利用開放原始碼之深度學習軟體評估二甲基亞碸對FMR1突變斑馬魚的治療效果

Abstract

背景簡介：FMR1 (fragile X mental retardation 1) 基因位於人類的Xq27.3基因座，當其CGG核苷酸重複序列 (nucleotide repeat) 過長時，將會造成X-染色體脆折症 (fragile X syndrome, 簡稱 FXS)。現已開發出可用於研究FXS的斑馬魚 (Danio rerio) 突變品系，相較於囓齒類動物，斑馬魚具有快速發育、幼體透明和高繁殖力等研究優勢。前人的囓齒類動物研究發現，FMR1突變與發炎反應間存有關聯，特別是壓力條件下的細胞激素 (cytokine) 和c-Fos的表現。已知二甲基亞碸 (dimethyl sulfoxide，簡稱DMSO）為一種的消炎藥 (anti-inflammation drug)，具有多種免疫調節效果和臨床應用。 本計劃藉由建立清晰的FMR1突變行為和基因表達表型後，來探討DMSO當作治療FXS藥物的可能性。研究方法：利用顯微鏡觀察受精後 (post-fertilization, dpf) 三天大幼魚的心率和體長，以評估DMSO處理對FMR1突變品系胚胎及幼體發育的影響。採用公開性的「深度學習軟件」，進行多動物同時自發性運動跟踪法 (multi-animal locomotor tracking)，分別對幼魚進行2分鐘 (n = 10) 及對成魚進行5分鐘 (n = 5) 的記錄及分析。通過觀察幼蟲的趨態性 (thigmotaxis) 和淺水內聚性 (shoal cohesion) 來評估幼魚的數焦慮行為 (anxiety-like behavior)，並使用新型水箱潛水模式 (novel tank) 對成魚進行焦慮評估。利用C-start反射評估幼魚的學習行為通過觀察幼蟲的趨態性 (thigmotaxis) 和淺水內聚性 (shoal cohesion) 來評估幼魚的焦慮程度，並使用新型水箱潛水模式 (novel tank) 對成魚進行焦慮評估。利用C-start反射評估幼魚的非聯結型學習行為 (non-associative learning)，而成魚則採用抑制性逃避模式 (inhibitory avoidance) 來評估學習反應。並以2項選擇模式 (two choice paradigm) 來評估成魚的社會興趣反應 (social interest paradigm)。最後透過定量聚合酶連鎖反應 (quantitative PCR) 評估全腦中FMR1和細胞激素 (cytokines) 的基因表現。實驗結果：突變品系成魚的行為表型分析顯示，同型合子 (homozygotes) 出現過動的反應 (hyperactivity)，異型合子 (heterozygotes) 對陌生魚的社會興趣 (allospecific social interest) 增加，同型合子中的焦慮反應及恐懼學習(fear learning)減少。DMSO的長期投予最佳濃度為0.05％，可恢復突變品系幼魚的趨態性(thigmotaxis)和淺灘凝聚性 (shoal cohesion) 行為表型。在5-dpf時觀察到誘導的C-起始反射 (strike induced C-start) 的減少，暗示該濃度的DMSO對毛細胞可能具有潛在的毒性，然而在7-dpf的幼魚身上，並未呈現空間運動的異常。該濃度的DMSO投予能夠改善突變品系成魚的焦慮和學習缺陷等行為表型。儘管DMSO處理不能使FMR1的表現恢復到正常水平，但能顯著改善c-Fos及適度改善細胞激素因子 (IL-1β，IL-6和IL-10) 的表現。結論：本計劃的結果顯示，1）FMR1突變品系的同型合子為適合的FXS動物模型，2）DMSO的使用可減少突變品系幼魚的異常行為，3）DMSO可降低突變品系成魚的異常行為，並使其腦中發炎反應基因 (inflammation genes) 表現減少。討論：突變品系成魚的運動，焦慮和恐懼學習結果與以前的囓齒動物和斑馬魚FXS模型大致相同。但在社會興趣的結果有差異，前人報導將突變品系對同種 (cospecific) 的興趣大於同種異體 (allospecific)。在幼魚實驗中的一些新發現，包括FMR1突變體在5-dpf時明顯的非聯結型學習 (non-associative learning) 障礙，焦慮以及淺灘凝聚力的增加。過去的文獻推測的腦部發炎反應基因增加，本計劃發現FMR1 KO樣本腦中神經炎症基因被下調。進一步探討特次腦區的特異性表現 (region-specific expression)，特別是在端腦內側和外側大腦皮層 (telencephalic medial and lateral pallium) ，可能會得到與囓齒類動物一致的結果。

Introduction: The fragile X mental retardation 1 (FMR1) gene is prone to developing a series of CGG repeats at locus Xq27.3 in humans, when sufficiently long, will lead to Fragile X syndrome (FXS). Zebrafish (Danio rerio) mutant models have been developed to study FXS, as they offer several advantages over rodents such as rapid development, transparent larvae, and high fecundity. The link between FMR1 mutation and inflammation has been described in rodent models, particularly regarding cytokine and c-Fos expression under inflammatory or otherwise stressful conditions. Moreover, DMSO is a known anti-inflammatory drug known to possess several immunomodulatory effects and clinical applications. This project aimed to explore the potential of DMSO as a therapeutic agent in the treatment of FXS by establishing clear FMR1 mutant behavioral and gene expression phenotypes.Methodology: The microscopic assessment of larval heart rate and body length were performed at three days post-fertilization (dpf) to evaluate the effect of FMR1 knockout and embryonic DMSO treatment on gross anatomical development. Multi-animal locomotor tracking was conducted for 2 minutes (larva; n=10) or 5 minutes (adult; n=5) using open-source, deep-learning software. Anxiety-like behavior was examined in larvae by observing thigmotaxis, shoal cohesion, and adults using the novel tank dive paradigm. Non-associative learning was evaluated in larvae by exploiting the C-start reflex and adults using the inhibitory avoidance paradigm. Social interest was evaluated in adults using a 2-choice paradigm. Whole-brain gene expression of the FMR1 and cytokines were evaluated by quantitative PCR.Results: Identification of adult mutant behavioral phenotypes revealed hyperactivity among homozygotes, increased allospecific interest among heterozygotes, and decreased anxiety and fear learning in homozygotes. Optimal DMSO exposure was found to be a chronic dose at 0.05% concentration, which recovered mutant larval thigmotaxis and shoal cohesion phenotypes. While a potential toxic effect was observed as a loss of strike-induced C-start reflex at 5-dpf, no deficit in reactivity, locomotion or spatial orientation was observed at 7 dpf. This treatment protocol was also able to recover diminished anxiety fear learning in mutant adults. Although DMSO treatment did not recover target gene expression back to WT levels, a significant improvement was seen in c-Fos expression as well as a modest improvement among cytokines IL-1β, IL-6, and IL-10.Conclusion: Evidence gathered during this project suggests that 1) FMR1 KO homozygotes are a suitable FXS knockout model, 2) Administration of DMSO improves abnormal behavior in larval mutants, and 3) DMSO improves abnormal behavior in adult mutants, for whom inflammation genes are strongly downregulated in the brain.Discussion: Locomotor, anxiety, and fear learning results in adult mutants were consistent with previous rodent and zebrafish FXS models. However, social interest results differ from previous reports describing mutants as interested in conspecific over allospecific fish. My larval experiments demonstrate several novel findings, including an apparent non-associative learning deficit among FMR1 mutants at 5-dpf, increased anxiety, and an open-source method of quantifying shoal cohesion. Contrary to the hypothesized increase in brain inflammation based on past literature, neuroinflammation genes were strongly downregulated in whole-brain FMR1 KO samples. Further investigation of region-specific expression, particularly at the telencephalic medial and lateral pallium, may yield results in line with rodent results.