand-binding Ors. Although the numbers of wingless and palaeopteran species, for which Or gene/transcript sequences are available, is still limited, a similar pattern begins to emerge within the group of Zygentoma Ors. This pattern on the one hand further supports the hypothesis that primitive Ors may function as true odorant receptors in these insect species, as they appear to follow an evolutionary process that is indicative of rapid evolution in response to the demands of the organism's ecological niche. On the other hand it suggests that Zygentoma may have a much richer species-specific chemosensory ecology than they have historically been given credit for.

But what does the Zygentoma Or signal transduction complex look like? Based on the finding that the firebrat genome harbors one Orco-like sequence that is more similar to neopteran Orcos than the others, the authors of a previous study concluded that the firebrat possesses an olfactory system akin to that of more derived insects consisting of one universal co-receptor and multiple ligand-binding Ors (Brand et al., 2018). Although such a molecular makeup of the zygentome olfactory system may indeed be true, some of our findings as well as results from a previous study suggest that the expression logic and molecular function of the zygentome Or signal transduction complex may differ from that of the Neoptera (Missbach et al., 2014). Firstly, in the firebrat, there is a stark contrast in the number of olfactory sensory units predicted from the Or gene number, when assuming a conventional Orx/Orco system with one universal co-receptor, i.e., 43 olfactory sensory neuron types, and those observed in physiological experiments, i.e., 13 functional units. Of those 13 functional units in the firebrat antenna, six display tuning profiles typically observed from Ir-expressing sensory neurons, leaving seven putative Or-expressing sensory neurons to house 43 possible Orx/Orco combinations predicted from genomic information. While it is possible that some of the firebrat Ors are expressed in different tissues or at different life stages, or that the electrophysiological screen simply missed some sensilla or sensory neuron types, this difference in numbers is striking and may hint at a different co-expression logic of primitive Ors than that observed in the Neoptera. Most primitive Ors display a higher degree of sequence conservation in their C-terminal region when compared to Orco than neopteran Orxs do. Many of them also retain some of the amino acid residues that have been identified to line the channel pore in Orco tetramers or possess chemically similar amino acid residues in the respective positions (Butterwick et al., 2018), which raises the possibility that some combinations of primitive Ors may be able to form ligand-gated cation channels independently from Orco. It is likely that the proteins most similar to neopteran Orco (TdomOrco, LsacOr1) act as true co-receptors for some of the Zygentoma Ors, most likely for those phylogenetically associating with neopteran Ors. However, the higher degree of conservation in the C-terminal region raises the possibility that some of the primitive Ors in bristletails, Zygentoma and mayflies may be components of an Orco-independent Or-based olfactory system that relies on stereotypical combinations of co-expressed primitive Ors and acts in parallel to the Orco-dependent system. Such an expression logic would be similar to that of the Gr gene family from within which Ors are thought to have originated (Dahanukar et al., 2007; Jones et al., 2007; Kwon et al., 2007; Jiao et al., 2008; Lee et al., 2009; Moon et al., 2009) and it could also accommodate the apparent lack of Orco in bristletails (Brand et al., 2018).

Including both Grs and neopteran Ors in our phylogenetic analysis allowed us to examine branching patterns within the insect Or multigene family at greater detail than has been possible previously. Although both primitive and neopteran Ors appear to evolve rapidly and presumably follow a birth-and-death evolutionary process, they were consistently resolved as two separate non-overlapping groups of genes in phylogenetic reconstructions including different sets of Grs and neopteran Ors. Our analyses thus suggests, that the Or multigene family found in neopteran insects is the result of not one, but two rapid expansions (Figure 5). The first “primitive Or” expansion presumably occurred in the common ancestor of insects in the mid-to-late-Ordovician (Rota-Stabelli et al., 2013; Misof et al., 2014). At this time the first vascular land plants had already emerged (Steemans et al., 2009) adding complexity and novel compound classes to the chemical landscape surrounding early terrestrial arthropods. Given the unclear positioning of the Orco clade in our phylogenetic reconstruction and because some Grs are able to form functional homomultimeric complexes in heterologous systems (Sato et al., 2011; Zhang et al., 2011), it is not clear at this point whether the Or gene family originated from a set of Gr genes, whose gene products functioned as heteromultimers, or from a single Gr gene. However, the second alternative seems attractive, because it can easily accommodate the as yet enigmatic absence of an Orco ortholog from the bristletail genome.

The co-receptor Orco most likely originated in the common ancestor of Zygentoma and winged insects (Brand et al., 2018). Given that just one Orco has been described from all winged insects (i.e., Palaeoptera and Neoptera) so far, the expansion of Orco-like sequences in the wider Orco clade potentially represents a derived character of the Zygentoma lineage and not the ancestral state of the insect olfactory system. The absence of Orco in bristletails, however, suggests that some primitive Ors may have functioned and may still be functioning without a co-receptor or in variable heteromultimers not only in bristletails, but also in Zygentoma and Palaeoptera. This notion is further corroborated by our inability to identify a clear Orco ortholog in T.gertschi, which suggests that this species may have secondarily lost its Orco gene. Although this observation needs to be treated with caution due to the fragmented nature of the T. gertschi transcriptome, it supports the idea that Orco is not absolutely essential to the function of all parts of the Zygentoma olfactory system.

Primitive Ors are mostly absent in damselflies and dragonflies apart from the clade closely associating with neopteran Ors, which may well-constitute the clade of ligand-binding Ors associating with these species' Orco proteins. However, they are still present with large expansions in the mayfly E. danica. According to our phylogenetic reconstruction, primitive Ors were entirely lost from insect genomes at the base of Neoptera in the early Devonian (Rota-Stabelli et al., 2013; Misof et al., 2014), leaving only Orco and one or a few Orx genes. The gradual loss of all Or genes except Orco and the Orxs forming complexes with Orco (i.e., the ancestral ligand-binding Ors) thus correlates with the advent of insect flight and the raised demands of in-flight odor detection. The gene products of T. domestica Orco and L. saccharina Or1 already possess some of the amino acid residues and motifs identified as targets of the molecular mechanisms modulating the sensitivity of the Orco/Or complex such as potential phosphorylation sites and a proposed calmodulin binding motif (Sargsyan et al., 2011; Mukunda et al., 2014). It is therefore tenable that the complex of Orco and the ancestral Orx replaced primitive Ors at the advent of flight, because it enabled early flying insects to detect the faintest whiff of an odor possibly indicating a valuable resource while still being able to adapt to the higher odor concentrations present at the odor source. After the loss of primitive Ors the radiation of neopteran insects in the Devonian then led to a second expansion of the Or multigene family giving rise to the large family of Ors observed in higher insects today.

In conclusion, we propose that the origin of the derived Orx/Orco-based insect olfactory system lies in a sequence of: (1) an initial rapid diversification of primitive Ors starting in the common ancestor of insects, which was presumably facilitated by the conquest of a terrestrial habitat and the advent of vascular plants. This system gave rise to the universal co-receptor and the ancestral ligand-binding Or/s at some later point; (2) pruning of this initial Or expansion was mediated by the demands of in-flight odor detection; (3) another rapid expansion and diversification of the remaining Orx lineage in Neoptera. Thus, both terrestrialisation and the advent of flight likely had dramatic effects on the molecular makeup of hexapod chemosensory systems. The proposed scenario will need to be put to the test by further increasing the phylogenetic resolution at the origin of the Or gene family with the addition of more genetic information on Ors from wingless insects and Palaeoptera and, importantly, Grs from non-insect hexapods, which may close the gap between the Gr and Or gene families. Aside from adding further sequence information, future studies should be directed at elucidating the expression logic and function of primitive Ors. The chemosensory systems of bristletails, silverfish and mayflies may well afford us with the rare opportunity to catch a glimpse of the evolutionary past and to understand the intermediate steps in the evolution of a multigene family not only from the sequence perspective, but also at a functional level.

Raw sequence reads have been submitted to the NCBI short read archive as BioProject PRJNA523997. Transcriptome assemblies will be provided upon request.

This study was conceived by all authors. Data collection and analysis was performed by MT, CM, and EG-W. Writing of the initial draft of the manuscript was performed by MT. All authors contributed to revising the manuscript and the acquisition of funding.