A Population Genetic Analysis of Chloroplast DNA in Phacella

Heredity 76 (1996) 143—155 Received 71 May 1995

A population genetic analysis of chloroplast DNA in Phacella

FOSTER LEVY*, JAN IS ANTONOVICSt, JOHN E. BOYNTONt & NICHOLAS W. GILLHAM Department of Biological Sciences, East Tennessee State University, Johnson City, TN 37614 and Departments of Botany and ZooIogy, Duke University, Durham, NC 27708, U.S.A.

Hierarchicalsampling from populations, incipient and recognized varieties within Phacelia dubia and P maculata has revealed high levels of intraspecific polymorphism in chloroplast DNA. Much of the variation is partitioned between populations as evidenced by population- specific variants at fixation in all three populations of P dubia var. interior and in both populations of P maculata. Nine of 16 populations were polymorphic for cpDNA haplotypes. A total of 16 haplotypes was found in a sample of 106 individuals; the most common occurred in eight of the 16 populations and in 31 per cent of the individuals in the entire sample. A phylogenetic analysis revealed four basic plastome types. The two major groups of plastomes were separated by four independent base-pair mutations which suggests an ancient split in the evolution of plastid genomes. Representatives from each major plastome division were found in each of five populations spanning two allopatric varieties of P dubia. The geographical distribution of haplotypes and lack of evidence for recent admixture argue against migration as a source of the polymorphism. It is more likely that the current taxonomic varieties are descendants of a polymorphic common ancestor.

Keywords:chloroplastDNA, genetic diversity, haplotype, intraspecific polymorphisms, population structure.

of these studies have been to outline the amount, Introduction geographical location and genetic nature of varia- Intraspecificvariation in chloroplast DNA (cpDNA) tion, often by comparing cultivars and wild relatives has now been widely documented (Harris & Ingram, (Kochert et al., 1991). However, population structure 1991; Soltis et a!., 1992). Population analyses based is best revealed via a hierarchical sampling design, upon the chloroplast genome are beginning to be e.g. individuals within populations and populations applied to problems in intraspecific phylogeography within species. in plants (Soltis et al., 1992; Hooglander et al., 1993; Widespread occurrence of intraspecific variation Mayer & Soltis, 1994) in ways analogous to the use in cpDNA has opened the door for studies designed of mitochondrial DNA (mtDNA) in studies of to infer mechanisms leading to population differ- animals (Avise et a!., 1987). In addition, data from entiation. Recent studies have used hierarchical cpDNA have been used to examine population sampling to reveal migration patterns that led to the structure within species (Neale et al., 1988; Milligan, recolonization of glaciated habitats in oaks (Petit et 1991; Soltis et a!., 1992; McCauley, 1994; Petit et a!., al., 1994), to provide evidence for isolation by 1994), haplotype diversity within and between popu- distance among populations of Eucalyptus (Byrne & lations (Milligan, 1991; Hong et a!., 1993; Byrne & Moran, 1994) and to show that populations of Tn- Moran, 1994), and to search for correlations fo!ium pratense appear to approach panmixia (Milli- between chloroplast variation and morphological gan, 1991). (Terauchi et a!., 1991) or breeding system (Wolff & The goal of this study was to analyse the popula- Schaal, 1992) characteristics. tion structure of a subspecifically differentiated Many detailed examinations of intraspecific varia- species and its nearest relative. The species, Phacelia tion in cpDNA concern agricultural crops. The goals dubia (L.) Trel., consists of three allopatric varieties and two incipient varieties, all of which differ *Correspondence morphologically and which show patterns of infertil-

1996 The Genetical Society of Great Britain. 143 144 F. LEVY ETAL. ity when intercrossed (Levy, 1991a,b). More specific- Carolina) between the known ranges of the recog- ally, we used data on population structure and nized varieties dubia and georgiana. 'Railroad' haplotype diversity to infer mechanisms underlying consisted of two large populations in southwestern plastome differentiation and to suggest possible Virginia, within the range of and in proximity to forces responsible for observed patterns. We exam- dubia populations (Levy, 1991a). Although partially ined evidence for gene flow among populations and reproductively isolated from dubia by a nuclear-cyto- whether the distribution of variants may best be plasmic barrier, these plants did not differ attributed to random genetic drift, selection, and/or morphologically from dubia (Levy, 1991a). historical factors such as random sorting of plastome Plants for cpDNA extraction were grown from lineages during taxon divergence. seeds which were collected in the field in 1987 and The study has uncovered some unexpected 1988. All plants were grown in the Duke University patterns. P dubia harbours a surprisingly high level Phytotron. Chloroplast DNA was analysed in plants of genetic diversity within each level (varieties, incip- from four populations of P maculata and from two ient varieties, populations) of the intraspecific hier- to four populations within each of the five varieties archy. There was evidence for an ancient plastome of P dubia (three recognized varieties: dubia, polymorphism whose current representatives coexist interior, and georgiana; two incipient varieties: 'rail- within several allopatric populations. A phylogeny of road' and 'imitator'). Population designations corre- plastid haplotypes and their distribution within taxa spond to those in Levy (1991a) with the addition of indicated taxon divergence probably proceeded from g-4, Douglas Co., GA (georgiana); m-3, Edgefield a polymorphic ancestor. Co., SC (P maculata); m-4, Gwyinette Co., GA (P maculata). Materials and methods Chiorop/astDNAisolation Study organism Themethods of cpDNA extraction and purification Phacelia dubia was chosen to examine the distribu- were similar to those described in Palmer (1982) and tion of intraspecific variation in cpDNA because Kemble (1987). Chloroplast DNAs that produced populations are spatially discrete and subspeciflc partial banding patterns suggestive of incomplete varieties express a broad range of geographical, digestion were further purified in a 1 mL column of ecological and morphological variation (Murdy, 100—200 mesh agarose beads (Bio-Rad A-50) 1966; Levy, 1991a). The varieties within P dubia followed by ethanol precipitation. constitute a monophyletic group which exhibits complete reproductive isolation from P maculata, Restriction their nearest relative (Gillett, 1964; Murdy, 1966). enzyme analysis Reproductive isolation among the subspecific taxa Restrictionfragments were separated on 0.8 per within P dubia is incomplete. cent agarose gels at 50 mA current for 12 to 16 h in The three allopatric, formally recognized taxo- 1 x TBE buffer. The gels were stained in 1 mg/L nomic varieties of P dubia are P dubia var. dubia ethidium bromide for 30 mm, destained in three (=dubia),P dubia var. georgiana McVaugh changes of distilled water for 30 mm each, and (=georgiana)and P dubia var. interior Fernald photographed on Kodak Tri-X film with a Polaroid (=interior).Based upon morphological, geographi- camera. Banding patterns were analysed visually. cal and reproductive relationships, dubia has been To facilitate visual scoring, a preliminary survey hypothesized as the progenitor of the other two var- conducted on a small subset of cpDNAs was carried ieties (Murdy, 1966; Levy, 1991a,b). out to identify restriction enzymes that (a) generated Two incipient varieties were also included in this banding patterns free of partially digested fragments study to determine (a) the extent of intrapopulation and (b) produced a moderate (15 to 30) number of polymorphisms, (b) levels of differentiation from fragments whose bands were evenly distributed recognized varieties, and (c) their phylogenetic rela- throughout the gel. Twenty-two enzymes were tionships to recognized varieties. The incipient vari- tested; 18 with six base-pair recognition sequences eties were designated P dubia 'imitator' (ApaI, BamHI, BclI, BglI, BglII, BstEI, ClaI, EcoRI, (='imitator')and P dubia 'railroad' (='railroad'). EcoRV, HindIII, KpnI, PstI, PvuII, Smal, SstI, SstII, Populations of 'imitator' are partially reproductively XbaI, XhoI), one with a five base-pair recognition isolated from both dubia and georgiana (Levy, sequence (Avail), and three with four base-pair 1991b) and they are located in a region (South recognition sequences (HaeIII, HhaI, HpaII). Five of

The Genetical Society of Great Britain, Heredity, 76, 143—155. CHLOROPLAST DNA IN PHA CELIA 145 these (BamHI, BglII, Hindili, SstI, XbaI) were used dent mutations and whether they mapped to neigh- to estimate genome size and length of the inverted bouring regions of the genome (as expected when repeat on each of four cpDNAs. Using six restriction length mutations cause the simultaneous loss or gain enzymes (Avail, BamHI, BglII, EcoRV, HhaI, of nearby restriction sites). Because each probe was Hindlil), plastome type was determined for all 122 derived from a variant fragment and each probe was cpDNAs and complete haplotypes were obtained for hybridized to samples digested with the panel of six 106 of the cpDNAs. restriction enzymes, variation was expected among cpDNAs digested with the restriction enzyme giving rise to the probe fragment. If hybridization of a Scoring variation particular probe also revealed variation in cpDNAs Base-pairsubstitutions were inferred when the digested with restriction enzymes other than the one combined length of two fragments present in one used to generate the probe, then the two variants cpDNA sample but absent in another, summed to must be in close physical proximity in the genome; the length of a newly visible fragment in the second that is, both variants were either encompassed by or sample. This pattern would correspond to a differ- adjoined DNA corresponding to the probe. ence of a single restriction site between the cpDNA in the two lanes. Several suspected base-pair substitutions were confirmed by filter hybridizations. Populationstructure When a fragment was either gained or lost, but the Ahierarchical sampling design (individuals within pattern corresponding to a simple base-pair substitu- populations; populations within varieties; varieties tion was not evident, this pattern was scored as a within species) was employed to facilitate an analysis presence/absence variant. Presence/absence variants of population structure. Between-population compo- may have arisen by length mutations such as inser- nents of haplotype diversity were estimated separa- tions or deletions but some were probably also the tely for each of the three recognized varieties of P result of base-pair substitutions in which one dubia as well as for the entire P dubia complex of cpDNA fragment was too small to detect. populations (including populations of the two putative varieties). The measure of population structure, theta (8), was estimated from a haploid model as Partialmapping of variant fragments outlined by Weir (1990). A mean value for theta and Apartial mapping procedure was used to distinguish its standard deviation were computed by jackknifing whether the changes associated with different vari- over populations using a computer program written ants were located in physical proximity to each other by B. Weir. on the genome. Two cpDNA samples may differ by several RFLPs, but if these variants map to a similar inference region of the genome it may indicate that they have Phylogenetic all arisen by a single mutation event, such as a large Phylogeneticrelationships among plastid haplotypes insertion or deletion. Five plastome-specific variants were assessed in a parsimony analysis using were partially mapped. In this procedure the largest unweighted characters in the branch and bound variant fragment in a particular enzyme digest was algorithm of PAUP (Swofford, 1991). The plastomes electroeluted either in dialysis tubing (slices of the in P maculata served as an outgroup for plastomes gels containing cpDNA fragments were electrophor- in P dubia. A goal of this study was to determine esed in 0.5 x TBE buffer for 1 h) or in an electro- whether intraspecific polymorphism occurred in eluter apparatus (IBI Model UEA) followed by cpDNA and if so, how the polymorphism may influ- purification (Elutip D columns, Schlicter & Schull) ence phylogenetic inferences. The chloroplast and ethanol precipitation. The recovered fragments genome has been believed to evolve more slowly were nick translated (BRL Nick Translation System than nuclear DNA (Wolfe et al., 1987) and to exhibit no. 8160SB) with 32P and used as probes. These little intraspecific variation (Palmer, 1987; Palmer et radioactively labelled probes were hybridized to al., 1988). Consequently, in many phylogenetic nitrocellulose filters onto which samples of cpDNAs studies using cpDNA, only one or a few plants of a encompassing several different variants were blotted species have been considered adequate as represen- following digestion of each with a panel of six tatives of a species. Two phylogenetic analyses were restriction enzymes. performed to investigate potential effects of limited The following strategy was used to determine intrataxon sampling. First, the plant with the most whether different variants resulted from indepen- frequent cpDNA haplotype within each recognized

The Genetical Society of Great Britain, Heredity, 76, 143—155. 146 F. LEVY ETAL.

Phacelia taxon (dubia, georgiana, interior, P macu- variant (Table 1; no. 6) restricted to plastome 0. In lata) was chosen as a representative. For example, addition to the five variants shared with plastome D, dubia was represented by haplotype D-4 (Table 4); plastome I differed from plastome D by two pres- in georgiana, 17 of the 27 individuals carried plas- cence/absence variants (Table 1; nos 12, 13). Radio- tome G and haplotype G-3 was chosen because it active probes derived from cpDNA of five was most common within plastome G. In an alter- plastome-specificfragments were sequentially native analysis, georgiana was represented by a plant hybridized to Southern blots of cpDNA digests carrying the most frequent haplotype in that variety produced by the same enzyme giving rise to the regardless of the plastome type (haplotype D-1; see respective variant fragments. Four of the probe Table 4). DNAs distinguished the D—I from the G—M lineage (Table 1; nos 1, 2, 4, 5) and one (Table 1; no. 9) separated plastome M from the other three plas- Results tomes. In four cases, the observed variation could be attributed to the loss or gain of a restriction site, Genome organization presumably by a single base-pair substitution. These The genome size was estimated as 129.9 kb and the five fragments encompassed 33.4 kb of the chlor- length of the inverted repeat was estimated as oplast genome (AvaIl, 9.0 kb and 2.8 kb; EcoRV, 20.0 kb, both of which are within the range of most 2.6 kb; HhaI, 13.5 kb; HindIlI, 5.5 kb). flowering plants (Palmer, 1985). For each hybridized probe, the banding patterns Each of 122 cpDNA samples was digested with six of representatives of each plastome were identical restriction endonucleases; the four enzymes with six except in hybridizations to digests by the enzyme base-pair recognition sequences (BamHI, Bg1II, which generated the probe fragment. These hybrid- EcoRV, HindIII) produced a total of 100 scorable izations indicated that variant fragments produced restriction fragments, the five-base cutter (Avail) by different restriction enzymes must have been generated 32 scorable fragments, and the four-base located in different regions of the chloroplast cutter (HhaI) generated 31 scorable fragments. genome. Hence, given the extent of the region used Approximately 884 bp were sampled which repre- as probe DNA (33.4 kb), the location of variant frag- sents 0.7 per cent of the 129.9 kb Phacelia chlor- ments in different regions of the genome, the lack of oplast genome. apparent major differences in genome size among these cpDNAs, and the evidence that four of these Plastomevariation variants were resulted from base-pair substitutions, it is unlikely that differences among the four plas- Eachof the 122 cpDNAs could be characterized as tomes resulted from a single rearrangement but possessing one of four basic restriction patterns; rather that they differed by several independent each of the four patterns is characterized by a point mutations and possible length mutations. unique suite of restriction fragments. Differences among these four plastid types (subsequently referred to as plastomes and designated by the Distributionof the p/astomes among taxa upper case letters 'D', 'I', 'G', and 'M') were Thedistribution and frequencies of the four plas- observed in digestions with each of the six enzymes tomes in the various taxa (Table 2) shows that plas- analysed. The thirteen variants that distinguished tome M was unique to P maculata and all plants in the four plastomes, and the nature of these variants that taxon carried plastome M. Plastomes D, G and are presented in Table 1. Plastome variants observed I were restricted to P dubia; plastome I was following digestion of cpDNAs with Avail are illus- restricted to and fixed within populations of interior; trated in Fig. 1. plastome D was fixed in the two putative varieties, A set of four base-pair substitutions (Table 1; nos 'railroad' and 'imitator'. Plastome D was found in all 1, 2, 4, 5) and one presumed length mutation (Table but one of the 28 representatives of dubia and it was 1; no. 3) divides the entire sample into two major present at varying frequencies in all four populations groups, one encompassing plastomes D and I and of georgiana (Table 2). Plastome G was represented the other consisting of plastomes G and M. Within in samples from all populations of georgiana (Table the G—M group, plastome M carried five additional 2). The wide range of plastome frequencies among variants that distinguished it from all other plas- georgiana populations may reflect variation arising tomes (three base-pair substitutions, two presence! from small sample sizes, but the presence of both absence bands) (Table 1; nos 7—11); there was one plastomes D and G within all sampled georgiana

The Genetical Society of Great Britain, Heredity, 76, 143—155. CHLOROPLAST DNA IN PHA CELIA 147

Table 1 Variants characteristic of the four Phacelia plastomes

Plastome

Code no. Enzyme M G D I Variant

1 AvaII* — — + + 5.3+3.7-+9.0 2 AvaII* — — + + 2.8-+2.4+0.4 3 BamHI — — + + 3.9-2.6 4 ECORV* — — + + 2.6-2.1+0.5 5 HindIII* — — + + 5.5-+5.2+0.3 6 HhaI — + — — 3.8—*2.2+1.6 7 AvaIl — + + + 4.0—4.3 8 EcoRV — + + + 14.0+9.0—dl.0+12.0 9 HhaI* — + + + 12.0-j-1.5--13.5 10 HhaI — + + + 9.8—8.0+1.8 11 HindlIl — + + + —2.5 12 AvaIl — — .— + 4.2—*4.3 13 BglII — — — + +2.0 The changes indicated under 'variant' refer to differences between plastome M and other plastomes. A plus indicates that the change listed under 'variant' occurs in a particular plastome. Single entries under 'variant' signify presence or absence of fragments not attributable to restriction site changes or obvious length mutations. DNA fragment lengths are in kilobase pairs. *Variants confirmed by Southern analysis.

populations (Table 2) suggests that both are 123456789 common throughout the range of georgiana. Taxonomicand geographical distribution of hap/otypes

•'ø— 23.1 Completehaplotypes were determined for 106 of 9.0 —s- •'— 9.0 the 122 samples. In addition to the 13 diagnostic -0--- 6.7 plastome variants, 16 additional variant banding 5.3 —s.- patterns (referred to as nonplastome variants and .ø— 4.4 designated by the lower case letters a—q) were 3.7 —s- detected (Table 3). One pair of these (a, e) co-occurred in 26 cpDNA samples from plants represent- 2.8 —s- ing members of five populations that spanned three 2.4 —0 '0— 2.3 taxonomic varieties (haplotypes D-2, D-3, D-4 in — 2.0 Table 4). A second pair (f, g) was only found in a single plant and to provide conservative estimates of

Fig. 1 Plastome-related variation observed following Avail digestion. Chloroplast DNAs from individual plants from population d-l (lanes 1—3), g-3 (lanes 4—6) and g-l (lanes 7—8). Plastome D has a 9.0 kb fragment (lanes 1, 2, 3, 7) which, in plastome G (lanes 4, 5, 6, 8) has an inter- -m--— 0.6 nal Avail recognition site splitting the 9.0 kb fragment into 5.3 kb and 3.7 kb fragments. Plastome G has a 2.8 kb fragment which is split into 2.4 and 0.4 fragments in plastome D. Note the intrapopulational plastome polymorphism in g-l (lanes 7, 8). Size markers are in lane 9.

The Genetical Society of Great Britain, Heredüy, 76, 143—155. 148 F. LEVY ETAL.

Table 2 Frequencies of the four plastomes in Phacelia by independent mutations. Sixteen different haplo- taxa and the frequencies of these plastomes in populations types were distinguished (designated by the upper polymorphic for two plastomes case plastome letter code and a genotype number): six were variants of the D plastome, four were vari- Plastome ants of the I plastome, there were four variants within the G plastome and two variants within the M Taxon N D G M I plastome (Tables 3, 4; Fig. 3a). P dubia var. dubia 28 0.96 0.04 — — In several instances, identical haplotypes (for Pdubiavar.interior 21 — — 1.0 example, haplotypes D-1 and G-2) occurred in indi- P dubia var. georgiana 29 0.34 0.66 — vidual plants collected from two different taxonomic P dubia var. 'railroad' 15 1.0 — — varieties (Table 4). One population within interior P dubia var. 'imitator' 16 1.0 — — (i-2) and one of P maculata (m-3) were fixed for P maculata 13 — — 1.0 unique variants, that is, all members of these populations carried a variant not encountered in any Polymorphic populations — other population. In a georgiana population (g-3), dubia d-2 8 0.87 0.13 which was polymorphic for plastomes D and G, all 9 0.22 0.78 — — georgiana g-1 seven individuals with plastome G carried a unique georgiana g-2 9 0.67 0.33 — georgiana g-3 9 0.11 0.89 — — variant (Table 4). georgiana g-4 2 0.50 0.50 — N =numberof individuals sampled. Populationstructure Nineof the 16 populations surveyed were polymorphic for chloroplast DNA haplotypes (Table 4). Five Table 3 Variation within plastomes in Phacelia populations (one of the three dubia populations Variant code (d-l) and all four georgiana populations) contained Enzyme Variant Plastome members with different plastomes; plastomes D and Avail a 2.6—*2.5 D G were represented in each of these five populations Avail b 2.0—1.6 I (Table 2). Six populations showed variation within a BamHI c 4.6-*2.4+2.2 G particular plastome and in three populations, there BglII d +3.8 G was a single individual with a unique variant BglII e 3.4+1.9—1.3 D (Table 4). BglII f 8.2+4.2-+15.0 G Considering the entire sample, 10 of the 16 haplo- EcoRV g 5.8+2.0-7.8 G types were rare, that is, they occurred with a EcoRV h 2.3—2.2 G frequency of less than 0.05 (Fig. 2). One haplo- EcoRV i +0.9 I type (D-1) was by far the most common EcoRV j 1.8—*1.0+0.8 D,I (frequency = and this haplotype was also the HhaI k 2.6—*2.3 I 0.31) most widely distributed, occurring in eight popula- HhaI 1 —5.0 D,G HhaI m —2.0 D tions that spanned two recognized and two incipient HhaI n 3.0—2.5 G varieties of P dubia (Table 4). In contrast, most (11 HhaI p 3.0—*3.2 D, I of 16) haplotypes were restricted to single popula- Hindu! q 2.6—2.7 D, M tions (Fig. 2). In P dubia, most of the haplotype diversity was These variants occur in addition to the respective distributed between populations (0 =0.61; diagnostic variants of each plastome. Single entries under SD = However, the three recognized varieties 'variant' signify presence or absence of fragments not 0.11). attributable to restriction site changes or obvious length showed very different levels of substructuring. mutations. DNA fragment lengths are given in kilobase Substructuring was greatest in var. interior where all pairs. haplotypes were population-specific(0 =0.86; SD =0.15). Substructuringwas lowest within var. dubia where a single haplotype (D-4) was the most divergence in phylogenetic analyses, was also frequent within each of the three populations assumed to be associated. No other pairs of variant (0 =0.19;SD =0.24).The level of substructuring fragments showed obvious associations. Conse- within var. georgiana was intermediate between the quently, each of the 14 remaining variants and the other two varieties (0 0.45; SD =0.24)which two associated pairs were considered to have arisen reflects a combination of unique haplotypes (G-3,

The Genetical Society of Great Britain, Heredity, 76, 143—155. Table 4 Distribution of cpDNA haplotypes in populations of Phacelia Taxon

dubia georgiana interior 'railroad' 'imitator' P maculata Haplotype totals

Population: d-1 d-2 d-3 g-1 g-2 g-3 g-4 i-i i-2 i-3 r-1 r-2 p-i p-2 rn-i m-2 No. of pops. No. of mdiv.

2. Haplotype C D-1 1 3 2 6 1 7 6 7 8 33 CD 7 2 9 D-3 2 1 2 D-4 3 3 8 1 4 15 D-5 1 1 1 2 1

2 1 2 1-2 7 1 7 1-3 5 1 5 1 1 1

1 3 1 3 5 1 4 2 5 1 7 1 1 1

M-1 7 1 7 M-2 4 1 4

Entries refer to the number of individuals of each haplotype. C-) I-I 0 0 >I- HC/) z0 z

A (0 150 F. LEVY ETAL.

0.4, 10 10

.280. 8 0 6 n04 4 E = 2 Fig. 2 (a) Haplotype frequencies in the 0 entire sample from Phacelia dubia. (b). <0.005 0.06- 0.11- 0.16- 0.21-0.26- 0.31- 1 2 3 4 5 6 7 8 0.10 0.15 0.20 0.250.30 0.35 Frequency distribution of haplotypes Haplotype frequency Number of populations among populations of P dubia.

G-4)and a haplotype (D-1) that was shared among (Scowcraft, 1979; Palmer & Zamir, 1982). More three of the four populations (Table 4). recent and thorough analyses of intraspecific cpDNA variation suggest that polymorphism is common, but that the level of variation depends Phylogeneticanalyses upon the particular species (Hong et a!., 1993). In Theparsimony analysis resulted in one shortest tree the present study, extensive variation was found which required 33 steps to assign 29 characters within P dubia and P maculata. Nine of 16 popula- (consistency index =0.88)(Fig. 3a). Four characters tions were polymorphic for cpDNA haplotypes showed homoplasy; three (1, p, q) resulted from (Table 4). The relatively high level of interpopula- parallelisms, one (j)wasa reversal. A suite of five tion polymorphism in Phacelia was comparable to mutations (Table 1, nos 1—5) separated plastomes D levels observed in Hordeum vulgare (Neale et al., and I from plastomes G and M. Five additional 1988; 10 of 19 populations polymorphic) and Euca- variants characterized plastome M. Haplotype- lyptus nitens (Bryne & Moran, 1994) but it was not specific apomorphies distinguished lineages within as high as levels observed in Tnfolium pratense plastome types (Fig. 3a). (Milligan, 1991). Lower levels of polymorphism The analysis using a single plant to represent each characterized European oaks (Petit et a!., 1994) and taxon (using the most frequent haplotype of the Lupinus texensis (Banks & Birky, 1985). most common plastome within each of the taxonom- A large proportion of the cpDNA diversity in 1? ically recognized taxa) indicated a sister group rela- dubia occurred as between-population differences tionship between georgiana and P maculata. Thus, (0 =0.61).A similar pattern of relatively high the taxon P dubia would be considered paraphyletic. genetic diversity primarily partitioned between The resultant tree of length 23 (six of 29 characters populations was also observed in H. vulgare where were constant) showed no homoplasy (Fig. 3b). Esti- 8 =0.67(Milligan, 1991, from data in Neale, 1988) mates of relatedness were drastically altered when and in bishop pine (0 >0.80; Hong et at., 1993). In the individual chosen to represent georgiana was one contrast, extremely low between-population differ- carrying the most frequent haplotype (regardless of entiation (0 =0.02)for cpDNA haplotypes was plastome type) in that variety, haplotype D-1. In this found in the highly polymorphic T pratense, a analysis, a lack of phylogenetically informative pattern which is suggestive of panmixia (Milligan, characters resulted in an unresolved polytomy (Fig. 1991). 3c). This simple comparison shows that tree topol- Haplotype frequency distributions also vary ogy and distances can differ greatly depending upon greatly among species: a single haplotype dominated the sampling scheme for individuals within taxa. the Lupinus survey (Banks & Birky, 1985), occurring in 88 per cent of the 100 individuals surveyed; in the current study, the most common of the 16 Phacelia Discussion haplotypes was found in 31 per cent of the 106 individuals haplotyped, and in a sample of 245 Intraspecific polymorphism plants of Hordeum, each of the three haplotypes Until recently, it has been an unwritten dogma that comprised 30—40 per cent of the sample (Neale et cpDNA shows little intraspecific variation, even at., 1988). The most frequent haplotype in Euca- though the earliest surveys of cpDNA variation in lyptus occurred in only 19 per cent of the individuals natural populations detected intraspecific variation (Bryne & Moran, 1994). Haplotype frequencies

The Genetical Society of Great Britain, Heredity, 76, 143—155. CHLOROPLAST DNA IN PHA CELIA 151

fl-I () D-2 fl-S fl-4 D-6

D-5

I—i

1-3 1-4 1-2

G-1 6-2 (3-3 6-4

M-1 M-2

• CI m •.l• dubla(D-4) (c) georgiana (0-I) 1213 ae Ini II I Interior (1-2) dubla(0-4) II12 13 •I georgiana (6-3) I I —— interior (1-2) 1 234 5 7 8 9 10 11 I I I I P. maculata (M-1) P. maculata (M- 1) I I I III I I I I

Fig.3 Cladograms showing relationships (a) among haplotypes; (b) among recognized taxa using the single most common cpDNA haplotype in each taxon to represent that taxon; (c) same as (b) but replacing the georgiana individual with one carrying the D plastome that occurred in georgiana populations. Variants that show homoplasy are shown as open (parallelisms) and gray-shaded (reversal) rectangles. Variant codes correspond to those in Table 1 (numeric) and Table 3 (alphabetic).

varied greatly among very local populations of Silene conclusion must be that intraspecific polymorphism alba (McCauley, 1994). Plastome variants that were can be found in most species for which extensive unique to single populations and fixed within those samples have been analysed. Species differ in the populations were found in two populations each of number of haplotypes and their partitioning within P dubia and I? maculata, also in two populations of and between populations as demonstrated by differ- Eucalyptus (Bryne & Moran, 1994), and in one ences among the closely related closed-cone pines in population of Lupinus (Banks & Birky, 1985). In California (Hong et a!., 1993). The amount of Hordeum, several populations lacked cpDNA haplo- genetic difference between haplotypes within species types that were common in populations less than may be small as is often observed, but in P dubia 30 km distant (Neale et al., 1988). and Coreopsis grandijiora (Mason et a!., 1994) haplo- Do data from the limited number of detailed types differed by as many as 14 and 19 mutations, intraspecific investigations suggest generalizations respectively. The species-specific nature of cpDNA regarding the amount and partitioning of cpDNA variation should not be surprising; levels of polymor- variation within species? The most noteworthy phism depend upon the mutation rate and effective

The Genetical Society of Great Britain, Heredity, 76, 143—155. 152 F. LEVYETAL. population size. Although the chioroplast genome is Co-occurrence of the highly divergent D and G plas- thought to evolve slowly, little is known about its tomes in different reproductively isolated varieties of mutation rate other than that there is an apparently P dubia suggests a long-standing polymorphism or higher rate for length variants compared to point recent migration. If the derivation of plastome types mutations (Hong et al., 1993). In contrast, extensive was indeed ancient, we must enquire why several data exist that show that species differ in all compo- populations remain polymorphic. Three alternatives nents (e.g. census size, mating system, migration should be considered: polymorphisms within popula- rate, etc.) of effective population size. In fact, the tions may result from recent migration, they may be species-specific nature of cpDNA variation is similar actively maintained by selection, or they may be to that found in analyses of nuclear-encoded allo- neutral. If neutral, polymorphisms are expected to zyme loci (Hamrick et al., 1979). Species specificity be transient and the frequency of a specific haplo- may preclude generalizations but it does facilitate type would reflect its point in a particular trajectory the use of cpDNA to infer population processes as of genetic drift. Drift trajectories within populations demonstrated below. are independent and we would not expect identical haplotypes to have similar frequencies in different populations. Although sample sizes within popula- Geographicalpatterns and genotypic distributions tions are small, inspection of the data shows that Aviseetal.(1987) distinguished five distributional frequencies of individual haplotypes differ greatly patterns of mtDNA in populations. In the most among varieties and among populations within vari- common (Category I), phylogenetic occurrence of eties (Table 4). The drift model cannot be ruled out. genomes was correlated with geographical separa- tion of these genomes. The restriction of Phacelia Migration. All four haplotypes in interior and both cpDNA plastomes I and M to populations of interior haplotypes in P maculata were population-specific and M maculata, respectively, exemplify this pattern and they were each fixed within their respective and it is analogous to the geographical partitioning populations. The uniqueness of each of these popu- of mtDNA haplotypes into isolated river drainages lations is maintained in spite of physical proximity to in several species of fish (Bermingham & Avise, other populations; within the respective taxa, popu- 1986). lations were only 10—30 km distant from other The observation of identical cpDNA haplotypes sampled populations. Furthermore, the 'private' (i.e. occurring as polymorphismswithin several restricted to a single population; Slatkin, 1985) geographically isolated populations is an unusual haplotypes represent a wide spectrum of population pattern (Category II of Avise et al., 1987). In the frequencies (Table 4). If these private haplotypes warmouth sunfish (Lepomis gulosus), Bermingham & represented recent mutations, most would be Avise (1986) found two individuals in the western expected to occur at low frequencies within their region of the range that had mtDNA typical of respective populations. Using Slatkin's (eqn 3 in eastern populations. In plants, coexistence of indivi- Slatkin, 1985) method based upon rare alleles, the duals with highly divergent chioroplast plastomes estimateof Nm =0.03 (N =populationsize; was recently reported in Coreopsis grandiflora m =migrationrate) for all populations within P (Mason et al., 1994). If we consider the possibility dubia is far below the value of one; it implies that that, at least for Phacelia, intraspecific polymor- migration is not sufficient to prevent population phism in cpDNA is the rule rather than an excep- differentiation, a conclusion that is consistent with tion, then several phenomena must be accounted the finding that most haplotype variation occurred for; the relative time and taxonomic origin of the between populations (0 =0.61).When Nm is esti- major plastome types, the current distribution mated from q (using eqn 3 in Ennos, 1994), the pattern of haplotypes within and between popula- estimate of Nm =0.16for all populations within P tions and the possible causes of monomorphism dubia is also sufficiently low to expect subdivision. within populations. Most strikingly, patterns in Phacelia are quite Plastome differences in Phaceliawerecaused by different from those in Trifolium where only 3 per several apparently independent base pair mutations. cent of the haplotype diversity occurred between In light of the relatively low evolutionary rate for populations and Nm 5.9 was consistent with higher cpDNA, and assuming a similar rate among these levels of gene flow (Milligan, 1991). closely related lineages, the number of mutations Although the possibility of historical introgression separating plastosomes strongly suggests a relatively can never be eliminated, morphological patterns, ancient origin for the D/I—G/M plastome split. reproductive relationships and analysis of nuclear-

The Genetical Society of Great Britain, Heredity, 76, [43—155. CHLOROPLAST DNA IN P/-IA CELIA 153 encoded allozymes provide further evidence that sizes, including data on spatial structuring within recent migration is an unlikely cause of the D—G populations and data on the relative fitnesses of plastome polymorphism (Murdy, 1966; Levy, individuals with different genomes under different 1991a,b). More likely explanations for the distribu- conditions would help distinguish between an tion of polymorphisms in P dubia should be based actively maintained polymorphism and one whose upon historical factors such as a relatively recent frequencies are governed primarily by drift. derivation of dubia and georgiana from a polymorphic common ancestor or hypotheses that postulate selective differences among plastomes. Phylogeneticimplications Ancesttyof putative varieties. The chloroplast Selection. Selective differences among plastomes genomes of both incipient varieties ('imitator' and may be hypothesized to account for the widespread 'railroad') of Phacelia dubia were clearly derived plastome polymorphism: for example, one plastome from a D plastome. The common cpDNA genotype may be favoured over the other in different micro- in 'railroad', found in all but one individual sampled, habitats or seasons. Is it plausible that selective occurred as part of a polymorphism in a nearby differences among organelle genomes underlie levels dubia population (d-2). The two 'railroad' popula- of polymorphism? There are no data that show tions have probably arisen from a dubia ancestor fitness differences among plants that differ in chlor- carrying the D plastome. In contrast, the two popu- oplast genotypes but recent analyses of mitochon- lations of 'imitator' may differ in their evolutionary drial DNA sequence diversity in Drosophila are not histories. Despite their relationship to individuals consistent with neutrality; the observed patterns are with D plastomes, the two populations did not share more suggestive of selection and variability of evolu- a cpDNA haplotype. However, the predominant tionary rates among mitochondrial lineages (Ballard haplotype in each population was also found in a & Kreitman, 1994; Rand et al., 1994). North Carolina population of dubia (population d-1). The common haplotype in 'imitator' population Neutral expectations. More generally, patterns of p-i was also the most common haplotype in the chloroplast genome partitioning and subsequent entire survey, occurring in two dubia and three lineage sorting following speciation can explain the georgiana populations (Table 4) but the haplotype plastome polymorphism within populations of dubia fixed in population p-2 was only found in a single and georgiana. Theoretical (Tajima, 1983; Nei, 1986) dubia population (population d-1). and numerical (Avise et at., 1984; Neigel & Avise, 1986) studies have shown that for animal mtDNA, Evolutionaiy relationships of recognized taxa. Al- polyphyly and paraphyly are not unexpected patterns though there have been relatively few detailed in recently diverged taxa. Polyphyletic patterns, in studies of intraspecific diversity in cpDNA, the avail- which a genome from one population is more closely able data suggest that population genetic character- related to a genome from a distant population than istics of the chloroplast genome are not unlike other it is to a genome of a neighbouring individual, do genomes. That is, not surprisingly, mutations prob- not necessarily imply admixture or migration ably arise randomly within populations, they spread (Tajima, 1983; Nei, 1986; Pamilo & Nei, 1988). If an via migration, and drift and/or selection govern ancestral stock was polymorphic, a natural progres- frequencies within populations. Furthermore, even sion is expected from polyphyly through paraphyly with a limited number of studies, intraspecific to monophyly (Tajima, 1983; Avise et al., 1987). A surveys clearly indicate that different species display progression similar to this was observed in Peromys- a wide spectrum of absolute levels of variation and cus (Avise et at., 1983) and Hawaiian Drosophila these species vary in the relative partitioning of vari- (DeSalle et at., 1986). ation within and between populations. However Historical forces and selection appear to be the trivial these conclusions may appear, they do high- two most likely causes of the current cpDNA poly- light the need for intraspecific analyses prior to morphisms in populations of Phacelia dubia. In analyses at higher taxonomic levels. In this regard, either case, evidence against either migration or cpDNA is not unlike any other taxonomic character. recent admixture suggests that polymorphism pred- The phylogeny of plastomes illustrated a clear ates taxon divergence. Fragmentation of a once evolutionary relationship in which plastomes D and I more widespread and continuous habitat may represent a distinct dade from plastomes G and M explain the current distribution of identical haplo- (Fig. 3a). However, caution must be exercised prior types within different populations. Larger sample to equating the plastome and taxon phylogenies.

The Genetical Society of Great Britain, Heredity, 76, 143—155. 154 F. LEVY ETAL.

Intraspecific variations, as exemplified by the D—G BANKS, 1. A. AND BIRKY, C. W., JR. 1985. Chloroplast DNA plastome polymorphism, demonstrated two serious diversity is low in a wild plant, Lupinus texensis. Proc. problems in taxonomic phylogeny reconstruction Nail. Acad. Sci. US.A., 82, 6950—6954. based solely upon cpDNA. Inadequate sampling BERMINGHAM, E. AND AVISE, S. C. 1986. Molecular zoo- could lead to an erroneous estimation of taxonomic geography of freshwater fishes in the southeastern United States. Genetics, 113, 939—965. relationships and, second, a phylogeny of genes is BYRNE, M. AND MORAN, ci. F. 1994. Population divergence not necessarily concordant with a phylogeny of taxo- in the chloroplast genome of Eucalyptus nitens. Heredity, nomic entities. The latter can be especially trouble- 73, 18—28. some when a phylogeny is derived from cpDNA DESALLE, R., GIDDINGS, L. V. AND KANESFIIRO, L. y1986. because the chloroplast genome is essentially asexual Mitochondrial DNA variability in natural populations and haplotypes are therefore best treated as alleles of Hawaiian Drosophila. II. Genetic and phylogenetic of a single locus (Doyle, 1992). If population surveys relationships of natural populations of D. silvestris and identify new haplotypes at a rate of one per eight to D. heteroneura. Heredity, 56, 87—92. 25 individuals, then sample sizes should reflect these DOYLE, j.i.1992. Gene trees and species trees: Molecular expected levels of polymorphism. Of particular rele- systematics as one-character taxonomy. Syst. Bot., 17, vance to the situation in R dubia is the fact that [44—163. discordant phylogenies are most likely to be ENNOS, R. A. 1994. Estimating the relative rates of pollen observed when the taxa are recently derived. Plas- and seed migration among plant populations. Heredity, 74, 250—259. tome polymorphisms within populations of dubia ciILLErr, G. w. 1964. Genetic barriers in the Cosmanthus and georgiana in addition to individuals with identi- Phacelias. Rhodora, 66, 359—366. cal cpDNA haplotypes occurring in each of these HAMRICK, J. L., LINHART, Y. B. AND MIlTON, J. B. 1979. two taxa suggest that the split in plastomes D and G Relationships between life history characteristics and probably predated divergence of the taxonomic vari- electrophoretically detectable genetic variation in eties dubia and georgiana. plants. Ann. Rev. EcoL Syst., 10, 173—200. HARRIS, S. A. AND INGRAM, R. 1991. Chloroplast DNA and biosystematics: the effects of intraspecific diversity and Acknowledgements plastid transmission. Taxon, 40, 393—412. HONG, Y.-P., I-HPKINS, V. D. AND STRAUSS, S. H. 1993. Chlor- Wethank B.Anderson,B.Burkhardt,L. Harris, J. oplast DNA diversity among trees, populations and P. Hosler, P. Liu, K. MacElwaine and D. Robertson species in the California closed-cone pines (Pinus for helpful advice and technical assistance. B. Weir radiata, Pinus muricata and Pinus attenuata). Genetics, kindly supplied a computer program to estimate 135, 1187—1196. coancestry. The manuscript benefited from M. HOOGLANDER, N., LUMARET, R. AND BOS, M. 1993. Inter- Uyenoyama's insightful comments and the sugges- intraspecific variation of chloroplast DNA of European tions of two anonymous reviewers. This research was Plantago spp. Heredity, 70, 322—334. supported in part by NIH grant no. 5 T32 KEMBLE, R. .1987.A rapid, single leaf, nucleic acid assay GM07754-09, NSF grant no. 333-0861 and a Grant- for determining the cytoplasmic organelie complement in-aid-of-Research from Sigma Xi. of rapeseed and related Brassica species. Theor Appi. Genet., 73, 364—370. KOCHERT, G., HALWARD, T., BRANCH, W. D. AND SIMPSON, C. References E. 1991. RFLP variability in peanut (Arachis hypogaea L.) cultivars and wild species. Theor App!. Genet., 81, AVISE,J. C., ARNOLD, S. BALL, R. M., BERMINGHAM, E., 565—570. LAMB, T., NEIGEL, S. E., REEB, C. A. AND SAUNDERS, N. C. LEVY, F. 1991a. Morphological differentiation in Phacelia 1987. Intraspecific phylogeography: The mitochondrial dubia and P maculata. Rhodora, 93, 11—25. DNA bridge between population genetics and system- LEVY,F. 1991b. A genetic analysis of reproductive barriers atics. Ann. Rev. Ecol. Syst., 18,489—522. in Phacelia dubia. Heredity, 67, 331—345. AVISE, S. C., NEIGEL, J. E. AND ARNOLD, i. 1984. Demo- MASON, R. J., HOLSINGER, K. E. AND JANSEN, R. K. 1994. graphic influences on mitochondrial DNA lineage survi- Biparental inheritance of the chloroplast genome in vorship in animal populations. I Mo!. Evol., 20, 99—105. Coreopsis (Asteraceae). J. Hered., 85, 171—173. AVISE, J. C., SHAPJRA, J. F., DANIEL, S. W., AQUADRO, C. F. MAYER, M. S. AND SOLTI5, i'. s. 1994. The evolution of AND LANSMAN, R. A. 1983. Mitochondrial DNA differ- serpentine endemics: A chloroplast DNA phylogeny of entiation during the speciation process in Peromyscus. the Streptanthus glandulosus complex (Cruciferae). Syst. Mo!. Biol. Evol., 1, 38—56. Bot., 19, 557—574. BALLARD, J. W. 0. AND KREITMAN, M. 1994. Unraveling McCAULEY, 0. E. 1994. Contrasting the distribution of selection in the mitochondrial genome of Drosophila. chloroplast DNA and allozyme polymorphism among Genetics, 138, 757—772. local populations of Silene alba: Implications for studies

The Genetical Society of Great Britain, Heredity, 76, 143—155. CHLOROPLAST DNA IN PHA GEL IA 155

of gene flow in plants. Proc. Nat!. Acad. Sci. US.A., 91, 1180—1206. 8127—8131. PAMIL0, P. AND NE!, M. 1988. Relationships between gene MILLIGAN, B. 6. 1991. Chloroplast DNA diversity within trees and species trees. Mo!. Bio!. Evol., 5, 568—583. and among populations of Trifolium pratense. Curt PETIT, R. J., KREMER, A. AND WAGNER, D. B. 1994. Genet., 19, 411—416. Geographic structure of chioroplast DNA polymorph- MURDY, w. i-i. 1966. The systematics of Phace!ia maculata isms in European oaks. Theor App!. Genet., 87, and Phace!ia dubia var. georgiana, both endemic to 122—128. granite outcrop communities. Am. .1.Bot., 53, RAND, D. M., DORFSMAN, M. AND KANN, L. M. 1994. Neutral 1028—1036. and non-neutral evolution of Drosophila mitochondrial NEALE, D. B., SAGHAI-MAROOF, M. A., ALLARD, R. W., DNA. Genetics, 138, 741—756. ZHANG, 0. AND JORGENSEN, R. A. 1988. Chioroplast SCOWCRAFT, w. R. 1979. Nucleotide polymorphism in DNA diversity in populations of wild and cultivated chioroplast DNA of Nicotiana debneyi. Theor App!. barley. Genetics, 120, 1105—1110. Genet., 55, 133—137. NE!, M. 1986. Stochastic errors in DNA evolution and SLATKIN, NI. 1985. Rare alleles as indicators of gene flow. molecular phylogeny. In: Gershowitz, H., Rucknagel, D. Evolution, 39, 53—65. L. and Tashian, R. E. (eds) Evolutionaiy Perspectives SOLTIS, D. E., SOLTTS, P. S., KUZOFF, R. K. AND TUCKER, T. L. and the New Genetics, pp. 133—147. Alan Liss, New 1992. Geographic structuring of chioroplast DNA geno- York. types in Tiarella trifo!iata (Saxifragaceae). P1. Syst. Evo!., NEIGEL, J. E. AND AvISE, J. c. 1986. Phylogenetic relation- 181, 203—216. ships of mitochondrial DNA under various demo- SWOFFORD, D. L. 1991. i'ui': Phylogenetic Analysis Using graphic models of speciation. In: Nevo, E. and Karlin, Parsimony, version 3.0. Computer program distributed S. (eds) Evo!utionaiy Processes and Theoiy, pp. 515—534. by the Illinois Natural History Survey. Academic Press, New York. TAJIMA, F. 1983. Evolutionary relationship of DNA PALMER, J. D. 1982. Physical and gene mapping of chlor- sequences in finite populations. Genetics, 105, 437—460. oplast DNA from Atriplex triangu!aris and Cucumis TERAUCJ-II, R., TERAUCHI, T. AND TSUNEWAKI, K. 1991. sativa. NucI. Acids Res., 10, 1593—1605. Intraspecific variation of chloroplast DNA in Dioscorea PALMER, i. ii 1985. Comparative arrangement of chior- bulbifera L. Theor App!. Genet., 81, 461—470. oplast genomes. Ann. Rev. Genet., 19, 325—354. WEIR, B. S. 1990. Genetic Data Analysis. Sinauer, Sunder- PALMER, J. ii 1987. Chioroplast DNA evolution and land, MA. biosystematic uses of chioroplast DNA variation. Am. WOLFE, K. H., LI, W.-H. AND SHARP, P. M. 1987. Rates of Nat., 130, S6—S29. nucleotide substitution vary greatly among plant mito- PALMER, J. D. AND ZAMIR, D. 1982. Chloroplast DNA chondrial, chioroplast, and nuclear DNAs. Proc. Nat!. evolution and phylogenetic relationships in Lycopersi- Acad. Sci. US.A., 84, 9054—9058. con. Proc. Nat!. Acad. Sci. US.A., 79, 5006—SOlO. WOLFF, K. AND SCHAAL, B. 1992. Chloroplast DNA varia- PALMER, J. D., JANSEN, R. K., MICHAELS, H. J., CHASE, M. W. tion within and among five Plantago species. .1. Evo!. AND MANHART, J. R. 1988. Chloroplast DNA variation Biol., 5, 325—344. and plant phylogeny. Ann. Mo. Bot. Gard., 75,

The Genetical Society of Great Britain, Heredity, 76,143—155.