BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research. Genetic Structure of Vaccinium parvifolium (Ericaceae) in Northern California Reveals Potential Systematic Distinctions Author(s): Jennifer DeWoody , Valerie D. Hipkins , Julie Kierstead Nelson , and Len Lindstrand III Source: Madroño, 59(4):196-210. 2012. Published By: California Botanical Society DOI: http://dx.doi.org/10.3120/0024-9637-59.4.196 URL: http://www.bioone.org/doi/full/10.3120/0024-9637-59.4.196 BioOne (www.bioone.org ) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/ terms_of_use . Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.
Transcript
BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofitpublishers, academic institutions, research libraries, and research funders in the common goal of maximizing access tocritical research.
Genetic Structure of Vaccinium parvifolium (Ericaceae) inNorthern California Reveals Potential Systematic DistinctionsAuthor(s): Jennifer DeWoody , Valerie D. Hipkins , Julie Kierstead Nelson , andLen Lindstrand IIISource: Madroño, 59(4):196-210. 2012.Published By: California Botanical SocietyDOI: http://dx.doi.org/10.3120/0024-9637-59.4.196URL: http://www.bioone.org/doi/full/10.3120/0024-9637-59.4.196
BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in thebiological, ecological, and environmental sciences. BioOne provides a sustainable onlineplatform for over 170 journals and books published by nonprofit societies, associations,museums, institutions, and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated contentindicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use.
Usage of BioOne content is strictly limited to personal, educational, and non-commercialuse. Commercial inquiries or rights and permissions requests should be directed to theindividual publisher as copyright holder.
LEN LINDSTRAND IIINorth State Resources, Inc., 5000 Bechelli Lane, Redding, CA 96002
ABSTRACT
Vaccinium parvifolium Sm. (Ericaceae) is an important understory shrub in conifer forests inwestern North America. Populations putatively classified as V. parvifolium in northern Californiadisplay alternate berry morphology, consistent with a possible phenotypic diversification or crypticspeciation. Identification of cryptic species or subspecies would influence management guidelinesgiven the limited range of some morphological variants. In order to inform management guidelines,two Vaccinium species were characterized via molecular genetic analyses. Plants of typical V.parvifolium morphology from the coastal areas of northwest California, western Oregon andWashington, atypical plants from Shasta County and the central Sierra Nevada, and one populationof V. deliciosum Piper, a congener, were assessed at five nuclear microsatellite loci. Analyses ofdifferentiation, admixture, and phylogenetic relationships indicated that populations displayingatypical morphology were more similar to V. deliciosum than to the typical V. parvifolium. Althoughadditional data are required to determine whether these differences warrant taxonomic treatmentwithin Vaccinium, management plans should consider three distinct gene pools among thesepopulations.
Key Words: Ericaceae, management units, microsatellites, population genetics, Shasta County,taxonomy.
Members of the genus Vaccinium (Ericaceae)are ecologically important shrubs in the forests ofthe Pacific Coast of North America, where theymay dominate understory vegetation especiallyin response to disturbance or thinning (Hanley2005). Vaccinium parvifolium Sm., the red huck-leberry, is a perennial understory shrub thatoccurs from Alaska south through British Co-lumbia and into California. The stems of V.parvifolium reach 4 m in height and are sharplyangled, giving shrubs a characteristic architec-ture, and the translucent red, edible berries arehighly distinguishing (Wallace 1993, accessed inRosatti 2003; Vander Kloet 2009). Plants are notrhizomatous but produce horizontal stems capa-ble of rooting, providing a means of vegetativereproduction which can maintain genets (geneticindividuals) over many years. The V. parvifoliumflower color is described as pink, bronze oryellowish green, and they produce red berries upto 10 mm in diameter (Vander Kloet 2009). Inaddition to having edible fruit, V. parvifoliumvegetation is highly palatable to deer (Vila et al.2004). As an angiosperm shrub in conifer forests,V. parvifolium likely plays a critical role in species
diversity and ecological systems (Wender et al.2004).
Populations of putative V. parvifolium in tworegions of California, Shasta County and thecentral Sierra Nevada, do not display the typicalmorphology (Fig. 1). Populations in the areaaround Shasta Lake, in Shasta County, Califor-nia display white to pink urceolate flowers withglaucous berries. Plants grow in distinct clumpsof or as individual ramets, typically 1 to 2 m inheight, but some reach up to 3 m tall. Plants areerect and do not root along stems. The Shastapopulations are found in conifer and hardwood-conifer forest, and chaparral habitats; typicallyin vegetation types classified as Douglas-fir,ponderosa pine, montane-hardwood-conifer,and mixed chaparral (Mayer and Laudenslayer1988). Shasta County Vaccinium are associatedwith riparian areas, springs and seeps, and mesicforest environments. These sites are distinguishedfrom the habitats of typical V. parvifolium bysubstantially higher summer temperatures andmore extended summer drought. Further, theShasta County populations are in areas charac-terized by acidic soil and/or water conditions. In
MADRONO, Vol. 59, No. 4, pp. 196–210, 2012
many situations the species grows immediatelydownstream of historic mines in the acidic soiland water. The species also regularly occurs inriparian areas with acid mine discharge waterchemistry. These inland Shasta County popula-tions are disjunct from the nearest known extantred huckleberry populations in the coastal regionby approximately 60 km, with the Trinity Alpsand other Klamath Ranges lying between them(J.K.N., personal observation and review ofherbarium records).
Vaccinium populations in the central SierraNevada display white urceolate flowers with blue-black berries that have a calyx ring, and plants aredecumbent and capable of rooting along branches
(Fig. 1). The Sierra Nevada plants grow alongdrainages in Pseudotsuga menziesii-Abies conco-lor-Calocedrus decurrens and in Abies concolor-Pinus lambertiana-Calocedrus decurrens/Cornusnuttallii/Corylus cornuta var. californica forestassociations, intermediate in moisture level be-tween the Shasta County form and the typicalform. The atypical populations in the SierraNevada are more than 250 km from the ShastaCounty populations. All specimens observed fromthe Shasta County and Sierra Nevada regionsdisplay the atypical forms (A. E. L. Colwell,National Park Service, personal communication).
The taxonomy of the Vaccinium genus is ofongoing study due to the wide range and high
FIG. 1. Morphological variation among Vaccinium parvifolium populations sampled for genetic analyses. Thetypical red-berry morphology observed in the coastal areas (top), purple-colored berry observed in Shasta County(middle), and a distinctive calyx ring on berries that mature to a blue-black color in the central SierraNevada (bottom).
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 197
levels of phenotypic variation observed withinspecies (Vander Kloet and Dickinson 1999;Powell and Kron 2002). Members of the Vaccin-ium genus are found in the Northern Hemispherecircling the Pacific Rim from Japan to Mexico,with a center of species and habitat diversityranging from Alaska to California. No cleargeographical pattern was distinguished in thegrouping of Vaccinium species, and the fullphylogeny of the clade has not been resolved,requiring greater sampling to better represent theapproximately 20 species of Vaccinium in NorthAmerica (Kron et al. 2002; Powell and Kron2002).
Given the ecological importance of Vacciniumshrubs in conifer ecosystems, the unresolvednature of the systematic classifications withinthe genus, and the morphological variationobserved in northern California, additional studyof V. parvifolium was warranted. Identificationof landscape patterns of genetic differentiation orcryptic species can influence estimates of speciesdiversity and local adaptation, and informconservation plans. Although most studies usingmolecular markers to identify cryptic species havefocused on animals (Bickford et al. 2006), the useof molecular markers in plant species is wellestablished in studies of population genetics(Hamrick and Godt 1996) and may aid effortsto resolve questions of speciation and divergence(Lawton-Rauh 2008).
Here, we used molecular data to investigatethe genetic relationship of Vaccinium popula-tions primarily from northern and central Cali-fornia. Vaccinium parvifolium populations fromnorthern California and western Oregon andWashington (the typical red huckleberry) werecompared to collections displaying atypicalmorphologies from Shasta County and thecentral Sierra Nevada. In addition, one popula-tion of the congener Vaccinium deliciosum Piperwas included from Shasta County. Vacciniumdeliciosum, the Cascade bilberry, is typicallyshorter than V. parvifolium but is rhizomatousand capable of forming dense clonal mats(Wallace 1993; Vander Kloet 2009). Vacciniumdeliciosum flowers range from creamy pink tored, and fruit are typically blue but may be blackor maroon or sometimes red. These species areclosely related, with V. parvifolium and its sistertaxa V. ovalifolium Sm. composing the ancestralclade to V. deliciosum (Powell and Kron 2002).Specifically, our objectives were to quantify thelevel of genetic differentiation among popula-tions and geographic regions, to determinewhether the populations from Shasta Countyand the Sierra Nevada are likely to be morpho-logical variants of V. parvifolium or may betaxonomically distinct, and to provide data toinform ongoing conservation efforts in ShastaCounty.
MATERIALS AND METHODS
Study Species and Collections
In order to investigate the genetic structure ofVaccinium parvifolium, five categories of popula-tions were sampled for genetic analyses (Table 1,Fig. 2). We sampled populations of atypicalmorphology from regions lacking the typicalred-berried V. parvifolium. These categories rep-resent distinct geographic regions with limited V.parvifolium distribution between. Five popula-tions displaying the unusual berry color andmorphology were sampled from the ShastaCounty area (category 1). Ten populationsdisplaying the typical V. parvifolium berry colorand morphology were sampled: eight fromnorthwestern California, relatively proximate toShasta County (category 2), and two from westernOregon and Washington (category 3). Fourpopulations of dark-fruited Vaccinium were sam-pled from the central Sierra Nevada (category 4).Finally, one population of V. deliciosum fromwestern Shasta County was sampled for interspe-cific comparison (category 5). Voucher specimenswere collected from each population in order toverify and document species identification.
Collections were made over three growingseasons (2004, 2005 and 2009). Up to 20 sampleswere collected from each population. Populationlocations were recorded in the field using GPS ortopographic maps. Samples from the coastal andShasta County regions (categories 1 and 2) werecollected from distinct shrubs randomly selectedfrom throughout the population. The Sierra Ne-vada samples (category 4) were spaced to minimizeduplicate collection of clones. Leaf material wascollected in the field and stored on wet ice and/orrefrigerated until shipped to the USDA ForestService National Forest Genetics Lab (NFGEL).
DNA Isolation and Analysis
Total genomic DNA was isolated from allsamples using the DNEasy-96 plant kit (Qiagen,Valencia, CA) following the liquid nitrogenprocedure as described in the manual. DNAconcentrations were quantified using a GeminiXPS Microplate Spectrofluorometer (MolecularDevices, Sunnyvale, CA) using PicoGreen dsDNAReagent (Invitrogen, Carlsbad, CA). Samples wereassayed for five microsatellite loci originallydescribed in Vaccinium corymbosum L.: CA23F,CA421, CA787F, VCC1_I2, and VCC1_J9(Boches et al. 2005). Amplification of the five locitook place in two multiplex reactions: CA421 withCA787F in one reaction, and CA23F, VCC1_I2,and VCC1_J9 in a second reaction. Amplificationconditions included 10 ng of template DNA, 2 mMof each primer, and 13 of Multiplex master mix.Both amplifications took place following the
198 MADRONO [Vol. 59
program provided with the Multiplex PCR Kit(Qiagen). Forward primers for all loci were labeledwith a fluorescent tag for visualization on anABI3130xl capillary electrophoresis system (Ap-plied Biosystems, Carlsbad, CA): CA23F andCA787F labeled with HEX, CA421 and VCC1_I2with 6FAM, and VCC1_J9 with NED. Electro-phoresis was conducted using a 1:100 dilution ofamplification products.
Data Analysis
In order to quantify the extent of vegetativereproduction in Vaccinium, multiple ramets of the
same genet were identified as matching multi-locus genotypes using the Multilocus Matchestool in GenAlEx v. 6 (Peakall and Smouse2006). Missing data were considered sufficientto identify a mismatch. The genet diversity ofeach population was quantified as the number ofunique genotypes observed per sample (G/N).Differences in genet diversity among geographicregions (coastal, Shasta, and Sierra) were deter-mined using univariate ANOVA as implementedin SPSS v. 17.0 (IBM SPSS, Chicago, IL). Theprobability of identity function was then usedto determine the expected number of unrelatedgenets with the same multilocus genotype in each
TABLE 1. LOCATION AND CLASSIFICATION OF 20 POPULATIONS OF VACCINIUM SPP. SAMPLED FOR DNAANALYSES. Populations were grouped into four categories according to morphology, geography, and speciesdesignation. aPopulation 2SR was collected from three sites forming one distinct population. The three collectionswere combined for all statistical analyses.
Abbrev. Population County, State nLatitude/Longitude Elev. (ft)
Atypical Vaccinium in Shasta County
1BH Bully Hill, Shasta Lake Shasta, CA 20 40.7882/2122.2087 11001FLM Friday-Lowden Mine,
Shoemaker GulchShasta, CA 26 40.7534/2122.4599 2400
1LB Little Backbone Creek,Shasta Lake
Shasta, CA 20 40.7611/2122.4376 1100
1SQC (Little) Squaw Creek,Shasta Lake
Shasta, CA 20 40.7397/2122.4689 1100
1ULB Upper Little BackboneCreek, Shasta Lake
Shasta, CA 20 40.77/2122.4444 2000
Typical V. parvifolium, Pacific coast of California
2ACF Arcata Community Forest Humboldt, CA 33 40.8772/2124.0799 412HC Happy Camp, Doolittle
Creek Drainage, ForestRoad 17N62
Siskiyou, CA 27 41.84/2123.45 3280
2HD High Divide, Hiouchi, CA Del Norte, CA 2 41.8029/2124.0625 422JS Jedediah Smith State Park,
Redwood National Park,Redwood State Park
Del Norte, CA 15 41.8083/2124.0894 42
2LD Low Divide, Hiouchi, CA Del Norte, CA 2 40.7798/2124.1144 412LR Lentell Road, Eureka, CA Humboldt, CA 2 41.9146/2124.1144 422SFM South Fork Mountain Trinity, CA 20 40.3474/2123.221 47002SRa Six Rivers National Forest Del Norte, CA 4 41.7683/2123.8852 30902SRa Six Rivers National Forest,
Camp SixDel Norte, CA 4 41.8268/2123.8729 3600
2SRa Six Rivers National Forest,Gordon Mtn.
Del Norte, CA 4 41.7884/2123.8713 3824
Typical V. parvifolium, western Oregon and Washington
3I Issaquah, Christmas Lake King, WA 1 47.432/2121.760 42503MP Mary’s Peak Recreation Area Benton, OR 23 44.4957/2123.5436 2575
Atypical Vaccinium in the central Sierra Nevada
4BC Big Creek, YosemiteNational Park
Mariposa, CA 22 37.5083/2119.6604 4630
4CC Clear Creek Road El Dorado, CA 20 38.6901/2120.6361 26004GH Greeley Hill Tuolumne, CA 19 37.7683/2120.0625 32814MC Moss Creek, Yosemite
National ParkTuolumne, CA 20 37.7638/2119.8332 5972
V. deliciosum
5VDE Shasta Bally, WhiskeytownNational Park
Shasta, CA 20 40.6011/2122.6437 5900
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 199
population. When multiple ramets were identifiedin a population, the data set was reduced tounique genets (one stem per genotype) andassessed for basic measures of allelic diversity,including the percent polymorphic loci (P), meanalleles per locus (A), effective alleles per locus (Ae)(which standardizes for varying sample sizes),observed (Ho) and expected heterozygosity (He),and the within-population fixation index (F).All measures were estimated using GenAlEx v.6 (Peakall and Smouse 2006). When sufficientsamples were present (10 genets), populationswere tested for an excess of heterozygotes relativeto mutation-drift equilibrium using the Wilcoxonsign-rank test under a two-phase mutation model
as implemented by Bottleneck (Cornuet andLuikart 1996).
When applying microsatellite markers to differ-ent species, mutation of the PCR primer site mayproduce null alleles that do not produce visibleproducts. These null alleles may inflate certaingenetic measures such as the fixation index (orestimated inbreeding) and decrease other measures(such as heterozygosity). As the markers used inthis study were not designed specifically for V.parvifolium or V. deliciosum, null alleles must betaken into account for some analyses. In order todetect null alleles, those populations with suffi-cient sample sizes (unique genotypes) were assess-ed using Micro-Checker v2.2.3 (van Oosterhout
FIG. 2. Vaccinium populations were sampled in western Oregon and western Washington, coastal California,Shasta County, and the Sierra Nevada. The Shasta County populations included one population of V. deliciosum(5VDE). Population abbreviations follow Table 1.
200 MADRONO [Vol. 59
et al. 2004). When null alleles were detected, theadjusted allele frequencies and genotypes (ran-domized within the population) were calculatedusing the method of Brookfield (1996) where nonull-null homozygotes are expected.
Both raw data and the null-adjusted genotypeswere then used to estimate measures of popula-tion differentiation using analysis of molecularvariance (AMOVA) over two models. The firstmodel assumed no hierarchical structure andestimated the differentiation of all populationsrelative to the total. The second model tested forhierarchical structure among geographic regionsof populations, and estimated differentiationamong populations within each region, andamong regions relative to the total collection.The hierarchical model grouped populationsgeographically and taxonomically into fourgroups: the atypical Vaccinium from ShastaCounty (category 1), typical V. parvifolium fromthe Pacific coast and western Oregon (categories2 and 3), the atypical Vaccinium from the centralSierra Nevada (category 4), and V. deliciosum(category 5). The two populations with a singlegenet sampled (2LD and 3I, Table 1) weredropped from the AMOVA. Significance ofdifferentiation was determined by 999 bootstrapreplicates, as implemented in GenAlEx v. 6(Peakall and Smouse 2006).
Allele frequencies (both raw and adjusted fornull alleles) were used to estimate Nei’s (1972)genetic distance for all pairs of populations. Thedistance matrices were then used to test if geneflow decreases as a function of geographicdistance (isolation by distance, IBD). The latitudeand longitude of each population was used tocalculate the linear distance between sites in km.The correlation between the genetic and geo-graphic distance matrices was assessed using theMantel Test procedure as implemented byGenAlEx v. 6 (Peakall and Smouse 2006).Significance of the correlation was assessed from999 permutations of the dependent (genetic)matrix relative to the independent (geographic)matrix. The distance matrices (raw and null-corrected) were also used to build consensuspopulation phenograms using Neighbor-Joiningmethods, with significant branches identified from100 bootstrap replicates following the extendedmajority rule option. Phenograms were builtusing the applications in PHYLIP (Felsenstein2005).
Raw multilocus genotypes were treated to twounsupervised multivariate analyses to identifygenetic structure not predicted a priori. First, thegenetic distance matrix for all pairs of populationswas treated to a Principle Coordinate Analysis(PCoA) as implemented by GenAlEx (Peakall andSmouse 2006). The analysis was conducted on thestandardized covariance matrix, with uniquegenotypes represented for each population.
Second, raw genotypes were assigned toanonymous genetic clusters using Markov ChainMonte Carlo simulations and Bayesian likelihoodmethods as implemented in the program Struc-ture v. 2.3.3 (Pritchard et al. 2000). This analysisis a robust method to identify clusters of similarmultilocus genotypes even in the presence ofadmixture (Pritchard et al. 2000; Falush et al.2003). To account for a low frequency of nullalleles, the analysis was conducted with theassumption of unreported recessive alleles, allow-ing for homozygous genotypes to be consideredas putative heterozygotes for an undetected (null)allele (Falush et al. 2007; Pritchard et al. 2007).For these data, simulations were undertaken totest for the most likely number of clusters forthe set K 5 {1:10}, over 100,000 replicationsfollowing a burn-in period of 50,000 replications,with no prior population information provided.The parameters were set to assume allele fre-quencies were uncorrelated among populations,which is feasible given the taxonomic andgeographic scale of this study. For example, atleast two distinct species have been sampled, andmany populations are separated by greaterdistances than the typical pollinator, bees, areexpected to travel regularly. Using the uncorre-lated frequencies may merge similar populationsand deflate estimates of K, while correlatedfrequencies may inflate K (Pritchard et al.2000). The parameter set did allow for admixturewithin populations and individuals, which canaccount for migration or hybridization events.Calculations are based on the mean likelihood offive simulations for each value of K. The numberof genetic clusters (K) was inferred from the teststatistic dK following Pritchard et al. (2000), andindividual assignments were assessed visually forthe simulation with the greatest likelihood value.In addition, the genetic relationship of theinferred genetic clusters was visualized as apopulation phenogram using Neighbor-Joiningmethods (Saitou and Nei 1987) visualized usingthe DrawTree application from PHYLIP (Fel-senstein 2005).
RESULTS
Matching multilocus genotypes were identi-fied in 12 populations (including V. deliciosum),indicating Vaccinium reproduces vegetativelywith some frequency. The majority of individualssampled had a unique multilocus genotype (meannumber of samples per genotype 5 1.24, sx 50.05), with a maximum of 11 stems found toshare one genotype (population 4BC, where allstems were in a single, large patch). Resultingmeasures of genet diversity (G/N) were high(mean 5 0.81, sx 5 0.05), with values rangingfrom 0.27 to 1.0 (Table 2). An analysis ofvariance indicated the levels of G/N to vary
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 201
among geographic groups (Shasta County, coast-al area, or Sierra Nevada) of putative V. parvi-folium (V. deliciosum excluded) (F2,16 5 4.66, P 5
0.026).The estimated probability that two unrelated
genets displayed the same multilocus genotype(probability of identity, PI) varied among collec-tions. The greatest PI occurred in populationswith small sample sizes (2LD, PI 5 0.05; 2HD, PI5 0.03) or in the population with the highestfrequency of clonality resulting in a low effectivesample size (4BC, PI 5 0.05). The remainingpopulations displayed lower PI values, rangingfrom 2.5 3 1026 to 0.008. When PI values wereused to estimate the predicted number ofunrelated genets with the same genotype in eachpopulation, based on sample sizes, only onepopulation (4BC) was expected to have oneduplicate genet (1.004), with all other values lessthan 0.2. These values indicate that while the fivemicrosatellite loci did not provide universally lowPI values in each population, the samplingstrategy was sufficient to conclude that duplicategenotypes are most likely ramets of one genet andnot a consequence of low statistical power.
All loci displayed allelic variation, althoughsome populations were fixed for single variants(Table 2, Supplemental Material). Allelic diversi-ty varied among populations, with the averagenumber of alleles ranging from 1.6 to 10.4 perpopulation. This range is likely influenced bydifferences in sample sizes, as a smaller range wasestimated for the effective alleles per population,
which accounts for differences in sample number(mean 3.3 alleles, range 1.6 to 6.2). Moderatelevels of heterozygosity were observed (mean Ho
5 0.464), with levels of fixation ranging from21.0 to 0.378 (mean 20.08).
No evidence was found of genetic bottlenecks.All tests for heterozygosity excess relative tomutation-drift equilibrium were non-significant.
Tests of per-locus excess homozygosity re-vealed the occurrence of null alleles to be variableamong loci and populations. Four populationscontained an insufficient number of uniquegenotypes for the Micro-Checker (van Oosterh-out et al. 2004) analysis: 2HD, 2LD, 2LR, and 3I.Evidence of null alleles was found in four of thefive loci. Null alleles were detected at locusCA421 in populations 2ACF, 2JS, 3MP, and4MC; at locus CA787F in populations 2ACF,4GH, at locus VCC1_I2 in populations 1FLMand 5VDE, and at locus VCC1_J9 in populations2HC, 2SR, 3MP, and 4CC (Supplemental Mate-rial). The frequency of the detected null rangedfrom low (0.071) to moderate (0.262), indicatingthat null alleles may have inflated fixation indices(indicating a deficit of heterozygotes) in thesecollections. Genotypes adjusted for null alleles ateach affected locus were randomized within eachpopulation and used for population-level analy-ses of differentiation and genetic distance.
Analysis of molecular variance over the null-corrected data revealed significant differentiationamong populations and categories of populationsin this collection of Vaccinium. The one-level
TABLE 2. GENOTYPIC AND ALLELIC DIVERSITY AMONG POPULATIONS OF VACCINIUM SPP. ASSESSED AT FIVE
MICROSATELLITE LOCI. G 5 Number of unique multilocus genotypes; n 5 Sample size; P 5 Percent polymorphicloci; A 5 Mean alleles per locus; Ae 5 Effective alleles per locus; Ho 5 Observed heterozygosity; He 5 Expectedheterozygosity; F 5 Fixation index; deviation from Hardy-Weinberg equilibrium may be affected by the presenceof null alleles.
model estimated 58% of the variance to becontained within populations and 42% of thegenetic variance to be partitioned among popu-lations (P , 0.001). Significant differentiationwas observed in the two-level hierarchical mo-del based on geographic location: 51% of thevariance was contained within populations, 10%of the variance was distributed among popula-tions within a region, and 39% among geographicregions (P , 0.001). Estimates of differentiationfrom the raw genotypic data were congruent tothose from null-corrected data (data not report-ed), indicating the occurrence of null alleles in thedata set was insufficient to bias the detection ofdifferentiation among these populations.
Levels of genetic distance calculated from thecorrected genotypes varied among pairs ofpopulations (Table 3). Genetic distances weregreater between populations in different geo-graphical regions, and smaller among popula-tions within geographic areas. This pattern wasconfirmed by tests for isolation by distance(IBD), which identified a significant positivecorrelation between genetic and geographicdistance among all pairs of populations for bothraw (RXY 5 0.38, P , 0.001) and null-correcteddata (RXY 5 0.39, P 5 0.003). This pattern isconsistent with the AMOVA analyses indicatinggreater differences among regions than amongpopulations within a region.
Population phenograms built from the raw ornull-corrected data were mostly congruent intheir topology but neither was statisticallysignificant (all branches were observed in lessthan 60% of bootstrap replicates). In bothphenograms, populations grouped by geography,with the coastal populations (categories 2 and 3)forming one clade, the Sierra Nevada populations(category 4) a second clade, and the ShastaCounty populations (category 1) a third clade(Fig. 3). Potentially the most informative differ-ence between the raw and null-corrected pheno-grams involved the placement of the singlepopulation of V. deliciosum (5VDE). In the rawdata phenogram, 5VDE is placed between thecoastal clade and the Shasta and Sierra clades,while in the null-corrected data, 5VDE is placedbetween the Shasta and Sierra clades (Fig. 3).Together, these phenograms indicate the atypicalpopulations from Shasta County and the centralSierra Nevada are more similar to the congenerV. deliciosum than the coastal populations of V.parvifolium.
While the phenograms represent the similari-ties among population averages, the PrincipleCoordinate Analysis (PCoA) maximizes differ-ences among individuals. The PCoA based on theraw genotypes clustered samples by geogra-phic collection and morphology on the first twoaxes, which explained 61% of the variationamong individuals (Fig. 4). Four putative genetic
clusters were indicated by the PCoA: V. parvifo-lium of typical morphology from the coastalareas, Vaccinium from the Sierra Nevada Moun-tains, Vaccinium from Shasta County, and thecongener V. deliciosum. The first axis distinguish-es the typical V. parvifolium from the CoastalRanges and western Oregon from the collectionsof atypical Vaccinium from Shasta County andthe Sierra Nevada. The single population of V.deliciosum is intermediate to these, but moresimilar to the atypical Vaccinium collections(Fig. 4). The second axis further separates theatypical Vaccinium collections by geographicregion.
The admixture analyses using Structure esti-mated the most likely number of genetic clustersin the Vaccinium data set to be five (dK5 < 1, allother dK < 0). Individual assignment to eachcluster roughly followed population categoriesper Table 1. The populations of V. parvifoliumof typical morphology were genetically similarand potentially admixed, being assigned to twoclusters in varying proportions (orange and red).The five populations of atypical morphologyfrom Shasta County were assigned to a singlecluster (blue). The atypical populations Vaccin-ium from the Sierra Nevada were assigned toanother cluster (purple), with some admixedwith the blue cluster. The single population ofV. deliciosum (5VDE) was distinct and assignedto a unique genetic cluster (green) (Fig. 5). Theseindividual assignments are concordant with thePCoA results.
The Neighbor-Joining phenogram of anony-mous genetic clusters identified in the Structureanalysis was concordant with the PCoA andpopulation phenograms. The cluster containingatypical Shasta County samples (blue) was moresimilar to the cluster containing atypical centralSierra Nevada collections (purple) and V. deli-ciosum (green) than to the clusters containingsamples of typical morphologies (orange and red)(Fig. 5). These results are likely not an artifact ofthe null alleles as nulls were accounted for in theStructure analyses. Together, these results indi-cated the genetic differences among populationscorresponded to geographic and morphologicalfactors.
DISCUSSION
Genetic Differentiation amongSampled Populations
The primary goal of this study was to establishwhether populations of Vaccinium from ShastaCounty and the Sierra Nevada displaying atyp-ical berry color and morphology were geneti-cally similar to typical V. parvifolium fromnorthwestern California and western Oregonand Washington. Analyses of five microsatellite
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 203
TA
BL
E3.
ES
TIM
AT
ES
OF
NE
I’S
(1972)
GE
NE
TIC
DIS
TA
NC
EA
MO
NG
20
PO
PU
LA
TIO
NS
OF
VA
CC
INIU
MS
PP.
Ab
bre
via
tio
ns
foll
ow
Tab
le1.
Valu
esb
elo
wd
iago
nal
are
fro
mra
wd
ata
;valu
esab
ove
dia
go
nal
fro
mn
ull
-co
rrec
ted
all
ele
freq
uen
cies
.
1B
H1F
LM
1L
B1S
QC
1U
LB
2A
CF
2H
C2H
D2JS
2L
D2L
R2S
FM
2S
R3I
3M
P4B
C4C
C4G
H4M
C5V
DE
1B
H0.0
69
0.0
83
0.1
56
0.0
94
0.8
33
0.9
49
0.7
20
0.9
51
0.8
88
1.3
47
0.9
01
0.9
13
2.0
71
0.9
22
0.3
84
0.4
31
0.5
36
0.4
46
0.6
35
1F
LM
0.0
60
0.0
65
0.1
54
0.0
81
0.8
85
0.8
79
0.7
26
0.9
16
1.0
36
1.4
25
0.8
81
0.8
13
2.7
64
0.9
22
0.3
15
0.4
01
0.4
99
0.3
87
0.6
55
1L
B0.0
83
0.0
51
0.1
10
0.0
16
0.7
97
0.9
13
0.7
53
0.8
02
1.0
76
1.4
96
0.9
31
0.8
37
2.1
58
0.9
06
0.2
87
0.4
47
0.4
68
0.3
66
0.6
75
1S
QC
0.1
56
0.1
28
0.1
10
0.0
77
0.9
19
1.3
13
0.9
41
0.9
37
1.4
53
1.9
18
1.3
26
1.0
51
2.5
18
1.0
74
0.5
39
0.5
80
0.6
70
0.5
67
0.6
66
1U
LB
0.0
94
0.0
59
0.0
16
0.0
77
0.7
95
1.0
33
0.7
66
0.8
21
1.1
90
1.5
15
0.9
99
0.8
82
2.3
65
0.9
51
0.4
01
0.5
02
0.6
11
0.4
96
0.6
66
2A
CF
0.7
71
0.7
84
0.7
44
0.8
83
0.7
49
0.2
05
0.2
81
0.0
83
0.4
13
0.4
38
0.1
76
0.1
25
0.7
62
0.1
09
1.1
40
1.4
04
1.5
24
1.2
17
1.5
00
2H
C0.8
81
0.8
30
0.8
46
1.2
31
0.9
61
0.1
86
0.3
05
0.1
91
0.3
05
0.4
36
0.1
37
0.1
14
0.8
59
0.1
11
0.8
41
1.0
45
0.9
33
0.8
93
2.4
77
2H
D0.7
20
0.7
05
0.7
53
0.9
41
0.7
66
0.2
63
0.3
05
0.3
14
0.3
18
0.5
08
0.3
49
0.2
89
1.2
35
0.2
69
1.0
25
0.9
98
1.5
66
1.2
91
2.4
08
2JS
0.9
46
0.8
87
0.7
92
0.9
28
0.8
16
0.0
91
0.1
89
0.3
14
0.4
38
0.4
12
0.1
97
0.0
95
0.6
43
0.0
86
1.2
64
1.3
73
1.7
82
1.2
93
1.3
99
2L
D0.8
87
1.0
58
1.0
76
1.4
53
1.1
90
0.3
88
0.3
08
0.3
18
0.4
43
0.5
26
0.3
21
0.3
40
1.0
30
0.2
91
1.2
67
1.1
09
1.5
85
1.2
89
2.3
39
2L
R1.3
47
1.4
49
1.4
96
1.9
18
1.5
15
0.4
23
0.4
34
0.5
08
0.4
16
0.5
26
0.3
66
0.3
57
1.1
54
0.3
82
1.8
61
1.4
60
1.9
64
1.8
04
1.2
62
2S
FM
0.9
01
0.9
04
0.9
31
1.3
26
0.9
99
0.1
52
0.1
39
0.3
49
0.1
99
0.3
21
0.3
66
0.1
02
0.7
97
0.1
12
0.9
62
1.1
35
1.5
56
1.1
68
1.6
46
2S
R0.9
06
0.8
03
0.8
25
1.0
37
0.8
69
0.1
08
0.1
02
0.2
93
0.0
91
0.3
44
0.3
59
0.1
02
0.6
68
0.0
65
1.0
33
1.1
65
1.5
15
1.1
63
1.7
94
3I
2.0
70
2.7
76
2.1
57
2.5
18
2.3
65
0.7
80
0.8
48
1.2
35
0.6
28
1.0
30
1.1
54
0.7
97
0.6
58
0.6
48
3.0
59
5.5
69
2.9
19
3.4
02
3.1
23
3M
P0.9
09
0.9
01
0.8
89
1.0
55
0.9
36
0.1
04
0.1
01
0.2
68
0.0
92
0.2
89
0.3
82
0.1
07
0.0
67
0.6
37
1.0
77
1.1
88
1.5
81
1.1
89
1.7
92
4B
C0.3
84
0.3
33
0.2
87
0.5
39
0.4
01
1.0
53
0.8
98
1.0
25
1.2
63
1.2
66
1.8
61
0.9
62
1.0
25
3.0
58
1.0
52
0.5
04
0.2
00
0.1
39
1.0
08
4C
C0.4
29
0.4
28
0.4
50
0.5
82
0.5
05
1.3
05
1.0
81
0.9
81
1.3
68
1.0
89
1.4
03
1.1
31
1.1
88
5.3
93
1.1
95
0.5
12
0.6
00
0.5
00
0.9
32
4G
H0.4
66
0.4
42
0.4
05
0.6
20
0.5
41
1.3
27
1.1
54
1.3
34
1.5
53
1.3
52
1.7
80
1.3
31
1.2
97
2.6
51
1.3
51
0.1
64
0.5
29
0.0
79
0.9
29
4M
C0.4
70
0.4
24
0.3
83
0.5
96
0.5
19
1.0
70
0.7
56
1.1
92
1.2
09
1.1
91
1.6
43
1.0
52
1.0
46
2.8
40
1.0
75
0.1
48
0.5
16
0.0
57
0.9
23
5V
DE
0.6
15
0.6
51
0.6
50
0.6
34
0.6
35
1.3
90
2.2
14
2.3
01
1.3
63
2.3
49
1.2
68
1.6
48
1.7
42
2.9
89
1.7
36
1.0
16
0.9
32
0.9
75
0.9
70
204 MADRONO [Vol. 59
loci revealed significant structure and differentia-tion along geographic and morphological groups.Four statistical analyses provide congruent evi-dence that the atypical populations of Vacciniumare genetically distinct from the V. parvifoliumlocated in the Pacific coastal areas, and are moresimilar genetically to V. deliciosum, indicatingadditional systematic consideration may be war-ranted.
The level of differentiation among regions inthis study was greater than the among-population
variation reported for other Vaccinium species.The level of differentiation observed amongpopulations within geographic regions in thisstudy (10%) is consistent with the other Vaccin-ium reports (Yakimowski and Eckert 2008; Bellet al. 2009; Debnath 2009). Levels of differenti-ation among geographic regions was much higher(39%), indicating gene flow is restricted betweenthe sampling categories (Table 1).
Varying levels of vegetative reproduction wereidentified across the collection. Vaccinium species
FIG. 3. Unrooted Neighbor-Joining consensus population phenograms from Nei’s (1972) genetic distances among20 populations of Vaccinium. (A) The phenogram built from the raw data. (B) The phenogram built from null-corrected allele frequency data. Low statistical support was found for the topologies, with all branches occurring in,60% of bootstrap replicates.
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 205
are widely reported to reproduce clonally (Kreheret al. 2000; Yakimowski and Eckert 2008; Bellet al. 2009). Vaccinium parvifolium is reported totypically reproduce sexually but capable ofproducing sprouts in response to disturbance ordamage (Wender et al. 2004) and in some casesvia rhizomes (Gehrung 2001). Here, most popu-lations displayed high levels of genet diversity(.0.8), but the collections from the SierraNevada displayed greater vegetative reproductionthan those from the Coastal areas as indicated bysmaller G/N ratios. Rhizomatous species mayshare resources among stems via a complex rootsystem, providing a mechanism of survival insuboptimal or heterogeneous habitats (Wenderet al. 2004). Thus, differences in vegetativereproduction in Vaccinium in different geograph-ic regions may reflect an adaptive response tohabitat quality or disturbance.
The presence of null alleles at four of the fiveloci may inflate measures of fixation and biasestimates of population differentiation, yet thelow frequency of most nulls and their unevendistribution among populations minimized theimpact on data interpretation. Null alleles areexpected to be higher in frequency when primersdesigned for one species are used to amplifymicrosatellite loci in a related species, as wasdone in this study. However, if the lack ofamplification resulted from fixed differencesbetween V. parvifolium and the species for whichthe primers were designed (V. corymbosum,
highbush blueberry, Boches et al. 2005), wewould expect null alleles to occur at highfrequency in the majority of populations. Threelines of evidence indicate the impact of the nullalleles was minimal and the genetic patternsobserved were robust. First, as the AMOVAconducted for both raw data and that adjustedfor null alleles produced congruent values, theresults are likely conservative estimates of differ-entiation among populations and regions. Sec-ond, the Neighbor-Joining trees built with bothraw data and null-corrected allele frequency datashowed highly concordant topologies. Third, theStructure admixture analysis and individualassignment tests were parameterized to accountfor low frequencies of null alleles (Falush et al.2007) and identified genetic structure concordantto those resolved with other analyses.
A high number of unique alleles were detectedin various Vaccinium populations. A total of 27alleles (26%) were only observed in a singlepopulation, with 12 of the 20 populationsdisplaying at least one unique allele (Supplemen-tal Material). Private alleles were more frequentin the typical coastal and atypical Sierra Nevadapopulations than the atypical populations ofShasta County, indicating gene flow is restrictedamong regions. A relatively small number ofmigrants per generation (Nem 5 4) is sufficient tomaintain genetic similarity between demes (Hartland Clark 2007). Restricted gene flow betweenthe Sierra Nevada and Coast Ranges seems
FIG. 4. Principal coordinate analysis reveals genetic structure among populations of Vaccinium. Each pointrepresents one genet, and symbols correspond to population. Color relates to morphological variation depictedin Fig. 1.
206 MADRONO [Vol. 59
FIG. 5. Patterns of genetic differentiation revealed by admixture analyses. A) The average likelihood of analysesidentified five genetic clusters (dK) in the collection. B) Assignment of individuals to the five genetic clusters. Eachvertical bar represents one sampled population. Color corresponds to each genetic cluster; bars composed ofmultiple colors represent potentially admixed populations. Refer to Table 1 for sample sizes, which vary amongpopulations. C) Neighbor-Joining phenogram of the genetic distances between the five genetic clusters. Colorsmatch genetic clusters from (B).
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 207
reasonable given the geographic distance betweenregions. A lack of gene flow between the CoastRanges and Shasta County may be moresurprising. Some populations of differing mor-phology are closer geographically (e.g., 1SQCand 2SFM separated by 77 km) than are twopopulations of the typical phenotype (e.g., 1SQCand 3MP separated by 425 km). Yet the distantpopulations of similar morphology are signifi-cantly more similar genetically. The lack ofprivate alleles in the Shasta County populationsmay be due to a number of demographic factorsrequiring additional study (e.g., inbreeding). Thelocus CA23F displayed fixed differences betweengeographic regions, with samples from the coastalareas being homozygous for a smaller allele(164 bp) and all other samples from the SierraNevada being homozygous for a larger allele(167 bp). Given the expected high rate ofmutation at SSR loci, the lack of variation atthis locus may indicate it is linked to a functionalvariant under strong selection. Additional studywill be required to determine if fixed allelicdifferences correspond to any adaptive variation.
Systematic Interpretations of Differentiation
The second objective of this study was todetermine whether any genetic differences be-tween populations might be indicative of greatersystematic divergence than currently described inthe taxonomic treatments. Morphologically, V.parvifolium forms a distinct clade separate fromthe V. deliciosum/ovalifolium complex (VanderKloet and Dickinson 1999). Vaccinium parvifo-lium is typically larger than related Vacciniumspecies, and is characterized by slower germina-tion and apical meristem development in seed-lings, and persistent juvenile leaf morphology,distinctive biochemical properties and a taprootsystem unique in the section. Phenotypic variationwithin V. parvifolium is well noted (Vander Kloetand Dickinson 1999; Gehrung 2001). Gehrung(2001) noted the rhizomatous growth and distinctberry color in the Sierra Nevada populations,and suggested elevation to subspecific status maybe appropriate upon further study. However, ifvariation is due to phenotypic plasticity amongpanmictic, we would expect low genetic differen-tiation between morphologically distinct popula-tions, which is not the pattern observed here.Further, phenotypic variation in the related V.membranaceum Douglas ex Torr.was shown to beenvironmentally induced (Schultz 1944 in VanderKloet and Dickinson 1999); common gardenexperiments are required to determine the geneticbasis of morphological variation in V. parvifolium.Vaccinium deliciosum, in contrast, is rhizomatous,forming dense stands of ramets, and displays verylittle phenotypic variation (Gerhung 2001). Ge-netic examinations have also separated these
species into different clades of section Myrtillus.Cytological evidence reported V. parvifolium fromthe Pacific Coast (British Columbia, Canada andWashington state) as diploid (n 5 12) while V.deliciosum was described as tetraploid (n 5 24)(Vander Kloet and Dickinson 1999). Combinedanalyses of nuclear ribosomal ITS and chloroplastmatK sequences clustered V. parvifolium with V.ovalifolium and indicated this clade to be directlyancestral the V. deliciosum clade (Powell andKron 2002).
The genetic differentiation observed in thisstudy indicates greater diversity is present in V.parvifolium than previously reported in NorthernCalifornia. The Shasta County and Sierra Ne-vada populations are more similar genetically tothe congener V. deliciosum than they are to thecoastal populations. Possible taxonomic revisionsmay involve expanding the definition of V.parvifolium to allow variation in berry color andhigh levels of genetic differentiation. Alternative-ly, a novel Vaccinium species or subspecies maybe necessary to accurately reflect the morpholog-ical and genetic variation. Given the divergenceof the Shasta County and Sierra Nevadapopulations from the typical V. parvifolium, thelatter option may be more accurate. Additionalmorphological and ecologic study of thesepopulations is warranted.
Ecological and Conservation Implications
Given the important role of Vaccinium shrubs inconiferous forests (Vila et al. 2004; Wender et al.2004; Hanley 2005), an accurate understanding ofthe species diversity and distribution is necessaryto understand the ecological role of this genus.The genetic differentiation observed among popu-lations of Vaccinium in northern California mayreflect functional or adaptive divergence amonggeographic regions, in particular with regard tofruit color. Avian preference of berry color mayvary temporally and with background foliagecolor (Burns and Dalen 2002; Honkavaara et al.2004). In Rubus spectabilis Pursh, seed germina-tion of different color morphs varies with soiltype, indicating fruit color may correlate withadaptive differentiation (Traveset and Willson1998). In addition to berry color, differences in theextent of clonal growth patterns can affect thefunctional structure of populations. Clonalgrowth forms may allow genets to survive over amuch longer timespan than individual ramets,and can significantly decrease effective populationsize (Persson and Gustavsson 2001). Observation-al and manipulative studies will be required toconfirm if these differences constitute ecologicallysignificant variation.
While V. parvifolium is not a species ofconservation concern, management plans shouldaccount for the distinct gene pools identified in
208 MADRONO [Vol. 59
these Vaccinium collections. The distinct mor-photype from Shasta County is of interest toUSDA Forest Service management plans as it isconsistent with growing evidence of a distinctivebiodiversity in the Shasta Lake region, particu-larly if distinct subspecies or cryptic species aredesignated. Given the significant genetic differ-entiation among geographic groups (coastal,Shasta County, and Sierra Nevada), artificialmovement of germplasm should be restricted towithin and not among regions. These observa-tions are interesting as the Vaccinium berries arefleshy and palatable, and provide forage fornumerous species in coniferous forests (Wenderet al. 2004), though the dispersal ranges are likelysmall relative to the habitat range described forV. parvifolium (Rosatti 2003).
ACKNOWLEDGMENTS
The authors thank A.E.L. Colwell for floral descrip-tions from living and herbarium specimens. A.E.L.Colwell and G. Wallace provided helpful comments onthe manuscript. J. Mello, R. Hernandez, R. Meyer, andR. Hanson conducted laboratory analyses. A.E.L.Colwell, D. Taylor, M. Hutten, T. Carlberg, K.Bainbridge, H. Kelly and E. Rentz conducted tissuecollections. P. Zika provided collections and photodocumentation. M. Goolsby provided assistance withArcGIS. We also thank the U.S. Bureau of Reclama-tion, Mid-Pacific Region, for their support.
LITERATURE CITED
BELL, D. J., L. J. ROWLAND, D. ZHANG, AND F. A.DRUMOND. 2009. Spatial genetic structure oflowbush blueberry, Vaccinium angustifolium, infour fields in Maine. Botany 87:932–946.
BICKFORD, D., D. J. LOHMAN, N. S. SODHI, P. K. L.NG, R. MEIER, K. WINKER, K. K. INGRAM, AND
I. DAS. 2006. Cryptic species as a window ondiversity and conservation. Trends in Ecology andEvolution 22:148–155.
BOCHES, P. S., N. V. BASSIL, AND J. ROWLAND. 2005.Microsatellite markers for Vaccinium from ESTand genomic libraries. Molecular Ecology Notes5:657–660.
BROOKFIELD, J. F. Y. 1996. A simple new method forestimating null allele frequency from heterozygotedeficiency. Molecular Ecology 5:435–455.
BURNS, K. C. AND J. L. DALEN. 2002. Foliage colorcontrasts and adaptive fruit color variation in abird-dispersed plant community. Oikos 96:463–469.
CORNUET, J. M. AND G. LUIKART. 1996. Descriptionand power analysis of two tests for detecting recentpopulation bottlenecks from allele frequency data.Genetics 144:2001–2014.
DEBNATH, S. C. 2009. Development of ISSR markersfor genetic diversity studies in Vaccinium angusti-folium. Nordic Journal of Botany 27:141–148.
FALUSH, D., M. STEPHENS, AND J. K. PRITCHARD.2003. Inference of population structure usingmultilocus genotype data: linked loci and correlat-ed allele frequencies. Genetics 164:1567–1587.
———, ———, AND ———. 2007. Inference ofpopulation structure using multilocus genotype
data: dominant markers and null alleles. MolecularEcology Notes 7:574–578.
FELSENSTEIN, J. 2005. PHYLIP (Phylogeny InferencePackage) version 3.2. Distributed by the author.Department of Genome Sciences, University ofWashington, Seattle, WA.
GEHRUNG, L. 2001. Investigations in the genusVaccinium, section Myrtillus, in northern Califor-nia and southern Oregon, including resolutions ofthe taxonomic status of Vaccinium coccineum Piperby RAPD analysis. M.Sc. thesis, California StateUniversity, Chico, CA.
HAMRICK, J. L. AND M. J. W. GODT. 1996. Effects oflife history traits on genetic diversity in plantspecies. Philosophical Transactions of the RoyalSociety B-Biological Sciences 351:1291–1298.
HANLEY, T. A. 2005. Potential management of young-growth stands for understory vegetation andwildlife habitat in southeastern Alaska. Landscapeand Urban Planning 72:95–112.
HARTL, D. L. AND A. G. CLARK. 2007. Principlesof population genetics. Sinauer Associates, Inc.,Sunderland, MA.
HONKAVAARA, J., H. SIITARI, AND J. VIITALA. 2004.Fruit colour preferences of redwings (Turdusiliacus): Experiments with hand-raised juvenilesand wild-caught adults. Ethology 110:445–457.
KREHER, S. A., S. FORE, AND B. S. COLLINS. 2000.Genetic variation within and among patches of theclonal species, Vaccinium stamineum L. MolecularEcology 9:1247–1252.
KRON, K. A., E. A. POWELL, AND J. L. LUTEYN. 2002.Phylogenetic relationships within the blueberrytribe (Vaccinieae, Ericaceae) based on sequencedata from matK and nuclear ribosomal ITSregions, with comments on the placement ofSatyria. American Journal of Botany 89:27–336.
LAWTON-RAUH, A. 2008. Demographic processesshaping genetic variation. Current Opinion in PlantBiology 11:103–109.
MAYER, K. E. and W. F. LAUDENSLAYER, JR. (eds.).1988. A guide to wildlife habitats of California.State of California, Department of Forestry andFire Protection, Sacramento, CA.
NEI, M. 1972. Genetic distance between populations.American Naturalist 106:283–292.
PEAKALL, R. AND P. E. SMOUSE. 2006. GENALEX 6:genetic analysis in Excel. Population geneticsoftware for teaching and research. MolecularEcology Notes 6:288–295.
PERSSON, H. A. AND B. A. GUSTAVSSON. 2001. Theextent of clonality and genetic diversity in lingon-berry (Vaccinium vitis-idaea L.) revealed by RAPDsand leaf-shape analysis. Molecular Ecology 10:1385–1397.
POWELL, E. A. AND K. A. KRON. 2002. Hawaiianblueberries and their relatives – a phylogeneticanlaysis of Vaccinium sections Macropelma, Myr-tillus, and Hemimyrtillus (Ericaceae). SystematicBotany 27:768–779.
PRITCHARD, J. K., M. STEPHENS, AND P. DONNELLY.2000. Inference of population structure usingmultilocus genotype data. Genetics 155:945–959.
———, X. WEN, AND D. FALUSH. 2007. Documenta-tion for structure software: Version 2.2. Distributedby the author. Department of Human Genetics,University of Chicago, Chicago, IL.
2012] DEWOODY ET AL.: GENETIC STRUCTURE OF VACCINIUM PARVIFOLIUM 209
ROSATTI, T. J. 2003. Jepson online interchange forCalifornia floristics. Fremontia 31(2):23–29.
SAITOU, N. AND M. NEI. 1987. The neighbor-joiningmethod: a new method for reconstructing phylo-genetic trees. Molecular Biology and Evolution4:406–425.
SCHULTZ, J. H. 1944. Some cytotaxonomic and germina-tion studies in the genus Vaccinium. Ph.D. disserta-tion. State College of Washington, Pullman, WA.
TRAVESET, A. AND M. WILLSON. 1998. Ecology of thefruit-colour polymorphism in Rubus spectabilis.Evolutionary Ecology 12:331–345.
VAN OOSTERHOUT, C., W. F. HUTCHINSON, D. P. M.WILLS, AND P. SHIPLEY. 2004. Micro-Checker:software for identifying and correcting genotypingerrors in microsatellite data. Molecular EcologyNotes 4:535–538.
VANDER KLOET, S. P. 2009. Vaccinium. Pp. 515–529 inFlora of North America Editorial Committee (eds.),Flora of North America North of Mexico. Vol. 8.Oxford University Press, New York, NY.
——— AND T. A. DICKINSON. 1999. The taxonomy ofVaccinium section Myrtillus (Ericaceae). Brittonia51:231–254.
VILA, B., F. TORRE, F. GUIBAL, AND J.-L. MARTIN.2004. Can we reconstruct browsing history andhow far back? Lessons from Vaccinium parvifoliumSmith in Rees. Forest Ecology and Management201:171–185.
WALLACE, G. D. 1993. Ericaceae. Pp. 544–567 in J. C.Hickman (ed.), The Jepson manual higher plantsof California. University of California Press,Berkeley, CA.
WENDER, B. W., C. A. HUARRINGTON, AND J. C.TAPPEINERII. 2004. Flower and fruit production ofunderstory shrubs in Western Washington andOregon. Northwest Science 78:124–140.
YAKIMOWSKI, S. B. AND C. G. ECKERT. 2008.Populations do not become less genetically diverseor more differentiated towards the northern limit ofthe geographical range in clonal Vaccinium stami-neum (Ericaceae). New Phytologist 180:534–544.
