Cotton is a high-value commercial crop in Australia. Although cotton is largely self-pollinating, previous researchers have reported that honeybees, Apis mellifera, can assist in cross-pollination and contribute to improved yield. Until recently, use of bees in cotton had, however, been greatly limited by excessive use of pesticides to control arthropod pests. With the widespread use of transgenic (Bt) cotton varieties and the associated reduction in pesticide use, I decided to investigate the role and importance of honeybees in Bt cotton, under Australian conditions. I conducted two major field trials at Narrabri, in the centre of one of Australia’s major cotton-growing areas, in the 2005-6 and 2006-7 seasons. In the first trial, I particularly assessed methods of manipulating honeybee colonies by feeding pollen supplements of pollen/soybean patties, and by restricting pollen influx by the fitting of 30% efficient pollen traps. I aimed to test whether either of these strategies increased honeybee flight activity and, thus, increased foraging on cotton flowers. My results showed that although supplementary feeding increased bee flight activity and brood production, it did not increase pollen collection on cotton. Pollen traps initially reduced flight activity. They also reduced the amount of pollen stored in colonies, slowed down brood rearing activity, and honey production. However, they did not contribute to increased pollen collection in cotton. In the second trial, I spent more time investigating honeybee behaviour in cotton as well as assessing the effect of providing flowering cotton plants with access to honeybees for different time periods (e.g. 25 d, 15 d, 0 d). In this year, I used double the hive stocking rate of (16 colonies / ha) than in the previous year, because in 2005-6 I observed few bees in cotton flowers. I also conducted a preliminary investigation to assess whether there was any gene flow over a 16 m distance from Bt cotton to conventional cotton, in the presence of a relatively high honeybee population. Both of my field experiments showed that honeybees significantly increased cotton yield via increased boll set, mean weight of bolls, number of seeds / boll, and weight of lint / boll. It was obvious that cotton flowers, and particularly cotton pollen, were not attractive to honeybees, and this was also reflected in the low proportion (5.3% w/w) of pollen from cotton collected in the pollen traps. However, flower visitation rate was generally above the 0.5% level regarded as optimal for cross-pollination in cotton, and this was reflected in increased yield parameters. I recorded a gene flow of 1.7 % from Bollgard®II cotton to conventional cotton, over a distance of 16 m. This is much higher than had previously been reported for Australia, and may have been a result of high honeybee numbers in the vicinity, associated with my managed hives. In an attempt to attract more honeybees to cotton flowers, I conducted an investigation where I applied synthetic Queen Mandibular Pheromone (QMP) (Fruit Boost®) at two rates, 50 QEQ and 500 QEQ / ha, and for two applications, 2 d apart. Neither rate of QMP increased the level of bee visitation to flowers, either on the day of application or the subsequent day. There was also no increase in boll set or yield in plants treated with QMP. My observations of honeybee behaviour in cotton brought some interesting findings. First, honeybees totally ignored extra floral nectaries. Second, most flower-visiting honeybees collected nectar, but the overwhelming majority of them (84%) collected floral nectar from outside flowers: this meant these bees did not contribute to pollination. Those nectar gatherers which entered flowers did contribute to pollination. However, they were observed to exhibit rejection of cotton pollen by scraping pollen grains from their body and discarding them, prior to returning to their hives. Pollen gatherers collected only small, loose pellets from cotton. SEM studies showed that cotton pollen grains were the largest of all pollen commonly collected by bees in my investigations, and that they also had large spines. It is likely that these characteristics make cotton pollen unattractive to honeybees. Another possible reason for the unattractiveness of cotton flowers was the presence of pollen beetles, Carpophilus aterrimus, in them. I conducted a series of studies to determine the role of pollen beetles in pollination of cotton. I found that they did not contribute to pollination at low levels; at high populations they damaged flowers (with ≥ 10 beetles / flower, no flowers set bolls); and that honeybees, when given the choice, avoid flowers with pollen beetles. Because the insecticide fipronil was commonly used in Australian cotton at flowering time, and because I had some experience of its toxic effects against honeybees in my field investigations, I conducted a series of laboratory and potted plant bioassays, using young worker bees. The studies confirmed its highly toxic nature. I recorded an acute dermal LD50 of 1.9 ng / bee, and an acute oral LC50 of 0.62 ppm. Fipronil’s residual toxicity also remained high for an extended period in both laboratory and potted plant trials. For example, when applied to cotton leaves in weather-exposed potted cotton plants, it took 25 d and 20 d for full and half recommended rates of fipronil, respectively, to become non- toxic to honeybees. I had previously investigated whether a shorter period of exposure of cotton plants to honeybees would contribute adequately to increased yield, and concluded that a 10 d window within a 25 d flowering period would contribute 55% of the increase in total weight of bolls contributable to honeybee pollination, but only 36% of the increase in weight of lint. Given the highly residual activity of fipronil I recorded, the only opportunity for an insecticide-free period during flowering would be at its commencement. I concluded that, while there is evidence that honeybees can contribute to increased cotton yield in Bt cotton in Australia, this is unlikely with the continued use of fipronil at flowering. / Doctor of Philosophy (PhD)

ROP GTPases are crucial regulators of pollen tube growth. The Rop GTPase family in maize consists of nine known rop genes, ropl-rop9. A subset of these genes (rop2, rop8, and rop9) are expressed in pollen. The rop2 and rop9 genes are a highly conserved duplicate gene pair of ancient origin. The rop2/rop9 duplicate gene pair displays differential expression in mature and germinated pollen, suggesting different roles for the genes in the process of male gametophyte development. To explore ROP2 function in maize, five Mutator transposon insertions in the rop2 gene were isolated (rop2::Mu alleles). I showed that three of the rop2::Mu alleles displayed reduced transmission through the male and were associated with reduced levels of ROP2-mRNA. Interestingly, the rop2::Mu male-specific transmission defect was apparent only when wild-type pollen was also present, an indication that the mutation reduces the competitive ability of the rop2 gametophytes. Dual pollination and pollen mixing experiments indicated that this competitive disadvantage is expressed by the majority of the mutant gametophytes, and that expression of the phenotype is associated with a delay in the ability to accomplish fertilization. Using the waxy phenotypic marker (linked to rop2 via a reciprocal translocation) to distinguish between rop2::Mu and wild-type pollen derived from heterozygous plants, I demonstrated that the delay is associated with a defect in early progamic development (i.e, germination and early pollen tube growth). The defect was detectable in vivo as early as 15 minutes after pollination. However, quantitative measurements provided no indication that the rop2 mutation affects pollen tube growth in the style. Finally, investigations focusing on the final stages of pollen function raise the possibility that a defect in the very last stages (i.e. either pollen tube guidance through the micropyle to the egg sac, or fertilization of the egg sac) may also contribute to the rop2 mutant delay. This work provides direct in vivo evidence confirming a role for Rop in male gametophyte development, and is the first study to demonstrate a role for Rop in the early stages of post-pollination gametophytic function. / Graduation date: 2005

"Why have separate sexes?" is a fundamental question in biology and has been investigated intensively since Darwin first proposed two hypotheses: Separate sexes evolve 1) to avoid detrimental effects of self-fertilization or 2) to improve allocation of scarce resources to each sex's reproductive function. In animal-pollinated plants dioecy is hypothesized to be favored when small, generalized pollinators predominate because they increase self-fertilization rates of plants. I concurrently tested Darwin's hypotheses using Echinocereus coccineus (Cactaceae). I found, as predicted, when specialized pollinators (hummingbirds) were excluded from plants and only generalized pollinators (bees) visited, selfing rates of plants increased 238%. High selfing rates produced by bees, along with high levels of inbreeding depression in E. coccineus populations, were sufficient to select for dioecy, even in absence of any reallocation of resources by the plants. Hummingbirds may also maintain high levels of gene flow among E. coccineus populations: populations blooming synchronously with the hummingbird migration had lower FST values (i.e. less genetic divergence) than asynchronously blooming populations. Population genetic data also suggest frequency of dioecious populations should be negatively correlated with both hummingbird abundance and elevation. Structural equation models of effects of hummingbird abundance, elevation and annual rainfall (resources) on dioecy across the range of E. coccineus supported this hypothesis. All predictor variables were either directly or indirectly negatively correlated with dioecy. Only models with hummingbird abundance as a direct predictor of dioecy significantly fit the data. Darwin's first hypothesis was again supported: as hummingbird abundance declines (and selfing rates increase), the incidence of dioecious populations increases. Finally, to determine how changes in pollinator type may be globally influencing the evolution of dioecy, I conducted a literature survey of studies noting changes in pollinator types along elevation gradients and compared these to studies of dioecy along elevation gradients. The abundance of specialized insect pollinators declines with elevation, often leaving only generalized pollinators at the highest elevations: Proportion of dioecious species in communities increases accordingly with elevation. I propose elevation may be useful as a worldwide predictor of dioecy because it effects pollinator community composition and amount of inbreeding depression expressed in plant populations.

Pollinator sharing in mixed species communities is expected to significantly contribute to mating patterns in contemporary populations but may also affect the evolutionary trajectory of traits associated with plant mating. In this thesis, I considered how the spring environment and pollinator sharing may contribute to the widespread convergence in traits among spring flowering species using comparative biology. The proposed correlation between a spring flowering phenology and white or light floral colour, fleshy fruits, woody growth forms and understory occupation is confirmed. In addition, I examined the effects of pollinator responses to community and population traits to determine the relative importance of inter- and intraspecific interactions in pollinator mediated reproductive success of a spring flowering species, Trillium grandiflorum. In this study, the reproductive success of T. grandiflorum was pollen limited. However, the magnitude of pollen limitation was influenced only by intraspecific density and varied independently of community diversity. The results of this thesis contribute significantly to our understanding of pollinator-mediated interactions in spring flowering communities but also highlight future avenues of investigation.

Pollinator sharing in mixed species communities is expected to significantly contribute to mating patterns in contemporary populations but may also affect the evolutionary trajectory of traits associated with plant mating. In this thesis, I considered how the spring environment and pollinator sharing may contribute to the widespread convergence in traits among spring flowering species using comparative biology. The proposed correlation between a spring flowering phenology and white or light floral colour, fleshy fruits, woody growth forms and understory occupation is confirmed. In addition, I examined the effects of pollinator responses to community and population traits to determine the relative importance of inter- and intraspecific interactions in pollinator mediated reproductive success of a spring flowering species, Trillium grandiflorum. In this study, the reproductive success of T. grandiflorum was pollen limited. However, the magnitude of pollen limitation was influenced only by intraspecific density and varied independently of community diversity. The results of this thesis contribute significantly to our understanding of pollinator-mediated interactions in spring flowering communities but also highlight future avenues of investigation.

Background and Aims Sexual deception is a species-specific pollination strategy commonly found in Orchidaceae. Sexually deceptive orchids lure male insect pollinators by mimicking the sex pheromones and/or appearance of female insects, which elicit copulatory behaviour with the flower by the male insects. This specialised pollination strategy has recently been found in a Pterostylis species in Australia. Pterostylis orchids also occur in New Zealand, although very few studies have been done on this genus, and no such specialised insect pollination strategy has been documented in New Zealand. Methods I investigated the breeding system and pollinators of three Pterostylis spp. to determine whether sexual deception may be operating in P. oliveri, P. irsoniana and P. venosa growing in native beech forests in Arthur's Pass. We also investigated the floral headspace volatiles of P. oliveri to determine which compounds are present, and which may be responsible for pollinator attraction. Key Results Breeding system experiments suggest that P. oliveri and P. irsoniana are self compatible, but exclusively dependent on insects for pollination. Only male fungus gnats (Diptera: Mycetophilidae) were found carrying pollinia attached to their thoraxes in traps set up over the flowers. Insect identification and ITS DNA analysis of the pollinia showed that each orchid species was pollinated by a specific fungus gnat species; Mycetophila latifascia males found with pollen of P. oliveri; Morganiella fusca males found with pollen of P. irsoniana; and Tetragoneura sp. males found with pollen of P. venosa. Field tests of an unidentified compound found in headspace volatiles of P. oliveri did not attract any Mycetophila latifascia males. Conclusions These results indicate that pollination via sexual deception may be operating in these three Pterostylis spp. However, further floral volatile analyses are required to confirm whether the flowers emit volatile compounds that resemble the sex pheromones of the specific pollinators.

Plants with similar traits compete for resources. If related taxa share similar traits, phylogenetic relationships may predict competitive outcomes. Although plants compete for pollinators, flowers are rarely considered in community-assembly theory. I tested the hypothesis that plant communities are structured by competition for pollination. I inventoried communities at three spatial scales, measured seven flower traits, and tested the observed patterns against those generated by a null model to judge if community members were more or less similar in floral traits than expected by chance. I also measured the phylogenetic relatedness of community members to gauge trait-conservatism. Clustering of visually attractive traits suggests they promote facilitation of pollinators while over-dispersion of morphological traits suggests they partition pollinators in to avoid competition. Communities were phylogenetically even, but relatedness did not explain floral trait patterns. I suggest that flowers represent an ecological niche through which species can be sorted.

Floral traits are typically associated with specific groups of pollinators. Yet, many flowering plants are pollinated by more than one group of pollinators. To explore the influence of multiple effective pollinators on floral traits, I examined how the pollinator assemblage of Polemonium brandegeei affects pollen movement and selection on floral traits. I documented phenotypic and genetic variation in floral traits, and quantified the effectiveness of floral visitors. In natural populations, I quantified pollen removal and deposition over two consecutive flowering seasons. I then performed a pair of novel array experiments to estimate selection through female (seeds set) and male (seeds sired) function on floral traits by two important pollinators, hawkmoths and hummingbirds. My analysis of field populations demonstrated that height and relative positioning of sex organs were important for the removal and deposition of pollen. Individuals of P. brandegeei displayed a large degree of continuous and heritable variation, particularly in the relative positioning of sex organs. Plants in the same population displayed stigmas recessed below (reverse herkogamy), or exserted above anthers (approach herkogamy). My array experiments determined that variation in herkogamy is likely maintained through contrasting selection, because hawkmoths selected for recessed stigmas but hummingbirds selected for exserted stigmas. While my results were the first to detect selection for reverse herkogamy by hawkmoths, I also identified selection for traits that are typically associated with both pollinators. For example, hawkmoths selected for narrow corolla tubes, and hummingbirds selected for longer corolla tubes. The selection I detected on floral traits through female function was generally stronger than through male function, which runs counter to traditional theory of gender-biased selection. My findings indicate that floral traits can conform to intermediate dimensions between the optima of two pollinators (herkogamy), or appear specialized to one pollinator (tube length). Therefore, cumulative arrangements of floral traits (floral design) can effectively function under pollination by two pollinators that exert different selection pressures on traits.

I investigated how orchid biology, floral morphology, and diversity of surrounding floral and pollinator communities affected reproductive success and hybridization of Cypripedium candidum and C. parviflorum. Floral dimensions, including pollinator exit routes were smallest in C. candidum, largest in C. parviflorum, with hybrids intermediate and overlapping with both. This pattern was mirrored in the number of insect visitors, fruit set, and seed set. Exit route size seemed to restrict potential pollinators to a subset of visiting insects, which is consistent with reports from other rewardless orchids. Overlap among orchid taxa in morphology, pollinators, flowering phenology, and spatial distribution, may affect the frequency and direction of pollen transfer and hybridization. The composition and abundance of co-flowering rewarding plants seems to be important for maintaining pollinators in orchid populations. Comparisons with orchid fruit set indicated that individual co-flowering species may be facilitators or competitors for pollinator attention, affecting orchid reproductive success.

Pollination and fruit set of four species of native New Zealand flowering plant species were examined through two field seasons. Bird exclusion, pollinator exclusion, natural and supplemental pollination treatments were initiated on individuals of Cordyline australis (Cabbage tree), Phormium tenax (Flax), Kunzea ericoides (Kanuka), and Pseudopanax arboreus (Five-finger). The species differed in the self-compatibility as well as in their floral syndrome. No species showed any evidence of pollen limitation, and two species. K. ericoides and P. arboreus set fruit from more than 70% of their flowers. The response of fruit set to treatment in C. australis varied from season to season, with birds appearing important to pollination in the first but not the second field season, while birds were important in pollination across both seasons for P. tenax. K. ericoides was resilient to treatment, setting high fruit set in every treatment, compared to P. arboreus which set high fruit set when pollinators had access, but low when all pollinators were excluded. No trends relating to fruit set or PLI and self-compatibility or floral syndrome were found. That there was no evidence of pollen limitation for any species, despite variation in fruit set from some treatments, indicates that these species are performing well and not at risk of decreased population size due to pollen limitation.