ABSTRACT The human language system is extraordinary, particularly in its capacity to handle more than one system of lexical and syntactic knowledge within a single individual. It is estimated that over half of the global population is bilingual or multilingual. However, the cognitive mechanisms that enable these individuals to monitor and choose the appropriate language for comprehension and production are not well understood. An important mechanism that contributes to efficient reading is predictive processing. Many studies have shown evidence for lexical-semantic prediction during sentence processing in constraining contexts in monolinguals and proficient bilinguals. But differences between monolingual and bilingual lexical-semantic prediction have been observed. Specifically, bilinguals will adjust their predictions based on the context of language use, the level of proficiency they have in each language, which language is being used and how recently it has been used (Valdez-Kroff & Dussias, 2022). Prior research has also shown evidence for syntactic prediction in both monolingual and bilingual language users. But it is currently not known whether bilinguals predict the word category of imminent words in the context. A previous study in English monolinguals used eye-tracking during sentence reading and showed that they do anticipate upcoming word forms, even when the critical words are semantically unpredictable in the sentence context (Brothers & Traxler, 2016). In the English language, the Subject-Verb-Object (SVO) order is a crucial factor in parsing sentences because this structure is highly fixed, allowing English speakers to predict whether an imminent word is likely to be a verb or a noun. Although Spanish also follows the SVO order, its richer morphological system makes word order more flexible. This thesis addresses the important question of whether Spanish-English bilinguals predict word category information in English as monolingual English speakers do, or if the greater flexibility of word order in Spanish influences their predictions in English. If Spanish-English bilinguals predict word category information in English like monolingual English speakers, then they should be able to anticipate whether an imminent word is a verb or a noun based on the fixed SVO order in English, resulting in similar reading behavior during sentence processing. In contrast, if the flexible word order in Spanish modulates Spanish-English bilinguals' predictions in English, then their anticipatory processing might show greater variability compared to monolinguals. This could manifest as a different pattern of eye movements when predicting upcoming words in English sentences. Understanding whether proficient Spanish-English bilinguals predict word category information in English like monolingual English speakers, or if their predictions are influenced by the flexible word order in Spanish, has significant implications for the study of bilingual sentence processing. It will contribute to our understanding of how bilinguals manage and switch between different syntactic systems. Furthermore, it can inform theories of predictive processing by highlighting the extent to which linguistic experience in one language affects processing in another, thereby contributing to a deeper understanding of the bilingual mind. In this dissertation, I will examine word form prediction during the reading of English sentences in proficient Spanish-English heritage speakers using the eye-tracking technique. I used the experimental design of Brothers and Traxler (2016) with sentence materials that were adapted for this thesis, and first established that the results of Brothers and Traxler were replicated in a new group of English monolinguals with these new sentence stimuli. I then performed the same experiment with a group of proficient Spanish-English bilinguals to examine if exposure to two languages that differ in flexibility of word order modulates the prediction of imminent nouns or verbs in sentence contexts. The organization of this thesis is as follows: In Chapter 1, I will begin with a review of general background literature on sentence processing, syntax, the processing of nouns and verbs, and predictive processing in monolinguals. This will be followed by a discussion of these topics in bilinguals. The chapter will then highlight the differences between Spanish and English word order and the morphological constraints that may affect noun or verb form prediction during sentence processing. Additionally, the chapter will introduce the eye-tracking methodology and discuss how it provides sensitive measures of prediction during natural reading of sentences. In Chapter 2, I tested a group of monolingual English speakers with the paradigm of Brothers and Traxler (2016). Specifically, I used the eye-tracking methodology with a gaze-contingent boundary and manipulated critical target words within sentences that were low in semantic predictability but syntactically constrained towards a noun or a verb, varying whether the parafoveal preview word was congruent or incongruent with the syntactic category of the critical target words. This study replicated the results of Experiment 2 from Brothers and Traxler (2016) in native English speakers using different sentence stimuli: readers again anticipated the identity of imminent nouns and verbs in sentence contexts. In addition to the importance of replicating this finding, the results of this study also provided the control study for the Spanish English bilinguals. In Chapter 3, I examined a group of proficient Spanish-English bilinguals using the same paradigm and stimuli as in Chapter 2. We observed notable differences in the eye movement patterns between bilingual and the monolinguals during reading. Skip rate, a primary measure of prediction, revealed that Spanish-English bilinguals did not show a significant difference in skipping critical target words based on the congruence of the preview with the word category, unlike monolinguals who had a higher skip rate when previews were congruent. However, for reading time measures we found shorter durations for congruent previews and longer durations for incongruent previews for first fixation durations and total reading time, just as for the monolingual English speakers in Chapter 2. In Chapter 4, I provided an overall discussion and conclusion of the results and provided some suggestions for future research. Although bilinguals skipped significantly less and had longer reading durations, the skip rates and reading durations were largely similar in monolinguals’ processing in L1. I concluded that the findings suggest that proficient bilinguals’ prediction mechanisms in L2 resemble those of monolinguals and can be shaped by experience and language exposure. This adaptability underscores the complex interplay between language and experience in bilingual processing and highlights the need to consider how language context may play a role in how bilinguals integrate and navigate different language structures during comprehension.

Streptomyces coelicolor is a soil-dwelling prokaryote representative of an order of Gram-positive filamentous bacteria designated Actinomycetales, the richest known bacterial source of bioactive natural products. These specialized / secondary metabolites (including macrolides, aminoglycosides, tetracyclines, cephalosporins and polyketides) often have potent anti-microbial activity and, hence, have been a staple of modern medicine. Genome sequencing has revealed that a given actinomycete genome encodes the potential biosynthetic capacity to produce from 10 to 30, or even more, distinct natural products. S. coelicolor is a model organism for studying actinomycete development, genetics, and specialized metabolism. However, in its natural environment, S. coelicolor must engage with a heterogenous and complex collection of other microorganisms, as well as with abiotic factors in soil. Therefore, in the research described in this dissertation, I explored the role of interspecies interactions in regulating the production of specialized metabolites by S. coelicolor. To obtain insight about how production of these compounds is induced, I exploited the fact that S. coelicolor produces two, readily visualized, red-colored antibiotics (actinorhodin and undecyl-prodigiosin), but only when placed in close contact to certain other bacterial genera. I used extensive analysis of gene expression (RNA-seq) during interspecies interactions and, thereby, identified involvement of conservon8 in controlling induction of the red-colored antibiotics. A conservon is an operon so named because it encodes a characteristic set of four proteins and the S. coelicolor genome encodes 13 such cassettes that are highly homologous to each other. I constructed single-gene deletions in conservon 8 and determined, using transcriptome analysis, that CvnA8 and CvnF8 seem to function together. On the basis of my data, combined with bioinformatic analysis, I propose that CvnA8 is a membrane-bound signaling protein and CvnF8, a GAF domain-containing protein (an element that typically binds ligands and allosterically regulates proteins), serves as a sensor for CvnA8. From the gene expression patterns of deletion mutants, I deduce that CvnC8 and CvnD8 likely function together and hypothesize that CvnC8 is a transcription factor that is negatively regulated by CvnD8, which resembles a small RAS-like GTPase. Similarly, I observed that a ∆cvnB8 mutation has an intermediate effect on red pigment production and genome wide expression patterns, suggesting that it may connect CvnA8/F8 and CvnC8/D8, perhaps through regulation of the CvnD8 GTPase. Finally, my expression data additionally demonstrate that these conservon 8 factors also control expression of the gene clusters for at least two other specialized metabolites aside from the two red pigments, including a cryptic cluster that is predicted to encode a lanthipeptide.

Wildfires affect soils in multiple ways, leading to numerous challenges for colonizing microorganisms. Fire increases soil hydrophobicity and reduces nutrient availability, imposing harsh conditions that foster intense competition. Several fire-adapted microorganisms are known to metabolize pyrolyzed organic compounds, and likely lie at the forefront of postfire ecosystem recovery. However, non-catabolic strategies that postfire microorganisms use to thrive in burned soils remain unexplored. We envision that microbial metabolites with specialized physicochemical properties, such as surface- or redox-activity, may be particularly primed to address postfire challenges. In this dissertation, we explore potential roles of specialized metabolites in facilitating bacterial adaptations to postfire conditions.

Through bioactivity screening of bacterial isolates from burned soils, we discovered that several Paraburkholderia spp. isolates produced a set of unusual rhamnolipid surfactants with a natural methyl ester modification. These rhamnolipid methyl esters (RLMEs) exhibited enhanced antimicrobial activity against other postfire microbial isolates, including pyrophilous Pyronema fungi and Amycolatopsis bacteria, compared to the typical rhamnolipids made by organisms such as Pseudomonas spp. Identification of the rhamnolipid biosynthesis genes in the postfire isolate Paraburkholderia kirstenboschensis F3 led to the discovery of rhlM, whose gene product is responsible for the unique methylation of rhamnolipid substrates. RhlM is the first characterized bacterial representative of a large class of integral membrane methyltransferases that are widespread in bacteria. RLMEs also showed enhanced surfactant properties and facilitated bacterial motility on agar surfaces. In vitro assays further demonstrated that RLMEs improved aqueous solubilization of polycyclic aromatic hydrocarbons, which are potential carbon sources found in char. These results indicate multiple roles for RLMEs in the postfire lifestyle of Paraburkholderia isolates, including enhanced dispersal, solubilization of potential nutrients, and inhibition of competitors. Beyond surface-active antibiotics from Paraburkholderia, we studied the specialized metabolite repertoire of a postfire Streptomyces isolate that likely represents a new species. This streptomycete produces a strong siderophore and potent redox-active antibiotic, which may facilitate inorganic nutrient acquisition and mediate interspecies competition. Altogether, our findings shed new light on the chemical adaptations that bacteria employ in order to navigate, grow, and outcompete other soil community members in postfire environments.

Microbes occupy diverse habitats, forming interconnected, dynamic communities. Microbiomes associated with various plant structures often contain members with the potential to make specialized metabolites, e.g. molecules with antibacterial, antifungal, or siderophore activities. However, when and where microbes associated with plants produce specialized metabolites, and the potential role of these molecules in mediating intra-microbiome interactions, is not well understood. Root nodules of legume plants are organs devoted to hosting symbiotic bacteria that fix atmospheric nitrogen and have recently been shown to harbor a relatively simple accessory microbiome containing members with the ability to produce specialized metabolites in vitro. Based on these observations, we sought to develop a model nodule microbiome system for evaluating microbial specialized metabolism in planta. Starting with an inoculum derived from field-grown Medicago sativa nodules, serial passaging through gnotobiotic nodules yielded a simplified accessory community composed of four members: Brevibacillus brevis, Paenibacillus sp., Pantoea agglomerans, and Pseudomonas sp.. Some members of this community exhibited clear cooperation in planta, while others were antagonistic and capable of disrupting cooperation between other partners. Using matrix-assisted laser desorption/ionization imaging mass spectrometry, we found that metabolites associated with individual taxa had unique distributions, indicating that some members of the nodule community were spatially segregated, accounting for their ability to co-exist. Finally, we identified two families of molecules produced by B. brevis in planta as the antibacterial tyrocidines and a novel set of gramicidin-type molecules, which we term the britacidins. Collectively, these results indicate that in addition to nitrogen fixation, legume root nodules are likely also sites of active antimicrobial production.