Geographical differentiation in floral traits across the distribution range of the Patagonian oil-secreting Calceolaria polyrhiza: do pollinators matter?

Abstract

Background and aims: The underlying evolutionary processes of pollinator-driven floral diversification are still poorly understood. According to the Grant-Stebbins model speciation begins with adaptive local differentiation in the response to spatial heterogeneity in pollinators. Although this crucial process links the micro- and macroevolution of floral adaptation, it has received little attention. In this study geographical phenotypic variation was investigated in Patagonian Calceolaria polyrhiza and its pollinators, two oil-collecting bee species that differ in body size and geographical distribution.

Methods: Patterns of phenotypic variation were examined together with their relationships with pollinators and abiotic factors. Six floral and seven vegetative traits were measured in 45 populations distributed across the entire species range. Climatic and edaphic parameters were determined for 25 selected sites, 2-16 bees per site of the most frequent pollinator species were captured, and a critical flower-bee mechanical fitting trait involved in effective pollination was measured. Geographical patterns of phenotypic and environmental variation were examined using uni- and multivariate analyses. Decoupled geographical variation between corolla area and floral traits related to the mechanical fit of pollinators was explored using a Mantel test.

Key results: The body length of pollinators and the floral traits related to mechanical fit were strongly correlated with each other. Geographical variation of the mechanical-fit-related traits was decoupled from variation in corolla size; the latter had a geographical pattern consistent with that of the vegetative traits and was mainly affected by climatic gradients.

Conclusions: The results are consistent with pollinators playing a key role in shaping floral phenotype at a geographical scale and promoting the differentiation of two floral ecotypes. The relationship between the critical floral-fit-related trait and bee length remained significant even in models that included various environmental variables and an allometric predictor (corolla area). The abiotic environment also has an important role, mainly affecting floral size. Decoupled geographical variation between floral mechanical-fit-related traits and floral size would represent a strategy to maintain plant-pollinator phenotypic matching in this environmentally heterogeneous area.

Keywords: Abiotic environmental gradients; Calceolaria; Patagonia; bee morphology; floral ecotypes; geographical range; local adaptation; oil-collecting bees; oil-offering flowers; phenotypic covariance; specialized pollination; speciation; vegetative morphology.

Fig. 1.

Map of the study area showing the 45 Calceolaria polyrhiza populations sampled. Populations pollinated by Chalepogenus caeruleus (grey circles) or Centris cineraria (grey squares) or those where pollinators were not observed (black dots) are also indicated (based on Cosacov, 2010). Dark and light shading indicate C. polyrhiza populations located in Andean-Patagonian forest or in the extra-Andean plain, respectively. Locality numbers correspond to those in Appendix 1 (N_loc). The inset depicts a shaded relief map of South America with the study area indicated with a box.

Fig. 2.

Schematic diagram of a flower, leaf and plant of Calceolaria polyrhiza showing linear morphometric measurements. (A) Lateral view of a flower showing fertile parts, elaiophore and the distance between the floral reward (elaiophore) and the fertile parts, i.e. throat length (ThroL). Black dashed line indicates the point where flowers were dissected to obtain parts shown in (B). (B) Floral traits: throat length (ThroL), elaiophore width (EW), corolla area (CA) indicated by the black surface, theca length (TheL), filament length (FL) and style length (SL). (C) Vegetative traits: total leaf length (TLL), leaf lamina length (LLL), leaf lamina width (LLW). (D) Plant traits: inflorescence height (IH), plant height (PH), plant maximum width (W1) and perpendicular width (W2).

Fig. 3.

Oil-collecting bee species pollinating the flowers of Calceolaria polyrhiza. (A) The large bee Centris cineraria seizing the flower with its mandibles and mid-legs, and collecting oil with the foreleg. Note the stigma and stamens making contact with the dorsal part of the bee's head (arrow). The critical trait involved in effective pollination, distance between the head and the middle legs (H-ML), is indicated. (B) The small bee Chalepogenus caeruleus. Note the stigma and stamens making contact with the dorsal part of the bee's thorax (arrow) and the pellet of pollen and oils accumulated on the hind legs.

Fig. 4.

Structure of phenotypic variation across the distribution range of Calceolaria polyrhiza. (A) Components of variance (CV) expressed as percentages of total variance among populations, among individuals within populations, and within individuals (error term). ***P < 0·001, **P  < 0·01, *P < 0·05. (B) Among-population CV/within-population CV ratio, showing main source of variation (local versus geographical) for each trait measured. Traits related to floral fit are indicated. Codes for the different morphological variables are provided in Materials and methods and in Fig. 2.

Fig. 5.

(A) Spatial autocorrelograms of Moran's I coefficient for ten floral and vegetative traits as a function of geographical distance among Calceolaria polyrhiza populations in Patagonia. Solid symbols are coefficients that differ significantly from zero; open symbols are non-significant coefficients. The shaded area indicates distance classes that should not be considered because they included fewer than ten pairs of points. (B) Geographical patterns of variation in floral and vegetative traits in 45 populations of C. polyrhiza. Only significant relationships are shown. Codes for the different morphological variables are provided in Materials and methods and Fig. 2.

Fig. 6.

Phenotypic covariation analysis. (A) Phenotypic correlations among traits measured. Positive correlations among traits are represented by solid lines connecting the traits; line thickness indicates the overall magnitude of each correlation. Dashed lines indicate negative correlations between traits. Floral mechanical-fit-related traits and corolla area are indicated. Codes for the different morphological variables are provided in Materials and methods and Fig. 2. (B) Results of Mantel test (r) indicating similarity between the empirical matrices (vegetative + floral traits or floral traits alone) and the respective theoretical matrices (development, functional and nested). Differences between theoretical matrices (d–f, d–n and f–n) in the power to explain the empirical matrix are also shown. **P  < 0·01, *P < 0·05, ^∧P = 0·05.

Fig. 7.

Biplot of the first two axes of the RDA ordination for 25 Calceolaria polyrhiza populations (black circles). The explanatory environmental factors (dashed arrows) that were significant (P < 0·05) determinants of floral (A) and vegetative (B) morphological variation (filled arrows) are shown. The eigenvalue associated with each axis is provided in parentheses. The explanatory variables are described in Materials and methods and their values are reported in Table S2. Population acronyms are given in Appendix 1.

Fig. 8.

(A) Geographical variation in the functional trait throat length of Calceolaria polyrhiza and in bee length of its pollinators, Centris cineraria and Chalepogenus caeruleus, in 25 populations. Values of these traits are proportional to the size of the symbols. (B) Relationship between throat length of C. polyrhiza and bee length of its pollinators.