Recognition of Pitcairnia corcovadensis (Bromeliaceae) at the Species Level
by Tania Wendt, Maria Bernadete Ferreira Canela, Jane Elizabeth Morrey-Jones, and Anaize Borges Henriques in Systemic Botany 25(3); 389-98. 2000
ABSTRACT. Pitcairnia flammea var. corcovadensis was formed by the reduction of P. corcovadensis. The re-¬establishment of the latter taxon is based on two points: (1) P. corcovadensis occurs sympatrically with P flammea in many locations, and (2) a morphometric analysis on a natural population involving these two taxa showed that they were distinct and no intermediate form was found despite their overlapping flowering seasons. Herbarium collections were examined to verify that this distinction is consistent throughout their geographic distribution. Despite the apparent reproductive isolation, artificial hybrids can be produced between P. cor¬covadensis, and P. flammea. Five morphological traits were recorded for 221 six-month old seedlings from artificial hybrids and parental taxa that were grown under the same controlled conditions. Analysis of var¬iance and discriminant analysis on seedlings showed the differences to be genetically based. A brief review of the synonyms related to P. flammea complex concluded that P. lancifolia var. lancifolia, P lancifolia var. minor, and P pumila are synonyms of P. corcovadensis. A re-description, an epitype establishment, geographic distri¬bution, description and list of material examined are provided for P. corcovadensis.
Pitcairnia L'Heritier (260 spp) is the largest genus in the subfamily Pitcaimioideae of the Bromeli¬aceae (Smith and Downs 1974). In Brazil, Pitcairnia (38 spp) form three distinct geographic groups (Amazon region, central Brazilian, Atlantic forest), and most of these species have a limited distribu¬tion (Wendt and Chamas 1997). Pitcairnia flammea Lind. is one the most morphologically variable Bra-zilian Pitcairnia species and is widespread along the Atlantic Forest. Smith (1943) reduced several spe¬cies of Pitcairnia to varieties of P. flammea, taxa that he considered to be inconsistent as herbarium col¬lections increased. Then he delimited five varieties in addition to the typical variety. One of these was formed by the reduction of P. corcovadensis Wawra. The keys proposed for varietal separation used terms such as "mostly" and "usually", and the measurement overlapped (Smith 1943; Smith and Downs 1974), reflecting the difficulties in delimit¬ing varieties. Two more varieties were described later (Smith and Reitz 1967; Pereira 1977) increas¬ing the taxonomic problems in this complex. Wendt (1994) revised the Pitcairnia species of Rio de Janei¬ro state and proposed five new synonyms to P flam¬mea, which easily fit in the wide morphological variation of this species. Wendt (1994) treated P. flam¬mea only in the broad sense due to the difficulties in recognizing some of the varieties, and pointed out the necessity for additional studies on this sub¬ject.
In some species, particularly those in which com¬plex patterns of variation are found, there is a gen¬uine utility in circumscribing infraspecific taxa (Stuessy 1989). Hamilton and Reichard (1992) pointed out that the vast majority of plant taxono¬mists perceive a need for infraspecific classification, although many authors abandon previously recog¬nized infraspecific taxa in order to increase clarity and simplicity of classification. Geographic distri¬bution is an important component in the recogni¬tion of infraspecific taxa. If the ranges of morpho¬logically distinct population systems are completely overlapping, they are probably reproductively iso¬lated, and hence, best viewed as good species. Sub¬species and varieties are worth considering when the distinctions are largely allopatric (Stuessy 1989). When Smith (1943) reduced P. corcovadensis to P. flammea var. corcovadensis, he apparently did not know that these taxa occur sympatrically in many localities throughout the range of P. corcovadensis. At Corcovado Mountain, the type locality of P. corco¬vadensis, they grow close together sometimes in mixed clumps. A study of reproductive biology conducted on these populations (T. Wendt, unpubl. data), showed that despite their very similar flower form and color and their overlapping flowering sea¬sons, they do not share the same pollinators. No intermediate forms between them were observed. Despite this ethological isolation, hybrids can be produced by hand-pollination. This illustrates the point that related species may be generally inter¬fertile, and their isolation in nature may depend largely on ecological and other external factors (Grant 1981).
The present work was undertaken to clarify the distinctiness of P. flammea var. corcovadensis within the P. flammea complex. The investigation involved: (1) a numerical analysis based on field studies of morphological characters on a mixed population of these taxa; (2) an examination of herbarium collec-tions to verify the consistency of morphological dif¬ferences along their geographic distribution; and (3) a morphometric analyses of seedlings of parents of the two taxa and their artificial hybrids. In the end, we decided to re-establish P. corcovadensis.
MATERIAL AND METHODS
Distinctions between P corcovadensis and P flammea.
We studied mixed populations of P. cor¬covadensis and P. flammea that grow side by side on a slope near the road Estrada do Redentor at Cor¬covado Mountain (22°57'S, 43°13'W). This is an tropical rain forest surrounded by urban develop¬ment and an attractive tourist site within the city of Rio de Janeiro. Trees dominate the area, but granite outcrops are common and colonized by sax¬icolous herbs, of which the bromeliads studied are important physiognomic elements. Forty-three in¬dividuals were randomly selected in different patches at the study site. Twenty-four belonged to P. corcovadensis and nineteen to P. flammea. Nine morphometric variables were chosen: corolla length, corolla width, pedicel length, bract length, pistil length, distance of stigma from anthers, scape length, number of flowers per inflorescence, and number of opened flowers per day. The corolla width was measured at the widest point of the dis¬tal end, where the petals are most divergent. The distance from stigma to anthers indicates how much the stigma exceeds the anthers. These mea¬surements were taken in the field in April 1996, when the two taxa were blooming together. Analysis of variance (ANOVA) was performed for each trait to detect differences between the taxa.
To estimate the consistency of differences throughout the geographic distribution of the species, we examined herbarium collections. Some of the traits are not preserved on dried material, thus we compared only the pedicel length, bract length, scape length, and number of flower per inflores¬cence. Approximately 120 herbarium specimens were examined from RB, R, HB, and GUA (Holm¬gren et al. 1990). Pitcairnia corcovadensis is known from three neighboring municipalities in the state of Rio de Janeiro, and P. flammea is widespread . throughout the Atlantic forest (Fig. 1). All speci¬mens collected of P. corcovadensis co-occurred with P flammea (Fig. 1, Table 1).
For the re-establishment P. corcovadensis, we pre¬pared a taxonomic treatment, which is given follow-ing the Discussion. No new description was pro¬posed for P. flammea, because of its large number of varieties not all of which have been re-studied.
Morphometric Analyses of Seedlings of P corco¬vadensis, P flammea, and their Artificial Hy-brids. Despite the absence of hybrid formation be¬tween P. corcovadensis and P. flammea in natural con¬ditions, they can be produced artificially by hand¬pollination (T. Wendt, unpubl. data). Ongoing reproductive research by the first author supplied the seeds to produce the seedlings that were used for the morphometric analyses presented here. Seedling from three types of crosses were com¬pared:
1. Seedlings grown from seeds from a single fruit formed by outcrossing hand-pollination on P. corcovadensis;
2. Seedlings grown from seeds from two fruits formed by crosses between P. corcovadensis (ovule donor) and P. flammea (pollen donor), and be¬tween P. flammea (ovule donor) and P. corcovaden¬sis (pollen donor);
3. Seedlings grown from seeds from a single fruit formed by outcrossing hand-pollination on P. flammea.
Fully mature seeds were harvested in June 1997 from partially opened fruit capsules of P. corcova-densis and P. flammea in a greenhouse. The seeds were wrapped in soft paper and stored in bottles kept at room temperature. Before germination tri¬als, seeds were removed from storage and washed initially in a solution with neutral detergent. They were disinfected by washing with 70% ethanol for 3 min followed by 0.6% sodium hypochlorite for 5 min in a polypropylene funnel (60 ml) with stainless steel mesh (160 µm pitch) at the end. The seeds were rinsed in a continuous flow of sterile distilled water for at least 3 min before transfer to the culture medium. The seeds were germinated in Petri dishes containing 40 ml of hormone-free medium with 1.3 µM tiamine, 30 µM piridoxine and 6 µM myo-inositol (Murashige and Skoog 1962). The pH was adjusted to 5.8 with KOH and the medium solidified with 0.9% agar before au¬toclaving at 101 lb, 121°C for 15 min. Cultures were maintained at 30°C under a 16 h photoperi¬od, with a light flux of 25 µmoles m-² s-¹. After two months, the seedlings were transplanted to bottles (12-24 seedlings per bottles) with the same culture medium, and kept at the same tempera¬ture and light condition. After four months the seedlings were harvested and the following traits were recorded: number of leaves, maximum width of the largest leaf, maximum length of the largest leaf, and dry weight. A total of 221 specimens rep¬resenting the three groups of seedling were mea¬sured. An analysis of variance (ANOVA) was per¬formed for each trait in each group of seedlings. Where the results of ANOVA were significant at P<0.05, a multiple comparison test (Tukey HSD) was undertaken to detect differences among groups of seedlings. Stepwise Discriminant Anal¬ysis was also performed to examine multivariate morphometric differences among the seedlings of P. corcovadensis, P. flammea, and hybrids. All statis¬tical tests were done following STATISTICA 4.2 procedures of Statsoft, Inc. (1993).
RESULTS
Individuals of P. corcovadensis and P. flammea growing in mixed clumps at Corcovado (Fig. 2A, B), were statistically different for seven of the nine morphological traits (corolla width, bract length, pistil length distance stigma-anther, scape length, number of flowers per inflorescence, and number of opened flowers per day). Only two traits (corolla and pedicel length) were not significantly different (Table 2). The morphological traits that distinguish P. corcovadensis from P. flammea (Fig. 3) in natural conditions at Corcovado Mountain were consistent with distinctions found in herbarium collections. No intermediate forms between these morphologi¬cal traits were found, although all material of P. cor-covadensis collected co-occurred with P. flammea, and in many localities the two species even grow side by side (Fig. 1; Table 1). The sole exception was P. flammea var. macropoda L. B.Sm. & Reitz which shows scape and floral bract length similar to P. corcovadensis, but differs from the latter in having longer pedicels (2.5-3.5 cm) and more than 12 flow¬ers per inflorescence. The former taxon does not co-occur with P. corcovadensis, and there were few spec¬imens of it in herbaria.
Morphological analysis of the seedlings (Fig. 2C) showed that mean values differed significantly among seedling groups for all the morphological attributes (Table 3). In three of the four traits, the mean values of P. corcovadensis and P. flammea dif¬fered significantly. The hybrid seedlings differed from their parental counterparts (mother or father) in at least two morphological traits (Table 3).
Scatter plots of the discriminant function results are presented in Fig. 4. In the seedling analysis, root 1 separates hybrid seedlings to some extent from parental seedlings, but the isolation is incom¬plete. Based upon discriminate function 1 and 2, P. corcovadensis is somewhat separated from both P. flammea and hybrid, but segregation is not com¬plete. The standardized coefficients for the canoni¬cal variable indicate that leaf length and leaf width contribute most to discriminate function 1. The dis¬criminant function 1 and 2 account for 100% of the variability. Even thought the analysis of seedling does not produce a scatterplot of completely isolat ed clusters, the classification results show an 87% correct identification. The discriminant function analysis, not including the hybrid data showed an increase of correct identification from 80 to 93% in P. corcovadensis, and from 77 to 85% in P. flammea (Table 4).
DISCUSSION
The morphological traits identified as distinctive to P. corcovadensis or P. flammea in a mixed popula-tion in the field were consistent in our examination of herbarium material. The adult specimens depos¬ited in herbaria are easily recognized without am¬biguity and no intermediate forms were observed. Discriminant function analysis based only upon vegetative data on young seedlings generated loose patterns of discrimination with overlap of individ¬uals among the taxa. However, some degree of seg-regation was observed with a high percentage of "grouped" cases correctly classified (95% to hybrid, 80% to P. corcovadensis, 77% to P. flammea; without hybrids, 93% to P. corcovadensis, and 85% to P. flam¬mea). Because these differences were shown by seedlings grown under the same conditions, we suggest that they are genetically determined and further support our recognition of P. corcovadensis at the species level.
There are many different species concepts. Be¬hind these definitions, there are correspondingly different goals (Templeton 1989). Thus, no single definition will be appropriate for all organisms (Quicke 1996). The "taxonomic species" is based on morphological likeness, being an assemblage of morphologically similar individuals that differs from other such assemblages. The criterion of mor¬phological distinction has the disadvantage that the amount of difference worthy of species rank cannot be prescribed objectively (Grant 1981). Despite the apparent limitations of the traditional taxonomic species concept, it remains the primary basis for circumscribing nearly all plant species. There is no way to decide objectively if a given taxon is a spe¬cies or a variety; however, it is widely advocated that the variety should be used for recognizing geo¬graphic variation of ordinary species, thus allopat¬ric distribution of different varieties is to be ex¬pected (Stuessey 1989). As the geographic distri¬bution of P. corcovadensis is completely contained within the range of P. flammea, and at many sites the plants grow side by side, we believe P. corcova¬densis is worthy of species-level recognition.
The most widely used framework for defining species is the biological species concept (Mayr 1942). It claims that a species is a group of inter¬breeding populations that are genetically isolated from other groups by reproductive isolating mech¬anisms. There are many difficulties in the applica¬tion of this concept (Cracraft 1989). Only rarely does the plant taxonomist have data on reproduc¬tive isolation and gene flow among populations. A strict application of the biological species negate the possibility of hybridization between species. Inter¬specific hybridization is widespread in plants, but logically cannot be rampant throughout the range, otherwise the distinctiveness of the taxa involved would dissolve to nothing. Evolutionary biologists have generally applied the term natural hybridiza¬tion to those cases involving crosses between indi¬viduals belonging to different species (Rieseberg 1997). Many experimental crosses involving different species result in viable hybrids (Stort 1984; Ba¬shaw, Hussey and Hignight 1992; Sorensson and Brewbaker 1994; Hodges, Burke and Arnold 1996). Thus, individuals from divergent lineages could form viable, fertile hybrids, and this would not rep¬resent a violation of biological species concept (Ar¬nold 1997). Pitcairnia corcovadensis and P. flammea are sympatric and, although they can hybridize by hand-pollination, no natural hybrids were found. This indicates that in all likelihood some barriers to hybridization exist. The artificial hybrids do not in¬validate our conclusion since, as Grant (1981) point¬ed out, breeding barriers need not be 100% effective and may often be indirect in the sense of being manifest in ecological or ethological isolation. Therefore, considering both the taxonomic and the biological species concept, we re-establish P. corcovadensis as a species.
Endemic to Rio de Janeiro State, growing on wet granite outcrops or on rocks near streams in tropical rain forest. Among species of Pitcairnia that occur in Rio de Janeiro State, P. corcovadensis is similar to P. flammea and P. carinata Mez in general appearance of flower shape and color. Pitcairnia flammea is highly variable in size showing plants with scape lengths 40-110 cm; in contrast, P. corcovadensis is less variable in size with scape lengths between 17-45 cm. Pitcairnia corcovadensis has few flowers per inflorescence (2-12) and flowers bracts (0.5-1.5 cm) not exceeding the pedicels length, while P. flammea has many flowers per inflorescence (30-120) and flowers bracts (1.5-4.5 cm) normally exceeding the pedicel length, except for P. flammea var. macropoda, which has long pedicels of 2.5-3.5 cm. Pitcairnia carinata can be as small as P. corcovadensis, but differ from it in having alate-carinate sepals.
The plate by Wawra (pl. 27, 1866), that typified P. corcovadensis, shows a plant with the lower floral bracts longer than the pedicels. However, other authors that described this taxon (Mez 1891-94, 1896, 1934-35; Baker 1889) described floral bracts shorter than the pedicels. This discrepancy was also noted by Smith (1943). Pitcairnia corcovadensis plants are usually shorter and have fewer flowers than what is presented on Wawra's plate. Because of the ambiguity, we indicate here an epitype based on material collected from the type locality that better exemplify the morphology of this taxon. —SeeSyst. Bot.
