真菌生物技術平台於蟲草素生產與菌絲體皮革前驅物開發之建立

Abstract

本研究以培養工程為共同核心，建立北蟲草功能性代謝物平台與紡織基材支撐菌絲體前驅物平台。第一部分依實驗時間軸整合王士誠（2020）之 purine 與 amino-acid metabolism 證據、本研究參與之未發表實驗室重分析、液態發酵、胞外分布、非標靶代謝體、EV 尺度顆粒及標的探索。既有體學資料用於選擇可測試變因；cysteine 的 wet-lab 反應才是後續硫化學實驗的直接觸發點，避免倒置因果。北蟲草部分以液相蟲草素濃度為主要生產指標，並將乾菌體量、biomass-normalized extracellular index 與 extracellular fraction 分開解讀。靜置培養、2 g/L cysteine 與低劑量 NaHSO₃ 提高液相濃度，但 biomass 與各終點的方向並不一致。非標靶代謝體未註解出蟲草素，卻保留條件分群與 purine/nucleoside 相關訊號，故定位為跨體學收斂而非正式、樣本配對的多體學整合或代謝通量證明。菌絲條件培養液經兩次 100,000×g 超高速離心取得 EV 富集製備物；NTA 支持 EV 相容粒徑。實際 HPLC chromatogram 在蟲草素保留區出現微量候選峰，來源試算為 (3,183×5−4,816.6)/18,155≈0.6113 ppm；此值低於有效校正範圍且每顆粒換算不一致，因此只能描述為低量、暫定存在，仍須標準品加標及 target LC–MS(/MS) 確認。果實體培養罐以無菌水淹沒浸泡一週後，回收液沿用相同超高速離心流程；同一批三個 NTA capture 的 FTLA 與 classical 輸出顯示近似粒徑，但濃度可靠性受報告警示，且不能由浸泡液推論主動分泌。AI 壓縮檔重建顯示其使用 DeepPurpose 的 MPNN_CNN_BindingDB 預訓練模型，以單一蟲草素 SMILES 對 54,825 條 UniParc 蛋白序列進行分批序列式 DTI 預測；程式碼與說明對 score 方向的描述互相矛盾，且缺少 protein_info 對照表與完整輸出，故候選蛋白保留為既有專案結果，不視為 docking 或已確認標的。第二部分建立 attachment–penetration–compaction 與 scale-up gate；非本實驗室或非本研究範圍之文獻比較、NTU Experimental Forest 與 TTRI 材料測試均移至補充資料。整體新意在於以可追溯的體學收斂選擇變因、分離相互衝突的 phenotype，並建立有明確證據邊界的決策框架。

This study establishes two cultivation-engineering platforms: functional-metabolite development in Cordyceps militaris and a textile-supported mycelium precursor. Part I preserves the experimental chronology by using Wang (2020) as the legacy purine/amino-acid rationale, a cysteine wet-lab response as the direct sulfur-follow-up trigger, and a later unpublished laboratory reanalysis as convergent support rather than retrospective cause.

Liquid cordycepin titer, dry biomass, biomass-normalized extracellular index, and extracellular fraction were interpreted as separate endpoints. Untargeted metabolomics did not annotate cordycepin but retained condition structure and purine/nucleoside associations; this is cross-omics convergence rather than a fully sample-matched multi-omics integration or flux analysis. Mycelial conditioned medium yielded EV-sized particles after two 100,000×g ultracentrifugation steps. The actual exploratory chromatogram contained a small retention-time-compatible peak; the source back-calculation, (3,183×5−4,816.6)/18,155≈0.6113 ppm, was outside the validated range and produced an unresolved per-particle discrepancy. Cordycepin is therefore described as provisionally present at low abundance pending spike-in and targeted confirmation.

A fruiting-body culture-jar soak was clarified and processed through the same ultracentrifugation workflow. FTLA and classical NTA analyses of the same three captures gave similar size estimates, but the report warned that concentration could be unreliable and the soak cannot distinguish active release from passive leakage. Reverse engineering of the target-screening archive showed a DeepPurpose MPNN_CNN_BindingDB sequence-based DTI screen pairing one cordycepin SMILES with 54,825 UniParc sequences. A score-direction conflict and missing mapping/output files prevent full ranking reproduction; the retained candidates are neither docking results nor confirmed targets.

Part II formalizes attachment–penetration–compaction and stage-gated scale-up for a textile-supported mycelium precursor. Literature comparisons and NTU Experimental Forest/TTRI measurements outside the thesis laboratory and project scope are supplementary. The thesis novelty is an evidence-bounded decision framework that turns independent omics layers into testable variables while demonstrating that growth, titer, extracellular recovery, and particle yield are separable response axes.