Invasive lumbricid earthworms in North America – different life-histories but common dispersal?

Abstract

Aim

Lumbricid earthworms are invasive across northern North America, causing notable changes in forest ecosystems. During their range expansion they encountered harsher climatic conditions compared to their native ranges in evolutionary short time. This study investigated if (1) dispersal barriers, (2) climatic selection, or (3) anthropogenic activities, such as fishing bait disposal, structure the dispersal of free-living earthworm populations.

Location

North America, forest habitats along former Wisconsinan glaciation line

Taxon

Lumbricus terrestris, L. rubellus

Methods

Lumbricus terrestris and L. rubellus co-occur in the same habitats but differ in ecology and use Conservation approach for goblin species were sampled in five transects ranging from the east to the west coast of northern North America, including major dispersal barriers, three different climate zones, and bait shops near sampling locations. Genetic diversity and structure were compared between the two species, and the presence of free-living bait shop genotypes was assessed using four markers (COI, 16S rDNA, 12S rDNA, H3).

Results

Populations of both species were genetically diverse with some geographic structure, which was more pronounced in L. terrestris than in L. rubellus. Common haplotypes were present in all regions, but locally restricted haplotypes also occurred. Further, two distinct genetic clades of L. terrestris co-occurred only in the two most distant transects (Alberta and Minnesota). Genotypes identical to bait individuals were omnipresent in field populations of L. terrestris.

Main conclusions

Genetic diversity was high in both species, and invasive populations represented a genetic subset of European earthworms. Geographic and climatic dispersal barriers affected the less mobile species, L. rubellus, resulting in differences in genetic structure between the two species. Our results indicate common long-distance dispersal vectors and specific vectors affecting only L. terrestris. The roles of climate and anthropogenic activities are discussed, providing additional explanations of dispersal and new insights into establishment of invasive earthworms.

Introduction

European lumbricid earthworms are among the most successful invasive species in North America (James and Hendrix, 2004). European settlers introduced them at the east coast about 400 years ago, both accidentally and intentionally (Gates, 1976). Similar to many invasive species living above the ground, earthworms substantially alter the functioning of invaded ecosystems (Scheu & Parkinson, 1994; Mooney and Hobbs, 2000; Bohlen et al., 2004; Eisenhauer et al., 2007; Hendrix et al., 2008). They change physical and biotic properties of the soil, which affects the density of other soil invertebrates, plant community composition, and aboveground food webs (Lee, 1985; Edwards & Bohlen, 1996; Eisenhauer et al., 2007, 2010a; Craven et al., 2017; Ferlian et al., 2018). In general, presence of earthworms beneficially affects plant growth (Scheu, 2003) and plant competition (Eisenhauer & Scheu, 2008) where they are native, but can exert contrasting effects on ecosystems that developed without their presence (Bohlen et al., 2004; Hale et al., 2005; Craven et al., 2017).

As successful invaders, earthworms possess high tolerance for a wide range of environmental conditions, though they prefer clay soils with near neutral pH that restricts their distribution (Laverack, 1961; Curry, 2004; Fisichelli et al., 2013). Due to their ability to tolerate disturbances, they also occur in agricultural fields and meadows, with varying frequencies and abundances (Hendrix et al., 1992). However, earthworms are susceptible to prolonged freezing periods, drought and geographic barriers like mountain ranges and large waterbodies, which usually restrict their natural dispersal range (Reynolds, 1994; Eggleton et al., 2009). Active dispersal of earthworms is slow, but they accomplished to spread across northern North America within a few hundred years by passive dispersal or repeated introductions, and today they are present in large areas from the east coast to the mid-west east of the Rocky Mountains in Canada, and the Pacific coast (Reynolds, 1977, 1994; Scheu & Parkinson, 1994). The pronounced ecological consequences of earthworm invasions in North America are well documented, making earthworms one of the best-studied invasive animal species living below the ground (Wardle, 2011) and thus, a unique model system for biological invasion and accompanying effects (Hendrix et al., 2008).

Earlier studies using molecular markers demonstrated that genetic diversity of European earthworm populations in eastern North America is similar or slightly reduced compared to European populations (Hansen et al., 2005; Gailing et al., 2012; Fernandez et al., 2015). Multiple introductions and human-mediated dispersal presumably contributed to this high genetic diversity (Keller et al., 2007; Holdsworth et al., 2007; Cameron et al., 2007; Cameron & Bayne, 2009). However, genetic diversity and structure of invading earthworm species in North America so far have only been studied at local or regional scales. The identified genetic diversity and human-mediated dispersal patterns likely also apply at larger scales, but it is unclear if one common invasion event or several independent local invasions are responsible for the fast spread of European earthworms in northern North America. In general, the following dispersal scenario of earthworms in North America is likely: European earthworms spread from east to west, following the European migrations and transport of goods, suggesting that earthworm populations are genetically related in large areas east of the Great Plains. This invasive front likely stopped at the North American Midwest, an area of dry grassland and intensive agriculture, and continental climate of severe frost in winter and dry summers. The origin and dispersal of earthworm populations west of the Great Plains remains unclear but may be based on three scenarios. First, populations introduced at the west coast expanded to the east, crossing the Rocky Mountains. Second, multiple independent introductions occurred west and east of the Rocky Mountains. Third, earthworms dispersed from the east of the Rocky Mountains to the west or from east of the Great Plains to the west by long-distance dispersal.

During their expansion across northern North America, European earthworms established in distinct climate zones that differ in the amount and distribution of precipitation across the year, as well as frost intensity and duration, two abiotic factors that are known to drive earthworm distribution (Holmstrup, 2003; Curry, 2004; Uvarov et al., 2011; Fisichelli et al., 2013). At the west coast, precipitation is high (1200 mm y^-1), mild frost occurs sporadically and lasts for only few weeks between December and January. By contrast, in the central plains of North America, precipitation is low (400-600 mm y^-1), and strong frost conditions typically persist between November and March, with occasional night frost already starting in late August and extending into early June. In the east, precipitation is intermediate (800-1000 mm y^-1), and frost conditions typically last from December to February. Given this great range in climatic conditions, and the fast and wide ranging colonization of North America by European earthworms, knowledge on genetic diversity and relationships of populations across North America is needed for understanding dispersal mechanisms and population establishment.

We investigated the genetic structure of Lumbricus rubellus and L. terrestris, two exotic earthworm species that are widespread and common across northern North America. Both feed on litter but have distinct ecological preferences and life histories (Sims & Gerard, 1999). Lumbricus rubellus lives in horizontal burrows near the soil surface, moves freely within the litter layer for foraging, prefers neutral to slightly acidic soils and generally has a higher pH and frost tolerance than L. terrestris (Tiunov et al., 2006; Addison, 2009). In contrast, L. terrestris prefers neutral to slightly alkaline soils, lives in permanent, vertical burrows of up to 2 m depth, and collects litter in the vicinity of its burrow entrance (Sims & Gerard, 1999; Tiunov et al., 2006; Addison, 2009). Active dispersal rates of the two earthworm species range between 2-4 m y^-1 for L. terrestris and 10-14 m y^-1 for L. rubellus (Marinissen & van den Bosch, 1992). Lumbricus terrestris is commonly used as fishing bait and sold in bait shops which likely facilitates its dispersal. By contrast, L. rubellus is rarely sold in bait shops (A. Klein, pers. obs.). Disposal of fishing baits contributes substantially to the introduction and establishment of earthworm populations in recreational and fishing areas (Holdsworth et al., 2007; Keller et al., 2007), but the long-term establishment of these populations and further dispersal in the field remain unclear.

We sampled earthworms from five transects of ~150 to 300 km length (north-south orientation) in three climatic regions in two provinces in Canada and three states in the USA: the warm and moist region of British Columbia, Canada (BC), the cold and dry regions of Alberta, Canada (AL) and Minnesota, USA (MN), and the cold and moderately moist regions of Michigan, USA (MI) and New York State, USA (NY), respectively. This is the first study investigating the invasion of detritivorous soil animals on continental scale, including two different dispersal barriers and distinct climate zones in its sampling design.

We tested three hypotheses to understand if climate (H1), dispersal barriers (H2), and/or human migrations and transport (H3) predominantly structured the distribution and establishment of European earthworm species in northern North America:

(H1) From a genetically diverse source population, distinct genetic clades established in the different climate zones. By environmental filtering, individuals that were better adapted to regional drought or cold conditions survived, resulting in monophyletic clades in the different regions. (H2) Earthworms were introduced independently in areas that are separated by major dispersal barriers (the Rocky Mountains and the Great Plains), resulting in distinct genetic clades in the west (BC, AL). In contrast, east of the Great Plains (MN, MI, NY) geographic dispersal barriers are less extreme and therefore, genetic structure is less pronounced or absent. (H3) Human-mediated dispersal of earthworms counteracts local selection and disregards dispersal barriers, resulting in diverse earthworm populations and genotypes that are represented in all regions without any local clades occurring.

To account for human-mediated dispersal by dumping of fishing baits, which is a severe problem in northern North America (Holdsworth et al., 2007; Hale, 2008; Seidl & Klepeis, 2011), we purchased earthworms from bait shops near sampling locations in all transects to test if bait genotypes contribute to free-living populations, thereby increasing local diversity.

Materials and Methods

Sampling design – dispersal barriers and climate

Between May and July 2014 and in June 2015, we collected L. terrestris and L. rubellus along five transects (regions) spanning from east to west of the northern North American continent, ranging in the USA from New York State (Adirondack Mountains, transect NY), to the Midwest, i.e. Michigan (upper peninsula, transect MI) and Minnesota (near Minneapolis/St. Paul, transect MN; Table 1). In Canada, we collected earthworms east and west of the Rocky Mountains in Alberta (south of Calgary, transect AL) and British Columbia (near Vancouver, transect BC). Distances among transects ranged between 700-1600 km, and within transects earthworms were collected at five sampling locations with north-south orientation that were 20-80 km apart. The two major dispersal barriers for plants and animals are the extensive dry grassland areas of the Great Plains extending between transects Minnesota (USA) and Alberta (Canada), and the Rocky Mountains, which separate the two Canadian transects Alberta and British Columbia. Climate in east and central northern North America is similar to continental climate in Europe, but seasonality in North America is harsher with hotter and drier summers, and longer and colder winters, which is most extreme in Alberta and Minnesota. Climate in British Columbia differs from that in Europe as three different climate zones (Mediterranean, Continental and Oceanic) co-occur in the Greater Vancouver area.

Table 1.

Overview of sampling area, abbreviations of sampling locations and climatic characteristics of each transect. AMP, annual mean temperature; AMT, annual mean temperature. See Appendix Table S5 for GPS coordinates.

sampling transect	climate zone	climate characteristics	sampling location
British Columbia (BC)	mixed Mediterranean, oceanic and continental	warm and moist AMP: ~1200 mm/year AMT: 6-16°C	Cypress Provincial Park	BC_I
			Golden Ears Provincial Park	BC_II
			Cultus Lake	BC_III
Alberta (AL)	cold continental	cold and dry AMP: 400-750 mm/year AMT: -2-9°C	Crandell Lake	AL_I
			Waterton Springs	AL_II
			Maycroft	AL_III
			Eden Valley	AL_IV
			Fish Creek Park, Calgary	AL_V
			Nose Hill Park, Calgary	AL_VI
Minnesota (MN)	cold continental	cold and dry AMP: 400-750 mm/year AMT: -2-9°C	Nerstrand	MN_I
			Wood-Rill SNA	MN_II
			Wolsfeld Wood SNA	MN_III
			Warner Nature Center	MN_IV
			Pine Needles Preserve	MN_V
			Rush City	MN_VI
Michigan (MI)	moderate continental	cold and moderately moist AMP: 800-1000 mm/year AMT: 0-9°C	Turner	MI_I
			Tawas City	MI_II
			Alpena	MI_III
			Gaylord	MI_IV
New York (NY)	moderate continental	cold and moderately moist AMP: 800-1000 mm/year AMT: 0-9°C	Hamilton	NY_I
			Norwich	NY_II
			Newcomb	NY_III
			Lower Saranac Lake	NY_IV
			Lake Placid	NY_V
			Portland Waterfront	NY_VI

Earthworms were collected in forests by turning over logs, hand sorting of litter, digging or applying mustard solution to extract earthworms from soil. We measured soil pH from sampling locations; seven locations in our study were sampled by cooperation partners and were not available for pH measurements. Additionally, we purchased earthworms sold as fishing baits in bait shops close to sampling locations; all bait shops exclusively sold L. terrestris, restricting the bait shop dataset to a single species. Earthworms were washed, stored in 75% ethanol in the field and later transferred in the laboratory into 95% ethanol and stored at 16°C. One centimetre of tail tissue of each individual was cut and shipped to the University of Göttingen (Germany) for molecular analyses; remaining body parts are stored as voucher specimens at the University of Minnesota (Minneapolis-St. Paul, MN) and the University of British Columbia (Vancouver, BC).

Genetic analyses

Genomic DNA was extracted with the Genaxxon DNA Tissue Mini Prep Kit (Genaxxon; Ulm, Germany) following the manufacturer’s protocol. Four molecular markers were amplified: the mitochondrial genes COI (~600 bp; Folmer et al., 1994), 16S rDNA (~750 bp; Pérez-Losada et al., 2009), and 12S rDNA (~400 bp; Simon et al., 1994), and the nuclear gene Histone 3 (~350 bp; Colgan et al., 1998). The PCR cycling conditions had an initial activation step at 95°C for 3 min, 40 amplification cycles (denaturation at 95°C for 30 s, annealing at 53°C for 60 s, elongation at 72°C for 60 s), and a final elongation step at 72°C for 10 min and were sequenced at the Göttingen Genome Sequencing Laboratory (Georg August University Göttingen) and SeqLab Göttingen (Microsynth; Balgach, Switzerland). Sequences were submitted to the GenBank databases under accession number (XXX - to be provided when available) (GenBank www.ncbi.nlm.nih.gov/genbank). Sequences were checked with Sequencher 4.9 (Gene Codes Corporation, USA), and ambiguous positions were coded as wobble bases. Consensus sequences of the individual genes were assembled in BioEdit 7.0.1 (Hall, 1999) and aligned with ClustalW. Genes were analysed individually and in a combined matrix of 2,150 bp; all positions with wobble bases were deleted for further analyses. Sequence alignments (single genes and combined) were collapsed into haplotype alignments using FaBox 1.41 (Villesen, 2007). The best-fit models of sequence evolution were estimated with TOPALi v2.5 (Milne et al., 2004) using the Akaike information criterion (AIC; Akaike, 1973). Trees were constructed using MrBayes 3.2. (Ronquist et al., 2012), partitioning the combined alignment to the following lset parameters for L. rubellus (COI: nst=2, rates=invgamma; 16S rDNA: nst=2, rates=invgamma; 12S rDNA: nst=6, rates=invgamma; H3: nst=1, rates=invgamma) and L. terrestris (COI: nst=6, rates=gamma; 16S rDNA: nst=6, rates=invgamma; 12S rDNA: nst=2, rates=invgamma; H3: nst=1, rates=equal). A mcmc run of 4 million generations with default settings was performed. We compared the North American haplotype identities with European earthworms with Bayesian phylogenetic trees of the COI and H3 datasets including sequences available from NCBI. A list of the data sources is found in Appendix S4 a) and b). Parameter settings were nst=6, rates=invgamma and default settings for the mcmc run.

Phylogeography and genetic differentiation across putative dispersal barriers

Spatial distribution of genetic clades was analysed with haplotype networks and constructed for 16S rDNA, which provided the most informative resolution. Median-joining (MJ) networks (Bandelt et al., 1999) were constructed with PopART (University of Otago, Dunedin, New Zealand) and edited using Inkscape (Software Freedom Conservancy, USA). Parameters were set to equal weights for all mutations and the epsilon parameter to zero to restrict the choice of possible links in the final network.

To test hypotheses about climatic and geographic dispersal barriers, we used analyses of molecular variance (AMOVA) and analysed genetic differentiation among populations using the distance method of Tajima & Nei, pairwise differences without Gamma correction, and pairwise genetic distances using Arlequin 3.5.2.2 (Excoffier, 2015). AMOVAs were calculated with COI, the most variable gene regarding nucleotide diversity (Table S3), and earthworm populations were assigned a priori according to our first hypotheses (H1) into climate zones separating populations from British Columbia (mixed climate), Alberta and Minnesota (cold continental climate), Michigan and New York (moderate continental climate). To test for the relevance of geographic barriers (H2), populations were analysed in three different combinations: Great Plains as main dispersal barrier (BC, AL vs. MN, MI, NY), Rocky Mountains as main dispersal barrier (BC vs. AL, MN, MI, NY), and Rocky Mountains and Great Plains as main dispersal barriers (BC vs. AL vs. MN, MI, NY). Human influence on reducing the effect of dispersal barriers was tested by comparing genetic variance among transects (BC vs. AL vs. MN vs. MI vs. NY). If human transport plays a significant role for earthworms across large geographic distances (H3), genetic variance should be similar among regions.

Climate data

The response of genetically diverse earthworms to ecological factors was inspected by using a multiple regression matrix (MRM). Bioclimatic data were retrieved from WorldClim v2 bioclimatic variables database (Fick & Hijmans, 2017) and had a spatial resolution of ~5 km². The response matrix compared genetic pairwise differences of the COI sequence data and was calculated with the Analysis of Phylogenetics and Evolution (ape) package (Paradis et al., 2004). Tested factors were (1) environmental abiotic parameters, i.e. annual mean temperature (BIO01), maximum temperature of the warmest month (BIO05), minimum temperature of the coldest month (BIO06), mean temperature of the wettest quarter (BIO08), mean temperature of the driest quarter (BIO09), annual precipitation (BIO12), precipitation of the driest month (BIO14), precipitation seasonality (BIO15), and (2) the geographical parameter spatial distance and elevation (Table S2). Data were transformed into scaled explanatory distance matrices for standardisation. The spatial distance between each pair of samples was calculated using the Geographic Distance Generator v1.2.3 (Ersts, 2014) with the World Geodetic System (1984) setting for the reference spheroid and then normalised by dividing the values by the maximum distance value, thus measuring the absolute but normalized distances. The MRM function was executed with the R package ecodist (Goslee & Urban, 2007).

Linear regression analyses

We analysed correlations between sampling success as proxy for earthworm abundance and genetic diversity (nucleotide diversity and number of genetic clades in transects) with environmental factors potentially affecting earthworm distribution and abundance (i.e., sampling location soil pH and human population density; Table 1) performing simple linear regression analyses in Microsoft Excel 2013. The number of genetic clades per transect referred to the clades of the COI phylogenetic trees (Fig. 1a, 2a). Human population densities for each sampling region (BC, AL, MN, MI, and NY) were calculated based on data obtained from the US Census Bureau (https://www.census.gov/) and Statistics Canada (http://www.statcan.gc.ca/) using the mean of counties (USA) or regional districts (Canada) for each sampling location.

Figure 1.

Bayesian phylogenetic tree based on a supermatrix of four genes (COI, 16S, 12S, H3) of 120 individuals of L. rubellus (a), and distribution and abundance of the four genetic clades in the five transects across northern North America (b). The corresponding clades of the haplotype network analysis based on 16S are provided next to each clade, the area of each circle is proportional to the numbers of individuals for each haplotype, the colour code refers to the five transects British Columbia (BC, red), Alberta (AL, orange), Minnesota (MN, green), Michigan (MI, violet), and New York (NY, blue). For abbreviations of sampling locations see Table 1, posterior probabilities of well-supported clades are highlighted in bold.

Figure 2.

Bayesian phylogenetic tree based on a supermatrix of four genes (COI, 16S, 12S, H3) of 122 individuals of L. terrestris (a), and distribution and abundance of the seven genetic clades in the five transects across northern North America (b). The corresponding clades of the haplotype network analysis based on 16S are provided next to each clade, the area of each circle is proportional to the numbers of individuals for each haplotype, the colour code refers to the five transects as in Fig. 1. For abbreviations of sampling locations see Table 1, posterior probabilities of well-supported clades are highlighted in bold.

Results

Sampling and genetic diversity

In total, 120 L. rubellus (LR) and 122 L. terrestris (LT) individuals were sampled from the 25 locations. The number of individuals per transect varied from 12 to 48 for L. terrestris and from 12 to 37 for L. rubellus (Fig. S1). Abundances of both species were similar in British Columbia and Michigan, but differed in the other transects with higher numbers of L. rubellus in Minnesota and New York, and L. terrestris being more common in Alberta (Fig. S1). Nucleotide (NUD) and haplotype diversity (HTD) was greater in L. rubellus and decreased in both species from COI to 16S rDNA to 12S rDNA to H3 (Table S3). Overall, nucleotide diversity of COI was two or three times higher in L. rubellus than in L. terrestris (Table 2) and varied among transects. Diversity of L. rubellus was highest in the New York transect, followed by Alberta, and transects from British Columbia, Michigan and Minnesota had similar diversity. In contrast, nucleotide diversity of COI in L. terrestris was highest in Minnesota and similar in transects British Columbia and Alberta; diversity was low in transects New York and Michigan. The combination of transects showed that a considerable fraction of polymorphism was overlapping among transects, except for the two rather distinct populations of L. rubellus from New York and L. terrestris from Minnesota. Genetic variance and diversity in bait (n=104) and field populations of L. terrestris were very similar.

Table 2.

Nucleotide diversity of Lumbricus rubellus and L. terrestris in each transect and in different combinations of transects to assess complementary diversity across the sampling area. Mean percentages [%] with standard deviation (SD)

transect	L. rubellus	L. terrestris
	nucleotide diversity (SD) [%]
BC	4.25 ± 2.19	2.62 ± 1.37
AL	5.13 ± 2.72	2.24 ± 1.14
MN	4.43 ± 2.21	2.82 ± 1.45
MI	4.14 ± 2.14	1.55 ± 0.83
NY	7.10 ± 3.51	1.40 ± 0.78
NY*	5.65 ± 2.87	/
combination
BC, AL	4.59 ± 2.30	2.67 ± 1.34
BC, MN	4.63 ± 2.29	3.00 ± 1.50
BC, MI	4.15 ± 2.07	2.32 ± 1.19
BC, NY	6.46 ± 3.16	2.61 ± 1.33
AL, MN	4.86 ± 2.40	2.47 ± 1.24
AL, MI	4.46 ± 2.24	2.43 ± 1.22
AL, NY	6.79 ± 3.33	2.47 ± 1.24
MN, MI	4.65 ± 2.30	2.64 ± 1.34
MN, NY	6.43 ± 3.14	2.76 ± 1.40
MI, NY	6.49 ± 3.18	1.55 ± 0.82
MN, MI, NY	6.11 ± 2.98	2.52 ± 1.27

^*nucleotide diversity in transect NY without sampling point E, which differed significantly in nucleotide composition.

Relatedness and spatial distribution

In both species, earthworms were closely related resulting in phylogenetic trees with a weakly supported backbone and clades with mixed geographic origin. Accordingly, phylogenetic and geographic structure was generally weak, in particular in L. rubellus. However, in both species, some populations formed well-supported clades (posterior probabilities: 0.95-1; Fig. 1a) that were also recovered by haplotype network analyses. In L. rubellus, two clades comprised closely related individuals from all transects (mixed clades 1 and 4 with 37 and 60 individuals, respectively). However, five individuals from Minnesota (clade 2, green) and 18 individuals from New York (clade 3, blue) were distinct and did not occur in other transects (Fig. 1b). All North American COI haplotypes of L. rubellus could be assigned to lineages from Europe (Sechi, 2013; Giska et al., 2015). Haplotypes of clade 4 corresponded to the widespread European lineages A1-A3. Haplotypes in clade 1 and 2 clustered with European lineages C and D from Eastern Europe (Poland, Hungary, Balkans), and haplotypes in clade 3 clustered with lineage H, which is restricted to Germany and Austria. We compared COI lineages with the H3 dataset to check if mitochondrial and nuclear markers corresponded. The North American haplotypes of the COI clades 1, 2 and 4 carried the same H3 lineage that is also common in Europe (Martinsson & Erséus, 2017). Three individuals from Michigan (clade 4) carried a different H3 lineage, which is undescribed in Europe. Clade 3 comprised several H3 haplotypes, one known from Europe (Martinsson & Erséus, 2017) and one also present as widespread H3 lineage in the common COI clade 4.

Genetic distances among populations of L. terrestris were less distinct but had more haplotypes separating into more clades than L. rubellus (Fig. 2a, b). The largest clade of L. terrestris (clade 2, 52 individuals) included haplotypes from all transects. The second largest clade (clade 1, 32 individuals) consisted of a haplotype predominantly found in Alberta (orange) and Minnesota (green) and in one individual from New York (blue). Further, haplotypes from Alberta also occurred in separate clades together with Minnesota (clade 4, 7 individuals), British Columbia (red, single individual) and Michigan (violet, clade 5, 16 individuals). Notably, Minnesota and British Columbia also had distinct haplotypes that formed isolated monophyletic clades (clades 6, 5 individuals and 3, 5 individuals).

Most haplotypes of L. terrestris from bait shops were identical to common and widespread haplotypes from field populations (Fig. 3). Only few haplotypes formed separate clades (mainly AL and BC) or were related to rare field haplotypes (BC) from the same sampling region. The North American COI and H3 haplotypes of L. terrestris were closely related or identical to haplotypes described from Europe or North America in previous studies (Fig. S2, S3, Table S4; Brown et al., 1999; King et al., 2008, 2010; James et al., 2010; Richard et al., 2010; Novo et al., 2011; Klarica et al., 2012; Souleman et al., 2016; Martinsson & Erséus, 2017).

Figure 3.

Haplotype network based on 16S of 122 individuals of Lumbricus terrestris sampled in 25 field locations and of 104 individuals purchased in nearby bait shops. Bait shop individuals are grey and labelled with transects of their origin. Colour codes of transects and field individuals correspond to transects in Figs. 1 and 2.

Genetic differentiation across putative barriers

Analysis of molecular variance (AMOVA) across all four genes showed that most of the molecular variance was at local scale (within sampling points = populations, Table S1), with ~92-94% of variance in L. rubellus and ~70-73% in L. terrestris in the most variable gene (COI, Table 3). In both species, molecular variance predominantly resided at population level but was much clearer in L. rubellus with only 3.75% of variance among populations compared to L. terrestris with 17.92% (Table 3). Analyses based on a priori assigned populations to test for effects of climate (H1: transects BC vs. AL, MN vs. MI, NY), geographic barriers (H2: Great Plains = transects BC, AL vs. MN, MI, NY; Rocky Mountains = transects BC vs. AL, MN, MI, NY; Great Plains and Rocky Mountains = transects BC vs. AL vs. MN, MI, NY), and distance (transects BC vs. AL vs. MN vs. MI vs. NY) on population structure also showed very little variance for L. rubellus within (3.75%-5.81%) and among geographic populations (0.71%-2.86%), thereby rejecting all hypotheses for this species. However, L. terrestris generally showed a higher genetic structure with 11.22% variance among climate regions (H1) followed by distance among regions (9.46%).

Table 3.

Analyses of molecular variance (AMOVA) of Lumbricus rubellus and L. terrestris assigned a priori into populations separated by climate regions, geographic barriers (Great Plains and Rocky Mountains) or by geographic distance between transects. Molecular variance is given in percent.

tested barriers	L. rubellus									L. terrestris
	within sp	d.f.	within gg	d.f.	among gg	d.f.	FCT	p-value		within sp	d.f.	within gg	d.f.	among gg	d.f.	FCT	p-value
Climate (BC vs AL,MN vs MI,NY)	93.32	104	4.92	17	1.76	2	0.02	0.022		70.58	102	18.20	16	11.22	2	0.11	0.001
Great Plains & Rocky Mountains (BC vs AL vs MN,MI,NY)	92.47	104	4.81	17	2.71	2	0.03	0.011		71.96	102	20.00	16	8.04	2	0.08	0.015
Great Plains (BC,AL vs MN,MI,NY)	93.48	104	5.81	18	0.71	1	0.01	0.157		72.04	102	22.44	17	5.51	1	0.06	0.036
Rocky Mountains (BC vs AL,MN,MI,NY)	92.52	104	5.43	18	2.06	1	0.02	0.057		72.47	102	24.84	17	2.69	1	0.03	0.224
Transect (BC vs AL vs MN vs MI vs NY)	93.39	104	3.75	15	2.86	4	0.03	0.001		72.62	102	17.92	14	9.46	4	0.09	0.006

Importance of bioclimatic factors

The MRM showed contrasting results for the two earthworm species; the permutation test indicated that 23% and 5% of the variance were explained by climatic variables for L. rubellus and L. terrestris, respectively (Table S2). Lumbricus rubellus correlated significantly (p<0.002) with all tested bioclimatic factors except for the minimum temperature in the coldest month (BIO06; p=0.755) and seasonality of precipitation (BIO15; p=0.084). In L. terrestris, correlations generally were not significant (p>0.130), except for the minimum temperature of the coldest month (BIO06; p=0.022). We repeated the analysis with reduced datasets containing only the widespread clades 1 and 4 of L. rubellus and clade 2 of L. terrestris. In the reduced dataset, the variance explained by climatic factors decreased strongly for L. rubellus, but increased for L. terrestris (Table 4). Temperature and seasonal precipitation explained 4% of the genetic variance of clades 1 and 4 of L. rubellus, and temperature of the coldest and wettest month and precipitation explained 23% of the variance of clade 2 of L. terrestris.

Table 4.

Genetic variance of Lumbricus rubellus and L. terrestris explained by bioclimatic factors, for local and widespread genetic clades.

bioclimatic factor	local		widespread
	L. rubellus	L. terrestris	L. rubellus	L. terrestris
annual mean temp.	n/a	0.278	0.391	0.490
max. temp. warmest month	n/a	0.855	0.004**	0.099
min. temp. coldest month	n/a	0.027*	0.001***	0.297
mean temp. wettest month	n/a	0.003**	0.011*	0.580
mean temp. driest month	n/a	0.844	0.001***	0.107
annual precipitation	n/a	0.57	0.868	0.848
precipitation driest month	n/a	0.001***	0.192	0.319
precipitation seasonality	n/a	0.007**	0.017*	0.575
	r2=n/a,	r2=0.23,	r2=0.04,	r2=0.11,
	p=n/a	p=0.001	p=0.001	p=0.001

*P > 0.05, **0.001 < P < 0.01, ***P < 0.001; n/a not available due to small sample size; r² standardized coefficient of a regression analysis indicating the influence of the bioclimatic factors (independent variable) on genetic variance (dependent variable).

Other environmental factors

The abundance of earthworms tended to correlate with soil pH with contrasting results for the two species. Lumbricus rubellus occurred more often in acidic soils (pH≤5.5, R²=0.715; Fig. S4, Table S5), and L. terrestris was more abundant in neutral soils (pH>6, R²=0.865; Fig. S4, Table S5). Human population density was positively correlated with nucleotide diversity in L. terrestris (R²=0.701; Fig. 4, Table S5), i.e. transects in British Columbia, Alberta and Minnesota had the highest nucleotide diversities (≥2.25%) and densities of >100 population/km². Transects in Michigan and New York had the lowest nucleotide diversities (≤1.5 %) and densities of <70 population/km². Nucleotide diversity of L. rubellus was weakly negatively correlated with human population density (R²=0.197; Fig. 4, Table S5); however, the negative correlation likely was due to a single sampling location in New York (clade 3; Fig. 1a, b) that included distinct haplotypes (Table 2).

Figure 4.

Correlation of nucleotide diversity (in percent) of Lumbricus rubellus and L. terresris with human population density in the five transects.

Discussion

Genetic diversity

This study shows that northern North American populations of the two earthworm species L. rubellus and L. terrestris share the same genetic lineages with populations of their native range in Europe. However, genetic diversity is lower in North America than in Europe, which is typical for invasive species (Sakai et al., 2001; Allendorf & Lundquist, 2003; King et al., 2008; Donnelly et al., 2013; Donnelly et al., 2014; Giska et al., 2015). Consistent with studies in Europe, genetic diversity in L. rubellus was higher than in L. terrestris (King et al., 2008; Martinsson & Erséus, 2017), and intraspecific genetic distances of COI were comparable with those reported from Europe (King et al., 2008; James et al., 2010; Klarica et al., 2012).

In North America, common and widespread haplotypes dominated in both species, but genetic and geographic structure differed. Among populations of L. rubellus, two genetic lineages predominantly occurred in each of the studied sampling regions, except in New York. Interestingly, haplotypes belonged to common and widespread European lineages (Sechi, 2013), indicating that North American L. rubellus populations represent a genetic subset of Europe’s diversity. The distribution and abundance of genetic lineages of L. rubellus in Europe was potentially shaped by survival of cold tolerant populations during the Last Glacial Maximum (~25,000-13,000 years ago) in northern and central Europe (Sechi, 2013). Further, clades with deeply divergent mitochondrial lineages were considered to be descendants from pre-glacial refuge populations that adapted to local climate conditions (Sechi, 2013; Giska et al., 2015). Similar to L. rubellus, all mitochondrial and nuclear lineages of northern North American L. terrestris corresponded with haplotypes described from Europe.

The co-occurring pattern of omnipresent lineages of L. rubellus and L. terrestris across northern North America suggests a common origin and mode of dispersal for both species. In particular road constructions, traffic, logging, fishing and agriculture have been identified as main drivers of earthworm range expansion in North America (Marinissen & van den Bosh, 1992, Dymond et al., 1997; Casson et al., 2002; Holdsworth et al., 2007; Gundale et al., 2005; Cameron et al., 2008; Cameron and Bayne, 2009) and certainly also apply here. Human-mediated long-range dispersal by passive transport is more likely for L. rubellus, which lives in leaf litter near or on the soil surface than for soil-dwelling endogeic or anecic species (Terhivuo & Saura, 1997). However, the presence of locally occurring lineages in L. rubellus and the distinct genetic assembly of L. terrestris in Alberta and Minnesota indicate that additional factors affected the dispersal and introduction of these two earthworm species.

The relevance of bait abandonment for the distribution of L. rubellus is difficult to assess. This species has been commonly used as fishing bait (Reynolds, 1977), but was not sold in any bait shops we purchased earthworms from. However, dispersal via bait abandonment in the past cannot be excluded. In contrast, L. terrestris is the most commonly sold live fishing bait in northern North America today, and a large fraction of individuals from bait shops and the field shared identical or closely related haplotypes, indicating that bait abandonment contributes significantly to the spread of L. terrestris. Historically, earthworms sold as fishing baits were collected from fields and sold locally, but establishment of refrigerated warehouses by large distributors selling pre-packed baits nationwide might additionally contribute to long-distance spread of genetic diversity. This assumption is supported by a study at local scale in Calgary, Alberta, that demonstrated the genetic relatedness of bait and field-populations with fine resolution markers (Klein et al., 2017). Here, at large scale, bait haplotypes from Alberta and British Columbia in part did not match the haplotypes of nearby field populations, but rather field populations of far distant transects. However, bait cannot be the only source and vector for dispersal of L. terrestris, since the two most distant transects of Alberta and Minnesota contained three genetic clades that occurred nowhere else, indicating the existence of a distinct dispersal vector that connects these two transects.

Climate and dispersal barriers

Genetic variation among regions was very low for L. rubellus, and bioclimatic factors or dispersal barriers did not explain the distribution of common lineages, which agrees with its higher tolerance to frost (Sims & Gerard, 1999; Tiunov et al., 2006; Fisichelli et al., 2013). The ability of epigeic earthworms to quickly adapt to cold and fluctuating temperatures through behavioural and physiological changes (Holmstrup, 2003), and their persistence to perturbations, such as heavy metal pollution by fertilizers and intoxication by pesticides, are well known (Kruse & Barrett, 1985; Levine et al., 1989; Edwards & Bohlen, 1996). Although consecutive summer droughts can have strong effects on epigeic earthworms (Eggleton et al., 2009), drought resistance of cocoons allows persistence through drought periods (Holmstrup & Loeschcke, 2003).

In contrast to L. rubellus, genetic variance in the common lineages of L. terrestris in part was related to climate factors, in particular frost, drought, and seasonality. These results corresponded to findings that anecic earthworm species are negatively affected by prolonged drought periods, high frequency of freeze-thaw cycles and low soil moisture during their prime reproductive periods in spring and autumn (Sims & Gerard, 1999; Curry, 2004; Addison, 2009). Conform to these findings, the distinct genetic composition of populations in Alberta and Minnesota correlated with the continental climate in both transects. However, if the more severe frost and drought periods in these regions facilitated genetic diversity by continuous extinctions and reintroductions, or if only climatically pre-adapted lineages were able to establish viable populations in these areas needs to be investigated under controlled experimental conditions (Holmstrup, 2003).

Correlations with other factors affecting population structure

According to our dataset of additional environmental factors, soil pH affected the abundance of both earthworm species and was consistent with their contrasting ecological preferences. However, L. rubellus and L. terrestris occurred at sites with high and low pH, indicating that pH did not directly affect earthworm distribution but rather nucleotide diversity, as larger populations likely contain more genetic variance, and therefore pH potentially influences the genetic structure of earthworm populations. However, effects of pH on earthworms are generally difficult to explain, as individuals can withstand soil pH values outside their optimal range, and earthworm activity may also alter soil pH (Drouin et al., 2016).

Human population density as proxy for human activities and anthropogenic dispersal of earthworms increased genetic diversity of L. terrestris but not that of L. rubellus. This apparently contrasting pattern can be explained by the nearly ubiquitous occurrence of haplotypes in L. rubellus, undermining the detection of any correlation. However, L. terrestris seems to be closer associated with human activity and transport. Transects closest to urban centres, such as Vancouver, Calgary, and Minneapolis-St. Paul, and agricultural land (Alberta and Minnesota) were genetically more diverse and appeared in the latter case to be connected by a common source population or a common distribution mode. Our continental data are in accordance with previous findings at the regional scale, which showed decreasing genetic variance outside the urban area of Calgary (Klein et al., 2017). Further, agricultural land often holds high population densities of earthworms and was source of non-native earthworm introduction in New York (Suarez et al., 2006) and in Alberta, earthworms occur more often in the south-western region with high agricultural activity (Cameron & Bayne, 2009). High densities and resilience of L. terrestris to mechanical disturbance increases the probability of cocoons being transported on tires of trucks or agricultural machinery or in potted plants (Marinissen & van den Bosh, 1992; Suarez et al., 2006).

Conclusions

Genetic diversity and structure of the two invasive earthworm species L. rubellus and L. terrestris was homogenous across all regions indicating a dominant common dispersal vector and the ability to adjust to most environmental conditions in northern North America. However, L. terrestris was genetically more structured, and here its genetic variance positively correlated with harsh climatic conditions in central North America as well as with human activities, such as traffic and land use. In contrast to L. rubellus, this species is common in arable fields with frequent disturbances, and distinctness of genetic lineages occurring predominantly in transects of Alberta and Minnesota could be explained by their position at the edges of the North American corn-belt. Overall, we did not find any support for a continuous invasive front spreading from the east to the west coast. Genetic patterns indicate that both species have common long-distance distribution vector(s) or even common source. For L. terrestris, nation-wide bait distributors potentially play a major role as dispersal agent of field populations. In the past two decades, the globalisation of economy has changed infrastructure, intensity and range of traffic including commercial distribution of soil-related goods, and potentially will increase dispersal of L. rubellus and L. terrestris.

Our present study exemplifies how earthworms as belowground invaders with substantial differences in life history traits can be used to test broad questions in invasion ecology, such as the genetic underpinnings of successful invasion events, geographic and climatic dispersal barriers, as well as the human role in ecologically relevant invasions.

Biography

Biosketch

Andreas Klein is interested in phylogeography, invasion biology and population genetics of lumbricid earthworms, with particular interest in dispersal vectors and climate adaptation. This study is part of his PhD work at the universities of Göttingen and Leipzig on the spread of European earthworms in North America.