Introduction

The genus Litsea Lam. includes approximately 400 species of dioecious trees and shrubs that mainly inhabit the tropical and subtropical regions of Asia, the Malayan archipelago, Australia, New Zealand, and other islands in the Pacific basin. A few species are distributed in continental America from the east coast of the United States, most of the Mexican territory (except for the California and Yucatan peninsulas) and in mountainous areas of Central America from Guatemala to Costa Rica (Mez 1889; Allen 1945; Rohwer 1993). Thus, the genus presents the classic pattern of intercontinental disjunction between the eastern areas of North America and Asia (Boufford and Spongberg 1983; Tiffney 1985a, b; Parks and Wendel 1990).

The American species of Litsea, except for L. aestivalis (L.) Fernald, grow in pine–oak forest, cloud forest, and rarely in the boundaries of tropical dry forest, between 1,000 and 3,000 m of altitude. Litsea aestivalis inhabits coastal areas of Alabama, North Carolina, South Carolina, Florida, Georgia, and Virginia, from 10 to 200 m of altitude (Van der Werff 1997).

Some efforts have been made to clarify the taxonomic history and phylogenetic relationships of Litsea (Hyland 1989; Li et al. 2004; Fijridiyanto and Murakami 2009), however, these issues are still confusing and controversial. Even in the small American group, taxonomic treatments over the years show strong discrepancies in the recognition of species and subspecies.

Since the revision of the Lauraceae by Meisner (1864), who placed the American species of Litsea in Tetranthera, and recognized five species, namely T. californica Hook. et Arn. (currently Umbellularia californica (Hook. et Arn.) Nutt.), T. geniculata (Walter) Nees, T. glaucescens (Kunth) Spreng. with four varieties (T. glaucescens var. subsolitaria Meisn. (hereafter cited as var. glaucescens, in conformity with nomenclatural rules), T. glaucescens var. subcorymbosa Meisn., T. glaucescens var. racemosa Meisn., and T. glaucescens var. major Meisn.), T. neesiana Schauer with two varieties (T. neesiana var. corymbifera Meisn. and T. neesiana var. villosa (M. Martens et Galeotti) Meisn.), and T. berteroi Spreng. (apparently T. laurifolia Jacq. (= Litsea chinensis Lam.), an introduced species from Asia (Mez 1892)), this group has undergone numerous changes and additions.

Mez (1889) distinguished, now in Litsea, six species for America; he retained L. geniculata (Walter) G. Nicholson (= L. aestivalis), L. glaucescens Kunth, and L. neesiana (Schauer) Hemsl., and added L. orizabae Mez, L. parvifolia Mez, and L. guatemalensis Mez as new species. Later, Bartlett (1909) recognized 11 species. To the six already recognized by Mez, he added L. flavescens Bartlett, L. novoleontis Bartlett, L. pedicellata Bartlett, L. pringlei Bartlett, and L. schaffneri Bartlett. The most recent treatment is that by Allen (1945), who accepted four species for Mexico and Central America: L. muelleri Rehder, L. pringlei (including L. novoleontis), L. parvifolia (including L. pedicellata), and L. glaucescens with three varieties—L. glaucescens var. glaucescens, L. glaucescens var. schaffneri (Bartlett) C. K. Allen, and L. glaucescens var. flavescens (Bartlett) C. K. Allen. She subsumed L. neesiana, L. guatemalensis, and L. orizabae as synonyms of L. glaucescens var. glaucescens.

Regardless of the last revision of the group, there is not yet any consensus on the number of Litsea species present in America. Local botanists still apply names to specimens of Litsea following Mez or Bartlett instead of Allen, showing they recognize more species than Allen did. In practice, three species complexes can be recognized: the L. glaucescens complex (L. glaucescens + L. guatemalensis + L. flavescens + L. neesiana + L. orizabae + L. schaffneri), the L. parvifolia complex (L. pedicellata + L. parvifolia), and the L. pringlei complex (L. pringlei + L. novoleontis). In addition, Litsea muelleri seems to be a species clearly different from the others.

Particularly complicated has been the distinction of species within the L. glaucescens complex, asserted to be the most widely distributed group of species in the continent (from northern Mexico to Costa Rica). Characters that have been used in several treatments to distinguish species in this group, mainly the form and size of the leaves, and the density of indument on different structures, are very variable. So, while some authors (Meisner 1864; Mez 1889; Bartlett 1909) recognize more than one species in the group, others consider it just as a highly polymorphic species (Allen 1945; van der Werff and Lorea 1997). A similar situation exists in the group of species of northwestern Mexico, where the size of the leaves, rather than the form, has been critical for the recognition of species; such is the case for L. pringlei, L. novoleontis, L. parvifolia, and L. schaffneri.

Although high morphological variability has been recognized as the main problem when discerning the number of taxa (Allen 1945, pages 407–409), it has not been examined rigorously. Recognition of species within the group is currently truly arbitrary. Thus, our focus in this study was to resolve the number, names, and distinguishing features of the American species of Litsea, on the basis of strict morphological analyses.

Materials and methods

A collection of 478 specimens borrowed from several herbaria (see Acknowledgments) was examined in this study. Studied material is listed in Appendix 1. This material covered almost the entire geographic range of the group (no samples from El Salvador and Honduras were available). A set of 36 morphological characters (vegetative and reproductive) was evaluated, including those traditionally used for the recognition of species. Further, a survey of leaf cuticle was carried out in order to consider features previously not assessed in the recognition of species in the group (particular results of this analysis will be published elsewhere). Thus, three characters of leaf epidermis were added to the study, making a total of 39 characters (Table 1).

Table 1 Morphological and anatomical characters of Litsea considered in the population aggregation analysis
Full size table

Values in the counts of secondary veins, flowers, stamens and staminodia are the average of at least three measurements (depending on available material). Percentage of indument density was calculated as the amount of foliar surface covered with indument in a given radial section of the field of the microscope. The continuous variables were measured with callipers or a ruler. Cuticle impressions were made with silicone fluid, following the technique described by Sandoval (2005).

Two different types of analysis were used to resolve the delimitation of American species in Litsea. A first approach was achieved applying a Population aggregation analysis (Davis and Nixon 1992). Then, with a different perspective, two multivariate analyses were performed. These had the purpose of assessing the correspondence of taxonomic schemes suggested by Mez (1889), Bartlett (1909), and Allen (1945), with the grouping produced by a phenetic clustering method. In addition, a taxonomic proposal of our own developed during the morphological review of specimens (which considers L. glaucescens, L. guatemalensis, L. muelleri, L. neesiana, L. orizabae, L. pringlei, L. parvifolia, and L. schaffneri as distinct species) was also evaluated.

Population aggregation analysis (PAA)

This method implies comparisons of the distribution of character states between populations and involves the search for fixed attributes that eventually, based on the differences among them, indicate the different species. To do so, 45 populations were defined across the geographic distribution of the genus using the boundaries of major river basins, plus other topographic features (like mountain ranges within basins) that may act as natural limits for distribution (Fig. 1). Profiles of character states for each population were constructed and were compared with each other (Table 2). In this case the whole set of 39 characters was used.

Fig. 1
figure 1

Map showing location of 45 populations of Litsea studied with PAA. The numbers correspond to the populations indicated in Appendix 1

Full size image
Table 2 Populations character profiles in Litsea
Full size table

Multivariate analyses

Principal components analysis (PCA)

This part of the study is based on a subset of the specimens examined in the previous analysis, which were chosen for being fertile and representative of the whole morphology to be compared (Appendix 1). Three matrices of data were constructed, because exclusive characters for each group were measured; one containing 85 flowering male specimens and 33 variables, a second matrix with 55 flowering female samples and 27 variables, and a third with 87 specimens with fruits and 28 variables (Table 3; data matrices are available from the corresponding author). The specimens examined represent all the entities proposed in the different taxonomic treatments (except for L. aestivalis), and covered almost all the geographical distribution registered for the group. The analysis was executed by use of the software Statistica v. 6.0 (StatSoft Inc. 1998) for each of the matrices. Because the variables were of different types, the data were transformed by columns to log10 (length measures), arc sine (ratio measures), or square root (counts), and the analyses were based on matrices of correlations.

Table 3 Morphological characters of Litsea evaluated in principal components analysis
Full size table

Non-metric multidimensional scaling (NMDS)

A basic matrix of 90 specimens and nine qualitative characters (Table 4; the matrix is available from the corresponding author) was constructed for this analysis. Later a similarity matrix was calculated using percentage disagreement as measurement of distance, because this is recommended for categorical variables. The analysis was executed by use of the software Statistica v. 6.0 (StatSoft Inc. 1998).

Table 4 Morphological characters of Litsea used for the non-metric multidimensional scaling
Full size table

The analyses were executed making no assumption about the identity of specimens. Later, on the resulting graphs, putative names according to the different taxonomic schemes were added for each specimen.

The differences among the resulting groups were subsequently tested for significance with ANOSIM (an analysis of similarity) using Primer v 5. The R statistic generated by ANOSIM is a relative measurement of separation of the a priori-defined groups and ranges from −1 to +1. A value of −1 indicates that all the samples within the groups are less similar to one another than to any other sample from different groups; zero (0) indicates that there are no differences between the groups; and a value of +1 indicates that all the samples within each group are more similar among themselves than to any other sample of a different group (Clarke and Gorley 2001). ANOSIM test of groups derived from PCA was based on the normalized euclidean distance matrix, whereas for groups obtained by NMDS it was based on the original distance matrix.

Results

Population aggregation analysis

Population profiles show that six of the characters considered are fixed—the deciduous character and membranous consistency of the leaves, the absence of a mucro, indument on the midvein only towards the base on the lower surface, indument on the proximal section of the petiole, and the difference in size of the epidermal cells between the upper and lower surfaces (Table 2). These attributes clearly separate L. aestivalis from the other species.

For the other records, the large number of polymorphisms in all populations and all characters is notorious. Characters such as the form of the lamina (2), the shape of the foliar base (3) and apex (4), indument type (9, 10) and color (12), and type of inflorescence (21), that have traditionally been used to distinguish species of Litsea in America are not fixed and were found to be very variable within populations. This situation made it impossible to group populations. Only populations 20 and 23, which correspond to the center of Hidalgo and some localities of Guerrero, State of Mexico and Morelos, were identical. No other group was found with this analysis.

Multivariate analyses

Principal components analysis

In this case the three matrices analyzed yielded similar results (Table 5), i.e. the first three components represented 50–54% of the variance. The first component explained between 21 and almost 24% in all cases, with a large to moderate contribution from characters concerning the length and density of trichomes. In particular, the length of the trichomes on the midvein on the lower surface of the leaves was the variable with the largest value in the three analyses. The second component explains 15–19% of the variance and encompasses large values for characters related to the size of leaves and petioles. Finally, the third component explained between 11 and 13% of total variance, where mainly reproductive characters contribute the most (Table 6).

Table 5 Eigenvalues of the first three components from principal components analyses for the three matrices analyzed in American Litsea
Full size table
Table 6 Results from principal components analysis
Full size table

Specimens identified with the same species name are hardly visualized as different groups in the corresponding PCA graphs for all the taxonomical schemes (Figs. 2, 3). However, the ANOSIM permutation test shows that when the groups obtained are evaluated for distinctiveness, those corresponding to the classification hypothesis advanced in this study are supported with the highest value of R for the three matrices analyzed. In contrast, Allen’s scheme is qualified with the lowest R figures (Table 7).

Fig. 2
figure 2

First two principal components for 27 quantitative variables of 55 female specimens of Litsea with mature flowers grouped under the Allen (1945) classification scheme. Open circles, L. pringlei; filled squares, L. parvifolia; open diamonds, L. muelleri; closed diamonds, L. glaucescens var. glaucescens; asterisks, L. glaucescens var. schaffneri

Full size image
Fig. 3
figure 3

First two principal components for 33 quantitative variables of 85 male specimens of Litsea with mature flowers grouped under the classification proposal advanced in this study. Open circles, L. muelleri; filled triangles, L. pringlei; filled squares, L. glaucescens; filled diamonds, L. guatemalensis; asterisks, L. schaffneri; open diamonds, L. neesiana; open squares, L. orizabae; filled circles, L. parvifolia

Full size image
Table 7 Results of the ANOSIM test for each classification scheme evaluated by PCA, showing global R values for each of the three matrices and the global significance level
Full size table

Non-metric multidimensional scaling

Global values of R and p derived from ANOSIM for the NMDS analyses again indicated that the classification hypothesis advanced here is the one with the greatest support (Table 8). The resulting diagram of the arrangement of the NMDS analysis for this hypothesis is given in Fig. 4, which with a stress value of 0.122, represents a good adjustment. In this case the ANOSIM permutation test significantly separated every pair of groups at the level p < 0.01 (Table 9), except for L. neesiana and L. orizabae, for which no significant differences were observed (R = −0.009, p < 0.426). The principal coordinates derived from this analysis are shown in Appendix 2.

Table 8 Results of the ANOSIM test for each classification scheme evaluated by NMDS, showing global R and p values
Full size table
Fig. 4
figure 4

Graph from NMDS analysis for nine qualitative variables of 90 specimens of Litsea grouped under the scheme advanced in this study. Open circles, L. muelleri; filled triangles, L. pringlei; filled squares, L. glaucescens; filled diamonds, L. guatemalensis; asterisks, L. schaffneri; open squares, L. neesiana and L. orizabae, filled circles L. parvifolia

Full size image
Table 9 Results of the ANOSIM test showing all pair differences between the eight Litsea species, for the classification scheme advanced here evaluated by NMDS
Full size table

Discussion

The results of the PAA revealed unequivocally that L. aestivalis is the most distinct species among the American group of Litsea. This result was expected because this species is ecologically and geographically isolated, growing in coastal environments at an altitude of 10–200 m (Van der Werff 1997), unlike the other species, which are distributed mainly in pine and oak forest at an altitude of 900–3,000 m. The populations of the remaining species are mostly sympatric and this condition diminishes the power of the PAA to distinguish, in this case, the other taxa.

As the ANOSIM test results show, the classification scheme advanced here is the best for describing the number of species recognized as Litsea in Mesoamerica. We consider that lower values of R obtained in the ANOSIM test for the other classification schemes evaluated mean that combinations of characters used for circumscription of the corresponding species do not categorize adequately the morphological variation observed in the group. So, Bartlett judged that some expression of character variation was of major importance in the identification of species and recognized twice as many as Mez did. Allen, on the other side, was right in pointing out, on the basis of the geography of character variation, that some of the species described by Bartlett were spurious. She, however, failed to distinguish other species and subsumed them in L. glaucescens, creating a very loose concept for this taxon.

The classification scheme advanced in this study suggested the recognition of eight species. Ordination of character variability through PCA and NMDS analyses, however, result in distinction of only seven additional species. There was no support to maintain L. neesiana distinct from L. orizabae. The accepted species for Mesoamerica are L. glaucescens, L. guatemalensis, L. muelleri, L. neesiana, L. parvifolia, L. pringlei, and L. schaffneri, which are discussed next.

Litsea muelleri, is a homogeneous species of restricted distribution, mainly from the area of the Cerro Potosí in Galeana, Nuevo León, in northeastern Mexico. In this study, nine specimens were considered, the majority from the locality mentioned, two from Hidalgo, and one from a small population within the Biosphere Reserve El Cielo in Tamaulipas. Litsea muelleri occurs close to L. guatemalensis and L. neesiana, because of its dense indumentum of long trichomes (Fig. 5c, d) but, unlike them, it always presents leaves almost round with cordate bases that do not exceed 5 cm in length (Fig. 6a).

Fig. 5
figure 5

Lower leaf indument. a and b Straight and long trichomes, L. guatemalensis (N. Jiménez P. et al. 1496); c and d long and curly trichomes, L. neesiana (N. Jiménez P. et al. 1504), and e and f erect and short trichomes, L. parvifolia (N. Jiménez P. et al. 1512)

Full size image
Fig. 6
figure 6

Leaf shape of a L. muelleri, b L. parvifolia, c L. pringlei, d L. schaffneri, e L. glaucescens, f L. guatemalensis, and g L. neesiana

Full size image

Litsea pringlei and L. parvifolia are clearly differentiated entities (Fig. 4, Table 9). They share coriaceous leaves with cordate, subcordate, obtuse, or rounded leaf bases (Fig. 6b, c), and their epidermal cells have cleft anticlinal walls (Fig. 7b). However, leaf blades in L. pringlei are always glabrous whereas in L. parvifolia they consistently have short erect, dispersed trichomes underneath (Fig. 5e, f). These two species are restricted to the Mexican states of Coahuila, Nuevo León, and Tamaulipas. Bartlett (1909) recognized L. pringlei and L. parvifolia as different species, in addition to L. novoleontis and L. pedicellata, all sharing the distinctive character of subcordate to rounded leaf base. He considered that L. novoleontis differed from L. pringlei, claiming the former presented a solitary, fasciculate inflorescence whereas the latter had terminal and axillary corymbs. He drew a similar conclusion for L. pedicellata and L. parvifolia; whereas L. pedicellata was distinguished for having paniculate inflorescences, L. parvifolia was characterized by bearing solitary or rarely fasciculate inflorescences. Although terminal inflorescences do not exist within the group, inflorescence structure is a very variable character both in single plants and among individuals of the same species.

Fig. 7
figure 7

Epidermal cell contour. a Polygonal cells with straight walls and sharp or rounded points, L. neesiana (N. Jiménez P. et al. 1506); b cleft cells with numerous deep, round teeth along walls, L. pringlei (N. Jiménez P. et al. 1527); c lobed cells with few shallow, round teeth along walls, L. glaucescens (N. Jiménez P. et al. 1546); d sinuous cells with wavy walls, L. schaffneri (Hinton 17495)

Full size image

Allen (1945) gave taxonomic importance to the form of the base and size of the leaves, and considered L. schaffneri a variety within L. glaucescens. The results of our study differ from Allen’s appreciation, and agree with the judgement of Bartlett in that these two taxa represent different species. L. schaffneri has narrow and coriaceous leaves (never rounded), glabrous, with acute base and epidermal cells with sinuous anticlinal walls (Figs. 6d, 7b). The species is found mainly in Hidalgo, Guanajuato, Querétaro, San Luis Potosí, and a few places in Tamaulipas.

Litsea glaucescens has been considered the most common and widely distributed Litsea species; it is also attributed with a considerable variation in vegetative characters, for example the type and density of pubescence on different parts of the plant and the form and size of leaves. Indeed, Allen (1945) considers it a too highly polymorphic species in which three varieties could be distinguished. An outcome of our study, perhaps the most unexpected, is that in contrast with the persistent concept of L. glaucescens, it was found that it forms a coherent entity differentiated from the rest of the species by the glabrous condition of its almost membranous leaves with petioles longer than 1 cm (Fig. 6e). The species exhibits a unique epidermal cell pattern with irregular lobed contours (Fig. 7c). Thus, circumscribed, the species is reduced in distribution to the environs of the Trans-Mexican volcanic belt, with some isolated locations in Querétaro and Jalisco. This is the most popularly used Litsea species (locally called “laurel”).

Litsea guatemalensis and L. neesiana are characterized by having the densest indument and the longest trichomes within the American species of Litsea. In the Mez (1889) and Bartlett (1909) classifications these taxa appear as three different species (because they considered L. orizabae different from L. neesiana), but Allen (1945) regarded them as synonyms of L. glaucescens var. glaucescens. Inspection of the individuals assembled in these species reveals they share epidermal cells with (mostly) polygonal contour (Fig. 7a). This type of epidermal cell pattern is similar to that found in L. aestivalis, and clearly distinguishes these from the other species. On the other hand, they differ in the morphological types of trichomes found in their indument—long and straight in L. guatemalensis and long and curly in L. neesiana.

Therefore, Litsea guatemalensis is here characterized by the presence of long, straight trichomes, of variable density (pubescence can be lost with age), leaves with acute to attenuate base, and long acuminate apex (Figs. 5a, b, 6f,). This species is actually, according to this study, the most widely distributed in America, found from northwest of Mexico through Central America. Litsea neesiana is distinguished by having a dense indument of long and curly trichomes (Fig. 5c, d), and mostly oblong, never long acuminate leaves (Fig. 6g). Thus circumscribed, the species would mainly be restricted to the surroundings of the volcano Pico de Orizaba, in Veracruz, Mexico, and to some isolated populations in the center of Chiapas and mountain areas of Oaxaca, Mexico. Analysis of the characters considered in this study does not support the separation of L. neesiana from L. orizabae (Table 9).

In short, the recognition of eight species of Litsea in America is supported by this work on the basis of analysis of qualitative and quantitative characters. Sympatric populations were found throughout the area of distribution, but principally in the environs of the Trans-Mexican volcanic belt and the Sierra Madre Oriental. In this respect, it is important to mention that local people use Litsea species, preferring, maintaining, and even propagating the glabrous forms, thus changing the original distribution of populations.

Finally, as a synthesis of the diagnostic characters of the species, a key for the identification of the American species of Litsea is provided.

figure a