Correction: Cryptic Species Due to Hybridization: A Combined Approach to Describe a New Species (Carex: Cyperaceae)

[This corrects the article DOI: 10.1371/journal.pone.0166949.].


Introduction
Around 2000 species have been recognized in the genus Carex L. (Cyperaceae; [1][2][3]) which is one of the largest genera among the angiosperms as the result of a relatively fast radiation mainly in temperate areas of the Northern Hemisphere [4,5]. The study of the genus as a whole has derived in taxonomic rearrangements based on phylogenetic studies [2]. Incomplete phylogenies of the genus Carex (e.g. [6,7]), or focusing on infrageneric taxa (subgenera or sections) (e.g. [8][9][10]), the restricted geographic coverage of the studies, mostly focused on Europe and North America, and the historical non-natural classifications of the genus, are some of the causes that have hampered an extensive revision of the genus [11]. Moreover, hybridization in Carex has been proposed to limit taxonomic delimitation of species [12][13][14].
Using molecular tools and combinations of different approaches have been demonstrated to be crucial to detect hybrid zones [15] and discover new cryptic species [16][17][18]. Hybridization and/or introgression cause reticulate evolution [18] which makes the establishment of limits among species difficult.
Carex section Glareosae G. Don has a circumboreal distribution and constitutes a monophyletic clade comprising 25 currently recognized species [7,14,15]. Species in Carex section Glareosae have experienced multiple taxonomic rearrangements (e.g. [8,[19][20][21]) due to its remarkable morphological, biogeographic and ecological variability [19][20][21]. The taxonomic identity of some species within the section is still unclear as it occurs with C. kreczetoviczii T.V. Egor. [8,19] and C. furva Webb [22]. The taxonomy of C. furva s.l., endemic to the Iberian Peninsula (Spain and Portugal), has been discussed in the past. This species was considered either a synonym of C. lachenalii Schkuhr, a subspecies or a variety of this species (see discussion in [22]). Nowadays, it has been broadly demonstrated from a morphological, phylogenetic and cytogenetic point of view, that C. furva s.l. and C. lachenalii are different species [8,22].
Within C. furva s.l., different morphogroups were detected by Gay [23] and later by Luceño [22]: one from the southern Iberian Peninsula, and another group constituted by central and northern populations of the species in the Iberian Peninsula. While Gay [23] considered central and northern morphotype of C. furva s.l. to be a subspecies of C. lachenalii [23], Luceño [22] did not. Intermediate morphologies in the southern Iberian Peninsula as well as the continuum of variation of diagnostic characters prevented him the consideration of this taxon. Nowadays, both morphotypes are considered a single species, C. furva [3,22]. Maguilla et al. [8] suggested the existence of an incipient speciation event involving these populations. Morphological data supports the presence of intermediate morphologies in the southern Iberian Peninsula [22] which could reflect hybridization processes between different taxa. Hybridization can act as a homogenization force of both genetic and morphological traits among species [24,25], and consequently C. furva s.l. could be hiding a cryptic taxon. Hybrids in Carex section Glareosae have been described to be mostly sterile. In fact, hybrid speciation seems not to be a major evolutionary force for Carex in general [12][13][14] or specifically for this section [8,19,20,26,27]. Previous studies by Maguilla et al. [8] and Luceño [22] suggest that hybridization could be avoiding the detection of a cryptic species within C. furva s.l.
The aim of this study is to delimitate the morphological variability of both previously detected genetic entities within C. furva s.l. and to decide whether to consider a new species, subspecies or variety. We have performed a combined approach based on statistical analyses of morphological and phylogenetic data to discriminate taxa that could have remained cryptic due to the existence of morphologically intermediate individuals.

Study species
Carex furva s.l. is a species endemic to the Iberian Peninsula that belongs to Carex section Glareosae. Previous phylogenetic studies have shown this species to be monophyletic [8]. The highly specific ecological requirements of soils on acid bedrocks and very cool environments [28] explain the distribution of the species in the highest mountains of the Iberian Peninsula, never occurring below 1800 m.a.s.l. [28]. Currently, C. furva s.l. has been found only in seven mountain ranges in the Iberian Peninsula (Sierra Segundera, Sierra del Cornón, Fuentes Carrionas, Sierra de Gredos, Sierra de Guadarrama and Sierra Nevada in Spain, plus Serra da Estrela in Portugal; [28,29]). Our sampling exhaustively covered the full range of the species (Fig 1). Field collecting permits were provided by Instituto de Conservação da Naturaleza e da Biodiversidade (ICNB, Portugal), Junta de Andalucía (Department of Environment, Spain) and Community of Madrid (Department of Environment, Local Government and Territorial Planning, Spain). Destructive sampling for DNA extraction was provided by UPOS herbarium.

Morphometric study
Sampling for morphological study included 43 herbarium specimens (M, SEV and UPOS herbaria [30]) and samples of inflorescences of 60 specimens collected in field trips and preserved Codes indicate sampled populations as follows: C1 = Serra da Estrela; C2 = Sierra de Béjar; C3 = Sierra del Barco; C4 = Picos de Gredos; C5 = Sierra de Guadarrama; N1 = Sierra Segundera; N2 = Sierra del Cornón; N3 = Fuentes Carrionas (Curavacas); S1 (C. furva s.s.) and S2 (hybrid) = Sierra Nevada. Created using country borders from Brummitt et al. [62] and elevation data from CGIAR [63]  in silica-gel. Thus, a total of 103 individual specimens were measured including six to 21 individuals per population (ten sampled populations) to reflect the full range of morphological variation in the species. The exceptions were populations in Fuentes Carrionas (N3; Fig 1) where only one specimen was available at UPOS herbarium, and population from Serra da Estrela (C1; Fig 1), with only one surviving individual [29]. Type specimen of C. furva s.s. [31] was visually inspected but not included in any of the analyses. Twenty-seven morphological variables were selected and measured based on characters used for the description of species in Carex sect. Glareosae in different flora ( [19][20][21][22]; Table 1 and Fig 2). Only one measurement per variable and specimen was taken (avoiding redundancy) from each specimen. We randomly selected a mature shoot per specimen. However, minimum and maximum values for each variable and individual were obtained measuring all mature shoots of each specimen. An ocular micrometer was used for characters shorter than 10 mm, and a 30-cm ruler when larger than 10 mm. Angles were measured with a standard angular encoder. Moreover, three new variables were calculated to represent the shape of the inflorescence and utricles: the ratio inflorescence length: inflorescence width (INFL/INFW), the ratio utricle length: utricle width (PERL/PERW) and the ratio utricle length: distance from the utricle base to its maximum width (PERL/PERMWD; Table 1). Minimum and maximum ranges of culm, leaf, inflorescence and spike lengths and widths were also measured. For the statistical analyses, culm length (CLML), ligule length (LIGL) and leaf characters (ILEAFL, ILEAFW, SLEAFL and SLEAFW; Table 1 and Fig 2) were excluded due to the lack of data in most of the samples from silica-preserved specimens. To avoid redundancy in statistical analyses, ratios (INFL/INFW, PERL/PERW and PERL/PERMWD; Table 1) were used only for species descriptions and not for statistical analyses. Accordingly, a total of 21 variables were included in statistical analyses.
Principal component analyses (PCA) were performed in IBM SPSS Statistics v.20 (IBM Inc., Chicago, IL, USA) rescaling variables to unit variance. We followed an analytical procedure based on Jiménez-Mejías et al. [32]. Characters reaching more than 0.6 of weigh in principal components as well as eigenvalues greater than 1 were used to perform a second PCA. Kaiser-Meyer-Olkin's measure of sampling adequacy (KMO) and Bartlett's test of sphericity were also estimated to evaluate the suitability of the data for finding structure in both approaches. PCAs were performed twice, including and excluding putative hybrids. Discriminant function analysis (DFA) was then performed in IBM SPSS Statistic v.20, using all variables included in the first PCA approach, to evaluate for taxonomic significance of two morphogroups as described in Valcárcel & Vargas [33], considering as potentially significant those groups correctly classified in 80% of excluded cases as established in Jiménez-Mejías et al. [32]. We randomly selected 70% of all samples to perform the DFA using a cross-validation of the model over these samples. Then, the remaining 30% was used for an additional validation. Based on the finding of intermediate individuals in Sierra Nevada by Luceño [22], populations S1 and S2 from Sierra Nevada (Fig 1) were studied very carefully. According to our own results (see below) we removed the population S2 from Sierra Nevada (Fig 1) from the subsequent analyses. Thus, we classified population S2 entirely as hybrid based on the consideration by Luceño [22] as intermediate morphology and the presence of sterile individuals detected (pers. obs.). All individuals from the hybrid population (S2; Fig 1) were unselected for the DFA and used only for the validation of the model using unselected cases, to test the placement of each individual from this population in any of the two groups. Additionally, univariate analyses were performed based on groups detected in the PCA to evaluate the characters that best allow the discrimination between the two species/taxa/morphologies. The Shapiro Wilk normality test showed non-normal distribution for most of the variables. The violation of the normality criteria in PCA and DFA analyses can be assumed when considering results as indicative and not a final evidence for taxonomic decisions [33], as these analyses are almost CLML-max Maximum culm length in a specimen when more than one fertile and mature culm is present (mm) CLML-min Minimum culm length in a specimen when more than one fertile and mature culm is present (mm)

CLMW
Width of the culm in the medial region (mm) CLMW-max Maximum culm width in a specimen when more than one fertile and mature culm is present (mm) CLMW-min Minimum culm width in a specimen when more than one fertile and mature culm is present (mm) insensitive to such violation [34]. Then, variation between groups was evaluated through a Kruskal-Wallis one-way ANOVA and a post-hoc Mann-Whitney U pairwise test to assess for significant differences between morphogroups for each character.

Phylogenetic analyses
Eight of our already published [8] sequences of the ITS, ETS and G3PDH nrDNA regions of C. furva s.l. as well as the matK cpDNA region were used for this study including two samples from southern Iberian Peninsula and six samples from central-northern populations (S1 File). Given that we aim to test for phylogenetic significance for the two detected morphogroups (see results), the population considered hybrid (S2; Fig 1) was excluded because we could not assign it to any of the two morphogroups. Additionally, sequences of these four markers were downloaded for species in Carex section Glareosae. According to the phylogeny in Maguilla et al. [8], the following outgroup species were included: Carex arctiformis Mack., C. billingsii Sequences were automatically aligned using MUSCLE [35] and concatenated to be analyzed using Bayesian inference (BI) and maximum likelihood (ML) as performed in Maguilla et al. [8]. Substitution models were calculated for each DNA region in jModelTest v.2.1.3 [36] and selected based on the Akaike's Information Criterion weights (AICw [37]). Gaps were encoded based on the "simple indel coding" criterion described by Simmons and Ochoterena [38] and analyzed using a F81-like substitution model as suggested by MrBayes manual [39].

Nomenclature
The electronic version of this article in Portable Document Format (PDF) in a work with an ISSN or ISBN will represent a published work according to the International Code of Nomenclature for algae, fungi, and plants, and hence the new names contained in the electronic publication of a PLOS article are effectively published under that Code from the electronic edition alone, so there is no longer any need to provide printed copies. In addition, new names contained in this work have been submitted to IPNI, from where they will be made available to the Global Names Index. The IPNI LSIDs can be resolved and the associated information viewed through any standard web browser by appending the LSID   Table 1.
contained in this publication to the prefix http://ipni.org/. The online version of this work is archived and available from the following digital repositories: PubMed Central, LOCKSS.

Morphometric study
The scatter plot of the two main principal components from the analysis using all variables (S2 File), as well as the analysis performed using only nine and 11 variables (including and excluding the intermediate population respectively ; Fig 3 and 19.86% variance explained using selected variables including intermediates; 20.21% when excluded). The first group was formed by central and northern populations of C. furva s.l. whereas the second was constituted by population S1 from the southern Iberian Peninsula (Figs 1 and 3), which definitely fits with the morphology of the type specimen of C. furva s.s. [31]. Individuals belonging to the population considered as hybrid were dispersed in the scatterplot of the principal components one and two (PC1 and PC2), with individuals nested in both morphogroups ( Fig 3A and  When compared with the type specimen of C. furva Webb [31], every sampled individual in population S1 fits definitively with this type material, whereas individuals from population S2 (Fig 1)

Phylogenetic analyses
Concatenated and aligned matrix of the ETS, ITS, G3PDH and matK DNA regions consisted of 29 sequences (S1 File) and 2212 sites which include the codification of four indels. The nucleotide substitution model that best fits each DNA region based on jModelTest results were: GTR+I (AICw = 0.4309) for ETS, GTR+G (AICw = 0.7193) for ITS, HKY (AICw = 0.4217) for G3PDH and GTR+I (AICw = 0.3015) in the case of the matK cpDNA region.
Bayesian inference and ML analyses supported the monophyly of C. furva s.l. with 1.0 posterior probability (PP) and 88% bootstrap support (BS) respectively (Fig 4). Within C. furva s.l. two main clades were significantly supported: one grouping southern individuals (0.95 PP / 72% BS) and the other formed by central-northern individuals (1.0 PP). The central-northern clade comprised also a subclade (0.95 PP / 96% BS; Fig 4) represented by one individual from Spain (Sierra de Béjar, population C2; Fig 1) and one individual from Portugal (Serra da Estrela, population C1; Fig 1).

Discussion
A new species hidden within C. furva s.l.
Consideration of two different species within C. furva s.l. was precluded based only on more traditional morphological studies because of the existence of morphologically intermediate individuals in the southern Iberian Peninsula [22]. In a molecular approach excluding the inferred hybrid population (Fig 4), two out of the three monophyletic clades significantly supported within C. furva s.l. have both geographical (Fig 1) and morphological (Fig 3) significance. Moreover, DFA analyses correctly classified 92% of unselected cases (S3 File), and ten out of the 21 measured characters presented significant differences between groups based on Mann-Whitney U test (S4 File). These evidences are enough for the consideration of two different species: C. furva s.s. (Fig 5) and a new species, C. lucennoiberica (Figs 6 and 7). This clear morphological and genetic differentiation between C. furva s.s. and C. lucennoiberica (Figs 3 and 4; S2-S4 Files) when excluding hybrid individuals from the analyses is in congruence with the observations by Luceño [22]. The new species fits the criteria of taxonomic [40] and phylogenetic [41,42] species. Moreover, the finding of sterile specimens occurring in the hybrid population (S2; Fig 1) suggests incipient reproductive isolation between C. furva s.s. and C.  lucennoiberica. Therefore, the two species might also fulfill the criteria to be biological species [43,44]. A third significantly supported clade is found within C. furva s.l. (0.95 PP / 96% BS; Fig 4) gathering a sample from Serra da Estrela (C1; Fig 1) and another from Sierra de Béjar (C2; Fig 1). The absence of morphological and/or geographic significance of this group (Fig 3) leads us to suggest the existence of a simple genetic structure within C. lucennoiberica.
Individuals in the intermediate population share morphological similarities with both species (Fig 3A), although most of them seems to be closer to C. lucennoiberica (most individuals fall within C. lucennoiberica morphospace in the PCA (Fig 3A) and 72.7% of individuals were assigned to C. lucennoiberica morphogroup in the DFA (S3 File)). In addition, the finding of sterile specimens and the classification of less than 80% of cases in one or another group (S3 File), justify the consideration of population S2 (Fig 1) in the southern Iberian Peninsula as hybrid population with morphological affinities to both species. When there is a hybrid zone -as it occurs in C. furva s.s. and C. lucennoiberica-, studies only based on morphology might fail in finding clear limits among the species involved. Luceño [22], in a morphology-based study without statistical methods behind it, highlighted that C. furva s.l. could constitute two independent biological entities. However, individuals with intermediate morphology found in the south of the Iberian Peninsula (Sierra Nevada) prevented him describing a new species. Descriptions of new Carex species based exclusively on morphology have been published recently (e.g. [45,46]). New animal species has even been described based only on molecular data, with neither morphological nor ecological traits differentiating each taxon (e.g. [47]; although similar cases have not been found in plants). Describing new species on the unique base of molecular data has been considered as something to avoid since molecular data should be used as an additional evidence for species delimitation [48,49]. Combined approaches of both morphological and molecular data and statistical analyses of those data are currently the most frequent practice for species delimitation and new species descriptions in botany as well as in zoology (e.g. [50][51][52][53][54][55][56]). The combination of morphological and molecular data has been previously shown to be a powerful tool to resolve the taxonomy in Carex (e.g. [10,57]). This highlights the utility of combined approaches in the detection and description of cryptic species even in countries or regions where the flora is very well studied and known.

Occurrence of C. lucennoiberica in the southern Iberian Peninsula
Whereas C. furva s.s is restricted to Sierra Nevada in the southern Iberian Peninsula (Fig 1), additional studied specimens (see paratypes) revealed that C. lucennoiberica is restricted to mountains in center-northern Iberian Peninsula (Sierra Segundera, Sierra del Cornón, Fuentes Carrionas, Serra da Estrela, Sierra de Gredos and Sierra de Guadarrama), but also present in the southern Iberian Peninsula (Sierra Nevada, Fig 1). One herbarium specimen in the southern Iberian Peninsula fits morphologically with C. lucennoiberica (see paratypes), whereas several studied materials from different herbaria present intermediate morphology. The observed intermediate morphologies (Fig 3A) and the existence of sterile specimens (pers. obs.) point to the coexistence of C. lucennoiberica and C. furva s.s. in the southern Iberian Peninsula, suffering active hybridization. Moreover, the reinterpretation of cytogenetic studies in C. furva s.l. [22] shows that C. Threats and conservation of the species The highly specific niche requirements of both species (C. furva s.s. and C. lucennoiberica) are the most limiting factors for their conservation, making them sensitive to climate change and habitat destruction (i.e. soil nitrification). Carex lucennoiberica is much endangered in Portugal, where only one individual occurs in Serra da Estrela (population C1 ; Fig 1), the only population in this country. In Spain, the most threatened population occurs in Sierra de Guadarrama (C5 ; Fig 1), where only seven individuals have been detected after several recent intensive searches. Carex furva s.s. seems to be also threatened by hybridization with its most closely related congener, C. lucennoiberica. Hybridization implies a serious threat for endangered species [58] and can affect the fitness of the species by genetic assimilation or outbreeding depression [59] as has been demonstrated in plants and animals [59][60][61]. In our study case, four out of 15 sampled individuals from the hybrid population in Sierra Nevada (population S2; Fig 1) showed aborted utricles (pers. obs.) which suggest outbreeding depression as the consequence of hybridization between C. lucennoiberica and C. furva s.s., which is an additional potential threat for the future conservation of C. furva s.s. Only the populations of C. lucennoiberica in Sierra de Guadarrama (C5 ; Fig 1), and C. furva s.s. in the southern Iberian Peninsula (Fig 1) are legally protected by the Spanish government, considered as "sensitive to habitat alteration" in the case of population from Sierra de Guadarrama (C5; Fig 1) and "Vulnerable" in Sierra Nevada. Nevertheless, all existing populations of C. furva s.l. occur in protected natural places, which is indirectly contributing to the conservation of both species. Diagnosis-Previously considered within the range of morphological variability of C. furva Webb. Differs from C. furva s.s. by the usually ovoid or shortly oblong, light brown inflorescence (instead of capitate-aggregated and rounded shape-and dark brown), its utricle beaks nearly appressed to the spike (instead spreading and prominent beaks in the outline of the inflorescence), utricles prominently veined (instead faintly veined), longer culms usually procumbent at maturity (instead shorter and erect), glumaceous or foliose lowest bract of the inflorescence (instead always glumaceous), glumes equal or slightly shorter than utricles (instead shorter than utricles), and utricles slightly smaller, usually the lower ones erect, rarely spreading (instead always spreading). See Table 2 for a detailed comparison of both species. Description-Rhizome lax or densely cespitose. Culms 3.7-23.8 (29.2) cm × 0.48-1.18 (1.42) mm, trigonous, with acute angles, smooth or slightly scabrid in the upper part, usually procumbent at maturity. Leaves 1.6-7.9 (13.8) cm × 0.64-2.57 mm, shorter than the culms, the longest ones reaching the inflorescences, rarely surpassing it, flat except at the apex, where it becomes trigonous, amphistomatic, smooth, antrorsely scabrid at the apex; ligule 0.26-1.28 (3.1) mm, usually as wide as the leaf blade, obtuse, rounded or emarginated; basal sheaths with blade absent, from entire to fibrous, brown. Lowest bract often linear or setaceous, with widened base having scarious margins, sometimes glumaceous, shorter than inflorescence, up to half its length, antrorsely scabrid on margins. Inflorescence 5.1-15 × 3-9.3 (9.7) mm, ovoid or shortly oblong, rarely capitate-aggregated and rounded shape-, usually light brown, consisting of 3-8 gynecandrous spikes 2.1-7.1 × 1.2-4.1 mm, ovoid to elliptic, erect to erect-spreading, overlapping, sometimes the lowermost distant, erect-spreading. Male glumes ovoidoblong, often with nerve prolonged in a mucro, sometimes the nerve scabrid; female glumes (0.92) 1.08-2 × 0.64-1.56 mm, as long as or shorter than the utricles, ovoid, with apex variable, often acute or subacute, 1-nerved, reddish-brown, sometimes with scarious-hyaline margins up to 1.52 mm wide. Utricles 1.48-2.37 mm × (0.56) 0.74-1.16 (1.26) mm, plano-convex to slightly biconvex, usually ellipsoid, prominently veined, erect to erect-spreading, the lower rarely spreading, greenish to brown at maturity, gradually attenuated into a beak (0.12) 0.2-0.58 (0.62) mm, truncate to slightly and irregularly bidentate, sometimes with a suture prolonging up to 0.6 (0.7) mm on the abaxial side, sometimes slightly curved at maturity, smooth or rarely with 1 (2) prickles. Achenes (0.98) 1.08-1.48 mm × 0.62-1.04 mm, biconvex or plano-convex, ± elliptical, with a persistent style base shortly cylindrical. 2n = 60 [22]. Phenology-Flowering and fructification occur from June to September. Etymology-The specific epithet-lucennoiberica-is an acronym of two words: "lucennoi" and "iberica". The first word-"lucennoi"-honors prof. Dr. Modesto Luceño (born in 1955). He is a Spanish caricologist who leads a research group at the Universidad Pablo de Olavide (Seville, Spain), focusing on the study of the evolution and systematic of the genus Carex (Cyperaceae). He is one of the authors of the most comprehensive taxonomic treatment for the whole Cyperaceae family for the Iberian Peninsula. He was the first who detected and published the presence of morphological variability within C. furva s.l. although the presence of intermediate individuals in Sierra Nevada prevented him to describe a new species. The second word-"iberica"-describes the distribution of the species endemic to the Iberian Peninsula (Spain and Portugal).

Taxonomic treatment
Nomenclatural note-Carex lagopina Wahlenb. var. baetica J.Gay was described in 1839 [23]. Because Carex baetica Auersw. ex Willk. was also already described in 1948, we cannot just combine the previously described variety at the rank of species.
Distribution and habitat-Endemic to Sierra Nevada, Granada province, Spain. Occurring in wet meadows, bogs, snowbeds, streams and lakes border of siliceous mountains.  [20,26,27] and hybrid speciation seems to be not relevant from an evolutionary point of view [19,20], hybridization can hinder the detection and characterization of incipient species. Combination of morphological and molecular data with differential treatment of hybrid populations has allowed the description of a new cryptic species endemic to the Iberian Peninsula, C. lucennoiberica.
Supporting Information S1 Table. Measurements of morphological variables. A priori and a posteriori identifications based on our results are shown, as well as label information and herbarium identifier, and population name as in