Urbanization drives genetic differentiation in physiology and structures the evolution of pace-of-life syndromes in the water flea Daphnia magna

Abstract

Natural and human-induced stressors elicit changes in energy metabolism and stress physiology in populations of a wide array of species. Cities are stressful environments that may lead to differential selection on stress-coping mechanisms. Given that city ponds are exposed to the urban heat island effect and receive polluted run-off, organisms inhabiting these ecosystems might show genetic differentiation for physiological traits enabling them to better cope with higher overall stress levels. A common garden study with 62 Daphnia magna genotypes from replicated urban and rural populations revealed that urban Daphnia have significantly higher concentrations of total body fat, proteins and sugars. Baseline activity levels of the antioxidant defence enzymes superoxide dismutase (SOD) and glutathione-S-transferase (GST) were higher in rural compared with city populations, yet urban animals were equally well protected against lipid peroxidation. Our results add to the recent evidence of urbanization-driven changes in stress physiology and energy metabolism in terrestrial organisms. Combining our results with data on urban life history evolution in Daphnia revealed that urban genotypes show a structured pace-of-life syndrome involving both life-history and physiological traits, whereas this is absent in rural populations.

1. Introduction

Cities play host to one of the most remarkable shifts in species occurrence and loss, community structure, biotic homogenization, and rapid adaptive evolution [1–6]. Multiple stressors, such as pollution, drought, heat and changed biotic interactions, act in concert in cities [7], generating an environment that is thought to select for those species and phenotypes able to cope with higher overall stress levels (e.g. [8–12]).

Maintaining homeostasis depends heavily on the balance between oxidative damage and antioxidant repair mechanisms [13,14]. Oxidative damage to lipids, proteins and DNA can be caused by pro-oxidative by-products of respiration (reactive oxygen and nitrogen species, ROS and RNS [14]). This leads to oxidative stress, a state in which cellular antioxidant and repair mechanisms fail to quench excess levels of ROS/RNS [13,14]. Natural stressors, such as predation, overcrowding (competition), food stress (low food quantity or quality), oxygen depletion and temperature extremes, are known to cause enzymatic and oxidative trait changes linked to stress resistance [14–18]. In cities, these environmental stress factors are disproportionally altered due to human activity [7]. Additionally, non-natural perturbations linked to pollution (e.g. metals, pesticides, noise, and light pollution) have been shown to change redox states in terrestrial and aquatic animals due their high oxidative reactivity with biomolecules [19–22]. Therefore, it might be expected that city populations feature adaptive changes in antioxidant defence mechanisms [10,23,24].

Studies on stress physiology of urban species and populations have only started to emerge, but evidence is rapidly growing that urbanization has a significant impact on the regulation of oxidative stress machinery (e.g. antioxidant enzymes such as superoxide dismutase—SOD, catalase—CAT, and glutathione-S-transferase—GST), immune function, stress-related behavioural changes, and energy metabolism [8,10,12,23,25–28]. For example, field-caught urban house sparrows have higher activity of antioxidant enzymes (SOD and GST), yet lower total antioxidant capacity (TAC) and higher levels of oxidative damage in their blood compared to rural birds [28]. Additionally, energy reserve compounds such as fat, sugars, and proteins can be an important component of stress responses, as they provide cellular energy that can be allocated to growth and reproduction, active dispersal, or an increase in physiological stress responses, and thereby provide the fuel to cope with stressful conditions and enable population survival [11,29]. For instance, field-caught juvenile house sparrows contain higher levels of body fat in urban than in rural areas [11].

Small aquatic ecosystems such as ponds located in the city are warmer [3,30] and often receive organic and chemical pollutants such as road salts, nutrients, metals and pesticides via surface run-off or atmospheric depositions [31,32]. In addition, they are often characterized by higher disturbance levels (e.g. water level control, leaf litter cleaning [31]; K.I.B. 2013–2015, personal observations). As populations of plants and animals inhabiting urban ponds might be exposed to greater environmental stress, we hypothesize that urban environments select for different coping mechanisms at the physiological level, causing genetic differentiation between rural and urban populations in traits such as energy storage molecules (e.g. fat, proteins and sugars), and oxidative stress enzymes (CAT, SOD, GST). Although some studies observed clear phenotypic upregulation of stress physiological parameters (e.g. GST [33]; Hsp70 [34]), evidence for an evolutionary shift in organismal stress physiology in cities is so far limited (but see [8,14]).

To quantify the degree of urbanization-driven genetic differentiation in stress physiology, we set out a common garden experiment with urban and rural genotypes of Daphnia magna, originating from replicated urban and rural populations (13 populations located across well-characterized urbanization gradients in Flanders, northern Belgium). Daphnia from city populations mature faster and at a smaller size, produce more offspring and release their juveniles sooner, and are characterized by a higher intrinsic population growth rate [35] (electronic supplementary material A, figure S3). In our common garden experiment, we measured seven physiological response variables that are either linked to energy metabolism (total fat, sugar and protein content) or oxidative stress responses, quantifying both antioxidative defence (activity of SOD, CAT and GST) and damage (concentration of malondialdehyde, a by-product of lipid peroxidation). We reared animals at two different temperatures (20°C and 24°C) to mimic the urban heat island reported to occur in these ecosystems (urban ponds are characterized by a higher mean, maximum, and minimum daily temperature compared to rural ponds, and this ‘urban hot-tub effect’ is most pronounced in summer; ΔT_mean-summer = 3.04°C; ΔT_max-summer = 3.69°C [3,30]; electronic supplementary material A, figure S1). This enabled us to quantify phenotypic plasticity across two temperatures and assess habitat-specific (urban versus rural) phenotypic plasticity (to warming) in the measured stress physiological responses. We first tested the hypothesis that there is genetic differentiation between urban and rural Daphnia populations for these physiological responses. More specifically, we expected that urban Daphnia might be characterized by higher amounts of storage molecules that can be invested in faster growth and development, or in detoxification processes. Stress physiological responses might include higher activity of antioxidant defence enzymes (SOD, CAT, GST), while suffering a similar or even reduced level of oxidative damage. We additionally tested for covariation patterns to explore to what extent the studied life-history and physiological traits are shaped by an overall pace-of-life syndrome [36]. The pace-of-life syndrome (POLS) hypothesis [36] predicts a concerted change in both life-history and physiological traits, as the physiological traits determine how much energy can be invested in life-history features [37]. The presence and structure of covariation patterns has been suggested to be environment-dependent and potentially under selection (e.g. [38,39]). Using multi-group structural equation modelling (SEM) on urban and rural Daphnia, we tested whether pace-of-life-syndrome covariation patterns are dependent on urbanization background (evolution) or rearing temperature (plasticity) [40–42]. More specifically, we hypothesize that especially in urban populations, a close link between life history and physiology might have evolved, where physiological processes (e.g. energy budget) are aligned to support their faster lifestyle.

2. Material and methods

(a). Urbanization gradient, model organism and experimental design

We sampled 13 Daphnia magna populations along multiple well-characterized urbanization gradients in Flanders (northern Belgium; see electronic supplementary material A, figure S2), with the percentage built-up area (BA; available via the Large-scale Reference Database at Flanders Geographical Information Agency) in the regional surroundings of the pond (3200 m radius) as a proxy for the degree of urbanization. Given that this measure does not include roads and parking lots, an area with a BA of 10% or more is considered highly urbanized. We classified all ponds according to their %BA into rural (less than 5%; n = 5) and urban (greater than 10%; n = 7). Rural ponds were additionally characterized by a minimum of 20% of biologically valuable area (3200 m radius) to ensure that they would be situated in natural areas and not in areas characterized by intensive agriculture. We performed all experiments with clones of these 13 populations, but due to accidental thawing we lost the set of samples of highly urbanized population Brxl, which prevented assessing physiological traits for this population.

Daphnia magna has a cyclic parthenogenetic life cycle; under favourable conditions all-female clonal lineages can be established and cultured over several generations. Six females from each population were isolated and reared under standardized laboratory conditions (dechlorinated tap water, 80% water replacement every 2 days, 20°C, 14 : 10 h L : D photoperiod, daily fed 10⁵ cells ml⁻¹ of the green algae Acutodesmus obliquus; for more detailed information on culturing conditions we refer to [3,43] and electronic supplementary material B). After one generation of culturing under standardized conditions, 2nd brood juveniles from each line were split up and divided over six replicated cultures (i.e. 36 cultures per population, total number of experimental units: 432), later to be assigned to the appropriate temperature treatment. We screened clonal identity of all lines using microsatellite markers (27 markers, detailed information in [3]), which was added as an additional level of within-population variation in the statistical analysis (i.e. ‘clone’, electronic supplementary material A, table S1). All population × line × temperature replicates were cultured for an additional two generations under the same standardized laboratory conditions to purge (grand)maternal effects, after which we transferred a cohort of 12 juveniles (less than 24 h old) of each line to the appropriate temperature treatment (20°C or 24°C).

We carried out a cohort life table experiment with all lineages at two temperatures, 20°C and 24°C (constant temperatures), reflecting the average summer temperature in rural and urban populations, respectively [30] (see electronic supplementary material A, figure S1). Cohorts were followed up to assess thermal tolerance and measure life-history traits (reported on in [43] and [35]). After release of the 2nd clutch, six animals from each experimental cohort were pooled and flash frozen in liquid nitrogen until sample processing. Body length (from centre of eye to base of tail spine) of three animals of each pooled sample was measured (Olympus X stereo microscope), and average length was translated into dry mass using a published length-weight regression [44] so as to standardize measured amounts of proteins, fat, carbohydrates and MDA by dry weight.

(b). Quantification of response variables

The pooled sample of animals per replicate of a clonal line-by-treatment combination was homogenized with a pestle in PBS buffer (Phosphate-Buffered Saline, 50 mM, pH: 7.4, 50 µl animal⁻¹) and centrifuged for 8 min at 13 000 r.p.m. and 4°C. Detailed sample analysis protocols are given in electronic supplementary material B. In brief, total protein content was quantified following the Bradford protocol [45], total fat content following Bligh & Dyer [46], and total sugar content (glucose + glycogen) following Stoks et al. [47].

The activity of superoxide dismutase (SOD), catalase (CAT), two key antioxidant enzymes in invertebrates [13], and glutathione-S-transferase (GST), a secondary antioxidative enzyme that protects against oxidizing and toxic substances by detoxifying ROS-damaged cellular components, was measured as an estimate of antioxidant defence. SOD and CAT activity was quantified using the protocol of De Block & Stoks [48], while GST activity was quantified following McLoughlin et al. [49]. Oxidative damage was quantified as malondialdehyde (MDA) concentration [50], a by-product of lipid peroxidation, following [51] (total fat content was used as a covariate in the statistical analysis).

(c). Statistical analysis

To test for differentiation in physiological traits between urban and rural Daphnia (categorical), and between animals reared at 20°C and 24°C (categorical), we first performed a MANOVA, followed by separate linear mixed-effect models (using R software [52]; packages car, lme4, mvnormtest), on all response variables scaled to a mean of 0 and standard deviation of 1 (fixed effects: urbanization, temperature, and their interaction). Population and clone within populations were implemented as random factors (clone nested in population, population nested in urbanization). Model specifications, random structure, computed test statistics and the validation of model assumptions are detailed in electronic supplementary material B. Significance of ‘clone’ reflects evolutionary potential within populations.

We tested for the impact of urbanization and/or temperature on covariation between life-history and physiological traits along the pace-of-life axis (cf. POLS hypothesis [36]) by combining the stress physiological traits with data on eight life-history traits measured on the same cohorts of individuals in a life table experiment [35]. Daphnia genotypes from urban populations show an evolutionary shift in lifestyle towards the fast end of the pace-of-life continuum compared to Daphnia genotypes from rural populations (see electronic supplementary material A, figure S3). Using structural equation modelling [53], we explored to what extent the shift towards a faster lifestyle in urban populations covaried with trait shifts in stress physiology according to an overarching pace-of-life syndrome [36], and whether urbanization (genetic differentiation) and/or temperature treatment (plasticity) shaped these POLS trait covariation patterns in a consistent manner [40,54].

Using multi-group SEM [55], we first evaluated the consistency of trait covariation patterns at both levels of biological organization separately (i.e. life-history trait covariation patterns and physiology trait covariation patterns, independent of each other; paths a–i and j–p, respectively, in electronic supplementary material A, figure S4). Second, we evaluated consistency of trait covariation in an overarching pace-of-life syndrome model integrating life history and physiology (paths q and r, electronic supplementary material A, figure S4). The rationale behind each SEM is that a ‘latent’ (i.e. non-measured) variable can be incorporated in the model (electronic supplementary material A, figure S4, ovals) that structures the covariation patterns observed in the measured variables (squared boxes). The suits of traits linked in a ‘syndrome’ given by the latent variable are indicated via significant pathways [53]. Each pathway in the SEM represents a linear model. For significant paths, a change in one variable (standardized units for all variables) coincides with a change in all other traits of the syndrome (according to path coefficients, which represent standardized model estimates). Multi-group SEM allows to compare models with factor loadings that are constrained or freely varying across temperature treatment, urbanization level, or a combination of both (detailed information in electronic supplementary material B). This results in different model hypotheses (models I to IV, electronic supplementary material, table S2). For example, if the best model fit is characterized by factor loadings being constrained across temperature treatments, but not urbanization level (model II), this indicates that trait covariation patterns significantly differed between urban and rural populations but were the same for animals reared at 20°C and 24°C (and vice versa for model III). This would imply that urbanization background, and thus evolution (or temperature and thus plasticity, for model III), significantly shaped trait covariation patterns. Models were ranked according to AIC scores, model weights (W), and evidence ratios (E.R.) (detailed in electronic supplementary material B). When models II, III or IV were selected as the best model (i.e. a non-fully restricted model), differences between groups were tested by comparing factor loadings (significance, direction and size) using the method of Zar [56] (see also [40,51,54] and in electronic supplementary material B for calculations). All models were fitted with the package lavaan [57] using the R software [52]. All variables were scaled and standardized prior to analysis.

3. Results

(a). Genetic differentiation in energy reserves and stress physiology

MANOVA indicated an overall significant effect of urbanization on the combined set of physiological response variables (p < 0.001; table 1). General linear-mixed effect models on each variable separately revealed that urban animals contained significantly more fat (p = 0.003; figure 1a), sugars (p < 0.001; figure 1b) and protein (p = 0.002; figure 1c) per µg dry mass (table 1), suggesting a higher amount of energy reserves in urban compared to rural populations. Animals reared at a higher temperature had less protein per µg dry mass (p = 0.02; table 1) but did not show differences for fat and sugar content (table 1).

Table 1.

Multivariate and univariate statistical results. Results of the MANOVA and univariate linear mixed-effect models, testing for the effect of urbanization level, treatment and their interaction on a set of seven physiological parameters linked to energy metabolism (total protein, fat and sugar content) and oxidative stress (antioxidant enzymes superoxide dismutase (SOD), glutathione-S-transferase (GST) and catalase (CAT); oxidative damage: lipid peroxidation (MDA)), measured in a set of urban and rural Daphnia magna genotypes grown under common garden conditions. Total fat content was used as a covariate in the MDA model. Significant results are given in italics (p < 0.05) or indicated with ‘·’ (p < 0.1). Numerator and denominator degrees of freedom are given by ndf and ddf respectively.

	urbanization				temperature				urbanization × temperature				clone (random)
MANOVA	F	ndf	pillai	p	F	ndf	pillai	p	F	ndf	pillai	p
all traits	7.820	7	0.543	<0.001	1.230	7	0.158	0.306	1.022	7	0.134	0.429	included as random error term
univariate ANOVAs	F	ndf	ddf	p	F	ndf	ddf	p	F	ndf	ddf	p	χ2	p
total protein content	16.265	1	10.090	0.002	5.486	1	353.07	0.020	0.174	1	353.06	0.677	2.648	0.050·
total fat content	14.671	1	9.68	0.003	2.985	1	352.40	0.085·	0.038	1	352.40	0.846	8.102	0.002
total sugar content	23.0736	1	9.980	<0.001	1.166	1	351.18	0.281	0.068	1	351.28	0.794	9.5027	0.001
MDA (lipid peroxidation)	0.045	1	10.68	0.835	0.579	1	355.25	0.447	0.238	1	354.37	0.626	0.106	0.372
SOD	8.808	1	9.85	0.014	12.779	1	358.84	<0.001	0.001	1	258.84	0.973	2.598	0.054·
CAT	0.435	1	10.01	0.525	29.439	1	356.850	<0.001	0.077	1	357.080	0.781	3.223	0.037
GST	7.282	1	9.79	0.023	0.141	1	357.89	0.707	1.419	1	358.01	0.243	5.447	0.010

Figure 1.

Energy storage compounds. Mean ± 1 s.e. (a) fat, (b) sugar and (c) protein content for urban (triangles) and rural (circles) animals reared at 20°C and 24°C. Significant differences between urban and rural animals are given by coloured symbols (urban: red; rural: blue).

Urban animals did not differ from rural ones in CAT activity (p = 0.525; figure 2a), but had a lower SOD (p = 0.013; table 1; figure 2b) and GST (p = 0.023; table 1, figure 2c) base line activity. MDA concentrations (lipid peroxidation) did not differ between urban and rural animals (p = 0.835; figure 2d). Elevated rearing temperature induced a significant increase in the activity of CAT (p < 0.001; figure 2a) and SOD (p < 0.001; figure 2b), but not in GST, and there were no differences in lipid peroxidation. We did not observe significant urbanization × temperature interaction effects for any of the traits.

Figure 2.

(a–c) Antioxidant enzymes (white) and (d) oxidative damage (grey). Mean ± 1 s.e. (a) CAT activity, (b) SOD activity, (c) GST activity and (d) lipid peroxidation, measured as MDA for urban (triangles) and rural (circles) animals reared at 20°C and 24°C. Significant differences between urban and rural animals are given by coloured symbols (urban: red; rural: blue). Non-significant differences for the effect of urbanization are depicted in black/white.

(b). Structural equation modelling: exploring the pace-of-life syndrome hypothesis

(i). Life-history trait covariation patterns

The path analysis on life-history traits revealed a syndrome structure with underlying life-history trait covariation differing according to model IV (electronic supplementary material A, table S3, factor loadings all free; electronic supplementary material, figure S5), indicating differently structured life-history trait covariation patterns across all four urbanization × temperature treatment groups. This implies that both urbanization background (evolution) and temperature treatment (plasticity) significantly shape life-history trait covariation (more detailed results are given in electronic supplementary material C).

(ii). Physiology trait covariation patterns

The best model explaining a structuring syndrome at the level of physiological traits was model I (electronic supplementary material, table S3, all factor loadings constrained to be equal: electronic supplementary material, figure S6), indicating a consistent trait covariation across all four urbanization × temperature groups (more detailed results are given in electronic supplementary material C).

(iii). Overarching pace-of-life syndrome

In the overarching POLS models, model IV was the best supported, indicating that covariation patterns differ among all groups (electronic supplementary material A, table S3). Both urbanization background (genetic differentiation) and temperature treatment (plasticity) shape trait covariation patterns. The difference in trait covariation patterns between urban and rural animals reared at 20°C reflects genetic differentiation for a POLS (figure 3a,b; significantly differing paths at the level of p < 0.05 are indicated with †; for the pattern at 24°C, see electronic supplementary material A, figure S7). The major difference between urban and rural animals at 20°C was that the overarching syndrome structure linking the physiological and life-history syndromes was significant in the urban but not rural animals (figure 3a,b; paths q and r in electronic supplementary material A, figure S4; both p < 0.001). This overarching POLS reflects that, within the set of urban animals, fast-paced (i.e. fast maturation, high fecundity, high population growth rate) animals have lower amounts of protein, fat, and sugars, but higher activity of oxidative stress enzymes and higher oxidative damage. This suggests that, within urban animals, shifts towards a faster lifestyle depletes energy resources and increases oxidative stress physiology. In the rural path model, no such link between physiology and life history, and thus no overarching POLS is present (figure 3b). Within the nested life-history syndrome, fecundity (number of juveniles in both clutch 1 and 2) significantly contributed to the syndrome structure in rural but not in urban animals (both p < 0.001; figure 3a,b). When looking at the nested physiological syndrome, MDA significantly covaried with other physiological traits in urban but not in rural animals (p < 0.001; figure 3a,b), whereas GST covaried with other physiological variables in rural but not in urban animals (p = 0.049; figure 3a,b). Of the significant paths shared between both covariation models (i.e. paths significantly contributing to the syndrome structure in both urban and rural animals), all path coefficients had the same direction in both groups, but significantly differed in size between urban and rural genotype sets for development time (age at maturity, age at release of first and second clutch), fecundity (both clutches), somatic growth, total protein content, MDA, and SOD (figure 3a,b; indicated with †, p < 0.05).

Figure 3.

Structural equation models. Path models showing the covariation pattern of life-history and physiological traits (i.e. measured traits, boxes) within a life history, physiology and overarching pace-of-life syndrome (POLS) (all modelled as latent variable, ovals) in (a) urban and (b) rural genotype sets of Daphnia magna when reared at 20°C. Differences in covariation structures between the two groups reflect genetic differentiation. Only paths significantly contributing to syndrome structure (i.e. significantly different from 0) are depicted (associated traits are in grey coloured boxes), and accompanied with path coefficients. Significance levels of path coefficients are depicted as ***p < 0.001, **p < 0.01, *p < 0.05. Path coefficients indicate the change (in SD units) of traits predicted to occur based on a 1 SD change in the underlying syndrome structure. Arrow thickness (dashed arrows: negative correlations, full arrows: positive correlations) is proportional to the strength of the path loading. Path loadings significantly differing between urban and rural genotype sets (either because of their significant contribution to a syndrome in one group, but not in the other, or because of a significant difference in direction or size of path loadings that significantly contribute to syndrome structures in both groups) are indicated with † (p < 0.05 after Bonferroni correction). Dashed double-headed arrows indicate correlated errors.

4. Discussion

The results of our common garden experiment reveal pronounced genetic differences between urban and rural populations of the water flea Daphnia magna for a set of physiological traits linked to stress resistance and energy reserves. Compared with rural populations, urban Daphnia have higher concentrations of protein, fat and sugars, reflecting a higher level of energy storage compounds available per unit dry mass. The activities of antioxidant enzymes superoxide dismutase (SOD) and glutathione-S-transferase (GST) were upregulated in rural animals, yet malondialdehyde (MDA) concentrations, a measure for lipid peroxidation and thus oxidative damage, did not differ between both genotype sets. These results suggest that urban Daphnia have evolved a more efficient stress resistance compared to rural animals. While raising the animals at 24°C compared with 20°C led to a reduction in total protein content and an increased activity of SOD and GST, we observed no urbanization × temperature interactions. Evolutionary differentiation thus involved constitutive trait changes but no evolution of plasticity. Integrating our data on physiological traits with existing data on life-history traits measured on the same set of genotypes reveals an overarching POLS, structuring covariation patterns between life history and physiology in urban, but not in rural animals. Our results thus show that urbanization not only induces evolution of mean trait values, but also shapes trait covariation patterns. More specifically, in urban animals life-history and physiology are integrated into a syndrome, covarying along the pace-of-life axis, while the two sets of traits were unlinked in rural animals. In the following paragraphs we discuss these results in more detail.

(a). Genetic differentiation in energy reserves and stress physiology

Most urban ecological studies reporting on stress physiology focus on terrestrial vertebrates such as birds and mammals [12,23,25–27] (but see [25] for amphibians). As most studies use field collected animals, they do not reveal whether or not the observed differences in physiological stress-coping mechanisms between urban and rural populations are driven by evolution (but see e.g. [58,59]). Here we show that populations of the water flea D. magna inhabiting small ponds in urban areas have genetically diverged from those in rural areas. Urban Daphnia genotypes have higher concentrations of protein, fat and sugars per unit of dry mass. Urban populations are therefore characterized by a higher available energy budget [60] than rural populations, which is important in determining their scope for growth under stress [60,61]. Urban populations are also characterized by a reduced activity of antioxidant enzymes (SOD and CAT), while oxidative damage to lipids did not differ, indicative of more effective defence responses or investment in other defence proteins. In general, these results are in line with earlier findings on the evolutionary and acclimatory responsiveness of Daphnia when exposed to human-associated toxicants such as heavy metals and pesticides [62,63] and natural toxicants such as cyanotoxins [64].

The results of our study reinforce earlier observations on genetic differentiation in D. magna populations in response to urbanization, which involved both heat tolerance [43] and life-history traits [35]. Combined, they suggest that urban populations of D. magna are genetically different from rural ones in multiple traits linked to warming, pace of life and overall stress. Traits range from heat tolerance (i.e. survival at higher temperatures) to energy metabolism, oxidative stress responses, and a diverse set of life-history features. The number of studies unequivocally showing evolution in response to urbanization is rapidly growing (e.g. [3,43,65–67]), but none involved as an extended set of traits as reported for Daphnia.

(b). The pace-of-life syndrome hypothesis: linking life history and physiology

Our earlier work on life-history differentiation indicated that urban D. magna populations showed a clear-cut shift to a faster pace-of-life compared to rural populations. Due to energy allocation trade-offs, stress physiological responses are often closely intertwined with higher-level response traits such as behavioural or life-history traits, potentially leading to a POLS [36] structuring covariation patterns between life-history and physiological traits. Strikingly, structural equation modelling revealed such an association between physiological and life-history traits in urban but not in rural populations. In urban populations, animals with a faster lifestyle, have a lower protein content, upregulated SOD and GST activity, and suffer higher lipid peroxidation (figure 3). These covariation patterns suggest that, within the set of urban genotypes, investment in a faster lifestyle depletes energy reserves (see electronic supplementary material A, figure S8) and increases antioxidant defence activity (SOD), yet still comes at a cost of increased lipid peroxidation (MDA). The fact that an underlying syndrome between life history and physiology is only present in the city populations is in line with the hypothesis that such strong associations between suits of life-history, physiological or behavioural traits are more likely to occur in low-quality habitats or stressful conditions (e.g. [38–40,68]).

While within urban genotypes the POLS reflects that a faster lifestyle comes at a cost of reduced energy reserves and higher oxidative damage, we observe the opposite relation when comparing mean traits of urban and rural populations. Here, the overall mean faster lifestyle of urban populations is associated with higher investment in energy storage (see also [69]) and with reduced investment in antioxidant defence. Similar to the observations made in many other studies (reviewed in [11]), the evolution of oxidative stress responses in urban populations is not consistent across all three enzymes tested (table 1 and figure 2). Overall, however, there is a reduction in oxidative stress enzyme activity in urban compared to rural populations, while there is no difference in oxidative damage, suggestive of a more efficient defence mechanism (figure 2). These results are at first sight counterintuitive given that urbanization induces a faster pace of life, which is expected to result in a higher metabolic rate and associated increase in oxidative stress [37,70]. Our results might, however, reflect that urban populations are adapted to an overall more stressful environment through a reduced metabolic rate and investment in other defence proteins (e.g. heat shock proteins) under baseline conditions, as has been repeatedly observed in a variety of organisms [70,71]. This might be achieved through the evolution of more efficient defence mechanisms or through trade-offs with other traits, such a antipredator defence or immune function [71–73]. Within the set of urban genotypes, however, variation in traits among genotypes does reflect a classic trade-off between faster life and reduced energy reserves and increased oxidative stress.

(c). Is stress in the city leading to adaptive evolution?

We documented genetic differentiation of D. magna populations along urbanization gradients, which represent multifarious selection gradients. City ponds are warmer [30] and likely suffer more mechanical disturbances linked to management [31] (K.I.B. 2013–2015, personal observations) than rural ponds, and both factors might select for a faster lifestyle [35]. The impact of temperature might be both direct as well as indirect, e.g. through increased feeding activity of invertebrate predators or changed biotic interactions (competition, parasitism). While urbanization potentially involves many environmental stressors, including pollution, an extensive field study on 81 ponds in the study region revealed only weak differences in nutrient concentrations (slightly lower nitrogen concentrations in urban compared with rural ponds [74,75]), no significant differences in food resources (chlorophyll a concentrations quantifying phytoplankton abundances [74,75]), and no significant differences between urban and rural ponds in overall metal contamination (26 metals) and pesticide concentrations (21 pesticides) between urban and rural ponds [75]). This suggests that differences in water quality or food resources are probably not a recurrent selection factor differentiating rural and urban ponds, although we interpret these data cautiously because these environmental data were collected during a single snapshot sampling campaign.

The genetic differences we observed between urban and rural populations are most likely to reflect adaptive evolution to cope with the higher temperatures and stress levels associated with anthropogenic disturbances in urban environments. Our study populations were scattered over a quite large region involving different cities and our results therefore suggest that city populations consistently behave differently from rural ones irrespective of the region from which they were collected. As we did not test for fitness differences between urban and rural populations, non-adaptive evolutionary processes can, however, not be ruled out as drivers of the observed genetic differentiation. Depending on the genetic architecture of the traits, genetic drift might generate a repeatable signal [76]. The broad set of responsive traits (physiology, life history and heat tolerance) observed in this study and previous work [35,43], and the fact that heat tolerance has a direct fitness link to temperature differences, provides further evidence for adaptive evolution as a driver of the observed genetic differentiation between urban and rural populations. Finally, the significant effect of genotype, nested within population, that we observe, indicates the presence of evolutionary potential for four out of seven traits measured.

5. Conclusion and future directions

We show that urban Daphnia populations differ genetically from rural ones for a suit of traits linked to energy metabolism and stress physiology. Additionally, urbanization structures the presence of an overarching POLS linking life history and stress physiology. Strikingly, the direction of evolutionary divergence in mean traits when comparing urban and rural populations is different from the evolved POLS-direction within the urban populations. The latter involves a classic trade-off between fast development, coinciding with energy depletion and higher oxidative stress levels. Possibly, the evolution of higher energy reserves and more efficient stress coping mechanisms in urban populations provides the needed energy to drive the evolution of a faster pace of life, which is then established via the evolved overarching POLS, but nevertheless comes at a cost of energy depletion and increased oxidative stress. Our results point to the need for more work on stress physiological responses and the mechanistic underpinning of trait covariation among and within sets of genotypes [77]. Additionally an effort should be made to include subcontinental (e.g. European) or global-scale studies as these will reveal whether our observations can be generalized at larger spatial scales [4,66]. While common garden experiments as used in the present study, involving urban and rural populations sampled across different urban cores in the same region (Flanders, Belgium), provide strong evidence for genetic differentiation in ecologically relevant traits, they are labour-intensive, which limits their application to large-scale survey studies. The capacity to detect urban evolution via genomic signatures would be a major step forward, given that patterns can then be compared across regions and taxa [58]. Finally, there is a strong need to quantify the consequences of urban evolution for ecosystem functioning and services to society, such as pollination, pest control and top-down control of primary producers. Additionally, given the strong temperature differences observed between urban and rural ponds, urban environments may provide a window into conditions pond-dwelling organisms will experience as the climate warms.

Data accessibility

Data are made available at Dryad Digital Repository [78].

Authors' contributions

K.I.B. and L.D.M. designed the experiment. K.I.B. performed the experiment and performed the statistical analyses with input of L.D.M. and R.S. K.I.B. wrote the first draft of the manuscript. R.S. and L.D.M. contributed to the subsequent versions of the manuscript.

Competing interests

The authors declare no competing interests.

Funding

This work was supported by Belspo (IAP SPEEDY). This work was also financially supported by KU Leuven Research Council funding PF/2010/07 and C16/17/002. K.I.B. acknowledges a FWO PhD (Aspirant) fellowship.