The genetics and epigenetics of animal migration and orientation: birds, butterflies and beyond

Migration is a common and critical behavioural adaptation for survival that has evolved in many animal taxa ranging from invertebrates to mammals. Animal migrations greatly vary in the distance travelled, from short trips as in the case of altitudinal migration (i.e. from lower to higher altitudes and back) to trans-continental migrations. All are, however, characterised by a seasonal movement to escape unfavourable environmental conditions and reach more-favourable sites during the inimical season, following precisely coordinated orientations and timing schedules. To accomplish such a remarkable navigational feat, migrants are equipped with a suite of adapted morphological, sensory, physiological and behavioural traits that are genetically encoded and turned on at the appropriate time of the year and/or under specific environmental conditions. Many studies have focused on elucidating the sensory cues and navigational strategies used for maintaining course orientation (Reppert et al., 2010; Reppert et al., 2016; Mouritsen, 2018), which includes the use of celestial compass cues (Kramer, 1950, 1952; Sauer, 1957; Schmidt-Koenig, 1958; Wiltschko et al., 1987; Perez et al., 1997) and the Earth's magnetic field (Wiltschko, 1968; Wiltschko and Wiltschko, 1972; Guerra et al., 2014). In contrast, the genetic architecture and molecular mechanisms that underlie the migratory phenotype, including flight direction/orientation and timing of their migration, remain poorly understood.

The explosion of technological advances in high-throughput sequencing (HTS) is starting to change this trend and holds great promise in applying non-biased approaches in the quest for de novo discovery of ‘migratory’ genes. Draft genome sequences of a few migratory species, including the monarch butterfly Danaus plexippus (Zhan et al., 2011), the Swainson's thrush Catharus ustulatus (Delmore et al., 2016), the willow warbler Phylloscopus trochilus (Lundberg et al., 2017), the stonechat Saxicola maurus (van Doren et al., 2017a), the rainbow trout Oncorhynchus mykiss (Berthelot et al., 2014) and the Chinook salmon Oncorhynchus tshawytscha (Christensen et al., 2018) are now available, and many more are on their way. This includes those of iconic migratory birds with well-characterised behaviours and evolutionary histories such as the European blackcap Sylvia atricapilla (K. E. Delmore and M.L., personal communication; Fig. 1). The use of whole-genome sequencing (WGS) or other HTS techniques, such as genotyping by sequencing (GBS) or restriction site associated DNA sequencing (RAD-seq) has enabled population genomics studies characterising migratory and non-migratory populations in the monarch and avian species or populations with different inherited migratory orientation (Zhan et al., 2014; Delmore et al., 2016). These studies revealed genomic regions with signatures of selection that may contain candidate genes for migration, cis-regulatory elements (CREs) and chromosomal inversions (Zhan et al., 2014; Delmore et al., 2015, 2016). RNA-seq is another approach that has been used to identify candidate migratory genes in the rainbow trout, the Swainson's thrush and the European blackbird Turdus merula by quantifying gene expression differences in brains and/or blood of individuals exhibiting variation in a migratory trait, such as migrants versus residents or populations of migrants varying in migratory orientation (Hale et al., 2016; Johnston et al., 2016; Franchini et al., 2017). A comprehensive understanding of the genetic basis of migration will, however, only be achieved through the use of combinatorial genomic and epigenomic (i.e. the study of modification in the genetic material due to environmental factors) approaches and the ultimate functional validation of candidate genomic regions through in vivo genetic disruption and behavioural assays in any given migratory species.

Here, we provide an overview of two iconic and complementary systems with well-characterised life history strategies and migratory patterns that are poised to increase our understanding of the genetics and epigenetics of animal migration: the European blackcap and the North American monarch butterfly. We also outline our vision of how technical advances and integrative approaches could be employed on these and other migratory species across both small and broad evolutionary and phylogenetic scales to rapidly move the field of genetics of animal migration forward.

Populations of European blackcaps, which are common breeding birds across Europe, exhibit a remarkably broad spectrum of behavioural variation in migratory status and orientation direction (Fig. 1A). Although selective breeding studies firmly established a genetic basis for migratory orientation decades ago (Helbig, 1991, 1996; Berthold et al., 1992), the recent development of genomics is starting to unlock the great potential of the extreme phenotypic variation in blackcaps to study migration genetics at the molecular level. Phenotypic variation across populations ranges from residents to short-, medium- and long-distance migrants (Berthold et al., 1990). In addition, migrants exhibit variation in migratory orientation strategies and form so-called migratory divides, i.e. areas where neighbouring populations breeding in close proximity have distinct migratory routes and non-breeding grounds (Fig. 1B). Blackcaps are distributed continuously across the central European migratory divide, with breeding populations in Europe and overwintering areas in southern Iberia and northern Africa. In autumn, populations breeding west of the divide migrate southwest (SW) around the Mediterranean Sea, while populations east of the divide migrate southeasterly (SE) (Fig. 1B). In addition, since the 1960s, a growing number of birds breeding in central Europe migrate northwest in autumn to overwinter in the British Isles (Berthold et al., 1992), possibly because of the milder wintering conditions and food availability in British gardens (Plummer et al., 2015).

A series of common garden experiments using blackcaps with different migratory phenotypes provided clear evidence that migratory traits like distance (Berthold and Querner, 1981) and direction (Helbig, 1991, 1996; Berthold et al., 1992) are genetically encoded and have the potential to drastically change within a few generations under strong artificial selection (Berthold, 1991). In these experiments, selection lines were first generated by mating SW migrants with SW migrants and SE migrants with SE migrants for several generations in order to obtain populations that are fixed for each migratory orientation (Fig. 1C). In cross-breeding experiments, SW migrants were then mated with SE migrants, and the progeny was reared in isolation from their parents to exclude the possibility that migratory direction could be learned. The intermediate orientation taken by hybrids unambiguously demonstrated the genetic inheritance of this migratory trait (Helbig, 1991, 1996; Berthold et al., 1992).

Like most songbirds, blackcaps migrate at night and mostly by themselves. Because parents often leave the breeding grounds earlier (Flack et al., 2018), young inexperienced birds on their first migratory journey have to find their way on their own. Despite heading towards a completely unfamiliar destination, they reach it with amazing precision. Behavioural experiments in the lab and in the wild have shown that migratory birds use various compasses for course orientation. Like other night navigators, blackcaps rely on a magnetic compass calibrated by sunrise/sunset polarized light cues and a star compass (Wiltschko and Merkel, 1966; Wiltschko and Wiltschko, 1972; Wiltschko et al., 1987; Phillips and Moore, 1992; Cochran et al., 2004). Birds are also equipped with an inherited time schedule and at least an initial migratory direction that are integrated into a spatiotemporal strategy signalling of when to leave the breeding grounds, and how far and in which direction to fly (Liedvogel et al., 2011).

The variability in migratory strategy that presumably reflects differences in direction and strength of selection make blackcaps an excellent model to study not only the genetic variation in migratory traits, but also the effects of selection on this complex phenotype (Fig. 1) (Liedvogel et al., 2011; Delmore and Liedvogel, 2016). The inability to estimate population structure using neutral markers across the genome (Helbig, 1994; Pérez-Tris et al., 2004; Mettler et al., 2013) suggests that migratory traits of neighbouring populations do not correlate with strong overall genetic differentiation. Migratory traits rather seem to be controlled either by selection processes on relatively few genomic regions and/or differences in gene expression. This hypothesis is supported by the finding that few genes appear to be involved in determining the expression of migratory traits (Helbig, 1996), and by the apparent strong genetic correlation between different migratory traits for which variation in one trait is often dependent on variation in another (Pulido et al., 1996; Pulido and Berthold, 1998). With a de novo assembly of the blackcap reference genome underway, and the technological advances in HTS, characterizing the genetic basis of the variation in migratory phenotypes at the genome-wide level will now be feasible. Whole-genome re-sequencing of blackcaps exhibiting the full spectrum of migratory phenotypes is anticipated to provide insights into the origins and maintenance of variation in the migratory behaviour and the identification of genetic variants underlying focal traits in migratory behaviour, including the propensity to migrate, orientation and distances travelled.

With established genomic resources that include tools for germline editing, the North American monarch butterfly is an equally compelling model and arguably the best suited to define the genetic and epigenetic basis of insect long-distance migration. Monarch butterflies from the eastern North American population undergo one of the most impressive annual long-distance migrations accomplished by any insects, have a rich and well-documented natural history, and progress has been made in understanding the neurobiological basis of their migration (Urquhart, 1960; Reppert et al., 2016). Each autumn, coincident with decreasing day length (i.e. photoperiod), monarchs east of the Rocky Mountains depart from their northern breeding sites, migrating south to reach their overwintering sites in Mexico, where they remain in a state of reproductive quiescence (i.e. diapause) until the spring (Fig. 2A). Then, with increased temperatures and photoperiod, migrants become reproductive, mate and re-migrate northward to the southern United States where females lay their progeny onto milkweed plants, the unique host on which larvae feed. Both autumn migrants and spring re-migrants use a time-compensated sun compass as a primary compass system to guide their course orientation (Perez et al., 1997; Mouritsen and Frost, 2002; Froy et al., 2003; Merlin et al., 2009) and may also take advantage of their ability to orient to the magnetic field when the sun is not visible (Guerra et al., 2014). Unlike birds, the re-migrant adults do not make the return back to the breeding sites. Instead, the completion of the annual migratory cycle takes at least two successive generations of reproductively competent spring and non-oriented summer butterflies that follow the northward progression of milkweed availability (Fig. 2A). The butterflies that emerge by late summer in the northern range of the breeding sites are then re-programmed into autumn migrants. Just like young inexperienced birds, autumn migratory monarchs travel thousands of kilometres on their virgin migratory journey and locate their overwintering grounds with astonishing precision, sometimes even clustering on the same trees as their great-grandparents. However, unlike birds, migratory direction cannot just be an inherited trait, as autumn migrants not only share the genetic make-up of their non-migratory parents but also reverse flight orientation by 180 degrees in the spring when they become remigrants.

The switch in migratory physiology and behaviour and the reversal in flight orientation that occur respectively in the autumn and spring in response to environmental changes suggest an epigenetic control of the eastern North American monarch migration (Fig. 2A). While seasonal changes in autumn photoperiod and temperature likely control the migratory switch, the trigger that reverses flight orientation from south in autumn migrants to north in spring remigrants has been identified as a prolonged exposure to low temperatures that mimic the coldness experienced by migrants at their overwintering sites (Guerra and Reppert, 2013). The ability to experimentally re-program the southward migratory orientation of autumn migrants into a northward orientation in controlled laboratory conditions provides a unique opportunity to define the epigenetic mechanisms underlying reorientation of the sun compass in the brain (Guerra and Reppert, 2013).

In fact, transient exposure of animals to environmental factors has previously been shown to induce and maintain behavioural states by changing the neuronal epigenetic landscape that regulates transcription of genome-wide gene expression over long periods of time (Bonasio, 2015; Yan et al., 2015). Reprogramming of gene regulatory networks could involve the alteration of chromatin structure via histone post-translational modifications, and/or the activity of specific transcription factors that bind to CREs to activate or repress gene expression (Bonasio et al., 2010). In contrast to birds, DNA methylation is unlikely to play a significant role in the monarch seasonal migratory behaviour, as this species appears to lack detectable patterns of de novo DNA methylation (Zhan et al., 2011; Bewick et al., 2017). Temperature-dependent post-transcriptional mechanisms involving regulation of gene expression by microRNAs, mRNA splicing or RNA editing are alternative mechanisms by which flight orientation could also be re-programmed, as temperature-dependent regulation of these mechanisms has previously been reported in other organisms (Low et al., 2008; Garrett and Rosenthal, 2012; Bizuayehu et al., 2015).

Despite being the focal population for studying the mechanisms that drive monarch migration, the North American population is not the only monarch population found around the globe. Other populations are present in Oceania, Central America, South America, the Caribbean, Europe and North Africa (Fig. 2B). While the Australian population also undergo a seasonal migration, although of shorter distances than the North American population (Dingle et al., 1999), all the others are felt to be non-migratory (Dockx, 2007; Altizer and Davis, 2010; Zhan et al., 2014). These differences in migratory behaviour between populations have also been leveraged to identify putative genes involved in monarch migration (see below).

As illustrated with the blackcap and the monarch butterfly, there are clear differences in the nature of the molecular basis (genetic and/or epigenetic) underlying migration between taxa that may be inherent to the biology of each species. The diversity of migratory phenotypes across bird species, together with the existence of populations with clearly defined and differing migratory orientations within a single species, makes them uniquely suited for studying the genetic basis of migration. The seasonal plasticity of monarch butterflies trans-generational migratory behaviour, on the other hand, offers opportunities to study the epigenetic basis of migration. With the ability to maintain colonies in the laboratory year-round, a fast generation time and its accessibility to in vivo genetic manipulation (Merlin et al., 2013; Markert et al., 2016; Zhang et al., 2017), the monarch is also uniquely suited to rapidly assess the function of candidate migratory genes (Fig. 3B). Using these two species as focal examples, we highlight how integrating a diverse array of approaches used in behavioural ecology, sensory biology, evolutionary biology, genetics and molecular biology can be leveraged to move the field of migration genetics forward. Ultimately, the discovery of generalizable mechanisms, or lack thereof, will, however, require the diversification of strategically chosen migratory animal models across taxa.

Regardless of the species or population of interest, accurate knowledge of the migratory phenotype and the ability to characterize and quantify it is an essential prerequisite to characterize the underlying molecular machinery that modulates the variation in migratory behaviour. Capture–mark–recapture methods and careful assessment of the individual's morphology and physiology associated with the migratory state (i.e. wing size and shape, fat stores) have provided useful information, with the exception of the detailed behavioural strategies of individual migrants.

Aided by the miniaturization of tracking devices, more advanced on-board tracking technologies can now inform both orientation and timing characteristics of migratory routes taken by individuals over long distances in their natural habitat. The most informative data come from studies using global positioning system (GPS) satellite transmitters (Perras and Nebel, 2012), where real-time recording of the signal by satellites enables the long-term monitoring of migratory movement around the globe without recapture of the tagged animal. Using this technology for real-time monitoring of smaller passerines such as the European blackcap is not yet possible due to the relatively large size of transmitters. Current alternative technologies include archival tags that record light-intensity data (i.e. light-level geolocators) as well as radio-transmitters and telemetry systems (Stutchbury et al., 2009; Kishkinev et al., 2016; Taylor et al., 2017). In contrast, no comparable devices exist yet for insects. Until the next revolution in miniaturization, when geolocators or GPS devices weighing only a few tens of milligrams could be developed, entomological radars will likely remain the method of choice to track flight paths and strategies of migratory insects in nature (Chapman et al., 2008, 2010, 2011; Bruderer, 2016; Woodgate et al., 2016).

Assessing migratory behaviour in laboratory conditions will be equally important to ultimately test gene function in vivo. A number of laboratory-based methods have been developed in birds and butterflies for use as proxies of migratory status and orientation. A commonly used method in birds relies on recording the characteristic migratory restlessness exhibited during the migratory season, even when kept in cages in controlled conditions (Kramer, 1949, 1950). Quantified timing and directedness of this behaviour coincide with timing and orientation of free-flying conspecifics (Gwinner, 1968, 1986; Berthold, 1973, 1995). Behavioural cages for testing bird migratory status and orientation, known as ‘Emlen funnel’ (Emlen and Emlen, 1966), are generally funnel-shape arenas. The walls are covered with coated and scratch-sensitive paper such that restless migratory birds leave scratch marks on the inclined walls. These marks can be quantified to measure activity levels and mean orientation bearings (Helbig, 1991; Wiltschko and Wiltschko, 1995; Mouritsen et al., 2009; Zapka et al., 2009). The precise monitoring of timing of migration, including onset, duration and intensity of migratory restlessness, can be achieved by fitting the cage with passive infrared detectors that record activity profiles (Gwinner, 1967). Related methodological approaches have been adapted to insects, where the orientation of a tethered insect is tracked in a flight simulator apparatus (Mouritsen and Frost, 2002). This cylindrical plastic barrel can be used outdoors under natural sky conditions to assess spontaneous migratory orientation of tethered insects that are free to fly in any direction on the horizontal plane (Mouritsen and Frost, 2002; Froy et al., 2003; Dreyer et al., 2018). Similar circular plastic arenas have also been adapted to track orientation in aquatic animals, including fish (Mouritsen et al., 2013a). Even though these approaches do not fully recapitulate the natural behaviour exhibited by a migrating animal (van Doren et al., 2017b), they will be valuable tools to ultimately test the function of migratory genes in vivo as genetically manipulated animals cannot be released into the wild.

Evidence suggests that migratory traits are genetically inherited and/or environmentally regulated in different taxa. The impressive series of common garden experiments with European blackcaps provided evidence for a genetic basis of migratory traits in birds (Fig. 1C) (Berthold et al., 1992; Helbig, 1991, 1996). Genetic inheritance of both timing and migratory direction was further supported by elegant displacement experiments in which both experienced adults that already had successfully completed a migratory journey and naïve juvenile birds on their first journey were displaced from their original location (Perdeck, 1958; Thorup et al., 2007). Inexperienced juveniles followed an innate clock and compass strategy (e.g. vector navigation), leaving at the right time and flying the correct distance in the inherited migratory direction. In contrast, adult birds that had the opportunity to learn landmarks on their previous trip were able to compensate for the displacement, actively navigating to their original wintering area (Perdeck, 1958).

In contrast to blackcaps, genetic inheritance of migratory traits in North American monarchs has not been established using breeding experiments, and the interpretation of displacement experiments remains controversial (Mouritsen et al., 2013b; Oberhauser et al., 2013). Nevertheless, the multigenerational nature of the monarch migratory cycle in the Eastern United States suggests that genes encoding migratory traits are encoded in the monarch genome, and thus heritable, but are turned on or off in response to seasonal environmental changes in the migratory generation.

Across taxa, the onset of migratory behaviour and departure from breeding grounds is tightly linked to the changing seasons. To precisely follow timing schedules, animals keep track of the seasons using an endogenous seasonal timer that measures photoperiodic changes.

In birds, migratory timing appears to be controlled in concert with timing of reproduction and moult by a circannual clock in both wild populations and caged birds (Gwinner, 2003; Visser et al., 2010). Correlative studies using a candidate gene approach identified the Clock gene as a possible candidate for migratory timing (Peterson et al., 2013; Bazzi et al., 2015, 2016; Saino et al., 2017; Contina et al., 2018). The otherwise highly conserved Clock gene shows length variation at a poly-glutamine repeat, and shorter alleles have been associated with earlier arrival times at the breeding grounds in barn swallows (Bazzi et al., 2015). In the same species, elevated methylation levels responsible for reduced transcription at the Clock locus have been shown to correlate with earlier spring migration and onset of breeding (Saino et al., 2017).

In monarchs, circadian clocks or clock genes in the brain have also been proposed to take part in photoperiodic measurement (Reppert et al., 2016; Denlinger et al., 2017). They could not only trigger the photoperiodic programming of the migratory state in the autumn but also time the departure of migrants from their breeding range. Although the role of circadian clocks in triggering the migratory switch remains to be explored, their importance for monarch migration is not unprecedented. Circadian clocks located in the antennae provide the necessary timing component that allows autumn migrants and spring remigrants to maintain course orientation by compensating for the daily changes of the position of the sun in the sky – the main compass cue used for navigation in this species (Perez et al., 1997; Mouritsen and Frost, 2002; Froy et al., 2003; Merlin et al., 2009; Guerra et al., 2012). Brain clocks are, however, most likely involved in the induction of the migratory traits and could impact migratory traits by regulating the transcription of clock genes and clock-controlled genes in this tissue (Hardin and Panda, 2013). With the availability of several clock gene knockouts in monarchs (Merlin et al., 2013; Markert et al., 2016; Zhang et al., 2017), functionally determining whether circadian clocks or clock genes play a role in the migratory switch should now be feasible, at least in this system.

The current revolution in HTS technologies and functional genomic tools and their applicability in non-model organisms provide unique opportunities to comprehensively dissect the genetic and epigenetic basis of migration (Fig. 3).

The use of WGS has already enabled population genomics studies between migratory and non-migratory populations for the identification of variation in the genome associated with variation in the phenotype. In monarchs, for example, re-sequencing of more than 100 individuals from migratory and presumably non-migratory populations from around the globe (Fig. 2B) identified genomic regions of divergent natural selection comprising ∼500 candidate genes proposed to be associated with shifts in migratory behaviour (Zhan et al., 2014). The interpretation of such genome-wide scans for a signature of selection is, however, often restricted to protein-encoding genes. Although variations within coding regions could have a fundamental impact on protein function, without functional evidence, variations in non-coding regions such as in CREs should not be excluded as they could affect differences in expression of genes located kilobases away. Quantification of differential gene expression between migratory and non-migratory forms can be achieved using RNA-seq, as exemplified by a few studies in fish and insects (Jones et al., 2008; Zhu et al., 2009; Hale et al., 2016). However, differential gene expression studies based on RNA-seq usually yield long lists of candidate genes and the prioritization process for further functional studies (discussed below) is not always straightforward. More-integrated approaches that consider not only loci under selection but also gene expression and its transcriptional or post-transcriptional control within a single species could help in narrowing down the list to a smaller number of promising candidates.

At the transcriptional level, gene expression is regulated by the binding of transcription factors to specific DNA sequences in CREs located in open chromatin regions, and by the recruitment of specific cofactors mediating histone post-transcriptional modifications (hPTMs) that regulate enhancer activity (Bonasio et al., 2010; Lelli et al., 2012) (Fig. 3A). The nature of the recruited cofactors, which include chromatin-modifying enzymes (Bonasio et al., 2010), determines the type of hPTMs that ultimately either inactivate or stimulate transcriptional activity. Identifying CREs and correlating their differential activity with the differential expression of the genes they regulate in a given tissue between migrants and non-migrants would be a powerful approach to pinpoint important candidate migratory genes and genomic regulatory elements. Of note, genomic variation associated to variation in migratory phenotypes in population genomic studies could occur in these open chromatin regions. DNase-seq, a method that identifies genome-wide nucleosome-free DNA regions that are hypersensitive to DNase I digestion (Song and Crawford, 2010), has been used extensively to identify gene regulatory elements in open chromatin regions. The large amount of starting material it requires may, however, be an important limitation on its application to migratory species. ATAC-seq (assay for transposase-accessible chromatin followed by sequencing), a technique relying on the integration of a mutated hyperactive transposase into open chromatin regions emerged as an attractive alternative as it requires 1000-times less starting material than DNase-seq (Fig. 3A) (Buenrostro et al., 2013, 2015). The activity status of CREs can be determined using chromatin immunoprecipitation (ChIP) of flanking histone marks followed by high-throughput sequencing (Fig. 3A). These are conserved across organisms, and antibodies against marks associated with active and poised/inactive promoters and enhancers (Creyghton et al., 2010; Shlyueva et al., 2014; Heinz et al., 2015) are commercially available.

DNA methylation is another gene regulatory mechanism by which behavioural and physiological plasticity can be regulated (Lyko et al., 2010; Foret et al., 2012; Herb et al., 2012; Pegoraro et al., 2016). Reprogramming of gene expression and gene regulatory networks between different migratory phenotypes could occur through changes in DNA methylation patterns, which can be studied using bisulfite-sequencing methods, even for species without a reference genome (Klughammer et al., 2015). The existence of such a mechanism may, however, be taxa-, lineage- or species-specific. While genome-wide patterns of DNA methylation occur broadly in vertebrates, including birds (Li et al., 2011; Laine et al., 2016), DNA methylation does not appear to be ubiquitous in insects (Bewick et al., 2017). Some species such as monarchs seem to lack detectable patterns of de novo DNA methylation (Zhan et al., 2011; Bewick et al., 2017).

Post-transcriptional regulation of gene expression by microRNA (miRNA), RNA editing and alternative splicing should also be considered as potential mechanisms by which steady state mRNA levels could be differentially regulated between migratory phenotypes and/or seasons. So far, with the notable exception of differential expression of miRNAs in migratory and non-migratory forms of the monarch butterfly (Zhan et al., 2011), to our knowledge, none of these mechanisms has been studied in any other migratory species. The current array of HTS approaches offers unprecedented opportunities to exploit all these possibilities, and should accelerate the identification of genes, epigenetic and post-transcriptional mechanisms underlying the migratory phenotype.

Ultimately, providing proofs of causality will be imperative. Genetically manipulating identified candidate genes and regulatory regions in vivo to assess how genetic disruption affects the migratory physiology and behaviour will undoubtedly be challenging, but should be facilitated by increasingly accessible functional genomic tools. Classical approaches that use RNA interference (RNAi) to knock genes down are possible ways to test gene function, but could have limited applications, as the gene knockdown is often incomplete. Unlike RNAi, targeted genome editing allows the generation of complete gene knockouts and the creation of mutations in CREs that would render them non-functional. The use of RNA-guided nucleases derived from clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system is arguably one of the most exciting avenues of investigation for the study of migration genetics because of its simplicity and applicability to virtually any species of interest. CRISPR/Cas9 enables site-specific genetic modifications by inducing DNA double-stranded breaks (DSBs) at genomic locations of choice. DSBs repair by the cellular machinery stimulates the introduction of small random insertions/deletions through non-homologous end joining (NHEJ) repair, or the introduction of donor constructs through homology-directed repair (Fig. 3B) (Gaj et al., 2013; Kim and Kim, 2014).

Thus far, CRISPR/Cas9 has been successfully applied to generate full knockouts and deletion mutants in the migratory monarch butterfly (Markert et al., 2016; Zhang et al., 2017), thereby unlocking its potential for the functional characterization of migratory candidate genes and CREs. Germline genome editing in this species has been achieved with high efficiency by injection of Cas9 mRNA and a single guide (sg)RNA into fertilized embryos within the first nuclei divisions (Fig. 3B) (Merlin et al., 2013; Markert et al., 2016). Similar strategies are commonly used in model species such as the zebrafish and mouse, in which embryos can be collected at an early development stage, and should therefore be transferable to other migratory species such as fish and mammals. In birds, efficient germline transformation has proved to be more challenging due to constraints associated with the structure of the zygote and the opacity of the oocyte (Dimitrov et al., 2016). However, the use of CRISPR/Cas9-mediated modification of primordial germ cells in vitro has been shown to overcome this limitation in chickens (Dimitrov et al., 2016; Oishi et al., 2016; Woodcock et al., 2017), making it a possible approach for gene editing in migratory bird species. In the event that breeding colonies in laboratory conditions preclude the implementation of similar approaches, strategies for genome editing in post-mitotic cells using virus-mediated delivery of CRISPR/Cas9 into the appropriate tissue/organ, depending on the presumptive function of the gene of interest, are attractive alternatives (Fig. 3B). Expanding genetic tools in migratory species for knocking genes out both in the whole organism and in a tissue-specific manner, as well as for knocking in reporters that can be used to probe circuit function, will be important milestones for the field. Combined with diagnostic physiological and behaviour assays, genetically engineered migratory organisms will ultimately facilitate a systematic dissection of the molecular and cellular mechanisms underlying animal migration and orientation.

The identity of genes underlying migratory orientation and how they code for this remarkable behaviour remains largely unknown. Applying cutting-edge HTS technologies and powerful behavioural approaches in the European blackcap and the North American monarch butterfly hold great promise to identify and functionally validate migratory genes and molecular pathways. However, restricting genetic and molecular approaches to only a few species can limit our ability to differentiate the molecular pathways that are species- or taxa-specific from those widely generalizable across taxa. Exploiting a larger number of strategically chosen migratory species for comparative studies across evolutionary scales and taxa will be key, and will be facilitated by the increasing accessibility of genomics and molecular tools in non-model organisms.

Among other migratory insect species, the nocturnal Australian bogong moth Agrotis infusa is an ideal complement to the diurnal monarch butterfly for molecular comparative studies. Each spring, migratory moths leave the breeding grounds in southern Queensland and northwestern New South Wales and migrate south to reach the Australian Alps, where they aggregate to aestivate (i.e. spend the summer in a state of dormancy) in cool mountain caves. In the autumn, the same individuals migrate back to the breeding sites to reproduce, and die soon after, leaving their progeny to become the new generation of southward migrants the following spring (Fig. 4A) (Heinze and Warrant, 2016; Warrant et al., 2016). The opposite flight orientations taken by bogongs and monarchs each autumn and spring in different hemispheres of the Earth, where seasons are reversed, may suggest that similar environmental cues (i.e. photoperiod and/or temperature changes) recalibrate compass orientation. Leveraging this parallel using genomic comparative approaches may accelerate the identification of a set of conserved molecules and pathways underlying migratory orientation in insects. All the genomic tools mentioned previously, including CRISPR/Cas9 that has been implemented in other moth species (Koutroumpa et al., 2016; Chen et al., 2018), should be easily transferable to the bogong.

Although the spectrum of migratory orientation is exceptionally diverse in blackcaps, other bird species also exhibit differences in migratory timing and direction. Two subspecies of willow warbler form a migratory divide in central Scandinavia, and WGS combined with SNP-chip data identified two chromosomal regions showing strong genetic differentiation between subspecies associated with differences in migratory behaviour (Lundberg et al., 2017). The Swainson's thrush is another compelling model forming a migratory divide in western North America, and hybrids in this system have been shown to take intermediate migratory routes (Delmore and Irwin, 2014). Variation in one genomic region linked to variation in migratory behaviour was identified, which included the Clock gene (Delmore et al., 2016). However, the overall lack of consistency between identified regions and genes linked to migratory traits across species may call for caution in assuming one common and simple genetic basis of migratory traits (Delmore and Liedvogel, 2016). Performing comparative analyses of gene expression and of the activity of CREs present in the regions under selection in different bird systems could help determine whether the molecular pathways responsible for flight orientation in birds are conserved or not.

Increasing the number of vertebrate taxa to include fish and mammals will be equally important to unravel molecular mechanisms underlying vertebrate migration and orientation behaviour. Mass seasonal migrations are also performed below the ocean surface by many fish species. Indirect evidence for a genetic basis of migratory orientation comes from a study on Atlantic eels. American (Anguilla rostrata) and European (Anguilla anguilla) eels both start their migratory journeys in the Sargasso Sea, but their migration differs in distance and direction. American eels migrate towards the coast of North America while European eels take a longer route to Europe. Icelandic eels, which have been characterised as possible hybrids (Albert et al., 2006), migrate using an intermediate orientation. The Salmonidae are another well-suited family to study the underlying genetics of fish migration and its spatiotemporal characteristics because of the high variability in migratory life history strategies within and among species (Dodson et al., 2013). In these species, the high variability in migratory phenotype is not mirrored in overall genetic differentiation (O'Malley and Banks, 2007). Variation within the Clock gene has however been shown to match a cline in spawning time in the Chinook salmon O. tshawytscha (O'Malley and Banks, 2008; O'Malley et al., 2010). The genetics of adult migration timing have been further characterized through genome-wide association mapping of migratory (steelhead) populations exhibiting two distinct timing strategies in the wild (Hess et al., 2016). This study identified a 46 kb region overlapping an oestrogen receptor cofactor GREB1, a relevant candidate gene since upstream migration happens during sexual maturation in steelhead and other salmonids (Choi et al., 2014). The same gene has recently been shown to have a major effect on adult migration timing in both O. mykiss and O. tshawytscha (Prince et al., 2017).

Research in O. mykiss, a species with both resident (rainbow trout) and migratory (steelhead) phenotypes, also identified several genomic regions associated with adaptations during smoltification, the process facilitating transition in preparation for seaward migration (Fig. 4B) (Nichols et al., 2008; Hecht et al., 2012; Hale et al., 2013). Epigenetic regulation of gene expression may contribute to controlling variation in migratory behaviour during the smoltification process, as differentially DNA methylated sites were found to distinguish migratory from resident lines, including in regions associated with circadian rhythm pathways (Baerwald et al., 2016). Correlating differential DNA methylation to differential gene expression in migratory versus resident forms of O. mykiss could be used to identify promising candidate genes underlying fish migration.

Bats living in temperate climate also embark on massive seasonal migrations with navigational abilities that are no less impressive than those of birds, insects and fish. Two examples of migratory bats with untapped potential for comparative genomics and molecular studies are the lesser long-nosed bat Leptonycteris yerbabuenae, and the Mexican free-tailed bat Tadarida brasiliensis. These bats, which are found in the southwestern United States and northern Mexico in the summer, migrate southward in the autumn to their respective overwintering sites in central or southern Mexico. Mating occurs in the early spring and pregnant females migrate north to give birth in maternity roosts before returning in the autumn, along with their progeny, to central Mexico (Fig. 4C) (Cockrum, 1967; Wilkinson and Fleming, 1996). Variation in migratory tendency (migratory versus residents) has also been reported in T. brasiliensis populations (Cockrum, 1967). Both the seasonal change in migratory direction and the variation in migratory phenotype could be exploited to uncover the molecular pathways underlying bat migration and orientation. Initiatives to sequence the genome of all bat species are underway (BAT 1K; http://www.bat1k.com). With blueprints in hand, applying cutting-edge genomic and epigenomic tools in these species (from blood samples, wing punches or even brain tissues in the case of the abundant T. brasiliensis) should take off.

The cyclic, mostly seasonal, nature of many species migrations up in the sky, on land and in the oceans has fascinated scientists and the public alike for decades. Its broad occurrence across the animal kingdom begs the question of whether common mechanisms or different molecular toolkits underlie migratory strategies. The involvement of genes associated with clock function in different taxa suggests that mechanisms timing the migration may use a similar molecular pathway. In contrast, the key genomic regions and molecular underpinnings responsible for variation in migratory orientation remain largely unknown as the regions identified in different species so far lack consistency and have not been validated functionally. The answers are out there somewhere in the DNA sequences, and systematically implementing integrative approaches in a diverse array of species should accelerate the pace at which we learn where to look to unveil some of the molecular secrets behind migratory behaviour.

We thank Gernot Segelbacher, Krista Nichols, MonarchWatch, Ajay Narendra, Jens Rydell, and Melinda Baerwald for allowing use of their photos in Figs 1, 2 and 4; Andrew Clarke for generating the map in Fig. 4B; and Aldrin Lugena for comments on the manuscript.