Salvia austriaca

Mátis, Attila, Malkócs, Tamás, Kuhn, Thomas, Laczkó, Levente, Moysiyenko, Ivan, Szabó, Anna, Bădărău, Alexandru S. & Sramkó, Gábor, 2023, Hiding in plain sight: Integrative analyses uncover a cryptic Salvia species in Europe, TAXON 72 (1), pp. 78-97 : 82-84

Distribution of the taxa of the Salvia austriaca group. —

Our collection of occurrence data for taxa of the Salvia austriaca group with two different pollen deposition modes (i.e., dorsal and lateral deposition) indicates strong eco-geographical differentiation with an almost complete vicariance of the two forms ( Fig. 3 ). The taxon with lateral pollen deposition (i.e., S. austriaca s.str.) occupies the forest-steppe zone of the Pannonian basin, Transylvanian basin, the Pontic steppe, the SE part of Crimea, and the Balkans, whereas Salvia FDPD is confined to the short-grass steppe zone (or true steppe zone) as defined by Lavrenko & al. (1991) of the Pontic steppe. The westernmost observations of Salvia FDPD are from northern Bulgaria (settlement: Bozhurluka), whereas the easternmost is from the city of Rostov ( Russia).

Morphological differentiation between Salvia austriaca and Salvia FDPD. — While processing the collected samples, we noted that several individuals exhibited reduced, shortened staminal levers, indicating that Salvia FDPD might be exhibiting gynodioecy, a dimorphic sexual system where populations are composed of both hermaphroditic and female individuals. Gynodioecy is known in S. austriaca , S. candidissima , S. cyanescens and S. pratensis as well ( Kaul, 1988; Zhang & Classen-Bockhoff, 2019). Measurements of quantitative floral traits were performed only on hermaphroditic flowers. Based on the results of the linear mixed-effects models, five of the six investigated floral traits differed significantly between the two taxa: calyx length (χ 2 = 34.70, p <0.0001), corolla height (χ 2 = 6.18, p = 0.0129), corolla length (χ 2 = 107.64, p <0.0001), absolute length of the abaxial lever arm (χ 2 = 269.67, p <0.0001), and absolute length of the adaxial lever arm (χ 2 = 10.25, p = 0.0014) had significantly greater values for Salvia FDPD compared to S. austriaca s.str. ( Table 1 ). In contrast, we only observed a marginally significant difference between the number of stem leaf pairs (χ 2 = 6.11, p = 0.0134), where the higher values also belonged to Salvia FDPD. There was no difference in the other vegetative traits between the two taxa (data not shown).

In the PCA, 74.6% of total variation in flower-traits was explained by the first two PC-axes (PC1 – 54.0%, PC2 – 20.6%). However, the two focal taxa were differentiated only along the first PC-axis ( Fig. 4 ).

Phylogenomics, species delimitation and divergence dating. — Our maximum likelihood (ML) phylogenetic tree reconstruction based on unlinked genome-wide SNPs produced a well-resolved phylogram ( Fig. 5A,B ), in which samples of the same a priori classified taxa are monophyletic and strongly supported (BS = 100), but some of the tip nodes received moderate support (i.e., 70 ≤ BS ≤ 99). The topology strongly supports the separation of the Salvia austriaca group from the rest of the samples (our a priori outgroup) along the longest branch on the tree ( Fig. 5A ). Within the latter clade, S. nutans is sister to the crown-clade formed by S. pratensis and S. nemorosa .

The species tree reconstruction based on coalescence analysis ( Fig. 6 ) yielded a fully congruent topology to the ML-tree, in which taxa of the Salvia austriaca group are sister to the lineage formed by S. nutans , S. pratensis and S. nemorosa . The next split within this lineage is much younger and is represented by S. nutans , which separated from the other two species during the Middle Pleistocene. The following isolation event is marked by the separation of the two taxa of the Salvia austriaca group, which was followed by the split between S. nemorosa and S. pratensis . All nodes were fully supported in the species tree (posterior probability [PP] = 1.0). The divergence dating analysis indicated that S. austriaca s.str. and Salvia FDPD diverged 0.197 mya (95% confidence interval [CI]: 0.302 –0.087 mya) ( Fig. 6 ).

The DAPC analysis attained the lowest BIC scores to models with 5 and 7 clusters (suppl. Fig. S1 ). At K = 5, all individuals were assigned into a cluster—according to existing classification and splitting the taxa of the Salvia austriaca group—with full posterior probability ( Fig. 5C ). Interestingly, models with higher number of clusters assigned individuals from S. nemorosa and S. pratensis into separate clusters, whereas the clusters formed by samples of the taxa of the Salvia austriaca group were not divided (suppl. Fig. S2 ). Similarly, at K = 4, S. nemorosa and S. pratensis were lumped together, whereas S. austriaca s.str. and Salvia FDPD were still conceived as separate groups (suppl. Fig. S2 ).

The BFD favored the alternative model, which considered Salvia austriaca s.str. and Salvia FDPD as separate entities. The null model, grouping all individuals into one species within the Salvia austriaca group (i.e., S. austriaca s.l.), performed worse in both repetitions with a mean BF value of 348.67 in favor of the alternative model ( Table 2 ) across the two repetitions. This is considered decisive according to Kass & Raftery (1995). Mean ESS was 336.34 (standard deviation = 200.74) across both models and all steps and repetitions, implying that the chain length was adequate to reach convergence.

Figures

Fig. 3. Distribution of Salvia FDPD and S. austriaca Jacq. Delimitation of the steppe zone follows Lavrenko & al. (1991). Delimitation of the Crimean Submediterranean forest ecoregion follows Olson & al. (2001). For geographic coordinates of each observation, see suppl. Table S1.

Fig. 4. PCA biplot of Salvia L. samples based on six quantitative floral characters, where the arrows represent loadings. Blue circles, S. austriaca Jacq. (N = 16); red triangles, Salvia FDPD (N = 26); ch, corolla height; cl, corolla length; cal, calyx length; a, absolute length of the abaxial lever arm; b, distance between joint and pollen-sacs; c, absolute length of the adaxial lever arm.

Fig.5. Phylogenetic relationships between the studied taxa. A, Phylogenetic tree shown as a phylogram resulting from a ML search in RAxML using the ascertainment bias correction option. Circles at nodes represent maximum bootstrap support (BS = 100), thick branches represent BS ≥70, and thin branches represent BS <70, where branch thickness increases in proportion to bootstrap support. The two main internal branches were shortened to condense the figure. B, Topology of the untruncated ML phylogram. C, Posterior membership probabilities based on the first two discriminant functions of four PCs using group assignment from k-means clustering at K = 5 in a DAPC analysis. The order of individuals follows the order of the tips of the phylogenetic tree. The beginning of an individual’s ID refers to the taxon: rev, Salvia FDPD; aus, S. austriaca Jacq. s.str.; prat, S. pratensis L.; nemo, S. nemorosa L.; nut, S. nutans L.

Fig. 6. The dated species tree of the studied Salvia L. taxa based on the best model of species delimitation as inferred by SNAPP. Blue bars represent 95% confidence intervals. The geological time chart below the chronogram is presented according to the International Union of Stratigraphy.

Fig. 1. A, Spatial distribution of sampled Salvia L. populations; B, Study area highlighted in red on a map of Europe; C, Transylvanian sites enlarged; D, Dobrogean sites enlarged. — White, Salvia austriaca Jacq. s.str.; black, Salvia FDPD; circle, genetic samples; diamond, morphometric samples; square: both genetic and morphometric samples; white lines represent country borders.

Fig. 2. Quantitative floral traits measured: absolute length of the abaxial lever arm (a), distance between joint and pollen-sacs (b), absolute length of the adaxial lever arm (c). Photo: Gabriel Gigea.