Introduction

Bacterial communities that reside within animal hosts are a powerful force influencing the biology of their hosts. The extent of bacterial communities’ influence on host life history, physiology, and behavior has long been a subject of investigation. Across the animal kingdom, host-associated bacteria may play a nutritional role [1], promote developmental processes in the host [2], and aid hosts in acclimating to ecological stressors such as low temperatures [3] and colonization resistance against invading natural pathogens [4]. Associated bacterial communities are also implicated in mediating their hosts’ adaptation to selection pressures through their variation across host genotypes [5, 6]. Host-associated microbes may be vertical transmitted directly through a variety of specialized mechanisms involving depositing microbes in or on the egg [7] or indirectly by parents altering the environment offspring are exposed to [8].

Despite the documented benefits of host-associated microbiota in a wide range of animal hosts, signatures of phylosymbiosis between host species and host-associated microbial community structure vary between different groups of animals. For instance, there is almost no association between microbial communities and avian host species or genotype structure [9] while marine invertebrate species show varying degrees of host genetic versus environmental forces influencing the structure of their associated microbial communities [10]. Genome-wide association studies in humans and mice have reported various genomic loci playing significant roles in host-associated microbial community composition [11, 12], which are generally involved in either host immunity or various lipid or carbohydrate metabolism pathways. In the aquatic snail Biomphalaria glabrata, genotypic variation at a single locus involved in pathogen recognition is responsible for an altered host-associated bacterial community [13]. However, these studies generally report varying degrees of host genetic versus environmental (e.g., diet) influence on host-associated microbial community structures. Hence, more studies in more diverse animal models are required to elucidate the role of deterministic forces such as host genetics in host-associated microbial community structure to better understand its role in host ecology and evolution [14].

The freshwater crustacean Daphnia is a well-established model system in ecology and evolutionary biology and is a particularly compelling system to investigate host bacterial community dynamics, both within an individual host and on host populations [15]. Studies in Daphnia magna indicate that host-associated bacterial communities are generally required for survival, reproduction, and nutrition [16]. Daphnia-associated bacteria also influence ecologically relevant traits such as tolerance to cyanobacteria [17, 18], host embryonic development [19], and overall ecological success [20]. The community composition of gut bacteria in D. magna is influenced by both the host genotype and by a variation in environmental bacteria [21, 22]. However, the extent of host genotype influence on Daphnia bacterial communities is debated, with a laboratory study revealing no variation in whole D. magna bacterial communities by host genotype [23] and a recent mesocosm study showing no variation by host genotype in D. magna gut bacterial communities raised in natural lake water [24].

Paleo-genetic reconstruction studies suggest that the genetic architecture of Daphnia populations varies by sediment depth and reflects responses to environmental stressors over time [27]. The bacterial communities associated with Daphnia-resting eggs deposited in the sediment have a beneficial effect on the survival of Daphnia hatched from those eggs, and may be required for the establishment of clonal lines [19]. However, no studies so far have utilized naturally occurring genetic variation among Daphnia across sediment depth to address host genotypic variation as a determinant of composition of bacterial communities in Daphnia. Also, studies investigating variation of the bacterial community associated specifically with the Daphnia tissue excluding the gut (e.g., filtering apparatus) are few, though these other associated bacteria may play a role in ecosystem level processes such as transfer of dissolved organic matter across the food web [28].

Here we investigate whether Daphnia galeata clonal lineages harbor genotype-specific bacteria after being reared in common garden in a laboratory setting for 5 years. We sequenced the bacterial communities associated with eight D. galeata genotypes hatched from two distinct sediment layers (1989 or 2009) of lake Greifensee, corresponding to different stages of reoligotrophication in that lake. Daphnia population structure changed considerably with eutrophication and reoligotrophication in Greifensee [27]. Genotypes isolated from the same sediment layer are genetically more similar than those from another sediment layer based on whole-genome sequencing data [29]. We determined whether bacterial communities of the (a) gut and (b) remaining body tissue of D. galeata differed between genotypes and sediment layers from which they were hatched (2 sediment layers × 4 genotypes) and further tested for a correlation between genetic distance among Daphnia genotypes and dissimilarity in their bacterial community composition. We also compared host-associated bacterial communities to those in the medium in which the Daphnia were reared to empirically validate the common garden used in this study. We hypothesized that Daphnia-associated bacterial communities would vary significantly by (a) the host genotype and (b) the sediment layer of origin, since sediment layers in our study represent distinct genetic clusters of Daphnia. Furthermore, we did not expect bacterial communities of water medium the Daphnia were kept in to differentiate between Daphnia genotypes or between sediment layers of origin.

Material and Methods

Daphnia galeata Culturing Conditions (for Bacterial Community Sequencing Experiment).

The eight D. galeata genotypes in this study each originated from resting eggs taken from a single sediment core collected from Greifensee (N 47° 20′ 41″, E 8° 40′ 21″) on 16 December 2014 (see also [29]). In the same week, intact resting eggs were collected from two separate sediment layers (corresponding to years 1989 and 2009; the sediment core dating method is described in [30]) and hatched simultaneously in 6-well plates containing Greifensee lake water filtered through a 0.45-µm mesh. Resting eggs were placed in separate wells (but not in a sterile environment). After hatching, the genotypes were maintained as clonal lineages in identical laboratory conditions for 5 years. Standard conditions were: animals were maintained at 12 °C in 100-mL filtered Greifensee lake water, routinely fed with the green algae Acutodesmus obliquus (formerly Scenedesmus obliquus) grown in a chemostat in WC medium [31] and diluted in filtered lake water before feeding. The amount of added food (three times a week) corresponded to 0.9 mg C/L per 10 animals. Medium was refreshed every 5–6 weeks. In preparation for an experiment, four genotypes (GR_020, GR_023, GR_024, GR_025) from the sediment layer dating to 1989 and four (GR_052, GR_053, GR_054, GR_055) from the layer 2009 were moved to 20 °C in October 2019, split into three replicate lines and maintained simultaneously for the experiment (see below).

Experimental Setup

Seven to eight females with eggs (8 genotypes × 3 replicates = 24 populations) were maintained together in 200-mL medium and fed 1.5 mg C/L daily, with medium change every alternate day. Twenty to 27 juveniles (experimental animals) produced by the females were moved to fresh 200-mL medium and fed 3 mg C/L with a medium change on alternate days. Jars with experimental animals were assayed twice a day for free-swimming juveniles of the next generation. Adult Daphnia were moved to fresh medium (~ 24 h after the first appearance of two to three free-swimming juveniles) to allow loosely associated-bacteria and food particles to diffuse away from the Daphnia before dissection. Twenty adult Daphnia per replicate were then dissected 52 ± 3 h after the first appearance of juveniles (or 26 ± 2 h after transfer to fresh medium). Two different batches of filtered lake water were used as Daphnia medium during the experiment.

Preparation of DNA Material

For gut samples, 20 Daphnia were dissected under a stereo microscope, each in individual droplets of nuclease-free water using sterilized forceps, and extracted guts were immediately moved to a 20-µL droplet of nuclease-free water. This pool of 20 guts was then transferred to a 1.5-mL microcentrifuge tube. For body samples, the remaining Daphnia tissue after the extraction of guts was pooled into a separate 1.5-mL microcentrifuge tube. Forceps were flamed between individual dissections to minimize cross-contamination between gut and body samples. For medium samples, 200 mL Daphnia medium from which the experimental animals were collected prior to dissection was filtered through a 0.22-µm filter using a sterile syringe. The filter was transferred to a 2-mL microcentrifuge tube. All samples were immediately stored at − 20 °C until further processing. Preparation of DNA material was done on nine non-consecutive days, and the order of processing was randomized across genotypes and replicates.

Bacterial Community Profiling

DNA was extracted using the Qiagen Blood & Tissue kit (Cat #69,506). Briefly, all samples were lysed at 56 °C for 4 h after which the recommended protocol for DNA extraction provided by the manufacturer of the Qiagen Blood & Tissue kit was followed for Daphnia gut and body samples. Modifications to the protocol were made in extraction reagent volumes for the medium samples according to [32] in order to maximize DNA recovery. All samples were eluted in 40 µL kit elution buffer for 20 min.

A nested PCR approach was done due to samples being of low biomass. Universal 16S primers 27F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGACTT-3′) were used to amplify the full-length 16S gene with the following cycling conditions 94 °C—30 s; 50 °C—45 s; 68 °C—90 s; 30 cycles [33, 34]. Amplified products were purified using the QiaQuick PCR purification kit (Cat# 28,106) before standard library preparation and amplicon metagenomics of the V3-V4 region using primers 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Library preparation and sequencing was done by Novogene UK (www.novogene.com Cambridge, UK).

Pre-Processing of 16S Sequencing Reads

Sequencing resulted in ~ 8.3 M reads (minimum = 55,804, maximum = 137,685 per sample). Raw reads were trimmed, quality filtered, and chimeras were removed. Amplicon sequence variants were clustered using UPARSE [35], denoised into Zero-radius OTUs (ZOTUs) based on 97% sequence similarity using UNOISE3 [36] and annotated using the non-Bayesian SINTAX classifier [37] and the Silva database [38]. Eight ZOTUs (4 of unidentified phylum, 4 chloroplasts) were filtered from the dataset. Samples were rarefied to an even depth of 55,000 reads. The rarefaction step resulted in the removal of one ZOTU, resulting in a total of 432 ZOTUs in the dataset. Whole-genome sequencing data processing of the eight genotypes in this study was performed separately [29].

Biodiversity Measures and Statistical Testing

All analyses were carried out in R v4.0.2 using the phyloseq package [39]. First, we investigated variation in bacterial community beta diversity between sediment layers of origin and among Daphnia genotypes. A two-way PERMANOVA of the β-diversity metric Weighted Unifrac distance was performed with genotype nested within sediment layer using the adonis function in the vegan package [40]. The Weighted Unifrac distance captures differential relative abundance as well as phylogenetic relatedness of ZOTUs. This was done separately for each sample type (i.e., Daphnia gut, body, and medium), as we were primarily interested in differences between genotypes and sediment layers. Finally, the capscale function was used to estimate the percentage of variation explained by PERMANOVA models.

Second, we tested for a correlation between Weighted Unifrac distances between bacterial communities (using the above 16S data) and genetic distances between host genotypes using distances based on 41,771 SNPs from Daphnia whole-genome sequencing data [29]. A Mantel test was performed between the dissimilarity matrices, separately for Daphnia gut and body bacterial communities using the mantel function of the vegan package (Spearman’s rho, 9999 permutations). Hierarchical clustering was performed on Weighted Unifrac distance of Daphnia gut and body-associated bacteria as well as genetic distances between host genotypes using the hclust function (method = Ward.D2) and customized using the dendextend package [41]. We also identified ZOTUs which could be indicative of specific Daphnia genotypes based on their abundance distributions using the Indicspecies package in R, separately for each tissue type. For this, we used the signassoc function (two-tailed test, 9999 permutations, corrected for multiple comparisons using the Sidak method) [42].

Third, we investigated the differential abundance of the dominant classes of bacteria in Daphnia tissues and medium between sediment layers. ZOTUs were aggregated at the class level using the tax_glom function. The bacterial classes that constituted < 1% of counts in the dataset (unless they were present in all samples) were classified as “Other” for visual representation and identification of differential abundance among dominant classes using the aggregate_rare function of the microbiome package (an extension of phyloseq). The DESeq2 package (Wald test) was used on agglomerated classes to identify differentially abundant classes [43]. Relative abundances were compared between sediment layers separately for each sample type (4 genotypes × 3 replicates = 12 samples per sediment layer). Then, relative abundances of major bacterial classes were also compared between sample types; i.e., across Daphnia gut, body, and medium (8 genotypes × 3 replicates = 24 samples per sample type). We then used the indicator species analysis described above to identify taxa that associate specifically with sample type.

Finally, alpha diversity metrics ZOTU richness and the Shannon Index were calculated for all samples using the estimate_richness function. ANOVAs were carried out for both alpha diversity measures, for all sample types together (three-way ANOVA with sediment layer, sample type, and genotype nested within sediment layer). This was followed by posthoc TukeyHSD tests to identify pairwise differences among (a) genotypes and (b) sample types (“sediment layer” was excluded from the posthoc test since it was not significant in the main test, see Results).

Results

Beta Diversity of Bacterial Communities Between Daphnia Genotypes

The Weighted Unifrac distance varied significantly by genotype but not sediment layer, for both Daphnia gut and body bacterial communities (Fig. 1a, b; Table 1). Medium bacterial communities varied neither by genotype nor sediment layer (Fig. 1c; Table 1). For gut bacterial communities, there was overlap between some Daphnia genotypes, but certain genotypes, e.g., GR055, GR053, and GR023 cluster separately from the rest (Fig. 1a). For Daphnia bodies, most genotypes formed separate clusters (Fig. 1b). We also found 38 ZOTUs that showed abundance distributions skewed towards specific genotypes in the Daphnia gut and body tissue. Similarly, 14 ZOTUs were associated with the media of specific Daphnia genotypes. However, only two of these ZOTUs were indicative of host genotype in the Daphnia gut or body. This suggests that differential abundance of ZOTUs in the media need not correspond with their differential abundance within Daphnia (see Fig. S3, Fig. S4 and Table S5).

Fig. 1
figure 1

PCOA plots of Weighted Unifrac distance between Daphnia a gut, b body, and c medium bacterial communities. Colors represent different genotypes; shades of blue are genotypes originating from 1989; shades of orange/red are those from 2009. Barplots adjacent to PCOA plots show % variation explained by each PCOA axis (Axes 1–10)

Full size image
Table 1 PERMANOVAs on Weighted Unifrac distances performed separately for gut (A), body (B), and medium (C) bacterial communities (9999 permutations). p < 0.05 are highlighted in bold. % variation column shows the % variation explained based on db-RDA using the capscale function
Full size table

Relationship Between Daphnia Genetic Distance and Bacterial Community Composition

We further tested for a correlation between the genetic distance among host genotypes based on 41,771 SNPs from whole genome sequencing data and the average Weighted Unifrac distance between the bacterial communities of these genotypes (Table 2). We found no significant correlation between host genetic distance and average Weighted Unifrac distance of bacterial communities at the ZOTU level in Daphnia guts (Mantel statistic based on Spearman’s rho r =  − 0.21, p = 0.889) or Daphnia bodies (r =  − 0.26, p = 0.935). Furthermore, neither Daphnia gut nor body bacterial communities clustered by sediment layer of origin, whereas Daphnia genotypes did (Fig. 2).

Table 2 Mantel test results based on Spearman’s rank correlation (9999 permutations), between Weighted Unifrac distances of Daphnia bacterial communities (based on 16S sequencing data) and genetic distances between Daphnia genotypes (based on whole genome sequencing data)
Full size table
Fig. 2
figure 2

Hierarchical clustering plots of pairwise distance matrices between Daphnia genotypes for bacterial communities and host genetic distances; colors represent sediment layer from which genotypes originated (blue–1989, red–2009). a Weighted Unifrac distance of average ZOTU counts between Daphnia gut bacterial communities, based on 16S sequencing data. b Weighted Unifrac distance of average ZOTU counts between Daphnia body bacterial communities, based on 16S sequencing data. c Genetic distances between Daphnia genotypes based on 41,771 SNPs from Daphnia whole genome sequencing data. p values for correlation between distance matrices were calculated separately using a Mantel test (see Methods)

Full size image

Bacterial Community Composition in Daphnia Between Sediment Layers of Origin

The Daphnia gut, body, and medium bacterial communities were comprised of 10 dominant classes of bacteria (Fig. 3). Comparisons between sediment layers were made separately for each sample type and averaged over genotypes (see Methods). Alphaproteobacteria was significantly more abundant in the guts of genotypes hatched from 2009 (9.01 ± 9.8%) compared to the guts of those from 1989 (3.89 ± 2.35%) while Acidimicrobiia was significantly more abundant in the bodies of Daphnia from 1989 (0.03 ± 0.02%) compared to the bodies of Daphnia from 2009 (0.01 ± 0.005%) (Supplementary Table 1). However, the higher abundance of Alphaproteobacteria in genotypes from 2009 is due to higher abundance in a single genotype, 2009_GR055 (Fig. 3) indicating that variation is primarily between Daphnia genotypes. Class Acidimicrobiia was rare within Daphnia tissue.

Fig. 3
figure 3

Bacterial communities in Daphnia a body, b gut, and c medium. Four columns on the left depict genotypes hatched from 1989 and the four on the right, from 2009. Bacterial classes constituting < 1% of the dataset and not present in every sample are classified as “Other.” The group “Other” is comprised of 23 bacterial classes

Full size image

Bacterial Community Composition Between Daphnia Tissue and Medium

The most abundant class, Betaproteobacteria, was significantly more abundant in the Daphnia gut (65.9 ± 15.6%, mean ± standard deviation) and body (56.6 ± 7.8%) than in the medium (46.5 ± 9.8). The next most abundant class, Flavobacteriia, was significantly more abundant in the medium (22.4 ± 7.3%) than in the Daphnia gut (9.8 ± 4.2%), with the Daphnia body having an intermediate relative abundance (13.4 ± 5.2). Sphingobacteriia was more abundant in the Daphnia body (13.7 ± 6.9%) than in the medium (10.8 ± 6.6%). The less common classes were also differentially distributed across Daphnia tissue: Gammaproteobacteria was significantly more abundant in the Daphnia gut than the body and medium. Alphaproteobacteria and Verrucomicrobiae were significantly more abundant in the Daphnia body than both the gut and medium, whereas Cytophagia, Actinobacteria and Acidimicrobiia were significantly more abundant in the medium than the Daphnia body and gut (Fig. 3, Supplementary Tables 2 and 3). We also found 73 ZOTUs that showed abundance distributions skewed towards specific sample types (see Fig. S5 and Table S6).

Alpha Diversity of Bacterial Communities Between Daphnia Genotypes and Tissues

We used ZOTU richness and Shannon Index as measures of alpha diversity in the samples (Fig. 4). There was no significant variation in ZOTU richness or Shannon Index between sediment layers (Table 3). ZOTU richness varied significantly across genotypes and sample types. Genotype GR053 had generally higher ZOTU richness, compared to bacterial communities of GR023, GR052, GR054, and GR055 (Supplementary Table 4). Tukey HSD pairwise comparisons showed that the ZOTU richness was higher in medium bacterial communities than the Daphnia body though only marginally significant (Supplementary Table 4). The Shannon Index (Fig. 4b) varied significantly only by sample type (Table 3); it was significantly higher for medium than for gut or body bacterial communities (Fig. 4, Supplementary Table 4) indicating that bacterial communities in the medium are more even in composition compared to Daphnia body and gut. Notably, ZOTU richness varied by genotype but the Shannon Index did not, suggesting that the differing ZOTUs across genotypes were likely rare in the dataset.

Fig. 4
figure 4

Alpha diversity of bacterial communities in Daphnia tissues and media. Each replicate is represented with a colored circle (gut-green, body-red, or medium-blue), and the cross signs represent arithmetic means. Top row includes genotypes originating from sediment layer 1989, bottom rows are those from 2009. a ZOTU richness of Daphnia gut, body, and medium bacterial communities

Full size image
Table 3 ANOVA of alpha diversity metrics across sediment layers, genotypes, and sample types. Genotypes are nested within sediment layer. Model used was variable ~ sediment layer/genotype + sample type. ANOVA, p < 0.05 are highlighted in bold
Full size table

Discussion

We compared the bacterial communities of Daphnia galeata guts, bodies, and culturing medium across genotypes simultaneously hatched from two different sediment layers of the same sediment core of Greifensee, and maintained in the laboratory under identical conditions for 5 years before collection of this 16S sequencing data. Daphnia genotypes hatched from the two sediment layers (1989 and 2009) formed distinct genetic clusters [29].We found that beta diversity of the gut and body bacterial communities differed significantly between genotypes, confirming our hypothesis that the bacterial community composition has a host-genetic component. However, this variation in beta diversity of bacterial communities did not correlate with genetic distance between their host genotypes. ZOTU richness varied by host genotype (but not by sediment layer of origin), mainly due to one divergent genotype, GR053 (Fig. 4), but the Shannon Index did not vary by genotype or sediment layer of origin. Overall, Daphnia bacterial community composition reported in this study is similar to those found in other Daphnia studies, particularly the dominance of Betaproteobacteria in Daphnia guts and Actinobacteria in the medium [22, 24, 26].

Previous laboratory studies have reported variation in Daphnia bacterial communities by genotype, but the extent of host genotype influence differed from study to study. For instance, D. magna gut bacterial communities did not vary by genotype in a mesocosm study [24]; similarly, in a laboratory study, whole D. magna bacterial communities did not vary by genotype but only by pond of origin [23]. In another experiment, D. magna genotype strongly influenced alpha and beta diversities of assembled bacterial communities when the genotypes were made germ-free first. However, the genotype effect was mainly driven by a large divergence between two specific genotypes that also originated from different ponds [26], consistent with a pond-of-origin effect reported elsewhere [23, 25]. When environmental factors such as temperature [25], diet [26], and the composition of environmental bacterial communities [22] were varied, all were found to play a significant role (comparable to or greater than host genotype) in shaping Daphnia bacterial communities. In contrast, our study compares bacterial communities of D. galeata genotypes originating from the same lake, replicated for genotype and distinct host genetic clusters, and we find that genotypes harbor both compositionally and phylogenetically distinct bacterial communities after being reared in a common garden for 5 years. We also found host genotypes belonging to distinct genetic clusters do not have divergent bacterial communities. Additionally, we empirically validated the common garden in the present study; bacterial communities in the culturing medium did not vary in alpha or beta diversity while Daphnia-associated bacterial communities did.

D. galeata genotypes isolated from distinct sediment layers in our study (1989 and 2009) differed compositionally in the abundance of some bacterial taxa. Specifically, the guts of Daphnia from 2009 contain a higher relative abundance of Alphaproteobacteria than the guts of those from 1989. However, there is no difference in alpha or beta diversity between sediment layers, and the higher relative abundance of Alphaproteobacteria in the guts of 2009 Daphnia is due to their higher relative abundance in a single D. galeata genotype, GR055 (Fig. 3). Thus, there is no consistency in bacterial communities associated with D. galeata genotypes belonging to two distinct host genetic clusters.

Despite the significant variation by host genotype, bacterial community composition in Daphnia guts and bodies did not correlate with genetic distance between the host genotypes [29] which is consistent with the absence of variation in bacterial communities of the different Daphnia genotypes between sediment layers of origin. In contrast, other aquatic hosts, e.g., sticklebacks [44] and sponges, [45] exhibit variation in bacterial communities between host genotypes and significant positive correlation between host genetic distance and divergence in bacterial community composition. Our results suggest that factors other than host genetics and environmental exposure may shape the structure of host bacterial communities. These could include stochastic processes [46] or interspecies interactions between members of the host’s bacterial community, which may range from co-operative to competitive [47]. Specific dispersal abilities of microbes may also determine their abundance within hosts [48].

The Daphnia body bacterial community in our study reflects bacterial groups associated with the filtering apparatus as well as epibionts on the Daphnia carapace. Bacterial communities isolated from Daphnia body also have differentially abundant associated taxa compared to the gut. Verrucomicrobiae associated with polysaccharide degradation [49] was significantly more abundant in the Daphnia body than both the gut and the medium (see Supplementary Tables 2 and 3). Thus, further “omics” studies targeting the transcriptome and metabolome would be required to assess the putative functional roles of such bacterial taxa associated differentially across Daphnia tissue [50, 51].

Several studies have demonstrated beneficial effects of Daphnia-associated bacteria such as tolerance to cyanobacteria [17], contribution to host development [19], and their general requirement for survival [52]. However, investigation on the adaptive significance and possible co-evolution between Daphnia and bacterial communities has yielded mixed results. One study reported no apparent benefit of long-term symbiosis with bacteria in Daphnia, or in other species of freshwater zooplankton such as various rotifers and crustaceans [53]. In contrast, D. magna benefit more from receiving sympatric vs. allopatric bacterial communities when exposed to environmental stressors such as toxic cyanobacteria [18] though such fitness benefits are also reportedly weaker in semi-natural settings [24]. Furthermore, it has also been suggested that zooplankton-associated bacteria essential to the host are functionally redundant [21, 53]. In this study, we show that while there is variation by host genotype in host-associated bacterial communities, this variation is not determined by genetic distances between hosts. If host-associated bacterial communities are beneficial and required for survival but also functionally redundant, the purpose of this genotype-specific diversity and the mechanism of co-evolution with bacteria among aquatic hosts remains unknown. Hence, a focus on the functional roles of host-associated bacteria and mechanisms of their vertical transmission in hosts could advance our understanding of their role in host ecology and evolution.