Canadian Journal of Botany 83(4): 386-398 (2006)
Molecular markers indicate that the narrow Québec endemics Rosa rousseauiorum and Rosa williamsii are synonymous with the widespread Rosa blanda.
Anne Bruneau, Simon Joly, Julian R. Starr, and Josée-Nadia Drouin

 s.l. = sensu lato (broad sense)
 s.s. = sensu stricto (narrow sense)

Abstract: Rosa rousseauiorum Boivin and Rosa williamsii Fern. are two rare roses in eastern Québec, whose taxonomic status is controversial. Morphological characters alone do not clearly differentiate these two taxa from each other or from the morphologically variable and widespread Rosa blanda Ait. We evaluated the taxonomic status of these two taxa, and of two other R. blanda segregates, Rosa subblanda Rydb. and Rosa johannensis Fern., through an analysis of RAPD, ISSR, and AFLP markers. We surveyed 86 individuals from 36 populations in eastern North America. Despite a high degree of polymorphism, principal coordinate analyses and the weighted pair group method with arithmetic averaging suggest no clustering of individuals that correspond to taxonomic boundaries. However, the closely related Rosa palustris Marsh. is clearly differentiated from the R. blanda s.l. taxa. When populations of R. blanda west of Québec are included, the principal coordinate analyses and Mantel tests indicate the presence of a significant east-west geographic gradient. Analyses of molecular variation suggest that most of the observed variation occurs within taxa, rather than among taxa. A weak inter-taxon variation is nonetheless significant for RAPD and ISSR data, and a weak pattern dependent on geographical location is evident within the province of Québec. In accordance with studies based on morphological characters, molecular data indicate that R. rousseauiorum and R. williamsii should not be considered as species distinct from R. blanda.

p. 396

Taxonomic status of Rosa rousseauiorum and R. williamsii

Our data suggest that R. rousseauiorum and R. williamsii cannot be considered as species distinct from R. blanda s.s. Here we consider species to be ecologically, morphologically, and (or) genetically cohesive groups of populations that evolve independently from other such groups. The absence of cohesion is indicated by results of the PCoA (Fig. 2) and WPGMA (Figs. 3 and 4) analyses, as well as by the AMOVA, which suggests most of the genetic variation occurs within rather than among R. blanda segregates (Table 3). This suggests an important degree of gene flow among taxa within R. blanda s.l. in Québec. Field observations and morphological analyses further support the molecular data. Despite intensive collecting in the regions where R. rousseauiorum and R. williamsii are endemic, few specimens could be unambiguously identified as belonging to either of these two taxa, as described in the taxonomic key given by Boivin (1945).

Doubts have long persisted in the botanical community regarding the species status of R. rousseauiorum and R. williamsii, as well as that of R. johannensis and R. subblanda, the other two species that at times have been segregated from R. blanda s.l. Gleason and Cronquist (1991) recognised only R. blanda but mention R. subblanda and R. johannensis as possible varieties. In contrast, in Gray’s Manual of Botany (Fernald 1950) and in the recent edition of La Flore Laurentienne (Marie-Victorin 1995), R. blanda, R. johannensis, R. rousseauiorum, and R. williamsii all are listed as good species. Similarly, Scoggan (1978) recognised R. williamsii, R. rousseauiorum, and R. blanda, the latter with two varieties and six forms. More recently, in Flora of New Brunswick, Hinds (2000) recognises only R. blanda (with R. johannensis as a synonym) but notes that a glabrous variety (var. glabra Crépin) is frequent in the region. Lewis (1957a, 1957b) in his taxonomic revision of North American roses did not recognise any of these species nor any infraspecific taxa, except for two forms: R. blanda Ait. f. alba (Schuette ex. Erl.) Fern. for a white-petaled variant and R. blanda Ait. f. carpohispida (Schuette) Lewis for a form with glandular–hispid hypanthia and pedicels. Likewise, Breitung (1952) in a study of native roses of Canada, recognised only R. blanda, describing the species as variable in terms of leaflet pubescence.

Rosa rousseauiorum and R. williamsii are distinguished from other R. blanda s.l. taxa by the presence of a large number of glandular trichomes on the lower surface of the stipules. Boivin (1945) described R. rousseauiorum to recognise large-stature plants that possessed stipules (>3.5 cm) and sepals (>1.5 cm) longer than those of R. williamsii. Unfortunately, the crucial differentiating character of sepal length is reversed in the discussion (i.e., <1.5 cm) relative to that given in the description and key (>1.5 cm), a situation that surely has added to the confusion associated with the taxonomic limits of R. rousseauiorum since its description. Although field observations support the idea that a population of R. blanda s.l. with glandular sepals and with plants of smaller stature does indeed occur in the Bic region of the lower St. Lawrence, as first suggested by Fernald (1918), we noted a high degree of variation in these characteristics. Plants can have more or less glandular stipules, often varying within a single individual, and sepal length likewise seems to represent a continuum from less than to greater than 1.5 cm. Lewis (1957b) also noted that across the range of R. blanda s.l., sepal length varied from 1.3 to 1.9 cm (mean 1.7 cm) and from the presence of glandular (47% of individuals measured) to non-glandular stipules. Similarly, leaflet pubescence was shown to vary, occurring in only 88% of specimens studied. In addition, Erlanson (1934) showed that individuals of R. blanda with erect sepals at maturity can have progeny with erect, spreading or sometimes reflexed sepals on mature hypanthia. All of these morphological analyses bring into doubt not only the taxonomic status of R. williamsii and R. rousseauiorum, but also of R. johannensis, R. subblanda, and most of the infraspecific taxa that have been described. Although our sampling for these other taxa is limited, our survey of molecular markers in specimens from Québec strongly suggests that on both molecular and morphological grounds, R. blanda should not be subdivided into several different species.

Biogeographical and conservation implications

The restricted geographical distribution of the R. rousseauiorum (Charlevoix and Lower St. Lawrence) and R. williamsii (Lower St. Lawrence River) variants suggests the possibility of a characteristic and well-defined biogeographic pattern within Rosa blanda s.l. in Québec. Molecular markers have been used in numerous phylogeographic studies to highlight such patterns (e.g., Tremblay and Schoen 1999; Abbott et al. 2000; Hagen et al. 2001; Stehlik 2002). Although in our analyses, no clear geographic pattern emerges from the PCoA, WPGMA analyses, or Mantel statistical tests, the AMOVAs suggest some partitioning of genetic variation relative to the four regions of Québec that we defined based on our sampling regime (Table 3). This suggests constraints on gene flow among regions in Québec, which may not be linearly related with the genetic distance and may therefore explain the lack of significance of Mantel tests. The St. Lawrence River may act as a barrier to gene flow across regions, resulting in differentiation between, for example, populations of Charlevoix, where most of the R. rousseauiorum variants occur, and Bic, where most of the R. williamsii variants are found. A stronger pattern of isolation by distance is evident when western R. blanda s.s. populations are included. Western R. blanda populations integrate with R. woodsii in regions where the ranges of the two species overlap (Manitoba, Minnesota, North and South Dakota) potentially adding increased genetic variability and differences to the western R. blanda gene pool (Lewis 1962; J.R. Starr, unpublished data). These analyses also suggest more gene flow among Québec populations than between Québec and the more western populations.

Both the R. rousseauiorum and R. williamsii variants tend to occur at the edge of the sea or in marsh habitats (Boivin 1945; Fernald 1950), but the latter seems to prefer a saline habitat (Fernald 1918). This led Erlanson (1934) to suggest that the latter taxon was a calciphile ecotype of R. blanda rather than a different species. Though restricted, the distribution of the R. rousseauiorum variant, in Charlevoix, the Gaspé Peninsula, the Bic region, and sporadically in the Gatineau valley, is nonetheless more widespread than that of the R. williamsii variant, which occurs almost exclusively in the Bic region (Fig. 1). Although there may indeed be a distinct morphotype of R. blanda in the Bic region with glandular stipules, small sepals, and an overall smaller stature, which merits conservation attention, all evidence suggests that R. williamsii should not be considered as a good taxonomic species. Rosa rousseauiorum, with its even less distinct phenotype and more widespread distribution, should simply be considered a synonym of the variable R. blanda.

Botanical Gazette 96(2): 197-251 (Dec 1934)
Experimental data for a revision of the North American wild Roses.
Eileen Whitehead Erlanson

R. blanda Ait. 1789. 2n=14.
Ecotype species: R. williamsii Fernald
Syn.: acicularioides Schuette, johannensis Fernald, palustriformis Rydb., subblanda Rydb.

Fernald (1918) Rosa williamsii and R. johannensis

Boivin (1945) Rosa rousseauiorum

Fernald (1948) Rosa williamsii, R. johannensis and R. rousseauiorum