City University of New York (CUNY) City University of New York (CUNY)

CUNY Academic Works CUNY Academic Works

Dissertations and Theses City College of New York

2018

Evolution of Floral Morphology and Symmetry in the Miconieae Evolution of Floral Morphology and Symmetry in the Miconieae

(Melastomataceae) (Melastomataceae)

Maria Gavrutenko City College of New York

How does access to this work benefit you? Let us know!

More information about this work at: https://academicworks.cuny.edu/cc_etds_theses/712

Discover additional works at: https://academicworks.cuny.edu

This work is made publicly available by the City University of New York (CUNY). Contact: AcademicWorks@cuny.edu

1

Evolution of Floral Morphology and Symmetry in the Miconieae (Melastomataceae)

Maria Gavrutenko

Thesis Advisor: Dr. Amy Berkov

A thesis submitted in partial fulfillment of the requirements for the

Degree of Master of Science in Biology

The City College of New York

The City University of New York

2018

2

Abstract

Analyses of evolution of floral morphology and symmetry broaden our understanding of the

drivers of angiosperm diversification. Integrated within a flower, labile floral characters produce

different phenotypes that promote variable interactions with pollinators. Thus, investigation of

floral evolution may help infer potential historic transitions in pollinator modes and ecological

pressures that generated present diversity. This study aims to explore morphological evolution of

flowers in Miconieae, a species-rich Neotropical tribe within family Melastomataceae. Despite a

constrained floral plan, Melastomataceae manage to achieve a variety of floral traits appealing to

diverse pollinator types, with majority of the species requiring specialized “buzz pollination” by

bees. However, previous research in Miconieae documented several instances of convergent

evolution of phenotypes associated with generalized pollination strategies. I explore

morphological evolution of Miconieae in a phylogenetic context to understand how

diversification relates to different phenotypes, and how common evolution of generalized

pollination systems is within this tribe. I reconstructed the largest species-level phylogeny of

Miconieae with maximum likelihood inference, and combined it with morphological data on a

variety of floral characteristics for over 350 species scored from field photographs. I analyzed

trait evolution with ancestral state reconstruction of discrete and continuous characters with R

packages ape and phytools. Trait correlation was estimated with Pagel’s statistical method for

discrete characters and with regression analysis of phylogenetically independent contrasts of

continuous characters. I analyzed diversification in the tribe with Bayesian Analysis of

Macroevolutionary Mixtures (BAMM) and explored the effect of character state evolution on

diversification with Binary State Speciation and Extinction (BiSSE) approach. My analyses

reveal rampant convergent and correlated evolution of multiple characters indicative of

3

pollinator-mediated selective pressures. I confirm several parallel trends in evolution of

generalized floral phenotypes. I find an association between generalization trends and increased

diversification rates that may be related colonization of highland environments within the tribe.

Key words: Miconieae, Melastomataceae, floral evolution, floral symmetry, pollination,

generalized pollination, pollination syndromes, diversification rates, BAMM, BiSSE.

Introduction

The flower is a complex reproductive structure and an integrated system of constituent

floral parts (Faegri and van der Pijl 1979). Floral morphology has understandably been a focus of

many evolutionary and taxonomic studies aimed at explaining observed diversity of angiosperms

(Anderson et. al 2002). Floral parts carry out distinct functions with a goal of successful sexual

reproduction, which makes the entire system subject to evolutionary forces. Exploring variability

in floral morphology in a phylogenetic framework may point to drivers of evolution within a

clade.

The idea that selective pressures to accommodate certain types of pollinators play an

important role in the evolution of flower morphology is at the heart of pollination syndrome

theory (Fenster et al. 2004). Different floral features promote interactions with particular pollen

vectors, while different pollinators in turn exert variable pressures on their associated plants.

These intricate plant-animal interactions lead to the development of specific pollination

syndromes, which are combinations of floral morphological traits appealing to a certain

pollinator guild (Rosas-Guerrero et al. 2014). Thus, analyzing variability in floral traits could

provide insight into the evolutionary history of plant-pollinator interactions. Such an

investigation is particularly intriguing within Melastomataceae, a plant family associated with a

4

wide variety of generalist and specialist pollinators despite conserved flower morphology (Brito

et al. 2016, Renner 1989).

Melastomataceae is a diverse family with over 4500 species (Renner 1993, Michelangeli

et al. 2004). Melastomataceae manage to achieve variety of floral traits necessary to appeal to

diverse pollinator types, despite a constrained floral plan (complete flowers, with free floral parts

and no fusion within or between whorls). The majority of Melastomataceae offer pollen as a

reward and require specialized “buzz pollination” by bees (Renner 1989). In such flowers, pollen

is enclosed in tubular anthers with minute terminal pores, and it is only released when a bee

grasps a stamen and vibrates the anther at a high frequency using its flight muscles (Renner

1989, Luo et al. 2008, Buchmann 1983). Other specialized modes of pollination have evolved

several times within the family, with some flowers providing nectar rewards to hummingbird,

rodent, and bat pollinators (Lumer 1980, Renner 1989, Judd 2007) or food body rewards to

passerines (Dellinger et al. 2014). However, several instances of independent evolution of

generalized pollination systems have been reported as well (Goldenberg 2008, Brito et al. 2016,

Brito et al. 2017, Kriebel and Zumbado, 2014). Generalist pollinators respond to nectar and

pollen rewards and include flies, wasps, butterflies, and non-buzzing bees.

Shifts in pollination mode have been suggested to be evolutionary drivers of variability in

floral morphology in Melastomataceae, and specifically of stamen diversification in Miconieae

(Brito et al. 2016, Goldenberg et al. 2008, Luo et al. 2009). Miconieae is a wide-ranging,

phylogenetically complex Neotropical tribe of approximately 1800 species (19-23 genera;

Michelangeli et al., 2008). Variation in stamen morphology within Miconieae is correlated with

shifts to generalized pollination systems (Brito et al. 2016). Interestingly, parallel evolutionary

trends in floral morphology seem to have occurred multiple times, suggesting that ecological

5

pressures have led to convergent evolution of similar phenotypes (Goldenberg et al. 2008). Shifts

towards generalized pollination strategies have often been interpreted as a means to increase

reproductive success in depauperate environments devoid of bees, such as highlands (Brito et al.

2017, Goldenberg et al. 2008, Varassin et al. 2008).

In this study, I explore morphological evolution of floral traits within Miconieae in a

phylogenetic context. I examine the phenotypic diversity of multiple floral characters that are

subject to pollinator-mediated selection. Anther anatomy (pore size, anther length) determines

the kinds of pollinators that can successfully extract pollen, such as buzzing bees in case of

minute pores or a wide range of non-buzzing animals in case of broad pores or slits. I examine

coloration of anthers and petals along with presence of color contrast, as these traits present

visual signals that may attract pollinators (Lunau 2006). Variation in the size and shape of petals

may bestow different functions on the corolla, from attracting a pollinator in the case of large,

colorful spreading petals to anther protection in cucullate (hood shaped) corollas. In addition to

these characteristics of individual floral parts, I also examine traits that characterize the entire

flower, such as symmetry of reproductive whorls (androecium and gynoecium) and herkogamy

(spatial separation of pollen and stigma to promote outcrossing).

I combine an extensive morphological dataset with a species-level phylogeny to

understand how diversification within Miconieae relates to suites of morphological features

within the flower. Better understanding of evolutionary history of labile floral characters helps to

infer potential historic transitions in pollinator modes. Floral morphology, symmetry, and the

variable pollinator interactions promoted by different floral phenotypes have repeatedly been

implicated in diversification rate variability in angiosperms (Givnish 2010). In this study, I

attempt to answer the following questions: (1) how is diversification in the tribe related to

6

different suites of morphological features? and (2) how common is the evolution of generalized

pollination systems in Miconieae? I expect the evolution of different traits to be correlated, and

anticipate detecting multiple instances of convergent evolution of different phenotypes.

Methods

Field work

In order to improve the previous knowledge of Miconieae systematics, I increased the

existing molecular and morphological datasets with new samples of Melastomataceae collected

during fieldwork in Peru. Collections were carried out in April 2016 in vicinity of Oxapampa

(Pasco Department) and Huancabamba (Piura Department) at elevations ranging from 500 m to

3700 m. Plants were photographed, and several specimens per collection were paper pressed for

further herbarium processing. Leaf tissues were silica-dried for DNA extraction. Flowers and

flower buds from flowering specimens were preserved in alcohol.

Molecular dataset

The molecular dataset was comprised of sequences available in GenBank and

supplemented with DNA sequences for previously unsampled species (Appendix 1). The

outgroup included 81 species from the tribe Merianieae (six genera), one species of Eriocnema

and three species of Physeterostemon (Table 1). I selected Merianieae as an outgroup, because it

was shown to be a sister group to the Miconieae in several previous studies (Michelangeli et al.

2004, Goldenberg et al. 2008, Martin et al. 2008). Eriocnema fulva also repeatedly came up as a

sister species to the Miconieae (Michelangeli et al. 2004, Goldenberg et al. 2008, Martin et al.

2008). The newly described genus Physeterostemon was added to the outgroup according to the

taxonomic assessment by Goldenberg and Amorim (2006) based on morphology and preliminary

phylogenetic analyses. The Miconieae were represented by 1083 species from 17 genera (Table

7

2). The molecular matrix included three nuclear (waxy, nrITS, nrETS) and six plastid (trnS,

psbK, accD, atpF, ndhF, rbcL) markers. Not all nine DNA regions were available for each

species (Appendix 1).

DNA from previously unsampled species was extracted from silica-dried leaf tissues

using Qiagen DNeasy plant mini-kit following the manufacturer’s protocols (Qiagen, Valencia,

California, USA) with some minor modifications that have shown to increase yield or quality for

the Melastomataceae (Martin et al. 2008). Four DNA regions were amplified with polymerase

chain reaction (PCR). PCR protocols for two nuclear ribosomal regions (nrITS1-nrITS2, nrETS)

and two chloroplast regions (psbKL, accD) are described in Table 3. Nuclear ribosomal regions

nrITS1 and nrITS2 were amplified simultaneously. Amplification results were visualized on 1%

agarose gels. I did not amplify or sequence the remaining five DNA markers (trnS, atpF, ndhF,

rbcL, waxy).

Sequencing was performed at the High-Throughput Genomics Center, Seattle, WA.

Sequence editing was done in Sequencher 5.4 (Gene Codes Corporation). Sequences were

aligned using MAFFT v.7 plugin in Geneious 9.0 (https://www.geneious.com, Kearse et al.,

2012). Phylogenetic relationships among species were reconstructed by Maximum Likelihood

using RAxML-HPC BlackBox through the CIPRES portal (phylo.org, Miller et al., 2010).

Phylogenies were first constructed for individual loci and inspected for consistency with

previously published results and for hard incongruences across gene trees. Suspicious sequences

were removed, and nine genetic markers were concatenated in Geneious into a molecular matrix

of 10851 bp. This final alignment was used to generate the phylogeny used for all downstream

analyses. Divergence times were estimated with chronos function from R package ape v. 4.1

(Paradis et al., 2004) with default molecular clock model (“correlated” rates among branches).

8

Due to the poor fossil record (only two fossils within the Melastomataceae, none of them within

Miconieae; Hickey, 1977, Renner et al., 2001), we could not use fossil priors to estimate

divergence times. We provided three secondary calibration points: 1) root age of minimum 25.78

and maximum of 34.6 million years for the common ancestor of Miconieae and Merianieae; 2)

minimum age of 14.47 and maximum age of 22.93 million years for the common ancestor of all

Miconieae and the genera Eriocnema and Physeterostemon; 3) minimum age of 9.67 and

maximum age of 17.57 million years for the divergence of Eriocnema and Physeterostemon from

Miconieae. These estimates are based on unpublished work by M. Reginato and F.A.

Michelangeli (Michelangeli et al., in preparation), and we consider them more accurate than the

previously estimated date of Miconieae origin (27.1 mya; Berger et al. 2016).

Morphological dataset

Morphological trait evolution was evaluated by scoring 358 species of Miconieae from

available flower images. Morphological characters were scored from field photographs of

flowers downloaded from the NYBG virtual herbarium (PBI Miconieae, Michelangeli et al.,

2009) and Melastomataceae de Centroamérica (melas-centroamerica.com, Kriebel, n.d.), as well

as new photographs taken during fieldwork. A morphological matrix was built in Xper3

(www.xper3.fr) and included 19 discrete and four continuous morphological characters

(Appendix 2).

Characters

I described floral symmetry using terminology summarized by Endress (1999).

1. Filament symmetry: polysymmetric (0); monosymmetric (1). Filament arrangement and shape

contribute to the androecium symmetry. In monosymmetric flowers, the filaments are oriented

9

towards one side of the flower, or arranged in two groups along a single axis of symmetry (Fig.

1A, M). Polysymmetric flowers exhibit multiple axes of symmetry (Fig. 1B-D).

2. Filament arrangement (in flowers with polysymmetric filaments): evenly spaced (0); grouped

in clusters of two (1). Filaments can be arranged along multiple axes of symmetry individually

(Fig. 1K, P), or in pairs (Fig. 1D). This character is inapplicable to monosymmetric filaments.

3. Filament arrangement (in flowers with monosymmetric filaments): all on one side (0); split

into two clusters (1). All filaments can occur on one side of the flower (Fig. 1L), or they can be

split into two groups on the opposite sides of the flower along a single axis of symmetry (Fig.

1M). This character is inapplicable to polysymmetric filaments.

4. Anther symmetry: polysymmetric (0); monosymmetric (1); asymmetric (2). Anther symmetry

is the second component of androecium symmetry and is usually, but not always, related to

filament symmetry. Polysymmetric anthers usually emerge on polysymmetric filaments (Fig.

1F). Monosymmetric anthers are usually the extension of monosymmetric filaments (Fig. 1A),

although in some species anthers resting on polysymmetric filaments acquire slight

monosymmetry during floral development (Fig. 1J). Asymmetric anthers are usually present on

monosymmetric filaments with few exceptions (Fig. 1G).

5. Anther curvature (in flowers with monosymmetric anthers): towards style/axis (0); away from

style/axis (1). Anthers can be curved towards the center of the flower, with anther pores opening

towards the style (Fig. 1A, M). They can also be curved towards the edge of the flower, with

pores facing away from the style (Fig. 1N). This character is inapplicable to flowers with

polysymmetric anthers.

6. Anther pore: minute (0); broad (1); rimose/slits (2). Minute pores are circular, with a diameter

that is much smaller than the diameter of the anther itself (Fig. 1E, L). Broad pores are circular,

10

but the diameter is much larger than that of the minute pores, and is close to the diameter of the

anther (Fig. 1K). Rimose pores are very large slits, have oval of irregular elongated shape, with a

diameter that is greater than that of the anther (Fig. 1P).

7. Style direction: upward (0); away from axis (1). Style can be directed upward (Fig. 1D, H), or

it can be directed away from the center of the flower starting at the ovary (Fig. 1B, F, N).

8. Style curvature: straight (0); curved (1). Style can be straight (Fig. 1B, E, H), or it may exhibit

bends or curves along its length (Fig. 1G, L).

9. Stigma direction (in flowers with curved styles): away from anthers (0); towards anthers (1).

Style may be curved in different directions, leading to two alternative stigma positions. Stigma

can be turned away from the anthers (Fig. 1O), or it can be facing the anthers (Fig. 1L). This

character is inapplicable to flowers with straight stigmas.

10. Number of petals: four petals (4); five petals (5); six or more petals (6). If the number of

petals varied among flowers in a photograph, the species was coded to have several character

states.

11. Number of stamens: diplostemonous (0); polystemonous (1). In diplostemonous flowers the

number of stamens is twice the number of petals (Fig. 1D, K). In polystemonous flowers, the

number of stamens is greater than twice the number of petals (Fig. 1 J).

12. Petal color: white (0); pink (1); yellow (2). Petals were coded as white only if no

pigmentation was observed (Fig. 1A, G, J). Pink petal color ranged from slightly pink, with only

partial pigmentation on otherwise white petals, to deep crimson (Fig. 1C, I, O). Yellow petals

ranged from pale yellow partial pigmentation on white petals to bright yellow (Fig. 1H, P). If

petal pigmentation varied among different flowers in a photograph (often only some flowers had

11

colorful petals, while the rest of the flowers were white), I coded the species as having colorful

petals.

13. Anther color: white (0); yellow (1); blue/purple (2). Anthers were coded as white only if no

observed flowers of the species exhibited any anther pigmentation (Fig. 1A, K). Yellow anthers

ranged from pale to bright yellow to nearly brown in some old flowers (Fig. 1L). Anthers were

coded as blue/purple if they exhibited any blue or red pigmentation (pink, dark red, light or dark

blue, purple; Fig. 1B, G, N).

14. Petal/stamen contrast: same color (0); different colors (1). Contrast between stamen color and

petal color. Species were coded as lacking contrast if both petals and anthers were white, yellow,

or pink (pink petals and blue/purple anthers; Fig. 1A, D, H, N).

15. Presence of herkogamy: non-herkogamous (0); herkogamous (1). Species were coded as

herkogamous if I observed spatial separation between anthers and stigma in at least some of the

flowers in a photograph.

16. Type of herkogamy: horizontal anther-stigma distance (arrangement or curvature) (0);

vertical anther-stigma distance (1); “poppy flower syndrome” (2). Horizontal distance between

anthers and stigma could be achieved through arrangement or curvature in filaments, anthers

and/or stigma, or through a combination of spatial arrangement and curvature (Fig. 1G, L, M).

Vertical distance between anther and stigma was usually due to a long exserted style extending

above the anthers (Fig. 1F), but sometimes it could be seen in protogynous flowers, where stigma

emerged from the bud before anthers began to unfold. Flowers with “poppy-flower syndrome”

resembled poppy flowers due to very broad lobed stigmas, where spatial separation was created

by the size of the stigma (Fig. 1J). Many flowers demonstrated combinations of herkogamy types

and were coded as having multiple character states.

12

17. Corolla plane: spreading (0); reflexed (1); upward (2). In spreading corollas petals were

positioned perpendicular to the floral axis (Fig. 1L, M). Reflexed corollas were composed of

petals curled back towards the peduncle (Fig. 1G, N). Upward corollas had petals hat did not

fully spread at anthesis and remained pointed up towards the style and stamens (Fig. 1C, D, I).

18. Corolla form (upward corollas): separate petals (0); tubular (1); cucullate (2). Separate petals

were directed upward, but they did not overlap (Fig. 1I). Tubular corolla was formed by

overlapping petals arranged into a tube (Fig. 1C). Cucullate corollas were formed by overlapping

petals curled into a hood-like structure over the stamens (Fig. 1H). This character is inapplicable

to flowers with spreading or reflexed corollas.

19. Stamen-style arrangement: opposite sides (0); same side (1). This character is only applicable

to flowers with monosymmetry in both androecium (through filament or anther shapes) and

gynoecium (through curvature or direction of the style). It reflects the relative positions of the

stamens and style (Fig. 1L: opposite sides; Fig. 1O: same side).

20. Minimum anther length (mm). Continuous character, sourced from literature (Almeda 2009,

Gamba and Almeda 2014, Judd 2007, Judd et al. 2013, Judd et al. 2014a, Judd et al. 2014b, Judd

et al. 2015, Kriebel 2016, Liogier 2000, Majure 2015, Michelangeli et al. 2005, Skean 1993,

Wurdack 1973 and 1980).

21. Maximum anther length (mm). Continuous character (see character 20 for sources).

22. Minimum petal length (mm). Continuous character (see character 20 for sources).

23. Maximum petal length (mm). Continuous character (see character 20 for sources).

Ancestral state reconstruction

To analyze trait evolution, I performed ancestral state reconstruction with function ace

from R package ape (Paradis et al. 2004) for discrete characters for three transition rate models

13

(equal transition rate between character states (ER), symmetric transition rate (SYM), and a

model with all rates different(ARD)). Model fit was evaluated with fitDiscrete function from ape

package, and the best model was selected based on the lowest AIC score. For continuous

characters, ancestral state was estimated with function contMap from the package phytools v.

0.5-64 (Revell, 2012). Character states for all examined traits were plotted on the dated

phylogeny to visualize patterns in the data.

Trait correlation

Correlation between discrete binary characters was analyzed with Pagel’s statistical

method (Pagel, 1994) implemented in phytools package in R (Revell, 2012). This method fits

two models to the data: a model in which traits evolve independently of each other, and a model

in which the evolution of two characters is correlated. A likelihood ratio test is then used to

assess if the difference between models is significant. Although it was designed to correct for

phylogenetic pseudoreplication, this method has been heavily criticized in the literature because

it often infers statistically significant correlations, even when the similarity in traits is an attribute

of common evolutionary origin (Maddison and FitzJohn, 2015). However, a more satisfactory

approach to eliminating pseudoreplication is yet to be put forward. The results of these tests

should be interpreted with caution, and we thus supplement statistical tests with visual inspection

of character state reconstructions mapped over the phylogeny to find biologically meaningful

associations. Test were carried out using fitPagel function with “fitMk” method for character

evolution, with “ARD” (all rates different) evolutionary model for each character. Two non-

binary characters (anther symmetry and pore size) were constrained to only two states per trait to

fit test requirements. I converted anther symmetry to a binary character by grouping

monosymmetric and asymmetric anthers into one category, because both character states may be

14

indicative of specialized pollination, as opposed to polysymmetric anthers. Pore size was

reclassified into minute pores versus large pores. I combined broad pores with rimose pores,

because neither of these character states requires specialized buzz-pollination.

Correlation between continuous traits was evaluated with regression analysis of

phylogenetically independent contrasts. Phylogenetically independent contrasts were calculated

following the comparative method introduced by Felsenstein (1985) to account for non-

independence of individual species within a phylogeny. The analysis was performed using pic

function in R package ape (Paradis et al. 2004).

Diversification analyses

I analyzed diversification in the tribe with Bayesian Analysis of Macroevolutionary

Mixtures (BAMM; Rabosky, 2014). BAMM models diversification rates with reversible-jump

Markov Chain Monte Carlo approach and is designed to detect rate heterogeneity and shifts in

evolutionary regimes within the system. I estimated a fine-scale sampling fraction based on

previously published monographs of the tribe and expert taxonomic knowledge (Renner 1993,

Goldenberg et al. 2008, Michelangeli et al. 2004 and 2008, Reginato et al. 2010, Reginato 2016;

Kriebel 2016, Martin et al. 2008, Michelangeli F.A., personal communication) to account for

non-random incomplete taxon sampling. Sampling probability was assigned to every species

based on the estimate of sampled diversity of each clade. Prior values for rate parameters were

scaled using setBAMMpriors function in package BAMMtools (Rabosky et al., 2014). Expected

number of shifts was set to 10, as recommended for large phylogenies, although we manipulated

different priors (1, 5, 10) and obtained similar results. I ran the analysis for 80 million

generations, sampling every 7000 generations. The first 10% of samples were discarded as burn-

15

in. The best rate shift configuration (sampled most frequently) was obtained with

getBestShiftConfiguration function.

To determine the effect of discrete trait evolution on diversification within the Miconieae,

and to estimate the rates of transition between character states, I implemented BiSSE (Binary

State Speciation and Extinction) approach pioneered by Maddison et al. (2007). BiSSE is a

likelihood-based approach that models trait evolution simultaneously with lineage diversification

within a phylogeny, and is a useful tool to test hypotheses about the effects of biological

characteristics of organisms on speciation and extinction rates (Maddison et al 2007, O’Meara

and Beaulieu 2016). Recent simulation studies have demonstrated that this approach has

alarming limitations, as it commonly attributes variation in diversification rates to an effect of a

character that in fact had no influence on evolution (Rabosky and Goldberg 2015, Beaulieu and

O’Meara 2016, Rabosky and Goldberg 2017). Beaulieu and O’Meara (2016) suggest that the

high “false positive” rate of BiSSE methods is the problem of falsely rejecting a trivial null

model in favor of a more complex but incorrect model.

I followed methods proposed by Beaulieu and O’Meara (2016) to generate a trivial null

BiSSE model as well as a more complex null HiSSE model (Hidden State Speciation and

Extinction) using function hisse from R package hisse (Beaulieu and O’Meara, 2016), and

estimate model fit with AIC. If trivial null BiSSE model had a better fit, I generated a suite of

eight BiSSE models of varying complexity using make.bisse function (R package diversitree;

FitzJohn, 2012). Model parameters (transition rates between states, speciation and extinction

rates for each character state) were allowed to vary between states, or were constrained to be

equal between states. Model fit was compared with AIC. I sampled parameter space of the best-

performing model with Marcov Chain Monte Carlo for 100 000 generations with an exponential

16

prior. This was done for every binary character in the morphological matrix. I also performed the

same series of analyses for categorical characters with more than two states: pore size and anther

symmetry. These characters were converted to binary according to their biological function

(broad and rimose pore were grouped together as opposed to minute pore; asymmetric and

monosymmetric anthers were grouped together as opposed to polysymmetric anthers).

Results

The phylogenetic relationships of 1167 species were reconstructed using maximum

likelihood to study the evolution of floral morphology (Figure S1). The phylogenetic tree was

used as input for all downstream analyses, but it was not the emphasis of my investigation. I do

not discuss taxonomic implications of these findings, but instead focus on implications for the

evolutionary history of morphological characters based on this phylogeny. Sampling spanned ten

clades following the taxonomic assignments of Goldenberg et al. (2008; see Table 4).

Examination of field photographs of over 300 Miconieae species has confirmed the

striking variability in floral traits for a tribe with conserved flower morphology (Appendix 2, Fig.

1). Petal and anther colors varied between white, yellow, and either pink or purple, which

allowed for many different combinations of colors across the tribe. Variability of shapes,

positions and sizes of reproductive organs allowed for various flower types, with variation in

forms of symmetry. While many species had herkogamous flowers, the same effect (separation

between stigma and anthers) was achieved through various combinations of anther and stigma

positions, shapes and lengths (compare Figs. 1F, J and L). Although the same suite of analyses

was performed for every encoded character, only the most interesting and relevant results are

presented here.

Ancestral state reconstructions and variation in floral morphology

17

Ancestral state reconstructions provided insight into the possible phenotype of the

ancestral flower of Miconieae. I infer that the most recent common ancestor of all Miconieae

likely had flowers with pink petals and pink, purple or blue anthers, as the likelihoods of these

states at the root of the phylogeny were greater than 50% (61% likelihood for pink petals and

55% for purple anthers; Fig. 2). Likelihoods of white being the ancestral state of either petals or

anthers were lower (white petals: 36%; white anthers: 44%). Likelihoods of yellow being the

ancestral state of either character approached 0 (yellow petals: 3%; yellow anthers: 1%). The

style of the ancestral flower was reconstructed as curved and slanted away from central axis with

likelihoods approaching 100% (curved style: 99%; slanted away from flower axis: 97%; Fig. 3).

Anthers of the ancestral flower probably had minute pores and were monosymmetrically

arranged (minute pore: 98%; monosymmetry: 99.94%; Fig. 4).

Visual inspection of character states mapped on the phylogeny along with ancestral state

reconstructions indicate that white petals are most common and evolved from pink several times

in different clades (Appendix 2, Fig. 2B). Although pink petals appear throughout the phylogeny

(clades Conostegia, Caribbean, Tococa, Centrodesma, Secundiflorae, some Clidemia), they

constitute only a quarter of the morphological dataset, and only two clades (Miconia thomasiana

clade within the Caribbean Miconieae, and Tococa) exhibit predominantly pink petals. White

petals appear in almost three quarters of all photographed species and are present in nearly every

clade within the phylogeny. Species within two large clades (Miconia and Cremanium) have

predominantly white petals. Yellow petals are very rare and appear in only six species (two

closely related species within Miconia, one in Caribbean, and three in Cremanium; Fig. 1H, P).

Yellow petals seem to have evolved from white.

18

Anther color appears to be a more labile character than petal color. Most of the clades

contain all three phenotypes, although pink, purple, or blue anthers are less frequent (Appendix

2, Fig. 2). White anthers are more common in Miconia (Fig. 1K) and Clidemia (Fig. 1A), while

nearly all species within Caribbean Miconieae and Conostegia have yellow anthers (Fig. 1J).

Ancestral state reconstructions suggest that white anthers evolve from pink or yellow, while

yellow anthers seem to descend from white.

Deviations from polysymmetry in gynoecium are frequent within the tribe (Fig. 3A and

B). Straight style evolved from curved style in several instances. Most of species within

Cremanium and Clidemia have flowers with a straight style (Fig. 3B). However, an upward style

is much less frequent, and only Cremanium clade contains mainly species with an upward style

(Fig. 3A). Most of the species in the phylogeny have styles that are directed away from the floral

axis, starting at the very base of the style where it connects to the ovary.

Androecium symmetry varies between polysymmetric, monosymmetric, and asymmetric;

polysymmetric androecia are less common than androecia with deviations from polysymmetry

(Fig. 4B). Asymmetric anthers appear in several species and seem to have evolved from the

monosymmetric character state. Polysymmetric anthers evolve from monosymmetric in many

clades, especially Cremanium and Leandra (within Clidemia; Fig. 1F, P).

Anther pore sizes vary between minute, broad and rimose (slits), with minute pores being

the most common (Appendix 2, Fig. 4A). Broad pores evolve within Miconia, but only appear in

a few species, while Cremanium contains mainly species with broad anther pores. Ancestral state

reconstructions indicate that rimose pore phenotype (slits) does not evolve from minute pores.

Flower size varies greatly within the tribe, with an evolutionary trend toward smaller

flowers occurring in several clades (Fig. 5). Petals of some species were only 1 mm in length,

19

while other species displayed large petals reaching a length of 20 mm. Anther sizes also fell

within a wide range, from 0.4 mm to 10 mm long. Small flowers evolve from large many times,

especially in Cremanium, Miconia and Leandra clades.

Trait correlation

Correlations between discrete characters were assessed with Pagel’s test for correlated

trait evolution and through visual inspection of character states mapped over phylogeny. Pore

size seems to be correlated with symmetry in both androecium and gynoecium, as well as with

contrast between petals and anthers of the flower. This is particularly evident when broad and

rimose pores are grouped together to convert pore size into a binary character (two character

states: minute and large pores). Large pores are more common in flowers with polysymmetric

anthers, straight upward style and lack of anther-petal contrast (especially when both anthers and

petals are white; Fig. 1K, P). These findings were supported by Pagel’s correlation test, which

showed significant correlation between all but two pairs of examined binary characters (Table 5).

Regression analysis of phylogenetically independent contrasts showed significant but

imperfect correlation between anther and petal sizes (p-value = 2.2*10-16, r2 = 0.506). Map of

continuous traits plotted against each other shows similar evolutionary trends towards size

reduction in both traits, but they are not precisely parallel (Fig. 5). In Clidemia and several

subclades of Miconia, petal size decreased with a simultaneous increase in anther length,

resulting in a peculiar flower phenotype with long, often colorful anthers and small, pale and

often reflexed petals (Fig. 1B, F, G, K).

Diversification analyses

Diversification rates within the entire tribe (1083 species) with one outgroup (4 species of

Eriocnemeae) were modeled with BAMM. There is a notable heterogeneity in diversification

20

rates through time and among different clades. The rate shift configuration with the maximum a

posteriori probability detects three diversification rate shifts (Fig. 6). The oldest shift occurs

approximately 20 mya, near the root of the Miconieae tribe, and denotes the split between the

Mirabilis clade (three samples in this phylogeny, out of nine known species) and the radiation of

the rest of Miconieae. Two other shifts happen in Cremanium circa 13 mya and in Leandra

(within Clidemia clade) 10 mya. Initial rapid diversification within these two clades gradually

slows down, but the current diversification rates remain higher than those of the rest of the tribe.

A suite of models generated for every trait under BiSSE (binary state speciation and

extinction) and HiSSE (hidden state speciation and extinction) scenarios provided evidence for

state-dependent diversification in two traits: anther symmetry and herkogamy. The best model

for anther symmetry has equal rates of speciation and extinction between two character states,

but the rates of transition from monosymmetric to polysymmetric anther arrangement are

significantly higher than the reverse (Fig. 7). The best model for herkogamy has equal extinction

rates for herkogamous and non-herkogamous flowers. While the rates of transition towards

herkogamy are significantly greater than the reverse (Fig. 8), plants with non-herkogamous

flowers seem to have higher speciation rates (Fig. 9).

Discussion

Examination of floral trait evolution using the largest species-level phylogeny of

Miconieae to date with morphological data on a variety of floral characteristics for over 350

species revealed rampant convergent and correlated evolution in most floral characters.

I detected correlated evolution of multiple characters indicative of pollinator-mediated selective

pressures. I confirm multiple parallel trends in evolution previously suggested for this

21

phylogenetically complex paraphyletic tribe (Goldenberg et al. 2008, Varassin et al. 2008, Brito

et al. 2016).

Floral morphological diversity

Miconieae is a species-rich tribe that exhibits a great floral diversity. I observed a great

amount of variation in individual floral traits: color and size of petals and anthers, shapes and

arrangement of anthers and style, shape and size of anther pores. Integrated within flowers, these

highly variable floral characters jointly produced a wide range of different flower morphologies.

Concurrent examination of evolutionary trends within individual characters informed my

understanding of the larger evolutionary patterns within the tribe.

Petal color is an important visual cue for both insect and bird pollinators (Lunau 2006,

Renner 1989) and helps infer pollination mode when viewed in combination with other

characters. Retention of pink petal color in some clades could indicate adherence to the

specialized buzz-pollination mode common in the family. Colorful petals attract bee pollinators

from a distance and provide a background for visual cues from anthers (Larson and Barrett 1999,

Lunau 2006). Moreover, some bee species exhibit spontaneous preference for yellow anthers

contrasted by a colorful rather than white corolla (Heuschen et al. 2005). Alternatively, pink

petals could indicate shifts to hummingbird pollination, especially when brightly colored pink or

deep-red petals form a tubular shape in a mature flower (Varassin et al. 2008). Bird pollination

has been observed in several nectariferous species in Miconieae (Miconia melanotricha,

Tetrazygia fadyenii, six species of Charianthus; Goldenberg et al 2008, Penneys and Judd 2003,

2005), three of which have been used in this study (M. melanotricha, T. fadyenii, C. purpureus:

see C. purpureus in Fig. 1C). The rarity of yellow flowers could be explained by selection

against a feature that involves costly production of yellow pigment while reducing efficiency of

22

bee pollination due to lack of anther-petal contrast. Out of six species coded as yellow, three are

not intensely colored (yellowish-green or white petals), and in two of them petals are short and

inconspicuous, suggesting that traces of yellow pigment do not participate in pollinator signaling.

Two closely related species in Cremanium clade (Miconia cernua: Fig. 1H; Miconia

corymbiformis: Fig. 1P) exhibit robust flowers with cucullate corollas, raising questions about

the possible pollinators of these species. Tetrazygia discolor also has an upward corolla and

appears to be closely related to the nectar-producing, hummingbird pollinated Charianthus clade.

Further investigation would be necessary to determine if T. discolor, M. cernua and M.

corymbiformis offer nectar rewards to attract vertebrate or generalist insect pollinators.

Loss of pink petal pigment and emergence of white corollas in Miconieae does not

necessarily indicate a trend towards generalization and should be interpreted jointly with other

floral characters. White petals may still provide a strong visual signal to bees, especially as a

backdrop to brightly colored anthers (Lunau 2006). In single flowers of understory plants, white

petal color may confer a selective advantage by making flowers more visible to pollinators

(Renner 1989). I also observed some striking floral displays created by white petals of small

flowers arranged in inflorescences. This suggests that these white petals serve as strong visual

cues to pollinators, although I did not include floral arrangement as one of the coded characters

in my analyses. Additionally, evolution of white petals from pink could be explained by reduced

role of petals in pollinator signaling in floral phenotype with small reflexed corollas and

conspicuous anthers (Fig. 1B, G, K).

Conversely, emergence of white petals could be evolutionarily favored in generalist

species catering to a wide range of pollinators, especially when an alternative reward such as

nectar is offered. In multiple documented cases of generalized pollination systems in Miconieae,

23

when a nectar reward supplements or substitutes pollen reward, petal color of the flower is white

(Miconia tonduzii and M. brevitheca (Kriebel and Zumbado 2014), M. hyemalis (Varassin et al.

2008), M. theizans (Brito et al. 2017), M. pepericarpa (Goldenberg and Shepherd 1998)). It

appears that in plant species with this pollination mode bright petal pigment no longer functions

in pollinator attraction and is selected against as an unnecessary expense. I thus conclude that

numerous instances of convergent evolution of white petals could be indicative of either

pollinator generalization or specialization, and are informative of shifts in pollination only when

analyzed collectively with other floral characters.

I attribute anther color variability and multiple historic changes in this trait to the role of

anthers in floral signaling. In tube-like poricidal anthers of Miconieae, the pollen is not visible to

the pollinator, and the anther itself assumes the function of a signaling device in species that

attract pollinators with a pollen reward (Lunau 2006). When the yellow color of pollen is

mimicked by the anther, it becomes an orientation signal to insect pollinators such as pollen-

collecting bees and pollen-eating flies. Bumblebees have been reported to instinctively respond

to colors other than yellow and to demonstrate learning behavior (Lunau 2006). In addition,

removal of yellow anthers, stamens or stamen appendages in species of Melastomataceae

resulted in reduction of visitation by bees (Larson and Barrett 1999, Luo 2008). As a short-

distance signaling cue for pollinators, anther color is an important and more reliable predictor of

the pollination mode than the petal color. Greater variability in anther color phenotype as

compared to petal color phenotype could be indicative of a greater role of anthers in floral

signaling in this tribe, and thus of stronger selective pressure. I consequently interpret emergence

of yellow petals in every major clade of Miconieae as a consequence of specialized pollinator

selection, whereas repeated evolution of white anthers might be indicative of generalized

24

pollination. It must be noted that species with recent switches to generalized pollination could

retain that feature due to phylogenetic inertia if they are nested within yellow-anthered clades.

Nearly identical patterns of historic diversification in androecium and gynoecium

symmetry demonstrate the integrated nature of floral function. Variations in anther and stamen

symmetry combined with shape and direction of style help produce diverse floral phenotypes

which can be interpreted from two related viewpoints: pollinator fit and herkogamy. During

vibrational action of the flight muscles of the bees the pollen is expelled from the anther and

deposited on the abdomen of the bee. The bee abdomen then touches the stigma of the next

visited flower, achieving successful transfer of pollen (Luo et al. 2008, Larson and Barrett 1999).

Deposition of pollen on a bee abdomen would reduce the amount of pollen reaching the stigma

of the flower, thus reducing the possibility of self-fertilization. In such specialized flowers,

positions and shapes of style and anthers would be the evolutionary result of adaptation to the

most effective pollinator species and could thus be interpreted as an indicator of specialized

pollination mode. However, these organs also play a role in achieving herkogamy, which is the

main mode of outcrossing in Melastomataceae (Renner 1989). While most of the examined

flowers exhibited some kind of herkogamy, this was often achieved with an exserted stigma on a

long or slanted style, while the anthers remained polysymmetric. Reduction of herkogamy has

been reported in species of Miconieae with generalized pollination syndromes (Miconia tonduzii

and M. brevitheca; Kriebel and Zumbado 2014). I conclude that evolutionary patterns of

androecium and gynoecium should be considered in unison to inform hypotheses about shifts in

pollinator modes. I detect several evolutionary shifts towards generalization in clades where

evolution of polysymmetric anthers co-occurred with emergence of a straight upward style.

25

I detected several instances of evolution of broad and slit-like pores within Miconieae,

and I interpret such pore broadening as an evolutionary trend towards generalization based on the

functional differences in pollen extraction from anthers with large pores as opposed to anthers

with minute pores. Convergent evolution of broad pores accessible to non-buzzing pollinators in

Miconia, Mecranium and Cremanium is consistent with previous findings by Brito et al. (2016).

Increase in the pore size is thought to have adaptive value, because it seems to expand the

number of pollinator species without excluding buzzing bees (Goldenberg et al. 2008. Varassin

et al. 2008, Kriebel and Zumbado 2014). Adaptive nature of generalist trends in pore size is

further supported by the findings of Larson and Barrett (1999), who report non-vibrational

expulsion of pollen from minute anther pores in marginal populations of Rhexia virginica

(Melastomataceae) in Ontario, where buzzing bee pollination may be less reliable than at the

center of the species’ range.

I was also able to confirm the previously reported convergence towards floral size

reduction in Miconieae (Goldenberg et al. 2008, Brito et al. 2016), specifically in Cremanium

and Miconia clades. These parallel trends seem to indicate generalization, even if they are still

restricted to bee pollination. In addition to large, buzzing bees that can vibrate and pollinate the

entire inflorescences of small flowers, such flowers can also be pollinated by buzzing bees that

are too small to be effective pollinators of larger flowers (Goldenberg et al. 2008). Species with

dense inflorescences of small flowers with broad pores can be pollinated by a variety of non-

buzzing insects, and warrant further inquiry to determine if evolution of nectar reward parallels

the reduction in size. Although I did not consider floral grouping in my analyses, arrangement of

flowers in large inflorescences demonstrates an additional advantage to size reduction in species

that present pollinators with a “big bang” floral display (Kriebel and Zumbado 2014). A

26

flowering strategy with many flowers open at once attracts pollinators while eliminating the need

for production of petal pigment, and demonstrates the need to take flower grouping into account

when analyzing pollinator-mediated evolution of flower characteristics. I attribute the mismatch

in the evolutionary trends in anther and petal length of some Leandra and Miconia species to a

reduced role of petals in floral signaling, with anthers assuming the function of pollinator

attraction and guidance.

Ancestral flower reconstruction

The reconstructed phenotype of the ancestral flower is consistent with the floral features

commonly observed in Melastomataceae (Renner 1989). I conclude that the ancestral flower was

probably medium-sized compared to most Miconieae flowers and had pink petals and anthers

(see Fig. 1N for this color combination). It likely displayed monosymmetry in both androecium

and gynoecium (see Fig. 1L for an example of such arrangement), with a curved style directed

away from the center of the flower accompanied by monosymmetric stamens with minute pores.

Pink petals are often observed in multiple species in the family, as are pink anthers, although

yellow anther color might be a more widespread feature (Larson and Barrett 1999). Floral

phenotype with colorful petals and anthers, monosymmetry in androecium and gynoecium

resulting in herkogamy and, most of all, minute pore is indicative of specialized buzz-pollination

which is the main mode of pollination in the family (Renner 1989, Buchman 1983). Although

such flower type is an entirely plausible ancestor of Miconieae tribe, the results of the

reconstructions should be interpreted with caution, because I did not estimate uncertainty during

these analyses.

Trait correlation

27

The highly significant correlation among floral traits is the logical consequence of the

integration of functional units within the flower. Inferring the taxon of a flower’s pollinator is

quite difficult using only a single character within my morphological framework. Even the most

distinctive character, pore size, would not help distinguishing between a highly specialized

(hummingbird) or a generalized pollination system, unless other floral characters are considered.

I report repeated emergence of a (suspected) generalist flower type with white petals,

polysymmetric white anthers and straight upward style in several species of Miconia and a large

number of species in Cremanium. Further fieldwork would be necessary to determine the

pollinators of these species.

Diversification analyses

The oldest shift in diversification dynamics detected with BAMM corresponds to the

diversification burst at the base of Miconieae clade first reported by Judd and Skean (1991) and

may be the reason for difficulties in morphological clade delimitation (Goldenberg et al. 2008;

but see Berger et al. 2016 for an unexpected alternative shift configuration). Two later shifts

occur in Leandra (within Clidemia) and Cremanium, two clades that share some floral

characteristics. Species in both clades exhibit predominantly polysymmetric anthers, white

petals, straight styles and trends towards size reduction (Figs. 2B, 3B, 4B, 5). However, the

majority of species in Leandra feature styles that are directed away from the flower center and

minute anther pores. Only one species on Leandra was coded to have broad pores (L.

ionopogon), and this assessment was based on a photograph of a single old flower, which may

have had damaged anthers. On the other hand, Cremanium species commonly have straight

styles and broad or rimose pores (Cremanium clade contains all but four species with rimose

pores assessed in this study). In addition to flower morphology, both Leandra and Cremanium

28

are distinct from the rest of the tribe, because the representatives of both clades are commonly

found at higher elevations than other Miconieae species. Most Leandra species are found in

eastern Brazil, at higher elevations than the rest of Miconieae (500-1000 m; Reginato 2016).

Species of Cremanium clade are found at even greater elevations in the Greater Antilles and the

Andes (Michelangeli et al. 2008, Goldenberg et al. 2008).

These increases in diversification rates concurrent with parallel patterns of floral

evolution within the clades suggest that ecology plays an important role in shaping species

diversification and pollinator interactions in the tribe. Heterogeneity of environmental conditions

in elevated habitats provides ecological opportunities that often lead to increased diversification

in plants (Hughes and Atchison 2014). Montane conditions are also associated with shifts in

pollinator mode due to the unpredictability of pollinators at higher elevations (Varassin et al.

2008, Cruden 1972, Dellinger et al. 2014). Such shifts can be related to specialized vertebrate

pollination, but it appears that shifts in Miconieae pollination strategies may also involve a

transition from specialized bee pollination to generalist pollination in elevated environments

where buzz-pollination in not reliable (Varassin et al. 2008, Brito et al. 2016). One should be

careful not to interpret the increased diversification rates in Leandra and Cremanium as the

consequence of evolution of polysymmetric anthers, white petals and straight styles in these

clades. Rather, these evolutionary trends in morphology and the rapid diversification are both the

products of the heterogeneity in environmental conditions observed in elevated habitats. This

observed association warrants a further investigation into the biogeography of these clades.

BiSSE analyses help detect interesting dynamics in Miconieae diversification with

relation to generalization of pollination modes. The faster rate of transition towards

polysymmetric anthers is likely the consequence of multiple trends towards generalization,

29

because anther polysymmetry is highly correlated with generalized pollination in my analyses as

well as in previous observations of such species (Kriebel and Zumbado 2014, Varassin et al.

2008, Brito et al. 2017, Goldenberg and Shepherd 1998). Faster transition towards herkogamy

corresponds to evolutionary advantages of outcrossing. Unexpectedly higher speciation rate in

non-herkogamous species warrants further investigation. I propose two scenarios in which loss

of herkogamy could lead to increased speciation: colonization of new environments or evolution

of physiological barriers to self-pollination. In the first scenario, plant species that are capable of

self-fertilization have adaptive advantages over species with obligate specialized pollinators that

may be unavailable during range expansion (Toräng et al., 2017). Species colonizing species

poor habitats devoid of specialized pollinators, such as newly emerged mountain ranges, islands,

or disturbed habitats may be more successful if they are able to reproduce asexually or through

self-pollination (Sandvik et al. 1999, Kriebel et al. 2015). Such colonization and subsequent

range expansion may be followed by speciation bursts due to environmental filters or emergence

of biogeographic barriers. In the second scenario, physiological self-incompatibility would

eliminate the need for herkogamy, while such generalized non-herkogamous phenotype would

increase the diversity of pollinators, resulting in a greater reproductive success. Self-

incompatibility has been documented in Miconieae, and Goldenberg and Shepherd (1998)

suggest that nearly a quarter of all Miconieae could be self-incompatible, but the current

knowledge of the issue in such a species-rich tribe is very limited.

Conclusions

My examination of the floral morphology in the Miconieae confirms that multiple traits

within the flowers evolve in a unified fashion due to the integrated functioning of these

angiosperm flowers. Correlated evolution of multiple floral traits emphasizes the need for

30

comprehensive trait evaluation in studies that focus on the processes that affect reproductive trait

morphology. I detect analogous trends towards generalized pollination in a range of characters

related to size, color, shape and disposition of reproductive structures within the Miconieae

flowers. I find an association between generalization trends and increased diversification rates

that may be related to adaptations to highland environments within the tribe.

Future directions

This study informs several future research directions within this study system. Future

fieldwork is necessary to investigate various aspects of pollination strategies employed by plants

that exhibit generalization trends in floral morphology. The tribe Miconieae is data-deficient in

knowledge of pollinator associations and self-compatibility. I also recommend further field work

to determine if plants that feature generalist trends in floral morphology (polysymmetry, loss of

pigment, pore broadening) lure pollinators with alternate rewards such as nectar or other

attractive cues such as scent. I also recommend considering other aspects of floral displays, for

instance floral grouping, in future analyses. It seems likely that individual characteristics of

flowers arranged into inflorescences are not equally informative for pollinator-imposed

evolutionary pressures to those of single flowers, because the entire inflorescence may act as a

floral guide in pollinator attraction.

In addition, further investigation into the role of apomixis in floral evolution is necessary.

Apomixis is an asexual seed reproduction strategy that is common in Miconieae (Maia et al.

2016, Goldenberg and Shepherd 1998), but I did not address the impact of this reproductive

mechanism in my analyses. Pollen sterility often observed in such species affects the pollinator

assemblages that visit these plants, and in obligatory apomicts that are not visited by pollinators

the evolutionary pressures in floral morphology remain unclear, especially because several

31

apomictic species assessed in this study still display herkogamy and/or anther pigmentation

(Leandra australis, L. purpurascens, Miconia albicans: Maia et al. 2015; Miconia stenostachya,

M. fallax: Goldenberg and Shepherd 1998).

Acknowledgements

I would like to thank the following people who supported my work on this thesis:

Dr. Fabián Michelangeli for being a great research advisor, funding my graduate research and

education, sharing his expertise in Melastomataceae and providing guidance and corrections

during the writing of the thesis;

Dr. Amy Berkov for being my graduate advisor and providing insights and comments during

thesis development;

Dr. David Lohman for being my graduate committee member and going through this manuscript;

Marcelo Reginato for providing guidance and assistance with all major analyses, helping me

develop necessary software and analytical skills and acquire data used in this research;

Ricardo Kriebel for sharing morphological data that expanded my dataset;

Asuncion Cano, Diego Paredes and Lizeth Cardenas for providing logistical support and help

during fieldwork;

Mathew Sewell for help with laboratory work at the New York Botanical Garden;

Staff at Herbarium HOXA and Herbarium USM for logistical help in the field.

References

32

Almeda, F. (2009). Melastomataceae. In Davidse, G., Sousa, S., Knapp, M.S., Chiang, S.Y.F.

(ed.), Flora Mesoamericana (Volume 4, part 1). Universidad Nacional Autónoma de

México, Ciudad Universitaria, Mexico.

Anderson, G.J., Johnson, S., Neal, P., Bernardello, G. (2002). Reproductive biology and plant

systematics: the growth of a symbiotic association. Taxon, 51: 637-653.

Beaulieu, J.M., O’Meara, B.C. (2016) Detecting hidden diversification shifts in models of trait-

dependent speciation and extinction. Systematic Biology, 65: 583-601.

Berger, B.A., Kriebel, R., Spalink, D., Sytsma, K.J. (2016). Divergence times, historical

biogeography, and shifts in speciation rates of Myrtales. Molecular Phylogenetics and

Evolution, 95: 116-136.

Brito, V.L.G., Fendrich, T.G., Smidt, E.C., Varassin, I.G., Goldenberg, R. (2016). Shifts from

specialised to generalised pollination systems in Miconieae (Melastomataceae) and their

relation with anther morphology and seed number. Plant Biology, 4: 585-593.

Brito, V.L.G., Rech, A.R., Ollerton, J., Sazima, R. (2017). Nectar production, reproductive

success and the evolution of generalised pollination within a specialised pollen-rewarding

plant family: a case study using Miconia theizans. Plant Systematics and Evolution, 303:

709-718

Buchmann, S.L. (1983) Buzz pollination in Angiosperms. Handbook of experimental pollination

biology. Van Nostrand Reinhold, New York, NY, USA.

Cruden, R.W. (1972). Pollinators in high-elevation ecosystems: relative effectiveness of birds

and bees. Science, 176: 1439-1440.

33

Dellinger, A.S., Penneys, D.S., Staedler, Y.M., Fragner, L., Weckwerth, W., Schönenberger, J.

(2014). A specialized bird pollination system with a bellows mechanism for pollen transfer

and staminal food body rewards. Current Biology, 24: 1615-1619.

Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A. (2012). Bayesian phylogenetics with

BEAUti and the BEAST 1.7. Molecular Biology and Evolution, 29: 1969-1973

Givnish, T.J. (2010). Ecology of plant speciation. TAXON, 59: 1326-1366.

Goldenberg, R., Penneys, D.S., Almeda, F., Judd, W.S., Michelangeli, F.A. (2008). Phylogeny

of Miconia (Melatomataceae): patterns of stamen diversification in a megadiverse

neotropical genus. International Journal of Plant Science, 169: 963-979.

Ellison, A.M., Denslow, J.S., Loiselle, B.E., Brenes, D.M. (1993). Seed and seedling ecology of

neotropical Melastomataceae. Ecology, 74: 1733-1749.

Endress, P.K. (1999). Symmetry in flowers: diversity and evolution. International Journal of

Plant Sciences, 160: S3-S23.

Faegri, K. and van der Pijl, L. (1979). Principles of Pollination Ecology. Pergamon Press,

Oxford.

Felsenstein, J. (1985). Phylogenies and the comparative method. The American Naturalist, 125:

1-15.

Fenster, C.B., Armbruster, W.S., Wilson, P., Dudash, M.R., Thomson, J.D. (2004). Pollination

syndromes and floral specialization. Annual Review of Ecology, Evolution and Systematics,

35: 375-403.

FitzJohn, R.G. (2012). Diversitree: comparative phylogenetic analyses of diversification in R.

Methods in Ecology and Evolution, 3: 1084-1092.

34

Gamba, D., Almeda, F. (2014). Systematics of the Octopleura clade of Miconia

(Melastomataceae: Miconieae) in tropical America. Phytotaxa, 179: 001-174.

Goldenberg, R., Amorim, A.M. (2006). Physeterostemon (Melastomataceae): A new genus and

two new species from the Bahian Atlantic forest, Brazil. Taxon, 55: 965-972.

Goldenberg, R., Penneys, D., Almeda, F., Judd, W., Michelangeli, F. (2008). Phylogeny of

Miconia (Melastomataceae): patterns of stamen diversification in a megadiverse Neotropical

genus. International Journal of Plant Sciences, 169: 963-979.

Goldenberg, R., Shepherd, G.J. (1998). Studies on the reproductive biology of Melastomataceae

in "cerrado" vegetation. Plant Systematics and Evolution, 211: 13-29.

Heuschen, B., Gumbert, A., Lunau, K. (2005). A generalised mimicry system involving

angiosperm flower colour, pollen and bumblebees' innate colour preferences. Plant

Systematics and Evolution, 252: 121-137.

Hickey, J.L. (1977). Stratigraphy and paleobotany of the Golden Valley Formation (early

Tertiary) of Western North Dakota. Geological Society of America, Memoir 150, Boulder,

Colorado, USA.

Hughes, C.E., Atchison, G.W. (2015). The ubiquity of alpine plant radiations: from the Andes to

the Hengduan Mountains. New Phytologist, 207: 275-82.

Judd, W.S. (2007). Revision of Miconia Sect. Chaenopleura (Miconieae, Melastomataceae) in

the Greater Antilles. Systematic Botany Monographs, 81: 1-235.

Judd, W.S., Bécquer, E.R., Majure, L.C. (2014a). Taxonomic studies in the Miconieae

(Melastomataceae). XI. A revision of Miconia sect. Calycopteris on Hispaniola. Brittonia,

66: 216-249.

35

Judd, W.S., Bécquer, E.R., Skean Jr., J.D., Majure, L.C. (2014b). Taxonomic studies in the

Miconieae (Melastomataceae). XII. Revision of Miconia sect. Miconiastrum, with emphasis

on the Miconia bicolor complex. Journal of the Botanical Research Institute of Texas, 8:

457-491.

Judd, W.S., Ionta, G.M. (2013). Taxonomic studies in the Miconieae (Melastomataceae). X.

Revision of the species of the Miconia crotonifolia complex. Brittonia, 65: 66-95.

Judd, W.S., Majure, L.C., Ionta, G.M., Neubig, K.M. (2015). Taxonomic studies in the

Miconieae (Melastomataceae). XIII. Systematics of Miconia subcompressa, a Hispaniolan

endemic comprised of three eco-geographic subspecies. Phytotaxa, 197: 061-083.

Judd W.S., Skean Jr., J.D. (1991). Taxonomic studies in the Miconieae (Melastomataceae). IV.

Generic realignments among terminal-flowered taxa. Florida Museum of Natural History,

Biological Sciences, 36: 25-84.

Kearse, M., Moir, R., Wilson, A., Stones-Havas, S., Cheung, M., Sturrock, S., Buxton, S.,

Cooper, A., Markowitz, S., Duran, C., Thierer, T., Ashton, B., Mentjies, P., Drummond, A.

(2012). Geneious Basic: an integrated and extendable desktop software platform for the

organization and analysis of sequence data. Bioinformatics, 28: 1647-1649.

Kriebel, R. (n.d.) Melastomataceae de Centroamérica. https://melas-centroamerica.com.

Kriebel, R. (2016). A monograph of Conostegia (Melastomataceae, Miconieae). PhytoKeys, 67:

1-326.

Kriebel, R., Michelangeli, F.A., Kelly, L.M. (2015). Discovery of unusual anatomical and

continuous characters in the evolutionary history of Conostegia (Miconieae:

Melastomataceae). Molecular Phylogenetics and Evolution, 82: 289-313.

36

Kriebel, R., Zumbado, M.A. (2014). New reports of generalist insect visitation to flowers of

species of Miconia (Miconieae: Melastomataceae) and their evolutionary implications.

Brittonia, 66: 396-404.

Larson, B.M.H., Barrett, S.C.H. (1999). The pollination ecology of buzz-pollinated Rhexia

virginica (Melastomataceae). American Journal of Botany, 86: 502-511.

Liogier, A.H. (2000). La flora de la Española. Instituto Tecnológico de Santo Domingo, Santo

Domingo, República Dominicana.

Lumer, C. (1980). Rodent pollination of Blakea (Melastomataceae) in a Costa Rican cloud

forest. Brittonia, 32: 512-517.

Lunau, K. (2006). Stamens and mimic stamens as components of floral colour patterns.

Botanische Jahrbücher, 127: 13-41

Luo, Z.L., Gu, L., Zhang, D.X. (2009). Intrafloral differentiation of stamens in heterantherous

flowers. Journal of Systematics and Evolution, 47: 43-56.

Luo, Z., Zhang, D., Renner, S.S. (2008). Why two kinds of stamens in buzz-pollinated flowers?

Experimental support for Darwin's division-of-labour hypothesis. Functional Ecology, 22:

794-800.

Maddison, W.P., FitzJohn, R.G. (2015). The unsolved challenge to phylogenetic correlation tests

for categorical characters. Systematic Biology, 64: 127-136.

Maddison, W.P., Midford, P.E., Otto, S.P (2007). Estimating a binary character's effect on

speciation and extinction. Systematic Biology, 56: 701-710.

Maia, F.R., Varassin, I.G., Goldenberg, R. (2016). Apomixis does not affect visitation to flowers

of Melastomataceae, but pollen sterility does. Plant Biology, 18: 132-138.

37

Majure, L.C., Judd, W.S., Michelangeli, F.A. (2015). Taxonomic revision of the Greater

Antillean Pseudolima clade of Miconia (Miconia sect. Krugiophytum: Miconieae:

Melastomataceae). Brittonia, 67: 11-28.

Majure, L.C., Neubig, K.M., Skean, J.D., Bécquer, E.R., Judd, W.S., editor: Alan W. Meerow.

(2015). Evolution of the sandpaper clade (Miconieae, Melastomataceae). International

Journal of Plant Sciences, 176: 607–626.

Martin, C.V., Little, D.P., Goldenberg, R., Michelangeli, F.A. (2008). A phylogenetic evaluation

of Leandra (Miconieae, Melastomataceae): a polyphyletic genus where the seeds tell the

story, not the petals. Cladistics, 24: 315-327.

Michelangeli, F.A. (2005). Tococa (Melastomataceae). Flora Neotropica Monographs, 98: 1-

114.

Michelangeli, F.A., Almeda, F., Goldenberg, R., Judd, W.S., Becquer-Granados, E.R, Tulig, M.

(2009 onward). PBI Miconieae: A complete web-based monograph of the tribe Miconieae

(Melastomataceae). The New York Botanical Garden, Bronx, NY.

http://sweetgum.nybg.org/melastomataceae

Michelangeli, F.A., Judd, W.S., Penneys, D.S., Skean, J.D., Bécquer-Granados, E.R.,

Goldenberg, R., Martin, C.V. (2008). Multiple events of dispersal and radiation of the tribe

Miconieae (Melastomataceae) in the Caribbean. The Botanical Review, 74: 53-77.

Michelangeli, F.A., Penneys, D.S., Giza, J., Soltis, D., Hils, M.H., Skean Jr., J.D. (2004). A

preliminary phylogeny of the tribe Miconieae (Melastomataceae) based on nrITS sequence

data and its implications on inflorescence position. TAXON, 53: 279-290.

38

Miller, M.A., Pfeiffer, W., Schwartz, T. (2010). Creating the CIPRES Science Gateway for

inference of large phylogenetic trees. Proceedings of the Gateway Computing Environments

Workshop, 14 Nov. 2010, New Orleans, LA pp 1 - 8.

O’Meara, B.C., Beaulieu, J.M. (2016). Past, future, and present of state-dependent models of

diversification. American Journal of Botany, 103: 792-795.

Paradis, E., Claude, J., Strimmer, K. (2004). APE: analyses of phylogenetics and evolution in R

language. Bioinformatics, 20: 289-290.

Penneys, D.S., Judd, W.S. (2003). The resurrection and lectotypification of Tetrazygia fadyenii

(Melastomatacaeae: Miconieae): a hummingbird pollinated treelet endemic to Jamaica. Sida,

20: 877-884.

Penneys, D.S., Judd, W.S. (2005). A Systematic revision and cladistic analysis of Charianthus

(Melastomataceae) using morphological and molecular characters. Systematic Botany, 30:

559-584

Pagel, M. (1994). Detecting correlated evolution on phylogenies: a general method for the

comparative analysis of discrete characters. Proceedings of the Royal Society B, 255: 37-45.

Paradis, E., Claude, J., Strimmer, K. (2004). APE: analyses of phylogenetics and evolution in R

language. Bioinformatics, 20: 289-290.

Proctor, M., Yeo, P., Lack, A. (1996). The natural history of pollination. Timber Press, Portland.

Rabosky, D.L. (2014). Automatic detection of key innovations, rate shifts, and diversity-

dependence on phylogenetic trees. PLOS ONE, 9: e89543.

Rabosky, D.L., Goldberg, E.E. (2015). Model inadequacy and mistaken inferences of trait-

dependent speciation. Systematic Biology, 64: 340-355.

39

Rabosky, D.L., Goldberg, E.E. (2017). FiSSE: a simple non-parametric test for the effects of a

binary character on lineage diversification rates. Evolution, 71: 1558-5646.

Rabosky, D.L., Grundler, M., Anderson, C., Title, P., Shi, J.J., Brown, J.W., Huang, H., Larson,

J.G. (2014). BAMMtools: an R package for the analysis of evolutionary dynamics on

phylogenetic trees. Methods in Ecology and Evolution, 5: 701-707.

Reginato, M. (2016). Taxonomic revision of Leandra sect. Leandra (Melastomataceae,

Miconieae). Phytotaxa, 262: 001-097.

Reginato, M., Michelangeli, F.A., Goldenberg, R. (2010). Phylogeny of Pleiochiton

(Melastomataceae, Miconieae): total evidence. Botanical Journal of the Linnean Society,

162: 423-434.

Renner, S.S. (1989). A survey of reproductive biology in Neotropical Melastomataceae and

Memecylaceae. Annals of the Missouri Botanical Garden, 76: 496-518.

Renner, S.S. (1993). Phylogeny and classification of the Melastomataceae and Memecylaceae.

Nordic Journal of Botany, 13: 519-540.

Revell, L.J. (2012) phytools: An R package for phylogenetic comparative biology (and other

things). Methods in Ecology and Evolution, 3: 217-223.

Rosas-Guerrero, V., Aguilar, R., Marten-Rodriguez, S., Ashworth, L., Lopezaraiza-Mikel, M.,

Bastida, J.M., Quesada, M. (2014). A quantitative review of pollination syndromes: do

floral traits predict effective pollinators? Ecology Letters, 17: 388-400.

Sandvik, S.M., Totland, O., Nyléhn, J. (1999). Breeding system and effects of plant size and

flowering time on reproductive success in the alpine herb Saxifraga stellaris. Arctic,

Antarctic, and Alpine Research, 31:196-201.

Sequencher version 5.4.6 DNA sequence analysis software, Gene Codes Corporation, Ann

40

Arbor, MI USA. http://www.genecodes.com

Shaw, J., Lickey, E.B., Beck, J.T., Farmer, S.B., Liu, W., Miller, J., Siripun, K.C., Winder, C.T.,

Schilling, E.E., Small, R.L. (2005). The tortoise and the hare II: relative utility of 21

noncoding chloroplast DNA sequences for phylogenetic analysis. American Journal of

Botany, 92: 142-166.

Skean Jr., J.D. (1993). Monograph of Mecranium (Melastomataceae-Miconieae). Systematic

Botany Monographs, 39: 1-116.

Toräng, P., Vikström, L., Wunder, J., Wötzel, S., Coupland, G., Ågren, J. (2017). Evolution of

the selfing syndrome: anther orientation and herkogamy together determine reproductive

assurance in a self-compatible plant. Evolution, 71:2206-2218.

Varassin, I.G., Penneys, D.S., Michelangeli, F.A. (2008). Comparative anatomy and morphology

of nectar-producing Melastomataceae. Annals of Botany, 102: 899-909.

Wurdack, J.J. (1973). Melastomataceae. In Lasser (ed.) Flora de Venezuela (No. 8). Instituto

Botánico, Ministerio de Agricultura y Cría, Caracas, Venezuela.

Wurdack, J.J. (1980) Melastomataceae. In Harling, Sparre (ed.) Flora of Ecuador (No. 13).

University of Göteborg and Riksmuseum, Stockholm.

Xper3 (n.d.) Laboratory of Informatics and Systematics, University Pierre et Marie Curie, Paris

6. http://www.xper3.fr

41

Table 1. Summary of taxon sampling (outgroup).

Genus Number of species

Meriania 28

Graffenrieda 22

Adelobotrys 10

Axinaea 9

Macrocentrum 9

Physeterostemon 3

Salpinga 2

Eriocnema 1

Table 2. A summary of taxon sampling (Miconieae).

Genus Number of Species

Miconia 589

Leandra 140

Clidemia 123

Ossaea 66

Conostegia 35

Tococa 29

Mecranium 22

Calycogonium 20

Tetrazygia 19

Pachyanthus 16

Pleiochiton 8

Charianthus 6

Sagraea 4

Killipia 2

Maieta 2

Anaectocalyx 1

Necramium 1

42

Table 3. Polymerase chain reaction protocols for gene amplification.

Marker Primers Reaction mixture Thermocycler protocol

nrITS1-

nrITS2

NY183

NY207

(Michelangeli et al.

2004).

0.7ul template DNA

0.75ul spermidine

2.05ul water,

2ul forward primer

2ul reverse primer

0.5ul of Green Taq

Denaturing: 2 min (94°C)

40 cycles: 30 sec (94˚C)

30 sec (52˚C)

40 sec (72˚C)

Final extension: 7 min (72˚C)

nrETS NY 1428

NY 320

(Nicolas,

unpublished; Kriebel

et al. 2015)

0.7ul template DNA

0.75ul spermidine

2.05ul water,

2ul forward primer

2ul reverse primer

7.5ul of Green Taq

Denaturing: 2 min (94°C)

40 cycles: 30 sec (94˚C)

37 sec (55˚C)

40 sec (72˚C)

Final extension: 7 min (72˚C)

psbKL NY 824

NY 825

(Reginato et al.

2010).

0.7ul template DNA

0.75ul spermidine

2.05ul water,

2ul forward primer

2ul reverse primer

7.5ul of Green Taq

Denaturing: 2 min (94°C)

40 cycles: 30 sec (94˚C)

30 sec (53˚C)

30 sec (72˚C)

Final extension: 7 min (72˚C)

accD NY826

NY827

(Shaw et al. 2005)

0.7ul template DNA

0.75ul spermidine

2.05ul water,

2ul forward primer

2ul reverse primer

7.5ul of Green Taq

Denaturing: 2 min (94°C)

40 cycles: 30 sec (94˚C)

30 sec (52˚C)

40 sec (72˚C)

Final extension: 7 min (72˚C)

Table 4. Sampling fraction of Miconieae. The number of sampled species is provided along with

the total number of described species in each clade.

Clade Number of species Sampling fraction

sampled total recognized

Miconia 267 396 0.67

Clidemia 243 414 0.59

Cremanium 242 544 0.44

Caribbean 118 196 0.60

Conostegia 99 130 0.76

Mecranium 58 94 0.62

Tococa 23 38 0.61

Secundiflorae 19 25 0.76

Centrodesma 11 17 0.65

Mirabilis 3 9 0.33

43

Table 5. Results of Pagel’s test for correlated evolution of discrete characters. Numbers are p-

values. Significant results are shown in bold.

Pore size Contrast Anther

symmetry

Style

curvature

Style

direction

Herkogamy

Pore size 7.79*10-07 0.00025 1

2.28*10-17

1.05*10-05

Contrast 7.79*10-07 0.00083 0.0104 1.36*10-06

0.937

44

Figure 1. Floral diversity within Miconieae exemplified by species with variable floral traits. A:

Clidemia capitellata. B: Clidemia ciliata. C: Charianthus purpureus. D: Maieta poeppigii. E:

Anaectocalyx bracteosa. F: Leandra australis. G: Miconia ampla. H: Miconia cernua. I:

Mecranium haemanthum. J: Conostegia macrantha. K: Miconia affinis. L: Pachyanthus

pedicellatus. M: Miconia cerasiflora. N: Miconia aureoides. O: Miconia brachycalyx. P:

Miconia corymbiformis. Scale bars = 5mm.

45

Figure 2. Ancestral state reconstructions of anther (A) and petal (B) color in Miconieae. Character states of the terminal branches are

at the tips. Node labels show likelihoods of ancestral states. Taxonomic assignment of species is indicated by arches with clade

numbers: (1) Cremanium; (2) Caribbean; (3) Conostegia; (4) Mecranium; (5) Miconia; (6) Clidemia; (7) Tococa + Secundiflorae +

Centrodesma.

46

Figure 3. Ancestral state reconstructions of style direction (A) and curvature (B) in Miconieae. Character states of the terminal

branches are at the tips. Node labels show likelihoods of ancestral states. Taxonomic assignment of species is indicated by arches with

clade numbers: (1) Cremanium; (2) Caribbean; (3) Conostegia; (4) Mecranium; (5) Miconia; (6) Clidemia; (7) Tococa + Secundiflorae

+ Centrodesma.

47

Figure 4. Ancestral state reconstructions of pore size (A) and anther symmetry (B) in Miconieae. Character states of the terminal

branches are at the tips. Node labels show likelihoods of ancestral states. Taxonomic assignment of species is indicated by arches with

clade numbers: (1) Cremanium; (2) Caribbean; (3) Conostegia; (4) Mecranium; (5) Miconia; (6) Clidemia; (7) Tococa + Secundiflorae

+ Centrodesma.

48

Figure 5. Map of trait evolution of maximum petal (left panel) and anther (right panel) lengths.

Trait values of terminal branches and estimated trait values of internal nodes are shown with a

color gradient. Terminal branches are aligned for a perfect match between species in the

phylogenies. Species names are omitted for clarity. Clade assignments: (1) Clidemia; (2)

Miconia.

49

Figure 6. Diversification in Miconieae and Eriocnemeae estimated with BAMM. Diversification

rates were estimated for a phylogeny of 1083 species of Miconieae, three species of

Physeterostemon and one species of Eriocnema (all other outgroups removed) and are shown

with a color gradient (number of lineages per million years). Red circles denote the most credible

set of diversification rate shifts. The bar on the bottom is a geological timescale in millions of

years.

Diversification rates

Leandra

Cli

de

mi

a

Cr

e

m

ani

u

m

50

Figure 7. Transition rates between character states under the best BiSSE model of anther

symmetry evolution (speciation and extinction rates are constant). Marginal distributions for the

two rates were sampled with Markov Chain Monte Carlo. Horizontal bars are 95% confidence

intervals.

51

Figure 8. Transition rates between character states under the best BiSSE model of herkogamy

evolution (extinction rates are constant). Marginal distributions for the two rates were sampled

with Markov Chain Monte Carlo. Horizontal bars are 95% confidence intervals.

52

Figure 9. Speciation rates of character states under the best BiSSE model of herkogamy

evolution (extinction rates are constant). Marginal distributions for the two rates were sampled

with Markov Chain Monte Carlo. Horizontal bars are 95% confidence intervals.