Population connectivity, dispersal, and swimming behavior in Daphnia

Abstract The water flea Daphnia has the capacity to respond rapidly to environmental stressors, to disperse over large geographical scales, and to preserve its genetic material by forming egg banks in the sediment. Spatial and temporal distributions of D. magna have been extensively studied over the last decades using behavioral or genetic tools, although the correlation between the two has rarely been the focus. In the present study, we therefore investigated the population genetic structure and behavioral response to a lethal threat, ultraviolet radiation (UVR), among individuals from two different water bodies. Our results show two genetic populations with moderate gene flow, highly correlated with geographical location and with inheritable traits through generations. However, despite the strong genetic differences between populations, we show homogeneous refuge demand between populations when exposed to the lethal threat solar UVR.

Daphnia has a high potential of movement and dispersal. Propagules can be transported by flowing water (Havel & Shurin, 2004), air (Caceres & Soluk, 2002;Pinceel et al., 2016;Vanschoenwinkel et al., 2008), and animals (van de Meutter et al., 2008;Waterkeyn et al., 2010). They can rapidly colonize new environments even at local scales (Louette & De Meester, 2004Michels et al., 2001). However, the distribution of species or populations is restricted over geographical scales by environmental selection and priority effects affecting genes under selection, gene flow among populations, and the genetic diversity within populations (Orsini et al., 2013). Moreover, isolation by distance occurs along geographical scales when spatial restrictions are coupled with long-term dynamic of distance-dependent dispersal (Fields et al., 2015).
Such escape behavior affects the vertical pattern through the water column (Ekvall et al., 2015;Hansson & Hylander, 2009b), impacting its refuge demand  and the ecology of the ecosystem (e.g., Grossart et al., 2010). Moreover, Daphnia can adapt to multiple environmental stressors by adjusting its morphology, swimming behavior, and life history traits within a few parthenogenic generations . In addition to phenotypic plasticity (Christjani et al., 2016), recent studies have suggested the occurrence of personalities among Daphnia individuals when exposed to UVR (Heuschele et al., 2017).
The concept of personality has been proposed to explain individual variance within ecological populations, for example, with respect to the strength of behavioral responses (Chapman et al., 2011).
However, although the fields of behavioral response and personalities have retained much attention lately, the genetic basis of behavioral responses is rarely addressed. Plaistow and Collin (2014) stressed the importance of investigating multivariate phenotypic plasticity (behavior, physiology, and morphology) associated with the genetic differences within clonal lineages. To our knowledge, only one study has investigated the genetic fingerprint associated with swimming behavior in response to stressors, showing an increased plasticity in accordance with adaptive evolution (Cousyn et al., 2001). Hence, there is a lack of understanding with respect to the genetic behavioral fingerprint in Daphnia and its sustainability within natural populations.
In this study, we assessed the behavioral responses in Daphnia magna associated with natural diel vertical migration in relation to their genetic fingerprints at the population and individual levels.
Because lakes in the Scania region have different history and are influenced by anthropogenic activity (Janson et al., 2014), we expected a presence of multiple genetic populations of D. magna in the investigated water systems, with little correlation over geographical location. In each water body, we also expected meta-population structure characterized by different genetic clusters with specific behavioral signatures associated with local adaptation (e.g., nutritional requirements (Maruki et al., 2019)). Moreover, we hypothesized that individuals belonging to the same population genetic cluster would exhibit similar swimming behavior and life traits, inheritable across generations.

| Organismal collection and cultivation
Daphnia magna (Figure 1) were collected in Lake Bysjön (55.67424 N-13.546529 E) and in a Sydvatten pond (55.660709 N -13.541140 E), located less than 3 km apart from each other, in southern Sweden ( Figure 2). Lake Bysjön is a hypertrophic seepage lake alimented by groundwater inflow (Enell, 1982;Persson & Svensson, 2004), whereas the human-made Sydvatten pond is part of a water network, involved in the production of drinking water for consumption. It receives filtered water (mesh 40 µm) from the nearby Lake Vombsjön (Sydvatten communication). Environmental parameters and information on the presence of fish are given in Table 1. Organisms were sampled using a 200µm meshed plankton net and individually isolated under a stereomicroscope (SZX7, Olympus).
Gender identification and taxonomy were determined using Infinity Analyse and Capture software (version 6.5.4) and reference taxonomic keys (Chiang & Du, 1979;Lilljeborg, 1901). A total of 35 individuals of D. magna were collected (Appendix S1) and grown at 18°C under a 14:10-hr light:dark cycle of 50 µmol photon m −2 s −1 in a 100-ml glass bottle (Heraco AB 6,212, Holmsund, Sweden) filled with 60 ml copper-free water and were fed every 2-3 days with 10 5 cells of pure culture of Scenedesmus sp. At each reproduction cycle (clutch), the mother was transferred to a new culture bottle and the offspring were grown for a week prior to being transferred to separate culture bottles, thereby constituting the next generation. Only females were kept for the analyses.

| Swimming behavior
In order to account for maternal effects, clonal lineages were grown for three consecutive parthenogenic generations prior to behavioral analyses (Lynch, 1985). Individual swimming behavior was monitored in a Plexiglas aquarium (0.2 × 0.2 × 0.85 m) filled with 30 L of copper-free water resulting in a water column of 0.75 m (Palmér et al., 2016) illuminated from above by LED-470 nm F I G U R E 1 Female Daphnia magna observed for sex and taxon determination using a stereomicroscope at 50 µmol photons m -2 s -1 to mimic nighttime and with addition of 150 mA UVR (UVA-340, Q-PANEL Co, USA) to mimic solar radiation during daytime (Ekvall et al., 2013). Individuals were labeled using fluorescent Quantum dots (585 ITK Carboxyl Quantum dot, Life Technologies, Prod. No. Q21311MP) and then transferred to the monitoring aquarium, where their swimming behavior was recorded under three consecutive illumination phases, including one minute of LED illumination (Acclimation phase), followed by one minute of UVR (Stimulation phase) and by a one minute of LED illumination (Recovery phase) . From these video recordings, we extracted the 3D position of each individual at six frame per second in MATLAB V1.7 using the method described in Palmér et al. (2016). We then calculated the swimming speed and the refuge demand (mm × s), that is, cumulative vertical position through the three-illumination phases . We calculated the individual mean values of swimming speed for each UVR phase for later statistical analyses. Dataset includes results from swimming behavior and refuge demand across generations for F I G U R E 2 Sampling location at Lake Bysjön (harboring mainly POP1 individuals) and Sydvatten Pond (harboring mainly POP2 individuals), Scania region, southern Sweden Lake Bysjön and at the third generation in the Pond (Appendix S1).
Statistical comparisons between genetic populations and lineages were performed in R v4.0.2 (R Core Team, 2020) using analysis of variance (ANOVA) on normal and homogeneous variance of residuals, and nonparametric tests for unbalanced design: the Kruskal-Wallis rank sum test (one-way analysis of variance), the Student two-samples t test for equal variances, and Welch test for unequal variances.

| Extraction of genetic material
After tracking, organisms were individually frozen at −20°C. Their genomic DNA was extracted following a modified protocol from Richlen and Barber (2005). Each individual was placed in 100 µl solution of sterile Chelex 10% (Chelex 100 Bio-Rad, diluted in double distilled water and autoclaved 121°C 30 min). Samples were vortexed for 5 s, centrifuged for 20 s (3,300 g), incubated at 95°C for 20 min, vortexed for 5 s, and spun down. The supernatant was collected for genetic analyses. Extracted DNA was quantified using dsDNA HS Qubit Assay Kit (Q32854, Life Technologies Corp., Eugene, Oregon, USA).

| Libraries and database preparation
Genetic amplifications using polymerase chain reaction (PCR) were performed in 12.5 µl final volume using a modified protocol from

| Population genetic analyses
Screening for scoring errors (large dropout, null alleles, stuttering) and allele distribution were made using Micro-Checker software version 2.2.3 (Van Oosterhout et al., 2003) with 95% confidence intervals, a Bonferroni correction (Dunn-Sidak), and 1,000 permutations (Appendix S3). We successively used the set of 13 neutral markers ( Table 2) to investigate population genetic structure and connectivity processes and the two adaptive markers (Table 2), under influence of fish predation stress (Locus B097) and general landscape stressors (fish predation, parasite infection, and land use stressors; Locus B087), to identify potential correlations with organismal swimming behavior in inferred populations.
Canonical factorial correspondence analysis was performed using the Genetix software, version 4.05.2 (Belkhir et al., 1994). The number of genotypes was inferred using the Gimlet software version 1.3.3 (Valière, 2002), the function "regroup genotypes," and the option "missing alleles as any other alleles." The number of populations and their genetic structure were investigated using the Structure version 2.3.4 (Pritchard et al., 2000) on strictly neutral markers with a burn-in period of 20,000 iterations, a MCMC after burn-in of 20,000 iterations, admixture model, a number of 20 replicates, and no prior information on location or ancestry. The number of populations was inferred using the Evanno method (Evanno et al., 2005).
Genetic diversity within and between populations was investigated using both Genetix and GenAlEx software version 6.51b2 (Smouse et al., 2017). Allele richness, heterozygosity, F-statistics, the number of private alleles, and fixation index across loci were calculated for codominant markers. The inbreeding coefficient was calculated within populations (F ST ) and individuals (F IS ), for small population size (G -statistics ), and with a stepwise mutation model approach (R ST ), with 1,000 permutations. The number of effective migrants (N m ) was estimated between populations using codominant data and Equation 1 (Wright, 1969). The Hardy-Weinberg equilibrium was tested in population with sample size ≥50 and expected numbers per classes ≥5. The degree of allelic association between locus was assessed by measuring linkage disequilibrium (1,000 bootstrap). (1) Environmental parameters collected at each sampling locations. The location name (Location), number of strains of Daphnia magna collected (N s ), the pH, conductivity, temperature, depth at the location, period of sampling (Period), and presence of possible fish predators   (Cornuet & Luikart, 1997) and the neutral markers. The two-phase model (TPM) for multiple-step mutations was applied to the data with default settings recommended for microsatellites analyses (i.e., a variance of geometric distribution of 30 and a proportion of stepwise mutations of 70%) and 1,000 permutations. A two-tailed Wilcoxon sign rank test (Luikart et al., 1998) was applied to each population to detect heterozygosity excess and recent bottlenecks.

| RE SULTS
In the two water systems, 35 strains (Lake Bysjön: 11 strains; Pond: 24 strains), represented by 137 individuals, were genetically characterized over 15 polymorphic microsatellite loci, including 13 neutral and two adaptive markers ( Table 2). Each locus was represented by a set of 2 to 4 alleles (Appendix S3).
The origin from Lake Bysjön (90%) belonged to POP1 and from the Pond (97%) to POP2 (Appendix S1). The remaining four individuals shared less than 80% of genetic similarity to either of the two genetic clusters (66%-73%; Appendix S1) and were classified as population Hybrids. This clear separation of individuals into population according to sample location indicates a strong influence of geographical barriers, despite the close distance between the locations (Figure 2

| 2879
TESSON aNd SHa (Holderegger et al., 2006). A total of 46 unique genotypes were identified. The majority of these genotypes belonged to one of the inferred genetic clusters, that is, 30% to POP1, 59% to POP2, and 4% to the Hybrid group (Appendix S1). Only three genotypes were shared between genetic clusters, that is, gr.9, gr.11, and gr.46, possibly due to the presence of null alleles (Appendix S5). Moreover, each genotype was strain-specific (Appendix S1, Appendix S5), with the exception of gr.2 (B36-P07).
Using prior information on generation, we investigated the population genetic inheritance and mutation pattern in 24 clonal lineages, Adaptive markers are utilized to infer adaptative or evolutionary potential of populations (Holderegger et al., 2006). Linkage disequilibrium between the two adaptive markers in the dataset was only marginal (p = .07; Appendix S4). However, they showed significant linkage disequilibrium with five neutral loci (p < .01; A002, B008, F I G U R E 5 Swimming speed (mm/s) in individuals of Daphnia magna within genetic clusters (POP1 (light gray) and POP2 (dark gray)) over the three-phase experiment in the presence (UV) and absence (before, after) of UV radiations. **indicates a significant difference in swimming speed (p < 0.01) between populations under threat, whereas ns stands for nonsignificant difference

| D ISCUSS I ON
High population diversity with consequent admixture events was expected because of the vicinity of the two water systems (less than 3 km apart) and of the intensive commercial and recreational use of aquatic environments in the region of Scania, southern Sweden.
Such patterns have previously been confirmed in, for example, fish (crucian carp; Carassius carassius) (Janson et al., 2014). However, with respect to Daphnia magna our results show two clearly separated genetic populations almost perfectly matching with geographical locations (94% of individuals).
Despite being less than 3 km apart, the two water systems are isolated from each other by landscape barriers (Figure 2). Using seed-banks of D. magna, a population genetic analysis demonstrated the existence of discrete genetic populations in isolated ponds, at different geographical scales (Orsini et al., 2013). Similarly, the absence of surface water connectivity between the two investigated water systems would explain why Lake Bysjön is mainly represented by POP1 individuals and the Sydvatten pond by POP2 individuals.
Isolation by distance promotes genetic differentiation between populations despite geographical barriers (Fields et al., 2015).
Therefore, knowing the high dispersal rate of propagules, one would expect a higher gene flow in closely connected (Fields et al., 2015) or interconnected ponds (Orsini et al., 2013). At local scale, genetic distance analysis suggests that the two isolated but neighboring water systems were connected by a non-negligible number of migrants and a moderate gene flow. It implies that Lake Bysjön and the Sydvatten pond are, with respect to Daphnia dispersal, in contact despite the absence of water connectivity.
Potential drivers for alternative ways of dispersal include transportation by wind, animals, and humans (Incagnone et al., 2014;Tesson et al., 2016) between different aquatic systems. Passive dispersal of zooplankters by natural vectors such as wind has been reported in previous studies (Havel & Shurin, 2004). The resistance of egg propagules (ephippia) to desiccation and freezing (Hebert, 1981) confers Daphnia with the ability to travel short distances (<5 m) up to several kilometers and to rapidly and intensively colonize isolated pools (Louette & De Meester, 2004Sirianni, 2016).
Ephippia can also be transported passively by animals such as aquatic insects (van de Meutter et al., 2008), fish (Mellors, 1975), and birds (Figuerola et al., 2003). A phylogenetic analysis showed the importance of bird-mediated dispersal and climate change in Arctic Daphnia species (Alfsnes et al., 2016;Haileselasie et al., 2016). The influence of such dispersal events on the genetic structure of zooplankton is rarely quantified, but at broader scale, our data suggest that although the effect is moderate since the two populations are still highly distinct, it may in the long run lead to both mixing of populations and hybridization.
Human can indirectly contribute to the dispersal of ephippia-mediated fish stocking and the use of construction materials (Havel & Shurin, 2004). Resting stages of Daphnia can resist passage through the digestive apparatus of fish (Mellors, 1975) and be dispersed by fish stocking (Havel & Shurin, 2004). This mechanism could have impacted Lake Bysjön where past events of species introduction have been recorded (Eklöv & Eklövs Fiske & Fiskevård, 2003;Persson & Svensson, 2004). However, the absence of fish in the investigated pond (Sydvatten communication) invalidates the hypothesis of fish-mediated transmission between the two locations. Alternatively, the use of working materials (e.g., engines, boots, and sands) can contribute to the dispersal of ephippia between water systems (Duffy et al., 2000;Hairston et al., 1995;Perrigo et al., 2012). However, under such dispersal we would have expected a higher admixed picture than the one observed in the present study. Ricklefs (1987) proposes that regional processes (speciation and dispersal) have a greater effect on unsaturated communities than local processes (competition, predation, adaptation). Local sorting can strongly affect the spatial distribution of zooplankters (Hessen et al., 2019) governed by widespread endemism, provincialism and allopatric speciation (Hebert & Wilson, 1994), and founder effects (De Meester et al., 2002) within restricted areas. Newly arrived eggs are affected by drift during colonization, local inbreeding, and local genetic selection (Haag et al., 2006). Moreover, each system possesses seed-banks of eggs that play a "rescue effect" (Gotelli, 1991) against local extinction and leads to competition with new incomers.
Altogether, local sorting often leads to a reduce level of gene flow between populations. The moderate gene flow retrieved in the present study would suggest that local sorting may have an impact on the two neighboring populations.
Among local processes, we investigated the behavioral response of D. magna individuals among populations to solar radiation (UVR threat). Solar radiation is a strong selection pressure that constrains diel vertical migration in Daphnia species. Their natural avoidance of UV radiation is a protective mechanism against the acquisition of photodamages (Williamson et al., 2001). Their behavioral response consists in the increase of their swimming speed and distancing from the UVR source (Ekvall et al., 2015). Recent studies have shown that the swimming speed and behavior under UVR threat are inherited in D. magna across generations (Sha et al., 2020, this study).
Moreover, the present study highlights a significant difference in swimming speed between populations under threat. The reason for the difference in swimming speed is beyond the scope of this paper.
But, the high variance ( Figure 5) and consistent individual response (Appendix S7) observed in swimming speed within each population indicate that different clones may respond differently to UVR threat.
Under the selection pressure inflicted by UVR on the swimming behavior of D. magna, populations follow the neutral phenotypic theory ( (Fisher, 1958) in (Spitze, 1993)  . We analyzed the adaptive genotypic diversity in D. magna using two markers associated with fish predation and general landscape stressors (i.e., fish predation, parasite infection, and land use) (Orsini et al., 2012). Our results showed that the adaptive genotypes were inherited within strain. However, deviations from this pattern included mutations principally located on the adaptive marker associated with fish predation. Moreover, high genotypic diversity over the fish predation marker occurred in particular in POP2 and over the general marker in POP1. Thus, the results indicate that genetic populations undergo different selection pressures in the two aquatic systems. The adaptive or evolutionary potential of populations is beyond the scope of this paper. But the significant difference in refuge demand in certain genotypes of Daphnia (Ba, Bd), principally associated with POP2 and the Pond, indicates strong selection for fish predation in POP1 and Lake Bysjön.
Results are consistent with inferred swimming behavior and with the environmental history in Lake Bysjön (recreational location for fishing) and the hand-made pond (exempt of fish but affected by anthropogenic activities). Sha et al. (2020) further demonstrated that D. magna adaption to different threats (predation, UVR) occurred within three generations. Lineage analysis showed that 75% of the neutral multilocus genotypes and 71% of the adaptive genotypes were inherited within strains down to four generations. Deviation from expectation was associated with the occurrence of mutations and possibly with null alleles. Null alleles are the result of consistent amplification failure in PCR over a particular locus and can introduce substantial assessment errors (Dakin & Avise, 2004). In strains B41 and B42 for instance, mothers had the same neutral genotypes than their descendants but were assigned to Hybrid instead of POP1 (Appendix S5). We maintained these individuals as Hybrids but discarded them from the population genetic analyses, thus not affecting the connectivity estimates. We also maintained the two individuals in strain P31 for which a transition POP1-POP2 was estimated, as the removal of the strain did not affect the genetic estimates. Globally, the observed mutations fall under microsatellite mutation pattern (i.e., frameshift, insertion/deletion, and principally step mutations) (Hancock, 1999;Selkoe & Toonen, 2006) with a net gain of heterozygosity as expected in asexual lines (Seyfert et al., 2008). Elevated microsatellite mutation rate would explain the punctual mutation events occurring in 13 out of 37 investigated strains. However, no clear mutation pattern was identified over the six inferred mutated lineages. Unexpectedly were the incongruent multilocus mutation patterns observed in both types of markers, at the origin of genotype and population cluster transitions (Appendix S5). Results call for genome-wise complement analyses and longer time series to further understand the effect of different environmental pressures (e.g., predation, UVR, and predation and UVR) on the swimming behavior of Daphnia and other relative quantitative traits (e.g., fitness, organismal size), using an exhaustive set of adaptive markers (Orsini et al., 2012) coupled with genome-wise (Bubac et al., 2020) and common garden experimentations.
In conclusions, our study unveils the existence of neighboring discrete populations connected by a moderate gene flow independent of water connectivity, highlighting the importance of wind-and animal-mediated dispersal. The populations were under different selection pressures that reflects in the genotypic fingerprint of the few adaptive markers investigated. Moreover, these populations harbored diverse inheritable individual personalities associated with swimming behavior. Furthermore, our results suggest that despite strong local sorting, regional processes have an impact on isolated favorable habitats at local scales.