Habitat Structure Influences Genetic Differentiation in the Pitcher Plant Midge

Molecular Ecology (2012) 21, 223–236 doi: 10.1111/j.1365-294X.2011.05280.x

From broadscale patterns to ﬁne-scale processes: habitat structure inﬂuences genetic differentiation in the pitcher plant midge across multiple spatial scales

GORDANA RASIC and NUSHA KEYGHOBADI Department of Biology, Biological & Geological Sciences Building, University of Western Ontario, 1151 Richmond Street, London, Ontario, Canada N6A 5B7

Abstract The spatial scale at which samples are collected and analysed influences the inferences that can be drawn from landscape genetic studies. We examined genetic structure and its landscape correlates in the pitcher plant midge, Metriocnemus knabi, an inhabitant of the purple pitcher plant, Sarracenia purpurea, across several spatial scales that are naturally delimited by the midge’s habitat (leaf, plant, cluster of plants, bog and system of bogs). We analysed 11 microsatellite loci in 710 M. knabi larvae from two systems of bogs in Algonquin Provincial Park (Canada) and tested the hypotheses that variables related to habitat structure are associated with genetic differentiation in this midge. Up to 54% of variation in individual-based genetic distances at several scales was explained by broadscale landscape variables of bog size, pitcher plant density within bogs and connectivity of pitcher plant clusters. Our results indicate that oviposition behaviour of females at fine scales, as inferred from the spatial locations of full-sib larvae, and spatially limited gene flow at broad scales represent the important processes underlying observed genetic patterns in M. knabi. Broadscale landscape features (bog size and plant density) appear to influence oviposition behaviour of midges, which in turn influences the patterns of genetic differentiation observed at both fine and broad scales. Thus, we inferred linkages among genetic patterns, landscape patterns and ecological processes across spatial scales in M. knabi. Our results reinforce the value of exploring such links simultaneously across multiple spatial scales and landscapes when investigating genetic diversity within a species.

Keywords: distance-based redundancy analysis, genetic structure, landscape genetics, Metri- ocnemus knabi, Sarracenia purpurea, spatial scale Received 10 December 2010; revision received 17 July 2011; accepted 20 July 2011

explored in ecological studies for more than two dec- Introduction ades (Wiens 1989; Kotliar & Wiens 1990; Holling 1992; Population genetic data are increasingly analysed Levin 1992; Wu & Loucks 1995; Wagner & Fortin 2005). within an explicitly spatial framework as more and Landscape and population genetics, however, have only more studies, largely in the growing ﬁeld of landscape recently seen a strong and growing focus on spatial genetics, relate the spatial organization of genetic varia- scale questions (Anderson et al. 2010; Cushman & tion to underlying ecological processes and associated Landguth 2010; Storfer et al. 2010). landscape and environmental variables (Guillot 2009; The scale at which samples for genetic analysis are Storfer et al. 2010). Issues of scale surrounding the col- deﬁned and collected is critical in determining the pat- lection and interpretation of spatial data have been terns observed, and the range of processes about which inferences can be made in population genetic studies Correspondence: Gordana Rasic, Fax: 519-661-3935; (Anderson et al. 2010). Both the extent and the grain of E-mail: [email protected] a study are important, where the extent represents the

2011 Blackwell Publishing Ltd 224 G. RASIC and N. KEYGHOBADI total area of genetic sampling and analysis, while the The purple pitcher plant S. purpurea is found within grain represents the smallest (elementary) sampling unit acidic bogs throughout Eastern North America. It has (Anderson et al. 2010; Cushman & Landguth 2010). We developed carnivory as an adaptation to the poor nutri- cannot make reliable inferences on patterns and pro- ent environment. However, the plant’s leaves are not cesses beyond the extent of our study, nor detect any only deadly traps for different arthropods but also rep- elements of a pattern below the grain (Wiens 1989). In resent the exclusive breeding habitat for the larvae of gene flow analysis for example, study area size (extent) several insect species (Addicott 1974; Miller et al. 2002; should be larger than the area occupied by the popula- Buckley et al. 2010). For example, larvae of the pitcher tion of interest and larger than expected dispersal dis- plant midge, Metriocnemus knabi Coquillett 1904, are tances, while sampling grain should generally be usually found at the bottom of the leaf where they feed smaller than the average home-range size or dispersal on the decomposing prey of the plant. Multiple leaves distance of the study organism (Fortin & Dale 2005; are found in each plant, plants are distributed in easily Anderson et al. 2010). identifiable clusters within each bog, and bogs are eas- Population genetic patterns result from a potentially ily delineated in a landscape. These levels of habitat complex combination of evolutionary, behavioural, eco- patches (leaf, plant, cluster of plants, bog, system of logical and stochastic processes operating at different bogs) not only represent scales separated by a certain spatial and temporal scales (Balkenhol et al. 2009; spatial distance (‘distance scales’), but are also hierar- Anderson et al. 2010). Furthermore, ecological processes chically nested (‘nested scales’). Thus, the insects that and environmental variables can influence genetic vari- are commensal inhabitants (i.e. ‘inquilines’) of the pur- ation differentially at different spatial scales (Lee-Yaw ple pitcher plant represent an excellent natural system et al. 2009; Murphy et al. 2010). For example, in the bor- for ecological and genetic studies across scales. The nat- eal toad, Bufo boreas, growing season precipitation and ural features of the system remove the need for an arbi- slope temperature–moisture affect genetic connectivity trary decision on focal scales, because they offer easily of populations across multiple spatial scales, while hab- detectable habitat patches that are hierarchically nested itat permeability is only important at a fine scale (Mur- at several spatial scales. For this reason, the system has phy et al. 2010). Finally, the spatial scales of dispersal been used in landscape ecological studies to understand and other relevant processes affecting genetic variation how local interactions in the pitcher plant communities may not be known a priori, particularly in organisms (Trzcinski et al. 2005), colonization patterns (Trzcinski that are very small or display cryptic behaviours. Thus, et al. 2003), species distribution (Krawchuk & Taylor there is great value in conducting population and land- 2003) and community composition (Harvey & Miller scape genetic analyses such that multiple spatial scales 1996; Buckley et al. 2010) vary across scales. of sampling are included (Diggle & Ribeiro 2007; Sch- Our first objective in this study was to examine popu- wartz & McKelvey 2009). lation genetic structure of one of the pitcher plant’s com- Although there are many reports of genetic structure mensal inhabitants, the pitcher plant midge M. knabi, across more than one spatial scale, the majority of studies across several, nested scales. By considering samples of include only up to three scales. For example, genetic midge larvae aggregated at each scale in the spatial hier- diversity was examined at: (i) fine, population and regio- archy (leaf, plant, cluster of plants, bog, system of bogs), nal scales (riparian and mountain) in Manchurian ash we essentially changed the grain of sampling and analy- Fraxinus mandshurica (Hu et al. 2010); (ii) rivers, among sis while keeping a constant extent that is largely rela- rivers and among regions on an island in the riparian tive to the expected dispersal ability of this species plant Ainsliaea faurieana (Mitzui et al. 2010); and (iii) pop- (Krawchuk & Taylor 2003). Our second objective was to ulation, watershed and drainage scales in steelhead On- test explicit hypotheses about the effects of landscape corhynchus mykiss (Nielsen et al. 2009). While the scales of variables on genetic structure of the midge across scales. analysis in these examples reflect natural hierarchies of Specifically, results obtained under our first objective spatial organization, such as river–watershed–drainage, suggested that broader-scale landscape variables related in many other studies, the scales of analysis are appar- to habitat amount and isolation may influence spatial ently arbitrary or based primarily on an anthropocentric patterns of genetic variation observed at both fine (leaf, perception of nature. In some cases, even political bound- plant) and broad (cluster, bog) scales. As such, we used aries may be used to define scales of sampling (Blanquer distance-based redundancy analysis to explicitly test the & Uriz 2010; Gonçalves da Silva et al. 2010). Here, we hypothesis that bog size, plant density, or isolation of take advantage of a unique study system associated with clusters influence patterns of genetic structure at a range commensal inhabitants of the purple pitcher plant, Sarra- of scales. Although there are many potential landscape cenia purpurea L., to examine patterns of genetic variation correlates of genetic structure (e.g. Murphy et al. 2010), across multiple, objectively defined, nested spatial scales. we focused on variables related to habitat amount, patch

2011 Blackwell Publishing Ltd GENETIC STRUCTURE OF M. KNABI ACROSS SCALES 225 size and patch isolation because these factors were sug- per plant. Two separate systems of bogs, located 26 km gested by our analyses of genetic variation across scales apart, were sampled in this way. The locations of all in M. knabi (under first objective) and because they have sampled plants were recorded to within 0.5 m using a previously been shown to influence pitcher plant midge high-accuracy GPS receiver (Trimble GeoXH, Sunny- larval densities (Krawchuk & Taylor 2003). vale, CA, USA). Larvae were sorted with forceps in a Petri dish and placed in absolute ethanol at )20 C until DNA extrac- Materials and methods tion. We used the DNeasy tissue kit (Qiagen, German- town, MD, USA) to extract genomic DNA from each Study area and species individual and genotyped them at 12 microsatellite loci Our study sites (bogs) were located in Algonquin Pro- (GenBank accession numbers FJ665262–FJ665273) devel- vincial Park, Ontario, Canada (UTM: 17N 687337E oped specifically for this species (Rasic et al. 2009). We 5046853N). The park is in an area of transition between followed amplification protocols and fragment analysis northern coniferous forest and southern deciduous for- methods described by Rasic et al. (2009), with the modi- est. These forests are dominated, respectively, by: (i) fication of performing multiplexed PCRs (details on white and red pine, poplar and white birch, and (ii) PCR conditions and primer concentrations found in sugar maple, American beech hemlock, white pine, red Table S2, Supporting information). oak, white oak and yellow birch. Bogs are found within In all analyses, we included only individuals with this forest matrix, and many of them contain Sarracenia complete genotypes. Our final data set consisted of 12 purpurea and its associated commensal arthropod inhab- loci scored for 740 individuals sampled from 24 clusters itants. Bogs represent peat-covered wetlands in which (clusters 1–24) in 8 bogs (SB, RSB, Bab, Min; WR, DL, the vegetation shows the effects of a high water table ML, BB) grouped in two systems (SYS1 & SYS2; and a general lack of nutrients. Owing to poor drainage Table S1, Supporting information). The average number and the decay of plant material, the surface water of of genotyped larvae per leaf (3.5) did not significantly bogs is strongly acidic. Dominated by sphagnum differ between the systems (t = 0.843, P = 0.397). mosses (peat) and heath shrubs (leather leaf, labrador tea, cranberries), the bogs also contain tamarack and Habitat mapping and landscape variables black spruce (Tiner 1999). The focal species in this study, Metriocnemus knabi We determined the density and distribution of pitcher (Diptera: Chironomidae), is expected to have one gener- plants within each bog by recording the number of ation per year at this latitude (Rango 1999). The midge plants within a 2-m-radius circle positioned every 10 m overwinters as a larva in the leaves of S. purpurea. along a linear transect extending from one edge of the Pupation, adult emergence, mating and oviposition bog to the other. The entire bog area was covered by occur during late spring and summer (Heard 1994; Ran- such transects, separated from each other by 5 m. We go 1999; Krawchuk & Taylor 2003). Adult midges exhi- used these point recordings and their UTM coordinates bit very cryptic behaviour and have body length of for the spherical kriging procedure performed in Arc- only 3 mm (Krawchuk & Taylor 2003), which makes it GIS 9.3 (ESRI, Redlands, CA, USA). From the resulting unfeasible to sample them in that life stage. Larvae on raster maps with the predicted plant distribution, we the other hand can be readily sampled from within estimated landscape variables related to habitat patch pitchers, which represent the exact spatial locations of density: bog plant density (average number of maternal oviposition. plants ⁄ m2 in each bog) and cluster plant density (average number of plants ⁄ m2 in each sampled cluster). Cluster connectivity that measures the connectivity of Sampling and genotyping each clusterP to all other clusters in a system was calcu- ) We sampled second instar larvae in August 2009 at five lated as exp( dij), where dij is a pairwise Euclidean nested spatial scales: leaf, plant, cluster, bog and system distance in km between centres of clusters j and i. Bog of bogs (Fig. 1). We balanced sampling so that (i) each area (m2) was measured in ArcGIS 9.3 from 30-m reso- system contained the four neighbouring bogs in a land- lution vector maps (Wetland class from Land Cover, scape (0.2–7.0 km apart in a mixed forest matrix), (ii) Circa 2000, AAFC) and Google Earth images. Bog plant within every bog, we randomly selected three clusters density and bog size were estimated for all eight bogs containing ten plants, (iii) we randomly chose three in the two systems, and cluster plant density and clus- plants within each cluster and (iv) we pipetted all lar- ter connectivity were estimated for all 24 clusters vae out of the bottom of three randomly selected leaves (Table S1, Supporting information).

2011 Blackwell Publishing Ltd 226 G. RASIC and N. KEYGHOBADI

N (a)

Algonquin Park

SYS1 SYS2 N Ontario

km 0 5 10 20 30 40 km 0 100 200 400

(b) N SYS2 N Min SYS1

WR BB Bab

DL ML

km km SB RSB 0 0.5 1 1.5 0 0.5 1 1.5

Fig. 1 Sampling locations and design. (a) The location of the two bog systems (SYS1 and SYS2) in Algonquin Park, Ontario, Canada. (b) The four neighbouring bogs were sampled in each system. (c) Three clusters (shown as circles) within each bog were sampled. Arrows represent the Euclidean distances between the clusters’ centres. (d) Three randomly chosen plants within each cluster were sampled. (e) Sarracenia purpurea. Three randomly chosen leaves within each plant were sampled.

Microsatellite variability and summary statistics Hierarchical AMOVA Genotypic data were initially tested for the presence of We investigated how genetic variation was partitioned null alleles and other scoring errors using MICRO- across all spatial scales using the hierarchical analysis CHECKER version 2.2.3 (van Oosterhout et al. 2004). of molecular variance (hierarchical AMOVA). The model Standard population genetic summary statistics were properties of hierarchical AMOVA correspond well to the generated, and tests for Hardy–Weinberg and linkage biological structural properties of this system (i.e. equilibrium performed in FSTAT version 2.9.3.2 (Gou- nested hierarchy). We employed the package HIERF- det 1995) for all scales of sampling; for simplicity, we STAT for the statistical software R (Goudet 2005), present results for the bog scale only. which computes variance components and moment

2011 Blackwell Publishing Ltd GENETIC STRUCTURE OF M. KNABI ACROSS SCALES 227 estimators of hierarchical F-statistics for any number of A linear pairwise geographic distance matrix was calcu- nested scales. We used 1000 randomizations to deter- lated as the Euclidean distance between UTM coordi- mine the statistical significance of genetic differentiation nates of sampled larvae. The spatial autocorrelation at a given scale (leaf, plant, cluster, bog, system), while coefficient (r) was calculated for several distance classes controlling for the effects at the other scales. For exam- that corresponded to the comparison of individuals at ple, testing for significant differences among plants the following scales: within a cluster, between the clus- (nested within clusters and above leaves) implies per- ters (within a bog) and among bogs. The statistical sig- mutating whole units of the scale ‘leaf’ among plants, nificance of the autocorrelation coefficient (r) was tested but keeping them within units defined by the scale with 999 permutations. To visualize the pattern of ‘cluster’. genetic structuring for bog and cluster samples, we also conducted principal coordinate analysis (PCoA) on the same mean genotypic distance values using GenAlEx Relationships between individuals across spatial scales ver. 6 (Peakall & Smouse 2006). Given that we defined samples starting at a very fine spatial scale (within leaves), we wanted to investigate Distance-based redundancy analysis the percentage of full-sib pairs sampled within leaves and more generally within each of the higher scales of We tested the effects of four landscape variables, (i) bog sampling. Although most studies try to avoid the inclu- size, (ii) average plant density within a bog, (iii) cluster sion of family groups, we were specifically interested in connectivity and (iv) average plant density within a how this variable would change with scale of sampling. cluster, on genetic differentiation of pitcher plant midge We used the data on the distributions of full-sib pairs larvae from different leaves, plants and clusters. To this over increasing aggregation scales to infer oviposition end, we used distance-based redundancy analysis behaviour of midge females. Maximum-likelihood esti- (dbRDA), a multivariate method that assesses the influ- mates of pairwise relationships between individuals ence of landscape data measured at distinct points on were obtained in ML-RELATE (Kalinowski et al. 2006). values in a dissimilarity (in this case, mean genotypic Relationships were tested between the following catego- distance) matrix (Legendre & Anderson 1999). Because ries: full-sibs (FS), half-sibs (HS), unrelated (U) and par- we had only four bogs per system, there was insuffi- ent–offspring (PO). Given that the PO relationship is cient power to test the effects of landscape variables at not possible between larvae collected in the same year, the bog scale. we treated those cases as FS (as in Savage et al. 2010). A matrix of mean genotypic distance values between A confidence set for the relationship between each pair individuals (Smouse & Peakall 1999) was calculated for of individuals was generated with 1000 randomizations leaves, plants or clusters, and each was used separately at the 99% confidence level. Every putative full-sib rela- in DISTLM forward (McArdle & Anderson 2001; Ander- tionship was then tested against each of the alternative son 2003) with all four predictor variables (bog size, plant relationships indicated by the confidence sets using density within bog, cluster connectivity and plant density likelihood ratio tests with 1000 simulated random geno- within a cluster) entered in each analysis. The multilocus type pairs (Kalinowski et al. 2006). Only full-sib pairs interindividual genetic distance measure of Smouse & that had significantly higher likelihood than the alterna- Peakall (1999) is commonly used in spatial autocorrela- tive relationships were further considered. The percent- tion analysis, does not require estimation of allele fre- age of full-sib pairs thus detected was plotted for each quencies from small samples and does not assume any level in the spatial hierarchy: within a leaf, between particular microevolutionary processes. Marginal tests leaves within a plant, between plants within a cluster (i.e. fitting of each variable individually, ignoring other and between clusters within a bog. It should be noted variables) were followed by the forward selection proce- that pairwise comparisons at lower levels were thus dure with conditional tests (i.e. fitting each variable one a removed as we analysed progressively higher levels. time, conditional on the variables already included in the model). The significance of the marginal tests was determined with 9999 permutations of raw data, while for the Spatial autocorrelation and PCoA conditional test, the program uses permutation of residu- To further examine spatial genetic structure, we als under the reduced model (Anderson 2003). Tests were employed spatial autocorrelation analysis for distance conducted separately for genetic distances measured at classes with equal sample sizes in GenAlEx ver. 6 each scale (between leaves, plants or clusters), allowing (Peakall & Smouse 2006). The program calculates a us to assess which landscape variables were important at matrix of mean genotypic distance values between all any given spatial scale. Separate analyses were con- pairs of individuals following Smouse & Peakall (1999). ducted within each system of bogs.

2011 Blackwell Publishing Ltd 228 G. RASIC and N. KEYGHOBADI

(F = 0.002, P = 0.014). Within both systems, Results system ⁄ Total variability was similarly partitioned among higher spatial scales (bog, cluster, plant). A difference between the Microsatellite variability two systems was detected at the leaf scale: in SYS1, it After the initial testing for scoring errors, we excluded contributed only 0.14% to the overall genetic variation, locus MK01 from further analyses owing to potential whereas in SYS2, this scale made up 2.9% of the total presence of null alleles (with estimated frequencies variation (Table 1). Consequently, significant structur- within bogs between 0.1 and 0.21). Within bogs, the ing was present at every hierarchical scale in SYS2, number of alleles per locus ranged from 2 to 20 and the whereas leaves were not structured within plants in average allelic richness ranged from 5 to 7.5. Observed SYS1 (Table 2). heterozygosity was significantly higher than expected in Overall, most of the genetic variation was contained two bogs in each system (P < 0.05; i.e. deviations from within individuals (error term), which is common for the Hardy–Weinberg proportions were owing to excess microsatellite markers (Hedrick 1999). Negative values heterozygosity). On average, six pairs of loci in system at the individual term (Table 1), which is an equivalent 1 (SYS1) bogs and 8.3 in system 2 (SYS2) bogs exhibited to an individual inbreeding coefficient, imply that indi- significant linkage disequilibrium. The results did not viduals were highly heterozygous. Consequently, hier- indicate consistent associations between any loci. archical F-statistics for individuals grouped at any of the higher scales (Findividual ⁄ level(i), FIS analogues) were negative as well (the last column in Table 2) and indi- Hierarchical structuring cated excess heterozygosity in groups of individuals A hierarchical AMOVA revealed that the two bog systems aggregated at any scale. were significantly differentiated from each other

Table 1 Hierarchical analysis of molecular variance in Metriocnemus knabi. The output from HIERFSTAT (Goudet 2005) contains overall variance components and percentage (%) of variation at each scale. SYS indicates the variance between systems. Results for all lower scales are shown separately for the two systems (SYS1 and SYS2)

Scale

SYS Bog Cluster Plant Leaf Individual Error

Variance components 0.014 SYS1 0.130 0.101 0.103 0.008 )0.333 5.838 SYS2 0.116 0.119 0.124 0.171 )0.423 5.784 % Variation 0.24 SYS1 2.22 1.73 1.76 0.14 )5.70 99.84 SYS2 1.97 2.02 2.10 2.90 )7.18 98.18

Table 2 Matrix of hierarchical F-statistics computed in HIERFSTAT (Goudet 2005). Each value in the table indicates differentiation among scales of the corresponding ‘column’ within scales of the corresponding ‘row’. Results are shown separately for each of the two systems. For example, the F-statistics measuring differentiation among clusters within bogs of system1 is 0.013. The most important values are found in the last line above the empty cells and are boxed for emphasis. Values within these boxes that are signiﬁ- cantly greater than zero are shown bold. The signiﬁcance of genetic differentiation at each scale (while controlling for the effects at all other scales) was determined using 1000 permutations

2011 Blackwell Publishing Ltd GENETIC STRUCTURE OF M. KNABI ACROSS SCALES 229

and 12%, respectively) and was omitted from the graph SYS1 16 as an outlier. Additionally, WR bog in SYS2 had a high number of FS at both the leaf and the plant scale 14 (10.3% and 8.4%, respectively). The bogs and clusters 12 in which we found these high levels of full-sib pairs are either characterized by low plant density (Bab and WR 10 bogs) or are distant from the main bog area (cluster 12 in Min bog), respectively (Table S1, Supporting infor- (%) 8

FS mation). 6 *

4 Spatial autocorrelation Bab 2 Min Correlograms in both systems were significant 0 RSB Bog (P = 0.01; Fig. 3), and all distance classes contained wl bl SB approximately 3000 comparisons each. Significant posi- bp bc tive spatial autocorrelation was detected at distance Scale classes within a bog (i.e. among plants and clusters) in both systems. r values remained significant and positive SYS2 at smaller between-bog distances (up to 1.5 km) in 16 SYS2. However, there was no significant positive auto- 14 correlation among bog samples in SYS1.

12 Principal coordinate analysis 10 When individuals were grouped by bog, the first three (%) 8 principal coordinate axes explained 78% of the genetic FS 6 variation among samples (Fig. 4a). We detected two highly differentiated bogs: Bab bog in SYS1 and WR in 4 WR SYS2. When individuals were grouped according to the 2 BB pitcher plant clusters from which they were sampled, 0 ML 68.7% of variation was explained and cluster 12 from Bog wl Min bog in SYS1 was also seen to be highly differenti- bl DL bp bc ated (Fig. 4b). These highly differentiated bogs and Scale clusters are the same ones in which we detected a high proportion of full-sib pairs within leaves and plants, Fig. 2 Percentage of Metriocnemus knabi full-sib pairs (FS) sampled at different spatial scales: within a leaf (wl), between leaves and they either have low abundance of pitcher plants within a plant (bl), between plants in the same cluster (bp) and (clusters 7–9 in Bab and 13–15 in WR) or are spatially between clusters in the same bog (bc). *For Min bog, cluster 12 isolated (cluster 12 in Min). (outlier) was excluded from the percentage calculations. Distance-based redundancy analysis First, at each scale (leaf, plant, cluster), the effect of each Full-sib pairs across scales predictor variable (bog size, bog plant density, cluster The pattern of full-sib distribution was quite different plant density, cluster connectivity) was tested individu- between the two bog systems (Fig. 2). In SYS1, the per- ally. In both systems and across all scales, the strongest centage of pairs of individuals that were FS was around predictor of genetic distance was bog plant density 0.2–4.8% across all scales, whereas in SYS2, significantly (P < 0.01, Table 3). In SYS1, this variable explained more full-sib pairs were found within a single leaf (5– 12.1%, 22.3% and 33.5% of variation in genetic dis- 15%) than at any higher spatial level (0.3–4.4%). There tances at the leaf, plant and cluster scales, respectively. were two exceptions to this general pattern in SYS1: (i) The amount of variation explained was similar in SYS2, Bab bog (clusters 7–9) showed higher number of FS in a going from 7.9% at the leaf, 23.5% at the plant, to single plant (12%) than in higher levels (0.3–3%) and 29.1% at the cluster scale. Bog size and cluster connec- (ii) cluster 12 in Min bog showed extremely high per- tivity were significantly associated with genetic patterns centage of full-sib pairs at the leaf and plant scale (53% at all scales in SYS1, while cluster plant density was not

2011 Blackwell Publishing Ltd 230 G. RASIC and N. KEYGHOBADI

Spatial autocorrelation analysis SYS1 0.12 wc bc bb

0.08

0.04 r 0.00

–0.04

–0.08

7 37 119 175 742 767 937 1440 1589 5413 5444 5974 6051 6366 6520 6838 Distance class (m)

Spatial autocorrelation analysis 0.12 SYS2 wc bc bb

0.08

0.04 r 0.00

–0.04

–0.08

9 155 181 740 852 1438 1471 1595 1685 1801 1970 2143 2278 2384 Distance class (m)

Fig. 3 Spatial genetic autocorrelograms showing average correlation coefficients between pairs of Metriocnemus knabi individuals (r) plotted against geographic distance classes in SYS1 (upper) and SYS2 (lower). Vertical dotted lines delineate distance classes contained within scales in the spatial hierarchy, namely: within a cluster (wc), between clusters in a bog (bc) and between bogs within a system (bb). Horizontal dashed lines represent critical values under the null hypothesis that genotypes are randomly distributed across a landscape (a = 0.05). Error bars represent 95% confidence intervals around each mean correlation coefficient. significant at any scale. In SYS2, cluster connectivity and bog size were only significant in the model at the and cluster plant density were significantly associated leaf scale in SYS2. with genetic patterns at plant and leaf scales (P < 0.05), but not at the cluster scale. Bog size was a marginally Discussion significant explanatory variable (P = 0.048) only at the leaf scale in SYS2. Analyses of genetic variation across multiple nested Sequential tests showed that bog size and bog plant scales revealed complex genetic structuring in the density jointly explained between 18.4% and 53.4% of pitcher plant midge. Comparing two systems of bogs, variation in genetic distances across scales in SYS1 the partitioning of variation was similar at the higher (Table 4), and cluster connectivity was only significant scales (among bogs, clusters and plants), but different in the model at the leaf scale. Two jointly significant at the finest scale (among leaves within plants). The factors in SYS2 were bog plant density and cluster con- percentage of full-sib pairs across hierarchical scales nectivity, explaining between 11.2% and 32.6% of varia- was quite different between the two bog systems, and a tion at the leaf and plant scales. Cluster plant density high proportion of FS were found within leaves and

2011 Blackwell Publishing Ltd GENETIC STRUCTURE OF M. KNABI ACROSS SCALES 231

not in SYS1, we observed a high proportion of FS within leaves of the same plant in SYS2 but not in SYS1 Bab WR (Fig. 2). The distribution of FS is a potential proxy for SB BB oviposition behaviour, given that chironomid females rarely mate with multiple males (Armitage et al. 1995). These results suggest that females of M. knabi in SYS1, RSB ML particularly in small bogs, leave smaller numbers of PCoA2 (21.62%) Min DL eggs within a single leaf and tend to distribute their eggs more equally across plants, while in SYS2, they tend to leave a large number of eggs (clutches) within a PCoA1 (42.84%) single chosen leaf. The high proportion of full-sib pairs found within

cl14 leaves and plants of clusters that were either highly iso- cl7 lated (cluster 12 in Min bog) or occured in bogs with low cl9 cl23 cl15 cl8 cl22 plant density (Bab and WR bogs) suggest a role of habitat cl2 cl3 cl1 cl24 patch isolation and habitat amount at these higher spatial cl5 cl19 cl21 cl6 scales in influencing the fine-scale (i.e. among leaves) ovi- cl4 cl11cl18 cl10 position decisions made by females. There are a number cl12 cl17 PCoA2 (20.27%) cl16 cl20 cl13 of reports of directional flight of the chironomid females prior to oviposition, but how they are able to select the correct site is not understood (Oliver 1971). The females PCoA1 (32.87%) of the pitcher plant midge appear to respond to leaf size Fig. 4 Plots of Eigen values for the first two components of (Paterson & Cameron 1982; Nastase et al. 1995), but ovi- the principal coordinate analysis (PCoA) performed on genetic position decisions may occur at several spatial scales distance matrix from bog (upper) and cluster (lower) samples (Trzcinski et al. 2003), as supported by our findings. of Metriocnemus knabi. n, SYS1; , SYS2. Given that this species is an extreme specialist with respect to oviposition sites, as the pitcher plant leaves plants that occur in isolated or low plant density represent the exclusive habitat for the larval develop- patches. Positive local spatial structure extended among ment, it would be highly advantageous to make active bogs in SYS2, but not in SYS1. Overall, dbRDA showed decisions about oviposition based on different character- that across several scales, a significant portion of genetic istics of the larval habitat at several spatial scales. structure in Metriocnemus knabi can be explained by bog size, bog plant density and cluster connectivity. Linking patterns and processes across scales In PCoA analyses, we found that the same clusters that Genetic structure across scales in M. knabi were characterized by a high proportion of full-sib pairs Overall, we detected significant structuring across multi- within leaves and plants (cluster 12 in Min bog, clusters ple nested scales, going from the system to the leaf scale. 7–9 in Bab bog and clusters 13–15 in WR bog) were The two bog systems were significantly differentiated identified as being highly differentiated from other clus-

(Fsystem ⁄ Total = 0.002, P = 0.014), which was expected as ters. This result is not surprising given that the inclu- they are 26 km apart. This small F value does not mean sion of highly related individuals within samples high genetic connectivity at this distance, but simply that inflates measures of genetic differentiation (Allendorf & the vast majority of the variation is contained within Phelps 1981; Anderson & Dunham 2008; Goldberg & lower scales in the hierarchy. In both systems, the parti- Waits 2010). However, this result is important because tioning of variability was similar at the bog, cluster and it indicates that the process of female oviposition occur- plant scales (Table 1), and genetic structuring was signif- ring at the finest scales (among leaves and plants) inter- icant at all of them (Table 2). A difference between the acts with the sampling design (in this case, collection of systems was revealed at the leaf scale, where structuring juveniles at fine spatial scales and before dispersal among leaves within plants was significant in SYS2 but events) to affect the output from the common popula- not in SYS1. A cruder grain in sampling would have tion genetic analyses conducted at broader scales. Our missed this component of the overall genetic pattern in results thus highlight that the scale of sampling, relative this species. to the scales of ecological ⁄ evolutionary processes, influ- Consistent with the finding that samples from leaves ences the conclusions that can be drawn in population within a plant were significantly different in SYS2 but and landscape genetic studies (Anderson et al. 2010).

2011 Blackwell Publishing Ltd 232 G. RASIC and N. KEYGHOBADI

Table 3 Distance-based redundancy analysis of genetic distances among Metriocnemus knabi samples performed at each scale (cluster, plant, leaf). Each predictor variable (bog size, bog plant density, cluster plant density, cluster connectivity) was tested separately. Signiﬁcant P values are bolded. ‘% Variation’ indicates the amount of variation in genetic distances explained by a particular variable

Marginal test Pseudo-F P % Variation

Scale Variable SYS1 SYS2 SYS1 SYS2 SYS1 SYS2

Cluster Bog size 2.75 0.44 0.023 0.863 21.6 4.2 Plant 6.50 1.00 0.001 0.428 16.5 2.8 Leaf 7.74 2.52 0.001 0.048 6.9 2.4 Cluster Bog plant density 5.04 4.10 0.001 0.005 33.5 29.1 Plant 9.48 10.47 0.001 0.001 22.3 23.5 Leaf 14.40 8.89 0.001 0.001 12.1 7.9 Cluster Cluster plant density 0.62 2.35 0.701 0.072 5.9 12.9 Plant 1.12 5.37 0.368 0.005 3.3 13.6 Leaf 1.45 4.46 0.261 0.003 1.4 4.1 Cluster Cluster connectivity 2.35 1.22 0.046 0.301 19.1 10.9 Plant 5.65 2.83 0.001 0.024 14.6 7.7 Leaf 6.38 2.77 0.001 0.022 5.7 2.6

Table 4 Forward selection procedure in distance-based redundancy analysis of genetic distances among Metriocnemus knabi samples performed at each scale (cluster, plant, leaf). Only signiﬁcant values in a combined model are reported. Bolded cummulative % indicates the total variation explained by combined variables in sequential tests. The top-down sequence of variables corresponds to the sequence indicated by the forward selection procedure, with the exception of the leaf scale in SYS2, where cluster connectivity pre- ceded bog size

Sequential test Pseudo-F P Cummulative %

Scale Variable SYS1 SYS2 SYS1 SYS2 SYS1 SYS2

Cluster Bog plant density 5.04 4.11 0.001 0.005 33.5 29.1 Bog size 3.84 — 0.006 — 53.4 — Plant Bog plant density 9.48 10.47 0.001 0.001 22.3 23.5 Bog size 8.14 — 0.001 — 38.1 — Cluster connectivity — 4.43 — 0.005 — 32.6 Leaf Bog plant density 14.40 8.89 0.001 0.001 12.1 7.9 Bog size 8.02 5.74 0.001 0.001 18.4 15.9 Cluster connectivity 3.74 3.88 0.021 0.006 21.2 11.2 Cluster plant density — 3.41 — 0.014 — 18.7

These findings also suggest, in this system, linkages tion was for cluster plant density in SYS1, which was among processes and patterns at different spatial scales. not significant at any scale. In the sequential tests, going Specifically, we hypothesize that the isolation and from the cluster to the plant and to the leaf scale, pro- amount of habitat at cluster and bog scales (broader- gressively more landscape variables were included in scale landscape patterns) lead females to aggregate their the significant models. Pitcher plant density within bogs eggs within leaves (fine-scale ecological process), which exhibited the strongest effect on genetic structure exaggerates genetic differentiation of larvae not only among leaves, plants and clusters (Table 3). When com- among leaves but also at higher scales (fine-scale to pared to other tested variables, it explained the largest broadscale genetic patterns). Based on this hypothesis, proportion of variation in genetic distances (between we would expect to see significant effects of broadscale 7.9% and 33.5%), and its effect was consistent in both landscape variables on genetic differentiation at all systems and across scales. Even when the other predic- scales, but more so at the finer scales. This is indeed tor variables were accounted for in the sequential mod- what our dbRDA analyses revealed. In the marginal els (Table 4), average plant density in a bog had a tests in both systems, significant effects were seen either pronounced effect on genetic distances among larval across all scales or only at finer scales. The only excep- samples, supporting the hypothesis that females are

2011 Blackwell Publishing Ltd GENETIC STRUCTURE OF M. KNABI ACROSS SCALES 233 more likely to aggregate eggs locally under conditions among European wolf populations can be attributed to of low plant density at the bog scale. vegetation types. Cluster connectivity measures Euclidean distance among sampling points in the entire landscape (within Dispersal and isolation by distance and among bogs) and therefore accounts for the importance of overall physical distance on the pattern of Ecological studies have made inferences about dispersal genetic distances. If lower connectivity of clusters is of pitcher plant midges based on spatial patterns of lar- associated with greater genetic distances between sam- val abundance (Miner & Taylor 2002; Krawchuk & Tay- ples, as we observed, this is analogous to isolation by lor 2003). These studies suggest that M. knabi distance and could simply indicate that spatially limited individuals are weak fliers, aggregate around plants dispersal of adult midges plays a role in determining and clusters and rarely move among bogs. Significant genetic patterns among larval samples. However, signif- genetic structure at the plant and cluster scales in our icant effects of cluster connectivity on genetic differenti- study largely supports these previous ecological infer- ation of midge larvae were observed at the finest scales, ences. However, our study also indicates that gene flow for leaves and plants, particularly in the sequential can occur among close bogs, as seen in our spatial auto- tests. Thus, it is likely that the significant influence of correlation analyses. Significant positive genetic correla- cluster connectivity on genetic differentiation is medi- tion was detected among bogs in SYS2, where distances ated to a large extent by the effects on female oviposi- between some bogs are relatively small, but not in SYS1 tion, as hypothesized, which should be observable at where the bogs are more distant from each other. Fur- the finest spatial scales. In contrast, if the influence of thermore, significant positive autocorrelation at short cluster connectivity was mediated simply by limited distances coupled with the significant negative autocor- dispersal of adults, we would expect to observe more relation at long distances is a pattern consistent with significant effects at the broader spatial scales. isolation by distance (Sokal & Oden 1991). Thus, in Bog size was the second most important landscape addition to the effects of female oviposition behaviour factor explaining genetic distances in SYS1 but was occurring at fine scales, the balance between restricted marginally significant only at the finest scale SYS2. This gene flow and genetic drift may also contribute to difference can be explained by the characteristics of the genetic structuring at broad spatial scales (i.e. among two systems: SYS2 contains only large bogs, whereas bogs). SYS1 contains bogs more variable in size (Table S1, Supporting information). This provided more size clas- High individual and group heterozygosity ses for the regression analysis in SYS1, making the pattern detectable. Limited variation in a predictor variable Excess heterozygosity (i.e. negative FIS) at neutral loci is reduces power in any analysis, and Short Bull et al. not often found in animal populations and is somewhat (2011), in their gene flow analysis in American black surprising for an insect that is considered to be a weak bears, found that landscape features had to be highly flier, dependent on a highly specific patchy habitat for variable in order to be supported in landscape genetic its development. Non-random mating, specifically out- models. Our results further reinforce the conclusion that breeding, is a frequent explanation for excess heterozy- landscape genetic studies should ideally incorporate gosity. However, looking at the individual inbreeding large variation in landscape attributes. The fact that the coefficient (calculated following Ritland 1996), we did larger bogs in SYS1 exhibited a pattern of full-sib distri- not observe significantly different values in the individ- bution similar to bogs in SYS2 (Fig. 2) points towards a uals we sampled as compared to individuals simulated critical bog size at which female oviposition behaviour under a random-mating scenario using our observed changes. allele frequencies (data not shown). Outbreeding is Overall, in SYS1, up to 54% of the variation in therefore not a likely explanation for the highly genetic distances was jointly explained by broadscale observed heterozygosities in our study. The excess het- variables in the dbRDA: bog size and plant density erozygosity we observed, coupled with significant dif- within bogs. The joint influence of bog plant density ferentiation among samples, is actually very similar to and cluster connectivity explains up to 33% of varia- the patterns observed in social mammals as well as tion across spatial scales in SYS2. The predictive power among communal hibernacula of the timber rattlesnake of only two habitat variables in each system is high (Anderson 2010). As pointed out by Anderson (2010), and comparable to the results from the study by Pilot these patterns counter the expectation of reduced het- et al. (2006). Their sequential tests in dbRDA revealed erozygosity within genetic ‘demes’ as a result of that after accounting for geographic distance, 53% of restricted gene flow (Wright 1969). However, such pat- variation in Nei’s genetic distance at microsatellite loci terns may be expected to arise when there is spatial

2011 Blackwell Publishing Ltd 234 G. RASIC and N. KEYGHOBADI clustering of individuals at some life history stage, in spatial scales (e.g. as observed through PCoA). Overall, combination with either sex-biased dispersal or a lim- the results of our study reinforce the value of consider- ited number of breeding adults (Anderson 2010), both ing patterns and processes across multiple spatial scales of which can lead to excess observed heterozygosity and in multiple landscapes when investigating genetic within samples (Prout 1981; Balloux 2004). diversity within a species.

Sampling considerations Acknowledgements It is important to note a caveat with respect to our hier- The authors would like to thank the Ontario Ministry of Natural archical sampling design: progressing from the leaf Resources and Harkness Laboratory of Fisheries Research for scale up to the bog and system scales, the size of each enabling research within the provincial park. Curtis Irvine, Jen- na Donald, Kevin Schreiber and Katelyn Weaver assisted sample increases while the total number of samples greatly in field work. Special appreciation goes to Igor Filipovic decreases. This could affect the power of analyses con- for programming the H-W simulation software, Jenna Donald ducted at each scale. Fewer individuals included in for the help with map making and Sheri A. Maxwell for the each sample at the leaf or plant scale would lead to assistance with DNA extractions. We are grateful to Dr Marc- more uncertainty associated with estimates of popula- Andre´ Lachance, Dr Lisette Waits and the anonymous review- tion genetic parameters (i.e. ‘noisier’ data) and poten- ers for the insightful comments on the earlier versions of the tially more outliers. At fine scales, we observed manuscript. Funding was provided by NSERC, the Canada Research Chair program and the University of Western Ontario. significant genetic differentiation as well as significant relationships between genetic and landscape variables. Thus, the decreasing sizes of samples at the finest scales References did not appear to limit our ability to detect significant effects. A small number of samples did however pre- Addicott JF (1974) Predation and prey community structure: an experimental study of the effect of mosquito larvae on the vent testing of such relationships at the bog scale. Repli- protozoan communities of pitcher plants. Ecology, 55, 475– cation at broad spatial scales is a pervasive challenge in 492. landscape and ecological studies. Even when there are Allendorf FW, Phelps SR (1981) Use of allelic frequencies to sufficient resources to sample multiple larger regions or describe population structure. Canadian Journal of Fisheries ‘landscapes’, each landscape may have its own history and Aquatic Sciences, 38, 1507–1514. and unique features, making their true replication tech- Anderson MJ (2003) DISTLM Forward: A FORTRAN Computer nically impossible (Hargrove & Pickering 1992). How- Program to Calculate a Distance-Based Multivariate Analysis for a Linear Model using Forward Selection. Department of ever, conducting research at multiple similar landscapes Statistics, University of Auckland, New Zealand. http:// can nonetheless be informative (Anderson et al. 2010; www.stat.auckland.ac.nz/~mja/Programs.htm/DISTLM_ Short Bull et al. 2011). Our sampling of two different forward3.exe. bog systems, which could be considered as different Anderson CD (2010) Effects of movement and mating patterns ‘landscapes’, proved to be very important given the dif- on gene flow among overwintering hibernacula of the ferent patterns of distribution of full-siblings between Timber Rattlesnake (Crotalus horridus). Copeia, 1, 54–61. the systems (Fig. 2). Nonidentical landscape attributes Anderson EC, Dunham KK (2008) The influence of family groups on inferences made with the program Structure. may limit generalizations and strict statistical infer- Molecular Ecology Resources, 8, 1219–1229. ences, but they provide the opportunity to detect plas- Anderson CD, Epperson BK, Fortin M-J et al. (2010) ticity of ecological processes and patterns. Considering spatial and temporal scale in landscape-genetic studies of gene flow. Molecular Ecology, 19, 3565–3575. Armitage PD, Cranston PS, Pinder LCV (1995) The Conclusion Chironomidae: The Biology and Ecology of Non-Biting Midges. We demonstrated that encompassing a large research Chapman & Hall, London. Balkenhol N, Gugerli F, Cushman SA et al. (2009) Identifying area (suitable extent), along with refining the resolution future research needs in landscape genetics: where to from of sampling (gradually changing the grain), reveals here? Landscape Ecology, 24, 455–463. links among processes and patterns across different Balloux F (2004) Heterozygote excess in small populations and spatial scales. Genetic differentiation at several scales in the heterozygote-excess effective population size. Evolution, Metriocnemus knabi is significantly associated with land- 58, 1891–1900. scape variables related to habitat size, abundance and Blanquer A, Uriz MJ (2010) Population genetics at three spatial spatial arrangement. These broader-scale landscape fea- scales of a rare sponge living in fragmented habitats. BMC Evolutionary Biology, 10, 13. doi: 10.1186/1471-2148-10-13. tures seem to influence the fine-scale process of female Buckley HL, Miller TE, Ellison AM, Gotelli NJ (2010) Local- to oviposition. This process, in turn, shapes the patterns of continental-scale variation in the richness and composition of genetic differentiation observed at both fine and broad

2011 Blackwell Publishing Ltd GENETIC STRUCTURE OF M. KNABI ACROSS SCALES 235

an aquatic food web. Global Ecology and Biogeography, 19, McArdle BH, Anderson MJ (2001) Fitting multivariate models 711–723. to community data: a comment on distance-based Cushman SA, Landguth EL (2010) Spurious correlations and redundancy analysis. Ecology, 82, 290–297. inference in landscape genetics. Molecular Ecology, 19, 3592– Miller TE, Horth L, Reeves RH (2002) Trophic interactions in 3602. the phytotelmata communities of the pitcher plant, Sarracenia Diggle PJ, Ribeiro PJ (2007) Model-Based Geostatistics. Series: purpurea. Community Ecology, 3, 109–116. Springer Series in Statistics, Springer-Verlag, Berlin. Miner JA, Taylor PD (2002) Effects of peatland size and Evanno G, Regnaut S, Goudet J (2005) Detecting the number of exposure on two species of Diptera inhabiting the pitcher clusters of individuals using the software STRUCTURE: a Sarracenia purpurea L. Ecoscience, 9, 347–354. simulation study. Molecular Ecology, 14, 2611–2620. Mitzui Y, Isagi Y, Setoguchi H (2010) Multiple spatial scale Fortin M-J, Dale MRT (2005) Spatial Analysis: A Guide for patterns of genetic diversity in riparian populations of Ecologists. Cambridge University Press, Cambridge. Ainsliaea faurieana (Asteraceae) on Yakushima Island, Japan. Goldberg CS, Waits LP (2010) Quantification and reduction of American Journal of Botany, 97, 101–110. bias from sampling larvae to infer population and landscape Murphy MA, Evans JS, Storfer A (2010) Quantifying Bufo boreas genetic structure. Molecular Ecology Resources, 10, 304–313. connectivity in Yellowstone National Park with landscape Gonçalves da Silva AS, Eberhard JR, Wright TF, Avery ML, genetics. Ecology, 91, 252–261. Russello MA (2010) Genetic evidence for high propagule Nastase AJ, De la Rosa C, Newell SJ (1995) Abundance of pressure and long-distance dispersal in monk parakeet pitcher-plant mosquitoes, Wyeomyia smithii (Coq.) (Diptera: (Myiopsitta monachus) invasive populations. Molecular Culicidae) and midges, Metriocnemus knabi Coq. (Diptera: Ecology, 19, 3336–3350. Chironomidae), in relation to pitcher characteristics of Goudet J (1995) FSTAT (version 1.2): a computer program to Sarracenia purpurea L. American Midland Naturalist, 133, 44–51. calculate F-statistics. Journal of Heredity, 86, 485–486. Nielsen JL, Byrne A, Graziano SL, Kozfkay CC (2009) Goudet J (2005) HIERFSTAT, a package for R to compute and Steelhead genetic diversity at multiple spatial scales in a test hierarchical F-statistics. Molecular Ecology Notes, 5, 184– managed basin: Snake River, Idaho. North American Journal of 186. Fisheries Management, 29, 680–701. Guillot G (2009) On the inference of spatial structure from Oliver DR (1971) Life history of the Chironomidae. Annual population genetics data. Bioinformatics, 25, 1796–1801. Review of Entomology, 16, 211–230. Hargrove WW, Pickering J (1992) Pseudoreplication: a sine qua van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P non for regional ecology. Landscape Ecology, 6, 251–258. (2004) MICRO-CHECKER: software for identifying and Harvey E, Miller TE (1996) Variance in composition of correcting genotyping errors in microsatellite data. Molecular inquiline communities in leaves of Sarracenia purpurea L. on Ecology Notes, 4, 535–538. multiple spatial scales. Oecologia, 108, 562–566. Paterson CG, Cameron CJ (1982) Seasonal dynamics and Heard SB (1994) Pitcher-plant midges and mosquitoes: a ecological strategies of the pitcher plant chironomid, processing chain commensalism. Ecology, 75, 1647–1660. Metriocnemus knabi Coq. (Diptera: Chironomidae), in Hedrick PW (1999) Perspective: highly variable loci and their southeast New Brunswick. Canadian Journal of Zoology, 60, interpretation in evolution and conservation. Evolution, 53, 3075–3083. 313–318. Peakall R, Smouse PE (2006) GenAlEx: genetic analysis in Holling CS (1992) Cross-scale morphology, geometry, and Excel. Population genetic software for teaching and research. dynamics of ecosystems. Ecological Monographs, 62, 447–502. Molecular Ecology Notes, 6, 288–295. Hu LJ, Uchiyama K, Shen HL, Ide Y (2010) Multiple-scaled Pilot M, Jedrzejewski W, Branicki W et al. (2006) Ecological spatial genetic structures of Fraxinus mandshurica over a factors influence population genetic structure of European riparian–mountain landscape in Northeast China. grey wolves. Molecular Ecology, 15, 4533–4553. Conservation Genetics, 11, 77–87. Pritchard JK, Stephens M, Donnelly P (2000) Inference of Kalinowski ST, Wagner AP, Taper ML (2006) ML-RELATE: a population structure using multilocus genotype data. computer program for maximum likelihood estimation of Genetics, 155, 945–959. relatedness and relationship. Molecular Ecology Resources, 6, Prout T (1981) A note on the island model with sex dependent 576–579. migration. Theoretical and Applied Genetics, 59, 327–332. Kotliar NB, Wiens JA (1990) Multiple scales of patchiness and Rango JJ (1999) Summer phenology of aquatic insect patch structure: a hierarchical framework for the study of communities inhabiting the leaves of the northern pitcher heterogeneity. Oikos, 59, 253–260. plant, Sarracenia purpurea L. Northeastern Naturalist, 6, 19–30. Krawchuk MA, Taylor PD (2003) Changing importance of Rasic G, Maxwell SA, Keyghobadi N (2009) Characterization of habitat structure across multiple spatial scales for three microsatellite loci for the pitcher plant midge, Metriocnemus species of insects. Oikos, 103, 153–161. knabi Coq. (Diptera: Chironomidae). Molecular Ecology Lee-Yaw JA, Davidson A, McRae BH, Green DM (2009) Do Resources, 9, 1388–1391. landscape processes predict phylogeographic patterns in the Ritland K (1996) Estimator for pairwise relatedness and wood frog? Molecular Ecology, 18, 1863–1874. individual inbreeding coefficients. Genetical Research, 67, 175– Legendre P, Anderson MJ (1999) Distance-based redundancy 185. analysis: testing multi-species responses in multi-factorial Savage WK, Fremier AK, Shaffer HB (2010) Landscape genetics ecological experiments. Ecological Monographs, 69, 1–24. of alpine Sierra Nevada salamanders reveal extreme Levin SA (1992) The problem of pattern and scale in ecology: the population subdivision in space and time. Molecular Ecology, Robert H. MacArthur award lecture. Ecology, 73, 1943–1967. 19, 3301–3314.

2011 Blackwell Publishing Ltd 236 G. RASIC and N. KEYGHOBADI

Schwartz MK, McKelvey KS (2009) Why sampling scheme Final PCR conditions and primer concentrations: uploaded as matters: the effect of sampling scheme on landscape genetic online Supporting Information. results. Conservation Genetics, 10, 441–452. Microsatellite data, sample locations and landscape variables: Short Bull RA, Cushman SA, Mace R et al. (2011) Why uploaded as online Supporting Information. replication is important in landscape genetics: American black bear in the Rocky Mountains. Molecular Ecology, 20, Algonquin Park (Ontario, Canada) forest types: http:// 1092–1107. www.mnr.gov.on.ca/en/Business/Forests/Publication/MNR_ Smouse PE, Peakall R (1999) Spatial autocorrelation analysis of E005106P.html; accessed on September 8, 2010. multi-allele and multi-locus genetic microstructure. Heredity, 82, 561–573. Sokal RR, Oden NL (1991) Spatial autocorrelation analysis as Supporting information an inferential tool in population genetics. The American Naturalist, 138, 518–521. Additional supporting information may be found in the online Storfer A, Murphy M, Spear SF, Holderegger R, Waits L (2010) version of this article. Landscape genetics: where are we now? Molecular Ecology, 19, 3496–3514. Fig. S1 The output from the STRUCTURE HARVESTER v0.6.5 Tiner RW (1999) Wetland Indicators: A Guide to Wetland (Evanno et al. 2005) of mean Log probability (±SD) for the Identification, Delineation, Classification, and Mapping. Lewis number of genetic clusters (K) for Metriocnemus knabi samples, Publishers, CRC Press, Boca Raton, Florida. inferred using the STRUCTURE analysis (Pritchard et al. 2000). Trzcinski MK, Walde SJ, Taylor PD (2003) Colonisation of Fig. S2 Assignment probabilities for individuals of Metriocne- pitcher plant leaves at several spatial scales. Ecological mus knabi from the STRUCTURE analysis. Entomology, 28, 482–489. Trzcinski MK, Walde SJ, Taylor PD (2005) Local interactions in Table S1 Locations of clusters (cl 1–24) where the pitcher plant pitcher plant communities scale-up to regional patterns in midge (Metriocnemus knabi) was sampled, in bogs from Algon- distribution and abundance. Environmental Entomology, 34, quin Provincial Park (Ontario, Canada). 1464–1470. Table S2 Multi-plex PCR combinations for microsatellite loci Wagner HH, Fortin M-J (2005) Spatial analysis of landscapes: of Metriocnemus knabi (Rasic et al. 2009). concepts and statistics. Ecology, 86, 1975–1987. Wiens JA (1989) Spatial scaling in ecology. Functional Ecology, Table S3 Summary of Structure results (Pritchard et al. 2000) 3, 385–397. in samples of Metriocnemus knabi, presented as the output from Wright S (1969) Evolution and the Genetics of Populations. the software STRUCTURE HARVESTER v0.6.5 (Evanno et al. University of Chicago Press, Chicago. 2005). Wu J, Loucks OL (1995) From balance of nature to hierarchical Table S4 Microsatellite data for each sample analyzed. patch dynamics: a paradigm shift in ecology. Quarterly Review of Biology, 70, 439–466. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be G.R. has recently obtained a Ph.D. degree in Biological Sciences directed to the corresponding author for the article. under the supervision of Professor N.K. and is interested in genetics and ecology of spatially structured populations and communities. The research interests of N.K. fall within the realm of molecular ecology, with the main focus in the field of landscape genetics.

Data accessibility

GeneBank accessions numbers for Metriocnemus knabi microsat- ellites: FJ665262–FJ665273; http://www.ncbi.nlm.nih.gov/ nuccore/224384722?report=genbank; accessed on August 20, 2010.

2011 Blackwell Publishing Ltd