Abstract

The rapid growth in plastic production and use has created major challenges for waste management. Mismanaged plastics pose serious threats to ecosystems and human health. Biological deconstruction of plastics has attracted increasing attention as a potentially more sustainable solution, but its typical agents, microbes and enzymes, often show low efficiency against non-hydrolysable plastics such as polyethylene (PE). In contrast, the yellow mealworm can ingest various plastics and has been reported to consume PE at considerable rates. Investigating mealworm-based plastic degradation holds promise for developing engineered systems to upcycle plastic waste. However, as an emerging field, it remains constrained by limited mechanistic insight and the lack of a consolidated evidence base. To accelerate progress, this dissertation benchmarks large language models (LLMs) as research tools, performs microbial profiling to in vitro gut enrichment and traces the PE-derived carbon in mealworm metabolites. The information retrieval capabilities of LLMs can assist researchers by aggregating and structuring literature-derived evidence. However, concerns about the trustworthiness persist, given their noisy training databases and tendency to hallucinate. Retrieval-Augmented Generation (RAG) can complement LLMs by grounding responses in an updatable knowledge base, thereby enhancing traceability to source evidence. Using a literature corpus on mealworm-based plastic degradation and a curated benchmark of 50 quantitative queries and 50 open-ended queries, this dissertation examined GraphRAG, LightRAG, and a traditional RAG with five LLM models (GPT-4o, GPT-5, Deepseek-V3.1, Qwen-plus, and Llama-3.3). Based on 1/(1+APE) and F1 scores, LightRAG, a knowledge graph-based pipeline, yielded the largest improvement in information extraction. Additional experiments validated the answers from the leading configuration LightRAG + Llama. Overall, this dissertation demonstrates a reliable application of advanced LLMs in the field of mealworm-based plastic degradation, while highlighting challenges from higher-order tasks such as experiment design. While LLM-assisted retrieval can organize prior evidence, experimental studies are essential for resolving mechanisms underlying mealworm-driven PE deconstruction. To date, microbial contributions have been investigated mainly in vivo. Here, this dissertation enriched gut microbes in vitro using alkanes to probe their functions relevant to mealworm-based PE deconstruction. Microbes derived from mealworms on a PE + oat diet showed robust growth on C12 and C16 alkanes but did not grow on C8 alkane. Full-length 16S rRNA gene sequencing identified abundant species enriched on mid-chain alkanes, including Pseudomonas aeruginosa. PICRUSt-based functional inference predicted an increased abundance of oxidation-related functions in alkane-enriched communities, which could plausibly contribute to early oxidative steps during PE deconstruction. Collectively, this dissertation advances understanding of microbial mechanisms by profiling communities enriched on PE proxies, thereby narrowing putative microbial entry points. The deconstruction of PE by mealworms has been confirmed through multiple analytical methods but evidence for the assimilation of PE-derived carbon into host metabolism is still inconclusive. To examine this, this dissertation compared central metabolites in mealworms fed on an oat diet supplemented with 13C-labeled PE with those in a control group fed on oats only. The gut and the body tissues of mealworms were separated for metabolite extraction and untargeted metabolomic profiling was performed using liquid chromatography (UltiMate 3000 UHPLC, Thermo Scientific, USA) coupled to mass spectrometry. After peak picking in Maven and isotope correction using IsoCor, no significant increases were observed in the labeled fractions of key intermediates in the tricarboxylic acid (TCA) cycle and glycolysis. This suggests that PE derived carbon did not contribute to accumulation of mealworm biomass in abundance at current experiment setup. Future research should aim to further resolve the microbial mechanisms and PE deconstruction pathways mediated by mealworms. To achieve and improve PE deconstruction, the principles underlying the construction of microbial consortia should be further investigated. Specifically, the alkane enrichments obtained in Chapter 2 should be tested for their capability to deconstruct PE and the species interactions within the microbial consortium deconstructing PE should be characterized using metagenomic sequencing data. In parallel, additional LC-MS analyses are required to identify the metabolite biomarker and enriched metabolic pathways associated with mealworm-mediated PE deconstruction. In particular, the application of advanced denoising and annotation methods may enable more robust metabolite identifications. Coupled with isotopic labeling, LC-MS analysis can help elucidate PE deconstruction pathways by tracking PE-derived carbon through associated metabolic networks. Moreover, metabolomic profiling of PE-deconstructing cultures can be used to refine the genome-scale metabolic models of microbial communities, thereby facilitating inference of the metabolic exchanges between species. Collectively, these efforts would provide deeper insights into the mechanisms of PE deconstruction and can support the future development of engineered systems for plastic upcycling.

Committee Chair

Yinjie Tang

Committee Members

Arpita Bose; Fangqiong Ling; Marcus Foston; Mark Blenner; Yinjie Tang

Degree

Doctor of Philosophy (PhD)

Author's Department

Energy, Environmental & Chemical Engineering

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

4-29-2026

Language

English (en)

Download

Available for download on Tuesday, June 15, 2027

Included in

Engineering Commons

Share

COinS