THE ECOLOGY AND EVOLUTION OF ARMAMENT DEFENSE VARIATION IN AFRICAN ACACIAS

By

MEGAN CATHLEEN GITTINGER

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2013

1

© 2013 Megan Cathleen Gittinger

2

To my mom, who always supports my endeavors with healthy skepticism

3

ACKNOWLEDGMENTS

I thank James Ekiru, James Olingnengero, John Lemboi, and Moso for research

assistance in the field, without Ekiru I may have been trampled by elephants; Mike

Littlewood for logistical support at Mpala Research Centre, cheery disposition and

endless stories from the earlier days of Mpala and Kenya; Kathleen Rudolph for

commiseration in the field research and for moral support along the way; Todd Palmer

for feedback and support throughout my dissertation and for the Penicillins; Craig

Osenberg for including me in lab meetings; the Research Reviews group for valuable

feedback on experimental design; Dr. David Augustine for insightful feedback on herbivore foraging and for allowing me access to long-term exclosure plots at Mpala;

Bernard Hauser for insights in plant development; Heather McAuslane for comments

and discussions on inducible plant defense; Pamela Soltis for lab-support and guidance on molecular phylogenetics; Julian Resasco and Schuyler van Montfrans for being there from the beginning; Matthew Smith and Adrian Stier for adopting me as friend; Ashley

Seifert for being my husband and for always finding what I have to say interesting; and my entire family for providing endless support and for being amazing role models for personal and professional life.

4

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 8

LIST OF FIGURES ...... 9

ABSTRACT ...... 10

CHAPTER

1 GENERAL INTRODUCTION ...... 12

Why Study Structural Defense Traits? ...... 12 Why Study Acacia Spines? ...... 12 General Overview ...... 14

2 HERBIVORES AND PLANT SIZE AS DETERMINANTS OF SPINE DEFENSE MORPHOLOGY IN AN AFRICAN ACACIA ...... 15

Background ...... 15 Methods ...... 17 Study Site ...... 17 Natural Variation in Spine Defenses ...... 18 Herbivory and Spine Defenses ...... 18 Analysis ...... 19 Results ...... 19 Natural Variation in Spine Defenses ...... 19 Herbivory and Spine Defenses ...... 20 Herbivory, Spine Defenses, and Plant Size ...... 21 Discussion ...... 22 Relaxation of Spine Defenses following Herbivore Exclusion ...... 22 Size-based Relationships in Spine Defenses ...... 23

3 SPINE VARIATION DETERMINES DEFENSIVE EFFICACY AGAINST BROWSERS OF DIFFERENT SIZES AND FEEDING STRATEGIES ...... 37

Background ...... 37 Methods ...... 40 Study Site and Feeding Trials ...... 40 Plant Selection and Branch Treatments ...... 41 Biomass Estimates ...... 41 Patterns of Biomass Removal ...... 43 Branch Type Characteristics ...... 44 Results ...... 44

5

Defense Efficacy of Spines Against Goats and Camels ...... 44 Patterns of Biomass Removal ...... 45 Branch Type Characteristics ...... 46 Discussion ...... 48 Ecological Function and Consequences of Spine Form and Investment ...... 48 Mechanistic Links to Spine Variation ...... 51 Evolutionary Significance of Armament Variation ...... 53

4 EVOLUTION OF SPINE FORM AND DOMATIA IN AFRICAN VACHELLIA ...... 59

Background ...... 59 Methods ...... 61 Species Sampling ...... 61 DNA Extraction, Amplification and Sequencing ...... 62 Phylogenetic Estimation ...... 62 Trait Evolution ...... 63 Results ...... 63 Phylogenetic Estimation ...... 63 Evolution of Spine Morphology ...... 64 Discussion ...... 65 Phylogenetic Relationships of African Vachellia ...... 65 Independent Origins of Spine Form and Domatia in Vachellia ...... 66 Evolutionary Convergence of Structural Defense Traits ...... 68

5 SPINES, PRICKLES AND THORNS, OH, MY! PHYSICAL ARMAMENTS AS STAND UP TRAITS FOR STUDYING PLANT DEFENSES ...... 80

Background ...... 80 Armament Morphology and Development...... 83 Prickles ...... 84 Spines ...... 85 Thorns ...... 86 Armament as an Effective Form of Resistance ...... 88 Defensive Benefits of Armament ...... 88 Effects of Armament on Herbivore Feeding Behavior ...... 90 Costs of Defensive Armament ...... 94 Allocation Costs ...... 94 Opportunity and Ecological Costs ...... 97 Armament Plasticity ...... 99 Responses as Induced Defenses ...... 100 Ontogenetic Shifts in Armament ...... 102 Physiological Changes and Resource Allocation ...... 104 Future Directions ...... 106

6 GENERAL CONCLUSIONS ...... 112

Determinants of Spine Defense Variation ...... 112

6

Spine Variation Affects Defense Effectiveness ...... 113 Evolution of Spine Form and Domatia in African Vachellia ...... 114 Physical Armaments as Model Traits for Studying Plant Defenses ...... 115

APPENDIX EFFECT OF SPINE LENGTH ON DEFENSE EFFICACY ...... 116

Methods ...... 116 Results ...... 117

LIST OF REFERENCES ...... 120

BIOGRAPHICAL SKETCH ...... 134

7

LIST OF TABLES

Table page

2-1 Eigenvalues and eigenvectors from principal components analysis...... 26

2-2 Regression analyses and statistics for tree diameter relationships ...... 27

2-3 ANCOVAs for herbivore treatment and tree diameter ...... 28

2-4 ANCOVA least squares parameter estimates and full model r2 ...... 30

4-1 Table of taxa, voucher information, accession numbers for GenBank ...... 70

4-2 List of African Vachellia in this study with spine form and domatia traits...... 74

A-1 ANOVA results for goat and camel trials...... 119

8

LIST OF FIGURES

Figure page

2-1 Pair-wise correlations of growth and defense variables ...... 31

2-2 Loading plots of variables on the first two principal components ...... 32

2-3 Relationships between tree diameter and spine traits ...... 33

2-4 Response of spine defenses to long-term herbivore exclusion ...... 34

2-5 The relationships of tree diameter with branch structure and spine metrics for V. etbaica in exclosure and control plots ...... 35

3-1 The beneficial effects of spine form against herbivory ...... 56

3-2 Patterns of biomass removal in feeding trials ...... 57

3-3 Dry weight defense investment and actual spine proportions ...... 58

3-4 Average total initial number of leaves between branch types ...... 58

4-1 Examples of spine morphology in Vachellia ...... 75

4-2 Maximum Likelihood phylogram of African Vachellia ...... 76

4-3 Ancestral state reconstructions of spine form and domatia presence ...... 77

4-4 Majority rule bootstrap consensus cladogram showing clades within Vachellia.. 79

5-1 Variation in armament morphology and arrangement...... 108

5-2 The potential effects of armament on bite size and rate ...... 109

5-3 Illustration of how tolerance or variation in resource acquisition and allocation can mask trade-offs between armament and growth ...... 110

5-4 Allometry and the cost of armament defenses ...... 111

A- 1 The effect of spine length on defense efficacy...... 118

9

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

THE ECOLOGY AND EVOLUTION OF ARMAMENT DEFENSE VARIATION IN AFRICAN ACACIAS

By

Megan Cathleen Gittinger

August 2013

Chair: Todd M. Palmer Major: Zoology

Plants have little option but to stand their ground in the face of herbivores. To counter these attackers, plants have equipped themselves with a range of defenses commensurate with the diversity of herbivores. Extensive research on defensive chemistry among plants has been key in revealing their adaptive nature, however we know relatively less about the ecology and evolution of structural defense traits such as

armament.

Plant armaments such as spines and prickles are widespread among plants.

While these traits are widely accepted as effective defenses against large vertebrate

herbivores, it is less clear why armament exhibits such marked variation. The central

focus of this research was to improve our understanding of the factors that drive

variation in armament using African acacias as model taxa. I show that spine traits can

show strong relationships with plant size in Vachellia etbaica, and that these plants are

capable of altering spine form and size in response to herbivory. I also tested the

defensive function of spine variation in feeding trials with two different-sized vertebrate

herbivores, and found that spine form but not length had significant effects on biomass

10

loss against both herbivores. In addition to examining ecological variation, I assessed evolutionary patterns of spine defenses in African Vachellia. Based on the constructed molecular phylogeny, spine form and domatia appear to have evolved multiple times within the clade, indicating convergence of these structural defense traits. Finally, I synthesize our current knowledge of the armament traits and develop future research avenues that will improve our understanding of armament defenses as well as broadly inform ecological and evolutionary defense theories.

11

CHAPTER 1 GENERAL INTRODUCTION

Why Study Structural Defense Traits?

Plant defenses have been and continue to be a topic of broad interest to ecologists and evolutionary biologists. Defenses are ecologically relevant as they mediate innumerable species interactions and can drive broader community-level patterns. Defense traits are thought to have evolved largely as a response to herbivore pressure, and the evolution of plant defenses has been posited to be a major driver in the diversification of both plants and herbivores (Ehrlich and Raven 1964). While most research has focused on understanding chemical defenses, structural traits may be model traits to examine ecological and evolutionary questions of defense adaptation.

The study of structural defenses has a number of advantages over other plant defense systems. Firstly, they are discrete characters that are easily measured in the field.

There is no need for laboratory analysis, only a measuring tape. Secondly, assessing the specific defensive functions of different structural traits is also tractable as they can be readily manipulated in the field (Milewski et al. 1991, Cooper and Ginnett 1998,

Young and Okello 1998, Goheen et al. 2007). In the same way, the mammals that are deterred by structural defenses are easy to manipulate, relative to pathogens or invertebrate herbivores (Young 1987, Young et al. 2003, Goheen et al. 2007). These properties make structural defenses model traits to study questions of plant defense.

Why Study Acacia Spines?

Structural defenses in the form of armament are widespread among plants, particularly those found in arid environments (Grubb 1992). Acacias (e.g. Vachellia and

Senegalia) are the most abundant woody plants in many African savannas (Osborne

12

2000) and are of significant economic and ecological importance, serving as an important food source for native and domestic browsers. All African acacias have physical armament capable of defending against vertebrate herbivores (Symon 1986,

Milewski et al. 1991, Cooper and Ginnett 1998). While the presence of physical armament is ubiquitous, morphology varies widely. Among African Vachellia, structures are developmentally derived from stipular tissue and show variation in spine length and form (e.g. recurved and straight). Previous work has documented changes in armament following natural or experimental herbivory, with herbivory increasing armament density

(Myers 1987, White 1988, Bazely et al. 1991, Midgley and Ward 1996, Obeso 1997,

Gomez and Zamora 2002, Cooper et al. 2003, Zhang et al. 2006) and length (Midgley and Ward 1996, Gowda 1997, Obeso 1997, Rohner and Ward 1997, Young et al. 2003,

Zhang et al. 2006, Zinn 2007). Support for these responses as an inducible defense are strongest in V. drepanolobium. Spine length changes in response to vertebrate herbivory are well documented (Young 1987, Young and Okello 1998, Young et al.

2003), and continued investment in spine defenses in the absence of herbivores has strong negative effects on V. drepanolobium reproduction (Goheen et al. 2007).

However, whether this finding holds among Vachellia is not well established.

The majority of acacias have either recurved or straight spine forms, however some Vachellia are polymorphic, having both morphologies. It has been suggested that armament forms (recurved versus straight) vary in effectiveness against smaller (leaf feeding) versus larger (branch tip pruning) vertebrate browsers (Cooper and Owen-

Smith 1986, Belovsky et al. 1991). This hypothesis was formulated from studies comparing total biomass loss and herbivore feeding behavior among plant species with

13

various armaments and using herbivores of different size and feeding strategy. Yet there are limitations to cross-species comparisons. Plant species vary not only in structural armaments, but also defensive secondary chemistry (Abdulrazak et al. 2000,

Seigler 2003, Mokoboki et al. 2005), leaf size (Campbell 1986), and nutritional quality

(Abdulrazak et al. 2000, Mokoboki et al. 2005), and all of these potentially influence

herbivore feeding (Belovsky and Schmitz 1994, Wilson and Kerley 2003). Armament

also tends to negatively co-vary with primary biomass among plants (Belovsky et al.

1991), which is important for two reasons. Animals feeding on plants with less primary

biomass should be expected to have lower feeding rates, as bite size is a key

determinant of feeding rate (Shipley 2007). Additionally, a plant that initially has less

primary biomass should also be expected to lose less biomass. Thus, the extent to

which the wide variety of observed armament forms confer distinct defensive benefits,

and perhaps reflect adaptive variation, is unclear.

General Overview

In Chapter 2, I examine how spine defenses relate to plant size, and using long-

term herbivore exclosures I test the hypothesis that spine form and length respond to

herbivory in Vachellia etbaica. In Chapter 3, I assess the defensive function of spine

form using browsers of two different sizes to test whether spine form reflects defense

efficacy. Since V. etbaica has recurved and straight spine forms, it provides a model

system to explicitly compare the defensive efficacy of spine morphologies. In Chapter 4,

I construct a molecular phylogeny of African Vachellia to test whether spine form and

domatium presence have evolved multiple times within the clade. Chapter 5 synthesizes

our current knowledge of plant armament traits and identifies future research avenues

that will improve our understanding of these defensive strategies.

14

CHAPTER 2 HERBIVORES AND PLANT SIZE AS DETERMINANTS OF SPINE DEFENSE MORPHOLOGY IN AN AFRICAN ACACIA

Background

African savannas hold much of the world’s vertebrate herbivore biomass

(Scholes and Archer 1997). Spinescent acacias are the dominant form of woody vegetation in these systems and serve as a key food source for wild and domestic herbivores (Osborne 2000). Armaments are effective at protecting plant tissues against vertebrate herbivores (Cooper and Owen-Smith 1986), but their production is costly

(Gomez and Zamora 2002, Goheen et al. 2007). Having a flexible (i.e., inducible)

defense strategy reduces costs by allowing plants to invest in defense only when herbivores are present (Karban and Baldwin 1997). For instance, previous work has shown that straight spine length in Vachellia (formerly Acacia subg. Acacia) is an induced response to herbivory (Milewski et al. 1991, Gowda 1997, Young et al. 2003).

Not all Vachellia share the same spine form (e.g., straight), and some species have

mixed-spine morphology with shorter recurved and longer straight spines. Moreover,

these two armament forms are thought to reflect defense effectiveness against

herbivores with distinct feeding strategies (e.g., giraffe versus impala; Cooper and

Owen-Smith 1986). If plants are capable of altering spine functional form in response to

herbivory, this could have important consequences for herbivore communities (e.g., the

induced spine form more negatively affects browsers with particular feeding modes). It

is not well established whether plants simultaneously alter spine form and straight spine

length in response to herbivory.

Defense theory predicts that plants are optimally defended in space and time

(Feeny 1976, Rhoades 1979), and spatial rather than temporal variation in herbivory is

15

suggested to underlie the evolution of inducible spine defenses (Young et al. 2003).

Large herbivores can exert strong effects on acacia reproduction (Goheen et al. 2007,

Young and Augustine 2007) and population dynamics (Augustine and McNaughton

2004, Maclean et al. 2011), and can exert strong effects on acacias at younger life

stages (Augustine and McNaughton 2004). However, the few studies that assessed

age- or size-related patterns in spine defense found conflicting relationships between

age and spine defenses in acacias (Brooks and Owen-Smith 1994, Rohner and Ward

1997, Gowda and Palo 2003, Rooke et al. 2004). For example, straight spine length can

be either longer (Brooks and Owen-Smith 1994) or shorter (Rooke et al. 2004) in

juveniles versus mature trees. Spine length has been documented as an induced

response to herbivory in some acacias (Young 1987, Milewski et al. 1991), although

these patterns could reflect past herbivory rather than age per se, as herbivore

presence was not controlled in these studies.

Plants often exhibit allometry in growth (Enquist and Niklas 2002), so size-based

metrics of spine characters may be related to plant size rather than ontogeny.

Ontogenetic shifts in spine defense allocation could be due to (1) shifts in trade-offs

(e.g., reproduction at maturity; reviewed in Boege and Marquis 2005) or (2) variation in

the risk of herbivory. In the one study using an allocation-based metric, juvenile trees

were shown to invest more in spine defenses relative to mature individuals (Gowda and

Palo 2003), suggesting an ontogenetic shift in defense allocation. However, the ubiquity

of this pattern among acacias is unknown. Additionally, there have been few studies

examining how plant growth and spine defense characteristics relate among a broad

range of size or age classes. Understanding how structural defense traits vary with plant

16

size would provide a valuable starting point for studies exploring how developmental or metabolic factors (i.e., resources, herbivores) influence defense allocation.

Vachellia etbaica (formerly Acacia etbaica) is a medium-sized tree that is widely distributed in East Africa (Kenya, Tanzania, Uganda, Sudan, Ethiopia, and Sudan).

Vachellia etbaica produces both straight and recurved stipular spines, with substantial variation among individuals in spine investment in the field (MG Seifert personal observation). The spine defense metrics varied with plant size (i.e., saplings to large trees). Using long-term ungulate exclusion plots (10 years) I tested whether V. etbaica alters spine morphology in response to mammalian herbivory. Specifically I asked: do straight spine length and spine form (e.g., recurved or straight) change following herbivore exclosure? Lastly, the effect of herbivory on plant size and spine defenses

was explored.

Methods

Study Site

Research was conducted at Mpala Research Centre in Laikipia, Kenya (0°17’ N,

37°52’ E) from 2009 to 2011. In this semi-arid savanna system, V. etbaica is one of the

dominant shrubs along with Senegalia mellifera, S. brevispica, and Grewia tenax

(Young et al. 1995). Wild browsers and mixed-feeders at this site include dik dik

(Madoqua spp.), impala (Aepyceros melampus), kudu (Tragelaphus strepsiceros),

giraffe (Giraffa camelopardalis), eland (Taurotragus oryx), and elephant (Loxodonta

africana) (Augustine and McNaughton 2004); the only domestic browser on Mpala is

camel (Camelus), which were brought to Mpala in 2010. All data were collected from

three paired herbivore treatment plots (e.g., exclosures, controls) that were established

in 1999 (Augustine and McNaughton 2004), each plot measuring 0.5 ha (70 x 70 m).

17

The exclosure fences have been shown to be effective at excluding browsers, ranging in size from dik dik to elephant (Augustine and McNaughton 2004).

Natural Variation in Spine Defenses

In one plot exposed to ungulate herbivores, we measured spine characteristics for 45 V. ebaica individuals representing a broad range of sizes (varying from 1.36 - 5.7 m in height and 2 - 23 cm in basal diameter) to explore the relationship between plant size, branch size and spine defense metrics. For each tree, we measured basal diameter 10 cm above ground level. Branch length was measured proximally to the most recent growth node, and branch diameter was measured just distal from this node.

Spine length of at most 5 straight spines (5 cm proximal to tip), total number of straight spines and total number of spines (recurved and straight) were measured on 8 branches haphazardly selected throughout the tree canopy between 1-2 m height.

Herbivory and Spine Defenses

Within each plot (n=6), 20 trees (1 - 3.1 m) were tagged and 8 branches from each tree were haphazardly selected throughout the canopy between 1-2 m in height.

Data were collected on straight spine length (for all straight spines), total number of straight spines, total number of spines (recurved and straight), branch length, and branch diameter. Straight spine density cm-1 was used as the metric for altered spine

form in response to herbivory. Branches were selected and measured as described

previously. The length of recurved spines was not measured, as there is little evidence

for their induction in related taxa (Rooke et al. 2004).

In one exclosure plot (i.e., paired with initial control plot used to examine natural

variation), trees (n=33) of a similar range of sizes (1.9 to 23.5 basal diameter) were

selected to assess whether herbivory influenced the relationships between plant size

18

and spine defenses. All trait measurements were taken as described above. As plant and branch size variables were highly correlated (Fig. 2-1), I used a subset of plant size and branch size metrics (i.e., tree diameter, branch length) that showed the strongest associations for analysis. Additionally, I used previous estimates of branch dry weight and spine dry weight (Chapter 3) to calculate spine defense allocation between these trees (i.e., spine dry weight g / total dry weight of spines plus branch).

Analysis

We used principal component analysis to explore the relationships of spine

defenses (straight spine length cm, total number of straight spines, total number of

spines) and growth characteristics (branch length, branch diameter) with plant size.

When necessary data were log or square root transformed to meet the assumption of

normality using a Shapiro-Wilk’s test (alpha 0.05). The effect of herbivore treatment on

spine form (i.e., straight spine density cm-1) was measured between plots using paired t- tests, with paired plot locations included as a block effect (N=3). As spine defense metrics were strongly correlated, multiple paired tests were not run. Finally, ANCOVAs were used to assess whether herbivore treatment influenced the relationship between plant size and spine defense traits (i.e., straight spine length cm, straight spine density cm-1, total spine density cm-1). Preliminary tests were done to ensure data met

assumptions of normality, linearity and homogeneity of variance.

Results

Natural Variation in Spine Defenses

Plant size was strongly and positively associated with branch structure and spine

variables; the only exception was a weak negative relationship for straight spine number

and tree diameter and that relationship was not statistically significant (Fig. 2-1). Thus,

19

as trees increased in size, absolute investment in spine defenses also increased. This relationship between plant size or growth and defense was heavily reflected in the first principal component (Z1), accounting for 46% of total variance (Table 2-1, Fig. 2-2). All

variables were strongly and positively related to Z1 except for straight spine number

(Table 2-1). Univariate regressions of spine and branch metrics with tree size further

illustrate these relationships (Fig. 2-3). Straight spine number was strongly and

positively correlated with the second principal component (Table 2-1), explaining 18% of

the variance. This likely reflects the induction of straight spines in response to herbivory,

rather than growth-related investment. Together, plant-size and straight spine number

explained 65% of the variance (Table 2-1). Principal components 3 (Z3) and 4 (Z4)

explained 16% and 11% of the remaining variance (i.e., cumulatively 92%; Table 2-1), respectively. These axes may represent trade offs between growth and spine investment. For instance, on the third principal component axes branch diameter

(positive) and spine length (negative) show opposite associations. Similarly, branch

diameter (positive) and total spine number (negative) have contrasting patterns with

principal component 4 (Table 2-1).

Herbivory and Spine Defenses

Long-term ungulate exclusion significantly reduced the density of straight spines

on the current-year’s growth of branches (t2 = 10.518, P=0.0089; Average density of

straight spines, EX = 0.06 ± 0.011, C = 0.20 ± 0.011; Fig. 2-4b), but did not reduce the

combined density of straight and recurved spines (EX= 1.83 ± 0.035, C= 1.82 ± 0.040;

Fig. 2-4b). In addition, herbivore exclusion reduced the length of the straight spines

(Figs. 2-3a,c; Spine length cm EX= 3.75 ± 0.292, C= 4.60 ± 0.245). The response of

straight spine length and density to herbivore absence differed. Straight spines in were

20

18% shorter where herbivores were absent (t(2) = 4.899, P=0.0392; EX= 3.75 ± 0.292,

C= 4.60 ± 0.245; Fig. 2-4a), while straight spine density was 70% lower (t(2) = 10.518,

P=0.0089; EX = 0.06 ± 0.011, C = 0.20 ± 0.011; Fig. 2-4b).

Herbivory, Spine Defenses, and Plant Size

Long-term herbivore removal significantly altered the relationship between tree basal area (i.e., stem diameter) and tree height (ANCOVA: EX/C* Diameterlog10 F(1,74)

2 = 4.172, p=0.045, n P=0.05), even though diameters of the trees sampled in both

browsed and unbrowsed treatments were similar (t(76)= -0.472, P=0.6386; Diameterlog10

EX= 0.85 ± 0.055, C= 0.87 ± 0.047). This finding suggests that smaller trees (i.e.,

diameter) were taller on average when herbivores were absent. Overall, tree diameter

was strongly and positively correlated with tree height (ANCOVA: Diameterlog10 F(1,74)

2 = 235.499, p<0.001, n P=0.76).

Herbivore exclusion significantly reduced the slope of the relationship between

tree diameter and branch length (Tables 2-4; Figs. 2-5d). Accordingly, the relationships between tree diameter and spine metrics were similar to those found for trees in the control plot (Figs. 2-3 and 2-5). Removal of herbivores reduced the length, number and density of straight spines between paired plots (Figs. 2-3 and 2-5; Table 2-3). Likewise, total spine number and density were unaffected by herbivore treatment (Table 2-3).

Spine dry mass allocation declined with increasing plant size (Table 2-3), with a trend

for spine allocation to be greater in the presence of herbivores, but this was not

statistically significant (Table 2-3, Fig. 2-5e; EX= 0.359 ± 0.0128, C= 0.381 ± 0.0110).

21

Discussion

Relaxation of Spine Defenses following Herbivore Exclusion

Long-term removal of browsing ungulates greatly reduced the density of straight spines of V. etbaica (Fig. 2-4). Previous studies in another mixed-spine morphology

acacia attributed changes in straight spine density to spine lengthening (V. tortilis;

Rohner and Ward 1997, Gowda et al. 2003); however, this was not the case for V.

etbaica. Rather, reductions in straight spine density were due to the production of fewer

straight spines and more recurved spines (e.g., change in spine form from straight to

recurved). Straight spine length was correlated with straight spine number (Figs. 2-1,4),

and in the presence of browsers, straight spines were longer (Fig. 2-4). This is the first

clear evidence that spine form and straight spine length are induced responses to

ungulate herbivory in Vachellia.

The rate of defense relaxation is predicted to track the risk of herbivory and the

signal reliability of herbivore presence (Karban and Baldwin 1997, Young et al. 2003).

For instance, when herbivory is high, the absence of herbivory is a reliable signal to

relax defenses, but when herbivory is low, the absence of herbivory is not a reliable

signal so relaxation is expected to be slow. Based on estimates at the same location of

this study, herbivory rates based on branch loss for V. etbaica are between 13-25% per

year. These rates are similar to those estimated for V. drepanolobium (10-20%), which

is also found at Mpala Research Centre (Young and Okello 1998). Likewise, the

estimated per-year relaxation rate for V. etbaica (straight spine density) and V.

drepanolobium (straight spine length) is similar, 7% and 5%, respectively. While this

lends some credence to the relationship between relaxation rate and herbivory rate, it is

unclear why straight spine length in V. etbaica does not show the same rate of

22

relaxation. Little is known about how spine induction is regulated, but disparate relaxation rates could suggest that the signal(s) for spine length reduction and straight spine determination are distinct or evolved with differing selection pressures.

The benefits of induced defense traits against herbivory should outweigh their cost (Karban and Baldwin 1997). Straight spines are larger and possibly more costly to produce than recurved spines. Even if straight spine length was reduced by half, altering the production of a spine from straight to recurved would still be greater savings in terms of tissue production. Consequently, reducing the number of straight spines may be “prioritized” over reductions in the length of straight spines. However it is not clear how having both spine forms is advantageous. Does having both forms provide two alternative defensive functions (e.g., small versus larger herbivores)? Or, is having both forms a cost saving measure (e.g., lending plants the ability to more quickly reduce investment in spine tissue production)? Future work examining the relative costs and benefits of these two spine forms and the effect of these different morphologies on feeding efficacy for different herbivores would help address these alternative hypotheses (see Chapter 3).

Size-based Relationships in Spine Defenses

My findings show that spine defenses were strongly correlated to plant and

branch size. Plant and branch size were strongly correlated (r2=0.39; Fig. 2-3d) and showed positive relationships with spine length and total spine number (Figs. 2-1,3).

These size-based relationships are well illustrated in the first principal component (Z1),

where all variables except straight spine number had strong positive loadings (Table 2-

1). Therefore, as plants increase in size, straight spine length and total spine number

also increase (Fig. 2-3). Brooks and Owen-Smith (1994) found the opposite pattern for

23

straight spine length in V. nilotica, and no difference for V. tortilis. It is possible that these different patterns reflect differential herbivory and subsequent induction rather than plant age, as they sampled plants exposed to herbivores. While it is unknown whether V. nilotica spines are inducible, there is evidence for spine induction in V. tortilis

(Gowda 1997, Rohner and Ward 1997). Overall armament density in Vachellia and

other plant taxa is typically reported to be higher in juvenile plants (Kozlowski 1971,

White 1988, Brooks and Owen-Smith 1994), as I show here for V. etbaica (Table 2-2).

Straight spine number, which loaded strongly (positive) on the second principal

component (Z2), was not significantly related to plant size (Table 2-1; Fig. 2-3c).

In V. tortilis straight spine density was higher in juvenile plants (Brooks and Owen-Smith

1994); however, as mentioned above their study did not control for herbivory. As my

data suggest, herbivory induces straight spines (Fig. 2-4), and it is possible that

differences in past herbivory may explain the variation found in straight spine number.

There was not a significant interaction between plant size and herbivore effect for different spine metrics (Table 2-3). This indicates that the relationships between spine defenses and plant size are unrelated to herbivory. However, herbivore treatment and tree diameter showed an interactive effect on branch length (Table 2-3). The slope between tree diameter and branch length was greater in the control plots than in the exclosure plots (Table 2-4; Fig. 2-5d). Compensation (i.e., enhanced shoot extension rates) following herbivory is well-documented in acacias (Du Toit and Cumming 1999,

Gadd 2001), and this interaction may reflect growth compensation in trees exposed to herbivores. Acacias exhibit strong re-growth capacity (i.e., tolerance; Gadd 2001,

Gowda and Raffaele 2004, Fornara 2007, Scogings and Mopipi 2008), and future work

24

considering both tolerance and resistance traits would reveal whether there is a defense trade-off.

Throughout a plant’s lifespan, developmental trends in defenses are predicted to

shift resource acquisition and allocation. Defenses are expected to be highest at the late

sapling to pre-reproductive stage because, as a plant ages, defense allocation is

predicted to decrease due to trade-offs with reproduction and/or senescence (Boege

and Marquis 2005). I found support for this pattern since overall spine allocation (i.e.,

spine dry weight / total branch and spine dry weight) decreased with plant size (Fig. 2-

5e). Herbivore treatment did not have a significant effect on overall allocation by plants

to spine production (Table 2-3; Fig. 2-5e). This runs counter to the patterns observed for spine relaxation. It may be that the lack of herbivore effects on spine allocation was driven by larger branches (i.e., more dry mass) on trees exposed to herbivores. Even if these branches had more straight spines, differences in branch weight between herbivore treatments could have reduced the allocation differences (e.g., branch mass overall is higher than spine mass). Additionally, leaf mass was not considered in allocation calculations. Leaf mass and leaf number have been shown to be negatively correlated with spinescence and herbivory (Obeso 1997, Rohner and Ward 1997).

Although there are alternative ways to measure spine allocation, the data do suggest that spine defense allocation decreases with plant age as predicted by theory (Boege and Marquis 2005).

25

Table 2-1. Eigenvalues and eigenvectors from principal components analysis. A) Eigenvalues for principal components. B) Eigenvectors with loadings.

A Z Eigenvalue Percent Cumulative Percent 1 2.7846 46.4 46.4 2 1.1277 18.8 65.2 3 0.9327 15.5 80.8 4 0.6814 11.4 92.1 5 0.3407 5.7 97.8 6 0.1328 2.2 100.0

B Variable Z1 Z2 Z3 Z4 Z5 Z6 Tree diameter (log10) 0.44 -0.28 -0.39 0.36 0.64 0.19 Branch length (log10) 0.56 -0.13 0.01 -0.23 -0.08 -0.79 Branch diameter (log10) 0.37 -0.09 0.55 0.64 -0.35 0.13 Spine length 0.36 0.41 -0.61 0.04 -0.53 0.22 Straight spine number (log10) 0.15 0.85 0.27 0.06 0.42 -0.09 Total spine number (log10) 0.46 -0.10 0.32 -0.64 0.04 0.52

26

Table 2-2. Regression analyses and statistics for tree diameter relationships. A) Regression analyses. B) Statistics. A Source df SS III MS F P Branch length (log10) 1 0.1512 0.151 27.808 <0.001 Residual 43 0.2338 0.005

Straight spine length 1 7.7849 7.785 8.351 0.006 Residual 42 39.1528 0.932

Straight spine number (log10) 1 0.0250 0.025 0.271 0.605 Residual 42 3.8768 0.092

Total spine number (log10) 1 0.0513 0.051 5.577 0.023 Residual 43 0.3952 0.00

B LS Estimate SE t Ratio P r2 Branch length (log10) Tree diameter (log10) 0.175 0.0332 5.27 <0.001 0.39 Intercept 1.325 0.0314 42.12 <0.001 Straight spine length Tree diameter (log10) 1.302 0.4506 2.89 <0.001 0.17 Intercept 3.061 0.4307 7.11 <0.001 Straight spine number (log10) Tree diameter (log10) -0.074 0.1418 -0.52 0.605 0.01 Intercept 0.720 0.1355 5.32 <0.001 Total spine number (log10) Tree diameter (log10) 0.102 0.0432 2.36 0.023 0.11 Intercept 1.641 0.0409 40.14 <0.001

27

Table 2-3. ANCOVAs for herbivore treatment and tree diameter. A) With an interaction term. B) Without an interaction term.

A 2 Source df SS III MS F P Effect Size, n P Branch length (log10) Herbivore trtmt (EX/C) 1 0.0493 0.049 9.339 0.003 0.11 Tree diameter(log10) 1 0.0880 0.088 16.687 <.001 0.18 Herb x TDiam 1 0.0266 0.027 5.045 0.028 0.06 Error 74 0.3903 0.005 Total 77 0.6049 Straight spine length Herbivore trtmt (EX/C) 1 5.2129 5.213 4.154 0.045 0.06 Tree diameter(log10) 1 4.4464 4.446 3.543 0.064 0.05 Herb x TDiam 1 1.3473 1.347 1.074 0.304 0.02 Error 68 85.3355 1.255 Total 71 99.3576 Straight spine number (log10) Herbivore trtmt (EX/C) 1 2.1003 2.100 20.978 <.001 0.24 Tree diameter(log10) 1 0.0351 0.035 0.351 0.556 0.01 Herb x TDiam 1 0.1389 0.139 1.387 0.243 0.02 Error 68 6.8083 0.100 Total 71 9.1254 Total spine number (log10) Herbivore trtmt (EX/C) 1 0.0212 0.021 2.552 0.114 0.03 Tree diameter(log10) 1 0.0166 0.017 1.996 0.162 0.03 Herb x TDiam 1 0.0193 0.019 2.320 0.132 0.03 Error 74 0.6150 0.008 Total 77 0.6894 Total spine density (sqrt) Herbivore trtmt (EX/C) 1 0.0104 0.010 0.918 0.341 0.01 Tree diameter(log10) 1 0.0724 0.072 6.404 0.014 0.08 Herb x TDiam 1 0.0016 0.002 0.143 0.706 0.00 Error 74 0.8364 0.011 Total 77 0.9387 Straight spine density (sqrt) Herbivore trtmt (EX/C) 1 0.4336 0.434 21.666 <.001 0.23 Tree diameter(log10) 1 0.0008 0.001 0.038 0.847 0.00 Herb x TDiam 1 0.0317 0.032 1.586 0.212 0.02 Error 74 1.4810 0.020 Total 77 1.9552

28

Table 2-3. Continued

2 Source df SS III MS F P Effect Size, n P Spine allocation Herbivore trtmt (EX/C) 1 0.0090 0.009 1.650 0.203 0.02 Tree diameter(log10) 1 0.0682 0.068 12.480 0.001 0.14 Herb x TDiam 1 0.0017 0.002 0.317 0.575 <.05 Error 74 0.4043 0.005 Total 77 0.4949

B 2 Source df SS III MS F P Effect Size, n P Straight spine length Herbivore trtmt (EX/C) 1 5.0128 5.013 3.990 0.050 0.05 Tree diameter (log10) 1 6.7796 6.780 5.397 0.023 0.07 Error 69 86.6828 1.256 Total 71 99.3576 Straight spine number (log10) Herbivore trtmt (EX/C) 1 2.1467 2.147 21.321 <.001 0.24 Tree diameter (log10) 1 0.0055 0.006 0.055 0.816 <.05 Error 69 6.9472 0.101 Total 71 9.1254 Total spine number (log10) Herbivore trtmt (EX/C) 1 0.0202 0.020 2.388 0.127 0.03 Tree diameter (log10) 1 0.0320 0.032 3.785 0.055 0.05 Error 75 0.6343 0.008 Total 77 0.6894 Total spine density (sqrt) Herbivore trtmt (EX/C) 1 0.0102 0.010 0.910 0.343 0.01 Tree diameter (log10) 1 0.0869 0.087 7.780 0.007 0.09 Error 75 0.8380 0.011 Total 77 0.9387 Straight spine density (sqrt) Herbivore trtmt (EX/C) 1 0.4400 0.440 21.813 <.001 0.23 Tree diameter (log10) 1 0.0072 0.007 0.358 0.551 <.05 Error 75 1.5128 0.020 Total 77 1.9552 Spine allocation Herbivore trtmt (EX/C) 1 0.0092 0.049 1.704 0.196 0.02 Tree diameter (log10) 1 0.0824 0.088 15.210 <.001 0.17 Error 75 0.4061 0.005 Total 77 0.4949

29

Table 2-4. ANCOVA least squares parameter estimates and full model r2.

LS estimate SE t Ratio P r2 Branch length (log10) Herbivore trtmt (EX/C) 0.025 0.008 3.06 0.003 0.35 Tree diameter(log10) 0.113 0.028 4.08 <.001 Herb x Tdiam(log10) 0.062 0.028 2.25 0.028 Intercept 1.353 0.025 53.41 <.001 Straight spine length Herbivore trtmt (EX/C) 0.271 0.136 2.00 0.050 0.13 Tree diameter(log10) 0.985 0.424 2.32 0.023 Intercept 3.075 0.396 7.77 <.001 Straight spine number (log10) Herbivore trtmt (EX/C) 0.178 0.038 4.62 <.001 0.24 Tree diameter(log10) 0.028 0.120 0.23 0.816 Intercept 0.451 0.112 4.03 <.001 Total spine number (log10) Herbivore trtmt (EX/C) 0.016 0.011 1.55 0.127 0.08 Tree diameter(log10) 0.065 0.033 1.95 0.055 Intercept 1.657 0.031 53.61 <.001 Total spine density (sqrt)\ Herbivore trtmt (EX/C) -0.012 0.012 -0.95 0.343 0.11 Tree diameter(log10) -0.107 0.038 -2.79 0.007 Intercept 1.459 0.036 41.04 <.001 Straight spine density (sqrt) Herbivore trtmt (EX/C) 0.076 0.016 4.67 <.001 0.23 Tree diameter(log10) -0.031 0.052 -0.60 0.551 Intercept 0.350 0.048 7.32 <.001 Spine allocation Herbivore trtmt (EX/C) 0.011 0.008 1.31 0.196 0.18 Tree diameter(log10) -0.104 0.027 -3.90 <.001 Intercept 0.461 0.025 18.64 <.001

30

Figure 2-1. Pair-wise correlations of growth and defense variables. Pearson product- moment correlations for each comparison are labeled. ***P<0.001, **P <0.01, *P <0.05.

31

Figure 2-2. Loading plots of variables on the first two principal components, together explaining 65% of the total variance. PCA 1 represents the positive association of plant size with branch structure and spine metrics. PCA 2 characterizes variance explained due to the straight spine investment.

32

A B

C D

Figure 2-3. Relationships between tree diameter and spine traits. A) Spine length. B) Total spine number. C) Straight spine number. D) Branch length. Points are raw or transformed data and lines are regression fits (see Table 2-1).

33

A B

C

Figure 2-4. Response of spine defenses to long-term herbivore exclusion. A) Average length of straight spines. B) Straight spine density. C) Total spine density. Bars are SE ±.

34

A B

C D

Figure 2-5. The relationships of tree diameter with branch structure and spine metrics for V. etbaica in exclosure and control plots. A) Spine length. B) Total spine number. C) Straight spine number. D) Branch length. E) Spine allocation for V. etbaica in exclosure and control plots. Points are raw or transformed data as indicate on axes and lines are best-fit linear regressions (see Table 2-3). When the interaction term was not significant (see Table 2-2), a common slope was used between herbivore treatments (see Table 2-3).

35

E

Figure 2-5. Continued

36

CHAPTER 3 SPINE VARIATION DETERMINES DEFENSIVE EFFICACY AGAINST BROWSERS OF DIFFERENT SIZES AND FEEDING STRATEGIES

Background

As sessile organisms incapable of running away from predators, plants have evolved a bewildering array of defenses. These defensive traits reduce the impact of herbivory and allow plants to persist in the presence of herbivores. While there has been considerable research exploring how various phytochemicals influence plant- herbivore interactions, relatively few studies focus on structural traits (Grubb 1992,

Hanley et al. 2007). Plant structural defenses in the form of armament are taxonomically and geographically widespread, and spiny plants are especially abundant in semi-arid environments where large mammalian herbivores occur (Campbell 1986, Milton 1991,

Grubb 1992). These structures are often cited as a broad-spectrum defense against mammalian browsers, yet it is unclear why this generalized defense exhibits such

marked morphological variation both among and within plant taxa. Herbivores and their

relative impacts on plant fitness are predicted to shape defenses over evolutionary time,

and variation in these traits may reflect herbivore specificity (e.g., species, taxon or

guild). It has been proposed that armament form (recurved, straight) reflects the type of

herbivore that a plant is defending against; straight forms defend against larger

browsers (e.g., kudu, giraffe), while recurved forms defend against smaller browsers

(e.g., impala, dik dik; Cooper and Owen-Smith 1986, Belovsky et al. 1991). However,

this has yet to be experimentally tested. Establishing the functional difference between

different structural traits would address a key gap in our understanding of the ecological

significance of armament variation.

37

African acacias are model plants to examine the function of structural defenses, as they are the dominant form of woody vegetation in many African savannas and are all equipped with some form of armament. The presence of spines has been shown to reduce plant biomass loss to wild and domestic vertebrate browsers (Cooper and

Owen-Smith 1986, Belovsky et al. 1991, Milewski et al. 1991). Yet African acacias vary widely in the size, form and arrangement of these defensive structures, rather than merely their presence or absence. It has been suggested that armament forms

(recurved versus straight) vary in effectiveness against smaller (leaf removal) versus larger (pruning) vertebrate browsers (Cooper and Owen-Smith 1986, Belovsky et al.

1991). This hypothesis was formulated from studies comparing total biomass loss and herbivore feeding behavior between plant species with various armaments and using herbivores of different size and feeding strategies. There are two limitations to cross- species comparisons. First, it is likely that plant species vary not only in armaments, but also secondary defensive chemistry (Abdulrazak et al. 2000, Seigler 2003, Mokoboki et al. 2005), leaf size (Campbell 1986), or nutritional quality (Abdulrazak et al. 2000,

Mokoboki et al. 2005), all of which influence herbivore feeding (Belovsky and Schmitz

1994, Wilson and Kerley 2003). Second, armament negatively co-varies with primary biomass in plants (Belovsky et al. 1991), which is important for two reasons. Animals feeding on a plant with less primary biomass should be expected to have lower feeding rates, as bite size largely determines feeding rate (Shipley 2007). Additionally, a plant that initially has less primary biomass should also be expected to lose less biomass.

While there is support for the general utility of armament presence, the extent to which

38

the wide variety of observed armament forms confer distinct defensive benefits and reflects adaptive variation remains unclear.

Because assessing the specific defensive function of armament variation

requires decoupling the effects of armament from other plant-specific traits, testing the functional significance of variation in armament within a species allows for a more direct test of defense efficacy. Vachellia etbaica (formerly Acacia etbaica) is a widespread and common acacia species that produces both straight and recurved spines in varying proportions, providing an ideal model system in which to examine the function of different defense morphologies. This species is a medium-sized tree common at Mpala

Research Centre (Kenya) and throughout East Africa (Kenya, Tanzania, Uganda,

Sudan, Ethiopia, and Sudan). The relative proportion of recurved to straight spines varies widely at Mpala Research Centre (Kenya), with the proportion of straight spines

(to total) ranging between 0 and 0.53 and averaging 0.11 (SE ± 0.007) on mature (2-5 m) trees. Condensed tannins, the primary form of chemical defense in African Vachellia, are low in V. etbaica relative to other acacias (T. P. Young, unpublished data) and do not co-vary with spine form (MG Seifert, unpublished data).

I used V. etbaica’s natural variation to examine the effect of spine form on

defensive function, comparing herbivory by both small (goat) and large (camel)

herbivores on three branch types that varied in the proportion of straight spines (to

total): low (“R”, 0%), medium (“LS”, 10-15%) and high (“HS”, 50%). The following questions were addressed: (1) Does variation in spine form affect defense efficacy? (2)

Does the relative efficacy of spines differ depending on herbivore size and feeding

39

strategy? (3) How does spine variation relate to dry weight defense investment and other branch-level characteristics?

Methods

Study Site and Feeding Trials

This research was conducted at Mpala Research Centre in Laikipia, Kenya from

May to August 2010. Mpala Conservancy manages livestock and throughout the experiment Ranch employees maintained care of goats and camels according to Kenya animal protocols. During the day livestock are herded to feed on native browse; in the evenings, livestock are kept in protective “bomas” and have no access to food. All feedings trials were performed in the morning, and to control the “motivation” of the animals only a limited number of trials took place every morning between 7:30-10:30 a.m. (a maximum of eight and four for goats and camels per morning, respectively).

Individual feeding arenas were constructed for camels and goats, such that during feeding trials the animal was separate from other individuals. At least two test trials were done for each animal for acclimatization. Feeding trials lasted 1 and 2 minute(s) for goats and camels, respectively, allowing enough time for animals to potentially eat the majority of the plant biomass (as goats feed more rapidly than camels). During a feeding trial, each animal was presented with one treatment branch at a time (see below for branch treatment details) and in between each treatment branch there was a 30- second break. Treatment presentation order was randomized for individuals and trials;

feeding trials were replicated twice for all individuals and response variables were

averaged between trials (goats, N=35; camels, N=8).

40

Plant Selection and Branch Treatments

To assess the relative defense efficacy of different spine forms, there were three branch treatments: (1) high density, straight spines (HS), (2) low-density, straight spines

(LS) and (3) recurved spines only (R). These branch types correspond to 50, 15 and 0% straight (to total) spines, and represent natural levels of proportional investment (e.g., high, average and low) found in V. etbaica. Each branch type was paired with a spine removal control to detect potentially confounding branch type characteristics that may

correlate with spine form, investment and patterns of biomass loss.

Each set of treatment branches (6) was selected from an individual tree and

branch diameter was controlled for within an individual tree. A total of 86 trees were

used for the experiment. All treatment branches were selected, marked and measured

at most 7 days prior to feeding trials. I always attempted to find terminal branches (30 cm and 50 cm for goats and camels, respectively) with exact spine form proportions as

noted above; however, in some cases I used non-terminal branches. In these cases and

where terminal branches had side shoots, I trimmed excess length and side shoots

immediately before feeding trials to limit water loss. Branches were collected the

morning of feeding trials, and were transported in 5-gallon plastic bins with lids. Spines

were carefully removed from control branches using clippers immediately before feeding

trials. For all branches, I recorded the diameter at the branch base, number of spine

pairs, number of straight spines and number of leaves; the same measurements were

recorded following each feeding trial.

Biomass Estimates

Because it was not possible to use an appropriately sensitive scale to measure biomass at the livestock arenas, I used the following technique to estimate biomass loss

41

due to herbivory. Three branches (30 cm) from 10 V. etbaica trees were sampled in

February 2011, each branch corresponding to similar spine form ratios as treatments

(HS=0.56, SE ±0.058; LS=0.20, SE ±0.019; R=0, SE ±0.000). I recorded the diameter at

the branch base, number of spine pairs, number of straight spines and number of

leaves. Plant parts (i.e., leaves, spines, branch) were separated, dried at 60°C for 48

hours and then weighed to the nearest 0.001 g. Average leaf dry weight was not

significantly different between branch types (F2,27 = 1.167, P=0.33), therefore I used the

average of branch types (0.0089 g, SE ±0.0004) to calculate the dry weight of leaves for

feeding trial treatment branches.

To account for branch tapering in estimating experimental branch dry weights, I calculated dry weight as a function of dry weight per volume (g cm-3) and frustrum cone

2 2 volume: Vbranch= ((π x L) /12) x (dtip + dtip x dbase + dbase ) (eqn 1), where V is branch

volume, L is branch length, dtip is diameter at branch tip and dbase is diameter at branch

-1 base. I calculated dtip by obtaining an estimate of change in diameter cm as a function

of dbase from a separate set of branches that were similar to experimental branches. For

these branches, basal diameter was linearly regressed against change (Δ) in diameter

cm-1 with the intercept set to zero. Different branch lengths were used for goat and

camel feeding trials, and to account for this I used two estimates of Δ diameter cm-1,

one for the 30-cm segments (e.g., proximal tip to 30 cm distally) and the other for 20-cm

segments (e.g., the basal 20 cm portion of 50-cm branches) (1-30 cm, Δ diameter cm-1

-1 = 0.0164 dbase, N=18, F1,17= 949.660, P<0.0001; 30-50 cm, Δ diameter cm = 0.0127

dbase, N=18, F1,17= 330.840, P<0.0001). I chose to estimate the segment volumes

42

separately to more accurately reflect branch tapering with increasing branch length and diameter.

For the collected and weighed branches, average dry weight per volume (g cm-3)

was not different between branch types (F2,9=1.679, P=0.21), so I used the average

among trees (0.0078 g cm-3, N=10, SE ± 0.00043). I calculated branch dry weight (DW,

-3 g) as DWbranch=0.0078g cm x Vbranch (eqn 2). Both equation 1 and 2 were used to

calculate the initial and final branch dry weights for experimental branches (e.g., dtip was

-1 estimated based on the final branch length, Δ diameter cm and dbase; volume and dry

weight were then calculated using equations 1 and 2).

Defense dry weight estimates were also obtained from the same set of branches

that I collected and weighed. The average dry weight of a recurved spine was not

different between branch types (N=30, F2,28=2.123, P=0.14), so I used the average

among branch types (DWrecurvedspine = 0.0051 g, SE ± 0.0004, N=10) to calculate the

mass of recurved spines between treatment branches. Straight spine dry weight

(DWstraight spine) was calculated as a function of straight spine length (Lstraight spine)

2 (R a=0.37, N=10, F=6.123, P=0.04; HS and LS spine measurements averaged within

tree) using equation 3: DWstraightspine = 0.0211 + (0.0118 x Lstraightspine).

Patterns of Biomass Removal

Feeding behavior (i.e., bite type and total feeding time) was assessed during all

feeding trials in order to calculate average biomass removed per bite (g bite-1) and rate

of biomass removal (g min-1). The primary observer assessed whether each bite was a

leaf bite (i.e., pick) or a branch bite (i.e., prune, strip), while a second person recorded

data. Only camels “stripped” branches, which I define as removing biomass by biting the

branch and subsequently stripping off leaf biomass towards the branch’s distal end;

43

stripping sometimes but not always resulted in branch removal (e.g., stripping

sometimes resulted in only leaf loss, and other leaf and branch loss). Total time spent

feeding was recorded, which included actively taking bites as well as handling time

(e.g., successful bites, attempted bites, time between bites).

Goat and camel data were analyzed separately to gain more information on

herbivore-specific responses to spine variation and due to differences in starting

biomass (e.g., 30 cm branches for goats, 50 cm for camels). Data among trials were

averaged for individuals. All data were analyzed using a mixed-model ANOVA in JMP

8.0 with individual animal as a random effect, and Spine presence, Branch type, and

Spine presence × Branch type as fixed effects. The interaction term (Spine presence ×

Branch type) represents the isolated effect of spine form. Data were logarithmically

(log10) or arcsine-square root transformed when necessary to meet assumptions of

normality; all figures and descriptive statistics are based on untransformed data.

Branch Type Characteristics

I assessed branch level characteristics (i.e., branch diameter, dry weights, spine length, leaf number) to explore whether aspects of branch morphology systematically varied with branch type. Branch data were analyzed separately for goats and camels,

and spine present/absent treatment branches were averaged. Main effects in the model

were branch type and individual (i.e., tree), the latter of which was treated as a random

effect to account for tree-level variation.

Results

Defense Efficacy of Spines Against Goats and Camels

Goats rarely removed plant biomass via pruning. For those instances where

goats pruned, the average amount of branch biomass removed was small (0.073 g, SE

44

±0.0142), representing only 4% (SE ±0.8, N=21) of initial branch biomass (5% SE ±1.1

for branches with spines absent, N=14; 3% SE ±0.8 for branches with spines present,

N=7). Initial leaf biomass, between branch types, represented on average 30% of total

initial primary biomass (SE ±0.55, N=207), and as this was the primary form of biomass

lost. Consequently, trials with pruned branches were excluded from subsequent

analyses. Branches with different spine forms differed in their protective effects (Spine

presence B ranch type ANOVA F2, 27=3.88, P=0.0230; Fig. 3-1a); spine presence and

branch type were also significant (Spine presence F1, 27=71.05, P<0.0001; Branch type

F2, 27=5.13, P=0.0071). Treatment branches with straight spines (HS and LS) were an

estimated 7% more effective at protecting leaf tissue relative to recurved-spined (R)

branches (Tukey’s hsd, α=0.05; Fig. 3-1a).

Spine form also affected proportional biomass removal in response to camel

feeding (Spine presence Branch type F2, 7=6.46, P=0.0041; Spine presence F1,

7=15.31, P=0.0004; Branch type F2, 7=9.31, P=0.0006; Camel F7=11.86, P<0.0001; Fig.

3-1b). High-density, straight-spined branches (HS) retained 11 to 19% more of their

biomass relative to LS and R branches, respectively. Low-density, straight-spine branches (LS) retained an estimated 8% more biomass relative to recurved branches; however, this was not statistically different in post-hoc comparisons (Fig. 3-1b).

Patterns of Biomass Removal

Branch bites or pruning rarely occurred throughout goat feeding trials, with only seven of the thirty-five individuals successfully pruning (e.g., removing branch tissue)

between both trials. Goats were more likely to attempt pruning when spines were

absent (Cochran-Mantel-Haenszel Test X2=9.62, P=0.0019), and were also more

successful at pruning in the absence of spines (Cochran-Mantel-Haenszel Test

45

X2=4.09, P=0.0431). For goats that did not prune, spine presence reduced biomass removal per bite by 20% (Spine presence F1, 27= 36.49, P<0.0001; Fig. 3-2a) and the rate at which biomass was removed by 24% (Spine presence F1, 27= 59.05, P<0.0001;

Fig. 3-2c) despite increasing the total time spent feeding (Spine presence F1, 7=16.03,

P=0.0001). The rate of biomass removal (g min-1) was also affected by branch type

(Branch type F2, 27=8.48, P=0.0003; Fig. 3-2c), such that high-density, straight-spined branches (HS) experienced 15% lower rates of biomass removal relative to R branches

(Tukey’s hsd, α=0.05; R=LS>HS). There was a similar but non-significant trend for biomass removed per bite (Branch Type F2, 27=1.94, P=0.1481; Fig. 3-2a). Spine form did not alter biomass removed per bite (Spine presence Branch type F2, 27=0.57,

P=0.5686), nor the rate of biomass removal (Spine presence Branch type F2,

27=0.607, P=0.5466).

Spine presence and branch type had significant effects on patterns of biomass removal. Spine presence reduced per-bite biomass removal by 41% (F1, 7=20.46,

P<0.0001; Fig. 3-2b) and the rate of biomass removal by 32% (F1, 7=18.45, P=0.0001;

Fig. 3-2d). The presence of spines also resulted in camels taking fewer pruning or

stripping bites, relative to the total number of bites (Spine presence F1, 7=15.23,

P=0.0004). Branch type affected the rate of biomass removal (F2, 7=10.86, P=0.0002),

with camels attaining 43% higher rates of biomass removal on recurved-spined versus

high-density, straight-spined branches (Fig. 3-2d).

Branch Type Characteristics

As expected, branch treatments differed in their proportion of straight spines to

total (Goat: Mean for HS=0.47, SE ± 0.0047; Mean for LS=0.15, SE ± 0.0031; Mean for

R=0, SE± 0; N=69; Camel: Mean for HS=0.50, SE ± 0.0069; Mean for LS=0.10, SE ±

46

0.0056; Mean for R=0, SE± 0; N=18; Figs. 3-3a,b), although LS branches in camel trials had a slightly lower proportion of long spines (10% versus 15%). Branch diameter was not different between branch types (Goat: Branch type F2,68=1.86, P=0.1640; Tree

F68=227.56, P<0.0001; Camel: Branch type F2, 15=0.44, P=0.6478; Tree F15=128.21,

P<0.0001). Branch types varied significantly in their dry weight defense investment

(Goat: Branch type F2, 68= 2468.80, P<0.0001; Tree F68=8.82, P<0.0001; Camel: Branch

type F2, 15= 654.45, P<0.0001; Tree F15=4.82, P<0.0001; Figs. 3-3a,b). High density,

straight-spined branches invested nearly 40% of their total dry weight in defense

structures (Goat: Mean of HS=0.38, SE ± 0.0084; Camel: Mean of HS=0.35, SE ±

0.015), low density, straight-spined branches (LS) invested about 20% (Goat: Mean of

LS=0.23, SE ± 0.006; Camel: Mean of LS=0.16, SE ± 0.008) and recurved-spined branches (R) approximately 8% (Goat: Mean of R=0.09, SE ± 0.003; Camel: Mean of

R=0.08, SE ± 0.003) of their dry weight (Figs. 3-3a,b). Average straight spine length

was not significantly different among branch types (Goat: Branch type F1, 68= 1.19,

P=0.28; Tree F68=4.12, P<0.0001; Camel: Branch type F1, 15= 0.736, P=0.4043; Tree

-1 F15=1.48, P=0.2248). However spine density (spines cm ) was strongly related to

branch type (Goat: Branch type F2, 68= 151.73, P<0.0001; Tree F68=5.51, P<0.0001;

Camel: Branch type F2, 15= 35.60, P<0.0001; Tree F15=2.68, P=0.0105), with R branches

having the highest spine densities and HS the lowest (Goat: Tukey’s hsd, α=0.05,

R>LS>HS; Mean for R=1.7 spines cm-1, SE ± 0.033; Mean for LS=1.6 spines cm-1, SE ±

0.031; Mean for HS=1.3 spines cm-1, SE ± 0.019; N=69; Camel: Tukey’s hsd, α=0.05,

R=LS>HS; Mean for R=1.7 spines cm-1, SE ± 0.045; Mean for LS=1.6 spines cm-1, SE ±

0.038; Mean for HS=1.3 spines cm-1, SE ± 0.044; N=16). There were also differences in

47

average leaf number among branch types (Goat: Branch type F2, 68=14.10, P<0.0001;

Tree F68=5.76, P<0.0001; Camel: Branch type F2, 15=5.56, P=0.0088; Tree F15=7.41,

P<0.0001), with R branches having the most leaves and HS branches the least (Figs. 3-

4a,b).

Discussion

Ecological Function and Consequences of Spine Form and Investment

Results from this study show that spine form dictates defense efficacy against

two vertebrate browsers and that the relative efficacy of spine form differs depending on

herbivore size and feeding strategy. Straight spines consistently were more effective

than recurved spines alone at protecting plant biomass (Figs. 3-1a,b). Against goats,

which primarily fed on leaf tissue, straight spines protected an estimated 7% more leaf

tissue than recurved spines alone (Fig. 3-1a). Independent of spine density (HS vs. LS)

straight spines conferred a greater benefit in terms of biomass protection against goats.

This was not the case for camels. Only high densities of straight spines (HS) enhanced

overall biomass protection (Fig. 3-1b), although there was a trend for low densities of

straight spines (LS) to be more effective relative to recurved spines only (R). These data

are consistent with previous hypotheses predicting that recurved armature protects leaf

tissue against picking, and straight armature protects branches against pruning (Cooper

and Owen-Smith 1986, Belovsky et al. 1991). However, straight spines are also more

effective than recurved at protecting leaf biomass against picking (Fig. 3-1a) and spines

are relatively more effective against larger browsers (i.e., proportionally more biomass is

protected against camel versus goat browsing; Figs. 3-1a, b).

Structural armament protects plant biomass by reducing rates of biomass

removal (Cooper and Owen-Smith 1986, Belovsky et al. 1991, Milewski et al. 1991,

48

Gowda 1996). Similarly, it was found that spines reduced the average amount of

biomass removed per bite and the overall rate of biomass removal by goats (leaf tissue)

and camels (leaf and branch tissue) (Figs. 3-2a-d). In the goat and camel feeding trials,

branch types differed in biomass removal rate (Figs. 3-2c, d), however different spine

forms among branch types was not the driving factor (e.g., Spine presence X Branch

type was not statistically significant). Herbivore feeding (e.g., bite size) can be

constrained by leaf size or available biomass (Burlison et al. 1991, Penning et al. 1991,

Laca et al. 1992), and the different branch types (R, LS, HS) used in this study varied in

the amount of starting leaf biomass. Branches with increasing densities of straight

spines had fewer leaves overall (Figs. 3-4a, b), and these differences paralleled per-bite

biomass removal for both goats and camels (Figs. 3-2a-d). If spine investment

consistently co-varies with available leaf or branch biomass, then disentangling the

function of different spine defenses will require taking these associations into account.

The fact that armament has been shown to correlate with a variety of growth-related

traits, such as leaf size or branch diameter (Cooper and Owen-Smith 1986, Belovsky et

al. 1991, Milewski et al. 1991, Gowda 1996, Ward 2010), underscores the complexity of

this relationship.

In V. etbaica, spine form was tightly linked to defense investment; a straight

spine of average length (5.5 cm, SE ± 0.31) weighed over 16 times more than a

recurved spine (0.086 g straight spine-1, SE ±0.0056; 0.005 g recurved spine-1, SE

±0.0004). Thus, an alternate explanation could be that spine investment rather than

spine form per se increases defense efficacy. Branch types (e.g., R, LS, HS) differed

substantially in their dry weight defense investment with recurved-spined branches

49

having the lowest levels of investment (8-9%; camel and goat branches, respectively) and straight-spined branches (HS) the highest (39-56%; goat and camel branches,

respectively; Figs. 3-3a,b). In this study, branches with many straight spines (HS)

retained 19% more of their leaf and branch biomass in camel feeding trials (e.g. 89% of biomass remaining with spines present vs. 70% when removed), whereas branches with few (LS) to no (R) straight spines only retained 8 and 0% more, respectively (Fig.

3-1b). Straight spines also conferred a greater benefit against goat feeding (Fig. 3-1a), but these differences were smaller in terms of total biomass protection (e.g., leaf and

branch tissue). Leaves represented an average 29% of total biomass between branch

types; therefore, a 7% reduction that straight spines confer against leaf removal translates to ~0.2% loss in total biomass. So, against camels, spine investment and form parallel defense efficacy, but against goats the protective effects of spine form and investment were less dramatic. In a separate experiment, I also found that straight spine investment (e.g., reducing spine length by approximately 50%) had no effect on defense efficacy against goats or camels (APPENDIX). These data suggest that while

investment is tied to spine form, investment does not necessarily enhance defense

efficacy. In V. etbaica, as well as other acacias, spine form appears to be consistently

linked to investment, and it is possible that armament form reflect a plant’s defense

strategy. Future studies decoupling form and investment, as well as modifying

investment in certain armament forms would shed light on these key distinctions.

The costs associated with losing branches versus leaf tissues are likely very

different, yet both forms of tissue loss have significant fitness consequences. Large

browsers, such as elephants and giraffes, feed primarily by removing branch tissue,

50

which significantly reduces plant growth and reproduction (Goheen et al. 2007). While the large-scale removal of branch tissue clearly has negative effects on plant fitness, the removal of leaf tissue alone can also be costly. Smaller vertebrate browsers such as dik dik feed via leaf removal, and reductions in leaf biomass within the feeding height of this animal (<0.5 m) have been shown to negatively impact acacia branch growth rates and shrub recruitment (Augustine and McNaughton 2004). In fact, at Mpala Research

Centre where dik dik occur at high density it appears that they exert some of the largest

effects on V. etbaica – second only to elephants – due to their “chronic” herbivory

(Augustine and McNaughton 2004). Severe defoliation, particular at young life stages,

increases mortality rates as a result of depletion of carbohydrate reserves (Schutz et al.

2011). Whether spine form is energetically effective will likely depend on the relative

costs of losing leaf or branch tissue, the length of time that these defensive benefits

accrue (i.e., spines protecting biomass over multiple seasons or years) and a plant’s

age (i.e., allocation, herbivore exposure).

Mechanistic Links to Spine Variation

Resources limit plant growth and differentiation, and if defense allocation trades

off with primary metabolic processes then there should be physiological trade-offs

associated with defense investment. For V. etbaica, there was a 16-fold difference in

the average dry weight investment between a straight and recurved spine, suggesting a

substantial allocation of resources to straight spine investment. The growth-

differentiation balance hypothesis (GBDH) predicts that defenses (differentiation) should

trade-off with growth (Herms and Mattson 1992), and data here are consistent with this

expectation, as spine investment (i.e., differentiation) was negatively associated with

leaf biomass (i.e., growth and primary metabolism) between branch types (Figs. 3-3a,b

51

and 3-4a,b). This finding also parallels previous work documenting reduced leaf size and increased spine investment in the presence of herbivores (Zhang et al. 2006, Zinn

2007). Source-sink dynamics can affect patterns of growth as well as defense investment (Haukioja 1990, 1991, Honkanen and Haukioja 1994, Honkanen et al. 1999,

Arnold et al. 2004), and it is interesting to consider the possible feedbacks of a negative relationship between primary production and defense investment. For example, if well- defended branches are smaller sources (i.e., lower photosynthetic output), then they may also be weaker sinks potentially affecting future growth if source-sink strength

regulates meristem activity and branch outgrowth. Herbivores could also modulate such

growth-related processes through differential rates of herbivory on less-defended,

higher leaf biomass versus well-defended, low leaf biomass branches. In this study,

even though branches with straight spines initially had less leaf biomass (Figs. 3-4a,b),

they had relatively more leaf biomass compared to recurved following feeding trials.

While the underlying developmental or physiological mechanisms regulating leaf and

spine biomass allocation, and their potential feedbacks, are unknown, data from this

work point to potential linkages between growth and spine defense.

Defense allocation can relate to allometry, ontogeny and environmental context

(i.e., herbivory, resources), and many of these have been shown to influence armament

variation. Plant age is often negatively associated with spine density (Brooks and Owen-

Smith 1994) and size (Brooks and Owen-Smith 1994, Gowda and Palo 2003); however,

the nature of this variation is equivocal given that in some cases no pattern (Rooke et

al. 2004) or the opposite pattern (e.g. positive association of defense with age) occurs

(Brooks and Owen-Smith 1994, Rooke et al. 2004). Size in plants is tightly linked to

52

developmental stage, making it difficult to disentangle allometric relationships from developmental or resource-driven effects. For example, previous work showing that resource availability can increase dry weight defense investment in armament (Gowda

et al. 2003) could be related to plant size rather than resources per se. Herbivores

undoubtedly play a role in shaping patterns of armament investment, and induced

responses of spines and prickles to herbivory have been well documented. High levels

of natural and experimental herbivory have been shown to increase spine, prickle and

thorn density or length (Myers 1987, Bazely et al. 1991, Gowda 1997, Rohner and Ward

1997, Gomez and Zamora 2002, Young et al. 2003, Zhang et al. 2006, Zinn 2007).

There are, however, cases where herbivory did not elicit an induced response in spine

investment (Gadd et al. 2001, Ward 2010). Together these data demonstrate that

armament structures can exhibit a wide range of responses and relationships to a

plant’s internal and external environment. Data also support trade-offs predicted by the

GDBH, and I suggest that the implicit link between structural armament and growth

(e.g., spines are produced through growth processes, such as cell division and

elongation) may result in stronger trade-offs in growth and differentiation or at least

make them more detectable. Future work examining how armament form and

investment vary with primary metabolism, as well as other defensive traits, will shed

light on the developmental and physiological trade-offs associated with structural

defenses and help inform broader-scale patterns of armament variation.

Evolutionary Significance of Armament Variation

Differences in the relative effectiveness of defense traits against herbivory lend

credence to interpretations of the adaptive nature of plant defenses (Karban and

Baldwin 1997). Structural defenses are considered to be broad spectrum, defending

53

against a wide array of vertebrate herbivores. Relative to highly specialized plant phytochemical-herbivore systems (Ehrlich and Raven 1964, Van Zandt and Agrawal

2004, Zangerl et al. 2008) armament is surely a more general form of defense. Results from the current study show that the defensive function of recurved spines is primarily for leaf protection, while straight spines are effective at protecting leaf and branch tissue. It is tempting to surmise that straight spines in Vachellia are an adaptation to

larger herbivores that feed primarily via branch removal given that: (1) straight spines

were more costly in terms of dry weight investment, (2) straight spines were relatively

more effective against camels and (3) increasing straight spine investment did not

confer additional benefits against goats. However, this is likely an overly simplified view

of the evolution of armament. Defense evolution is not just a consequence of protection

at the scale of a branch; it arises through its influence on herbivore foraging behavior.

Both goats and camels were less likely to prune or strip branches when spines were

present, and also had smaller bite sizes and slower feeding rates (Fig. 3-2). The effects

of spines on animal foraging will likely have important opportunity costs, especially for

wild browsers (e.g. vigilance, mate competition). Thus, even small differences in terms

of defense efficacy could have large effects in terms of regulating browser populations.

Plant defenses are complex traits reflecting not only defensive function but also

evolutionary and environmental context (Ehrlich and Raven 1964, Feeny 1976, Coley et

al. 1985, Armbruster 1997, Fine et al. 2006). The presence of armament among

numerous plant taxa and the association of armed plants with arid environments have

been suggested as correlative evidence for the convergence of structural defense traits

(Campbell 1986, Milton 1991, Grubb 1992). Convergence would indicate that selection

54

for these traits is strong and that structural defenses are evolutionarily labile; however, this hypothesis has yet to be formally tested. There is also a hindrance to broadly testing the nature of convergence since there has been little work establishing armament homology, and the evolution of similar traits via different pathways has different implications than the evolutionary gain or loss of traits via the same pathway. A recent study found that the prickles of two closely related plant species differed in their cellular anatomy (Kellogg et al. 2011), which may mean that even phenotypically similar armament structures may be developmentally distinct among species. Comparative

work focusing on the development of defense structures is necessary to elucidate the

evolutionary novelty or homology of armament among plant taxa. Intra-taxon studies,

however, may be key in elucidating how armament variation has evolved and the

extents to which these structural defenses are associated with other plant

characteristics (e.g., growth rate, palatability, chemical defences, plant architecture;

Agrawal and Fishbein 2006).

55

A B

Figure 3-1. The beneficial effects of spine form against herbivory. A) Goat feeding trials. B) Camel feeding trials. Note the different scales on y axes. Bars are SE ±. Letters denote Tukey’s HSD significant differences (α=0.05).

56

A B

C D

Figure 3-2. Patterns of biomass removal in feeding trials. The presence of spines affected the amount of biomass removed per bite for A) goats (Spine presence F=36.49, P<0.0001) and B) camels (Spine presence F=20.46, P<0.0001). C) The rate at which goats were able to remove leaf biomass was effected by spine presence as well as branch type (Spine presence, F=59.052, P<0.0001; Branch type, F=8.477, P=0.0003). D) A similar pattern in manifested in camel feeding trials, with a non-significant trend for spine form influencing the rate of biomass removal (Spine presence, F=18.45, P=0.0001; Branch type, F=10.86, P=0.0002; Spine presence x Branch type, F=2.52, P=0.0954). Bars are SE ±. Letters denote Tukey’s HSD significant differences (α=0.05).

57

A B

Figure 3-3. Dry weight defense investment (primary y-axis, grey bars) and actual spine proportions (secondary y-axis, points) between branch types. A) Goat feeding trials. B) Camel feeding trials. Bars are SE ±.

A B

Figure 3-4. Average total initial number of leaves between branch types. A) Branches used in goat trials. B) Branches used in camel feeding trials. Bars are SE ±. Letters denote Tukey’s HSD significant differences (α=0.05).

58

CHAPTER 4 EVOLUTION OF SPINE FORM AND DOMATIA IN AFRICAN VACHELLIA

Background

A key evolutionary question is how natural selection can shape particular

organismal traits, and robust phylogenies provide the framework in which to examine this process. Plant defense morphology is an ideal system to explore evolving traits, as plants are equipped with a range of defensive armaments that display inter- and intraspecific variability. Physical armament, along with toxic chemicals, digestion- inhibiting compounds, and minerals that wear down herbivore mouthparts (e.g., silica), are all examples of the artillery that plants employ to deter herbivores (Rosenthal and

Berenbaum 1992b). Defense traits are thought to have evolved largely as a response to

herbivore pressure, and the evolution of plant defenses has been posited to be a major

driver in the diversification of both plants and herbivores (Ehrlich and Raven 1964). With

the increasing accessibility of molecular techniques, there has been a resurgence of

comparative studies examining macroevolutionary patterns of plant defenses. However,

there has been an almost exclusive focus on chemical defenses (but see Armbruster

1997, Agrawal and Fishbein 2008), leaving structural defense traits have been largely

ignored. Because structural defenses are almost as widespread among plant taxa

(Grubb 1992), there exists a large gap in our understanding of the broad-scale patterns

of defense trait evolution (Agrawal 2007).

Morphological traits such as spines and thorns are modifications of primary plant

structures (e.g., leaf and branch tissue, respectively), and each exhibits interspecific

and intraspecific variation, making them model traits to examine evolutionary questions

of trait adaptation. Structural defenses in the form of armament are widespread traits

59

among plant taxa, particularly in arid environments (Grubb 1992). Acacias (e.g.

Vachellia and Senegalia) are some of the most abundant woody plants in African

savannas (Osborne 2000), and all are equipped with physical armament capable of

defending against vertebrate herbivores (Symon 1986, Milewski et al. 1991, Cooper and

Ginnett 1998). There is, however, a wide range of interspecific and intraspecific variation in armament form (Fig. 4-1), and this variation has been linked to defense effectiveness (Cooper and Owen-Smith 1986, Belovsky et al. 1991). In addition to defense structures, some Vachellia species also develop domatia, which are plant- produced hollow structures that may house mutualist-ant defenders. This additional form of defense via mutualist-ant partners is a highly effective deterrent against large mammalian as well as invertebrate herbivores (Stapley 1998, Goheen and Palmer

2010, Campbell et al. 2013). Because of the complex nature of defense traits, phylogenetic conservatism predicts that defense type or form will be restricted to closely related taxa (Janzen 1974b, Berenbaum 1983, Armbruster 1997, Farrell 1998).

Alternatively, similar biotic and abiotic selection pressures may result in convergence of plant defenses (Kursar 2003, Agrawal and Fishbein 2006, Fine et al. 2006). Although conflicting patterns have been found in studies examining chemical defense evolution

(Wink 2003, Liscombe et al. 2005), there have been few studies explicitly addressing these hypotheses for structural defense traits.

In this study, I critically evaluate a long-held hypothesis that plant defense traits are phylogenetically conserved using the novel system of spine morphology among

Vachellia. The wide variation in defense morphology among Vachellia and the potential adaptive nature of armament form make this a model group to examine the evolution of

60

structural defenses. A plastid phylogeny was constructed to test whether spine form and domatia presence have evolved more than once among 39 of the 74 African Vachellia taxa, with the a priori prediction that (1) spine form will show constraint in the evolution of recurved forms, and (2) domatia have evolved independently multiple times.

Methods

Species Sampling

Vachellia comprises 162 species: 74 species in Africa, 60 in the Americas, 36 in

Asia, and 9 in Australia. Defense structures are believed to be stipular in origin but vary in spine form as well as domatia presence (Fig. 4-1; Janzen 1974a, Vassal 1981,

Pedley 1986, Coe and Beentje 1992). Vachellia has been resolved as monophyletic

(Clarke et al. 2000, Luckow et al. 2003, Miller and Bayer 2003, Bouchenak-Khelladi et al. 2010), with the African and American clades sister to one another (Luckow et al.

2003, Miller and Bayer 2003, Fig. 2; Bouchenak-Khelladi et al. 2010). However, there are gaps in species sampling for African Vachellia, particularly when considering the range of defense morphologies within the clade. To fill this gap, additional species were sampled using field-collected and herbarium specimens (Table 4-2). Field collection was restricted to species present in Kenya due to logistical constraints. All samples were collected in affiliation with the National Museums of Kenya (NMK) East African

Herbarium (EAH). For each species and at each location for each species, 3 leaf samples and 2 voucher specimens were collected from individual trees, and voucher specimens were deposited at the Florida Museum of Natural History. Each leaf sample was preserved in silica gel desiccant and was transported to the University of Florida for

DNA extraction, amplification, sequencing and subsequent analysis. This sampling

effort increased the total species coverage within Vachellia to 53%, including 50% of all

61

species with domatia, 100% of all species with recurved spines and 52% of all species with straight spines (Table 4-2).

DNA Extraction, Amplification and Sequencing

Plastid DNA are widely used in the construction of molecular phylogenies and contain accepted markers for resolving phylogenetic relationships (Soltis et al. 1998).

Three plastid regions (matK, psbA, trnL) were sequenced from field and herbarium specimens. Tissue was homogenized with Zirconium beads and suspended in CTAB buffer. DNA was extracted using a CTAB method and purified with phenol-chloroform according to current methods (Doyle and Doyle 1987, Cullings 1992). Primers used for sequence amplification followed methods previously established in Vachellia (Miller and

Bayer 2001). All lab-based work was performed in the Laboratory of Molecular

Systematics and Evolutionary Genetics at the Florida Museum of Natural History, and

DNA was sequenced at the Interdisciplinary Center for Biotechnology Research,

University of Florida.

Phylogenetic Estimation

Sequences were edited and assembled in Sequencher 4.2 (Gene Codes Corp.

2005), aligned using MUSCLE and manually checked in SeaView 4.3 (Galtier 1996,

R.C. 2004, Gouy 2010). Plastid data were analyzed together as they individually

exhibited low levels of sequence divergence. Phylogenetic relationships were estimated

using Maximum Likelihood with a GTR Gamma model of molecular evolution (RAxML;

Stamatakis 2006). Twenty ML searches on 20 randomized stepwise addition parsimony

trees were run to find the ML tree, and 1000 bootstrap replicates were run for bootstrap

support values (Felsenstein 1985a).

62

Trait Evolution

Ancestral state reconstructions were done in Mesquite 2.75 using parsimony and

ML methods on the Majority Rule Bootstrap Consensus tree with branch lengths set to 1

(50%) (Maddison 2007). ML ancestral state reconstruction was done using a Markov k-

state 1 parameter model (Mk1 model; Lewis 2001). Significance for likelihood scores

was set at a difference of 2 likelihood units (Edwards 1972). Spine form has two states; however, some species contain both states (e.g, recurved and straight spines). While a species with mixed-spine morphology may be considered polymorphic, there is no a

priori expectation for different transition rates between singular states and the

polymorphic state. Therefore, spine form was assigned as 1 of 3 possible states:

straight, recurved, or mixed. Domatia had two possible character states: presence or

absence. Spine form and domatia were discrete and unordered characters, such that

transitions between character states are treated equally likely and unconstrained. Both

traits are derived from stipular tissue, and there is no current evidence suggesting these traits should be restricted in their directionality.

Results

Phylogenetic Estimation

The complete data set included 38 species of Vachellia plus two species as an

outgroup (Parkia timoriana, Mimosa tenuiflora). Of 4584 characters, 1161 were potentially parsimony informative; the proportion of gaps and completely undetermined characters was 30.13%. The best scoring ML tree (-11006.91) is shown in Fig. 4-2. The amount of sequence divergence was low overall as indicated by branch lengths, and many of the relationships within Vachellia were not well resolved (Fig. 4-2). There were however groups of species that formed distinct clades (Fig. 4-4). Clade A included V.

63

erioloba, V. haematoxylon, V. bussei, and V. lahai and was sister to the rest of African

Vachellia. Previous work based on molecular data showed a similar pattern, however V. erioloba was the only species of Clade A included (Bouchenak-Khelladi et al. 2010).

Clades B (V. reficiens, V. drepanolobium, V. etbaica, V. elatior, V. grandicornuta, V. luedertzii, V. sieberiana, V. abyssinica, V. tortils, V. hebeclada, V. rhemanniana, V. gerrardii, V. robusta, V. stulmannii, V. nubica), C (V. kosiensis, V. karroo, V. ormocarpoides, V. sekhukheniensis, V. robbertsii) and D (V. permixta, V. exuvialis, V. nebrownii) also showed strong support and corroborate previous groupings

(Bouchenak-Khelladi et al. 2010). While Clade C was previously nested within another

clade that included Clade D and V. xanthophlea and V. kirkii (Bouchenak-Khelladi et al.

2010), our analysis showed low support for that topology.

Evolution of Spine Morphology

Of the 38 species, 6 had recurved or mixed spine morphologies (16%), and five

of these species were within Clade B (Fig. 4-4). Parsimony and ML character

reconstruction yielded similar results and suggest that species with recurved and mixed

spine morphologies arose from ancestors with straight spines (Fig. 4-3). Results of

parsimony reconstruction show these morphologies exhibiting maximum homoplasy

(Steps=6; Consistency Index=0.3, Retention Index=0.0), and the results of ML are

comparable (Mk1 est.: rate=0.062, -log Likelihood=23.37; Figs. 4-3b). Both parsimony

and ML indicated a high degree of uncertainty for the ancestral states of 2 nodes (V.

hebeclada and V. rhemanniana; V tortilis and sister clade).

Seven species (18%) had domatia and were restricted to three clades (Clade A,

D, and E; Fig. 4-3b). Parsimony (Steps=6; Consistency Index=0.3, Retention Index=0.0)

and ML (high degree of uncertainty in ML values. (Mk1 est.: rate=0.139, -log

64

Likelihood=19.01) character reconstructions both showed a high degree of homoplasy in domatia presence. Parsimony suggested a single origin for domatia in Clade F (V. zanzibarica and V. seyal); however, the difference in likelihood values was not significant in ML reconstruction (-log L domatia presence=19.14, domatia

absence=21.12). There was ambiguity for domatia presence or absence in the common

ancestor of Clade A in the ML reconstruction (Fig. 4-3b).

Discussion

Phylogenetic Relationships of African Vachellia

While the plastid phylogeny was unable to resolve many of the relationships

among African Vachellia, there were a number of clades and interspecific relationships

that showed strong support (Figs. 4-2 and 4-4). The identification of these clades in

previous molecular work corroborates these relationships (Bouchenak-Khelladi et al.

2010). While plastid DNA can provide useful markers for resolving phylogenetic

relationships (Soltis et al. 1998), single genome trees can be biased due to

introgression (Rieseberg and Soltis 1991, Heede et al. 2003). As Vachellia hybrids have been previously described (Ali and Qaiser 1980, Ebinger and Seigler 1992), it is possible that at least some of the relationships in this phylogeny do not accurately reflect interspecific relationships. The internal transcribed spacer (ITS) region of nuclear ribosomal DNA (nrDNA) has been shown to be a useful nuclear marker in resolving interspecific relationships (Baldwin 1992, Brown et al. 2010). These data are not currently available for Vachellia, and this was outside the scope of this study. Future work incorporating ITS and/or other nuclear markers will help address whether plastid

and nuclear trees are congruent.

65

Independent Origins of Spine Form and Domatia in Vachellia

Data from this study suggest that spine form (e.g., straight, recurved and mixed

spine morphologies) and domatia presence have arisen multiple times in African

Vachellia, with straight spines and domatia absence as the likely ancestral states.

Incorporating evolutionary history (i.e., phylogeny) is fundamental to assessing where and when traits have evolved, since character states in related species may not be independent (Felsenstein 1985b). Models of ancestral trait reconstruction taking topology and time into account allow for more precise estimates of how characters have evolved in related taxa (Schluter et al. 1997). Previous work has theorized convergence among ant-acacias in the Americas (Janzen 1974a), and this study is the first to explicitly test domatia evolution in Vachellia within a phylogenetic framework.

Additionally, distinct spine morphologies exist among African Vachellia, and these

appear to have evolved more than once. Spine form and domatia presence are

associated with defense effectiveness against vertebrate browsers (Cooper and Owen-

Smith 1986, Belovsky et al. 1991, Stapley 1998, Goheen and Palmer 2010), and data

from this study support a pattern of convergence for these structural defense traits

within African Vachellia.

Straight spines are the most common spine morphology among African

Vachellia, and the results presented here suggest that straight spine morphology is the

ancestral state. Recurved spines are not present in American Vachellia (Vassal 1981)

nor in related taxa with stipular spines within Mimosoideae (Robinson and Harris 2000).

This provides further support for the hypothesis that recurved spines evolved from

ancestors with straight spines. V. reficiens is the only species with only recurved spines,

and the other 5 species have mixed, straight and recurved forms. Based on the current

66

phylogeny, it is unclear whether V. reficiens may have evolved from an ancestor with mixed or straight spines. Overall there were 6 independent origins of recurved or mixed spine forms estimated in this study, but most species with recurved spines were restricted to one clade (Clade D; Fig. 4-3a), the only exception being V. ancistroclada

(Clade E; Fig. 4-3a). Ancestral state reconstructions are sensitive to the frequency of state changes (Schluter et al. 1997), and as spine form states tended not to cluster within Vachellia this may affect the reliability of our estimates. Additionally, the molecular phylogeny of Vachellia lacked resolution, and the resulting Majority

Consensus tree contained multiple polytomies. Ancestral state reconstruction in

Mesquite treats polytomies as hard polytomies; in other words, it treats these nodes as actual relationships rather than reflecting uncertainty. Future work utilizing molecular markers with greater sequence divergence (e.g., nuclear DNA) may reveal more robust interspecific relationships within Vachellia, allowing for a more precise assessment of how many times recurved spine forms have evolved.

The convergence of ant-associated domatia in American Vachellia has been previously proposed (Janzen 1974a), and the data here support a similar pattern for the

African clade. Domatium presence showed a more equal coverage relative to spine form within Vachellia, with representatives spread roughly evenly between three clades

(Fig. 4-3b). Nevertheless, ancestral reconstructions estimated domatia arising 6 or 7

times, which suggests maximal or nearly maximal homoplasy. The ancestor of Clade F

is ambiguous. Parsimony reconstruction supported a common origin for species of

Clade F, and while ML estimates showed a similar pattern (moderate support for an

ancestor with domatia, 70% proportional likelihood) this was not significant. As with

67

spine form, lack of phylogenetic resolution may reduce the accuracy of trait evolution estimates. While this study increased sampling of Vachellia with domatia, there are 8

other species that have domatia. Incomplete taxon sampling can lead to inaccurate

phylogenetic estimation (Zwickl and Hillis 2002, Wiens 2006), which could also lead to

inaccurate ancestral state reconstruction. This possibility cannot be rejected here and

future studies with increased taxon sampling would resolve this potential source of

error. Although domatia morphology was not considered in this study, there is

phenotypic variation in domatia morphology among species (i.e., fusiform, globular). As

domatia presence is clustered among different clades, it would be worthwhile to

examine whether phenotypic variation is more similar in closely related species

compared to more distantly related clades.

Evolutionary Convergence of Structural Defense Traits

Plants are expected to converge on defense traits in response to similar abiotic

and biotic selection pressures (Janzen 1974b, Agrawal and Fishbein 2006, Fine et al.

2006). Structural defenses such as armament are more generalized defenses, broadly

effective against large vertebrate herbivores. However, not all spine forms are equal in

their defense effectiveness (Cooper and Owen-Smith 1986, Belovsky et al. 1991).

Straight spines are thought to be more effective against larger browsers (e.g., giraffe),

while recurved spines are more effective against smaller browsers (i.e., dik dik, impala).

If the evolution of spine form is related to these functional differences, then differences

among herbivore communities may help to explain the distribution of plants with distinct

spines forms. Large herbivores have been shown to influence Vachellia distribution

(Greve et al. 2012), but the relationship between variation in herbivore communities and

spine defenses among Vachellia has not be tested. Given that Vachellia recurved spine

68

forms are substantially smaller (i.e., mass) than straight forms, it may be that armament form is related to defense strategy (e.g., recurved spines are low investment) rather than herbivore-specificity per se. Similarly, domatia and the defensive ants associated with them are thought to be particularly effective against mammalian herbivores, but the costs associated with ant partners can be very high (Goheen et al. 2007). If variation in spine form and domatia reflects alternative investment strategies (e.g., domatia > straight spines > recurved spines), then resource availability along with variation in herbivore communities may better explain variation in spine traits (Coley et al. 1985,

Fine et al. 2006). Future work examining how spine defenses (e.g., spine form, spine size, domatia presence) relate to distribution of Vachellia along resource gradients would lend credence to the role of abiotic factors in shaping spine defenses.

69

Table 4-1. Table of taxa, voucher information, accession numbers for GenBank (asterisks are those taxa sequenced for this study).

Voucher Voucher Voucher Species trnL-F matK psbA-trnH (herbarium) (herbarium) (herbarium) Vachellia abyssinica Benth. * * * * * * Vachellia ancistroclada Brenan * * * * * * Vachellia arenaria OM1048 OM1048 Schinz (JRAU) GQ872257 OM1048 (JRAU) GQ872212 (JRAU) GQ872302 Vachellia borleae Burtt Davy RL1308 (JRAU) GQ872258 RL1308 (JRAU) GQ872214 RL1308 (JRAU) GQ872303 Vachellia bussei Sjostedt * * * * * * Vachellia davyi N.E.Br. RL1319 (JRAU) GQ872263 RL1319 (JRAU) GQ872219 RL1319 (JRAU) GQ872308 Vachellia drepanolobium Sjostedt * * * * * * Vachellia elatior Brenan * * * * * * Vachellia erioloba (E. Mey.) P. Hurter RL1298 (JRAU) GQ872265 RL1298 (JRAU) GQ872221 RL1298 (JRAU) GQ872310 Vachellia etbaica Schweinf. * * * * * * Vachellia exuvialis Verd. RL1284 (JRAU) GQ872267 RL1284 (JRAU) GQ872223 RL1284 (JRAU) GQ872312 Vachellia gerrardii Benth. RL1321 (JRAU) GQ872269 RL1321 (JRAU) GQ872225 RL1321 (JRAU) GQ872314

70

Table 4-1 continued. Voucher Voucher Voucher trnL-F matK psbA-trnH (herbarium) (herbarium) (herbarium) Vachellia grandicornuta RL1286 RL1286 Gerstner (JRAU) GQ872271 RL1286 (JRAU) GQ872227 (JRAU) GQ872316 Vachellia gummifera Willd. - - - - Aparaicio. s.n. HE602486.1 Vachellia haematoxylon CANB 615730 Willd. (JM386) EU440024.1 Unpublished AF523189.1 - - Vachellia OM1034 OM1034 hebeclada DC. (JRAU) GQ872272 OM1034 (JRAU) GQ872228 (JRAU) GQ872317 Vachellia hockii De Wild. * * * * * * Vachellia karroo Hayne JM255 AF522972 CANB 615590 AF274137 JM255 AF524992 Vachellia kirkii RL1307 RL1307 Oliv. (JRAU) GQ872275 RL1307 (JRAU) GQ872231 (JRAU) GQ872319 Vachellia kosiensis P.P.Sw. ex Coates RL1305 RL1305 Palgr. (JRAU) GQ872276 RL1305 (JRAU) GQ872232 (JRAU) GQ872320 Vachellia lahai Benth. * * * * * * Vachellia RL1285 RL1285 luederitzii Engl. (JRAU) GQ872278 RL1285 (JRAU) GQ872234 (JRAU) GQ872322 Vachellia nebrownii Burtt OM1050 OM1050 Davy (JRAU) GQ872281 OM1050 (JRAU) GQ872236 (JRAU) GQ872325 Vachellia nilotica RL1302 (L.) Delile (JRAU) GQ872283 RL1302 (JRAU) GQ872238 BREMN1 FJ808555 Vachellia nubica Benth.. * * * * * *

71

Table 4-1 continued. Voucher Voucher Voucher Species trnL-F matK psbA-trnH (herbarium) (herbarium) (herbarium) Vachellia ormocarpoides RL1293 RL1293 P.J.H.Hurter (JRAU) GQ872284 RL1293 (JRAU) GQ872239 (JRAU) GQ872327 Vachellia permixta Hurter J.2 Hurter J.2 Hurter J.2 Burtt Davy (JRAU) GQ872285 (JRAU) GQ872240 (JRAU) GQ872328 Vachellia reficiens RL1297 RL1297 Wawra (JRAU) GQ872287 RL1297 (JRAU) GQ872242 (JRAU) GQ872330 Vachellia rhemaniana Schinz - - RL1288 (JRAU) GQ872243 - - Vachellia robbertsei P.P.Sw. RL1289 RL1289 ex Coates Palgr. (JRAU) GQ872288 RL1289 (JRAU) GQ872244 (JRAU) GQ872331 Vachellia robusta subsp. clavigera RL1316 RL1316 (E.Mey.) Brenan (JRAU) GQ872289 RL1316 (JRAU) GQ872245 (JRAU) GQ872332 Vachellia sekhukhuniensis RL1296 RL1296 P.J.H.Hurter (JRAU) GQ872291 RL1296 (JRAU) GQ872247 (JRAU) GQ872334 Vachellia seyal Delile * * * * * * Vachellia OM1029 OM1029 sieberiana DC. (JRAU) GQ872294 RM02 (JRAU) GQ872250 (JRAU) GQ872337 Vachellia RL1294 RL1294 stuhlmannii Taub. (JRAU) GQ872296 RL1294 (JRAU) GQ872251 (JRAU) GQ872339 Vachellia swazica RL1327 RL1327 Burtt Davy (JRAU) GQ872297 RL1327 (JRAU) GQ872252 (JRAU) GQ872340 Vachellia tortilis (Forssk) P. Hurter RL1290 RL1290 & Mabb. (JRAU) GQ872298 RL1290 (JRAU) GQ872253 (JRAU) GQ872341

72

Table 4-1 continued. Voucher Voucher Voucher Species trnL-F matK psbA-trnH (herbarium) (herbarium) (herbarium) Vachellia xanthophloea RL1291 RL1291 Benth. (JRAU) GQ872300 RL1291 (JRAU) GQ872255 (JRAU) GQ872343 Vachellia zanzibarica (S.Moore) Taub. * * * * * * Parkia timoriana DM 265 DM 265 (DC.) Merr. (MELU) AF195682 DM 265 (MELU) AF523091 (MELU) AF195719 Mimosa tenuiflora (Willd.) Poir. 615541 (CANB) AF522943 615541 (CANB) AF274120 615541 (CANB) AF524963

73

Table 4-2. List of African Vachellia in this study with spine form and domatia traits.

# Species Spine Form Domatia 1 Vachellia abyssinica Benth. Straight Absent 2 Vachellia ancistroclada Brenan Mixed Absent 3 Vachellia arenaria Schinz Straight Absent 4 Vachellia borleae Burtt Davy Straight Absent 5 Vachellia bussei Sjostedt Straight Present 6 Vachellia davyi N.E.Br. Straight Absent 7 Vachellia drepanolobium Sjostedt Straight Present 8 Vachellia elatior Brenan Straight Present 9 Vachellia erioloba (E. Mey.) P. Hurter Straight Present 10 Vachellia etbaica Schweinf. Mixed Absent 11 Vachellia exuvialis Verd. Straight Absent 12 Vachellia gerrardii Benth. Straight Absent 13 Vachellia grandicornuta Gerstner Straight Absent 14 Vachellia gummifera Willd. Straight Absent 15 Vachellia haematoxylon Willd. Straight Absent 16 Vachellia hebeclada DC. Mixed Absent 17 Vachellia hockii De Wild. Straight Absent 18 Vachellia karroo Hayne Straight Absent 19 Vachellia kirkii Oliv. Straight Absent 20 Vachellia kosiensis P.P.Sw. ex Coates Palgr. Straight Absent 21 Vachellia lahai Benth. Straight Absent 22 Vachellia luederitzii Engl. Mixed Present 23 Vachellia nebrownii Burtt Davy Straight Absent 24 Vachellia nilotica (L.) Delile Straight Absent 25 Vachellia nubica Benth. Straight Absent 26 Vachellia ormocarpoides P.J.H.Hurter Straight Absent 27 Vachellia permixta Burtt Davy Straight Absent 28 Vachellia reficiens Wawra Recurved Absent 29 Vachellia rhemaniana Schinz Straight Absent 30 Vachellia robbertsei P.P.Sw. ex Coates Palgr. Straight Absent 31 Vachellia robusta subsp. clavigera (E.Mey.) Brenan Straight Absent 32 Vachellia sekhukhuniensis P.J.H.Hurter Straight Absent 33 Vachellia seyal Delile Straight Present 34 Vachellia sieberiana DC. Straight Absent 35 Vachellia stuhlmannii Taub. Straight Absent 36 Vachellia swazica Burtt Davy Straight Absent 37 Vachellia tortilis (Forssk) P. Hurter & Mabb. Mixed Absent 38 Vachellia xanthophloea Benth. Straight Absent 39 Vachellia zanzibarica (S.Moore) Taub. Straight Present

74

Figure 4-1. Examples of spine morphology in Vachellia. Top panel from left to right: Recurved, V. reficiens; Recurved and straight, V. tortilis; Straight, V. hockii. Bottom from left to right: Swollen spines, V. zanzibarica, V. horrida, V. drepanolobium.

75

Figure 4-2. Maximum Likelihood phylogram of African Vachellia. Branch lengths represent sequence divergence. Bootstrap values higher than 50% are shown at nodes. Branch lengths are proportional to the amount of change that occurs on them.

76

Recurved Mixed Straight

A

Figure 4-3. Ancestral state reconstructions of spine form and domatia presence. Parsimony (left panel) and maximum likelihood with Mk1 model (right panel) reconstructions of A) spine form and B) domatia presence based on majority rule bootstrap consensus tree. Nodes of uncertain states are denoted by multiple colors, with likelihoods illustrated as proportional likelihoods.

77

Present Absent

B

Figure 4-3. Continued

78

Figure 4-4. Majority rule bootstrap consensus cladogram showing clades within Vachellia. *Recurved or mixed spine morphology, † Domatia presence.

79

CHAPTER 5 SPINES, PRICKLES AND THORNS, OH, MY! PHYSICAL ARMAMENTS AS STAND UP TRAITS FOR STUDYING PLANT DEFENSES

Background

Physically armed plants are well-represented among embryophytes, and if you wished, you could choose a fern, bryophyte, cycad, water lily, monocot, or eudicot to prick your finger on, albeit more painfully in some. While these structures can have physiological roles (e.g., ameliorate heat stress) or provide mechanical support (e.g., vine growth), they most often function to defend against herbivores. Defenses buffer plants against consumers, and in doing so mediate numerous plant-animal and plant- plant interactions. This along with their adaptive nature has made defense traits a template for understanding wide-ranging aspects of community ecology and evolutionary biology (Rosenthal and Berenbaum 1992a, b). Despite the fact that armament defenses are geographically and taxonomically widespread, there exist major and fundamental gaps in our understanding of these traits: How are they shaped by abiotic and biotic factors? What are their costs relative to the benefits? Why is there morphological variation in armament? Do different sizes or forms reflect defensive function or herbivore specificity? Do resistance or tolerance traits consistently co-vary with armament presence or investment? The answers to these questions can shed light on the extent to which these traits influence interspecific and community-level interactions, and would broadly inform ecological and evolutionary defense theories.

Armaments are pointed, physical structures that develop from various plant tissues or organs. Prickles, spines and thorns are all examples of armament, and their defensive roles are primarily against vertebrate herbivores, the focus of this chapter. In contrast, some structures such as trichomes are small, epidermal structures that can

80

defend against invertebrate herbivores (review in Hanley et al. 2007) and that have well- established roles in ecophysiology (Ehleringer et al. 1976, Rodríguez et al. 1984,

Southwood 1986). These and other invertebrate defenses are beyond the scope of this chapter and the reader is directed to a recent review on the topic (for an excellent review, see Hanley et al. 2007).

Plants with armament are found in a broad range of terrestrial environments, from temperate to tropical latitudes, and can comprise the majority of some woody plant communities (Cowling and Campbell 1980, Milton 1991, Young et al. 1996). The high incidence of armed plants in arid areas is readily apparent in the savannas of East or

South Africa, the deserts of North America (Sonoran, Chihuahuan, Mojave), the caatinga of Brazil, the heathlands of the Mediterranean, the desert or scrubland of

Southern Asia (Thar desert, Deccan Thorn Scrub Forest), or the kwongan heathlands of

Western Australia. While there have yet to be formal phylogeographic tests, there are several explanations for why this pattern may emerge. Arid environments often have low canopy cover, with nutrients and water limiting growth more than photosynthesis, thus the cost of carbon-based defenses should be minimal (Bryant et al. 1989). Plants growing in low resource environments are predicted to have slower growth rates and higher costs of tissue replacement, and therefore relatively high levels of constitutive and immobile defenses (Coley et al. 1985, Herms and Mattson 1992). Also, in arid environments if plants are more limited by nutrients than carbon, then plants should have carbon-based defenses (Bryant et al. 1983). However, these hypotheses and predictions do not always suffice for explaining the distribution of armed plants. Armed plants can be found in moist and relatively resource rich environments or microhabitats

81

(Campbell 1986, Milton 1991, Grubb 1992), and often exhibit plasticity. Armed plants can also have relatively fast growth rates (Bryant et al. 1989, Grubb 1992), suggesting competition may be an important selective force (Herms and Mattson 1992). Finally, the

absolute and relative costs of armament are not well understood; given that the

development of these structures would necessitate both carbon and nitrogen, it is not

apparent how nutrient and water limitation should influence defense allocation. Grubb

proposed the Scarcity-Availability Hypothesis (SAH) to explain the incidence of

armament, as well as chemical defenses; the SAH predicts that defenses will be

present based on the temporal or spatial scarcity and the accessibility of nutritious

growth to herbivores (Grubb 1992). While SAH qualitatively aligns with armament

incidence, there have been no other studies utilizing the SAH framework. On the whole,

our knowledge of how armament conforms to classic theories of plant defense is poor,

and there is a need to better understand the abiotic and biotic factors that influence and

select for armament traits.

Armament deters and reduces biomass loss primarily against vertebrate

herbivores. As a resistance trait, armament is certainly more generalized than

numerous chemical traits, but there is evidence that armament form may reflect

herbivore specificity based on size or feeding strategy (Cooper and Owen-Smith 1986,

Belovsky et al. 1991, Seifert and Palmer in press). Herbivores have been shown to

exert selective pressure on both specialized (furanocoumarins, Zangerl et al. 2008) and

broad-spectrum (thorns, Gomez and Zamora 2000) resistance traits. Armaments should

be as relevant as phytochemicals to understanding ecological and evolutionary

perspectives of plant defense. Armament traits also offer advantages for two basic and

82

intrinsic reasons: they are easily measured and readily manipulated. For ecological studies, this allows for manipulative experiments ranging from questions about herbivore-specificity to broader, community-level responses. Furthermore, herbivores affected by armament are themselves relatively easy to manipulate (Cooper and Owen-

Smith 1986, Young 1987). Armament may also provide an historical record of past defense investment, as these structures are often retained on plants for several years.

Finally, the ability to accurately record and quantify defense traits is advantageous in both ecological and evolutionary work. Armament traits are also a prime candidate model system for addressing patterns of defense evolution at a range of taxonomic scales, as armament is widespread and highly variable among plant taxa and highly variable. Defensive structures, such as armament, are highly tractable traits to test plant

defense theory, and there exist substantial gaps in our knowledge about them.

The intent of this chapter is: (1) to provide background on armament traits, (2) to review our current knowledge of armament traits, as they relate to the ecology and evolution of plant defense, (3) to identify key gaps in our knowledge and (4) to develop hypotheses for future research on armament within the framework of plant defense theory.

Armament Morphology and Development

While armaments have been categorized based on their developmental origins a

detailed tissue analysis would offer a welcome resolution to the relationships among

these structures. Nevertheless, generally speaking, prickles are defined as outgrowths

of the epidermis; leaves are modified to form spines, and thorns are derived from

branch meristems. Plants are widely appreciated for their diversity in form, and the

range of observed armament morphologies and arrangements exemplifies this (Fig. 5-

83

1). As there is little known about the molecular or genetic basis of armament development, here I review the anatomical and morphological characters associated

with armament morphogenesis and development.

Prickles

Prickles are the most taxonomically widespread form of armament, with

examples in basal (ferns, cycads, water lily) as well as more derived plant lineages

(eudicots). Prickles are epidermal outgrowths and can be present on the surface or

edges of leaves, reproductive parts, and stems (Fig. 5-1). At maturity, these structures

are highly lignified and non-vascularized, and can be hollow or solid depending on plant species or prickle location (Briand and Soros 2001, Kellogg et al. 2011). In some cases,

cell wall thickening or sclerification only occurs in the apical and adjacent epidermal

cells rather than throughout (Carlquist 1962). Prickle morphogenesis appears to initiate

with a proliferation of cells in either epidermal or cortical layers with cell proliferation and

expansion thought to subsequently push cells outwards. Towards the final stages of development cells are elongate and lignify along the distal-proximal axis (Carlquist

1962, Kellogg et al. 2011). While the signals responsible for initiating and regulating prickle development are not documented, prickles have in some cases been linked to non-glandular as well as glandular trichome development (Carlquist 1962, Kellogg et al.

2011). Interestingly in Rubus, prickles only develop on cultivars with glandular trichomes

(Coyner 2005). Based on distal to proximal differentiation of prickles (i.e., lignification),

Kellogg et al. (2011) suggest that directional signaling within developing glandular trichomes may signal tissues to ultimately develop into prickles. This explanation might also apply to prickles of Lobeliaceae, which form from the ground meristem below unicellular trichomes (Carlquist 1962). In cases where prickles are derived from

84

trichomes, studies exploring whether similar developmental mechanisms hold for prickles would be of particular interest (e.g., patterning; Schnittger et al. 1999). The link between trichomes and prickles is currently anecdotal as there have been few comparative anatomical or developmental studies, and even closely related species show significant differences in underlying prickle anatomy (i.e., rose and raspberry prickles contain epidermal and cortical tissue, while blackberry prickles are epidermal only; Kellogg et al. 2011). This underscores the need for thorough cross-species examinations of prickle cellular anatomy and development, despite the superficial similarities of these structures.

Spines

There are myriad leaf or leaf-derived structures (e.g., stipules, petals, bracts, cataphyll) that form the armaments collectively referred to as spines. Spines can be entire, in that the whole organ is morphologically a spine, or marginal (marginal spines as outgrowths within the form of a leaf [e.g., Ilex] versus prickles as superficial epidermal outgrowths; Fig. 5-1). The few studies examining spine morphogenesis and development have focused on plants with entire spines. Among species, spines have shared properties including: terete cross-sectional shape, acropetal growth, asymmetric cell division (longitudinal along proximal-distal axis), basipetal differentiation, simplified cellular anatomy relative to vegetative leaves and a high degree of sclerification at maturity (Mauseth 1977, Boke 1980, Pabón-Mora and González 2012). There are also interspecific differences. For example, cactus spines are highly simplified; all protodermal and mesophyll cells develop into fibres, and at no point during development are vasculature or other specialized cells apparent (e.g., guard cells, mucilage cells;

Mauseth 1977, Boke 1980). Berberis spines on the other hand have defined spongy

85

and palisade mesophyll layers, as well as vasculature, and many of these cells become sclerified throughout maturity (Pabón-Mora and González 2012).

The stipular spines of acacias are somewhat intermediate, with vasculature but

no apparent differentiation between mesophyll cells (MG Seifert unpublished data). As

these distinctions are related to cell fate, this may suggest that the timing or type of

signals following initial basal meristem growth may direct cell differentiation throughout spine development. Nevertheless, the parallels between species provide insight into potential regulation of spine development (Pabón-Mora and González 2012). Of additional interest is how and when spine identity is determined? Plants with entire spines also produce vegetative leaves, although some are microscopic (e.g., cactus).

Based on hormone-induced changes (e.g., cytokinin induces short shoot growth with vegetative leaves, and gibberellic acid induces short shoot growth with spines),

Mauseth (1977) suggested that hormones act on undetermined meristems to induce changes in the meristem (i.e., mitotic activity) that subsequently produce a leaf or spine.

Understanding the molecular pathways that regulate the development of spines versus leaves will be important to elucidating how plants compartmentalize the development of two very different structures from similar meristems, and would also inform how defense

signaling may affect these processes where spines are inducible.

Thorns

Derived from shoots, thorns are distinct from spines and prickles as they are

formed from indeterminate meristems. The phenomenon of switching from determinate

to indeterminate growth is a topic of broad interest in developmental biology, and within

plant development much work has focused on indeterminate tendril and floral

meristems. Interestingly, thorns provide an alternative and fairly simple model to further

86

our understanding of this developmental problem. There are many thorny plants and numerous that are well known: Gleditsia, Citrus spp., Bougainvillea spp., Ulex

europaeus. At maturity, thorns are composed of branch-like cellular components, and in

some cases with vascular traces apparent from undeveloped leaf primordia (Blaser

1956, Bieniek and Millington 1967, Posluszny and Fisher 2000). However, in

comparison to non-thorny branches there are differences in the size of certain tissue

layers (i.e., pith) or cellular components (i.e., thinner vessels, thicker cell walls of vessel

fibers) (Posluszny and Fisher 2000). The timing of the switch from determinate to

indeterminate growth (i.e., mitotic activity in meristem) appears to vary with plant

species. In Ulex europeans, apical meristem activity appears to cease relatively early

(Bieniek and Millington 1967), while Gleditsia thorns show high levels of mitotic activity

throughout the meristem; as the shoot tip becomes conical due to growth, meristem

activity ceases (Blaser 1956). As with spines and prickles, elongation and sclerification

progress basipetally or from the leaf apex toward the base (Blaser 1956, Bieniek and

Millington 1967, Posluszny and Fisher 2000).

Histological analysis of armament anatomy and descriptive morphogenesis along

with known plant developmental mechanisms can provide a valuable starting point for

future research avenues on armament development. For example, entire spines share

similarities in their underlying anatomy to leaf counterparts; however, their form is

radialized. Delimitation of adaxial and abaxial identity plays a key role in laminar leaf

and stipule development, with the loss of adaxial identity (i.e., adjacent to the shoot

apical meristem) resulting in the absence of adaxial cell types and ultimately an

abaxialized, radially symmetric structure (Waites and Hudson 1995, Tattersall et al.

87

2005). Similar mechanisms could drive spine formation, as they lack laminar outgrowth and are radial in symmetry. Understanding the mechanisms underlying the development of these structures is a key component in assessing how environmental changes may

drive phenotypic change in armament structure. Moreover, as these developmental

mechanisms are explored in the context of natural inter- and intraspecific variation a

deeper understanding of how environmental factors have shaped these defensive traits

among plant taxa can be achieved.

Armament as an Effective Form of Resistance

Armament is widely cited as an effective form of defense against vertebrate

herbivores. However, we know relatively little about the functional significance of the

wide variation found within and among plant species (Fig. 5-1). Here, I review our

current understanding of the benefits associated with armament as well as how

herbivores respond to armament.

Defensive Benefits of Armament

Establishing the functional role and fitness consequences of various forms of

armament is fundamental to understanding its role in defense. Fitness benefits are

preferably measured in terms of reproduction; however, growth or measures of relative

biomass loss have more often been used as surrogates in studies on armed plants.

Studies manipulating the presence and absence of resistance traits still provide clear

evidence for their defensive function against herbivory. Feeding trials as well as in situ

manipulations show a positive effect of armament (e.g., presence vs. absence) in

reducing biomass losses to vertebrate herbivores in tropical and temperate plant

species (Cooper and Owen-Smith 1986, Milewski et al. 1991, Rohner and Ward 1997,

Stapley 1998, Takada et al. 2003, Wilson and Kerley 2003, Cash and Fulbright 2005,

88

Seifert and Palmer in press, but see Rafferty et al. 2005). Armament morphology is highly variable among and within plant species, and there is evidence that this morphological variation (e.g., density, size or form) is related to resistance. Studies examining various plant species have documented reductions in biomass loss or browsing rates with increasing spine densities (Gowda 1996), spinescence (e.g., spine length and number; Obeso 1997) and spine morphological combinations (Seifert and

Palmer in press). These types of variation may also reflect defense investment as they are usually correlated with increasing armament biomass (i.e., higher spine densities will also be greater in mass cm-1). Contrary to this expectation, Seifert and Palmer (in

press) found that straight spine length (i.e., investment) had no effect on biomass

removal against two different browsers. Disentangling trait variation from investment will

be important for understanding whether distinct morphologies (e.g., recurved, straight;

single, bi-furcating) reflect herbivore specificity.

Cross-species comparisons of armed and unarmed plant taxa confirm the defensive benefits of armament. Armed plants tend to lose less biomass to herbivory

than unarmed plants (Belovsky et al. 1991), and herbivory differs among plant species

that vary in armament (Cooper and Owen-Smith 1986, Belovsky et al. 1991, Wilson and

Kerley 2003). These studies typically quantify herbivore responses (i.e., bite size, bite

rate, instantaneous intake rate), and in some cases total biomass loss. Yet assessing

the defensive function or relative benefits of armament in terms of biomass loss may be

problematic. Armed plants tend to have less vegetative (leaf) biomass per branch

diameter (Belovsky et al. 1991), and within species, higher levels of armament (more

biomass) are often negatively related to primary biomass (Obeso 1997, Seifert and

89

Palmer in press). Thus, lower levels of biomass loss could reflect armament effectiveness or differences due to initial biomass. For example, herbivores select smaller diameters on thorny relative to non-thorny species; however, thorny plants have

smaller diameters on average relative to non-thorny species (Belovsky et al. 1991). It is

not fully understood whether the negative relationship between armament and primary

biomass is defensive (e.g., reducing the “value” of branches or plants to herbivores; see

“Effects of Armament on Herbivore Feeding Behavior”) or simply reflective of internal

trade-offs (see “Allocation Costs”). Nonetheless, interspecific comparisons of defensive

forms among multiple herbivore species have led to the hypothesis that armament form

is adaptive, as recurved spine forms tend to be more effective against smaller leaf-

feeding herbivores while straight spines are more effective against larger pruning

herbivores (Cooper and Owen-Smith 1986). A recent test of this hypothesis yielded

mixed results. Branches of V. etbaica with straight spine forms retained proportionally

more biomass following herbivory by small and large herbivores (Seifert and Palmer in

press). While straight spines were relatively more effective against a large herbivore

(e.g., camels remove more biomass overall relative to goats), straight spines also

reduced proportional leaf loss against smaller herbivores. Knowledge of the relative

fitness costs of different types of biomass loss would improve our understanding of

armament variation in relation to herbivore specificity. Additionally, studies assessing

how plant size and age influence susceptibility to different browsers will help illuminate

the temporal contexts in which armament is most beneficial.

Effects of Armament on Herbivore Feeding Behavior

Armament primarily functions as a defense against large vertebrate herbivores

(but see Cooper and Ginnett 1998), which exert strong effects on plant biomass,

90

productivity and composition (reviewed in Augustine and McNaughton 1998). Herbivore foraging is hierarchical, and bite and patch-level feeding can have significant consequences for larger-scale foraging patterns (Shipley 2007). Bite size and bite rate

jointly determine an animal’s instantaneous intake rate (IIR, g min-1), and armament can

affect both. Armament structures most often reduce herbivore bite size, g bite-1 (Pellew

1984, Cooper and Owen-Smith 1986, Belovsky et al. 1991, Illius et al. 2002, Wilson and

Kerley 2003), although there can also be negative effects of armament on handling time

(e.g., increase cropping time, min bite-1 or decrease processing rate, g min-1; Illius et

al. 2002). There is also variation in the context in which armament affects herbivores.

First, herbivore responses depend on armament variation and herbivore size or feeding

mode (Belovsky et al. 1991, Illius et al. 2002, Wilson and Kerley 2003). Second,

herbivores forage based on ingestion and digestion constraints, and the nutritional

quality of armed branches or plants explain at least some of the variation in herbivore

feeding (Belovsky et al. 1991, Wilson and Kerley 2003). Lastly, the relative abundance

and distribution of plants with different armament forms influence patch-level foraging,

but there have been few theoretical or empirical tests assessing how this heterogeneity

influences relative or total utilization of armed versus unarmed plants (but see Owen-

Smith 1993, Belovsky and Schmitz 1994).

Armament directly influences bite size or rate if (a) herbivores are avoiding the

painful effects of armament or (b) armament structurally separates biomass in a way

that restricts accessibility in a given bite. Accounts of scarring in the mouthparts and

upper digestive organs (i.e., esophagus) of wild browsers suggest potential negative

effects of ingesting armed branches (Barnard and Hassel 1981, Hübschle 1988).

91

Further, herbivores are often cited as manipulating bites in a way that reduces the contact of spine tips and mouthparts (Cooper and Owen-Smith 1986). Armament has also been shown to provide modular separation of biomass (e.g., herbivores can only remove leaves at one node rather than two; Gowda 1996). Armament can influence bite size distinctly depending on herbivore size. In response to armament, large and medium-sized herbivores remove branches relatively smaller in diameter than they would otherwise select (Cooper and Owen-Smith 1986, Belovsky et al. 1991, Milewski et al. 1991). Medium to small herbivores often alter their feeding mode in the presence of armament, removing leaf clusters rather than taking branch bites (Cooper and Owen-

Smith 1986, Belovsky et al. 1991, Seifert and Palmer in press). Small herbivores can also be restricted in the amount of leaf mass they are able to remove per bite (Gowda

1996, Seifert and Palmer in press). Armament is not, however, always effective at reducing bite size. For example, smaller herbivores can maneuver around armament structures that are not densely arranged, allowing them to compensate for smaller bite sizes via higher bite rates (Gowda 1996, Illius et al. 2002). Likewise, shorter or recurved armaments are not always effective at reducing bite sizes of large herbivores (Seifert and Palmer in press). These observations, along with comparisons of the relative costs and benefits of armament (e.g., dry weight spine investment, dry weight biomass saved), have led to the suggestion that armament variation (e.g., form, density, size) may reflect potentially adaptive defense against certain herbivores (Cooper and Owen-

Smith 1986). While reasonable, this is likely an oversimplification. For instance, long straight spines are thought to be more effective against larger herbivores. However,

Seifert and Palmer (in press) found that goats were not able to compensate for reduced

92

bite sizes via increasing bite rate, and had lower IIRs when straight spines were present on V. etbaica branches. While the effects on herbivore IIR will be most pronounced when armament reduces both bite size and rate (Figs. 5-2c, f), the inability to compensate for reductions in either will still have substantial effects on IIR (Figs. 5-2b, e). Additionally, the potential bite size of an armed plant relative to an herbivore’s maximum or optimal bite size may influence the degree to which an herbivore is affected by armament (Figs. 5-2a-c vs. d-f). The context in which armament affects bite size and rate, collectively determining IIR, will better inform hypotheses of how armaments may align with herbivore size or feeding strategy.

The nutrient content and digestibility of plant material also affect herbivore foraging behavior, as energy intake must meet daily metabolic demands. If there are characteristics of armed branches that alter nutritional quality, then this may have additional effects on herbivore behavior. Here, I refer to these as “indirect effects”, as armament per se is not influencing herbivore responses. At a given branch diameter, armed branches tend to have fewer leaves and less leaf mass than unarmed branches, and as branch tissue is higher in fiber this would reduce the overall digestibility.

Belovsky et al. (1991) demonstrate how this could explain lower minimum acceptable branch diameters on armed versus unarmed plants among a range of herbivore sizes

(i.e., constraining digestion via lower turnover). These effects are important to consider particularly in cross-species comparisons when leaf size and digestibility co-vary with armament, and may help explain variability in herbivore responses among armed species (e.g., Cooper and Owen-Smith 1986). Ultimately, the efficacy of armament will depend on whether plants benefit directly or indirectly via reduced biomass losses. This

93

may be highly contingent on environmental context, as habitat heterogeneity can have strong effects on herbivore functional response (Spalinger and Hobbs 1992). For instance, spatial patterning of plants can influence herbivore intake rates, due to trade- offs in searching time and ingestion (Spalinger and Hobbs 1992), and the effect of spatial patterning may interact with patch level intake rates. Hobbs et al. (2003)

demonstrate that plant characteristics such as architecture or leaf mass affect bite size

and bite rate (e.g., cropping), which in turn determine intake rate. However, they also

show that plant spacing becomes an important determinant of intake rate above a

certain distance threshold (i.e., chewing time is less than time to encounter the next

bite). When intake rates are low within a patch (e.g., small bite sizes), the effects of

spatial patterning on foraging rates are expected to manifest at smaller spatial scales.

Plant species composition, palatability, and spatial association of palatable and

unpalatable plants also influence the relative consumption of plants (Owen-Smith 1993,

Rautio et al. 2008). Studies addressing the effects of spatial and temporal heterogeneity

would greatly improve our understanding of the efficacy of armament. Herbivore

functional responses and optimal foraging models provide ideal complementary

theoretical frameworks to make and test predictions of how herbivore foraging may

respond to spatial, temporal and morphological variation in armament.

Costs of Defensive Armament

Allocation Costs

Plants have access to finite resources, and allocation to secondary function

(defense, storage) should compete with primary metabolism (growth, reproduction) via

developmental or physiological mechanisms (Bryant et al. 1983, Coley et al. 1985,

Haukioja 1990, Herms and Mattson 1992). Reported correlations between growth-

94

related traits and armament span the entire spectrum, from positive (Gowda et al. 2003,

Ward et al. 2011), to negative (Obeso 1997, Rohner and Ward 1997, Zinn 2007), to

neutral (Hanley and Lamont 2002, Ward and Young 2002, Ward 2010). Correlations

among resistance traits are similarly variable: positive (Ward and Young 2002), negative

(Björkman and Anderson 1990, Twigg and Socha 1996, Hanley and Lamont 2002), or

neutral (Ward 2010, Ward et al. 2011). There are several key potential explanations for

this variation: (1) univariate trade-offs often reveal inconsistent patterns, (2) tolerance

can mask trade-offs between growth and resistance traits, and (3) more vigorous plants

often have higher levels of defense (Koricheva et al. 2004). A failure to account for

these factors may explain why the relationship between armament and growth can run

counter to theoretical predictions in some cases. For example, African acacias often

have strong re-growth responses suggesting that tolerance may be important in these

plants’ overall defense strategy (du Toit et al. 1990). If armament or reproduction comes

at the expense of tolerance (or storage), then a positive correlation between armament

and growth could emerge (Fig. 5-3b). In the case of plant vigor, positive correlations can

manifest despite underlying costs if variation in resource acquisition is more variable

than resource allocation (Fig. 5-3c; van Noordwijk and de Jong 1986). Finally, many

armed plants are long-lived, and it may be that trade-offs are not apparent on relatively

short temporal scales. In support of this notion, the most convincing evidence of the

allocation costs of defenses comes from longer-term datasets on defense and

reproduction (Gomez and Zamora 2002, Goheen et al. 2007). For example, V.

drepanolobium (formerly Acacia drepanolobium) trees that maintained high levels of

spine investment (i.e., spines on average >2 cm) had depressed levels of reproduction,

95

even decades after herbivore exclusion (Goheen et al. 2007). Micro- and macro- evolutionary approaches are greatly needed to examine the potential trade-offs associated with armament.

Plant armament structures can be extreme. In V. etbaica (formerly Acacia etbaica), I have measured stipular spines up to 15 cm in length with basal diameters of

the same size as the subtending internode (MG Seifert personal observation).

Furthermore, some branches have up to 80% of their dry weight invested in spines (MG

Seifert unpublished data). As armament production involves the same physiologically

demanding processes of growth, requiring carbon, water and nitrogen, armament allocation should reduce resources available for other forms of growth. However, to some extent armament will to be positively associated with plant growth and plant size

(Chapter 2). So, a key issue is whether the allometry of armament investment scales

differently than does, for example, total photosynthesis as plants grow larger. This will

be important to consider in metrics of armament investment and allocation (Figs. 5-a-d),

and deviations from these allometric relationships may also reveal allocation trade-offs

as well as developmental or biomechanical constraints of armament development

(Enquist and Niklas 2002). In addition to allocation costs and potential constraints,

armament may incur physiological costs after production. For example, in species with

prickles on leaf surfaces, photosynthetic rates may be reduced, and negative

relationships between prickles and leaf size may exacerbate this cost (Bazely et al.

1991). The extent to which armament allocation negatively affects plant growth and

physiology will have important implications for plant populations investing heavily in

these defense traits.

96

Opportunity and Ecological Costs

Opportunity and ecological costs are other potential costs of plant defense traits,

but there has been little research examining these costs in armed plants. Opportunity

costs reflect the competitive costs of investing in defense traits (i.e., competitive

displacement due to defense investment) relative to the cost of herbivory (i.e., replacing

lost tissue). These costs should be high (a) when competition between plants is high,

(b) for plants with fast growth rates and (c) in areas of high resource availability (Coley

et al. 1985, Herms and Mattson 1992), as it would be less costly for these plants to

replace tissue and more costly in terms of competitive displacement. Conversely, the

cost of herbivory for slow-growing plants in low resource environments will be high as

tissue replacement is resource limited. In these instances, opportunity costs are low and

plants invest heavily in defense. There is some evidence that spiny plants of eutrophic

(i.e., nutrient rich) savannas have higher growth rates than plants of dystrophic (i.e.,

nutrient poor) savannas (Bryant et al. 1989). However, whether competitive interactions,

herbivore pressure and resources interact to drive differences in armament defenses is

unknown. Studies performing reciprocal transplants would help assess the relative costs

of armament under different resource conditions (e.g., Fine et al. 2006), particularly if

both armament and herbivores are manipulated. There are defined predictions for such

an experiment. For instance, in a landscape with patchy resources, armed plants in high

vs. low resource patches should differ in growth, defense and herbivory: (1) all plants

growing in high resource patches should have higher growth rates but species that

normally grow in low resource areas will have lower growth rates, (2) species that

normally grow in low resource areas should also have higher levels of defense

compared to high resource species, (3) in the presence of herbivores and in the high

97

resource area, high resource plants should have high levels of tolerance and should out

compete transplanted slower-growing plants that normally grow in low resource areas,

(4) in low resource areas and in the presence of herbivores, transplanted plants that are

normally present in high resource areas will be outcompeted by slower-growing, low resource plants, and (5) in the absence of herbivores and independent of resources, high resource plants will out compete slower growing, low resource plants. Additionally, armament removal would assess the relative importance of these defenses between patch types.

Unlike opportunity costs, ecological costs are external costs resulting from a defense trait negatively affecting beneficial interactions. For example, resistance traits can negatively affect plant-pollinator interactions and consumer-specific induced resistance can affect subsequent interactions with a different subset of herbivores (Heil

2002). It is not well established whether these costs result in divergent selective pressures, however it is sensible to expect that this would depend on the ecological costs of defense relative to their benefits against herbivores. Little work has been done assessing the potential ecological costs of armament, but there is some evidence armament traits affect floral traits and pollination. Floral spines of Centaurea solstitialis

reduced visitation of nectar robbing pollinators, but not legitimate pollinators, which

resulted in lower seed set on flowers where spines were removed flowers (Agrawal et

al. 2000). In another study examining dual selection pressures of herbivores and

pollinators, populations of Hormathophylla spinosa show different patterns of selection

depending on herbivore pressure. Under high ungulate herbivory, selection for

increased thorn density was found, but in areas experiencing low herbivory pollinators

98

exerted selection for larger floral displays. Thorn density did not appear to reduce pollinator abundance but floral display increased rates of herbivory (Gomez and Zamora

2000), indicating that armament did not directly affect pollinators (i.e., no ecological

cost). However the negative relationship between floral display and thorn density

suggest that defense adaptations can be at odds with traits associated with positive

interactions. In a macro-evolutionary study of Hakea, Hanley et al. (2009) found that the

presence of spines was associated with pollinator identity (e.g., insect pollinated plants

possessed spines, while bird-pollinated plants possessed red flowers and cyanide-

based defenses), but that this did not appear to constrain selection (e.g., pollinators vs.

herbivores). Together these studies show that there is potential for armament to affect

pollinators, particularly when floral and armament traits are associated. The fact that

armament can be manipulated with relative ease makes these systems amenable to

experimental tests of the ecological costs of defense traits with respect to well-known

beneficial organisms (e.g., pollinators, rhizobia, mycorrhizae).

Armament Plasticity

Armament often shows plastic responses to herbivory, and there are alternative

mechanisms for how these responses may occur: (1) induction via plant defense

signaling, (2) ontogenetic shifts, and (3) altered physiology (e.g., CNBH, Source/Sink

Hypothesis). Inducible defenses are thought to have evolved as ways for plants to

minimize the cost of defense, particularly in cases where there is spatial or temporal

variation in herbivory (Karban and Baldwin 1997). Ontogenetic reversion in defense

strategy (i.e., investment) may reflect differential costs of defense or variation in

susceptibility throughout a plant’s lifetime (Boege and Marquis 2005). Trait plasticity

may also be the result of altered internal resource pools. These responses reflect how

99

internal resources pools in turn influence allocation (e.g., growth, reproduction, storage, defense), and can be associated with optimal defense allocation within a plant

(Rhoades 1979, Bryant et al. 1983, Coley et al. 1985, Haukioja 1990, Herms and

Mattson 1992).

Responses as Induced Defenses

Inducible strategies are expected to be less costly than those that are constitutive, and three criteria define inducible defense: herbivory induces a response in a resistance trait, the response alters herbivore feeding and reduces biomass loss and the response subsequently results in higher plant fitness (Karban and Baldwin 1997).

Numerous studies have documented changes in armament in response to natural or experimental herbivory, with herbivory increasing armament density (Myers 1987, White

1988, Bazely et al. 1991, Midgley and Ward 1996, Obeso 1997, Gomez and Zamora

2002, Cooper et al. 2003, Zhang et al. 2006), length (Midgley and Ward 1996, Gowda

1997, Obeso 1997, Young et al. 2003, Zhang et al. 2006, Zinn 2007) or form (Rohner and Ward 1997, Chapter 2). There have not been studies eliciting induced responses and examining subsequent differences in herbivory, but available data show that armament variation can increase defense efficacy (see “Defensive Benefits of

Armament”), supporting the criterion that induced responses must alter herbivore feeding behavior and biomass loss. Induced changes in armament are thought to be localized rather than systemic (Milewski et al. 1991, Young et al. 2003), and some have suggested that spatial rather than temporal variation in herbivory have led to the evolution of localized inducibility (Young et al. 2003). The fitness benefits and costs of armament induction are not well documented; there are currently two published studies.

Gomez and Zamora (2002) found that induced thorn densities in Hormathophylla

100

spinosa reduced fruit set and seed production. They did not detect a significant increase

in reproductive fitness in the presence of herbivores, but the authors suggest this was

due to low sample size and variation in degree of herbivore damage in their study.

Induced spine length responses are well documented in V. drepanolobium (Young

1987, Young and Okello 1998, Young et al. 2003), and continued investment in spine

defenses in the absence of herbivores had strong negative effects on V. drepanolobium

reproduction. Together these studies provide evidence that induced armament

responses exact fitness consequences. Disentangling the effects of biomass loss, tolerance or induced responses on plant fitness continues to be a challenge particularly in long-lived plants in the aforementioned studies. Armed plants with shorter life spans would be more amenable to isolating these distinct costs.

Inducible defenses are likely regulated via injury-induced stimulation of developmental signaling pathways, and linking armament responses to these pathways would lend credence to these as active defense responses (Rhoades 1979, Haukioja and Neuvonen 1985). There is no published work on defense hormones and armament induction, but there is anecdotal evidence that jasmonic acid regulates spine length induction (J.R. Goheen and T.M. Palmer pers. comm.). Studies examining armament responses to defense hormones as well as internal hormone concentrations in developing armament structures are greatly needed. These data may also help explain the spatial and temporal variation of armament induction as well as the asymmetry between induction and relaxation timing. Armament can be rapidly induced and requires a relatively long period of time to relax (Gomez and Zamora 2002, Young et al. 2003).

Temporal lags in relaxation have been suggested to be a function of herbivore risk

101

(Karban et al. 1997, Young et al. 2003), which may be related to herbivory or abiotic

cues indicating herbivore presence. Growth hormones are also presumably involved in

armament responses. For example, gibberellic acid and auxin (NAA) influence thorn

formation (i.e., apical conversion) in Ulex europaeus (Bieniek and Millington 1968).

Cytokinins have also been found to induce vegetative development of Opuntia

primordia, and in the same study gibberellic acid-treated primordia developed spines

(Mauseth 1977). While growth-related hormones are expected to be involved in

armament development, it is unknown how herbivore-induced changes in armament

growth and development are regulated

Ontogenetic Shifts in Armament

Throughout a plant’s lifetime there can be large shifts in resource allocation and herbivore pressure, and these may drive ontogenetic changes in defense allocation

(Boege and Marquis 2005). The association between plant age and armament

structures is well established. For example, juvenile Citrus are very thorny with thorns

somewhat persistent on mature trees towards the tree base (Cameron and Frost 1968).

For many armed taxa, the juvenile stage (i.e., sapling) is reported as “spinier” relative to mature plants (Peterken and Lloyd 1967, Bieniek and Millington 1968, Kozlowski 1971,

Boke 1980, White 1988, Brooks and Owen-Smith 1994, Rohner and Ward 1997, Gowda

and Palo 2003), although there are exceptions (Rooke et al. 2004). When assessing

ontogenetic shifts in defense, it should be noted that different relationships might

emerge depending on the metric used. For instance, in V. etbaica straight spine length

was positively correlated with plant age (i.e., size) and straight spine production (i.e.,

relative to recurved) was unrelated to plant age (Chapter 2). However, total spine

allocation (i.e., percent spine dry weight investment) tended to decrease with plant age

102

(Chapter 2); this finding supports the pattern of higher levels of defense allocation at the pre-reproductive stage. Future work examining armament over plant ontogeny in various taxa may substantiate previous observations of decreasing defenses from the sapling to adult stages, but should include metrics of allocation rather than investment alone.

An alternative hypothesis of altered defense expression following large-scale herbivory is that plants are reverting to their spiny, juvenile strategy (Kozlowski 1971).

Isolating the effect of ontogeny from other physiological or defense-related factors is difficult in natural systems. Recent work suggests that induction of straight spine length and straight spine form in V. etbaica may be independent of plant age (Chapter 2). If plastic responses of spine traits were related to plant age (i.e., reversion), then the magnitude of the response would be expected to change over plant ontogeny. However, this was not the case in V. etbaica. There is some work from the horticultural literature that provides anecdotal evidence of age-related changes in thorn development. Grafting from budwood is a common method in Citrus propagation, where budwood selected from one individual is grafted to the rootstock of another. When budwood is taken from thorn-less branches, this results in nearly thorn-less scions (Cameron and Frost 1968).

As budwood meristems are robust to large physiological changes (e.g., grafting of budwood to new root stock), this may indicate that meristem identity (thorny or thorn- less) is related to ontogeny (i.e., aging) rather than physiology in Citrus. Ontogenetic patterns of plant defense allocation are just beginning to be explored (Boege and

Marquis 2005), and studies manipulating armament presence or investment among life

103

stages will contribute to our understanding of the costs and benefits of herbivory

throughout ontogeny.

Physiological Changes and Resource Allocation

Biomass removal and resource availability can have complex effects on plant

physiology resulting in altered growth and allocation patterns. Therefore, changes to

defense allocation following herbivory or resource addition may be the result of altered

physiology and internal trade-offs rather than defense induction or ontogeny (Bryant et

al. 1983, Haukioja 1990, Herms and Mattson 1992). The propensity for plants to alter

armament depending on resource availability is variable; resources can increase

(fertilization or light; Gibson et al. 1993, Fisher et al. 2002, Gowda et al. 2003, Ward et

al. 2012), decrease (fertilization or grass competition; Bazely et al. 1991, Scogings and

Mopipi 2008), or have no effect (fertilization or water; Myers 1987, Myers and Bazely

1991, Cash and Fulbright 2005) on armament. As discussed above, herbivory also has

variable effects on armament allocation.

The “Carbon-Nutrient Balance Hypothesis” (CNBH) posits that biomass removal

alters a plant’s carbon to nutrient (CN) ratio, and if defense allocation is flexible this will

result in differences in the amount of carbon-based defenses present in new tissues.

For plants storing carbon in leaves (e.g., evergreen plants), biomass removal should

decrease CN ratio and carbon-based defenses will decrease (as there is no “extra”

carbon to allocate). Alternatively, biomass loss should increase the CN ratio in plants

with large below-ground carbon storage resulting in increased levels of carbon-based defenses, and this should particularly pronounced when nutrient supply is low (Bryant et al. 1983). Conversely, fertilization or shade should decrease CN ratio in plants with

belowground stores, and in this case decrease carbon-based defense allocation.

104

Studies explicitly considering allocation (e.g., level of armament biomass relative to primary biomass) even in related species have yielded conflicting results. For example,

(Gowda et al. 2003) found that V. tortilis (formerly Acacia tortilis) seedlings increase the relative mass of long spines to twigs and leaves with increasing resources; twig mass explained much of the variation in total spine mass (i.e., positively). Conversely, V. karroo (formerly Acacia karroo) showed no response to resource addition (Scogings and Mopipi 2008), although between their treatments shoot mass was positively correlated with spine allocation. Pisani and Distel (1998) found no effect of fertilization or water to spine allocation in two Prosopis species, although spine allocation was low overall (i.e., e.g., spines were 0.2-3% of shoot mass). Plastic responses to resources are contingent on whether plants have fixed or flexible allocation strategies (Stamp

2003), so different responses between species could reflect variation in allocation strategies. When a nutrient response was detected, spine allocation showed the opposite pattern of CNBH predictions (e.g., increased rather than decreased carbon- based defense with decreasing CN ratio). It may be that plastic responses to altered CN ratios per the CNBH are not as relevant for armament, as spine growth and development involve both carbon and nitrogen in which case CN ratio should have little impact in defense allocation. An additional consideration for future work is plant age, as large physiological shifts associated with ontogeny could alter the way in which plants respond to resource manipulation.

GDBH predicts a humped shaped relationship between resource availability and defense based on shifts between source and sink limited growth phases: when resources are low, photosynthesis (NAR) and relative growth rate (RGR) are limited by

105

allocation, as resources increase NAR is high but RGR is still limited by resources so defense allocation increases (e.g., resource “excess”), and at high resources RGR is high and resource allocation in defense should be low (Herms and Mattson 1992).

Depending on experimental ranges of resource availability relative to NAR and RGR, it is difficult to interpret whether armament shows this pattern. Such studies are broadly needed, as there have been relatively few studies attempting to fully assess GDBH (but see Stamp et al. 2004, Glynn et al. 2007). The Source/Sink Hypothesis (S/SH) proposes an alternative way in which plant physiology influences defense allocation

(Haukioja 1990, Honkanen et al. 1994): source/sink dynamics influence growth potential or activity (e.g., sink strength) and growth potential could affect defense production.

There are no specific predictions about how or in what circumstances defenses should change. There is some evidence for source-sink mediated changes in chemical defense production (Honkanen et al. 1999, Appel et al. 2012). There are no studies that have discussed or assessed how source-sink relationships influence armament responses, but these mechanisms should not be discounted. One of the major issues in predicting plastic responses of armament is that most theoretical frameworks are based on defensive chemical traits, as noted by other authors (Grubb 1992, Hanley et al. 2007).

Until we know more about the internal trade-offs regulating physiological processes affecting armament it will be difficult to distinguish between mechanisms regulating armament responses to abiotic and biotic factors.

Future Directions

Armament traits are key mediators of plant-herbivore interactions and are an effective form of defense against vertebrate herbivores. The relative ease with which armament can be measured and manipulated makes armed plants model organisms to

106

examine the ecology and evolution of plant defenses. There is a need to contextualize research on armament traits with respect to current plant defense theories, and when needed, expand these frameworks. Here, I propose a number of future research directions that would address current gaps in our knowledge and broadly advance our understanding of plant defenses.

1. What regulates armament development? Are these mechanisms conserved among taxon? Armament structures are derived from epidermal, leaf and branch tissue but how these modifications have occurred is unknown.

2. Is there a genetic basis for armament? There have been no studies assessing whether variation in armament is heritable, a precondition for their adaptive evolution. An armed plant with a short life span should be identified for this work.

3. What are the functional consequences of armament variation? Armament traits can be highly variable within species, however the relative costs and benefits associated with this variation are not well understood.

4. What drives plastic responses in armament? Multiple factors have been shown to influence the expression of armament traits: herbivory, resource availability, ontogeny, and allometry. Controlled tests assessing the contribution of each of these are needed.

5. Do herbivore pressure and armament effectiveness change through plant ontogeny? Data suggest shifts in armament traits over a plant’s lifetime, but there have been no studies testing whether the benefits of these defenses parallel allocation patterns.

6. How do community context and spatial scale affect the responses of wild and domestic browsers to armament traits? Herbivore foraging decisions are contingent on available plant material, and armament traits may be more or less effective depending on the traits of neighboring plant species. Additionally, few studies have examined whether bite and patch level decisions affected by armament defenses translate into differences in daily intake rates.

7. What are the macroevolutionary patterns of armament defenses? Do armament traits co-vary with other suites of defense traits or life history characteristics? The evolution of armament traits among plant taxa has not been explicitly tested.

107

Figure 5-1. Variation in armament morphology and arrangement.

108

Figure 5-2. The potential effects of armament on bite size and rate. Open circles are plants or branches with armament removed, and closed circles represent armed branches. The shift from open to closed represents how armament affects herbivore bite size or bite rate, and ultimately intake rate. Grey lines denote different intake rates (g min-1). Intake rate is (a, d) fully or (b, e) partially maintained if herbivores can increase bite rate to compensate for smaller bite sizes. (c, f) If armament reduces both bite size and rate, then the effect on intake rate will be greatest. (a-c versus d-f) The bite size in the absence of armament relative to an herbivore’s minimum or maximum bite size may influence the magnitude of the response to armament presence in different plant species.

109

A B

C

Figure 5-3. Illustration of how tolerance or variation in resource acquisition and allocation can mask trade-offs between armament and growth. A) Data show a positive relationship between armament and growth, rather than the predicted negative relationship. B) However, this could be because armament or growth trade-off with tolerance (or storage), rather than each other. C) A positive relationship between growth or reproduction and defense can emerge if a plant varies more in its acquisition (Aq) than in allocation (Al) (modified from van Noordwijk and de Jong 1986).

110

A B

C D Figure 5-4. Allometry and the cost of armament defenses. Unarmed and armed plants show the same allometric relationship in branch diameter and branch mass (Belovsky et al. 1991), but armed plants have smaller branch diameters. Armed plants also have relatively lower leaf mass. A) It could be that armed plants have smaller branch diameters, resulting in lower leaf mass (e.g., same allometric relationship as unarmed plants). B) However, if leaf mass relative to branch diameter is lower unarmed plants (e.g., lower intercept), this could suggest that armament reduces leaf growth. The dashed represents armament and leaf mass, which if the same as unarmed plants would further support this trade-off. C) Armament investment could be related branch diameter, which could reflect optimal investment or could be due to biomechanical constraints of supporting large armaments. D) Alternatively, if branches with high investment have proportionally lower leaf mass, this could suggest an allocation cost of between armament and leaf production.

111

CHAPTER 6 GENERAL CONCLUSIONS

In the studies described herein, I have examined ecological and evolutionary

aspects of structural defense trait variation using African acacias as model taxa.

Determinants of Spine Defense Variation

Using a long-term dataset I examined how herbivory and plant age influence spine defenses. In areas where herbivores have been excluded for 10 years, straight spine length and straight spine density were greatly reduced in V. etbaica (18% and

70%, respectively). These data support previous work showing that straight spine length is inducible in Vachellia, and demonstrate for the first time that a plant with mixed-spine morphology is capable of altering the production of different spine forms in response to herbivory. Spine form is thought to reflect defensive function against herbivores of disparate feeding strategies (Cooper and Owen-Smith 1986, Belovsky et al. 1991), and induction of straight spine forms following herbivory may have asymmetric effects on herbivores. Spine defense metrics showed varied relationships with plant size. Straight spine length increased, total spine density decreased and straight spine density was not correlated with plant size. In part, these patterns were explained by covariation between branch growth and spine defense traits. Smaller individuals had smaller branches with shorter straight spines and fewer spines overall but with higher total spine densities cm-

1; the opposite was true for larger plants. Lastly, spine defense allocation in V. etbaica

decreased with plant age, which supports theoretical predictions of ontogenetic shifts in

defenses (Boege and Marquis 2005). Acacias, such as V. etbaica, are widespread

across African savannas, and their defenses influence interactions between both wild

and domestic browsers. Understanding the factors that influence spine traits has broad

112

implications for the plants investing in these costly defenses as well as the herbivores that are deterred by them.

Spine Variation Affects Defense Effectiveness

Establishing the functional role and fitness consequences of various forms of armament is fundamental to understanding its role in defense. I explicitly tested the defensive function of spine variation through feeding trials with two different-sized vertebrate herbivores, and found that spine form had significant effects on proportional biomass loss against both herbivores. Straight spines have been considered to be effective defenses against branch-related biomass loss (pruning), yet our data demonstrate that a mixed strategy of straight intermingled with recurved spines can also be effective at protecting leaf tissue (against picking) even when straight spine density is low. Spine investment between branch types was correlated with leaf biomass, and these differences paralleled rates of biomass removal by goats and camels. The negative relationship between leaf and spine investment suggests a physiological trade- off between growth and differentiation, and also highlights that importance of decoupling the effects of defensive traits from primary growth characteristics that influence herbivore feeding. Spines, prickles and thorns are considered to be a general form of defense against vertebrate herbivores, but results from this work suggest that these traits may be functionally more complex than currently recognized. Structural defenses, such as spines, may be analogous to secondary chemistry; however, it is possible that there are distinct developmental and physiological constraints when defensive traits involve growth processes (e.g. cell division, expansion). A more detailed examination of this is warranted to explore how structural defense traits align with ecological and evolutionary predictions of plant defense theory.

113

Evolution of Spine Form and Domatia in African Vachellia

Plant defenses are thought to have evolved largely as a response to herbivore pressure, and the evolution of plant defenses has been posited to be a major driver in the diversification of both plants and herbivores (Ehrlich and Raven 1964). With the increasing accessibility of molecular data, there has been a resurgence of comparative studies examining macroevolutionary patterns of plant defenses. However, there has been an almost exclusive focus on chemical defenses (but see Armbruster 1997,

Agrawal and Fishbein 2008). I critically evaluated a long-held hypothesis that plant defense traits are phylogenetically conserved using the novel system of spine morphology and ant-associated domatia within Vachellia. The evolution of spine form and domatium presence in 38 species of African Vachellia was examined following phylogenetic reconstruction using plastid DNA. Both defense traits were highly homoplasious, suggesting multiple independent origins of spine morphology and domatia in this clade. While the convergence of domatia among ant-acacias has been previously proposed, this is the first explicit test of domatium evolution in Vachellia incorporating phylogenetic history. Given the level of phylogenetic uncertainty, it is suggested that patterns of spine defense evolution be re-considered with additional molecular data. Nevertheless, structural defenses have rarely been considered in tests of plant defense evolution, and this study presents evidence that structural defense in the form of armament shows convergence in African Vachellia. The biotic and abiotic contexts in which alternative spine forms or domatia have arisen are not clear, and future work taking Vachellia distribution, resource availability and herbivore community into account would help elucidate these factors.

114

Physical Armaments as Model Traits for Studying Plant Defenses

Defenses buffer plants against consumers, and in doing so mediate numerous plant-animal and plant-plant interactions. This along with their adaptive nature has made defense traits a template for understanding wide-ranging aspects of community ecology and evolutionary biology (Rosenthal and Berenbaum 1992a, b). Despite the fact that armament defenses are geographically and taxonomically widespread, there exist major and fundamental gaps in our understanding of these traits: How are they shaped by abiotic and biotic factors? What are their costs relative to the benefits? Why is there morphological variation in armament? Do different sizes or forms reflect defensive function or herbivore specificity? Do resistance or tolerance traits consistently co-vary with armament presence or investment? The relative ease at which armament can be measured and manipulated makes armed plants model organisms to examine the ecology and evolution of plant defenses. I provide the first comprehensive synthesis of our knowledge of the armament traits, and identify future research avenues that will improve our understanding of these widespread defenses as well as broadly inform ecological and evolutionary defense theories.

115

APPENDIX EFFECT OF SPINE LENGTH ON DEFENSE EFFICACY

Methods

Feeding trials were performed with goat (N=12) and camels (N=8) to assess the

effect of straight spine length on defense efficacy and herbivore feeding. Trials took

place at Mpala Research Centre, under the same conditions as spine form experiments,

in August 2010. Treatment branches were collected in pairs from different trees, and all branches had approximately 50% straight spines (i.e., similar to HS branches in spine form feeding trials). Clippers were used to reduce spine length, and cuts were made at

an angle so spine tips were still sharp. There were two treatments, Long (“L”) and Short

(“S”): L branches had just the tip of all straight spines modified, S branches were clipped

at an angle to approximately 2 cm. Long-spined branches had an average straight spine

length of 5.7 (SE ± 0.29, N=11) and 5.5 (SE ± 0.41, N=4) in goat and camel feeding

trials, respectively. Branch order was varied among trials, with two trials per individual.

Goat and camel data were analyzed separately using individuals as a blocking

factor. As branch order was not randomized (e.g., Trial 1, order L, S; Trial 2, order S, L),

I tested for a significant interaction between branch treatment and trial for each

response variable. In no case was the interaction significant, so I averaged responses

among trials for each individual. For camels, several individuals were removed from the

analysis due to interruptions during trials that resulted in little to no feeding. Only one

goat pruned throughout all trials, and I removed this individual from the analysis to

present results similar to spine form data. All data was analyzed in JMP 8.0 using a

mixed model ANOVA, with Branch Type as a fixed effect and Individual as a random

116

effect. I present ANOVA results using Methods of Moments (EMS), which did not differ from REML.

Results

No effect of straight spine length on defense efficacy against goats (Mean for

L=0.25, SE ± 0.0353; S=0.26, SE ± 0.039) or camels (Mean for L=0.83, SE ± 0.059;

S=0.87, SE ± 0.026; Fig S1, Table S1) was found. Similarly, biomass removed per bite and feeding trial were not different between treatments (Goats: Mean bite size for

L=0.019, SE ± 0.0012; S= 0.019, SE ± 0.0016; Mean feeding rate for L= 0.54, SE ±

0.040; S= 0.50, SE ± 0.044; Camels: Mean bite size for L= 0.056, SE ± 0.0186; S=

0.045, SE ± 0.0066; Mean feeding rate for L= 0.65, SE ± 0.137; S= 0.51, SE ± 0.033;

Table S1).

117

Figure A-1. The effect of spine length on defense efficacy against goats and camels. A) Goat feeding trials. B) Camel feeding trials. Note the different scales on y-axes. Bars are SE ±.

118

Table A-1. ANOVA results for goat and camel trials. Branch treatment (Long and Short spines) as a fixed effect and Individual as a random effect.

Source SS MS df F P Goats (N=11) Proportion Leaf Biomass Remaining 0.0014 0.0014 1 0.3822 0.5502 Individual 0.2656 0.0266 10 7.1674 0.0023 Bite size, g <.001 <.001 1 0.0185 0.8946 Individual <.001 <.001 10 5.0097 0.0089 Intake Rate, g min-1 0.0073 0.0073 1 2.0529 0.1824 Individual 0.3551 0.0355 10 9.9489 0.0006 Camels (N=5) Proportion Biomass Remaining 0.0026 0.0026 1 0.8976 0.3971 Individual 0.0583 0.0146 4 5.0317 0.0733 Bite size, g 0.0007 0.0007 1 0.0120 0.9179 Individual 0.3423 0.0856 4 1.5245 0.3464 Intake Rate, g min-1 0.0121 0.0121 1 0.5708 0.4920 Individual 0.1153 0.0288 4 1.3581 0.3870

119

LIST OF REFERENCES

Abdulrazak, S. A., F. T., J. K. Ondiek, and Ørskov. 2000. Nutritive evaluation of some Acacia tree leaves from Kenya. Animal Feed Science and Technology 85:89-98.

Agrawal, A. A. 2007. Macroevolution of plant defense strategies. Trends In Ecology & Evolution 22:103-109.

Agrawal, A. A. and M. Fishbein. 2006. Plant defense syndromes. Ecology 87:S132- S149.

Agrawal, A. A. and M. Fishbein. 2008. Phylogenetic escalation and decline of plant defense strategies. Proceedings of the National Academy of Sciences of the United States of America 105:10057-10060.

Agrawal, A. A., J. A. Rudgers, L. W. Botsford, D. Cutler, J. B. Gorin, C. J. Lundquist, B. W. Spitzer, and A. L. Swann. 2000. Benefits and Constraints on Plant Defense against Herbivores: Spines Influence the Legitimate and Illegitimate Flower Visitors of Yellow Star Thistle, Centaurea solstitialis L. (Asteraceae). The Southwestern Naturalist 45:1-5.

Ali, S. I. and M. Qaiser. 1980. Hybridization in Acacia nilotica (Mimosoideae) complex. Botanical Journal of the Linnean Society 80:69.

Appel, H. M., T. M. Arnold, and J. C. Schultz. 2012. Effects of jasmonic acid, branching and girdling on carbon and nitrogen transport in poplar. New Phytologist 195:419-426.

Armbruster, W. S. 1997. Exaptations link evolution of plant-herbivore and plant- pollinator interactions: A phylogenetic inquiry. Ecology 78:1661-1672.

Arnold, T., H. Appel, V. Patel, E. Scotum, A. Kavalier, and J. Schultz. 2004. Carbohydrate Translocation Determines the Phenolic Content of Populus Foliage: A Test of the Sink-Source Model of Plant Defense. New Phytologist 164:157-164.

Augustine, D. J. and S. J. McNaughton. 1998. Ungulate Effects on the Functional Species Composition of Plant Communities: Herbivore Selectivity and Plant Tolerance. The Journal of Wildlife Management 62:1165-1183.

Augustine, D. J. and S. J. McNaughton. 2004. Regulation of shrub dynamics by native browsing ungulates on East African rangeland. Journal Of Applied Ecology 41:45-58.

Baldwin, B. G. 1992. Phylogenetic Utility of the Internal Transcribed Spacers of Nuclear Ribosomal DNA in Plants: An Example from the Compositae. Molecular Phylogenetics and Evolution 1:3-16.

120

Barnard, B. J. and R. H. Hassel. 1981. Rabies in kudus (Tragelaphus strepsiceros) in South West Africa/Namibia. J S Afr Vet Assoc 52:309-314.

Bazely, D. R., J. H. Myers, and K. B. Dasilva. 1991. The Response of Numbers of Bramble Prickles to Herbivory and Depressed Resource Availability. Oikos 61:327-336.

Belovsky, G. E. and O. J. Schmitz. 1994. Plant Defenses and Optimal Foraging by Mammalian Herbivores. Journal of Mammalogy 75:816-832.

Belovsky, G. E., O. J. Schmitz, J. B. Slade, and T. J. Dawson. 1991. Effects of Spines and Thorns on Australian Arid Zone Herbivores of Different Body Masses. Oecologia 88:521-528.

Berenbaum, M. R. 1983. Coumarins and caterpillars – a case for coevolution. Evolution 37:163–179.

Bieniek, M. E. and W. F. Millington. 1967. Differentiation of Lateral Shoots as Thorns in Ulex europaeus. American Journal of Botany 54:61-70.

Bieniek, M. E. and W. F. Millington. 1968. Thorn Formation in Ulex europaeus in Relation to Environmental and Endogenous Factors. Botanical Gazette 129:145- 150.

Björkman, C. and D. B. Anderson. 1990. Trade-off among Antiherbivore Defences in a South American Blackberry (Rubus bogotensis). Oecologia 85:247-249.

Blaser, H. W. 1956. Morphology of the Determinate Thorn-Shoots of Gleditsia. American Journal of Botany 43:22-28.

Boege, K. and R. J. Marquis. 2005. Facing herbivory as you grow up: the ontogeny of resistance in plants. Trends In Ecology & Evolution 20:441-448.

Boke, N. H. 1980. Developmental Morphology and Anatomy in Cactaceae. BioScience 30:605-610.

Bouchenak-Khelladi, Y., O. Maurin, J. Hurter, and M. van der Bank. 2010. The evolutionary history and biogeography of Mimosoideae (Leguminosae): An emphasis on African acacias. Molecular Phylogenetics and Evolution 57:495.

Briand, C. H. and C. L. Soros. 2001. Spatial Variation of Prickle Abundance on Leaves of the Devil's Walking Stick (Aralia spinosa; Araliaceae) during the Trunk-Building Phase. Journal of the Torrey Botanical Society 128:219-225.

Brooks, R. and N. Owen-Smith. 1994. Plant defences against mammalian herbivores: are juvenile acacia more heavily defended than mature trees? Bothalia 24:211- 215.

121

Brown, G. K., C. Clowes, D. J. Murphy, and P. Y. Ladiges. 2010. Phylogenetic analysis based on nuclear DNA and morphology defines a clade of eastern Australian species of Acacia s.s. (section Juliflorae): the 'Acacia longifolia group'. Australian Systematic Botany 23:162-172.

Bryant, J. P., F. S. Chapin, and D. R. Klein. 1983. Carbon Nutrient Balance of Boreal Plants in Relation to Vertebrate Herbivory. Oikos 40:357-368.

Bryant, J. P., P. J. Kuropat, S. M. Cooper, K. Frisby, and N. Owen-Smith. 1989. Resource availability hypothesis of plant antiherbivore defence tested in a South African savanna ecosystem. Nature 340:227-229.

Burlison, A. J., J. Hodgson, and A. W. Illius. 1991. Sward canopy structure and the bite dimension and bite weight of grazing sheep. Grass Forage Science 46:29-38.

Cameron, J. W. and H. B. Frost. 1968. Genetics, Breeding, and Nucellar Embryony.in M. Nemeth, editor. The Citrus Industry. University of California, Division of Agricultural Sciences, Davis, CA.

Campbell, B. M. 1986. Plant spinescence and herbivory in a nutrient poor ecosystem. Oikos 47:168-172.

Campbell, H., I. R. Townsend, M. D. E. Fellowes, and J. M. Cook. 2013. Thorn-dwelling ants provide antiherbivore defence for camelthorn trees, Vachellia erioloba, in Namibia. African Journal of Ecology:n/a-n/a.

Carlquist, S. 1962. Ontogeny and Comparative Anatomy of Thorns of Hawaiian Lobeliaceae. American Journal of Cardiology 10:413-&.

Cash, V. W. and W. E. Fulbright. 2005. Nutrient enrichment, tannins, and thorns: Effects on browsing of shrub seedlings. Journal Of Wildlife Management 69:782-793.

Clarke, H. D., S. R. Downie, and D. S. Seigler. 2000. Implications of chloroplast DNA restriction site variation for systematics of Acacia (Fabaceae : Mimosoideae). Systematic Botany 25:618-632.

Coe, M. and H. Beentje. 1992. A Field Guide to the Acacias of Kenya. Oxford University Press, Oxford.

Coley, P. D., J. P. Bryant, and F. S. Chapin. 1985. Resource Availability And Plant Antiherbivore Defense. Science 230:895-899.

Cooper, S. M. and T. F. Ginnett. 1998. Spines protect plants against browsing by small climbing mammals. Oecologia 113:219-221.

Cooper, S. M. and N. Owen-Smith. 1986. Effects Of Plant Spinescence On Large Mammalian Herbivores. Oecologia 68:446-455.

122

Cooper, S. M., M. K. Owens, D. E. Spalinger, and T. F. Ginnett. 2003. The Architecture of Shrubs after Defoliation and the Subsequent Feeding Behavior of Browsers. Oikos 100:387-393.

Cowling, R. M. and B. M. Campbell. 1980. Convergence in vegetation structure in the mediterranean communities of California, Chile and South Africa. Vegetatio 43:191-197.

Coyner, M. A. 2005. Thornlessness in Blackberries A Review. Small fruits review 4:83- 106.

Cullings, K. W. 1992. Design and testing of a plant-specific PCR primer for ecological and evolutionary studies. Molecular Ecology 1:233-240.

Doyle, J. J. and J. L. Doyle. 1987. A rapid DNA isolation protocol for small quantities of fresh leaf tissue. Phytochemical Bulletin 19:11-15.

Du Toit, J. and D. Cumming. 1999. Functional significance of ungulate diversity in African savannas and the ecological implications of the spread of pastoralism. BIODIVERSITY AND CONSERVATION 8:1643-1661. du Toit, J. T., J. P. Bryant, and K. Frisby. 1990. Regrowth and Palatability of Acacia Shoots Following Pruning by African Savanna Browsers. Ecology 71:149-154.

Ebinger, J. E. and D. S. Seigler. 1992. Ant-Acacia Hybrids of Mexico and Central America. The Southwestern Naturalist 37:408.

Edwards, A. W. F. 1972. Likelihood. Cambridge University Press, Cambridge.

Ehleringer, J. R., O. Björkman, and H. A. Mooney. 1976. Leaf pubescence: effects on absorptance and photosynthesis in a desert shrub. Science 192:376-377.

Ehrlich, P. R. and P. H. Raven. 1964. Butterflies and Plants: A Study in Coevolution. Evolution 18:586-608.

Enquist, B. J. and K. J. Niklas. 2002. Global Allocation Rules for Patterns of Biomass Partitioning in Seed Plants. Science 295:1517-1520.

Farrell, B. D. a. M., C. 1998. The timing of insect–plant diversification: might Tetraopes (Coleoptera: Cerambycidae) and Asclepias (Asclepiadaceae) have co-evolved? Biol. J. Linnean Soc. 63:553–577.

Feeny, P. P. 1976. Plant apparency and chemical defense. Pages 1-40 in Biochemical interaction between plants and insects. Plenum Press.

Felsenstein, J. 1985a. Confidence Limits on Phylogenies: An Approach Using the Bootstrap. Evolution 39:783.

123

Felsenstein, J. 1985b. Phylogenies and the Comparative Method. American Naturalist 125:1-15.

Fine, P. V. A., Z. J. Miller, I. Mesones, S. Irazuzta, H. M. Appel, M. Stevens, H. H., I. Sääksjärvi, J. C. Schultz, and P. D. Coley. 2006. The Growth-Defense Trade-off and Habitat Specialization by Plants in Amazonian Forests. Ecology 87:S150.

Fisher, J. B., U. Posluszny, and D. W. Lee. 2002. Shade promotes thorn development in a tropical liana, Artabotrys hexapetalus (Annonaceae). International Journal of Plant Sciences 163:295-300.

Fornara, D. D. T., JT. 2007. Browsing lawns? Responses of Acacia nigrescens to ungulate browsing in an African savannah. Ecology 88:200-209.

Gadd, M. E., T. P. Young, and T. M. Palmer. 2001. Effects of simulated shoot and leaf herbivory on vegetative growth and plant defense in Acacia drepanolobium. Oikos 92:515-521.

Gadd, M. Y., TP; Palmer, TM. 2001. Effects of simulated shoot and leaf herbivory on vegetative growth and plant defense in Acacia drepanolobium. Oikos 92:515- 521.

Galtier, N., Gouy, M. & Gautier, C. 1996. SEAVIEW and PHYLO_WIN: two graphic tools for sequence alignment and molecular phylogeny. Computer Application in the Biosciences 12:543-548.

Gibson, D., D. R. Bazely, and J. S. Shore. 1993. Responses of Brambles, Rubus vestitus, to Herbivory. Oecologia 95:454-457.

Glynn, C., D. A. Herms, C. M. Orians, R. C. Hansen, and S. Larsson. 2007. Testing the growth-differentiation balance hypothesis: dynamic responses of willows to nutrient availability. New Phytologist 176:623-634.

Goheen, J. R. and T. M. Palmer. 2010. Defensive Plant-Ants Stabilize Megaherbivore- Driven Landscape Change in an African Savanna. Current Biology 20:1768.

Goheen, J. R., T. P. Young, F. Keesing, and T. M. Palmer. 2007. Consequences of herbivory by native ungulates for the reproduction of a savanna tree. Journal Of Ecology 95:129-138.

Gomez, J. M. and R. Zamora. 2000. Spatial variation in the selective scenarios of Hormathophylla spinosa (Cruciferae). American Naturalist 155:657-668.

Gomez, J. M. and R. Zamora. 2002. Thorns as an induced mechanical defense in a long-lived shrub (Hormathophylla spinosa, Cruciferae). Ecology 83:885-890.

124

Gouy, M. G., S. & Gascuel., O. 2010. SeaView version 4 : a multiplatform graphical user interface for sequence alignment and phylogenetic tree building. Molecular Biology and Evolution 27:221-224.

Gowda, J. and E. Raffaele. 2004. Spine production is induced by fire: a natural experiment with three Berberis species. Acta Oecologica-International Journal of Ecology 26:239-245.

Gowda, J. H. 1996. Spines of Acacia tortilis: What do they defend and how? Oikos 77:279-284.

Gowda, J. H. 1997. Physical and chemical response of juvenile Acacia tortilis trees to browsing. Experimental evidence. Functional Ecology 11:106-111.

Gowda, J. H., B. R. Albrectsen, J. P. Ball, M. Sjoberg, and R. T. Palo. 2003. Spines as a mechanical defence: the effects of fertiliser treatment on juvenile Acacia tortilis plants. Acta Oecologica-International Journal of Ecology 24:1-4.

Gowda, J. H. and R. T. Palo. 2003. Age-related changes in defensive traits of Acacia tortilis Hayne. African Journal of Ecology 41:218-223.

Greve, M., A. M. Lykke, C. W. Fagg, J. Bogaert, I. Friis, R. Marchant, A. R. Marshall, J. Ndayishimiye, B. S. Sandel, C. Sandom, M. Schmidt, J. R. Timberlake, J. J. Wieringa, G. Zizka, and J.-C. Svenning. 2012. Continental-scale variability in browser diversity is a major driver of diversity patterns in acacias across Africa. Journal of Ecology 100:1093-1104.

Grubb, P. J. 1992. A Positive Distrust In Simplicity - Lessons From Plant Defenses And From Competition Among Plants And Among Animals. Journal Of Ecology 80:585-610.

Hanley, M. E. and B. B. Lamont. 2002. Relationships between physical and chemical attributes of congeneric seedlings: how important is seedling defence? Functional Ecology 16:216-222.

Hanley, M. E., B. B. Lamont, and W. S. Armbruster. 2009. Pollination and plant defence traits co-vary in Western Australian Hakeas. New Phytologist 182:251-260.

Hanley, M. E., B. B. Lamont, M. M. Fairbanks, and C. M. Rafferty. 2007. Plant structural traits and their role in anti-herbivore defence. Perspectives In Plant Ecology Evolution And Systematics 8:157-178.

Haukioja, E. 1990. Induction of defences in trees. Annual Review of Entomology 36:25- 42.

Haukioja, E. 1991. The influence of grazing on the evolution, morphology and physiology of plants as modular organisms. Philosophical Transactions of the Royal Society of London 333:241–247.

125

Haukioja, E. and S. Neuvonen. 1985. Induced Long-Term Resistance Of Birch Foliage Against Defoliators - Defensive Or Incidental. Ecology 66:1303-1308.

Heede, C. J. V. d., R. L. L. Viane, and M. W. Chase. 2003. Phylogenetic Analysis of Asplenium Subgenus Ceterach (Pteridophyta: Aspleniaceae) Based on Plastid and Nuclear Ribosomal Its DNA Sequences. American Journal of Botany 90:481.

Heil, M. 2002. Ecological costs of induced resistance. Current Opinion in Plant Biology 5:345 - 350.

Herms, D. A. and W. J. Mattson. 1992. The Dilemma of Plants: To Grow or Defend. The Quarterly Review of Biology 67:283-335.

Hobbs, N. T., J. E. Gross, L. A. Shipley, D. E. Spalinger, and B. A. Wunder. 2003. Herbivore Functional Response in Heterogeneous Environments: A Contest among Models. Ecology 84:666-681.

Honkanen, T. and E. Haukioja. 1994. Why does a branch suffer more after branch-wide than tree-wide defoliation? Oikos 71:441–450.

Honkanen, T., E. Haukioja, and V. Kitunen. 1999. Responses of Pinus sylvestris branches to simulated herbivory are modified by tree sink/source dynamics and by external resources. Functional Ecology 13:126–140.

Honkanen, T., E. Haukioja, and J. Suomela. 1994. Effects of simulated defoliation and debudding on needle and shoot growth in Scots Pine (Pinus sylvestris) - implications of plant source-sink relationships for plant-herbivore studies. Functional Ecology 8:631-639.

Hübschle, O. J. B. 1988. Rabies in the Kudu Antelope (Tragelaphus strepsiceros). Reviews of Infectious Diseases 10:S629-S633.

Illius, A. W., P. Duncan, C. Richard, and P. Mesochina. 2002. Mechanisms of Functional Response and Resource Exploitation in Browsing Roe Deer. Journal of Animal Ecology 71:723-734.

Janzen, D. H. 1974a. Swollen-Thorn Acacias of Central America. 13, Smithsonian Institute, Washington, D.C. .

Janzen, D. H. 1974b. Tropical Blackwater Rivers, Animals, and Mast Fruiting by the Dipterocarpaceae. Biotropica 6:69.

Karban, R., A. A. Agrawal, and M. Mangel. 1997. The benefits of induced defenses against herbivores. Ecology 78:1351-1355.

Karban, R. and I. T. Baldwin. 1997. Induced responses to herbivory. The University of Chicago Press.

126

Kellogg, A. A., T. J. Branaman, N. M. Jones, C. Z. Little, and J. D. Swanson. 2011. Morphological studies of developing Rubus prickles suggest that they are modified glandular trichomes. Botany-Botanique 89:217-226.

Koricheva, J., H. Nykanen, and E. Gianoli. 2004. Meta-analysis of trade-offs among plant antiherbivore defenses: are plants jacks-of-all-trades, masters of all? American Naturalist 163:E64-75.

Kozlowski, T. T. 1971. Growth and development of trees. Academic Press, New York.

Kursar, T. A. a. C., P.D. 2003. Convergence in defense syndromes of young leaves in tropical rainforests. Biochem. Syst. Ecol. 31:929– 949.

Laca, E. A., E. G. Ungar, N. Seligman, and M. W. Demment. 1992. Effects of sward height and bulk density on bite dimensions of cattle grazing homogeneous swards. Grass Forage Science 47:91-102.

Lewis, P. O. 2001. A likelihood approach to estimating phylogeny from discrete morphological character data. Systematic Biology 50:913-925.

Liscombe, D. K., B. P. MacLeod, N. Loukanina, O. I. Nandi, and P. J. Facchini. 2005. Evidence for the monophyletic evolution of benzylisoquinoline alkaloid biosynthesis in angiosperms. Phytochemistry 66:1374-1393.

Luckow, M., J. T. Miller, D. J. Murphy, and T. Livshultz. 2003. A phylogenetic analysis of the Mimosoideae (Leguminosae) based on chloroplast DNA sequence data.in B. B. Klitgaard and A. Bruneau, editors. Advances in Legume Systematics, Part 10: Higher Level Systematics. Royal Botanical Gardens. , Kew.

Maclean, J. E., J. R. Goheen, D. F. Doak, T. M. Palmer, and T. P. Young. 2011. Cryptic herbivores mediate the strength and form of ungulate impacts on a long-lived savanna tree. Ecology 92:1626-1636.

Maddison, W. P. a. D. R. M. 2007. Mesquite: a modular system for evolutionary analysis.

Mauseth, J. D. 1977. Cytokinin- and Gibberellic Acid-Induced Effects on the Determination and Morphogenesis of Leaf Primordia in Opuntia polyacantha (Cactaceae). American Journal of Botany 64:337-346.

Midgley, J. J. and D. Ward. 1996. Research note: Tests for inducible thorn defences against herbivory must consider plant growth strategies. African Journal of Range & Forage Science 13:75-77.

Milewski, A. V., T. P. Young, and D. Madden. 1991. Thorns as Induced Defenses - Experimental-Evidence. Oecologia 86:70-75.

127

Miller, J. T. and R. J. Bayer. 2001. Molecular phylogenetics of Acacia (Fabaceae : Mimosoideae) based on the chloroplast matK coding sequence and flanking trnK intron spacer regions. American Journal of Botany 88:697-705.

Miller, J. T. and R. J. Bayer. 2003. Molecular phylogenetics of Acacia subgenera Acacia and Aculeiferum (Fabaceae : Mimosoideae), based on the chloroplast matK coding sequence and flanking trnK intron spacer regions. Australian Systematic Botany 16:27-33.

Milton, S. 1991. Plant spinescence in arid southern Africa: does moisture mediate selection by mammals? Oikos 87:279-287.

Mokoboki, H. K., L. R. Ndlovu, J. W. Ngami, M. M. Malatje, and R. V. Nilolova. 2005. Nutritive value of Acacia tree foliages growing in the Limpopo Province of South Africa. South African Journal of Animal Science 35:221-228.

Myers, J. H. 1987. Nutrient Availability and the Deployment of Mechanical Defenses in Grazed Plants - a New Experimental Approach to the Optimal Defense Theory. Oikos 49:350-351.

Myers, J. H. and D. Bazely. 1991. Thorns, spines, prickles and hairs: Are they stimulated by herbivory and do they deter herbivores? Pages 325-344 in D. W. Tallamy and M. J. Raupp, editors. Phytochemical induction by herbivores. . John Wiley, New York.

Obeso, J. R. 1997. The Induction of Spinescence in European Holly Leaves by Browsing Ungulates. Plant Ecology 129:149-156.

Osborne, P. L. 2000. Tropical ecosystems and ecological concepts. Tropical ecosystems and ecological concepts. Cambridge University Press.

Owen-Smith, N. 1993. Evaluating optimal diet models for an African browsing ruminant, the kudu: How constraining are the assumed constraints? Evolutionary Ecology 7:499-524.

Pabón-Mora, N. and F. González. 2012. Leaf Development, Metamorphic Heteroblasty and Heterophylly in Berberis s. l. (Berberidaceae). The Botanical Review 78:463- 489.

Pedley, L. 1986. Derivation and Dispersal of Acacia (Leguminosae), with Particular Reference to Australia, and the Recognition of Senegalia and Racosperma. Botanical Journal of the Linnean Society 92:219-254.

Pellew, R. A. 1984. Food Consumption and Energy Budgets of the Giraffe. Journal of Applied Ecology 21:141-159.

128

Penning, P. S., A. J. Parsons, R. J. Orr, and T. T. Treacher. 1991. Intake and behaviour responses by sheep to changes in sward characteristics under continuous stocking. Grass Forage Science 46:15-28.

Peterken, G. F. and P. S. Lloyd. 1967. Ilex Aquifolium L. Journal of Ecology 55:841-858.

Pisani, J. M. and R. A. Distel. 1998. Inter- and Intraspecific Variations in Production of Spines and Phenols in Prosopis caldenia and Prosopis flexuosa. J Chem Ecol 24:23-36.

Posluszny, U. and J. B. Fisher. 2000. Thorn and Hook Ontogeny in Artabotrys hexapetalus (Annonaceae). American Journal of Botany 87:1561-1570.

R.C., E. 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research 32:1792-1797.

Rafferty, C., B. B. Lamont, and M. E. Hanley. 2005. Selective feeding by kangaroos (Macropus fuliginosus) on seedlings of Hakea species: Effects of chemical and physical defences. Plant Ecology 177:201-208.

Rautio, P., K. Kesti, U. A. Bergvall, J. Tuomi, and O. Leimar. 2008. Spatial scales of foraging in fallow deer: Implications for associational effects in plant defences. Acta Oecologica 34:12-20.

Rhoades, D. F. 1979. Evolution Of Plant Chemical Defense Against Herbivores. Pages P3-54 Rosenthal, G. A. And D. H. Janzen (Ed.). Herbivores: Their Interaction With Secondary Plant Metabolites. Xiii+718p. Academic Press, Inc.: New York, N.Y., Usa. London, England. Illus.

Rieseberg, L. H. and D. E. Soltis. 1991. Phylogenetic Consequences Of Cytoplasmic Gene Flow In Plants. Evolutionary Trends In Plants 5:65-84.

Robinson, J. W. and S. A. Harris. 2000. A plastid DNA phylogeny of the genus Acacia Miller (Acacieae, Leguminoseae). Botanical Journal of the Linnean Society 132:195-222.

Rodríguez, E., P. L. Healey, and I. Mentha. 1984. Biology and chemistry of plant trichomes. Plenum Press, New York.

Rohner, C. and D. Ward. 1997. Chemical and mechanical defense against herbivory in two sympatric species of desert Acacia. Journal of Vegetation Science 8:717- 726.

Rooke, T., K. Danell, R. Bergström, C. Skarpe, and J. Hjältén. 2004. Defensive traits of savanna trees -/ the role of shoot exposure to browsers. Oikos 107:161-171.

129

Rosenthal, G. A. and M. R. Berenbaum, editors. 1992a. Herbivores Their Interactions With Secondary Plant Metabolites Second Edition Vol. I The Chemical Participants. Academic Press, Inc. , San Diego.

Rosenthal, G. A. and M. R. Berenbaum, editors. 1992b. Herbivores Their Interactions With Secondary Plant Metabolites Second Edition Vol. II: Ecological and evolutionary processes. Academic Press, Inc., San Diego.

Schluter, D., T. Price, A. Ø. Mooers, and D. Ludwig. 1997. Likelihood of Ancestor States in Adaptive Radiation. Evolution 51:1699-1711.

Schnittger, A., U. Folkers, B. Schwab, G. Jürgens, and M. Hülskamp. 1999. Generation of a Spacing Pattern: The Role of TRIPTYCHON in Trichome Patterning in Arabidopsis. The Plant Cell 11:1105–1116.

Scholes, R. J. and S. R. Archer. 1997. Tree-grass interactions in savannas. Annual Review of Ecology and Systematics 28:517-544.

Schutz, A. E. N., W. J. Bond, and M. D. Cramer. 2011. Defoliation depletes the carbohydrate reserves of resprouting Acacia saplings in an African savanna. Plant Ecology 212:2047–2055.

Scogings, P. F. and K. Mopipi. 2008. Effects of water, grass and N on responses of Acacia karroo seedlings to early wet season simulated browsing: Aboveground growth and biomass allocation. Journal of Arid Environments 72:509-522.

Seigler, D. S. 2003. Phytochemistry of Acacia - sensu lato. Biochemical Systematics and Ecology 31:845-873.

Shipley, L. A. 2007. The influence of bite size on foraging at larger spatial and temporal scales by mammalian herbivores. Oikos 116:1964-1974.

Soltis, D. E., P. S. Soltis, and J. J. Doyle. 1998. Molecular systematics of plants, II: DNA sequencing. Molecular systematics of plants, II: DNA sequencing. Kluwer Academic Publishers; Kluwer Academic Publishers.

Southwood, T. R. E. 1986. Plant surfaces and insects—an overview. Edward Arnold, London.

Spalinger, D. E. and N. T. Hobbs. 1992. Mechanisms of Foraging in Mammalian Herbivores: New Models of Functional Response. The American Naturalist 140:325-348.

Stamatakis, A. 2006. RAxML.

Stamp, N. 2003. Out of the quagmire of plant defense hypotheses. Quarterly Review Of Biology 78:23-55.

130

Stamp, N., M. Bradfield, S. Li, and B. Alexander. 2004. Effect of Competition on Plant Allometry and Defense. American Midland Naturalist 151:50-64.

Stapley, L. 1998. The interaction of thorns and symbiotic ants as an effective defence mechanism of swollen-thorn acacias. Oecologia 115:401-405.

Symon, D. E. 1986. A Survey of Solanum Prickles and Marsupial Herbivory in Australia. Annals of the Missouri Botanical Garden 73:745-754.

Takada, M., M. Asada, and T. Miyashita. 2003. Can spines deter deer browsing?: a field experiment using a shrub Damnacanthus indicus. Journal of Forest Research 8:321-323.

Tattersall, A. D., L. Turner, M. R. Knox, M. J. Ambrose, T. H. N. Ellis, and J. M. I. Hofer. 2005. The Mutant crispa Reveals Multiple Roles for PHANTASTICA in Pea Compound Leaf Development. The Plant Cell 17:1046–1060.

Twigg, L. E. and L. V. Socha. 1996. Physical versus Chemical Defence Mechanisms in Toxic Gastrolobium. Oecologia 108:21-28.

van Noordwijk, A. J. and G. de Jong. 1986. Acquisition and Allocation of Resources: Their Influence on Variation in Life History Tactics. The American Naturalist 128:137-142.

Van Zandt, P. A. and A. A. Agrawal. 2004. Specificity of induced plant responses to specialist herbivores of the common milkweed Asclepias syriaca. Oikos 104:401- 409.

Vassal, J. 1981. Acacieae. In "Advances in Legume Systematics, Part 1.". Royal Botanical Gardens, Kew, Richmond, Surrey, UK.

Waites, R. and A. Hudson. 1995. phantastica: a gene required for dorsoventrality of leaves in Antirrhinum majus. Development 121:2143-2154.

Ward, D. 2010. The effects of apical meristem damage on growth and defences of two Acacia species in the Negev Desert. Evolutionary Ecology Research 12:589– 602.

Ward, D., M. K. Shrestha, and A. Golan-Goldhirsh. 2011. Evolution and ecology meet molecular genetics: adaptive phenotypic plasticity in two isolated Negev desert populations of Acacia raddiana at either end of a rainfall gradient. Annals of Botany.

Ward, D., M. K. Shrestha, and A. Golan-Goldhirsh. 2012. Evolution and ecology meet molecular genetics: adaptive phenotypic plasticity in two isolated Negev desert populations of Acacia raddiana at either end of a rainfall gradient. Annals of Botany 109:247-255.

131

Ward, D. and T. P. Young. 2002. Effects of large mammalian herbivores and ant symbionts on condensed tannins of Acacia drepanolobium in Kenya. J Chem Ecol 28:921-937.

White, P. S. 1988. Prickle Distribution in Aralia spinosa (Araliaceae). American Journal of Botany 75:282-285.

Wiens, J. J. 2006. Missing data and the design of phylogenetic analyses. Journal of Biomedical Informatics 39:34-42.

Wilson, S. L. and G. I. H. Kerley. 2003. Bite diameter selection by thicket browsers: the effect of body size and plant morphology on forage intake and quality. Forest Ecology and Management 181:51-65.

Wink, M. 2003. Evolution of secondary metabolites from an ecological and molecular phylogenetic perspective. Phytochemistry 64:3-19.

Young, T. P. 1987. Increased Thorn Length in Acacia drepranolobium - an Induced Response to Browsing. Oecologia 71:436-438.

Young, T. P. and D. J. Augustine. 2007. Interspecific variation in the reproductive response of acacia species to protection from large mammalian herbivores. Biotropica 39:559-561.

Young, T. P. and B. D. Okello. 1998. Relaxation of an induced defense after exclusion of herbivores: spines on Acacia drepanolobium. Oecologia 115:508-513.

Young, T. P., N. Patridge, and A. Macrae. 1995. Long-Term Glades in Acacia Bushland and Their Edge Effects in Laikipia, Kenya. Ecological Applications 5:97-108.

Young, T. P., M. L. Stanton, and C. E. Christian. 2003. Effects of natural and simulated herbivory on spine lengths of Acacia drepanolobium in Kenya. Oikos 101:171- 179.

Young, T. P., C. H. Stubblefield, and L. A. Isbell. 1996. Ants on swollen-thorn acacias: species coexistence in a simple system. Oecologia 109:98-107.

Zangerl, A. R., M. C. Stanley, and M. R. Berenbaum. 2008. Selection for chemical trait remixing in an invasive weed after reassociation with a coevolved specialist. Proceedings of the National Academy of Sciences of the United States of America 105:4547-4552.

Zhang, Z., S. P. Wang, P. Nyren, and G. M. Jiang. 2006. Morphological and reproductive response of Caragana microphylla to different stocking rates. Journal of Arid Environments 67:671-677.

132

Zinn, A. D. W., D.; Kirkman, K. 2007. Inducible defences in Acacia sieberiana in response to giraffe browsing. African Journal of Range and Forage Science 24:123-129.

Zwickl, D. J. and D. M. Hillis. 2002. Increased Taxon Sampling Greatly Reduces Phylogenetic Error. Systematic Biology 51:588-598.

133

BIOGRAPHICAL SKETCH

Megan Gittinger was born in Christiansburg, VA and was raised in Mt. Pleasant,

SC. She earned a Bachelor of Degree in wildlife science at Virginia Polytechnic Institute in 2005. It was during her undergraduate research experience in Dr. Lynn Adler’s laboratory that she became interested in plant adaptation, particularly in how plants evolve in response to herbivore pressure. Before pursuing a doctorate, Megan gained professional experience in conservation biology at the Smithsonian Institute’s

Conservation Research Center in Front Royal, Virginia and at WildAid Foundation in

Bangkok, Thailand. In August 2007, Megan entered graduate school at The University of Florida in the Department of Zoology and received her Doctor of Philosophy in

August 2013.

134