A Review on Dispersal Polymorphism in Wing-Dimorphic, Mono-Morphic, Wingless, and Range-Expanding Insects, and Experimental Designs for Sorting out Resident and Disperser Phenotypes

Dispersal represents a key life-history trait with several implications for the fitness of organisms, population dynamics and resilience, local adaptation, meta-population dynamics, range shifting and biological invasions. Plastic and evolutionary changes of dispersal traits have been intensively studied over the past decades in entomology, in particular in wing-dimorphic insects for which literature reviews are available. Importantly, dispersal polymorphism also exists in wing-monomorphic and wingless insects, and except for butterflies, fewer syntheses are available. In this perspective, by integrating the very last research in the fast moving field of insect dispersal ecology, this review article provides an overview of our current knowledge of dispersal polymorphism in insects. After having provided a definition of the main terms characterising insects’ movements, some of the most often used experimental methodologies for the separation of dispersers and residents in wing-monomorphic and wingless insects are presented. The existing knowledge on the morphological and life-history trait differences between resident and disperser phenotypes is then synthetized. The fourth part examines the effects of range expansion on dispersal traits and performance, in particular for insects from range edges and invasion fronts. Finally, some research perspectives are proposed in the last part of the review.


Introduction
Dispersal corresponds to the movement of individuals or populations from the natal (natal or postnatal dispersal, see [1][2] for examples) or breeding (breeding dispersal, see [3] for example) habitat to another breeding habitat [4][5]. This type of animal movement is of particular importance for the stabilization of population demography and dynamics, by balancing the departure of individuals with arrival of other conspecifics (reviewed in [6]). Variations in dispersal success and rates are common among individuals and species, and plastic and evolutionary changes of dispersal traits are often the consequence of spatio-temporal variations in the fitness performance [7]. Variability of dispersal traits can result from different mechanisms, including the risk of inbreeding, which can be deleterious for fitness performance ([8]; Inbreeding depression: [9]), decreases kin competition [10] or competition for resources [5]. Dispersal also shapes the spatio-temporal distribution of the genetic diversity of species, in parallel to increasing the proportion of the total genetic diversity of populations [11]. Of note, by reducing genetic drift, dispersal may save populations from local extinction [12], and allow the survival of low density populations (Rescue effect: [13]).
Dispersal capacities are of significant importance for setting and reshuffling the geographic distribution of species. Nowadays, climate change has led to poleward shifts across many taxonomic groups, both on land and in the oceans [14][15], and upwards range shifts within montane ecosystems [16][17]. The extents of these shifts are, at least partially, strongly supported by dispersal capacities of individuals [18][19]. In addition, dispersal enhances the persistence of populations thriving in fluctuating environments [20], and is expected to be more frequent in disturbed or stochastically variable habitats [21][22]. This can be illustrated by the propensity for ballooning dispersal of spider mites, which is increased 5.5 times in specimens from disturbed habitats as compared with their relatives from stable habitats [23]. Importantly, as the increased gene flow allowed by individual dispersal can be random or non-random (see for instance [24]), this can have significant cascading effects on local adaptations of populations. For instance, immigration of insecticide-resistant mosquitoes increases the frequency of resistant alleles within the population, while influxes of susceptible or less adapted mosquitoes would decrease their frequencies (see the review of Miller and Sappington [25]), in turn lowering the resilience of the population. Finally, the ecological importance of insect fluxes also scales up to higher organizational levels, by having for instance significant positive consequences for ecosystem services [26].
All animal species can disperse, but the dispersal rate, the frequency of this event, and the geographical distances that can be covered, can greatly vary among individuals, populations and species. For instance, individuals of a population of the damselfly Coenagrion mercuriale can exhibit either short-or long-distance dispersal strategies [27]. Often in animal populations, several individuals remain at their natal site (i.e. philopatric individuals), and a decreasing number of individuals successfully settles outside their natal habitat or home range (i.e. dispersers). Maintaining high dispersal capacity performance can be costly for individuals, and dispersal should be advantageous if benefits (increased fitness of the individual) in the new breeding habitat exceed the costs resulting from dispersal thus conferring a selective advantage [22] (reviewed in Bonte et al. [28]; but see [29] who suggested that benefits could be null for dispersing individuals, and only in favour of residents). Consistently, it is often assumed that individuals having the highest dispersal capacities should benefit from the weaker competition in their new breeding habitat [30][31], thus favouring a higher reproduction, growth and developmental performance.
Given the importance of dispersal in driving the ecology and evolution of organisms and populations in the ever-changing environmental conditions, the causes, mechanisms, consequences and costs of dispersal have been studied in a variety of insect models [3,28,[32][33][34][35]. Several literature reviews have been published in this field in entomology; yet, these valuable studies are most often species or genus specific [36][37], or focus on the costs of dispersal [28], on trade-offs with reproductive traits [38], or on available methods for the monitoring of insect dispersal [39]. Moreover, since the publication of the book 'Dispersal in ecology and evolution ' [6], several new studies have been conducted on this fast moving topic. Thus, the present review, which integrates the more recent advances in the field of insect dispersal ecology, aims at giving an overview of our current knowledge of dispersal polymorphism in wing-dimorphic, monomorphic, wingless, and range-expanding insects. By focusing on active dispersal, i.e. insects dispersing by walking, flying, or swimming, the first part of the article briefly summarizes the different wording and definitions used for defining insects' dispersal. While working with wingdimorphic insects eases the differentiation of disperser versus resident individuals, this distinction is less obvious for wing monomorphic insects, and can be even more difficult for non-flying (or poorly flying) ones. The different experimental systems that have been designed for separating disperser and resident insects are thus reviewed, and this second part includes some suggestions of the parameters that could be considered in future studies for manipulating dispersal propensity, rate and success. As insect dispersers are not a random subset of their population, and rather exhibit a suite of traits which offset dispersal costs and increase the probability of dispersal being successful [28,41], the existing knowledge on the morphological and life-history trait differences between resident and disperser phenotypes is synthetized in a third part of the article. Finally, dispersal is studied extensively in the context of range-expanding insects, including biological invasions, and there is supporting evidence that spatial sorting may contribute selecting dispersive phenotypes at the front distribution margins. The knowledge of these ecological differences, possibly related to enhanced dispersal capacities at range edges, is discussed in a fourth part.

The different terminologies used for describing the movements of insects
The geographical scale at which dispersal occurs can be highly variable, and greatly varies among populations and species. The large variation of dispersal distances has often made it difficult to clearly define dispersal from other types of movements [ '. In several studies, the definition and use of the terms dispersal and migration is debated. To try to solve the inconsistent use of these two terms, Dingle [59] suggested the use of 'ranging' instead of 'dispersal', and defined 'ranging' as 'movement over a habitat to explore it, and movement ceases when a suitable home range is located'. However, despite this suggestion, and even if there is no simple definition of dispersal, the most often adopted one in entomological studies presents dispersal as 'any movement of individuals or propagules with potential consequences for gene flow across space ' [4], thus corresponding to insects moving beyond their neighbourhood. In this line, Renault et al. [40] also suggested that dispersal can be functionally discriminated from the two other forms of movement, i.e. routine/home range movements and migration. Specifically, these authors reported that dispersal is being characterised by a distinct frequency over the life of an organism, as compared with the two other movement types, has an effect on gene flow (as defined by Ronce [4]), little link with seasonality, and moderate preparatory physiological changes.
Two main types of dispersal can be observed, as mentioned in the introduction: (a) natal or pre-breeding dispersal, when the insect leaves its birth habitat to reach a distinct habitat for reproduction (new breeding site) [1][2], and (b) (post)breeding dispersal, when the individual leaves the habitat where it was reproducing and reached another reproduction site (movement from one breeding site to another) [3], thus resulting in a relocation of the reproductive habitat of the individual [60]. In both cases, truly active dispersal encompasses three main phases : (i) departure (or emigration), i.e. the decision of the insect to leave a patch, (ii) the transfer phase, corresponding to the path taken by the individual, with the mosaic of biotic and abiotic parameters it will encounter, and (iii) the settlement (or immigration) phase in which the new habitat of the individual is determined [5, 40].

Experimental methods for separating resident from dispersers in wing-monomorphic or wingless insects by direct observations
In many populations, most individuals of the population remain at their natal site (philopatric insects), and few insects successfully move and establish outside their natal (or former breeding) habitat or home range (i.e. dispersers) [4,22]. Dispersal polymorphism has thus been commonly observed in insects, enhancing the performance of (a) fecundity and growth (philopatric insects) or (b) dispersal capacities (dispersers), in turn resulting into dispersal-related life-history tradeoffs (see the example of the butterfly Melitaea cinxia, [61], and see [28, 62] for reviews).
In wing-dimorphic insects, dispersing and resident phenotypes can be easily separated (dispersers are winged, or they are long-winged, while residents are unwinged, or they are shortwinged). Conversely, the distinction of disperser and resident phenotypes, or highly mobile versus less mobile insects, is less evident in wing monomorphic or wingless insect species. As a result, different techniques have been elaborated for their separation, both in the field (landscape scale studies) and under controlled conditions [39]. An overview of the field techniques for measuring insect dispersal can be found in Feldhaar and Schauer [63], with saproxylic species as insect models. Indirect observations can be conducted with molecular techniques, as for instance done by Suchan et al. [64] who undertook a metabarcoding work on pollen collected from butterflies to determine the source origin of the individuals according to the patches they visited before they were sampled. For experiments under controlled conditions, the methodology based on tethered flight mills, a system that is used for measuring flight behaviour and dispersal capacities in insects, has been recently reviewed (see [65][66]). In the below sections, the article will rather focus on the other experimental systems allowing the assessment of insect dispersal propensity and polymorphism in the laboratory at small spatial scales.

Assessing dispersal polymorphism under controlled conditions: photos and videos
The movements of insects, either they are corresponding to foraging, dispersal or migration, can be observed by visual monitoring, including photos, or by videos. These experiments allow assessing dispersal tracks by regular records of the position of the insect. For instance, video records have been realized for measuring the movements of nymphs and adults of the brown marmorated stink bug Halyomorpha halys in Petri dish arenas (diameter: 10 cm) at 25 °C in the dark (fluorescent lights were used for facilitating the records) over a period of 1h [67]. A similar design has been used by Socha and Zemek [34] who investigated the walking patterns of the bug Pyrrhocoris apterus, except that the arena consisted of a white formica cylinder having a diameter of 125 cm and a height of 62 cm; in addition, each insect was allowed to acclimate for 10 min before the record started for 1h at 26 °C. Matsumura and Miyatake [68] compared the walking activity of Tribolium castaneum by recording the movements of the insects placed in a Petri dish of 35 mm diameter and 10 mm height over 30 min at 25 °C; before each record, the individual was acclimated for 2h.
There are two possible issues with these direct observations. First, by manipulating the insects when they are transferred to the arenas, their subsequent movement behaviour may be altered (the same remark applies to the experimental designs presented below). Second, when working with photos and videos, our capacity to (automatically) discriminate individuals from their medium can be a technical issue. Contrasting coloured substrate / medium can be used to ease the discrimination, but this procedure is then likely to overstimulate insects' movements.

Assessing dispersal polymorphism under controlled conditions: experimentally connected patches in the absence of hostile conditions
By browsing the available literature, it appears that we are missing a consensus regarding the type of apparatus that is being used for separating dispersers and residents, in particular when different studies are working with the same genus or species, as for example in Tribolium sp. [69][70][71][72][73]]. An experimental system for investigating the movement of adult Tribolium castaneum, which flies but also moves by walking [74], was designed by Prus [69]: two containers (A and B) were bridged with a plastic tube of 4.5 mm inner diameter (ID); container A contained medium, while no medium was added to container B. The distance of the connecting tube from the bottom of container B made it impossible for the insects to move again to container A. Dispersal was checked every 24h over 10 days at 29 °C. Later, Łomnicki [75] worked with a chain of five containers, four of them consisted of 157 mL glass bakers (A to D, A containing an old medium with eggs; C a new fresh medium, and no medium in containers B and D), and the terminal one (E, containing fresh medium) being a 50 mL plastic beaker. This suite of containers was connected with a glass tube of 4 mm I.D., and the experiment lasted five weeks, i.e. the time necessary to obtain imagoes that could disperse. After the seventh generation, and thus the seventh dispersal assay, the proportion of red flour beetles that dispersed from beaker A ranged from 0.60 to 0.95 in dispersing insects, and from 0.05 to ca. 0.60 in their resident relatives [75].
Working with the fruit fly Drosophila melanogaster, Edelsparre et al.
[76] used a containerto-container system consisting of two 50 mL containers filled with 2 mL of rearing media or agar, joined by a 1 mL tip which prevented the flies to return to the former container, and maintained at 24 °C. Dispersal was assessed after 6h in 32 5-to 7-day-old flies for each assay. Tung et al. [77] also used two containers (1.5 L each) system; the source container was empty, and the destination one contained wet cotton. The containers were connected by a 2 m-long tube (transparent plastic tube of 1 cm I.D.) whose length was progressively increased so that it reached 10 m at the 33 rd fly generation. The system was designed to limit as much as possible the backflow of the flies. At the start of the experiment, 2400 flies were placed in the source container, and the assay was run for 6h at 25 °C. In D. melanogaster, dispersal seems to be linear over time, as Tung et al. [77] reported that ca. 25% of the flies dispersed after 3h, and 50% after 6h, a conclusion which is also in line with the results obtained by Arnold et al.
[72] on T. castaneum. Importantly, photoperiod, which is not always reported in dispersal studies, may have a significant role on flight initiation in T. castaneum, as shown by Drury et al. [74] who tested the effect of light on the propensity of the beetles to fly.

Assessing dispersal polymorphism under controlled conditions: experimental systems incorporating hostile conditions
In natural environments, dispersers will likely have to cross physical barriers (climb, obstacles, wind, etc…), and will have to deal with different soil substrates that could make walking more difficult. The introduction of a hostile matrix in the dispersal path connecting the two containers (patches) has thus been considered some experimental systems (see [78]), as it increases the likelihood that movement among patches is indeed dispersal (and not foraging). In this perspective, the selection of an adequate hostile matrix should supported by the existing knowledge of the biology and ecology of the tested organisms.
Working with T. castaneum, Arnold et al.
[72] used three containers (A, B, C) of 70 mL sequentially connected with a flexible tube of 4 mm I.D. Containers A and C were filled with 15g of flour, and container B had a filter paper only. These authors then designed five experiments, differing in terms of distance among the containers (70, 120, 165, 310 or 620 mm), and angle made by the flexible tube connecting the containers (angle of the tube of 4, 8, 16, 24 or 55° in between two containers), so that the dispersal difficulty was increased. Dispersal assessment was monitored twice a day over four days; the temperature was not mentioned, but was most probably the one that was used for insect rearing (29.5 ± 1 °C). For each experiment, 50 adults (males or females) aged of three days were used. The main findings of this study were: (a) a higher angle of the connecting tube, i.e. a more difficult dispersal path, lowered the dispersal success of males and females of T. castaneum, and (b) the increased length of the tubes connecting the containers increased the duration necessary for reaching the terminal container. Morrison et al.
[79] modified the Arnold's et al. [72] system by working with two containers separated by 25, 75 or 175 cm, and connected by PVC pipes of 5 mm I.D.; the first container had no food, whereas the second one was filled with 20g of organic flour. The experiment was run with 20 insects of a single sex for the beetle T. castaneum or of mixed sex for the beetle Rhyzopertha dominica for 48h at 30 °C. In this study, mainly designed for investigating the effects of short exposures to pyrethroid (long-lasting insecticide-incorporated netting), the authors reported a lower dispersal ability of the treated insects in comparison with control ones.
In addition to the creation of a hostile matrix connecting the two patches, some studies may require the introduction of biotic cues that will also increase ecological realism. In their study, Fronhofer et al.
[78] added a visual, chemical or auditory cue as a proxy of predation risk. For the butterfly P. brassicae, the cages were connected by a S-shaped corridor having hostile conditions for the insect (dark, warm, and poorly vegetated); the presence of two toads in the pond of the departure cage created visual and olfactory predatory cues, in addition to the presence of two crushed butterflies in a tube. The same cages and hostile matrix was used for the damselfly Platycnemis pennipes, but the predatory cue was obtained by adding two frogs which created auditive, chemical and visual signals for the insects. In these two experiments, the dispersal was monitored daily over 4 and 5 days, and the temperature of the experiment was not mentioned [78]. In the same study, the dispersal of the marsh cricket Pteronemobius heydenii was also assessed with an initial and a destination patch consisting plastic containers (130 L), connected with a narrow plastic pipe of 4 cm I.D. covered by a thin layer of soil (hostile matrix). Lizard olfactory cues were used for simulating predation risk, and the dispersal was monitored over five days at 16 -25 °C with daily observations. In the three experiments, the authors also tested the effect of the availability of trophic resources on dispersal propensity, by having a low and a high resource treatments [78].
Finally, the assessment of the flight endurance of the insects subjected to stressful conditions has been measured with a vortex system in butterflies, [80][81]. In this procedure, butterflies are assessed individually for mobility performance in a 250 x 100 x 100 mm plastic container; they were acclimated for 30 s before being vortexed for 60 s at 25 °C. The time spent flying for each individual during this stressful minute was recorded, and represents a good correlate of dispersal ability, i.e. dispersers are characterized with good abilities to maintain flight in these stressful conditions, while residents have a lower flying performance [82].

Standardization of the experimental systems with an emphasis of some important parameters affecting dispersal results
The investigations conducted on the beetle T. castaneum and on the fly D. melanogaster (see sections 3.2 and 3.3) example the lack of consensus regarding the different experimental systems employed in entomological studies in the topic of dispersal ecology. While the use of a source container (patch), a dispersal path (corridor), and a destination container (path) are all components of the experimental dispersal apparatus under controlled conditions, the number of destination containers, the length of the dispersal path and its degree of hostility for the tested insect, the number of insects in the source container, the duration of the assay, and the presence / absence of food / medium in the containers can greatly vary among studies, in spite of the critical effect these factors can have on dispersal propensity and success.
In order to get a first overview of the variables of potential interest when designing an experimental system assessing dispersal propensity and performance in entomological studies, a short experiment was conducted for the purpose of this review. The biological model for this trial was the tenebrionid beetle Alphitobius diaperinus. Two containers (patches) of 110 mL were used, connected by a 1.6 m long plastic tube of 13 mm I.D. making an angle of 15 ° from the source container (containing wheat bran and one piece of carrot -used as an oviposition site for the species) to the destination container (containing a layer of 0.5 cm of sand at the bottom of the container). Before the containers were connected, the beetles were acclimated for 24h in the source container at 18 °C. Then, the assay was run for 24h at 18 °C, and 19-20 unsexed insects were placed in the source container. In this experiment, the numbers of dispersing, in transit, and resident individuals were 32 ± 5, 24 ± 5, and 43 ± 6 % (the proportions were marginally significantly, ⲭ²=5.37 ; 2 ddl ; N=213 ; P=0.07). In another test, the angle in between the two containers was removed, the effects of temperature (18 versus 25 °C), duration of the assay (24 versus 8h), nature of the medium (sand versus wheat bran) and number of insects (20 versus 40 adults) in the source container were assessed. As expected, temperature had a prominent effect on the dispersal propensity, with almost 100 % of the 20 beetles that dispersed at 25 °C, and many of them (about 16 insects over the 20) reached the destination container in all replicates (N = 7) after 24h; by reducing the duration of the assay to 8h, the number of beetles that reached the destination container was around 10 in all replicates. When assessing the effects of the number of beetles in the source container, we found that the proportion of dispersing insects was two times higher in assays that contained 40 beetles as compared with those having 20 beetles (note that in this assay, the distance in between the container was 2.4 m). Importantly, the nature of the medium used in the source container made no differences in dispersal rates.
In sum, the results available from the literature and the complementary tests conducted for this review indicate that it would be valuable to get additional methodological information in in the published literature. Some of the parameters that should be mentioned and that can be manipulated, together with their possible effects on dispersal, are presented in Table 1. Particular attention should be given to the distance among the source and destination container, as this dispersal path distance and the duration of the assay contribute to determine if the insects will be sorted out according to their mobility (foraging movement, vagrancy) or according to their dispersal capacities (movements that drive gene flow, as defined by Ronce [4]). Moreover, the harshness (permeability) of the dispersal path (or matrix) is a an important factor that can modulate both dispersal success and dispersal syndrome characteristics of dispersers versus resident organisms, as recently reported in ciliates [102]. The hostile matrix can be of different nature: absence of (trophic) resources, low to null humidity conditions, shaded / exposed / dark dispersal paths, temperature lower / higher than the thermal preferendum of the species, olfactory cues repulsive / attractive for the species, angle and length of the tubes connecting the containers, internal diameter of the connecting tube. A primary 'control quality' of the experimental systems designed for sorting out dispersers and residents can be obtained by observing dispersal rates (propensity). In order to increase the ecological realism of the study, the dispersal propensity should range from 10 to 50% of the individuals leaving the initial container / patch; in butterflies Stevens et al. [103] reported that dispersal propensity had an average of 34%.

Variable of interest
For the insect, the variable has an effect on    of the butterflies they studied from a managed forest could represent dispersal phenotypes. 129 130

Main characteristics of disperser and resident insects
Interestingly, colour patterns of the wings have recently been reported as a morphological 131 proxy that could be used for sorting out noctuid moth populations with long dispersal distance 132 capacities (high variation in colour patterns among individuals) in comparison to their resident 133 conspecifics (little variation in wing colouration among individuals) [151]. As more variable colour 134 patterns has also been found to increase population abundance and stability in noctuid moths 135 [152], changes in colour pattern variability along an invasion gradient could strengthen the range 136 expansion dynamics of invasive insects. Additional observations of possible correlations among 137 wing colour patterns, thermal tolerance, dispersal motivation and success should be further 138 examined, as it could have several implications for improving our predictions of range expansion 139 in contexts of climate change and habitat fragmentation. 140 141 Differences in dispersal capacity may also be supported by distinct behaviours of the 142 insects according to the environmental conditions of their habitats. In a spider species, for 143 example, there is little or no detectable phenotypic difference between dispersers and residents, 144 but the conditions encountered in the natal patch during the juvenile stage influence the dispersal 145 strategy of the individuals. Specifically, some wolf spiders will disperse over long distances using 146 silk threads (dispersers) and others will remain relatively sedentary by only moving over short 147 distances while abseiling along the plants [153]. Social behaviour, in correlation with the 148 morphology of the insect, can also have links / consequences on dispersal patterns [154]. In the 149 beetle Librodor japonicus, there are three main body sizes in males (small, medium and large 150 males); large males disperse earlier from the patch than the two other morphs [155]. In parallel, 151 these males are more aggressive, i.e. they fight more frequently than males of smaller sizes who 152 are sneaking. The authors concluded that resources allocation differ among the three morphs, 153 with more investment into testes, wings, and mandibles for the small, medium and large L. 154 japonicus males, respectively [155]. A similar finding has been reported from the armed beetle 155 Gnatocerus cornutus, whose males having a smaller weapon size (lower enlargement of the 156 mandibles) dispersed more and have a higher spermatogeny expenditure than their counterparts 157 having enlarged mandibles [156]; in the latter phenotype, the rate of remating with the same 158 female is higher and likely explains this lower investment in testis size and volume of sperm 159 production. 160 161

Phenotypic differences in dispersers and residents from wing-monomorphic and wingless 162
insects: reproduction and fecundity 163 164 The mating latency (age at reproduction of males and females) and duration of copulation can be 165 affected by dispersal capacities, as reported in the butterfly    Sci. 2008, 105, 19060-19065. 443