Mechanisms of non-classical multidrug resistance in cancer

三种主要的方法被用来定义癌症中的非经典多药耐药。首先，我们分离了KB细胞（HeLa的一个亚克隆），对顺铂（CP-r）水平升高具有耐药性，并证明对亚砷酸盐和镉、甲氨蝶呤和核苷类似物具有多药耐药性。这种交叉抗性模式是由于每一种药物的摄取减少，因为它们的受体已经从细胞表面重新定位到细胞的细胞质中。表面转运蛋白的这种重新定位似乎是由于顺铂耐药细胞中细胞骨架的改变影响了内吞循环区，从而改变了这些转运蛋白的再循环。最近对来自顺铂耐药人类癌症的顺铂耐药细胞系的研究表明，顺铂积累减少并不是耐药肿瘤的必然特征。我们已经证实，在顺铂耐药的KB细胞中，特异性microrna (miRNAs)（如miRNA-181）的变化一直存在，并且它们对耐药的贡献已通过miRNA模拟物和抑制剂的表达得到证实。此外，对逆转KB-CP-r细胞顺铂耐药的mirna的高通量分析发现，WEE1和CHK1是顺铂耐药的重要因素。miRNA155和miR-15家族成员是通过WEE1和CHK1表达影响顺铂耐药的mirna。我们对检查点激酶及其在耐药性中的作用的兴趣使我们开发了蛋白磷酸酶2A （PP2A）抑制剂LB100，目前处于乳腺癌的I期临床试验中。由于PP2A控制许多dna损伤反应（DDR）基因的磷酸化状态，我们假设LB100会使卵巢癌细胞对顺铂敏感。我们证明了LB100抑制PP2A使细胞（OVCAR8和SKOV3）对顺铂敏感，LB100诱导dna损伤反应通路中Chk1和其他基因的过度磷酸化，阻止顺铂诱导的G2阻滞，迫使细胞进入有丝分裂，导致细胞凋亡。我们发现腹腔注射skov3 -荧光素酶细胞的小鼠与对照组相比，LB100对顺铂（3mg /kg）敏感。我们最近在暴露于顺铂的细胞中完成了RNAi筛选，以确定与顺铂敏感性相关的基因。如果暴露于亚毒性顺铂的细胞在特定基因缺失时发生细胞死亡，可以假设抑制该基因靶点可能被证明是铂化疗的有用辅助。当DNA损伤修复基因（包括磷酸蛋白磷酸酶）被沉默时，观察到最强的致敏效应，其中一些正在研究它们在顺铂耐受性中的作用。在这种筛选背景下，我们发现需要确定一种适合溶解顺铂进行筛选的溶剂。我们最近表明，DMSO灭活了所有临床和实验铂复合物测试的生物活性。此外，对顺铂文献的回顾显示，约有三分之一的研究论文使用溶解在DMSO中的顺铂，这使这些论文的数据和结论受到质疑。这对相当一部分文献的可靠性具有重要意义，并为在研究中适当使用铂类药物指明了道路。第二种方法是评估黑色素瘤细胞导致多药耐药的独特特征。黑色素瘤细胞的一个明显特征是黑色素小体，这是一种溶酶体衍生的细胞器，色素的形成发生在其中。我们已经证明，顺铂被隔离在这个细胞器中，与黑色素形成的程度无关，并与黑色素小体一起挤进培养基，减少了这种抗癌药物的核积累。ABCB5是一种与ABCB1同源的转运蛋白，在黑色素细胞和黑色素瘤等色素细胞中高水平表达，目前正在进行研究，以确定ABCB5是否有助于黑色素瘤中所见的黑素体隔离。大约15%的人类黑色素瘤携带ABCB5突变，这表明ABCB5缺陷与黑色素瘤的进展有关。在另一种方法中，我们开发了一种Taqman低密度阵列（TLDA）微流控芯片来检测380种不同的假定耐药基因的mRNA表达，并证明它是一种敏感、准确、可重复和可靠的方法来测量肿瘤样本中的mRNA水平。我们实验室以前的工作表明，耐药基因水平的mRNA测量可以初步预测耐药机制的功能表达。这种耐药芯片已被应用于人类癌症的分析。该分析的一个结果是，现有的癌细胞系并不能模拟实际人类癌症的表达模式，为TLDA分析选择了380个假定的耐药基因，而在3D培养中培养细胞的简单权宜之计并不能纠正这个问题。这表明需要更好的体外癌细胞模型来研究多药耐药。另一个结论是，我们研究的11个MDR基因的特征可以预测非积液性卵巢癌的不良反应，18个MDR基因的不同亚群可以预测积液性卵巢癌的不良反应。对于肝癌，两种不同的MDR基因表达特征与预后较差和预后较好的肝癌相关。特异性药物组蛋白去乙酰化酶抑制剂可将预后不良的肝癌基因表达模式转化为预后改善的肝癌。对于急性髓性白血病（AML），化疗缓解后的疾病复发与多种不同的MDR基因表达模式相关，提示AML获得性耐药可能是多因素的。这些结果的验证表明耐多药在临床癌症中是复杂和多因素的，将需要开发可靠的体外培养模型，并且正在使用基于网络理论的数学模型来解释这些数据。为了实现这一目标，我们开发了一种生物反应器，可以模拟毛细管输送（通过硅水凝胶）氧气到3D悬浮液中生长的细胞。我们已经证明了生理氧梯度和癌细胞的改变生长更接近体内表型。该生物反应器可以扩大规模，用于原发癌细胞或培养癌细胞的多种培养物的生长，以确定生长条件是否在影响耐药模式中起主要作用。由于体内耐药机制的明显复杂性，我们重新检查了基于更均匀系统预测耐药发展的现有数学模型。与马里兰大学的Doron Levy合作，我们制定了一个新的数学模型，该模型考虑了肿瘤的异质性和与体内系统相关的其他特征。这个模型可以预测化疗应该如何靶向以优化杀死耐药癌细胞。此外，我们在数学上表明，培养细胞中药物杀伤曲线的形状是由局部细胞相互作用和细胞密度变化引起的。

Three major approaches have been taken to define non-classical multidrug resistance in cancer. In the first, we isolated KB cells (a subclone of HeLa) resistant to increasing levels of cisplatin (CP-r) and demonstrated multidrug resistance to arsenite and cadmium, to methotrexate, and to nucleoside analogs. This cross-resistance pattern is due to reduced uptake of each of these agents because their receptors have been relocalized from the cell surface into the cytoplasm of the cell. This relocalization of surface transporters appears to be due to altered recycling of these transporters due to alterations in the cytoskeleton that affect endocytic recycling compartments in cisplatin-resistant cells. Recent studies on cisplatin-resistant cell lines derived from cisplatin-resistant human cancers indicate that reduced cisplatin accumulation is not an obligatory characteristic of resistant tumors. We have demonstrated changes in specific microRNAs (miRNAs), such as miRNA-181, consistently seen in cisplatin-resistant KB cells, and their contribution to drug resistance has been demonstrated by expression of miRNA mimics and inhibitors. In addition, a high throughput analysis of miRNAs that reverse the cisplatin resistance of KB-CP-r cells has identified WEE1 and CHK1 as essential elements of resistance to cisplatin. miRNA155 and miR-15 family members are miRNAs whose expression affects cisplatin resistance through WEE1 and CHK1. Our interest in the checkpoint kinases and their role in resistance led us to the protein phosphatase 2A (PP2A) inhibitor LB100, currently in Phase I clinical trials for breast cancer. As PP2A controls the phosphorylation status of a number of DNA-damage response (DDR) genes, we hypothesized that LB100 would sensitize ovarian cancer cells to cisplatin. We demonstrated that inhibition of PP2A by LB100 sensitized cells (OVCAR8 and SKOV3) to cisplatin, and that LB100 induces hyperphosphorylation of Chk1 and other genes in the DNA-damage response pathway, preventing cisplatin-induced G2 arrest and forcing cells into mitosis, resulting in apoptosis. We showed that mice injected intraperitoneally with SKOV3-luciferase cells were sensitized to cisplatin (3 mg/kg) when treated with LB100 compared with control. We recently completed an RNAi screen in cells exposed to cisplatin, in order to identify genes associated with cisplatin sensitivity. If cells exposed to sub-toxic cisplatin undergo cell death when a particular gene is deleted, one can hypothesize that inhibition of this gene target might prove to be a useful adjuvant for platinum chemotherapy. The strongest sensitizing effects were observed when DNA damage repair genes (including a phosphoprotein phosphatase) were silenced, and several of these are now being investigated for their role in cisplatin tolerance. In this screening context, we found a need to identify a solvent appropriate for dissolving cisplatin for screening. We recently showed that DMSO inactivated the biological activity of all clinical and experimental platinum complexes tested. Furthermore, a review of the cisplatin literature revealed that about a third of all research papers have used cisplatin dissolved in DMSO, calling into question the data and conclusions of those papers. This has important implications for the reliability of a significant portion of the literature and points the way for appropriate use of platinum drugs in research. A second approach is to evaluate the unique features of melanoma cells that contribute to multidrug-resistance. One obvious feature of melanoma cells is the melanosome, a lysosome-derived organelle in which pigment formation takes place. We have shown that cisplatin is sequestered in this organelle, independent of extent of melanin formation, and extruded with melanosomes into the medium, reducing nuclear accumulation of this anti-cancer drug. Studies are underway to determine whether ABCB5, a transporter homologous to ABCB1, expressed at high levels in pigmented cells such as melanocytes and melanomas, contributes to the melanosomal sequestration seen in melanomas. Approximately 15% of human melanomas carry mutations in ABCB5, suggesting that defects in ABCB5 are linked to melanoma progression. In another approach, we have developed a Taqman Low Density Array (TLDA) microfluidic chip to detect mRNA expression of 380 different putative drug resistance genes and demonstrated that it is a sensitive, accurate, reproducible, and robust way to measure mRNA levels in tumor samples. Previous work from our laboratory indicates that mRNA measurements of levels of drug-resistance genes are, to a first approximation, predictive of functional expression of drug-resistance mechanisms. This drug-resistance chip has been applied to analysis of human cancers. One result from this analysis is that existing cancer cell lines do not mimic the expression patterns of actual human cancers for the 380 putative drug resistance genes chosen for the TLDA analysis and the simple expedient of growing cells in 3D culture does not correct this problem. This suggests the need for better in vitro cancer cell models to study multidrug resistance. Another conclusion is that a signature of eleven MDR genes we have studied predicts poor response in non-effusion ovarian cancer, and different subsets of 18 MDR genes predict poor response in ovarian cancer with effusions. For hepatoma, two different MDR gene expression signatures are associated with poor prognosis and better prognosis hepatoma. Specific drugs that are histone deacetylase inhibitors can convert the pattern of gene expression of poor prognosis hepatomas into cancers with improved prognosis. For acute myeloid leukemia (AML), recurrence of disease after remission induced by chemotherapy is associated with multiple different patterns of MDR gene expression, suggesting that for AML acquired resistance may be multifactorial. Validation of these results, indicating that MDR is complex and multifactorial in clinical cancers, will require the development of reliable in vitro culture models, and interpretation of these data using mathematical models based on network theory is proceeding. Towards this goal, we have developed a bioreactor that mimics capillary delivery (through silicon hydrogels) of oxygen to cells grown in 3D suspension. We have demonstrated physiological oxygen gradients and altered growth of cancer cells more closely approximating in vivo phenotypes. This bioreactor can be scaled up for growth of multiple cultures of primary cancer cells or cultured cancer cells to determine whether growth conditions play a primary role in affecting patterns of drug resistance. Because of the apparent complexity of drug resistance mechanisms in vivo, we have re-examined existing mathematical models that predict the development of drug resistance based on more homogeneous systems. In collaboration with Doron Levy (University of Maryland), we have formulated a new mathematical model that takes into account tumor heterogeneity and other features associated with in vivo systems. This model allows a prediction of how chemotherapy should be targeted to optimize killing of drug-resistant cancer cells. In addition, we have shown mathematically that the shape of drug killing curves in cultured cells results from local cell interactions and variations in cell density.