Litron Laboratories, Rochester, New York, 14623.
Toxicol Sci. 2019 Aug 1;170(2):382-393. doi: 10.1093/toxsci/kfz123.
A tiered bioassay and data analysis scheme is described for elucidating the most common molecular targets responsible for chemical-induced in vitro aneugenicity: tubulin destabilization, tubulin stabilization, and inhibition of mitotic kinase(s). To evaluate this strategy, TK6 cells were first exposed to each of 27 presumed aneugens over a range of concentrations. After 4 and 24 h of treatment, γH2AX, p53, phospho-histone H3 (p-H3), and polyploidization biomarkers were evaluated using the MultiFlow DNA Damage Assay Kit. The assay identified 27 of 27 chemicals as genotoxic, with 25 exhibiting aneugenic signatures, 1 aneugenic and clastogenic, and 1 clastogenic. Subsequently, a newly described follow-up assay was employed to investigate the aneugenic agents' molecular targets. For these experiments, TK6 cells were exposed to each of 26 chemicals in the presence of 488 Taxol. After 4 h, cells were lysed and the liberated nuclei and mitotic chromosomes were stained with a nucleic acid dye and labeled with fluorescent antibodies against p-H3 and Ki-67. Flow cytometric analyses revealed that alterations to 488 Taxol-associated fluorescence were only observed with tubulin binders-increases in the case of tubulin stabilizers, decreases with destabilizers. Mitotic kinase inhibitors with known Aurora kinase B inhibiting activity were the only aneugens that dramatically decreased the ratio of p-H3-positive to Ki-67-positive nuclei. Unsupervised hierarchical clustering based on 488 Taxol fluorescence and p-H3: Ki-67 ratios clearly distinguished compounds with these disparate molecular mechanisms. Furthermore, a classification algorithm based on an artificial neural network was found to effectively predict molecular target, as leave-one-out cross-validation resulted in 25/26 agreement with a priori expectations. These results are encouraging, as they suggest that an adequate number of training set chemicals, in conjunction with a machine learning algorithm based on 488 Taxol, p-H3, and Ki-67 responses, can reliably elucidate the most commonly encountered aneugenic molecular targets.
描述了一种分层生物测定和数据分析方案，用于阐明导致化学诱导的体外非整倍性的最常见分子靶标：微管蛋白不稳定、微管蛋白稳定和有丝分裂激酶抑制。为了评估该策略，首先将 TK6 细胞暴露于 27 种假定的非整倍体化合物中的每一种，浓度范围从低到高。处理 4 和 24 小时后，使用 MultiFlow DNA 损伤检测试剂盒评估 γH2AX、p53、磷酸化组蛋白 H3（p-H3）和多倍体化生物标志物。该测定法鉴定出 27 种化学物质均具有遗传毒性，其中 25 种表现出非整倍体特征，1 种具有非整倍体和断裂剂特征，1 种具有断裂剂特征。随后，采用新描述的后续测定法来研究非整倍体试剂的分子靶标。对于这些实验，将 TK6 细胞暴露于 26 种化学物质中的每一种中，并加入 488 紫杉醇。4 小时后，裂解细胞，用核酸染料染色游离核和有丝分裂染色体，并标记针对 p-H3 和 Ki-67 的荧光抗体。流式细胞分析显示，仅观察到与 488 紫杉醇相关荧光的变化与微管蛋白结合剂有关——微管蛋白稳定剂增加，微管蛋白稳定剂减少。具有已知 Aurora 激酶 B 抑制活性的有丝分裂激酶抑制剂是唯一显著降低 p-H3 阳性核与 Ki-67 阳性核比值的非整倍体。基于 488 紫杉醇荧光和 p-H3:Ki-67 比值的无监督层次聚类清楚地区分了具有这些不同分子机制的化合物。此外，基于人工神经网络的分类算法被发现可以有效地预测分子靶标，因为留一法交叉验证与先验预期一致，准确率为 25/26。这些结果令人鼓舞，因为它们表明，足够数量的训练集化学物质，结合基于 488 紫杉醇、p-H3 和 Ki-67 反应的机器学习算法，可以可靠地阐明最常见的非整倍体分子靶标。
