Creating outbred and inbred populations in haplodiploids to measure adaptive responses in the laboratory

Abstract Laboratory studies are often criticized for not being representative of processes occurring in natural populations. One reason for this is the fact that laboratory populations generally do not capture enough of the genetic variation of natural populations. This can be mitigated by mixing the genetic background of several field populations when creating laboratory populations. From these outbred populations, it is possible to generate inbred lines, thereby freezing and partitioning part of their variability, allowing each genotype to be characterized independently. Many studies addressing adaptation of organisms to their environment, such as those involving quantitative genetics or experimental evolution, rely on inbred or outbred populations, but the methodology underlying the generation of such biological resources is usually not explicitly documented. Here, we developed different procedures to circumvent common pitfalls of laboratory studies, and illustrate their application using two haplodiploid species, the spider mites Tetranychus urticae and Tetranychus evansi. First, we present a method that increases the chance of capturing high amounts of variability when creating outbred populations, by performing controlled crosses between individuals from different field‐collected populations. Second, we depict the creation of inbred lines derived from such outbred populations, by performing several generations of sib‐mating. Third, we outline an experimental evolution protocol that allows the maintenance of a constant population size at the beginning of each generation, thereby preventing bottlenecks and diminishing extinction risks. Finally, we discuss the advantages of these procedures and emphasize that sharing such biological resources and combining them with available genetic tools will allow consistent and comparable studies that greatly contribute to our understanding of ecological and evolutionary processes.


| INTRODUC TI ON
Understanding the processes that shape individual traits and ecological processes in natural populations is arguably the ultimate aim of evolutionary ecology. This can be achieved by studying populations in their natural environment (Arnold, 1983). However, this approach suffers from the difficulty in controlling several environmental variables simultaneously (Lauder, Leroi, & Rose, 1993). Laboratory studies, in contrast, while allowing for controlled variables, are often criticized for not being representative of the processes occurring in natural populations (Aguilar, Dong, Warr, & Dimopoulos, 2005;Calisi & Bentley, 2009;Melvin & Houlahan, 2012). This is partly because it is not possible to recreate the complexity of the natural environment in the laboratory (Calisi & Bentley, 2009;Carpenter, 1996).
Another reason is that laboratory populations often do not harbor sufficient variability to produce representative responses. Indeed, some studies have shown that laboratory populations have lower genetic variability than natural populations (Bian, Gao, Lamberton, & Lu, 2015;Norris, Shurtleff, Touré, & Lanzaro, 2001;Stohler, Curtis, & Minchella, 2004), possibly due to bottlenecks during the establishment and maintenance of the population, or to the long-term adaptation to the same environment, that is, the laboratory conditions (Aguilar et al., 2005;Matos, Rose, Pité, Rego, & Avelar, 2000;Santos et al., 2012;Stohler et al., 2004). However, the lack of representativity of laboratory populations may also be related with the origin and the procedures involved in the creation of such populations (Berthier et al., 2010). Natural populations of the same species may be genetically differentiated and/or harbor different genetic compositions, shaped by different geographic and environmental factors (Aguilar et al., 2005;Bian et al., 2015;Langley et al., 2012;Nunes, Neumeier, & Schlotterer, 2008). Thus, laboratory populations founded by individuals collected from a single-field population may not produce representative responses, even if the sample size is enough for the sample to be representative of that population.
To ensure that data obtained in the laboratory reflect the range of possible responses found in natural populations of the species under study, the ancestral population should reflect the variability found in the field (Faria & Sucena, 2017;MacDonald & Long, 2004;Nunes et al., 2008). Several studies have used laboratory populations founded by a large number of individuals collected in the field and maintained at high numbers in the laboratory (e.g., Magalhães, Fayard, Janssen, Carbonell, & Olivieri, 2007;Martins, Faria, Teixeira, Magalhães, & Sucena, 2013;Mery & Kawecki, 2002;Teotónio, Chelo, Bradić, Rose, & Long, 2009). However, this method falls short of accounting for potential geographic variation in trait values across populations. To ensure this, some authors have produced outbred populations by merging clones, inbred lines (King et al., 2012;Kover et al., 2009;Zbinden, Haag, & Ebert, 2008), or field populations (Fricke & Arnqvist, 2007;Tucic, Milanovic, & Mikuljanac, 1995) collected at different locations. This option increases the chance of obtaining a population containing genotypes from different environments, thus potentially representing different subsets of the genetic variability of a species. Yet, this procedure does not preclude the possibility that one (or a set of) genotype(s) from a particular environment is overrepresented in the final population. To circumvent this caveat, in sexual organisms, it is desirable to create an outbred population with an equitable representation of the genotypes present in several field populations, which can be achieved by performing controlled crosses between individuals of different populations.
Using outbred populations not only increases the representativity of the observed responses but also the robustness of comparisons between studies performed in different laboratories, if these populations are shared (Churchill, Gatti, Munger, & Svenson, 2012).
Thus, with the exception of studies on local adaptation, such as common garden and reciprocal transfers experiments (Blanquart, Kaltz, Nuismer, & Gandon, 2013;Kawecki & Ebert, 2004), and of other studies aiming at comparing populations, most types of laboratory studies may benefit from using outbred populations. In particular, understanding the process of adaptation to specific selective pressures, in controlled laboratory conditions, with experimental evolution and quantitative genetic methodologies, requires the usage of populations with large amounts of variability (Kawecki et al., 2012;Svenson et al., 2012).
Experimental evolution follows adaptation of populations exposed to specific selection pressures in real time (Gibbs, 1999;Kawecki et al., 2012). Hence, it allows measuring the process of adaptation itself instead of inferring it based on observed patterns, and to infer causality. This method consists in deriving populations from a common ancestral and exposing them to specific controlled environments during several generations, which enables (a) knowledge of the ancestral state of populations (i.e., the ancestral population from which all others were derived), (b) the possibility to define and control the environments that populations are exposed to, and (c) replication at the population level (Magalhães & Matos, 2012).
The explanatory power of experimental evolution can be used to unravel how populations adapt to environmental changes, to the presence of antagonists or to different population structures (Kawecki et al., 2012;Macke, Magalhães, Bach, & Olivieri, 2011;Rodrigues, Duncan, Clemente, Moya-Laraño, & Magalhães, 2016;Zélé, Magalhães, Kéfi, & Duncan, 2018). Additionally, this method can be used to measure convergent evolution of different populations to a common environment (e.g., the laboratory; Fragata et al., 2014;Simões et al., 2008). In any case, adaptation of nonmicrobial organisms to rapid environmental changes relies mostly on the standing genetic variation present in a population, rather than on the arrival of new mutations (Barrett & Schluter, 2008;Hermisson & Pennings, 2005;Sousa et al., 2019). Thus, for the establishment of experimental evolution populations, it is crucial to generate and maintain populations with large genetic variability in the laboratory, being the above-mentioned outbred populations an excellent tool for that purpose. Moreover, some populations may crash during the evolution process. Therefore, it is useful to design methods that maximize the prevention of such events.
Quantitative genetics uses several designs to evaluate the genetic versus environmental contribution to a particular phenotype (Falconer & Mackay, 1996). In such studies, it is important that the population used to infer these contributions is sufficiently variable. Some designs rely on a panel of inbred lines, which allows identifying any quantitative trait loci involved in one phenotype (Mackay, 2004).
To ensure that this panel is composed of different genotypes, it is important to derive it from a highly outbred population (e.g., King et al., 2012;Mackay et al., 2012). Such panel can then be used to measure the broad-sense heritability of a given trait, as well as genetic correlations between traits (e.g., Travers, Garcia-Gonzalez, & Simmons, 2015;Howick & Lazzaro, 2017;Wang, Lu, & Leger, 2017;Lafuente, Duneau, & Beldade, 2018;Everman, McNeil, Hackett, Bain, & Macdonald, 2019. Although the most famous and complete panels are found in Drosophila (DGRP-Mackay et al., 2012;DSPR-King et al., 2012;GDL-Grenier et al., 2015), this resource has also been used in plants (Kover et al., 2009;Wills et al., 2013) and other animals (Table 1). Outbred populations themselves may also be useful in quantitative genetic designs (Solberg Woods, 2014;Svenson et al., 2012). In contrast with inbred lines, where each line represents a fixed allelic combination, individuals from outbred populations are maintained in randomized recombinant crossings. Therefore, from an outbred population one can retrieve a much higher amount of allelic combinations, allowing a fine mapping of complex phenotypes (Solberg Woods, 2014;Svenson et al., 2012).
Here, we describe the creation of the above-mentioned biological tools, outbred populations and inbred lines, using protocols focused on haplodiploid systems. The creation of hybrid populations using controlled crosses in haplodiploids has an extra layer of complexity as compared to diploid species. This is because in these systems, females stem from fertilized eggs whereas haploid males stem from unfertilized eggs. Thus crosses between different genotypes/ populations only generate hybrid diploid females, and hybrid males can only stem from unfertilized eggs of these hybrid females.
As a case study, we describe the creation of outbred populations for Mus musculus "BXD ARI" founded from 2 laboratory strains (after 9-14 generations of intercrossing, followed by at least 14 generations of inbreeding 46 Peirce, Lu, Gu, Silver, and Williams (2004) two species of haplodiploid spider mites. This was done by performing controlled single crosses between individuals of different populations, within each species, in round-robin and matched crosses designs. From one of these outbred populations, we also created inbred lines through 15 generations of sib-mating, for which we also present a method to calculate the coefficient of inbreeding through time, adapted to haplodiploid species, as well as a method to calculate the probability of having a fully inbred line. Finally, we provide a general description of an experimental evolution protocol, which includes a backup for each experimental population that can be used to replenish the population when needed, maintaining a constant population size at each transfer and minimizing the risk of extinction of such populations. With this work, we aim to provide the community with protocols that can be easily applied, not only to this, but to other systems.

| COLLEC TI ON OF FIELD P OPUL ATI ON S
In order to maximize the representativity of responses observed in laboratory studies, field populations used to create outbred populations should be sampled at different locations. Here, we used spider mites (Acari: Tetranichidae), which are haplodiploid pests widespread in many agricultural crops (Migeon, Nouguier, & Dorkeld, 2010).
Given their small size and life-cycle characteristics, these species are easily reared and maintained in high numbers in the laboratory. We Subsequently, each population was identified at the species level by performing a multiplex PCR on a pool of 50-100 spider mites , detailed in Appendix S1). A total of 27 populations were collected in 24 different locations ( infesting tomato plants, but on 4 of those, spider mite populations were found on neighboring plants (Table 2).

| CHAR AC TERIZ ATI ON OF FIELD P OPUL ATIONS
In several organisms, different features of the field-collected populations can lead to reproductive incompatibilities between different populations/genotypes and may hamper the maintenance of variability along the creation of outbred populations. Identifying the source of such incompatibilities, and excluding or avoiding them, is thus a prerequisite to the successful creation of outbred populations.
In many arthropod species, including spider mites, the presence of maternally inherited bacterial endosymbionts may hamper the viability of offspring from interpopulation crosses (Duron et al., 2008;Engelstädter & Hurst, 2009;Telschow, Hammerstein, & Werren, 2002). Therefore, we assessed infection by three of the most common reproductive manipulators found in arthropods (Weinert, Araujo-Jnr, Ahmed, & Welch, 2015; including spider mites, e.g., Zélé, Santos, et al., 2018), namely, Wolbachia, Cardinium, and Rickettsia, in most of the field-collected populations (cf. Table 3). Using a multiplex PCR method developed by  detailed in Appendix S1), we found Wolbachia in 6 out of the 14 populations screened, whereas the remaining populations were free of symbionts (Table 3). Subsequently, to avoid incompatibilities among populations due to the presence of endosymbionts, a subset (N > 300 females) of each population selected to create the outbred populations (see below) was cured from endosymbiont infection by heat shock (continuous exposure to 33°C) for 6 generations, a method previously used in T. urticae for the same purpose (van Opijnen & Breeuwer, 1999). Due to potential side effects of the heat shock treatment, this procedure was used for all selected populations, independently of whether they were initially infected by symbionts. All populations were retested after the heat shock treatment to confirm the absence of symbionts.
Reproductive incompatibilities due to genetic differentiation among populations of the same species are a common feature in  Knegt et al., 2017). To avoid such incompatibility, we sequenced the ITS of T. evansi populations (Table 3; detailed in Appendix S1) and used only the populations with ITS type T1, corresponding to clade I, to create the outbred population.

| CRE ATI ON OF OUTB RED P OPUL ATIONS IN HAPLODIPLOIDS
Using different field-collected populations to create outbred labo-  (Table 2; hereafter labeled A to D and E to H, respectively). The populations were merged by performing interpopulation crosses in a controlled match design, to avoid overrepresentation of genotypes from a given population (Figure 2 Table 1). Each step represents the production of offspring to use in the crosses for the following generation: females were obtained from crosses between different genotypes and males from virgin females of a given genotype. Bold arrows represent the development of the offspring forming the next generation, and dashed arrows represent the use of hybrids for the subsequent crosses within a generation parental crossing. The outbred population was founded with 51 females from each of 6 different combinations, corresponding to a total of 306 females. Because the total number of males was low (N = 197), we opted to use them all, even though the number across genotypes was not even. Therefore, the frequencies of each type of cross, at each generation (t + 1), are given by the following equations:

| CRE ATI ON OF INB RED LINE S IN HAPLOD IPLOIDS VIA FULL S IB -MATING
The coefficient of inbreeding (f t ), which corresponds to the probability that two alleles at one locus are identical by descent (Wright, 1921), is subsequently given by the following equation: Alternatively, for full sib-mating in haplodiploids, this coefficient can also be obtained directly as: where the first two terms correspond to the probability of both alleles coming from the grandmother, being the alleles equal (first term), so that f t = 1, or different (second term), so that f t is equal to that of the grandmother f t−2 , and the third term corresponds to the probability of one allele coming from the grandmother and the other from the grandfather, so that f t is the same as that of the mother f t−1 .
Both methods yield the same result, and assuming that generation 0 starts with a [xy] female mated with a [z] male (i.e., A 0 = 1 with the first method and f 1 = f 2 = 0 with the second method), we obtain a coefficient of inbreeding of 95.1% after 15 generations (Figure 3).
However, the first method also allows estimating the probability of having a fully inbred line, which is given by the frequency of individuals stemming from fully homozygous crosses (D t ). Again, assuming the most heterozygotic scenario, we obtain a probability of having a fully inbred line of 93.6% after 15 generations (Figure 3).
To create inbred female lines from the T. evansi outbred population, we randomly sampled 450 mated females, 2 generations after the creation of the population. These females were installed individually on leaf patches, where they laid eggs for 48 hr. The offspring of each female was then allowed to develop until adulthood (10-12 days) and to mate on that patch (i.e., sib-matings). After 14 days, 3 mated females from each patch were isolated on 3 new patches and the same procedure was repeated. On the following generation, 3 sib-mated females from one of the three patches only were isolated on 3 new patches and allowed to oviposit for 48 hr. The entire procedure was then repeated for 15 discrete generations. Having 3 replicates per line decreases the chances that lines are lost at each generation. However, despite this, many lines were lost due to the death of the female, null fecundity, no egg hatching, or no female or male offspring produced by a given female (Figure 4).

| E XPERIMENTAL E VOLUTI ON PROTO CO L
Experimental evolution not only is a powerful method to detect adaptation to specific controlled factors but can also be combined with next-generation sequencing techniques in order to identify and quantify individual loci contributing to adaptation (Magalhães & Matos, 2012;Schlötterer, Kofler, Versace, Tobler, & Franssen, 2015).
For this purpose, several parameters of the experimental design, such as the number of founders, the number of generations, and the number of replicates per selection regime, must be carefully considered according to the system involved (Kofler & Schlötterer, 2014).
However, during the course of the experiment these parameters may be affected due to predictable and unpredictable events (e.g., the loss of replicates due to the extinction of an experimental population). Here, we present a protocol that helps maximizing the prevention of such events. This experimental evolution protocol consists in transferring 220 randomly collected females from the outbred population to a box corresponding to a given selection regime. This number ensures >200 living (due to mortality in the transfer <5%) females founding each experimental population, which is the number needed to maximize the probability of detecting and quantifying responses to selection in spider mites (Sousa et al., 2019). This

| D ISCUSS I ON
We describe the creation of biological resources (outbred and inbred populations) that maximize the maintenance of standing genetic variation in laboratory populations and, thus, increase the representativity of the responses described in laboratory conditions. As a case study, we present the creation of outbred populations for two spider mite species, T. urticae and T. evansi, by performing controlled crosses between recently collected field populations. In addition, we report a procedure to calculate the inbreeding coefficient, but also the probability of having a full inbred line when performing full sibmating in haplodiploids and apply it to the creation of inbred lines of T. evansi. Finally, we provide an outline of an experimental evolution protocol allowing the maintenance of constant densities across generations of selection, thereby reducing the risk of bottlenecks.
Even though the methods we described are applied to haplodiploid species, they can easily be adapted to diploid systems as well. Thus, these protocols may be used in any species that can easily be sampled, have a relatively short generation time and can be maintained in the laboratory conditions in high numbers.
Undoubtedly the method we present here is time and work consuming. However, we believe that the advantages of creating such powerful tools, as we describe here, compensate the effort in the before testing adaptive responses to other environments (Fragata et al., 2014;Matos, Avelar, & Rose, 2002;Simões et al., 2008).
Moreover, while developing these tools, the populations collected may be thoroughly characterized, providing a preview of the variability expected in the derived outbred population and inbred lines.
Here, to illustrate the application of these methods, populations were collected from nearby locations. Therefore, the resulting populations do not encompass large areas of potential geographic variation, unlike, for example, the populations used to create the DSPR and the GDL panels, where the founding genotypes have distant geographic origins (Grenier et al., 2015;King et al., 2012). This may limit the standing genetic variation available, because of similar environmental conditions, and/or migration among populations. However, this is not very likely in the case of spider mites, as (a) experimental evolution studies performed with populations from a single location Gen t+1 = 1 + N t * Gen t + N t−1 * Gen t−1 + N 0 * Gen 0 N total F I G U R E 5 Estimated number of effective generations of selection during experimental evolution of spider mite populations exposed to different environments. Populations were exposed to an environment similar to that of the ancestral population (control), to a new environment, or to a mixture of both (heterogeneous environment). Because individuals from the t − 1 and base populations were often added in the selection regime corresponding to a new environment, the estimated number of generations decreases considerably relative to the other selection regimes. However, this procedure allowed populations to overcome the initial reduction in population size and to subsequently adapt to the selection regime imposed have repeatedly shown responses to selection (reviewed in Sousa et al., 2019), and (b) populations of spider mites show high genetic differentiation even within small geographic scales (e.g., Bailly, Migeon, & Navajas, 2004;Carbonnelle et al., 2007). Additionally, by founding the outbred populations with more than 500 individuals, the chances of obtaining large amounts of standing genetic variation are high.
Performing controlled crosses among field populations maximizes the chances of obtaining a highly outbred laboratory population. Indeed, this method allows controlling for assortative mating and for differences in fitness and/or mating competitive ability be-  (Fricke & Arnqvist, 2007;Tucic et al., 1995).
Using outbred populations increases the chance that the responses observed are representative of the study species, which is a common shortcoming of laboratory studies. For example, Vala, Egas, Breeuwer, and Sabelis (2004) found that T. urticae females that are not infected with Wolbachia prefer uninfected over infected males, thereby potentially reducing the costs of incompatible mattings. However, this result was based on a single line, whereas a later study using an outbred population stemming from several field populations does not recapitulate this result (Rodrigues, Zélé, Santos, & Magalhães, 2018). Another example concerns the interaction between T. urticae and tomato plant defences. Although this herbivore generally induces plant defences, some field-collected lines were shown to suppress them instead (Kant, Sabelis, Haring, & Schuurink, 2008). Therefore, capturing and maintaining natural variation in laboratory studies is highly relevant for understanding the ecology and evolution of the interaction between study organisms, such as spider mites, and many environmental factors, such as symbionts or plants. A particular example of studies that may profit from using outbred populations is those using experimental evolution. As genetic variance is the raw material for selection to act upon, having a highly outbred population to initiate experimental evolution will increase the chances of observing fast responses to selection. However, many experimental evolution studies have been performed with populations or strains collected from a single location, and in some cases from a small number of individuals (but see Fricke & Arnqvist, 2007;Zbinden et al., 2008). Spider mites are no exception to such contingencies (reviewed in Sousa et al., 2019).
Therefore, the responses obtained may be idiosyncratic of the genetic background used. Providing the community with highly outbred laboratory populations may be very useful to test the generality of the responses reported and to perform future studies on other topics with a larger representation of the genetic variation of the species.
Moreover, in experimental evolution studies the initial diversity harbored by the ancestral populations may be quickly lost due to selection and/or stochastic events, leading to the extinction of experimental populations. Indeed, in environments that impose a strong selection pressure there is a high probability that the populations adapting to those conditions crash in a few generations.
Additionally, unpredictable logistical problems that lead to the loss of experimental populations may occur. Here, we outline an experimental evolution protocol that allows using populations from the previous generation of selection (backup t − 1 populations) to ensure that the total population size remains constant across generations, thereby allowing populations to overcome the initial reduction in population size. In this way, it is possible to avoid losing replicates, as commonly occurs in experimental evolution studies (Cooper & Lenski, 2010;Schlötterer et al., 2015;Simões, Rose, Duarte, Gonçalves, & Matos, 2007). Such populations can thus be rescued and subsequently adapt to the selection regime imposed ( Figure 5).  (King et al., 2012;Mackay et al., 2012). The main advantage of this method is that the high genetic variability of the outbred population can be maintained among the inbred lines, while keeping the same genetic background. In particular, studying these lines allows a clear understanding of the phenotypic and genotypic variability for traits that may be relevant in many different contexts. Importantly, inbred lines can also be used to assess genetic correlations and trade-offs between different traits (e.g., Everman et al., 2019;Howick & Lazzaro, 2017;Lafuente et al., 2018;Travers et al., 2015;Wang et al., 2017), including those measured in different environments (Howick & Lazzaro, 2014;Ørsted, Rohde, Hoffmann, Sørensen, & Kristensen, 2018;Unckless, Rottschaefer, & Lazzaro, 2015). Indeed, because all individuals of a given inbred line represent roughly the same genotype, responses of each genotype can be measured in different contexts. Additionally, such inbred lines can be used as a fixed genetic background against which the response of another population is studied. This may be particularly useful in the context of the evolution of biological interactions. For example, unraveling the evolution of sexual conflicts can be done by exposing individuals of the evolving sex to inbred lines of the nonevolving sex (e.g., Macke, Olivieri, & Magalhães, 2014).
Finally, having the same background in the outbred population and the inbred lines allows comparing results stemming from both types of populations when tackling a common question.
The power of the biological resources described here, which can easily be adapted to other organisms, can be further potentiated if they are shared with collaborative laboratories and combined with increasingly fast advances on the genetic and genomic resources available. This will allow consistent and comparable studies that unquestionably will provide great advances in many different frameworks.

ACK N OWLED G M ENTS
We

DATA AVA I L A B I L I T Y S TAT E M E N T
The dataset has been deposited in Figshare repository (https://doi. org/10.6084/m9.figsh are.12263777).