Geography is essential for reproductive isolation between florally diversified morning glory species from Amazon canga savannahs

The variety, relative importance and eco-evolutionary stability of reproductive barriers are critical to understanding the processes of speciation and species persistence. Here we evaluated the strength of the biotic prezygotic and postzygotic isolation barriers between closely related morning glory species from Amazon canga savannahs. The flower geometry and flower visitor assemblage analyses supported pollination by the bees in lavender-flowered Ipomoea marabaensis and recruitment of hummingbirds as pollinators in red-flowered Ipomoea cavalcantei. Nevertheless, native bee species and alien honeybees foraged on flowers of both species. Real-time interspecific hybridization underscored functionality of the overlap in flower visitor assemblages, questioning the strength of prezygotic isolation underpinned by diversification in flower colour and geometry. Interspecific hybrids were fertile and produced offspring in nature. No significant asymmetry in interspecific hybridization and hybrid incompatibilities among offspring were found, indicating weak postmating and postzygotic isolation. The results suggested that despite floral diversification, the insular-type geographic isolation remains a major barrier to gene flow. Findings set a framework for the future analysis of contemporary evolution of plant-pollinator networks at the population, community, and ecosystem levels in tropical ecosystems that are known to be distinct from the more familiar temperate climate models.

For comparative purposes, all flowers are to the same scale, bar = 1 cm. Flowers were photographed in the wild, in between 7 am to 12 pm, thus the variation is unlikely due to the incomplete opening or senescence. Furthermore, in ex situ collection flowers did not change their shape from 7 am to 15 pm. For example, the cup shaped flowers of pink flowered EB139 do not change (Figure 2a), indicating that shown variation is more genetically determined than environmentally controlled. Flowers were arranged according to the curvature of the limb. At the top-left image is a flower shape that we operationally call a cup. The flower series ends with the bottom-right shape called here an umbrella. All flowers were aligned by the proximal end of their corollas. Thus, the next trait the viewer can easily apprehend is variation in flower tube length. Note, that the widest part of the tube is not at a flower throat, as usual in Ipomoea, but closer to the middle. In our sampling of 498 flowers, all shapes were not unique in populations. The selected shapes are part of a continuum comprising intermediate shapes, thus we avoid assigning exact shape frequencies.
Supplementary Figure S4. Flower shape variation in I. marabaensis.
All flowers are to the same scale, bar = 2 cm. Images were selected from a sample of I. marabaensis n = 472 from all cangas. Other details as in Supplementary Figure 3.
Supplementary Figure S5. Descriptive statistics of the canga-delimited trait diversification illustrated with boxplots.
The band inside the box is a median value; box spans the upper and lower quartiles; boxconnected lines are whiskers showing the lowest and highest datum still within 1.5 interquartile range; dots are outliers. Letters above each plot represent statistically significant differences at p < 0.05 according to the Wilcoxon post-hoc test. Individual boxplots correspond to cangadelimited samples, the ID's of which are listed along X axis as: cangas N1 to N9; S11 is S11 plateau canga; SO -Sossego granitic inselberg; Tcanga Tarzan. Cangas N4 and N5 are sympatric and host putative hybrids "N4H", "N5H"; I. cavalcantei, "N4C", "N5C"; and I. marabaensis, "N4M", "N5M". Labels "M1" and "M2" identify cangas Morro 1 and Morro 2. Sample species identity is shown by box-plot colouring: red -I. cavalcantei; blue -I. marabaensis; blackputative hybrids that here are treated as an independent taxon.

(a)
Flower tube length. Flower sepals, filaments and styles were removed. The corolla tube length was measured as a distance in cm for the landmark at corolla bottom to the flower throat. Tube length limits access to nectar between visitors that cannot enter the tube, differentiating by the length of tongues.

(b)
Herkogamy. Shown are the distances from stigma to the distal tip of the longest stamen in the flower. Values > 0 or < 0, are the ascending and descending herkogamy, respectively.
Analysed genes were for subunits of RNA polymerase II, RPB2-1, RPB2-2, RPB3; granulebound starch synthase, WAXY. These results complete those shown the Fig. 3  Close up of the black eye phenotype. In a flower to the right, all reproductive organs are absent, a fully developed black eye. In a flower to the left, a style and two filaments are gone. Remaining three filaments are detached at their bases and are being in a process of removal from the flower.

(i)
Trigona pallens bee and perforation at the flower tube of hybrid EB081 at ex situ collection.

(j)
Trigona pallens at a flower throat. Flower is sterilized by these small bees that remove styles and filament-anthers organs. We explain this behavior as nectar robbing by legitimate access to the flower and followed by breaking into the nectar chamber by demolishing chamber entrance that is formed by the filament bases as described in footnotes to Supplementary Table S14.

(k)
I. marabaensis flower and T. pallens bees. The species flower tubes are broader than in I. cavalcantei, and easily accommodate three bees here. Note two filaments-anthers pushed out of the flower.

(l)
T. pallens bee legitimately access EB081 hybrid flower. Only two anthers and style are remaining at that moment. (m, o) Flowers of EB081 produced on a given day. All four flowers tube bases have perforations made by T. pallens bees, i.e. four flower a day -all damaged. (n, p) Nectar wells. Three flowers of I. cavalcantei pink-flowered plant EB139. In (p), nectar wells (our term) are near the flower bases of two flowers to the left. This is a third route, which we think is the smartest, used by the T. pallens bees to access the nectar. Bees carve out flower tissues through the sepals. Once a bee reaches the nectar chamber, it stops, at least we have not seen any damage to the carpels. Many insects, including honeybees, visit nectar wells made by T. pallens. The nectar wells are very common and corresponding circular scars are easily recognizable on fruit capsules. Figure S10. Honeybee race identification.

Supplementary
Other abbreviations and column identifiers as in the Supplementary Table S12. produced by approximately 12 hours-old flowers at anthesis. Average values (AV). In Ipomoea flowers, the bases of five filaments bulge out from the adaxial corolla surface touching style. Those morphological features separate the proximal tube space, forming what is known as a nectar chamber. Five triangular openings to nectar chamber are small, ca. 1 mm at widest side in I. cavalcantei and I. marabaensis, and are further occluded by well-developed trichomes, likely discriminating nectar foragers by the dimensions of their tongues and other mouth parts.
In some flowers, nectar can overspill from the chamber into the tube proper. This nectar fraction is more easily accessible to visitors such as beetles or wasps, which were numerous. The flower calix, a whorl of sepals, covers the corolla tube parts that comprise nectar chamber. b I. cavalcantei (C); I. marabaensis (M), colour variants (pink) and interspecies hybrids (H; magenta) were analysed. . Flowers were placed in plastic tubes filled with water to just cover the flower peduncle for water supply and maintenance of turgor. After the transport to the laboratory, nectar volumes were measured in the afternoon, ca. 7 hours after collection in a wild. Average value (AV) is shown in the bottommost row. The difference in average here as compared to Supplementary Table 11, suggests that either some nectar has been taken by the flower visitors between 5:00 am and 7:30 am, or/and physiological differences between plants from canga and plants in ex situ collection that is positioned on land cleared from rain forest. It was not feasible to bag flower buds in wild populations. b This column indicates whether beetles were present (YES or NO) in the flower at the time of nectar measurement. The species of beetles (not identified in this work, body length ca. 1 cm) were commonly found in the tubes of I. cavalcantei flowers. It is unlikely that the architecture of beetle mouth is suitable for the animal to reach nectar concealed within nectar chamber. Beetles often aggregated up to ten individuals in a single flower tube, possibly mating. Flower tube is also a good place to avoid predation. We did not find indications that I. cavalcantei beetles were feeding of flower tissues. This is different from I. marabaensis flowers in which smaller (2-5 mm body length) species (at least 4 species) of beetles were very common. Beetle infestation correlated with flower tube discoloration due to animal grazing on adaxial flower tissues, which however did not lead to perforations in a tube.