Evolution of oil-producing trichomes in Sisyrinchium (Iridaceae): insights from the first comprehensive phylogenetic analysis of the genus

Abstract

Background and aims: Sisyrinchium (Iridaceae: Iridoideae: Sisyrinchieae) is one of the largest, most widespread and most taxonomically complex genera in Iridaceae, with all species except one native to the American continent. Phylogenetic relationships within the genus were investigated and the evolution of oil-producing structures related to specialized oil-bee pollination examined.

Methods: Phylogenetic analyses based on eight molecular markers obtained from 101 Sisyrinchium accessions representing 85 species were conducted in the first extensive phylogenetic analysis of the genus. Total evidence analyses confirmed the monophyly of the genus and retrieved nine major clades weakly connected to the subdivisions previously recognized. The resulting phylogenetic hypothesis was used to reconstruct biogeographical patterns, and to trace the evolutionary origin of glandular trichomes present in the flowers of several species.

Key results and conclusions: Glandular trichomes evolved three times independently in the genus. In two cases, these glandular trichomes are oil-secreting, suggesting that the corresponding flowers might be pollinated by oil-bees. Biogeographical patterns indicate expansions from Central America and the northern Andes to the subandean ranges between Chile and Argentina and to the extended area of the Paraná river basin. The distribution of oil-flower species across the phylogenetic trees suggests that oil-producing trichomes may have played a key role in the diversification of the genus, a hypothesis that requires future testing.

Fig. 1.

Floral morphology of the four genera in tribe Sisyrinchieae included in the present study: (A) Orthrosanthus monadelphus (SP174), (B) Solenomelus pedunculatus (SP221), (C) Olsynium scirpoideum (SP009), (D) Sisyrinchium tinctorium (SP240), (E) S. convolutum (SP108), (F) S. cuspidatum (SP038), (G) S. scariosum (SP087), (H) S. micranthum morphotype T (SP054) and (I) S. demissum (SP034).

Fig. 2.

Main geographical patterns of distribution within the range area of Sisyrinchium: (A) subandean ranges, between Argentina and Chile; (B) extended Paraná river basin area; (C) central area circumscribed between the south-western United States to Bolivia; (D) north American area. The red area included in Area A corresponds to the distribution areas of S. arenarium, S. chapelcoense, S. convallium, S. cuspidatum, S. graminifolium, S. macrocarpum, S. nervosum, S. papillosum, S. pearcei and S. striatum.

Fig. 3.

Maximum-likelihood topology obtained from the full combined molecular dataset and generated under the GTR + I + Γ model with PhyML 3·0 online web server (Guidon et al., 2005) showing branch lengths (–ln likelihood = 25978) for 101 Sisyrinchium, six Olsynium and two Solenomelus accessions, and Orthrosanthus monadelphus as outgroup species. The major clades are labelled I to IX. Thick red branches indicate support of at least 80 % parsimony bootstrap (PBS), 80 % likelihood bootstrap (LBS) and 0·95 posterior probability (PP). Thick green lines indicate branches in which only two of the three support values reach these minimum values and the third support value is at least 50 % (PBS), 60 % (LBS) and 0·95 (PP). Thick blue lines indicate branches in which only one of the three support values reaches the minimum values set above and the two other support values are at least 50 % (PBS), 55 % (LBS) and 0·95 (PP). Accessions of the genus Sisyrinchium all belong to the clade included in the coloured rectangle.

Fig. 4.

Consensus tree based on the strict consensus tree of the parsimony ratchet, the estimated maximum-likelihood tree and the Bayesian 50 % majority-rule consensus tree obtained from the analysis of the full molecular dataset. Support values along nodes are parsimony bootstrap (PBS, above branches), likelihood bootstrap (LBS, on the left, below branches) and Bayesian posterior probability (PP, on the right, below branches). The major clades are labelled I to IX. Thick red branches indicate support of at least 80 % (PBS), 80 % (LBS) and 0·95 (PP). Thick green lines indicate branches in which only two of the three support values reach these minimum values and the third support value is at least 50 % (PBS), 60 % (LBS) and 0·95 (PP). Thick blue lines indicate branches in which only one of the three support values reaches the minimum values set above while the two other support values are at least 50 % (PBS), 55 % (LBS) and 0·95 (PP).

Fig. 4.

Consensus tree based on the strict consensus tree of the parsimony ratchet, the estimated maximum-likelihood tree and the Bayesian 50 % majority-rule consensus tree obtained from the analysis of the full molecular dataset. Support values along nodes are parsimony bootstrap (PBS, above branches), likelihood bootstrap (LBS, on the left, below branches) and Bayesian posterior probability (PP, on the right, below branches). The major clades are labelled I to IX. Thick red branches indicate support of at least 80 % (PBS), 80 % (LBS) and 0·95 (PP). Thick green lines indicate branches in which only two of the three support values reach these minimum values and the third support value is at least 50 % (PBS), 60 % (LBS) and 0·95 (PP). Thick blue lines indicate branches in which only one of the three support values reaches the minimum values set above while the two other support values are at least 50 % (PBS), 55 % (LBS) and 0·95 (PP).

Fig. 5.

Consensus trees based on the strict consensus tree of the parsimony ratchet, the estimated maximum-likelihood tree and the Bayesian 50 % majority-rule consensus tree resulting from the analysis of the plastid + mitochondrial datasets on one hand (left tree) and the nuclear dataset on the other hand (right tree). The major clades are labelled I to IX. Red lines connect S. jamesonii (A) and S. chilense (B), the placement of which differs between the two topologies. Thick red branches indicate support of at least 80 % parsimony bootstrap (PBS), 80 % likelihood bootstrap (LBS) and 0·95 posterior probability (PP). Thick green lines indicate branches in which only two of the three support values reach these minimum values and the third support value is at least 50 % (PBS), 60 % (LBS) and 0·95 (PP). Thick blue lines indicate branches in which only one of the three support values reaches the minimum values set above and the two other support values are at least 50 % (PBS), 55 % (LBS), 0·95 (PP).

Fig. 6.

Mirror trees showing the optimization of the type of nuptial glandular trichomes (Tree A) and the distribution of nuptial trichomes (Tree B) on the consensus tree (see Fig. 4), with maximum parsimony. External nodes are coloured according to the character state observed, internal nodes are coloured according to the ancestral state inferred. Multicoloured branches correspond to unresolved ancestral character states under parsimony analysis. The evolutionary history of both characters was also traced using maximum likelihood. Likelihood supports are indicated for seven internal nodes (2–7 and 5·1). The absence of a square in front of the external nodes corresponds to missing data.

Fig. 7.

Nuptial trichomes of Sisyrinchium striatum (clade III). (A) Cross-section of the flower showing the glabrous staminal column and the distribution of oil-glandular trichomes on the adaxial surface of tepals (arrows). (B and C) SEM micrographs showing the distribution and the morphology of oil-glandular trichomes on tepals at anthesis (adaxial surface): (B) overview of the basal part of the inner tepal; (C) oil-glandular trichome topped by the secretory cavity. (D) Differential interference contrast imaging of a nuptial trichome showing the oil blister (arrow). (E and F) Fluorescence micrographs under 450–500 nm light excitation of the same trichome stained with Nile Red: (E) red emission observed with a 610 LP emission filter showing accumulation of polar lipids in a subcuticular space (oil blister); (F) yellow emission observed with a 535 ± 20 emission filter – neutral lipids are concentrated in the oil blister.

Fig. 8.

Nuptial trichomes of S. tenuifolium (A and B), S. macrocarpum subsp. macrocarpum (C and E) and S. claritae (F and G). (A) Flower of the Central American S. tenuifolium (clade II). (B) Distribution of nuptial glandular trichomes (devoid of lipidic secretions) along the filamental column. (C) Overview of the andro-gynoecial apparatus of S. macrocarpum subsp. macrocarpum (clade III): oil-glandular trichomes are located all along the fused and free parts of the filaments. (D) Nile Red fluorescence (red emission) observed on trichomes with a 610 LP emission filter (polar lipids are concentrated in the oil blister). (E) Nile Red yellow emission observed on trichomes with a 535 ± 20 emission filter (neutral lipids). (F) Arrangement of oil-glandular trichomes along the filamental column of S. claritae (clade V). (G) Nile Red fluorescence (red emission) observed on trichomes located on the lower part of the filamental column with a 610 LP emission filter (polar lipids).

Fig. 9.

Nuptial trichomes of S. setaceum (A and C), S. micranthum (D and F) and S. angustifolium (G and H). (A) Lateral view of the S. setaceum (clade V) flower showing the distribution of oil-glandular trichomes along the filamental column. (B) Nile Red fluorescence (red emission) observed on trichomes with a 610 LP emission filter showing polar lipids accumulation in the oil blister. (C) Nile Red yellow emission observed on trichomes with a 535 ± 20 emission filter showing neutral lipids accumulation in the oil blister. (D) Arrangement of oil-glandular trichomes along the carboy-shaped filamental column of S. micranthum (clade VII). (E) Trichomes stained with Nile Red: polar lipids observed with a 610 LP emission filter are concentrated in the oil blister. (F) Nile Red yellow emission observed on trichomes with a 535 ± 20 emission filter showing neutral lipid accumulation in the oil blister. (G) Location of nuptial trichomes (devoid of lipidic secretions) on the lower half part of the filamental column of S. angustifolium (clade IX). (H) SEM micrograph showing the arrangement and the morphology of glandular trichomes on the filamental column.

Fig. 10.

Optimization of geographical distribution on the consensus tree (see Fig. 4), with maximum parsimony. External nodes are coloured according to the state observed for the current distribution area of each sample, internal nodes are coloured according to the ancestral state inferred. Multicoloured branches correspond to unresolved ancestral states. The most-parsimonious states are indicated above nine internal nodes (2–10).

Fig. 11.

Major clades (I to IX) identified within the consensus phylogenetic tree for Sisyrinchium (see Fig. 4). The subgenera (1) proposed by Goldblatt et al. (1990) and the sections (2) defined as subdivisions of the genus by Ravenna (2000, 2002, 2003b) are indicated on the right side of the figure. Species are designated for subgenus Sisyrinchium (red colour) and subgenus Echtronema (green colour), according to Bentham and Hooker (1883), Cholewa and Henderson (1984), Kenton et al. (1986), Rudall et al. (1986), Goldblatt et al. (1989, 1990) and Cocucci and Vogel (2001). Species labelled with the two colours were alternatively classified in the two subgenera. Uncertainties are coloured in grey.