Phylogenetic Relationships and Infraspecific Variation in Canadian Arctic Poa Based on Chloroplast DNA Restriction Site Data

Color profile: Disabled Composite Default screen

679

Phylogenetic relationships and infraspecific variation in Canadian Arctic Poa based on chloroplast DNA restriction site data

Lynn J. Gillespie and Ruben Boles

Abstract: Infraspecific variation and phylogenetic relationships of Canadian Arctic species of the genus Poa were studied based on chloroplast DNA (cpDNA) variation. Restriction site analysis of polymerase chain reaction amplified cpDNA was used to reexamine the status of infraspecific taxa, reconstruct phylogenetic relationships, and reexamine previous classification systems and hypotheses of relationships. Infraspecific variation was detected in three species, but only in Poa hartzii Gand. did it correspond to infraspecific taxa where recognition of subspecies ammophila at the species level is supported. Additional variation in P. hartzii ssp. hartzii is hypothesized to be the result of hybridization with Poa glauca in the High Arctic and subsequent introgression resulting in repeated transfer of P. glauca DNA. The variation in Poa pratensis L. had a geographical rather than taxonomic basis, and is hypothesized to correspond to indigenous arctic versus introduced extra-arctic populations. In P. glauca Vahl cpDNA variation was detected only in western Low Arctic and boreal populations and may represent greater variation where the species survived the Pleisto- cene glaciations. Cladistic parsimony analysis of cpDNA restriction site data mostly confirms recent infrageneric classification systems. Poa alpina L., along with the non-arctic Poa annua L. and Poa sect. Sylvestres, formed the basalmost clades. The remaining taxa group into two main clades: one consisting of Poa sects. Poa, Homalopoa, Madropoa and Diocopoa; the second, of Poa sects. Secundae, Pandemos, Abbreviatae and Stenopoa. Poa sect. Poa, comprising Poa arctica R. Br. and P. pratensis, is a strongly supported monophyletic group, not closely related to P. alpina. Poa hartzii is confirmed as a member of a paraphyletic or weakly supported P. sect. Secundae. Poa glauca and Poa abbreviata R. Br. are distinct members within a generally unresolved Poa. sect. Stenopoa–Abbreviatae complex Key words: Poa, Canadian arctic, chloroplast DNA, restriction site analysis, infraspecific variation, phylogeny. Résumé : En se basant sur la variation de l’ADN chloroplastique, les auteurs ont étudié la variation infraspécifique et les relations phylogénétiques d’espèces du genre Poa de l’Arctique canadien. Il ont utilisé l’analyse des sites de restriction de l’ADN, amplifié par PCR, afin de réexaminer les systèmes de classifications antécédents et les hypothèses de relations. On décèle de la variation infraspécifique chez trois espèces, mais seulement chez le Poa hartzii correspond- elle à des taxons infraspécifiques où la reconnaissance de la sous-espèce ammophila est bien supportée. On formule l’hypothèse que la variation additionnelle repérée chez le P. hartzii subsp. hartzii résulterait de l’ hybridation avec le Poa glauca dans le haut-arctique et d’une introgression subséquente conduisant à un transfert répété de l’ADN de le P. glauca. La variation décelée chez le Poa pratensis a une base géographique plutôt que taxonomique et on propose l’hypothèse qu’elle correspond à des populations indigènes vs des populations extra-arctiques introduites. Chez le P. glauca, on ne décèle une variation du cpADN que chez les populations boréales et du bas arctique occidental, ce qui pourrait représenter une variation plus importante là où les espèces ont survécu à la glaciation du Pléistocène. L’analyse cladistique en parcimonie, des données sur les sites de restriction du cpADN, confirment largement les systè- mes de classification infragénériques récents. Le Poa alpina,aveclePoa annua et les Poa sect. Sylvestres non- arctiques, forment les clades les plus fondamentaux. Les autres taxons se regroupent en deux clades principaux, un comprenant les Poa sects. Poa, Homalopoa, Madropoa et Diocopoa, et le second les Poa sects. Secundae, Pendemos, Abbreviatae et Stenopa. Les Poa sect. Poa, comprenant le P. arctica et le P. pratensis, constituent un groupe monophy- létique fortement supporté, pas étroitement relié au P. alpina.LeP. hartzii se voit confirmé comme membre du groupe, paraphylétique ou faiblement supporté, Poa sect. Secundae.LeP. glauca et le Poa abbreviatae sont des membres dis- tincts à l’intérieur du complexe généralement irrésolu Poa sect. Stenopoa-Abbreviatae. Mots clés : Poa, Arctique canadien, ADN chloroplastique, analyse des sites de restriction, variation infraspécifique, phylogénie. [Traduit par la Rédaction] Gillespie and Boles 701

Received March 08, 2001. Published on the NRC Research Press Web site on June 6, 2001. L.J. Gillespie1 and R. Boles. Research Division, Canadian Museum of Nature, P.O. Box 3443, Station D, Ottawa, ON K1P 6P4, Canada. 1Corresponding author (e-mail: [email protected]).

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:32 AM Color profile: Disabled Composite Default screen

680 Can. J. Bot. Vol. 79, 2001

Introduction The first floristic study of the entire Canadian Arctic Ar- chipelago treated eight species of Poa and an additional The bluegrass genus Poa L. in the Canadian Arctic has been three infraspecific taxa (Porsild 1957, 1964). Poa abbreviata studied recently as part of floristic (McLachlan et al. 1989; R. Br., Poa arctica R. Br., Poa glauca Vahl, Poa hartzii Aiken et al. 1996a, 1996b) and taxonomic studies (Soreng Gand., and Poa alpigena var. colpodea (Th. Fr.) Schol. were 1991b). This research, based on morphology, has highlighted considered widespread in the Flora area; Poa arctica ssp. areas of persisting taxonomic confusion that warrant further caespitans (Simm.) Nannf., and Poa arctica var. vivipara study. The present investigation uses restriction site analysis of Hook., as primarily eastern Arctic; Poa alpigena Lindm. and chloroplast DNA (cpDNA) amplified by polymerase chain re- Poa alpina L., as Low Arctic, reaching only the southernmost action (PCR) to address these taxonomic problems and to study Arctic islands; and Poa flexuosa Sm. and Poa nascopieana phylogenetic relationships and biogeographic hypotheses Polunin, with restricted distributions in the eastern Low Arc- among Canadian Arctic species of Poa. The genus is very tic. This work was partly based on previous regional floristic widespread in the Canadian Arctic, from sea level to close to studies of the Canadian Arctic Islands (Polunin 1940, 1955). the upper altitudinal limit of vegetation, and as far north as Five taxa of Poa (P. abbreviata, P. alpigena var. colpodea, northern Ellesmere Island. Species occupy a diversity of habi- P. arctica, P. glauca, and P.“×hartzii”) were included by tats, including densely vegetated mesic tundra, dry barren McLachlan et al. (1989) in their treatment of grasses of the slopes and ridges, sand plains and hills, and mesic and wet Queen Elizabeth Islands, the northernmost island group. The mossy meadows. Ecologically important as a component of two infraspecific taxa of P. arctica treated by Porsild (1957, successional vegetation, several species grow in naturally dis- 1964), subspecies caespitans and variety vivipara, were both turbed and often enriched sites such as riverbanks, eroding synonymized under P. arctica, whereas P. hartzii was treated slopes, areas around animal burrows and remains, ancient as a hybrid species. However, the status of each of these dwelling sites, and recent human-disturbed sites. three taxa plus that of P. alpigena var. colpodea was consid- With over 500 species, Poa is renowned to be a taxonomi- ered to be uncertain. This study was expanded by Aiken et cally difficult genus. In North America alone the genus com- al. (1996a, 1996b) to cover all the Canadian Arctic Islands. prises 71 species, 37 infraspecific taxa, and 3 named hybrids In this study, P. hartzii was not formally treated as a hybrid (Soreng and Kellogg 2001). Species are often difficult to de- taxon but was discussed as a possible hybrid, and P. fine and distinguish from closely related congeners. High alpigena var. colpodea was not considered distinct from the levels of polyploidy, high incidence of asexual reproduction, Low Arctic P. alpigena, which was treated here as Poa and frequent occurrence of hybridization and introgression pratensis ssp. alpigena (Lindm.) Hiitonen. In addition, the are some of the factors believed to be responsible for this Low Arctic species P. alpina was included, and P. arctica taxonomic challenge (Clayton and Renvoize 1986; Kellogg ssp. caespitans was treated as a distinct taxon following R.J. 1987). The paucity of clearly distinguishing, discrete mor- Soreng (herbarium annotations, 1990). The two additional phological characters has led to a reliance on suites of over- species mapped by Porsild (1957, 1964) as occurring on lapping quantitative characters for identification, and, as a Baffin Island, P. flexuosa and P. nascopieana, were ex- consequence, unsatisfactory keys. Furthermore, many spe- cluded, as unconfirmed or of uncertain status, respectively. cies of Poa appear to vary morphologically depending on the The authors remained uncertain of the status of many of microenvironment; arctic species, in particular, have been these taxa and recommended that more detailed studies be described as phenotypically plastic with broad ecological carried out to resolve the persisting taxonomic uncertainties. tolerances (Murray 1995). The mainland Canadian Arctic has been treated in large Poa has been considered “an extremely uniform genus for part by Porsild and Cody (1980) and also by Cody (1996). which there is no satisfactory infrageneric classification” In addition to the taxa treated above, Poa ammophila A.E. (Clayton and Renvoize 1986, p. 101). Classifications have Porsild is included as a locally common, N.W.T. endemic been based primarily on morphological characters, and to a species, found primarily along the western arctic coast. Sev- lesser extent on ecological and cytological characters, most eral alpine species are indicated as reaching the alpine–arctic lack a cladistic framework and, due to the size of the genus, transition zone of the northernmost cordilleras in the Yukon are usually regionally based. We follow here the most recent and westernmost N.W.T. but are not considered to be arctic infrageneric classification of Poa in North America (Soreng species and will not be further discussed here. These include 1998) that makes use of cpDNA restriction site data in a Poa abbreviata ssp. pattersonii (Vasey) A. Löve, D. Löve & cladistic framework in addition to morphology, ecology, and B.M. Kapoor (syn. Poa jordalii A.E. Porsild), Poa arctica cytology. Other recent classifications pertaining to arctic and ssp. lanata (Scribn. & Merr.) Soreng (syn. Poa lanata northern temperate species are also discussed (Edmondson Scribn. & Merr.), Poa leptocoma Trin., Poa paucispicula 1978, 1980; Tzvelev 1983). Scribn. & Merr., and Poa pseudoabbreviata Roshev. Like- wise, Poa eminens J. Presl, a seashore plant found primarily within the boreal ecozone, just reaches the Low Arctic in Taxonomic history of Poa in the Canadian Arctic northern Labrador and northwestern Quebec (Cayouette and Although much of the Canadian Arctic flora has been Darbyshire 1993) and will also not be considered here. treated in several regional floristic studies, there has been no comprehensive floristic treatment of the entire arctic region of Canada, defined as the area north of the latitudinal tree- Species of Poa in the Canadian Arctic line. Outlined below is a brief taxonomic history of Poa in The species of Poa found in the Canadian Arctic are the Canadian Arctic. described in detail below, including their distribution, taxo-

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:32 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 681

nomic history, infraspecific taxa, and any systematic prob- with the Eurasian Poa bulbosa L. species complex lems or taxonomic uncertainties. (Edmondson 1978, 1980; Tzvelev 1995); in its own section, Alpinae (Hegetschw. ex Nyman) Soreng, along with several Poa abbreviata closely related European species (Soreng 1998); and in its Poa abbreviata is an arctic–alpine circumpolar species own subsection, Caespitosae V. Jirásek, in a broadly defined (although rare and with a very scattered distribution in east- P. sect. Poa, which also includes Poa pratensis L., P. ern arctic Russia), which was described from Melville Is- arctica, and P. bulbosa (Tzvelev 1983). Phylogenetically, land. Of the three subspecies recognized in North America the species has been considered closely related to P. (Soreng 1991b), only P. abbreviata ssp. abbreviata is found pratensis and P. arctica (Nannfeldt 1940; Tzvelev 1983); to in the Canadian Arctic. Considered to be a High Arctic cir- the P. bulbosa species complex (Edmondson 1980; Tzvelev cumpolar taxon, this subspecies is primarily restricted to the 1983, 1995); or as the core species of a small isolated, early arctic islands in Canada, with few known localities on diverging species group, near P. sect. Ochlopoa Asch. & coastal mainland N.W.T. (Porsild and Cody 1980). Poa Graebn. (Soreng 1990, 1998). abbreviata ssp. pattersonii, distinguished by the presence of a well-developed web and shorter lemma pubescence, is pri- Poa arctica marily a North American Rocky Mountain alpine taxon, Poa arctica is a widespread circumpolar arctic–alpine spe- reaching the arctic only on Wrangel Island in Far Eastern cies, considered to be highly polymorphic (Nannfeldt 1940; Russia (Tsvelev 1983 as P. abbreviata ssp. jordalii (A.E. Edmondson 1980; Tzvelev 1983, 1995). In the Canadian Porsild) Hultèn; Soreng 1991b). Poa abbreviata ssp. Arctic the species is currently considered to comprise two abbreviata, the only subspecies to be considered here, is subspecies based primarily on growth form and anther steril- morphologically distinct from other arctic taxa and not ity (Aiken et al. 1996a, 1996b). Poa arctica ssp. arctica is highly variable. The species has been placed in its own rhizomatous, with a dispersed or loosely turf-forming small, mainly Beringian section, Abbreviatae Nannf. ex growth form, while P. arctica ssp. caespitans is considered Tzvelev, by all recent authors, although Tsvelev (1995) does to be caespitose or densely tufted, with rhizomes few and suggest that it is closely allied to P. sect. Oreinos Asch. & short or absent, and apomictic with sterile pollen (Porsild Graebn. and Cody 1980; Aiken et al. 1996b). Although their ranges overlap considerably in the Canadian Arctic, P. arctica ssp. Poa alpina caespitans appears to be more common in the High Arctic Poa alpina is an (sub)arctic–(sub)alpine species with an and less so in the Low Arctic than P. arctica ssp. arctica interrupted circumpolar–circumboreal distribution (absent (Aiken et al. 1996b). Nannfeldt (1940) originally described from most of arctic eastern Siberia). In the Canadian Arctic P. arctica ssp. caespitans from Ellesmere Island as a non- the species is found in the eastern Low Arctic (eastern viviparous taxon with sterile anthers, but with viable seed, Baffin Island, Southampton Island, northern Quebec, and the with a distribution from eastern Arctic Canada to Arctic Eu- Hudson Bay region), as well as in the subarctic across north- rope and Novaya Zemlya. In the Russian Arctic, P. arctica ern Canada and south in the western mountains to northern ssp. caespitans is treated as a separate species under the New Mexico. Restricted to the Low Arctic and south in name Poa tolmatchewii Roshev. and is thought to be of prob- North America (primarily so in Greenland), in Eurasia it able hybrid origin with P. arctica and P. glauca as parental commonly occurs in the High Arctic zone of Svalbard and species (Tzvelev 1983, 1995). Soreng and Kellogg (2001) on Novaya Zemlya. Canadian material was considered to be also comment that the subspecies may be the result of hy- highly variable by Polunin (1940). Although considering bridization between these two species. Rønning (1996) con- most of this variation to consist of intergradating ecological siders P. arctica ssp. caespitans to be clearly distinguishable forms, he did treat two varieties (P. alpina var. frigida in Svalbard, whereas Edmondson (1980) describes a more (Gaudin) Rchb. and P. alpina var. bivonae (Parl. ex Guss.) broadly defined P. arctica as a laxly caespitose, variable spe- St. John) and one forma (P. alpina f. brevifolia (Gaudin) cies and recognizes no infraspecific taxa in Europe. Recent Polunin). Subsequent authors have not considered any studies in Greenland have either treated P. arctica ssp. infraspecific taxa as occurring in the Canadian Arctic. The caespitans as P. arctica var. caespitans (Simm.) Nannf. species has been described as often viviparous (vivipary is (Böcher et al. 1968) or not recognized it (Fredskild 1996), used here in the sense of vegetative proliferation of florets), and R.J. Soreng (personal communication) considers a previ- particularly in the eastern Canadian Arctic (Porsild 1957, ously recognized species, Poa filipes Lange (=trichopoda 1964; Porsild and Cody 1980). However, Polunin (1940) Lange), to be clearly synonymous with the subspecies. In comments that no viviparous forms are found in this area addition, Fredskild (1996) mentions that P. arctica is diffi- and Aiken et al. (1996a, 1996b) mention that plants with cult to distinguish from P. pratensis ssp. alpigena in West vegetatively proliferating inflorescences have not been col- Greenland, where material is highly variable with numerous lected in the Canadian Arctic Archipelago. Viviparous plants intermediates. are recognized as a distinct variety or subspecies in Green- Poa arctica var. vivipara was described from Somerset Is- land, Svalbard, and Russia (Böcher et al. 1968; Tzvelev land in Arctic Canada and is reported to occur sporadically 1983, 1995; Rønning 1996) but generally not in North around the arctic. Polunin (1940) cited several collections, America (Porsild 1957; Porsild and Cody 1980) and some- whereas Porsild (1957) and Porsild and Cody (1980) times not in Europe (Edmondson 1980). mapped the variety in 18 different localities in eastern Arctic Poa alpina has been classified in the following widely Canada. Porsild (1955) had previously commented that the divergent ways: in P. sect. Bolbophorum Asch. & Graebn. viviparous High Arctic Poa, which most previous authors

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:33 AM Color profile: Disabled Composite Default screen

682 Can. J. Bot. Vol. 79, 2001

had referred to P. pratensis or P. alpigena, seems best widely introduced in North America and, according to placed in P. arctica as var. vivipara but gave no specific Tzvelev (1995), even more closely related to the Siberian reasons. More recently, Aiken et al. (1996a, 1996b) did not species Poa botryoides (Trin. ex Griseb.) Roshev. and Poa consider P. arctica var. vivipara to be a distinct taxon in the stepposa (Krylov) Roshev. Intermediates between P. glauca Canadian Arctic Islands, whereas Cody (1996) treated it as and P. nemoralis or Poa palustris L. are common where the occurring rarely in the Yukon from the arctic coast south to species are sympatric (Soreng and Kellogg 2001). In Green- 64°N. R.J. Soreng (personal communication), in a recent land, forms that appear transitional with P. nemoralis have study of Alaskan Poa collections (ALA), found that vivipa- been treated as Poa glauca var. glaucantha (Gaudin) Blytt rous P. arctica specimens were restricted to alpine areas pri- (Böcher et al. 1968). Recent suggestions considered for the marily south of 66°N. In the Russian arctic, the rarely found Flora of North America were the treatment of P. glauca as P. arctica var. vivipara (Novaya Zemlya and the Chukokta either synonymous with (proposed by E.A. Kellogg) or as a Peninsula) is considered to be a possible hybrid between P. subspecies of P. nemoralis (R.J. Soreng, personal communi- arctica and P. pratensis s.l., and “is not always distinguish- cation). able from P. pratensis ssp. colpodea” (Tzvelev 1983, p. 689, 1995). The presence of viviparous P. arctica in the Cana- Poa hartzii dian Arctic will be discussed in more detail in the Discus- Poa hartzii is an arctic species found sporadically from sion. Svalbard westward across arctic North America to Wrangel Poa arctica is most often classified with P. pratensis plus Island, with areas of greatest abundance in northeast Green- a number of closely related Eurasian species in P. sect. Poa land and the Canadian High Arctic. A detailed account of its (Soreng 1998) or subsect. Poa (Tzvelev 1983). In contrast, taxonomic history and systematic problems is given in Edmondson (1978, 1980) treats P. arctica with the European Gillespie et al. (1997); only a brief summary will be pro- Poa cenisia All. in P. sect. Cenisia Asch. & Graebn., re- vided here. Two subspecies of Poa hartzii are currently rec- stricting P. sect. Poa to the P. pratensis complex. ognized in the Canadian Arctic. Poa hartzii ssp. ammophila, endemic to the mainland N.W.T Beaufort Sea coast, was Poa glauca originally described as P. ammophila by Porsild (1943) and Poa glauca, a common and widespread species in the Ca- treated as such by Porsild and Cody (1980) but was consid- nadian Arctic, has a circumpolar, arctic–alpine distribution ered to be included within P. hartzii by Polunin (1959) and extending southward into the boreal zone. Among several Scoggan (1978) and as a subspecies by Soreng (1991b). Poa infraspecific taxa recognized by Simmons (1906) as occur- hartzii ssp. hartzii is widespread but scattered on the Arctic ring on Ellesmere Island, variety tenuior Simmons was de- Islands with one disjunct population on the arctic coast of scribed based on slender, depauperate specimens from northern Quebec (Cayouette 1984) and two on the Beaufort Ellesmere Island, but with the exception of a very tentative Sea coast of mainland N.W.T. (Soreng 1991b; Table 1, recognition by Polunin (1940), none have been recognized voucher Nos. 6397, 6398). Viviparous plants (i.e., with vege- as distinct by subsequent authors. Subsequent to Simmons, tatively proliferating florets) of P. hartzii ssp. hartzii were botanists have been frustrated in their attempts to subdivide first collected from Ellesmere Island (Simmons 1906) and P. glauca in the Canadian Arctic and, for the most part, did considered as a distinct variety, var. vivipara Polunin, or not; as Polunin (1940, p. 68) mentions “transitional forms form, f. prolifera (Simmons) B. Boivin (based on P. glauca were so abundant and the characters so unstable that the task var. atroviolacea f. prolifera Simmons). Although P. hartzii seemed futile.” Most recently, Soreng and Kellogg (2001) f. prolifera was recognized by Boivin (1967) and Scoggan treat the western North American alpine Poa glauca ssp. (1978), neither taxon was treated or discussed by other rupicola (Nash) W.A. Weber as a separate, but weakly de- authors of Canadian Arctic floras (Porsild 1957, 1964; marcated taxon, and Soreng (herbarium annotations, 1999) McLachlan et al. 1989; Aiken et al. 1996a, 1996b). Vivipa- considers it to likely occur in the Canadian Arctic. Tzvelev rous plants have recently been collected from Axel Heiberg, (1983) recognizes three subspecies in Russia and considers Melville, and Victoria islands (herbarium specimens at P. glauca ssp. glauca to be “highly polymorphic…, probably CAN, Table 1, voucher Nos. 6623, 6624) and are found in divided into several still insufficiently studied taxa” (1983, populations comprised entirely of proliferating plants or, p. 718). Tzvelev (1995) recognized numerous infraspecific more frequently, mixed with non-proliferating plants. Out- and segregate taxa in Russia, of which two, Poa anadrica side the Canadian Arctic viviparous plants are known from Roshev. and Poa bryophila Trin. (treated as varieties of P. Greenland where they are treated as P. hartzii f. prolifera glauca ssp. glauca in Tzvelev 1983), were considered as ap- (Bay 1993) and Wrangel Island, Russia, where they are con- parently or probably occurring in the Canadian Arctic. How- sidered either as a distinct species, Poa vrangelica Tzvelev ever, recent studies in Siberia argue against the subdivision (Tzvelev 1983) or variety, P. hartzii var. vrangelica of P. glauca (Olonova 1993). (Tzvelev) Prob. (Probatova 1984). Poa glauca belongs to the primarily Eurasian P. sect. Poa hartzii ssp. hartzii has previously been suggested to Stenopoa Dumort. Tzvelev (1995, p. 212) considers this sec- be a hybrid between P. abbreviata and either P. glauca or P. tion to comprise numerous closely related species that are arctica (Scholander 1934; Nannfeldt 1935; Edmondson tied closely to one another by transitional populations, “... so 1980). In our previous study investigating the role of hybrid- that under a broader species concept they might all be con- ization in the evolution of this species, P. hartzii ssp. hartzii sidered subspecies of a single gigantic polytypic species was determined to be a morphologically distinct apomictic (Poa nemoralis L. s.l.).” Poa glauca is considered to be taxon with two very different cpDNA haplotypes (Gillespie closely related to P. nemoralis, a species both native to and et al. 1997). The presence of these two haplotypes, deter-

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:34 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 683

mined to be identical to the P. glauca and Poa secunda J. Soreng and Barrie (1999) consider the P. pratensis com- Presl. species complex haplotypes, suggested that hybridiza- plex to be extremely variable in morphology and cytology tion had been involved in its evolution. Two hypotheses with frequent asexual reproduction and discuss the history of were formulated as a result of these findings. The first sug- treating taxa in this complex as species versus subspecies or gests that P. hartzii is a stabilized intersectional hybrid be- varieties. Poa pratensis, the type species of the genus, is tween P. glauca and a species in the P. secunda complex. classified in the mostly Eurasian P. sect. Poa (subsect. Poa in The second hypothesis suggests that an ancestral P. hartzii, Tzvelev 1983). closely related to P. secunda, “captured” P. glauca cpDNA by hybridization and introgression. Poa flexuosa and P. nascopieana Poa hartzii was placed in P. sect. Abbreviatae by Tzvelev Poa flexuosa was considered by Porsild (1957) and (1983) and, following him, by Soreng (1991a), considered Porsild and Cody (1980) to occur on eastern Baffin Island, an apparently stabilized intersectional hybrid (probably northern Labrador, and along the western shore of Hudson between sects. Stenopoa and Abbreviatae) by Edmondson Bay. The taxon was treated as Poa laxa Haenke ssp. flexuosa (1980) and most recently considered a member of P. sect. (Sm.) Hyl. by Scoggan (1978). In contrast, Aiken et al. Secundae V. Marsh ex Soreng (Soreng 1991b; Soreng and (1996a, 1996b) excluded the taxon from the Arctic Island Kellogg 2001). flora as unconfirmed, whereas Soreng (1991b) considered most of the previously annotated collections to be referable Poa pratensis to P. glauca s.l. The status of populations attributed to P. Three taxa within P. pratensis are currently recognized in flexuosa by Porsild will not be addressed here but will be the the Canadian Arctic. Poa pratensis ssp. pratensis, a wide- focus of a future study. spread temperate–boreal species thought to be introduced Polunin (1940) described P. nascopieana based on a sin- from Eurasia, reaches the arctic at several localities in north- gle collection from the Pangnirtung area on Baffin Island. ern Quebec (Porsild and Cody 1980) and along the mainland The species was treated by Porsild (1957, 1964) with one Beaufort Sea coast (based on keys in Porsild and Cody additional locality in the Clyde River area, Baffin Island, but 1980; Cody 1996). Poa pratensis ssp. alpigena is an indige- was excluded from the Arctic Archipelago flora by Aiken et nous circumpolar arctic–alpine taxon. Although previously al. (1996a, 1996b). These collections have abnormally de- treated as the distinct species P. alpigena (Porsild 1957, veloped inflorescences making it difficult to be sure of their 1964; Porsild and Cody 1980; McLachlan et al. 1989), the identity. R.J. Soreng (personal communication) considers taxon has been most recently treated, in Canada, as a sub- these collections to be most like P. glauca in vegetative fea- species of P. pratensis (Aiken et al. 1996a, 1996b; Cody tures and has recently recognized a third collection from 1996) following Tzvelev (1983) and Soreng (1991b). In the Lake Hazen, Ellesmere Island (Murray & Yurtzev 10133 Canadian Arctic, P. pratensis ssp. alpigena is found primar- ALA), that is similar to the type. The species remains an un- ily in the Low Arctic with scattered populations on Prince certain taxon, possibly synonymous with P. glauca, and will Patrick, Melville, Ellesmere (primarily on the Fosheim Pen- not be further dealt with in this study. insula) and Axel Heiberg islands in the High Arctic. The third taxon, Poa pratensis ssp. colpodea (Th.Fr.) Chloroplast DNA has been shown to be useful in defining Tzvelev, was recognized as P. alpigena var. colpodea by plant taxa and to infer both infrageneric and infraspecific re- Porsild (1957, 1964), Porsild and Cody (1980), and lationships (Palmer et al. 1988; Soltis et al. 1992, and refer- McLachlan et al. (1989), and as P. pratensis ssp. alpigena ences therein). In the genus Poa, Soreng (1990) successfully var. colpodea by Soreng (1991b) but was not considered dis- used cpDNA restriction site variation, mapped for the whole tinct by Aiken et al. (1996a, 1996b). Tzvelev (1983) consid- cpDNA genome, to reconstruct phylogeny, provide an inde- ered the taxon as P. pratensis ssp. colpodea, a treatment pendent test of relationships inferred from morphological followed by Edmondson (1980) and Soreng and Kellogg data, and serve as markers of geographic radiation. Gillespie (2001). In Canada, this circumpolar taxon is primarily re- et al. (1997) determined PCR-amplified cpDNA restriction stricted to the Arctic Islands, with several coastal mainland site data from two intergenic regions to be sufficiently vari- arctic populations (Porsild and Cody 1980). Soreng (1991b) able within Canadian Arctic Poa to be useful in characteriz- considered the taxon to be a High Arctic small viviparous ing species and also in detecting hybridization in P. hartzii. form of P. pratensis ssp. alpigena, suggesting the possibility Although the need for greater taxon sampling has been dis- that vivipary may be a fixed or plastic response of the taxon cussed previously (Soltis et al. 1992), most cpDNA phylo- to the extreme High Arctic environment. A second vivipa- genetic studies assume that infraspecific variation is absent rous form, P. pratensis ssp. alpigena var. vivipara or negligible and utilize a single individual to represent a (Malmgren) Schol., considered a distinct taxon in Svalbard species. The difficulties in adequately characterizing species (Rønning 1996), may occur on Ellesmere (based on two and infraspecific taxa of Poa based on morphology provide Simmons collections recently examined at O) and Melville is- even more reason to broadly survey genetic variation infra- lands (based on recent field collections). Tzvelev (1995) com- specifically. In our previous study (Gillespie et al. 1997) we ments that these two viviparous taxa are hardly distiguishable surveyed for infraspecific variation prior to phylogenetic anal- and considers most specimens of the latter taxon to likely be ysis. Whereas the majority of arctic species were found to be the result of a hybrid cross between P. alpigena and P. characterized by a single unique haplotype, infraspecific arctica, a hypothesis also considered by Soreng (1991b) for P. cpDNA variation was detected in High Arctic P. hartzii (a pratensis ssp. colpodea. case involving hybridization) and Low Arctic P. glauca.

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:35 AM Color profile: Disabled Composite Default screen

684 Can. J. Bot. Vol. 79, 2001

In the present study, restriction site analysis of PCR- increase the geographical coverage (particularly within the amplified cpDNA is used to address problems of taxonomic arctic but also from elsewhere in North America where ap- status and to infer phylogenetic relationships in Canadian plicable). Subsequently, additional collections were added Arctic Poa. This study expands on our previous study of P. for arctic species in which infraspecific cpDNA variation hartzii based on the variation found in two cpDNA regions was detected. Our sampling strategy focused on covering a (Gillespie et al. 1997). Three additional cpDNA regions are broad geographic range, rather than on extensive within pop- examined here, and additional extra-arctic Poa species, out- ulation sampling. Note that each voucher collection number groups, and collections of arctic taxa are evaluated. represents a single individual. Field work was undertaken in In the first step of the study, infraspecific cpDNA varia- the Canadian Arctic to obtain material for DNA analysis as tion was surveyed within each arctic species, with the goals outlined in Gillespie et al. (1997) during the summers of of determining the presence and degree of such variation, 1997 and 1999. Observations were also recorded on the dis- elucidating the status of infraspecific taxa, and detecting hy- tribution, ecology, reproduction, and morphological varia- brid taxa. We were particularly interested in examining the tion of plants in each population sampled. All arctic Poa status of subspecies in P. arctica, P. hartzii, and P. pratensis; preliminary identifications were made by the first author; investigating the possible hybrid origin of some of these these were then rechecked and, if necessary, redetermined taxa; and further elucidating the origin of P. hartzii by test- by R.J. Soreng (Smithsonian Institution). ing the two hypotheses formulated in Gillespie et al. (1997). Twenty-one Poa species from outside the arctic region, A secondary objective was to determine if sufficient infra- primarily from North America, were chosen to represent the specific variation exists in the cpDNA regions examined to morphological diversity found within north temperate Poa be useful in phylogeographic studies, especially to detect and to represent most major sections (Table 1). Many of areas of greater infraspecific genetic diversity that may pro- these collections were provided by R.J. Soreng. The subge- vide evidence for refugia where species survived the Pleisto- neric classification followed here is that of Soreng (1998). cene glaciations. Also included in Table 1 are the six arctic genera used as The second step involved reconstructing the phylogenetic outgroups in the cladistic analysis. Arctagrostis Griseb., relationships of arctic Poa species taking into account this Arctophila (Rupr.) Rupr. ex Andersson, Dupontia R.Br., infraspecific variation. Our goals were to examine the evolu- Puccinellia Parl., and Phippsia (Trin.) R. Br. belong to the tion of arctic species in the broader context of the genus and same tribe, Poeae, as Poa, whereas Hierochloë R. Br. be- to provide a test of both traditional classification schemes longs to tribe Aveneae Dumort. (following the system of and previous phylogenetic hypotheses. Specific questions of Clayton and Renvoize 1986). Puccinellia has been used as interest here include the phylogenetic position of P. alpina, an outgroup in previous Poa cpDNA analyses (Soreng 1990; the relationship of P. arctica to P. pratensis, and the status Gillespie et al. 1997). Arctagrostis and Dupontia are sug- and affinities of P. glauca. Whereas the focus of this paper is gested to be very closely related to Poa based on phylogen- on Canadian Arctic species, the phylogenetic analysis pre- etic analyses of subfamily Pooidae using cpDNA and sented includes a representative sample of north temperate, morphological data (Soreng et al. 1990; Soreng and Davis primarily North American, species of Poa chosen to repre- 2000), whereas Arctophila has not previously been surveyed. sent the majority of Poa sections. This phylogenetic ap- proach provides a useful framework for studying a geographically restricted group of species in the broader DNA extraction and PCR amplification context of the evolution of the genus. Given the large size DNA was extracted from leaf tissue of individual plants and taxonomic complexity of Poa, it would be impractical as and subjected to RNAase following the methods outlined in a first step to include all or most species and unwise to ana- Gillespie et al. (1997). The extraction methods were varia- lyze a taxonomic subdivision of the genus. The phylogeny tions of Doyle and Doyle’s (1990) modification of the hexa- presented here should be considered as a preliminary work- decyltrimethylammonium bromide (CTAB) total DNA ing hypothesis and represents one step towards producing a extraction method of Saghai-Maroof et al. (1984). The pri- more comprehensive hypothesis of phylogenetic relation- mary method used was one modified for extraction from 10– ships in the genus Poa. 50 mg of silica-gel dried leaf tissue in 1.5-mL Eppendorf tubes, which yielded ample DNA (10–90 ng/µL) for PCR Materials and methods amplification. Five regions of the chloroplast genome located within the Taxa and populations sampled large single copy region were amplified via PCR. Three re- The taxa and populations sampled are listed in Table 1. gions, trnF–trnV, trnV–trnL and trnH–trnK, are new to this Ten arctic taxa of Poa, representing six species and an addi- study, whereas two, trnT–trnF (Taberlet et al. 1991) and tional four subspecies, were sampled. A total of 9–57 indi- rbcL–ORF106 (Arnold et al. 1991), were examined in our viduals of each arctic species (with the exception of P. previous study (Gillespie et al. 1997). The amplification re- alpina) were examined to cover the geographic range and the action mix and program for these latter two regions were de- morphological variation found in the Canadian Arctic. Only scribed in that study. Modifications made in this study are three individuals of P. alpina were sampled because of field- 0.2 mM of each dNTP and an annealing temperature of work limitations and since it is primarily a subarctic–alpine 57°C for the trnT–trnF region, 1 µΜ of each primer, 0.2 mM taxon reaching the Low Arctic only in eastern Canada. Using of each dNTP, 20 µL “Q” solution (Qiagen, Mississauga, the collections sampled in our previous study (Gillespie et Ont.), and 1 unit of Taq DNA polymerase for the rbcL– al. 1997) as a starting point, we added collections initially to ORF106 region.

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:35 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 685

Table 1. Taxa and populations of Poa and outgroups Arctagrostis, Arctophila, Dupontia, Hierochloë, Phippsia, and Puccinellia sampled for DNA analysis.

Taxon Section ETU Population location Vouchers Arctic taxa Poa abbreviata R. Br. Abbreviatae abbreviata Ellesmere I., Lake Hazen, 81°49 ′N, 71°20 ′W 6028 Ellesmere I., Tanquary, 81°24 ′N, 76°53 ′W 5957, 5959* Ellesmere I., Eureka, 80°00 ′N, 85°57 ′W 5724, 5731 Cornwallis I., Resolute Bay, 74°15 ′N, 94°50 ′W 5810, 5814, 5816*, 5818 Poa alpina L. Alpinae alpina Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 5717, 5723 U.S.A., Colorado 6299* Poa arctica R. Br. ssp. arctica Poa arctica Ellesmere I., Lake Hazen, 81°49 ′N, 71°20 ′W 5781* Ellesmere I., 79°44 ′N, 83°10 ′W (mound form) 6647* Baffin I., Pond Inlet, 72°47 ′N, 77°00 ′W 6045, 6055 Baffin I., Arctic Bay, 73°02 ′N, 85°10 ′W 6062*, 6071*, 6072* Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 5701, 5702*, 5705 5706, 5709 Victoria I., Cambridge Bay, 69°07 ′N, 5830, 5842 105°03 ′W NWT, Tuktoyuktuk, 69°26 ′N, 133°03 ′W 5941* NWT, Mackenzie Delta, 69°28 ′N, 134°35 ′W 6435* Poa arctica ssp. caespitans Poa caespitans Ellesmere I., Tanquary, 81°24 ′N, 76°53 ′W 5964 (Simm.) Nannf. Baffin I., Arctic Bay, 73°02 ′N, 85°10 ′W 6068*, 6069* Baffin I., Pond Inlet, 72°47 ′N, 77°00 ′W 6041*, 6044* Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 5704, 5722 Victoria I., Cambridge Bay, 69°07 ′N, 5843 105°03 ′W Poa glauca Vahl Stenopoa glauca1, glauca2, glauca3 Ellesmere I., Lake Hazen, 81°49 ′N, 71°20 ′W 6005 (1), 6006 (1), 6007 (1) Ellesmere I., Tanquary, 81°24 ′N, 76°53 ′W 5963 (1)* Ellesmere I., Ridge Lake, 79°56 ′N, 84°40 ′W 5802 (1), 5804 (1)*, 5805 (1) Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 5700 (1), 5715 (1) Baffin I., Ogac Lake, 62°50 ′N, 67°22 ′W 6755-1 (1)* Victoria I., Cambridge Bay, 69°07 ′N, 5823 (1), 5831 (1), 5834 105°03 ′W (1), 5841 (2) NWT, Nicolson I., 69°53 ′N, 129°02 ′W 5863 (2), 5873 (1)*, 5877 (2)* NWT, Cape Dalhousie, 70°14 ′N, 129°40 ′W 5913 (1) NWT, Stanton, 69°48 ′N, 128°42 ′W 5897 (1)* NWT, Kitigazuit, 69°21 ′N, 133°41 ′W 5931 (1) NWT, MacKenzie Delta, 69°04 ′N, 134°17 ′W 6452 (1)* NWT, Prelude Lake, 62°34 ′N, 113°55 ′W 6352 (3)*, 6353 (1)* U.S.A., Colorado 6303 (1)* Poa glauca × P. hartzii glaucaxhartzii Baffin I., Pond Inlet, 72°47 ′N, 77°00 ′W 6054* Poa hartzii Gand. ssp. hartzii Secundae hartzii1, hartzii2 Ellesmere I., Lake Hazen, 81°49 ′N, 71°20 ′W 5771 (1), 5783 (1), 5997 (2), 6000 (1), 6016 (1), 6017 (1), 6020 (1), 6024 (1) Ellesmere I., Tanquary, 81°24 ′N, 76°53 ′W 5740 (2), 5945 (2), 5952 (1), 5960 (2), 5988 (2), 5990 (2) Ellesmere I., Eureka, 80°00 ′N, 85°57 ′W 5725 (1), 5738 (2), 5726 (1), 5729 (1) Ellesmere I., Hot Weather Creek, 79°58 ′N, 6130 (2)*, 6146 (2)* 84°26 ′W Ellesmere I., Ridge Lake, 79°56 ′N, 84°40 ′W 5807 (2) Axel Heiberg I., 80°33 ′N, 90°41 ′W (viviparous 6623-1 (2)*, 6623-2 (2)*, population) 6623-3 (2)*,

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:36 AM Color profile: Disabled Composite Default screen

686 Can. J. Bot. Vol. 79, 2001

Table 1 (continued).

Taxon Section ETU Population location Vouchers 6623-4 (2)*, 6623-5 (2)*, 6624-1 (2)* Axel Heiberg I., Depot Point, 79°37 ′N, A91-036 (1) 86°30 ′W Axel Heiberg I., Expedition Fiord, 79°24 ′N, 6118 (1)*, 6121 (1)*, 6124 90°48 ′W (1)* Victoria I., Cambridge Bay, 69°07 ′N, 5824 (2)*, 5833 (2), 5835 105°03 ′W (2), 5849 (2), 6319 (2)*, 6323 (2)*, 6333 (2)*, 6351 (2)* NWT, Cape Dalhousie, 70°11 ′N, 129°41 ′W 6397 (2)*, 6398 (2)* Poa hartzii ssp. ammophila Secundae ammophila1, ammophila2 NWT, Nicolson I., 69°53 ′N, 129°02 ′W 5851 (1), 5869 (1)*, 5870 (A.E. Porsild) Soreng (1), 5882 (1), 5883 (1)*, 5890 (1)*, 5892 (1)* NWT, Cape Dalhousie, 70°14 ′N, 129°40 ′W 5908 (1), 5909 (2), 5910 (1)*, 5911 (1), 5912 (1)*, 5915 (1)*, 6403 (1)*, 6405 (1)* NWT, Kitigazuit, 69°21 ′, 133°41 ′W 5916 (1), 5921 (1), 5933 (1), 6448 (1)*, 6451 (1)* Poa pratensis L. ssp. pratensis Poa pratensis1, pratensis2 Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 6701-1 (1)* NWT, Nicolson I., 69°53 ′N, 129°02 ′W 5852 (1), 5866 (1) NWT, Stanton, 69°48 ′N, 128°42 ′W 5901 (1)* NWT, Inuvik, 68°21 ′N, 133°43 ′W 6358 (1)* Quebec, Gaspé Pen. 6455 (2)*, 6458 (2)* U.S.A., Colorado 6291 (2)* U.S.A., Colorado 6310 (2)* Poa pratensis ssp. agassizensis Poa agassizensis U.S.A., Colorado 6309* (B. Boivin & D. Löve) R.L. U.S.A., New Mexico S5805* Taylor & MacBryde Poa pratensis ssp. alpigena Poa alpigena Ellesmere I., Ridge Lake, 79°56 ′N, 84°40 ′W 5801, 5803 (Lindm.) Hiitonen Baffin I., Apex, 63°43 ′N, 68°27 ′W 6790-1* Victoria I., Cambridge Bay, 69°07 ′N, 5837 105°03 ′W NWT, Nicolson I., 69°53 ′N, 129°02 ′W 5858, 5880 NWT, Kitigazuit, 69°21 ′N, 133°41 ′W 5927 Poa pratensis ssp. colpodea Poa colpodea Ellesmere I., Tanquary, 81°24 ′N, 76°53 ′W 5951 (Th. Fr.) Tzvelev Cornwallis I., Resolute Bay, 74°15 ′N, 94°50 ′W 5820 Baffin I., Pond Inlet, 72°47 ′N, 77°00 ′W 6043* Extra-Arctic species Poa annua L. Ochlopoa annua1, annua2 Ontario 6284 (1)* British Columbia 6285 (2)*, 6288 (1)* Poa arachnifera Torrey Dioicopoa arachnifera U.S.A., Oklahoma S5801* Poa arida Vasey Secundae arida U.S.A., Oklahoma S5802* Poa autumnalis Muhl. ex Elliot Sylvestres autumnalis U.S.A., Maryland S4680* Poa bulbosa L. Bolbophorum bulbosa U.S.A., Nevada S5814* Poa chaixii Vill. Homalopoa chaixii Russia, St. Petersberg S4677* Poa chambersii Soreng Homalopoa chambersii U.S.A., Oregon S5858* Poa compressa L. Stenopoa compressa NWT, Prelude Lake, 62°34 ′N, 113°55 ′W 6352* U.S.A., Colorado 6289* Poa cusickii Vasey ssp. cusickii Madropoa cusickii1, cusickii2 U.S.A., Nevada S5829 (1)* U.S.A., Nevada (“hansenii form”) S5830 (2)* Poa cuspidata Nutt. Homalopoa cuspidata U.S.A., Maryland S4679* Poa fendleriana (Steudel) Vasey Madropoa fendleriana1, fendleriana2 U.S.A., Colorado 6292 (1)* U.S.A., Colorado 6302 (1)* U.S.A., Colorado 6308 (2)*

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:37 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 687

Table 1 (concluded).

Taxon Section ETU Population location Vouchers Poa macrantha Vasey Madropoa macrantha U.S.A., Oregon S5861* Poa napensis Beetle Secundae napensis U.S.A., California S2926a Poa nemoralis L. Stenopoa nemoralis1, nemoralis2 U.S.A., Maryland S4682 (1)* U.S.A., Oregon S5856 (2)* Poa nervosa (Hook.) Vasey Homalopoa nervosa U.S.A., Oregon S5849* Poa palustris L. Stenopoa palustris Ontario 6461* Poa secunda J. Presl. ssp. Secundae secunda1, secunda2 U.S.A., Washington (“canbyi” form) Native Plants Inc. acces- secunda sion POCO4247 (1)a* U.S.A., Nevada S5813 (2)* Poa secunda ssp. juncifolia Secundae juncifolia U.S.A., Colorado S5809* (Scribner) Soreng U.S.A. (“ampla” form) Sharp Bros. Seed Co.a U.S.A. (“nevadensis” form) Davis et al. s.n.a Poa sylvestris A. Gray Sylvestres sylvestris U.S.A., Maryland S4678* Poa trivialis L. Pandemos trivialis U.S.A., Maryland S4681* Poa wheeleri Vasey Homalopoa wheeleri U.S.A., Nevada S5825* Poa wolfii Scribner Sylvestres wolfii U.S.A., Missouri S5800* Outgroup taxa Arctagrostis latifolia (R.Br.) Arctagrostis Axel Hieberg I., 79°56 ′N, 87°15 ′W 6586*, 6587* Griseb. Arctophila fulva (Trin.) Rupr. Arctophila Banks I., 73°46 ′N, 119°57 ′W A99-230* Dupontia fisheri R.Br. Dupontia Axel Heiberg I., 79°56 ′N, 87°15 ′W 6589* Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 6699* Hierochloë pauciflora R.Br. Hierochloe Cornwallis I., Resolute Bay, 74°41 ′N, 94°50 ′W 6248* Phippsia algida (Sol.) R.Br. Phippsia Devon I., Dundas Harbour, 74°31 ′N, 6668-1* 82°33.5 ′W Baffin I., Nanisivik, 73°02 ′N, 84°33 ′W 6253* Puccinellia andersonii Swallen =Puccangustata Ellesmere I., Lake Hazen, 81°49 ′N, 71°20 ′W 5790 Puccinellia angustata (R. Br.) Puccangustata Ellesmere I., Lake Hazen, 81°49 ′N, 71°20 ′W 5784 5786 Rand. & Redfield Ellesmere I., Eureka, 80°00 ′N, 85°57 ′W 5732 5733 6159 Victoria I., Cambridge Bay, 69°07 ′N, 5836 105°03 ′W Puccinellia arctica (Hook.) =Puccangustata Victoria I., Cambridge Bay, 69°07 ′N, 5844 Fern. & Weath. 105°03 ′W Puccinellia borealis Swallen =Puccangustata NWT, Mackenzie Delta, 69°04 ′N, 134°17 ′W 6453* Puccinellia bruggemannii =Puccangustata Cornwallis I., Resolute Bay, 74°15 ′N, 94°50 ′W 5813 Sorensen Puccinellia phryganodes (Trin.) Puccphryganodes Victoria I., Cambridge Bay, 69°07 ′N, 5850 Scribn. & Merr. 105°03 ′W Puccinellia poacea Sorensen =Puccangustata Ellesmere I., Tanquary, 81°24 ′N, 76°53 ′W 5744 Puccinellia vahliana (Liebm.) Puccvahliana Ellesmere I., Ridge Lake, 79°56 ′N, 84°40 ′W 5808 Scribn. & Merr. Devon I., Dundas Harbour, 74°31 ′N, 6682* 82°33.5 ′W Baffin I., Iqaluit, 63°45 ′N, 68°31 ′W 6794-1* Note: Canadian arctic species of Poa are listed first, followed by extra-Arctic Poa species and outgroup taxa; within a taxon, populations are arranged geographically from north to south and east to west. The sectional classification followed here is that of Soreng (1998). Canadian Arctic Island populations are located in Nunavut; detailed locality information is given only for Nunavut and the Northwest Territories (NWT). Voucher numbers refer to collections made by Gillespie (deposited at CAN); S.G. Aiken, prefixed with an “A” (CAN); and R.J. Soreng, prefixed with an “S” (US). Each voucher number represents an individual plant within the specified population; those with an asterisk are new to this study. Values in parentheses are the evolutionary taxonomic unit (ETU) designations in those cases where a taxon has more than one ETU. aDNA obtained from R.J. Soreng and cited in his cpDNA study (Soreng 1990).

The trnF–trnV, trnV–rbcL, and trnH–trnK regions were Dumoulin-Lapègue et al. 1997). Amplification reactions amplified using primers homologous to, respectively, por- were performed in 100-µL volumes containing 0.5 µMof tions of two transfer RNA genes trnF and trnV, portions of each primer, 0.2 mM of each dNTP, 10 µL reaction buffer, µ the transfer RNA gene trnV and the gene coding for 1.5 mM MgCl2 (contained in the reaction buffer), 0.5 Lof ribulose-1,5-bisphosphate carboxylase, and portions of trans- template DNA, and 1 unit of Taq DNA polymerase. To fa- fer RNA genes trnH and trnK (Demesure et al. 1995; cilitate amplification, 20 µL “Q” solution (Qiagen) and an

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:38 AM Color profile: Disabled Composite Default screen

688 Can. J. Bot. Vol. 79, 2001

additional 0.5–1 mM MgCl2 were often added to the trnV– random taxon addition sequence and 100 replications. rbcL reaction. Reaction mixes were overlayed with two or Characters were treated either as normal reversible (Fitch three drops of mineral oil. The amplification program for the parsimony, all character transformations equally likely) or as first two regions consisted of an initial denaturation step of “Dollo” characters (parallel restriction site gains not al- 4 min at 94°C; 25 (trnF–trnV) or 35 cycles (trnV–rbcL) of lowed). The outgroup taxa Arctagrostis, Arctophila, 45 s at 92°C, 45 s at 57°C, 4 min at 72°C; and a final exten- Dupontia, Hierochloë, Phippsia, and Puccinellia were in- sion step of 10 min at 72°C. The program for the trnH–trnK cluded in all analyses and used to root the cladograms a pos- region consisted of the same initial and final steps and 30 teriori. Support for the cladistic relationships was assessed cycles of 45 s at 92°C, 45 s at 62°C, 2 min at 72°C. Length using bootstrap analysis (Felsenstein 1985). Bootstrap analy- of the amplified products was estimated by comparison with ses were performed with 100 replications using the heuristic known marker DNA ladders in 1.1% agarose gels stained search option and default settings. with ethidium bromide (Sambrook et al. 1989). Results Restriction site analysis Twenty-five restriction endonuclease enzymes were used PCR product length variation in the initial screening for variable restriction fragment pat- All five cpDNA regions showed at least some length vari- terns. Twenty four of these enzymes, having four to six base ation among genera and species. Of these, only the trnF– pair recognition sequences, are listed in Gillespie et al. trnV and the rbcL–ORF106 regions showed length variation (1997). One additional enzyme, PvuII, with a six base pair among species of Poa that was large enough to have a pro- recognition sequence (Promega, Fisher, Ottawa), was used in nounced effect on fragment patterns. this study. The amplification products were digested with The trnF–trnV region amplified product was found to vary one to four units of each restriction enzyme for 2 h follow- from approximately 3100 to 3350 base pairs (bp) in length. ing the manufacturer’s specifications. Restriction fragments Among arctic species, P. arctica and P. pratensis had the were separated by electrophoresis in agarose gels (Sambrook shortest amplification product (-3100 bp); followed by P. et al. 1989) ranging in concentration from 1.1 to 2.3%, de- abbreviata, P. glauca, P. hartzii, Arctagrostis, Arctophila, pending on the length of the fragments. The gels stained Dupontia, and Hierochloë (-3180 bp); then P. alpina (-3280 with ethidium bromide were then photographed on an ultra- bp); and Puccinellia and Phippsia with the longest amplifi- violet light source. cation product (-3350 bp). For the three cpDNA regions new to this study, trnF–trnV, The rbcL–ORF106 region amplified product varied in trnV–rbcL and trnH–trnK, 10 DNA samples representing all length from 2250 bp for P. alpina to 2550 bp for P. arctica arctic Poa species, including both haplotypes of P. hartzii and P. hartzii (haplotypes hartzii2 and ammophila) among and three species of Puccinellia, were digested with each arctic species of Poa (as previously described in Gillespie et restriction enzyme to screen for variable restriction sites. al. 1997). The six outgroup genera had longer amplification Likewise, the trnT–trnF and rbcL–ORF106 regions were products at approximately 2700–2750 bp (this estimate is screened for the PvuII enzyme. In addition, the trnT–trnF re- somewhat longer than previously described for Puccinellia). gion was rescreened for several enzymes (EcoRI, NciI, RsaI) Two extra-arctic species of Poa (P. annua and Poa using 10 species (six Poa species of which four were arctic, trivialis L.) and one species of Puccinellia (Pu. borealis two Puccinellia species, Phippsia, and Hierochloë), where Swallen) were not successfully amplified for this region, in initial screening results were somewhat unclear and (or) in- addition to the five species of Puccinellia that we were un- sufficient species had been sampled. able to amplify in our previous study. We suspect that muta- Following determination of the restriction enzymes that tions have occurred in one of the primers of these species produced variable restriction fragment patterns for each that block annealing. cpDNA region, all collections new to this study were pro- The trnV–rbcL amplified product of species of Poa and cessed for each of these useful region–enzyme combina- the six outgroup genera varied only slightly between 3900 tions. For collections previously studied (Gillespie et al. and 4000 bp in length. The trnH–trnK region amplified 1997), a representative sample (three to six individuals) from products of all seven genera were between 1950 and 2000 each taxon was processed for the three new DNA regions. bp in length. TrnT–trnF length variation of arctic species of Where variation was detected within a taxon, the remaining Poa and Puccinellia has previously been described collections of that taxon were then processed for all region- (Gillespie et al. 1997); Phippsia is similar in length to enzyme combinations variable within Poa. Unusual infra- Puccinellia (-1750), whereas the other outgroup genera vary taxon variation was rechecked by amplifying and digesting between 1800 and 1850 bp in length. the DNA a second and often a third time. Restriction fragment patterns were interpreted in terms of Restriction site analysis restriction site presence (character state 1) versus absence (character state 0) (Dowling et al. 1990). The binary data Restriction fragment patterns matrix (Table 2) was analyzed by cladistic parsimony meth- The five cpDNA regions sampled for the presence of ods using the program PAUP* version 4.0 beta 5 for Win- restriction site variation comprise approximately 10% (-13 450 dows (Swofford 1998). The heuristic search option was used bases) of the chloroplast genome. A total of 43 region–en- with default settings, including a simple taxon addition se- zyme combinations were found to have variable and inter- quence and tree bisection–reconnection (TBR) branch swap- pretable restriction fragment patterns (Table 2). The three ping. Subsequently, analyses were performed using a cpDNA regions new to this study provided 30 useful region–

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:39 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 689

enzyme combinations, whereas rescreening of the trnT–trnF was detected in four Poa taxa known from the arctic (P. region yielded an additional three enzymes (EcoRI, NciI, glauca, P. hartzii ssp. ammophila, P. hartzii ssp. hartzii and RsaI) for a total of eight for that region. P. pratensis ssp. pratensis). Note that only arctic Poa taxa Interpreting restriction fragment patterns in terms of re- were sufficiently sampled to allow detection of infrataxon striction site presence versus absence was relatively straight- variation. However, despite the poor sampling within extra- forward for the three cpDNA regions (trnT–trnF, trnV–rbcL, arctic Poa taxa, restriction site variation was detected in five trnH–trnK) having minimal length variation. For the rbcL– of these taxa (P. annua, Poa cusickii Vasey ssp. cusickii, ORF106 and trnF–trnV regions, it was necessary to take into Poa fendleriana (Steudel) Vasey, P. nemoralis, Poa secunda account amplification product length variation in the inter- J. Presl. ssp. secunda). Restriction site variation among arc- pretation of fragment patterns. In this way, fragment pattern tic taxa at the species and generic levels is given in more de- variation resulting from restriction site differences was dis- tail below. tinguished from the sometimes considerable pattern variation because of length differences (e.g., fragment pattern Infraspecific variation variation resulting from digestion of trnF–trnV with HinfI Infraspecific variation in restriction sites was detected in was determined to be due solely to length variation). three arctic Poa species (Tables 1 and 2). Poa pratensis com- Differences in interpretation of restriction fragment pat- prises two haplotypes differing in three restriction sites terns between our present study and the previous one within the rbcL–ORF106 region. All arctic collections of the (Gillespie et al. 1997) are outlined below. Fragment pattern species (including all collections of P. pratensis variation resulting from digestion of the rbcL–ORF106 re- sspp. alpigena and colpodea and only arctic collections of P. gion with MspI enzyme was previously coded as pattern pratensis ssp. pratensis) share the haplotype pratensis1, rather than site differences. In this study with additional pat- whereas the six extra-arctic collections examined (including tern variation, we were able to interpret this variation in plants identified as Poa pratensis ssp. agassiziensis (B. terms of restriction sites. Upon closer examination of the Boivin & D. Löve) R.L. Taylor & MacBride and extra-arctic trnT–trnF patterns for P. alpina, a consistent pattern was no- collections of P. pratensis ssp. pratensis) share the haplotype ticed across several enzymes, which led us to hypothesize pratensis2. Poa glauca also exhibited within species restric- that insertion–deletion events, rather than restriction site dif- tion site variation. The haplotype glauca2, represented by ferences, were responsible for the somewhat different frag- three collections from two western arctic populations, dif- ment patterns. Several characters previously considered fered in a single trnT–trnF restriction site from the dominant unique to P. alpina were deleted and other characters haplotype (glauca1), which is found throughout the Cana- recoded. Following a recheck of the data, Poa secunda ssp. dian arctic and also from outside the arctic in the N.W.T., juncifolia (Scribner) Soreng was determined to have only and Colorado. A third unique haplotype was found in one one haplotype, contrary to the two recorded previously. individual from boreal N.W.T. (from the same population as an individual with the glauca1 haplotype) and differs from Restriction site variation glauca1 in four sites (the above trnT–trnF site, 2 trnF–trnV, Collections (corresponding to individual plants) within and 1 trnV–rbcL sites). each taxon were grouped according to their restriction site Differences between the two haplotypes of P. hartzii ssp. profile. Each group characterized by an identical haplotype hartzii, as previously determined (Gillespie et al. 1997), was considered as a separate evolutionary taxonomic unit were further emphasized here. An additional six restriction (ETU). Table 1 provides a summary of the taxa, the ETUs, site differences were found, for a total of 15 sites, distributed and the collections included within each ETU. The data ma- across all five cpDNA regions examined but with the great- trix comprising 52 ETUs and 114 restriction site characters est number, seven, found in the trnT–trnF region. Poa hartzii is given in Table 2. Because six species of Puccinellia (Pu. ssp. hartzii (hartzii2 haplotype) and P. hartzii ssp. ammo- andersonii Swallen, Pu. angustata (R. Br.) Rand & Redfield, phila were previously found to have identical haplotypes Pu. arctica (Hook.) Fern. & Weath., Pu. borealis, Pu. (Gillespie et al. 1997). New to this study are restriction site bruggemannii Sorensen, and Pu. poacea Sorensen) had iden- differences between these two P. hartzii subspecies and a tical haplotypes and also contained missing data for the low level of cpDNA variation within P. hartzii ssp. ammo- rbcL–ORF106 region, one ETU (Puccangustata) was used to phila. The dominant haplotype of P. hartzii ssp. ammophila, represent all six species. ammophila1, was found to differ in one trnT–trnF site and The trnF–trnV region was found to be by far the most four trnF– trnV sites from hartzii2. Four of these restriction variable region for the taxa and enzymes sampled, with a site differences are apomorphies unique to hartzii2. The total of 51 variable and interpretable restriction sites (Ta- infrasubspecific variation detected within P. hartzii ssp. bles 2 and 3). Among the remaining regions, trnT–trnF ammophila consisted of a single individual differing in one yielded 20 variable restriction sites, trnV–rbcL yielded 18 trnF–trnV site from all other individuals examined. This re- sites, the rbcL–ORF106 region 14 sites, whereas only 11 striction site character state is shared with both haplotypes variable sites were found within the trnH–trnK region. of P. hartzii ssp. hartzii. The majority of taxa examined consisted of a single ETU; that is, all individuals examined had identical haplotypes. Infrageneric and intergeneric variation The following six arctic Poa taxa are represented by a single A summary of the cpDNA variation characterizing Poa ETU: P. abbreviata, P. alpina, P. arctica ssp. arctica, P. and Puccinellia is provided in Table 3. Few sites were found arctica ssp. caespitans, P. pratensis ssp. alpigena and P. to be unique to each of the genera. Poa was defined by only pratensis ssp. colpodea. In contrast, restriction site variation three unique restriction sites, as was Puccinellia (although

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:39 AM Color profile: Disabled Composite Default screen

690 Can. J. Bot. Vol. 79, 2001 0 01 e trnH–trnK d trnV–rbcL c trnF–trnV and outgroup evolutionary taxonomic unit. 001110 0010011101 1000110110 0011011001 0101111010 1011010000 1011011000001111 1000000011001111 0010000101 1000 0010000101 1110110101 1110110101 0010001000 0010001000001000 0111110010 0111110010 0000010010 0010010010 0010010010 0100001110 1111111001 0010010100 1111111001 1001001011 0110101000 1001001011 0000000100 1010 1000 0000101000 1011000001 0101 Poa b ?????????? ???? ???????????????????? ???? ???? ?????????? ???? rbcL–ORF106 10 20 30 40 50 60 70 80 90 100 110 114 a trnT–trnF Data matrix of restriction site characters for each Table 2. arcticacaespitanspratensis1pratensis2agassizensis 0010101110 0010101110alpigena 0010101110colpodea 0010101110 1000100111 1000100111 0010101110arachnifera 1000100111chaixii 1000100111 1010010101 1010010101 0010101110 1000100111chambersii 0010101110 0010010100cusickii1 0010101110 1011010101 0110101110 0110101110cusickii2 1011010101 1000100111 0010101110 1000100111cuspidata 0010101110 0100100111 0010011101 0010101110 0010011101 0010101110fendleriana1 0010101110 0010010100 0010011101 0010101110 1000010110fendleriana2 0010010100 1000010110 0100100111 0010011101 0010010101 0100100111 0010101110 0010011101macrantha 1000010110 0010101110 0010101110 0010011001 0010101110 0010011001 0100100111nervosa 1000010110 0010101110 1000010110 0010010101 0100001110 0010101110 0010011001 1010010101 0010011101 0100100111wheeleri 1001110010 1001110010 0010011001 0010011101 0100100111 0100100111 0010011001glauca1 0010011101 1010010101 1001110010 0010101110 1000010110 0100001110 0011011010 0011011010 0100100111 0100001110glauca2 0010010101 1001110010 1000010110 1001110010 1000110110 0011011010 1010010101glauca3 1010010101 0010101110 0010011101 0010011001 0100001110 0010011101 0100000111 0011011010 1111011000 0010101110 1111011000 0010011001glaucaxhartzii 0010010101 0011011010 0100001110 0011011001 1001110010 0010011101 1000110110 1111011000compressa 1000110110 0100001110 1011101001 0100001110 0000100111 1001110010 0010010101 0010011101 1111011000 0001110010nemoralis1 0100100111 1010101001 1011101001 0101000111 0100001110 1111011000 0101000111 0011011010 1000110110 0011011001 0010011101 0010011101 0011011001nemoralis2 1010101001 0011011010 0101000111 1100101110 1000110110 0011010010 0010010101 0010011101 0100001110 0001110010 0011011001palustris 0101000111 0010010101 1010101001 0001110010 1000110110 1000110110 1100101110 1100101110 1000 1111011000 1000 0101000111 0011011001abbreviata 1010101001 1111011000 1000110110 0000011101 1100101110 1000 0011000101 1111011000 0011010010 0001110010 0100001110 0011010010 0011011001 0011011001trivialis 1010101001 0100001110 1000 1100101110 0011000101 0011000101 0001110010 0101000111 1000 0011011001 1000110110hartzii1 0011010010 0010011101 1100101110 0001110010 0001110010 0011010101 0101000111 1111011000 0100001110 1101000111 1010101001 1111011000 0010011101hartzii2 0011010010 1110101001 1100101110 0001110010 0011011001 0100001110 0100001110 0011010101 1000110110 1000 0011010010ammophila1 0011010010 0010011101 1111011000 0011010101 1000110110 0100001110 1000 1101000111ammophila2 0011010010 1000 1100101110 0010101101 0001110010 0010011101 0010011101 1111011000 1101000111 0011011001 0011010101 1100101110 0110001110 1001110110arida 1111011000 1111011000 1011011011 1011101001 0010011101 1101000111 0110001110 0010101110 0011010010 1001110110 1001110110 0001110010juncifolia 1111011000 0010101110 1000 0011010101 0010011101 1100000111 0011011001 1000 1101000111 0110001110 0010101110 0011010101 0001110010 1001110110napensis 0010011101 1100101110 0011011001 0011011001 1101000110 1101000111 0011010010 1000 1111011000 1100100111 1001110110 0101011010secunda1 0010011101 0011010010 1100110111 0110001110 0011011001 1101000111 1000 1001110110 0110001110 1100100111 0101011010 0101011010secunda2 0011010010 0010101110 0011000101 0011011001 1111011000 1000 1000 1001110110 0010011101 ?101111010 0010101110alpina 0011010101 1101000111 1111011000 0010011100 0011011001 0011010010 0011010101 0011010010 1000 0010101110bulbosa 0011010101 0101111010 0011011001 1001110110 0011010010 1011011000 0010101110 0100001110 1100100111 0101111010 1101000111 1001110110annua1 0110001110 1100100111 1000 1011011000 1011011000 0010101110 0110001110 1101000111 0011010010 0101111010annua2 1100100111 0110001110 0011011001 0010011101 0011010010 0011011001 1011011000 0010011100 1100100111 0011010101 0000000011autumnalis 1000 0011011101 0011010101 0011010010 0000100010 0010011101 1011011000 0101111010 1001110110 1100100111 1000 0000000011 0000000011sylvestris 0000100010 0101111010 0011010101 1001110110 1011011000 1001110010 1000000011wolfii 0011010101 0110001110 1000 0000010010 0011010010 1001110110 0011011001 0110001110 1011011000 0011010010 0111100001 0011011001Phippsia 0010010101 1000000011 0000101010 1000 100 0000010010 0111100001 0110001110 0001011001 0010011101 ?000000011 0011011001 0101011010 0010011100Puccangustata 0110001110 1000 1011011000 0101111010 0011100100 0000101010 1011011000 0010011101 1000000010 0101111010 1010110100Puccphryganodes 0110001110 1001110110 1000 1100100111 1010110100 0101111010 0011100100 0011010010 1001110110 0010011101Puccvahliana 0000101010 0011010010 0000101010 1000 1001110110 0011010010 0000101010 0010011101 1000000011 0011011001 1100100111 0000101010Arctagrostis 0011010010 1000000011 1000 0100010110 0011011001 1001110110 0010110101 1011011000 0100010110Arctophila 1011011000 0011011001 1100000110 10011

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:42 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 691

the Puccinellia + Phippsia group was characterized by 17 unique sites). Since the remaining outgroups were not included in the initial screening for restriction fragment pattern variation, the number of restriction sites defining these outgroups is likely to be greatly underestimated, and as a result, their relationships poorly resolved. For this reason these I sites; Nos. 18 and outgroups are not included in this comparison. I sites; Nos. 54–62 are Rsa I sites; and Nos. 84–85

cII sites; Nos. 98–100 are Sites variable within Poa comprised, by far, the largest Hha Taq I sites; and No. 114 is an

Hin category, with 80 variable sites (of a total of 114), of which 70 were cladistically informative (i.e., sites with gains or Msp I sites. losses that are shared by two or more ETUs). In sharp con- Sin trast to this high degree of cpDNA variation in Poa is the extremely low degree of variation found within Puccinellia. Six of the eight arctic Puccinellia species sampled share identical haplotypes. Only Puccinellia phryganodes (Trin.) I site; Nos. 16 and 17 are I site; Nos. 77–83 are Scribn. & Merr. and Puccinellia vahliana (Liebm.) Scribn. RI sites; Nos. 50–53 are Nci Sin & Merr. were found to differ by two and one unique sites, Eco

III sites; Nos. 96 and 97 are respectively. Puccinellia vahliana shares a single site with III site; Nos. 112–113 are

Hae ETUs outside of the genus (a parallel loss), but no Hae I sites; and Nos. 32–34 are cladistically informative sites are shared among species of

Msp Puccinellia.

96I site; No. 76 is a Cladistic analysis of restriction site data RI sites; No. 15 is an

Sau The data matrix used in the cladistic analysis comprises Eco I sites; Nos. 47–49 are 52 ETUs and 114 restriction site characters (Table 2). Dde I site; Nos. 93–95 are Ninety-eight characters are cladistically informative, i.e., RV sites; No. 111 is an

Dde shared by two or more ETUs, whereas 16 sites define single

Eco ETUs (Table 3). Data was scored as missing (?), signifying

I sites; Nos. 30 and 31 are “state unknown,” for the following reasons: cpDNA region

I sites; No. 75 is a was not successfully amplified (rbcL–ORF106 region for Hha

Rsa four ETUs), or interpretation of fragment patterns was am- I sites, Nos. 11–14 are biguous for a particular site because of length variation of Dra I site; No. 92 is a amplified product (trnF–trnV region). The eight outgroup

Bgl ETUs listed in Table 2 were included in all analyses. Note OI site; Nos. 45 and 46 are

Bst that branch lengths are underestimated (sometimes considerably so) and relationships poorly resolved in outgroup

I sites; Nos. 109 and 110 are clades, since the outgroup taxa, with the exception of Pucci- III sites; Nos. 26–29 are II site; Nos. 72–74 are

Dde nellia, were not screened for variable restriction sites. Hae Pvu Parsimony analysis of this data matrix, with characters I sites; Nos. 7–10 are I sites; No. 91 is a treated as normal reversible (Fitch parsimony), resulted in a Dde Apa II site; No. 44 is a single most parsimonious tree 157 steps in length with a

Bgl consistency index (CI) of 0.73 and a rescaled consistency in- I sites. dex (RC) of 0.68 (Fig. 1). Within the genus Poa, P. sect. Sin Sylvestres V. Marsh ex Soreng formed the basalmost clade,

I site; No. 71 is a followed by a clade comprising P. annua (sect. Ochlopoa)

Pst and P. alpina (sect. Alpinae)+P. bulbosa (sect. Bolbo- phorum). This latter clade has particularly long branch

I site. lengths compared with the P. sect. Sylvestres clade. The remaining species group into two clades. The first comprises Xho

OI restriction sites; Nos. 23–25 are P. sects. Poa, Dioicopoa E. Desv., Homalopoa Dumort., and I restriction sites; Nos. 107 and 108 are OI restriction sites; Nos. 3–6 are Bst I restriction sites; Nos. 89–90 are I restriction sites; No. 43 is a Madropoa Soreng. Within this clade, P. sect. Poa forms a Alu Bst I sites; No. 70 is a Alu Alu strongly supported clade, whereas the other three sections I site; and Nos. 102–103 are Nci together form a weakly supported clade with no resolution Pst of the three sections. The second main clade is comprised of ). P. sect. Secundae forming a basal paraphyletic complex (i.e., not defined by any shared restriction sites), P. trivialis (sect. Pandemos) and a clade comprising P. sects. Abbreviatae and concluded 3AI sites; and No. 20 is an

( Stenopoa. The ETU acronyms are explained in Table 1. The order of ETUs approximates that of the cladogram (Fig. 1). I sites. Sau Constraining restriction site gains as uniquely derived I sites; Nos. 63–69 are Xho fI sites; No. 101 is a I site. Character Nos. 1 and 2 are Character Nos. 21 and 22 are Character Nos. 86–88 are Character Nos. 35–42 are Character Nos. 104–106 are (Dollo parsimony) resulted in two most parsimonious trees Note: a b c d e Table 2 DupontiaHierochloe 0000100010 0000100010 1100100111 1100100111 0110111001are 0110111001 1110001100Hin 1110001100 1010000101 1010000??1 1000110110 1000110110 0011011000 0011011000 0101111000 0101111000 0100110011 0100110011 1111111000 1111111000 1001000011 1001000011 0010 0010 Msp Nci 19 are 171 steps long (CI = 0.67, RC = 0.65). Figure 2 illustrates

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:42 AM Color profile: Disabled Composite Default screen

692 Can. J. Bot. Vol. 79, 2001

Table 3. Amplification product length, total number of variable restriction sites, and total number of cladistically informative sites (i.e., shared by two or more ETUs) for the five cpDNA regions examined.

trnT–trnF rbcL–ORF106 trnF–trnV trnV–rbcL trnH–trnK All regions examined PCR product length (base pairs) 1750–1850 2250–2750 3100–3350 3900–4000 1950–2000 -13 450 Total no. of variable sites 20 14 51 18 11 114 Total no. of informative sites 18 12 42 17 9 98 Poa, no. of unique sites 0 2 1 0 0 3 Poa, no. of variable sites 20 8 35 10 7 80 Poa, no. of informative sites 18 8 29 9 6 70 Puccinellia, no. of unique sites 0 0 2 1 0 3 Puccinellia, no. of variable sites 0 0 2 0 2 4 Puccinellia, no. of informative sites 0 0 0 0 1 1 Puccinellia + Phippsia, no. of unique sites 0 1 7 6 3 17 Note: For Poa and Puccinellia, the number of unique sites defining the genus and the number of variable and cladistically informative sites within the genus are given.

only those parts of the cladogram that differ from the clado- genus is genetically diverse. The majority of species exam- gram described above (Fig. 1). The two Dollo trees differ ined could be distinguished readily by cpDNA restriction from each other only in the order of branching of taxa within site characters and were characterized by unique cpDNA the P. sect. Poa clade. Overall the cladograms produced by haplotypes. Exceptions were primarily species pairs and the Fitch and Dollo parsimony analyses are very similar complexes (e.g., Poa sylvestres A. Gray and Poa autumnalis (Figs. 1 and 2). Apart from differences in the lengths of Muhl. ex Elliot; P. napensis and P. secunda) that belong in branches, they differ only in the branching order within P. sect. the same section and are considered to be closely related. Secundae and in the uncertainty in branching order in the P. The clade comprising P. sects. Homalopoa, Madropoa, and sect. Poa clade. The P. napensis and P. secunda ETUs form a Dioicopoa was particularly poorly resolved with members of clade in the Dollo analysis but not in the Fitch analysis. different sections sometimes having identical haplotypes. It must be noted that extra-arctic species were not included in Discussion the initial screening for variable fragment patterns, so rescreening with enzymes not used in the study may provide Variation among cpDNA regions additional sites. Soreng (1990) also determined the degree of In terms of numbers of variable and informative restric- cpDNA variation in Poa to be high enough to adequately de- tion sites the trnF–trnV region was, by far, the most useful, fine most taxa and resolve higher level phylogenetic rela- whereas the trnH–trnK region was the least useful (Table 3). tionships. This study was also unable to resolve relationships However, in addition to differences in number of sites, there within the Homalopoa, Madropoa, and Dioicopoa clade. may also be a large difference in the information content of The very low level of cpDNA variation found in arctic each region; one of the more evident examples of this con- Puccinellia is in sharp contrast to the relatively high level cerns P. sect. Poa and the rbcL–ORF106 region. All varia- found in arctic Poa. The majority of Puccinellia species ex- tion detected between P. arctica and P. pratensis and within amined share identical cpDNA haplotypes. Only two spe- P. pratensis was found only in this region. No differences cies, Pu. phryganodes and Pu. vahliana differed from this were detected based on the other regions sampled nor were pattern, each differing in two sites, with only one of these any found in a previous study (Soreng 1990), which sampled sites being potentially cladistically informative. These two the entire cpDNA genome but with fewer enzymes. Variation species are also morphologically divergent from other arctic in the rbcL–ORF106 region also accounted for the infra- Puccinellia species (Consaul and Gillespie 2001); Pu. specific variation detected in P. cusickii, P. fendleriana, and vahliana often has been treated in the primarily Asian genus P. secunda. Two sites (Table 2, sites 21 and 24) were partic- Colpodium Trin. (Porsild and Cody 1980; Tzvelev 1983), ularly variable, and are perhaps more useful at the infra- whereas Pu. phryganodes is the only species to have a rhizo- specific level than above. matous habit. Choo et al. (1994), in a study that included Differences between the phylogenetic analysis of species representative of the morphological diversity within Gillespie et al. (1997) and the one presented here are also Puccinellia, also found a low degree of restriction site varia- due to the number of cpDNA regions sampled (the previous tion. While 15 restriction sites varied within Puccinellia in analysis was based on a subset of the regions used here) and this study, only three were cladistically informative. the different information content of each region. For exam- ple, the strongly supported monophyly of P. sect. Poa in this Why is there such a great difference in degree of cpDNA analysis was due to shared restricton sites in the trnF–trnV, infrageneric variation between Poa and Puccinellia? One trnV–rbcL, and trnH–trnK regions. In our previous study, reason may be very different evolutionary histories. The ge- which did not include these regions, P. sect. Poa was para- nus Puccinellia may have evolved and diversified much phyletic. more recently than Poa. Species of Puccinellia in the Cana- dian Arctic and possibly also in North America are perhaps Chloroplast DNA variation much more closely related to one another and form a single or few clades, whereas Canadian Arctic and North American Poa versus Puccinellia Poa species belong to many different clades (and sections) Despite the apparent morphological uniformity in Poa, the and are much more representative of global generic diver-

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:43 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 693

Fig. 1. Analysis of relationships in Poa based on cpDNA restriction site data showing the single most parsimonious tree (CI = 0.73, RC = 0.68) from the Fitch parsimony analysis. Tree is shown as a phylogram, with branch lengths proportional to the number of restriction site changes. Bootstrap values of 50% and over are shown at the nodes. Sections of Poa are given on the right.

sity. Alternatively, cpDNA may evolve at a much slower that infraspecific cpDNA variation is relatively common and rate in Puccinellia than Poa. But it is unclear why this might not rare as is often assumed. This finding would suggest that be so, as the two genera share similar life-history character- phylogenetic studies at the genus level or below that use istics, with the main difference being a preference for saline these cpDNA regions or others having similar levels of vari- habitats by Puccinellia. ation need to sample more than one individual per species for adequate representation of a species. Poa infraspecific variation Although infraspecific restriction site varation was detected, The finding of infraspecific variation in three of six arctic this variation generally did not correspond with recognized Poa species examined and five of six extra-arctic Poa spe- subspecies or varieties in either arctic or extra-arctic species cies (the latter represented by low sample sizes and includ- of Poa. The exception to this is P. hartzii, in which the two ing only those for which more than one individual was subspecies examined, ammophila and hartzii, were distin- examined) corroborates Harris and Ingram’s (1991) finding guished by five restriction sites (comparing the dominant

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:47 AM Color profile: Disabled Composite Default screen

694 Can. J. Bot. Vol. 79, 2001

Fig. 2. Analysis of relationships in Poa based on cpDNA restric- clades supported in the cladistic analyses presented here and tion site data. Two most parsimonious trees (CI = 0.67, RC = the sectional classification of Soreng (1998). Poa sects. Poa 0.65) from the Dollo parsimony analysis, showing only the two and Sylvestres are resolved as strongly supported and clades that differ from the cladogram in Fig. 1. The two Dollo weakly supported clades, respectively, supporting the status trees differ from each other only in the P. sect. Poa clade, as of each of them as distinct monophyletic groups. Poa annua shown at top. and P. trivialis, the only representatives examined of P. sects. Ochlopoa and Pandemos, respectively, are fairly isolated taxa on long branches, suggesting that these sections are also distinct taxa. Poa sect. Secundae appears as a paraphyletic complex basal to the Stenopoa–Abbreviatae– Pandemos clade. Poa sects. Stenopoa and Abbreviatae together form a strongly supported clade that is poorly resolved internally. Likewise, Poa sects. Homalopoa, Diocopoa, and Madropoa together form a clade, but with no resolution of the individual sections. Poa alpina and P. bulbosa, the only representatives of Poa sects. Alpinae and Bolbophorum examined, respectively, were found to share an identical cpDNA haplotype and formed a strongly supported clade distinguished by numerous apomorphies. Spe- cific relationships of the arctic taxa and congruence with the classifications of Soreng (1998) and others are discussed in the following section. Although our previous cladistic analysis of arctic Poa (Gillespie et al. 1997) involved fewer sites, taxa, and outgroups, there is much congruence between the two analyses. In both, P. alpina is the basalmost arctic taxon and P. sects. Stenopoa, Abbreviatae, and Secundae form a clade with sect. Secundae basal and paraphyletic. The main difference in- volves P. sect. Poa, which was paraphyletic with P. pratensis and P. arctica as separate but consecutive branches in our previous study (branches separated by a single restriction site, which P. pratensis shared with the Stenopoa– Abbreviatae–Secundae clade, but not with P. arctica). Adding more outgroups (particularly several that appear to be more closely related to Poa than Puccinellia), more taxa (such as those in the Homalopoa–Diocopoa–Madropoa complex) and examining additional cpDNA regions resulted in a haplotypes ammophila1 and hartzii2). The restriction site strongly supported, monophyletic P. sect. Poa. variation detected in P. pratensis and P. secunda did not Our results are, in general, consistent with the results of correspond to currently recognized subspecies. No variation the only other cladistic analysis of the genus, Soreng’s was found in P. arctica for which two subspecies were ex- (1990) cpDNA restriction site study. Soreng’s (1998) sec- amined, whereas variation was detected in several species, tional classification is based partly on the results of his such as P. glauca, for which subspecies were either not rec- cpDNA cladistic analysis. Allowing for the very different ognized or not examined. sets of Poa taxa included, the overall structures of the result- Poa infraspecific (or infrasubspecific in P. hartzii) varia- ing cladograms are very similar in the two phylogenetic tion often represented variation within a population. Multi- studies. Both studies indicate that P. alpina, P. annua, and P. ple haplotypes were detected in three of six populations of P. sect. Sylvestres are basal taxa in the genus, with P. sect. glauca, one of three populations of P. hartzii ssp. ammophila Sylvestres (plus P. eminens of P. sgen. Arctopoa (Griseb.) and three of eight populations of P. hartzii ssp. hartzii Prob., unpublished data) as the basalmost clade. The remain- (counting only those populations in which two or more indi- ing taxa group into two main clades, each having a similar viduals were examined). Intrapopulation variation was also structure and constituent taxa (at least at the sectional level). detected in P. annua, the only extra-arctic species for which Although relationships of specific species are generally not more than one individual per population was examined. This comparable, because of the very different set of taxa used, variation may be the result of multiple lineages originating our study did provide much better definition and resolution from a polymorphic ancestor, novel mutations within a pop- of taxa within P. sect. Poa. Unfortunately, because of the dif- ulation, or hybridization and introgression of closely related ferent data collection methods, the cpDNA data from the taxa (Harris and Ingram 1991). In the exceptional case of P. study by Soreng (1990) cannot be combined with this study. hartzii ssp. hartzii, the variation appears to be the result of hybridization and introgression of two distantly related taxa. Arctic Poa species and their phylogenetic relationships Poa phylogeny and sectional classification Poa abbreviata In general, there is reasonably close correspondance of Poa abbreviata ssp. abbreviata is characterized by a unique

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:49 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 695

and invariant haplotype, which is consistent with its distinct appears to be, at least, two very different types of densely morphology and low level of morphological variation in the tufted Poa in the Canadian Arctic. One type, generally Canadian Arctic. Our data suggest a strong affinity with spe- treated as P. arctica ssp. caespitans, comprises mostly ro- cies in P. sect. Stenopoa and, although not inconsistent with, bust plants with broad flat leaves and often tall flowering does not provide support for classification of the species in a culms generally found in mesic stony habitats, whereas a separate section, Abbreviatae. Relationships of taxa in the second type comprises very low mound-forming plants with Stenopoa–Abbreviatae clade (Figs. 1 and 2) are poorly re- very short culms and is characteristic of cold moist to wet solved in our analyses and additional data are needed to pro- meadows at higher elevations (Table 1, voucher No. 6647). vide resolution here. Other subspecies of P. abbreviata plus This suggests that, in the Canadian Arctic, plants having a additional species of the primarily Beringian P. sect. tufted growth form represent at least two different lineages Abbreviatae should be examined. Soreng (1990) included and not a single genetically distinct taxon. two different species of P. sect. Abbreviatae in his cpDNA Poa arctica is a highly polymorphic species in the Cana- study and similarly found both to be aligned with P. sect. dian Arctic. Since this variation is far from being discrete Stenopoa in an unresolved relationship. Soreng (1991a) also morphologically or geographically, we question the validity comments that the two sections are morphologically allied and utility of recognizing infraspecific taxa in this area. and that intersectional hybrids are known. Much of this variation seems to be correlated with habitat and thus taxa, may perhaps be better described as different Poa alpina ecotypes. Poa arctica ssp. caespitans accounts for only a The two very distant populations sampled (Baffin Island small part of this variation. We tend to agree with Polunin and alpine Colorado) shared an identical cpDNA haplotype. (1940) that there appears to be so much variation, and plants More populations, including those from the subarctic, need vary “in such a mixed and baffling manner that I have given to be examined for an understanding of infraspecific varia- up on sorting out the more obvious traits.” Further molecular tion. Poa alpina is the most phylogenetically basal and iso- studies using more variable DNA regions are needed to de- lated of the arctic species, a finding consistent with Soreng’s termine whether the different morphological types of P. study (1990). A phylogenetic position close to P. bulbosa arctica should be recognized as distinct taxa. but distant from all other species examined in this study is strongly supported by our data. This is consistent with Reports concerning a viviparous rhizomatous form of P. Edmondson’s (1978) classification, which groups the two arctica in the Canadian Arctic may be mostly erroneous. Al- species together in P. sect. Bolbophorum. In contrast, though Porsild and Cody (1980) mapped P. arctica var. vi- Tzvelev’s (1983) classification, which suggests a close rela- vipara in 18 different localities throughout the Eastern tionship with P. pratensis and P. arctica (see also Nannfeldt Arctic, we have not been able to verify any of these records. 1940), in addition to P. bulbosa, is not supported by our Upon examination of herbarium specimens at CAN and data. Since P. alpina and P. bulbosa were found to share an DAO, none were located at CAN, whereas the several speci- identical cpDNA haplotype, Soreng’s (1998) classification of mens at DAO were redetermined as P. pratensis ssp. the two species in separate sections is also not supported. In colpodea. Porsild (1955) previously commented that most our analysis, P. alpina + P. bulbosa formed a strongly sup- viviparous High Arctic Poa seem best placed in P. arctica as ported clade with another basal but derived species P. annua (P. var. vivipara but, unfortunately, provided no explanation. It sect. Ochlopoa), whereas in Soreng’s (1990) analysis P. alpina appears that he may have reidentified collections of P. and P. annua are separate, but successive basal branches. pratensis ssp. colpodea as P. arctica var. vivipara in his subsequent works (Porsild 1957; Porsild and Cody 1980). Based Poa arctica on our examination of field and herbarium specimens, P. Our results suggest that P. arctica is very closely related pratensis ssp. colpodea appears to be the common vivipa- to P. pratensis, differing in only two or three restriction sites rous rhizomatous Poa in the Canadian Arctic Islands and P. in the rbcL–ORF106 region. Together, they form a strongly arctica var. vivipara is either rare or not a distinct taxon. Our supported clade (six synapomorphic sites), supporting their cpDNA results of viviparous rhizomatous Poa (identified as traditional classification together in P. sect. Poa (Tzvelev P. pratensis ssp. colpodea) having a P. pratensis haplotype 1983 as subsect. Poa; Soreng 1998) and confirming previous and not the P. arctica haplotype, so far, support this conclu- molecular evidence (Soreng 1990). These results do not sup- sion. One of the localities sampled, Pond Inlet, was indi- port the statement that “P. arctica shows only a very remote cated by Porsild and Cody (1980) as a locality for P. arctica affinity to the group P. pratensis s.l.” (Tzvelev 1995, p. 197). var. vivipara but not P. pratensis ssp. colpodea. Interestingly, The two subspecies of P. arctica examined (arctica and there also appears to have been some confusion between the caespitans) were not differentiated in our study; all 18 indi- two taxa in the Russian Arctic (Tzvelev 1983, 1995). Given viduals examined share an identical cpDNA haplotype. that spikelets are typically incompletely developed in plants Thus, we found no evidence for P. arctica ssp. caespitans as with vegetatively proliferating inflorescences, one reason for a distinct taxon nor as a hybrid between P. arctica and P. this confusion may be the lack of important spikelet charac- glauca. Recent field observations also do not support the teristics often used to distinguish the two taxa. Additional concept of two morphologically distinct subspecies in the collections of viviparous rhizomatous Poa should be ana- Canadian Arctic. Regarding the main character (distinctly lyzed to adequately address this question. tufted versus rhizomatous habit) used to distinguish the two Populations of viviparous P. arctica ssp. arctica have also subspecies, there appears to be a continuum in degree of been found in coastal arctic Yukon. While these have been tuftedness, with intermediate conditions common, such as referred to as P. arctica var. vivipara (Cody 1996), these loosely tufted with some short rhizomes. In addition, there plants are much more robust than what is typically referred

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:50 AM Color profile: Disabled Composite Default screen

696 Can. J. Bot. Vol. 79, 2001

to as that variety and most likely do not represent the same son, many currently recognized Poa species differ in only taxon. Other viviparous Canadian Arctic P. arctica include one or two sites from closely related species (e.g., most spe- collections of P. arctica ssp. caespitans from Baffin Island cies in P. sect. Secundae or the Homalopoa–Madropoa– that developed vegetatively proliferating inflorescences un- Dioicopoa clade), whereas some species pairs or complexes der greenhouse conditions, but were not observed to be vi- were found to share identical haplotypes, e.g., P. autumnalis viparous in the field. and P. sylvestris, P. napensis and P. secunda (secunda1 and juncifolia haplotypes). This comparatively large restriction Poa glauca site difference between P. hartzii sspp. ammophila and Three haplotypes were detected in P. glauca. The majority hartzii suggests that they may be better recognized at the of individuals examined, including those from boreal and al- species level (following Porsild 1943; Porsild and Cody pine areas in addition to the Arctic, had the same dominant 1980; Cody 1996). In contrast, examined viviparous individ- haplotype. A second very similar haplotype (glauca2) was uals of P. hartzii ssp. hartzii had the hartzii2 haplotype, thus detected only in the western Low Arctic, in two mixed popu- providing no evidence for their recognition as a separate tax- lations. A third haplotype (glauca3) was detected in a single onomic entity. individual from boreal N.W.T., and is intermediate between The occurrence of both subspecies at Cape Dalhousie on the two main P. glauca haplotypes and those of the P. the mainland N.W.T. Beaufort Sea coast, which was sug- nemoralis, P. palustris and Poa compressa L. complex. This gested on the basis of morphological evidence (Soreng individual was identified as P. glauca ssp. glauca but be- 1991b), is verified by our restriction site study. This locality longed to a heterogeneous population that included plants represents a western range extension for P. hartzii ssp. identified as P. glauca sspp. glauca, rupicola, and glauca hartzii in North America and is one of only three known on tending towards Poa nemoralis ssp. interior (Rydb.) W.A. the Canadian Arctic mainland. Cape Dalhousie is also the Weber (Soreng, herbarium annotations 1999). Its intermedi- only site where cpDNA variation was detected in P. hartzii ate haplotype may be the result of phylogenetic infraspecific ssp. ammophila. The one restriction site difference found in variation or, perhaps more likely, due to hybridization and a single individual is shared with P. hatrzii ssp. hartzii introgression with the sympatric P. palustris or P. nemoralis. (along with most Poa taxa), perhaps suggesting a low level In contrast, a second individual examined from this popula- of hybridization and cpDNA exchange between the two sub- tion and identified as P. glauca ssp. rupicola was found to species (although this would infer an atypical mode of have the dominant P. glauca haplotype. Additional boreal cpDNA inheritance or gene exchange). and alpine collections of P. glauca need to be examined for a Soreng (1991b, p. 407) suggested that P. hartzii ssp. more thorough investigation of infraspecific cpDNA varia- ammophila may prove “to be a hybrid between P. secunda tion in P. glauca and to determine whether the intermediate and P. hartzii.” Poa hartzii ssp. ammophila has a cpDNA glauca3 haplotype occurs only in this population or is more haplotype distinct from both putative parents, suggesting ei- widespread. ther that it is a distinct taxon not of recent hybrid origin or No firm evidence of hybridization was detected in P. that a taxon or population within the P. secunda complex not glauca on the basis of our cpDNA results. Although the spe- yet examined is a parent. Based on our results the dominant cies is hypothesized to hybridize and introgress with P. haplotype of P. hartzii ssp. ammophila was found to be iden- hartzii ssp. hartzii (as discussed under P. hartzii), cpDNA of tical to Poa arida Vasey known from the Great Plains and that taxon was not detected in individuals of P. glauca. Canadian Prairies. However, these two taxa are morphologi- Therefore, introgression of cpDNA appears to be unidirec- cally quite distinct and have never been thought to be related tional from P. glauca to P. hartzii ssp. hartzii. previously. Interestingly, P. arida is itself hypothesized to be Poa glauca belongs to P. sect. Stenopoa, which (together a stabilized hybrid between P. secunda and either P. with P. sect. Abbreviatae) is not well resolved in our analy- pratensis or P. arctica (Soreng and Kellogg, in press). Addi- ses (Figs. 1 and 2). Most of the species examined have iden- tional collections of P. arida need to be examined, since the tical or very similar cpDNA haplotypes. Interestingly, P. single individual examined is not morphologically typical of glauca appears to be the exception, with two very similar the species and may itself be the result of hybridization pos- haplotypes that differ considerably from other species (apart sibly with P. secunda (identification by R.J. Soreng). from the single glauca3 individual discussed above). The Two hypotheses concerning the origin of P. hartzii ssp. suggestion that P. glauca should be treated as a subspecies hartzii were formulated by Gillespie et al. (1997) and out- of the widespread Eurasian species P. nemoralis (as pro- lined in the introduction. These hypotheses were based on posed in a recent draft manuscript for Flora of North Amer- the presence of two very different cpDNA haplotypes, one ica; Soreng and Kellogg 2001) is not supported by our data. identical to a haplotype of P. glauca, the other to taxa in P. The two haplotypes of arctic P. glauca differ from those of sect. Secundae. The present study confirms that the hartzii1 P. nemoralis in four to six restriction sites in three or four re- haplotype is identical to the dominant haplotype of P. gions, a greater difference than between many currently rec- glauca. In contrast, the hartzii2 haplotype, although con- ognized species of Poa. firmed as part of the P. sect. Secundae complex, was determined to be a unique haplotype characterized by four Poa hartzii uniquely derived restriction sites. These results are consis- The two subspecies of P. hartzii examined, subspecies’ tent with the hypothesis of P. hartzii ssp. hartzii as a preex- ammophila (the dominant ammophila1 haplotype) and isting taxon that “captured” P. glauca cpDNA via hartzii (hartzii2 haplotype), differ in five restriction sites, of hybridization and introgression. The alternate hypothesis of which four are uniquely derived in ssp. hartzii. In compari- P. hartzii ssp. hartzii as a stabilized intersectional hybrid be-

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:51 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 697

tween P. glauca and a species in P. sect. Secundae is not several different localities; the one examined here was supported. All northern Canadian taxa belonging to P. sect. determined to have the glauca1 haplotype. Although recent, Secundae, plus other more southerly species, were examined ongoing introgression is the most likely explanation, an al- in an attempt to determine a second parental species. No ternative possibility is stabilized introgressed populations taxon within P. sect. Secundae, nor any other species exam- having two codominant haplotypes that have reached a state ined, was found to have a haplotype identical to or even of equilibrium. closely resembling the hartzii2 haplotype. Thus the hartzii2 The above hypothesis is complicated by the fact that P. haplotype was determined to be unique to P. hartzii ssp. hartzii ssp. hartzii appears to be apomictic with sterile pol- hartzii, strongly suggesting that it is the original or ancestral len (Soreng 1991b, Gillespie et al. 1997). If this taxon occa- haplotype of that subspecies. sionally bears fertile pollen, then it could act as the pollen Furthermore, the hartzii2 haplotype was found in popula- donor in the initial hybridization event, with P. glauca acting tions in all three geographical regions sampled (High Arctic, as the maternal parent and plastid donor. On the other hand, Victoria Island, and mainland N.W.T.), whereas the hartzii1 if P. hartzii ssp. hartzii never bears fertile pollen, then P. haplotype was detected only in High Arctic populations. Al- glauca would be both the pollen donor and plastid donor, though P. hartzii ssp. hartzii and P. glauca co-occur in all which would imply a paternal or biparental mode of plastid three regions, our cpDNA evidence suggests that hybridiza- inheritance (modes of inheritance reviewed for angiosperms tion and introgression resulting in the transfer of P. glauca by Harris and Ingram (1991)). Introgression via backcross- cpDNA to P. hartzii ssp. hartzii is taking place only in the ing of the hybrid with P. hartzii ssp. hartzii, likewise, would High Arctic. imply either the latter as the pollen donor (and the hybrid as In the High Arctic both haplotypes of P. hartzii ssp. hartzii the plastid donor) or the hybrid as both the pollen and appear to be equally frequent (15 versus 16 individuals) and plastid donor. were found to co-occur in three of the four most extensively Counterfeit hybridization may provide an alternative ex- sampled populations (four to eight individuals). However, planation of cpDNA exchange between P. hartzii ssp. hartzii there appears to be considerable variation in the presence of and P. glauca. In this gene transfer process, chromosomally haplotypes and their frequency within a population. For ex- nonreduced eggs are stimulated to develop parthenogeneti- ample, seven of eight individuals at Lake Hazen had the cally by pollen of another species (pseudogamy), and al- hartzii1 haplotype, whereas only one of six individuals at though normal transfer of chromosomes from pollen to egg Tanquary had this haplotype. The occurrence and degree of does not take place, gene transfer can occur (de Wet et al. introgression is hypothesized to be dependent on the pres- 1984; de Wet 1986). As previously described in Gillespie et ence, proximity and size of P. glauca populations to P. al. (1997), pollen of P. glauca may stimulate P. hartzii ssp. hartzii ssp. hartzii populations. Although their habitat pref- hartzii ovules to develop (pseudogamy). If this process is erences overlap considerably, P. glauca appears to be absent leaky and some cpDNA transfer occurs, then P. glauca from, or very sparse on, coarse sand or marine clay deposits cpDNA could be introduced into P. hartzii ssp. hartzii di- where P. hartzii ssp. hartzii often thrives. The latter occurs rectly via counterfeit hybridization, without the need for the more commonly by itself at Tanquary, where these habitats intermediate step of true F1 hybrids followed by subsequent are much more common, than at Lake Hazen. Introgression backcrossing. was not detected in the most extensively sampled Axel Heiberg population, a coastal viviparous population growing Poa pratensis on a sand substrate in the absence of P. glauca. In contrast, Poa pratensis has infraspecific cpDNA variation with a only the glauca1 haplotype was detected in the other two, distinct geographical pattern. Restriction site data separates albeit poorly sampled, Axel Heiberg populations where P. collections examined into two groups, an arctic complex of glauca was also present. Habitat differences might be P. pratensis comprising subspecies’ alpigena, colpodea and thought to explain the lack of the hartzii1 haplotype in the arctic individuals of subspecies pratensis, and a non-arctic Low Arctic populations sampled. However, this is not the complex comprising P. pratensis ssp. agassiziensis and non- case, since P. glauca individuals were found growing with or arctic individuals of subspecies pratensis. The two cpDNA near P. hartzii ssp. hartzii in both Low Arctic populations groups, each characterized by a unique haplotype, differ in sampled. three restriction sites. This level of variation is greater than The geographical distribution of the hartzii1 haplotype that found within other species complexes or between many described above suggests multiple recent hybridization and species pairs of Poa. We hypothesize that the two cpDNA introgression events in the High Arctic. In stabilized intro- groups may represent a fundamental division between an in- gressed populations, one haplotype is thought to eventually digenous arctic complex and a primarily or entirely intro- dominate becoming the only haplotype in the population or duced non-arctic complex. species (Reiseberg and Soltis 1991). The presence of both The cpDNA variation detected in P. pratensis does not the original and introgressed haplotypes throughout the High correspond to the subspecific classification as currently ac- Arctic and in three of the four most extensively sampled cepted. Individuals identified as P. pratensis ssp. pratensis populations suggests populations that have not stabilized had one of two haplotypes depending on whether they were with recent and perhaps continuing introgression. The pres- arctic or extra-arctic. This may reflect our poor understand- ence of putative hybrids that appear to be morphologically ing of subspecies boundaries rather than being real infra- intermediate between P. hartzii ssp. hartzii and P. glauca subspecific variation. Arctic individuals keyed out as P. also suggests that these events are recent and ongoing. Al- pratensis ssp. pratensis lacked features generally used to though rare, these hybrids have been collected recently at characterize P. pratensis ssp. alpigena, such as hairs on me-

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:52 AM Color profile: Disabled Composite Default screen

698 Can. J. Bot. Vol. 79, 2001

dial nerves of lemma (Porsild and Cody 1980) and only two len fertility (Muntzing 1933; Bashaw and Hanna 1990). branches at the lowest node of the panicle (Edmondson Kellogg (1987) describes this species (and P. secunda,an- 1980; Cody 1996; Soreng and Barrie 1999). However, upon other difficult species complex) as often appearing to com- closer examination these individuals were found to be inter- prise distinct taxonomic units at a local scale but with mediate in overall morphology between the two subspecies. boundaries that blur at a broader scale. The ability to hybrid- All are from the Low Arctic and mostly from sites having ize with other, sometimes distantly related species of Poa varying degrees of human disturbance. Based on our cpDNA further complicates the picture (Soreng and Kellogg 2001). results these individuals are likely indigenous and most Given this evolutionary complexity the detection of genetic closely allied with P. pratensis ssp. alpigena. They may be markers distinguishing two groups in North America is an the result of hybridization of the two subspecies with the interesting finding and is currently being explored further. cpDNA of the native P. pratensis ssp. alpigena dominating. Our results are consistent with a classification treating P. The two arctic subspecies examined, P. pratensis pratensis in a section together with P. arctica but separate sspp. alpigena and colpodea, share an identical haplotype, from any other species examined here, as discussed under suggesting a close relationship but providing no evidence the heading Poa arctica above. supporting the recognition of the latter as a distinct subspecies. Although cpDNA of the viviparous form of P. pratensis Biogeography and Pleistocene survival ssp. alpigena was not examined, recent observations on Mel- Phylogeographic studies involving the examination of ville Island suggest that vivipary in this subspecies may be geographic variation in infraspecific genetic diversity have environmentally induced and that P. pratensis subsp. been used to detect areas of greater genetic diversity that alpigena var. vivipara is likely not a distinct taxon in arctic may provide evidence of where species survived the Pleisto- Canada. Culms with vegetatively proliferating florets were cene glaciations (Tremblay and Schoen 1999; Abbott et al. always found growing within patches of non-viviparous P. 2000). Although our analysis of cpDNA variation did not pratensis ssp. alpigena and, whenever the delicate rhizomes detect extensive infraspecific variation nor was it an in depth were successfully excavated, they were found attached to the phylogeographic study, several of our results are interesting same rhizome as non-viviparous culms (e.g., Gillespie & from a biogeographic perspective. In P. glauca, infraspecific Consaul specimen No. 6942). variation was detected in the western Low Arctic, but not In North America extra-arctic P. pratensis has generally from the High Arctic or eastern Low Arctic, suggesting that been treated under P. pratensis ssp. pratensis, but Soreng Canadian Arctic populations may derive from populations and Kellogg (2001) consider these populations to represent that survived the Pleistocene glaciation in a Beringian re- several European taxa (subspecies’ angustifolia and irrigata, fugium. More extensive sampling, particularly of eastern in addition to subspecies pratensis), a complex of cultivars Low Arctic, boreal and alpine populations is needed to test produced mostly in North America putatively from Euro- this hypothesis. In general, however, cpDNA restriction site pean stock (primarily subspecies pratensis) and a putative data detected insufficient infraspecific variation in arctic Poa native taxon. The latter taxon, P. pratensis ssp. agassizensis, for phylogeographic studies and other methods of DNA was only recognized in 1960 from Manitoba (Boivin and analysis are needed for such studies. Löve 1960). Individuals identified as this subspecies share a Our cpDNA results provide further insight into the bio- haplotype with extra-arctic P. pratensis ssp. pratensis. This geography of P. hartzii and the evolution of its subspecies as poorly understood subspecies, although perhaps once a ge- discussed previously in Gillespie et al. (1997). Poa hartzii netically distinct native form, may now be swamped with the sspp. hartzii and ammophila (and the Alaskan endemic ssp. cpDNA from widespread and pervasive introduced forms alaskana Soreng) were hypothesized to have originated by and cultivars. range fragementation and isolation in widely separated Interestingly, extra-arctic P. pratensis shares more restric- refugia during the Pleistocene. In this scenario the western tion sites with P. arctica than with arctic P. pratensis. All Arctic P. hartzii sspp. ammophila and alaskana diverged fol- restriction sites distinguishing these three haplotypes are lo- lowing isolation in Beringian refugia, whereas the primarily cated in one cpDNA region examined, rbcL–ORF106. Three High Arctic P. hartzii ssp. hartzii likely originated in High restriction site losses distinguish arctic from extra-arctic P. Arctic refugia. Given the restricted geographical distribution pratensis. In addition, the length of the rbcL–ORF106 region of the P. glauca haplotype in P. hartzii ssp. hartzii, hybrid- in arctic P. pratensis was found to be distinctly shorter than ization and introgression appears to have played a later role in both extra-arctic P. pratensis and P. arctica. The latter two in the evolution of this subspecies. An alternative hypothesis haplotypes have a similar PCR product length and differ suggested by their very distinct haplotypes is that P. hartzii from each other in only two restriction sites. Given this sig- sspp. ammophila and hartzii are not most closely related and nificant length difference, the restriction site losses may be diverged prior to the Pleistocene. the result of an insertion–deletion event, specifically a deletion in arctic P. pratensis, rather than the result of base sub- Conclusions stitutions. Poa pratensis in North America is considered to be a large Restriction site analysis of specific cpDNA PCR products polymorphic complex of native, introduced, and cultivated was found to be a useful method for examining both species forms that are questionably distinguishable (Soreng and delimitation and phylogeny in the genus Poa. Extensive Kellogg 2001; also cf. Soreng and Barrie 1999). The species within-species sampling of Canadian Arctic species was de- comprises numerous facultatively apomictic and pseudo- termined to be both informative and necessary to understand gamous races, having varying degrees of apomixis and pol- species delimitation and in detecting hybridization. Arctic

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:52 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 699

species were, in general, characterized by one or more Aurora Research Institute, Nunavut Research Institute, and unique cpDNA haplotype(s). Although infraspecific cpDNA Ellesmere Island National Park. We especially thank Rob variation was detected in three arctic species, it corre- Soreng for stimulating discussions, providing DNA of extra- sponded to subspecific taxa in only one species. Poa hartzii arctic species and determining Poa collections. Laurie sspp. hartzii and ammophila were each characterized by Consaul, Rob Soreng, Claus Vogel, and Françoise unique haplotypes that are as or more different than many Chatenoud are gratefully acknowledged for their enthusiastic closely related species, suggesting recognition at the species assistance in the field and Tina Saffioti, Izabella Szymanska, level. In contrast, the variation detected in P. pratensis had a and John Coltess for their assistance in the laboratory. Re- geographic rather than taxonomic basis and is hypothesized search was funded by grants from the Canadian Museum of to correspond to indigenous arctic versus introduced non- Nature and PCSP grants from the Canadian Museum of Na- arctic populations. Variation in P. glauca was detected only ture and PCSP. This paper is Polar Continental Shelf Project within western Low Arctic and subarctic populations and Contribution number 008-01. may be an historical consequence of greater genetic diversity where the arctic populations survived the Pleistocene glaciations. Our cpDNA analysis did not detect infraspecific References variation in P. arctica and was, therefore, not informative in examining the status or possible hybrid origin of its infra- Abbott, R.J., Smith, L.C., Milne, R.I., Crawford, R.M.M., Wolff, K., specific taxa. and Balfour, J. 2000. Molecular analysis of plant migration and The results of our expanded cpDNA study are consistent refugia in the Arctic. Science (Washington, D.C.), 289: 1343–1346. with one of the two hypotheses concerning the evolution of Aiken, S.G., Consaul, L.L., and Dallwitz, M.J. 1996a. Grasses of P. hartzii outlined in Gillespie et al. (1997). Populations of the Canadian Arctic Archipelago: a Delta database for interac- P. hartzii ssp. hartzii were determined to contain two very tive identification and illustrated information retrieval. Can. J. different haplotypes, one identical to P. glauca and the sec- Bot. 74: 1812–1825. Aiken, S.G., Consaul, L.L., and Dallwitz, M.J. 23 December ond a unique haplotype within P. sect. Secundae. This is 1996b. Grasses of the Canadian Arctic Archipelago: descrip- consistent with the hypothesis of P. hartzii ssp. hartzii as a tions, illustrations, identification, and information retrieval. preexisting taxon that has captured P. glauca cpDNA via hy- (accessed October 2000). bridization and introgression. Given that the P. glauca Arnold, M.L., Buckner, C.M., and Robinson, J.J. 1991. Pollen- haplotype was found only in High Arctic populations, hy- mediated introgression and hybrid speciation in Louisiana irises. bridization appears to be taking place only in this region. No Proc. Natl. Acad. U.S.A. 88: 1398–1402. individuals of P. glauca were found to have the P. hartzii Bashaw, E.C., and Hanna, W.W. 1990. Apomictic reproduction. In ssp. hartzii haplotype, suggesting that introgression is unidi- Reproductive versatility in the grasses. Edited by G.P. Chapman. rectional, from P. glauca to P. hartzii ssp. hartzii. Cambridge University Press, Cambridge, U.K. pp. 100–130. The phylogenetic analysis presented here of Canadian Bay, C. 1993. Taxa of vascular plants new to the flora of Green- Arctic and selected extra-arctic species is, in general, consis- land. Nord. J. Bot. 13: 247–252. tent with recent phylogenetic analyses and infrageneric clas- Böcher, T.W., Holmen, K., and Jacobsen, K. 1968. The flora of sification systems. Poa sect. Sylvestres was determined to be Greenland. English edition. P. Haase & Son, Copenhagen, Den- the most plesiomorphic of the taxa examined, whereas P. mark. alpina, P. bulbosa, and P. annua are basal but derived spe- Boivin, B. 1967. Énumération des plantes du Canada. VI cies. All remaining species examined grouped into two main Monopsides (2ème partie). Nat. Can. 94: 471–528. clades, one comprising P. sects. Poa, Homalopoa, Madro- Boivin, B., and Löve, D. 1960. Poa agassizensis, a new prairie poa, and Dioicopoa, and the second composed of P. sects. bluegrass. Nat. Can. 87: 173–180. Secundae, Pandemos, Abbreviatae, and Stenopoa. Poa Cayouette, J. 1984. Additions and extensions d’aire dans la flore glauca and P. abbreviata were each found to be distinct taxa vasculaire du Nouveau-Québec. Nat. Can. 111: 263–274. with characteristic cpDNA haplotypes within a generally un- Cayouette J., and Darbyshire, S.J. 1993. The intergeneric hybrid resolved P. sect. Stenopoa–Abbreviatae complex. Poa hartzii grass “Poa labradorica.” Nord. J. Bot. 13: 615–629. is confirmed as belonging to P. sect. Secundae. Poa sect. Choo, M.K., Soreng, R.J., and Davis, J.I. 1994. Phylogenetic rela- Poa, consisting of P. arctica and P. pratensis, is a strongly tionships among Puccinellia and allied genera of Poaceae as in- supported monophyletic group, not closely related to P. ferred from chloroplast DNA restriction site variation. Am. J. alpina. Where previous phylogenetic hypotheses or classifi- Bot. 8: 119–126. cations differ, our results were often informative in provid- Clayton, W.D., and Renvoize, S.A. 1986. Genera Graminum: ing support for a particular hypothesis. grasses of the world. Kew Bull. Add. Ser. XIII. Two additional collections of Poa alpina were examined Cody, W.J. 1996. Flora of the Yukon Territory. NRC Research Press, Ottawa, Ont. (Gillespie et al. 6749-1 from Ogac Lake, Baffin Island, and Consaul, L.L., and Gillespie, L.J. 2001. A re-evalution of species Djan-Chekar 00-126 from Newfoundland). Both collections limits in Canadian Arctic Island Puccinellia (Poaceae): resolv- have the alpina haplotype, identical to the three collections ing key characters. Can. J. Bot. 79. In press. of P. Alpina that were examined in this study. Demesure, B., Sodzi, N., and Petit, R.J. 1995. A set of universal primers for amplification of polymorphic non-coding regions of Acknowledgements mitochondrial and chloroplast DNA in plants. Mol. Ecol. 4: 129–131. We gratefully acknowledge the field support provided to de Wet, J.M.J. 1986. Hybridization and polyploidy in the Poaceae. us by the staff at Polar Continental Shelf Project (PCSP), In Grass systematics and evolution. Edited by T.R. Soderstrom,

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:53 AM Color profile: Disabled Composite Default screen

700 Can. J. Bot. Vol. 79, 2001

K.W. Hilu, C.S. Campbell, and M.E. Barkworth. Smithsonian Porsild, A.E., and Cody, W.J. 1980. Vascular plants of continental Institution Press, Washington, D.C. pp. 188–194. Northwest Territories, Canada. National Museum of Natural Sci- de Wet, J.M.J., Newell, C.A., and Brink, D.E. 1984. Counterfeit ences, National Museums of Canada, Ottawa, Ont. hybrids between Tripsacum and Zea. Am. J. Bot. 71: 245–251. Probatova, N.S. 1984. New taxa of the Poaceae from the far east of Dowling, T.E., Moritz, C., Palmer, J.D., and Rieseberg, L.H. 1990. the USSR [in Russian.] Bot. Zh. (Kiev), 69: 251–259 Nucleic acids III: analysis of fragments and restriction sites. In Reiseberg, L.K., and Soltis, D.E. 1991. Phylogenetic consequences Molecular systematics. 2nd ed. Edited by D.M. Hillis, C. of cytoplasmic gene flow in plants. Evol. Trends Plants, 5: 65– Moritz, and B.K. Mable. Sinauer Associates, Sunderland, Mass. 84. pp. 249–320. Rønning, O.I. 1996. Flora of Svalbard. 3rd ed. Norsk Polarinstitutt, Doyle, J.J., and Doyle J.L. 1990. Isolation of plant DNA from Oslo. Polarhåndb. No. 10. fresh tissue. Focus, 12: 13–15. Saghai-Maroof, M.A., Soliman, K.M., Jorgensen, R.A., and Allard, Dumoulin-Lapègue, S., Pemonge, M.-H., and Petit, R.J. 1997. An R.W. 1984. Ribosomal DNA space-length polymorphism in bar- enlarged set of consensus primers for the study of organelle ley: Mendelian inheritance, chromosomal location, and popula- DNA in plants. Mol. Ecol. 6: 393–397. tion dynamics. Proc. Natl. Acad. U.S.A. 81: 8014–8018. Edmondson, J.R. 1978. Infrageneric taxa in European Poa L. Bot. Sambrook, J., Fritsch, E.F., and Maniatis, T. 1989. Molecular clon- J. Linn. Soc. 76: 329–334. ing: a laboratory manual. 2nd ed. Cold Spring Harbour Labora- Edmondson, J.R. 1980. Poa L. In Flora Europaea. Vol. 5. Edited by tory Press, Cold Spring Harbor, N.Y. T.G. Tutin, V.H. Heywood, N.A. Burges, D.M. Moore, D.H. Scholander, P.D. 1934. Vascular plants from northern Svalbard. Valentine, S.M. Walters, and D.A. Webb. Cambridge University Skr. Svalbard Ishavet. 62: 1–153. Press, Cambridge, U.K. pp. 159–167. Scoggan, H.J. 1978. Gramineae (Grass family). In The flora of Felsenstein, J. 1985. Confidence limits on phylogenies: an ap- Canada. Natl. Mus. Nat. Sci. (Ottawa) Publ. Bot. No. 7. proach using the bootstrap. Evolution, 39: 783–791. pp. 218–337. Fredskild, B. 1996. A phytogeographical study of the vascular Simmons, H.G. 1906. The vascular plants in the flora of plants of West Greenland. Medd. Grønl. 45: 1–157. Ellesmereland. Report of the Second Norwegian Arctic Expedi- Gillespie, L.J., Consaul, L.L., and Aiken, S.G. 1997. Hybridization tion in the “Fram” 1898–1902. No. 2. The Society of Arts and and the origin of the arctic grass Poa hartzii (Poaceae): evidence Sciences of Kristiania. Kristiania, Norway. from morphology and chloroplast DNA restriction site data. Soltis, D.E., Soltis, P.S., and Mulligan, B.G. 1992. Infraspecific Can. J. Bot. 75: 1978–1997. chloroplast DNA variation: systematic and phylogenetic consider- Harris, S.A., and Ingram, R. 1991. Chloroplast DNA and ations. In Molecular systematics of plants. Edited by P.S. Soltis, D.E. biosystematics: the effects of intraspecific diversity and plastid Soltis, and J.J. Doyle. Chapman & Hall, New York. pp. 117–150. transmission. Taxon, 40: 393–412. Soreng, R.J. 1990. Chloroplast-DNA phylogenetics and biogeography in a reticulating group: study in Poa (Poaceae). Am. J. Bot. 73: Kellogg, E.A. 1987. Apomixis in the Poa secunda complex. Am. J. 1383–1400. Bot. 74: 1431–1437. Soreng, R.J. 1991a. Systematics of the “Epiles” group of Poa McLachlan, K.I., Aiken, S.G., Lefkovitch, L.P., and Edlund, S.A. (Poaceae). Syst. Bot. 16: 507–528. 1989. Grasses of the Queen Elizabeth Islands. Can. J. Bot. 67: Soreng, R.J. 1991b. Notes on new infraspecific taxa and hybrids in 2088–2105. North American Poa (Poaceae). Phytologia, 71: 390–413. Müntzing, A. 1933. Apomictic and sexual seed formation in Poa. Soreng, R.J. 1998. An infrageneric classification for Poa in North Hereditas, 27: 131–154. America, and other notes on sections, species, and subspecies of Murray, D.F. 1995. Causes of arctic plant diversity: origin and evo- Poa, Puccinellia,andDissanthelium (Poaceae). Novon, 8: 187–202. lution. In Arctic and Alpine Biodiversity. Edited by F.S. Chapin Soreng, R.J., and Barrie, F. 1999. Proposal to conserve the name Poa and C. Korner. Springer-Verlag, Berlin. pp. 33–43. pratensis (Poaceae) with a conserved type. Taxon, 48: 157–159. Nannfeldt, J.A. 1935. Taxonomical and plant-geographical studies Soreng, R.J., and Davis, J.I. 2000. Phylogenetic structure in Poaceae in the Poa laxa group. Symb. Bot. Ups. 1(5): 1–113. subfamily Pooidae as inferred from molecular and morphological Nannfeldt, J.A. 1940. On the polymorphy of Poa arctica R.Br. with characters: misclassification versus reticulation. In Proceedings, 3rd special reference to its Scandanavian forms. Symb. Bot. Upsal. International Symposium on Grass. Systematics and Evolution, 4(4): 1–85. 1998. Edited by S.W.L. Jacobs and J. Everett. Commonwealth Sci- Olonova, M.V. 1993. Phenogeographical study of Poa glauca Vahl. entific and Industrial Research Organization, Collingwood, Austra- Sib. Biol. Zh. 6: 70–74. lia. pp. 61–74. Palmer, J.D., Jansen, R.K., Michaels, H.J., Chase, M.W., and Soreng, R.J., and Kellogg, E.A. 2001. Poa. In Flora of North America. Manhart, J.R. 1988. Chloroplast DNA variation and plant phy- Edited by Flora of North America Editorial Committee. Oxford Uni- logeny. Ann. Mo. Bot. Gard. 75: 1180–1206. versity Press, New York. In press. Polunin, N. 1940. Botany of the Canadian Eastern Arctic. Part 1. Soreng, R.J., Davis, J.I., and Doyle, J.J. 1990. A phylogenetic anal- Bull. Natl. Mus. Can. No. 92. ysis of chloroplast DNA restriction site variation in Poaceae Polunin, N. 1959. Circumpolar arctic flora. Clarendon Press, Ox- subfam. Pooideae. Plant Syst. Evol. 172: 83–97. ford, U.K. Swofford, D.L. 1998. PAUP* phylogenetic analysis using parsi- Porsild, A.E. 1943. Materials for a flora of the continental North- mony (*and other methods), version 4 edition. Sinauer Associ- west Territories of Canada. Sargentia, 4: 1–79. ates, Sunderland, Mass. Porsild, A.E. 1955. The vascular plants of the western Canadian Taberlet, P., Gielly, L., Pautou, G., and Bouvet, J. 1991. Universal prim- Archipelago. Bull. Natl. Mus. Can. No. 135. ers for amplification of three non-coding regions of chloroplast Porsild, A.E. 1957. Illustrated flora of the Canadian Arctic Archi- DNA. Plant Mol. Biol. 17: 1105–1109. pelago. Bull. Natl. Mus. Can. No. 146. Tremblay, N.O., and Schoen, D.J. 1999. Molecular phylogeography Porsild, A.E. 1964. Illustrated flora of the Canadian Arctic Archi- of Dryas integrifolia: glacial refugia and postglacial re- pelago. 2nd ed. revised. Bull. Natl. Mus. Can. No. 146. colonization. Mol. Ecol. 8: 1187–1198.

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:54 AM Color profile: Disabled Composite Default screen

Gillespie and Boles 701

Tzvelev, N.N. 1976. Grasses of the Soviet Union [in Russian.] Tolmachev. USSR Academy of Sciences, Komarov Botanical Zlaki SSSR, Nauka, Leningrad, Russia.[Translation from Rus- Insititute [Translated from Russian by G.C.D. Griffiths. Edited by sian by Amerind Publishing Co., New Dehli, 1983]. J.G. Packer. University of Alberta Press, Edmonton, 1995]. Tsvelev, N.N. 1964. Poa L.—bluegrass. In Flora of the Russian Arc- tic. Arkticheskaya flora SSSR, Vol. 2 (Gramineae). Edited by A.I.

J:\cjb\cjb79\cjb-06\B01-036.vp Thursday, May 31, 2001 8:19:54 AM