1. Introduction
Globularia L. is a genus of angiosperms composed mostly of small evergreen and perennial shrubs, recognizable by spherical inflorescences of blue-violet flowers [
1,
2,
3,
4]. The number of species belonging to this genus and their taxonomic status have been interpreted in several ways by the authors who have studied these plants over the years. Currently, 27 taxa are included in this genus [
5], of which many are narrow endemics [
1,
6,
7]. Recent molecular studies have demonstrated that
Globularia, which was traditionally included in a separate family (Globulariaceae) along with
Poskea Vatke, actually belongs to Plantaginaceae [
8]. Otto Schwarz proposed the first extensive classification of
Globularia, recognizing 22–25 taxa [
3,
9]. Subsequently, after years of discordant or incomplete taxonomy, the genus underwent taxonomic modifications by several authors [
4,
6,
10]. In the current classification of the genus
Globularia, eight sections are recognized:
G. section
Lytanthus (Wettst.) O.Schwarz,
G. section
Polycephalium O.Schwarz,
G. section
Carradoria (A.DC.) Wettst.,
G. section
Hellenion O.Schwarz,
G. section
Globularia (Syn.: section
Aphyllanthes O.Schwarz),
G. section
Alypum (Fisch.) O.Schwarz,
G. section
Empetron O.Schwarz, and
G. section
Gymnocladium O.Schwarz [
4]. Molecular data have provided evidence for Miocene origin of
Globularia, and the independent evolution during the Pleistocene in three European Alpine and two Mediterranean groups [
11,
12,
13,
14]. Accordingly, representatives of the genus are widespread in most of Europe. However, the highest concentration of taxa is found in Central and Southern Europe, Anatolia, Northern Africa, and Macaronesia [
1,
5]. Disjunct populations are also found in the Atlantic and Swedish islands and around the Volga [
5]. Most species are adapted to dry and open habitats, often on calcareous soils, in meadows or on bare rocks [
6].
The typical life-forms of
Globularia include hemicryptophytes and chamaephytes, with a primary root dominating the root system in most species during the whole life of the plant, with the exception of the
G. cordifolia species complex (Figures 1 and 8) [
4]. Stems are made of branches and stolons, along with flowering scapes, which often possess one to many bracts [
4]. When in small number, such bracts on the flowering scape are known as scales [
2]. True leaves, which are simple and mostly evergreen, can be arranged in several ways, but are generally sparse or fused in a basal rosette or on several bundles [
4]. Flowers are collected in solitary capitula, with individual flowers clustered on a common receptacle. Such receptacles can be hairy, conical, cylindrical, or globose. The external flower bracts forming the envelope are more or less similar in shape to the internal ones, and both type of bracts often persist on the receptacle after flowering. Fruits are achenes enclosed in a persistent calyx, usually 1–1.5 × 0.5 mm [
6,
15].
Some
Globularia species are well known for their medicinal use [
16,
17]. Accordingly, several species of
Globularia have been phytochemically investigated, given their contents in iridoids and phenolic compounds with potential biological applications [
18]. Conversely, in the last decades, the genus has been scarcely studied from a taxonomical point of view, with some exceptions [
1,
7,
19,
20,
21]. During the years, some species have always been clearly recognizable from a taxonomic point of view (e.g.,
Globularia alypum L.,
Globularia nudicaulis L., or
Globularia incanescens Viv.), whereas there has always been great confusion on
G. cordifolia L. and allied species [
6,
21,
22,
23,
24].
Globularia section
Empetron includes
G. cordifolia,
Globularia meridionalis (Podp.) O.Schwarz,
Globularia repens Lam., and the narrow endemic
Globularia neapolitana O.Schwarz, which can all be found in Italy [
2,
4,
25]. All of these species share most of their morphological character-states, being woody shrubs no taller than 25 cm, living in the same habitats, that is, calcareous rocks from sea level to high elevations (
Figure 1) [
2,
6].
Globularia cordifolia shows the largest area of occurrence, from Turkey to Spain, whereas
G. repens is limited to the western Mediterranean, from Spain to northern Italy [
6].
Globularia meridionalis is restricted to the Balkans, from Greece and Bulgaria to Croatia, Slovenia, and Austria, and extends along the Italian Apennines [
3]. In the southern Apennines, the narrow endemic
G. neapolitana is known for some populations in the Sorrento-Amalfi peninsula and the Island of Capri [
26,
27,
28], although some populations have been reported in the nearby Taburno-Camposauro complex in sympatry with
G. meridionalis [
29,
30]. According to early karyological data,
G. meridionalis is diploid with 2
n = 16 chromosomes, although autopolyploids exist with 2
n = 32 chromosomes, whereas
G. cordifolia (2
n = 32) is an allotetraploid allegedly originating from
G. repens (2
n = 16) and
G. meridionalis [
9,
31]. However, according to Milletti [
6], there is no evidence of true karyological differences between
G. cordifolia and
G. meridionalis, given that tetraploid level 2
n = 32 is reported for both species, as confirmed also by other authors [
22]. On the other hand,
G. repens is diploid with 2
n = 16 chromosomes [
10], and
G. neapolitana has been reported as tetraploid (2
n = 32) or even 2
n = 16 in some cases [
23]. Thus, with the exception of
G. repens, all species in
G. section
Empetron are, or can be, tetraploid; share the same habitats; and largely overlap in their distribution.
Concerning the morphological features that have been used to distinguish these species, Milletti [
6] made an extensive survey of the different criteria used throughout the years, which can be summarized as (a) the general habitus of the plant, including its life-form and overall size; (b) the leaf size and its apex shape; (c) the scale number on the flowering scape; (d) the shape of the outer and inner bracts, including the presence and abundance of hairs; and (e) the calyx shape, with particular regard to the ratio between the teeth and the tube length. Pignatti [
2] used a combination of the aforementioned features to circumscribe the species complex within
G. section
Empetron, but according to Milletti [
6] there is too much variability in most of these features, with the exception of the calyx. According to Ravnik [
21,
22], populations from
G. cordifolia species complex can be different from one another in the size and shape of their leaf blades, the serration of the leaf edge, the type of the flower calyx and corolla, and the shape and sharpness of the outer bracts. However, they never differ in all of these features, whereas different types of the same feature can, at the same time, be present in an individual plant. The same author suggested these differences to be consequences of polymorphisms and proposed to consider the sole taxon
G. cordifolia [
21,
22].
Evidence from recent molecular data support the hypothesis that
G. section
Empetron is one of the latest branching lineages in the evolution of
Globularia, and that
G. meridionalis, G. cordifolia, and
G. neapolitana are close to
G. repens [
1,
5]. According to Milletti [
6], such large variability could be explained by the wide area of occurrence of the species and their habitat fragmentation, noting a large polymorphism and quoting Willkomm [
32] for the explanation, “
Hic speciei typus per innumeras formas intermedias transit in varietatem” (English: “
This type of species passes from variety to variety through countless intermediate forms”). Admittedly, mechanisms of speciation are still poorly understood in
Globularia, and fine-scale distributional, population genetic, morphological, and reproductive data are needed to further clarify the evolutionary history of these plants [
5].
Yet, there is a large amount of recent literature exploring morphological variations in plants as dependent on environmental factors. For instance, sea daffodils (
Pancratium maritimum L.) vary their morphological features according to the bioclimatic variables to which they are exposed, including precipitation, average annual temperatures, and strong irradiance [
33]. Subtler climate effects were observed in
Pinus flexilis E.James, where there was no macroscopic variation in morphology on a large elevation gradient (1600–3300 m), yet stomatal density was inversely correlated with elevation [
34]. Climate has been shown to have strong potential effects on flower morphology as well [
35]. Undoubtedly, the availability of global-scale bioclimatic variables [
36] and the introduction of novel statistical techniques can help to shed new light to ecological and biological phenomena [
37]. Yet, the study of morphology remains a key-instrument in plant systematics and ecology, both with traditional approaches [
38,
39,
40] or with modern geometric morphometrics [
41,
42,
43].
Thus, in our research, we tested the hypothesis that the variability in shape and size observed in
G. cordifolia species complex could be explained by the impact of bioclimatic variables. In order to do so, we used both classical and geometric morphometrics, by extracting principal components (PCs) from multivariate datasets and by using them as outcomes, as well as bioclimatic variables as predictors, in order to test the effect of climate on size and shape variability. Although our final outputs were based on a comparatively small sample size (10 populations;
Figure 2 and Table 3), the data were derived from a large number of individuals and organs per populations, and were based on robust, cross-validated statistical models.
3. Discussion
Our morphometric approach confirmed the great variability in
G. cordifolia species complex, both in size and shape. A set of morphometric features that would reliably discriminate
G. cordifolia,
G. meridionalis, and
G. neapolitana was not observed, supporting the results from earlier morphological investigations of the same species [
6,
21,
22]. All morphometric variables, when considered at a population level, did not show significant differences, with the exception of LMA. Pignatti [
2] suggested that the key diagnostic characteristic is the ratio between calyx teeth and tube, with the tube being much longer in
G. cordifolia compared to that in
G. meridionalis, in which teeth and tube should be sub-equal, whereas in
G. neapolitana, teeth are longer than the tube. Our results did not confirm such a pattern, given that teeth are clearly longer than the tube only for
G. cordifolia (C-VA) and
G. meridionalis (M-FV, M-MU, M-CI), whereas all other populations showed teeth slightly longer than, or sub-equal to tubes (C-FZ, M-CS, and M-FI). Similarly, Milletti [
6] found that the calyx teeth were always longer than the tube, and rarely sub-equal. Concerning the separation of
G. cordifolia from
G. meridionalis, a distinction on outer bracts was also proposed, specifically on the presence of hairs and the shape of the bracts characterized by the maximum width positioned at the base or at the center of the bract [
2]. We did observe the presence of hairs in some populations of
G. meridionalis (see
Supplement 2), yet with a great degree of variability in abundance, and we agree with Milletti [
6] that this character can be deemed unreliable. Outer bracts were reported as strongly mucronate in
G. neapolitana [
2], a feature that we did notice in most specimens from N-MO (see
Supplement 2), although there was no difference in their size compared with those of
G. cordifolia and
G. meridionalis. Milletti [
6] observed that the only true morphological differences existing between
G. repens and
G. cordifolia concern the leaves not exceeding 2 × 0.5 cm, always with entire apex, in the former. Conversely,
G. meridionalis was considered a heterotypic synonym of
G. cordifolia, whereas
G. neapolitana was retained as a subspecies of
G. cordifolia [
6,
23], being distinguished by leaf shape, that is, lamina from spatulate-obovate to suborbicular with crenate-undulate margin vs. lamina from oblanceolate-obovate to spatulate-cuneiform with entire margin [
6].
Globularia neapolitana was also reported as bearing nude flowering scapes [
6].
On the basis of such a confused taxonomic picture, our novel approach shed new light on the morphological variability in
G. cordifolia species complex, showing that ecological predictors could explain both size and shape, revealing an apparent lack of taxonomical differences among species. Most of the variation in size, involving leaf and bract morphological variables, could be explained by a combination of solar radiation, temperature seasonality, and annual precipitation. Among the several available environmental predictors, we found that elevation was not significant in any model. Accordingly, previous evidence showed that elevation could not influence morphology in
G. cordifolia [
6], although there has been no in-depth study on
Globularia morphology on elevation gradients. The evidence from literature highlighted that elevation might cause a variation in a clinal pattern, as was observed with
Penstemon sp. pl. [
44], or even no variation on strong altitudinal gradients, as observed in
Sesleria rigida Heuff. ex Rchb. [
45]. In our case, variation on classical morphometric PC1 was shown to be correlated with a combination of solar radiation and annual precipitation, with noticeable changes in flower and receptacle variables. Solar radiation, especially UV-B radiation, can trigger a broad range of responses in plants at the molecular, cellular, and organism level, including the structure of the inflorescences [
46,
47]. In bromeliads from xeric environments, variations in the shape of the rosette, leaf color, and size of the leaf sheath and blade were shown to be correlated with solar radiation [
48]. As for annual precipitation, anatomical differentiation of populations of
S. rigida were found to be significantly correlated with annual precipitation and the precipitation of the wettest month, whereas temperatures were not significant [
45]. Conversely, there is a general consensus that mean annual temperatures are significantly more strongly correlated with plant traits than mean annual precipitation, although such evidence can be biased due to the weak link between mean annual precipitation and the availability of water to plants [
49].
We presented, for the first time, extensive LMA measurements for
G. cordifolia species complex. Pierce et al. [
50] reported specific leaf area only for
G. cordifolia as 6.3 mm
2 mg
−1. This value corresponds to LMA value of 158.7 g/m
2, which is in line with our measurements. From a life-form perspective, LMA values like those we have found in our samples are reported as typical for evergreen shrubs, although they are overlapping also with evergreen gymnosperms and succulent plants [
51]. From an ecological point of view, values of 200 g/m
2 are typical for shrubland or desert species [
51]. These data confirm the known ecology and evolutionary history of
Globularia section
Empetron as adapted to arid rocky xeric environments [
1,
5,
12], but do not contribute to their taxonomy. LMA was shown to be significantly higher for
G. meridionalis than the other two species, but LMA was also the driving variable on PC2 for classical morphometrics and PC1 for geometric morphometrics which, in turn, were correlated with temperature seasonality. Temperature seasonality is a measure of temperature change over the course of the year and it is computed as standard deviation of the 12 mean monthly temperature [
52]. This variable has been shown to be an important ecological predictor for the distribution of many plant species, either rare or widespread [
53,
54,
55,
56,
57]. This predictor clearly helped us to distinguish some populations (M-GP, M-MU, and M-FV) that are above 6.2 °C temperature seasonality, exhibiting the highest values of most morphological variables, as well as also explaining most of the variation in shape. Comparing the values of temperature seasonality with LMA, M-MU, M-SF, and M-FV showed LMA values above 200 g/m
2 and up to 260 g/m
2 for M-GP. According to ecological literature, these LMA values indicate that either drought, nutrient limitation, or both can have a limitative effect on plant growth [
51], which can give insights to a morphological adaptation to specific stresses.
We cannot be completely sure on how these morphological variations can be beneficial for the life of
Globularia populations as a response of ecophysiology. Undoubtedly, plants can show a surprising degree of morphological plasticity at either small or large scale. As a matter of fact, in two populations of
Prunus serotina Ehrh., one from a xeric and one from a mesic environment, sun leaves from the xeric population had greater thickness, specific mass, and guard cell length than the sun leaves from the mesic population, yet no morphological difference was found in shade plants in either population [
58]. Plants that have evolved in arid environments, such as
G. cordifolia, can show morphological changes to environmental modifications at a comparatively fast rate. For instance, two desert shrubs from Central Asia responded to recent increase in precipitation both in their morphology and physiology [
59].
Changes in calyx and receptacle size could also reflect changes in pollinators, given that coevolution with pollinators can generate within-species geographic variation in the morphology of plant species, eventually leading to plant speciation [
60]. Pollinators can be “drivers” of variation, such as in the case of
Arum maculatum L., a widespread species with a specialized pollination system [
61]. In this plant and its pollinators (two species of flies), a geographically structured variability in pollinators was found, with increasing proportion of one species with higher annual precipitation and lower precipitation in the warmest trimester, two features typical of the Mediterranean zone [
61].
Globularia flowers are bilabiate and protogynous, and observed pollinators have been butterflies, bees, beetles, and syrphids [
6,
62,
63], although detailed study on pollinators in this genus, especially in
G. cordifolia species complex or along elevation gradients, are scarce and outdated [
64]. Thus, it cannot be ruled out that the trends in calyx and receptacle morphology can be a consequence of a shift in pollinators on climatic gradients.
Our study on size was accompanied by a geometric morphometrics approach on leaves. Our test for allometry showed that there was a strong effect of size onto shape. Allometry can be a pivotal component of shape variation, and generalized procrustes analysis (GPA) removes isometric effects of size on shape, but not allometric effects [
65]. Thus, our approach of performing shape analysis on size-corrected shapes allowed us to interpret only the variation on shape once allometric affects have been taken into account [
66,
67]. Given the lower amount of variation of modelling with PC1 of geometric morphometrics, the variation of shape was smaller than size and less correlated with ecological predictors. We found that temperature seasonality can explain most of the variations in shape, in the same way as it explained variation in PC2 for classical morphometrics. Leaf shape variations have been shown to be a functional response to altitude and longitude at regional scales rather than to temperature-related factors such as latitude [
68]. Environmental gradients in shapes, with special regard to elevation, were also found in
Sophora davidii Franch., providing strong evidence that variations in morphological and genetic parameters reflect morphological and genetic adaptation to native habitats, highlighting ecological and evolutionary consequences along altitudinal gradients of mountainous ecosystems [
69]. The shape of the leaf apex was widely indicated in the past as diagnostic [
2,
15,
70]. Nevertheless, the shape of the leaf apex (rounded vs. mucronate vs. three-toothed) has been considered a largely inconstant character, greatly varying among and within populations (see Figure 9) [
6,
21,
22]. Ravnik [
21,
22] also reported that different leaf shapes, that, according to Schwarz [
3] are characteristic for
G. cordifolia or
G. meridionalis, can be present in a single plant at the same time. We found no relevant variation in the leaf-apex landmarks in our shape analysis.
Despite a relatively small sample size compared to the area of occurrence of the studied plants, which might limit the breadth of this research, we provided evidence that the variability in size and shape within G. cordifolia species complex could be connected to environmental factors. Accordingly, we did not observe any reliable morphological pattern that would support the current classification into different species, albeit our research was ecologically oriented and not a taxonomic study. Admittedly, the observed absence of morphological evidence that would support current classification G. cordifolia species complex might simply be a consequence of low statistical power due to our insufficient sample size, along with the underrepresentation of G. cordifolia (two samples) and the narrow endemic G. neapolitana (one sample). Nevertheless, our analysis was supported by an underlying large sample size per individuals as well as robust evidence from cross-validation of the models.
Moreover, it should be noted that lack of characteristics that would support the separation of
G. cordifolia and
G. meridionalis into different species was recently recorded by Friščić [
71], who studied different aspects of
Globularia species from Croatia. On the basis of molecular [
1,
5], phytochemical [
18,
72,
73], and karyological data [
6,
22,
71], as well as previous morphological data [
6,
21,
22]
, G. meridionalis formed a continuum with
G. cordifolia and might be considered without any taxonomical value as indicated also by earlier studies of plant material from Italy, Slovenia, Croatia, Bosnia and Herzegovina, Serbia, and Macedonia [
6,
21,
22,
23,
24]. Concerning
G. neapolitana, the single population that we sampled (N-MO) falls within the same morphological pattern, and might therefore be included as a heterotypic synonym of
G. cordifolia. On the basis of our evidence, the reported sympatry between
G. meridionalis and
G. neapolitana [
30] would be unrealistic. According to molecular data,
G. neapolitana forms a single lineage with
G. cordifolia, with the only difference that possible ancestral ranges for
G. cordifolia should be from the Circumboreal Region, whereas
G. neapolitana belongs to the Mediterranean basin [
5]. Nevertheless, literature data suggest that a population referred as
G. neapolitana from the island of Capri, given its isolation and low-elevation, shows morphological differences substantiating its taxonomic position as a distinct unit [
6]. Such peculiarity is confirmed by anomalies in the chromosome counts and by cultivation experiments, where plants retained their morphology, suggesting a genetic stabilization of characters [
6,
23]. A herbarium sheet by G. Gussone dating back to 1808 is present in the Herbarium Neapolitanum (NAP!). The original label refers to
G. bellidifolia Ten. with a vague geographical reference to Capri, with no indication of precise toponyms. The sheet was later revised by M. Ricciardi in 1976, who identified the samples as
G. neapolitana. The conservation status of the herbarium sheet, the low number of organs, and the impossibility of evaluating clone sampling prevented us from considering this herbarium sheet in our analysis. Despite field research, we were unable to retrieve any
Globularia populations in Capri, but we cannot exclude that, at least in this population, there could be a patroendemic systematic unit [
74].
Given its ploidy level, recent origin, and distribution,
G. cordifolia is likely a result of glaciations [
12]. Polyploid species are favored under stressful environmental conditions [
75], and patterns of morphological clines within and among plant species are known in the Mediterranean basin, as in the case of
Pinguicula [
76],
Soldanella [
77]
, Fritillaria [
78]
, and other mountain-Mediterranean taxa, including
Globularia sp. pl. from other sections besides
G. section
Empetron [
5,
13,
14]. Although there is a need for a larger scale genetic and morphometric analysis in
G. cordifolia species complex [
1], along with an in-depth cultivation study to assess if there is a potential genetic basis for plant traits [
79] and/or any effect of soil, our morphological analysis, along with ecological statistics, helped to bring about a new perspective this puzzling group of species.