Cancer Classifiers Based on RNA Sensors in Living Cells

摘要 在活的哺乳动物细胞中存在对RNA传感器的迫切需要。随着单细胞RNA的出现 通过测序，任何细胞类型的转录组即使尚未可用，也很容易获得。相比之下， 仍然迫切需要一种通用方法来作用于这种转录组信息。如果我们能通过基因表达 根据其转录标记，在特定细胞类型中的任意效应蛋白，我们将转化 大量的基础研究和生物医学应用，如免疫学、神经科学和癌症 疗法此外，我们希望这种传感器是可编程的，并在转录后水平上运行。 受益于这种传感器的一个有希望的用例是癌症消融，使用一种方法 被称为“电路作为药物”，其中编码整个“电路”的遗传载体（比喻为一个集合， 被工程化以相互调节并实现特定功能的生物分子）在细胞内递送。 该电路将基于癌症的标志感测细胞状态（即，特定RNA的过度表达 或特定突变的存在），处理信号，并相应地递送特定的治疗有效载荷 在癌细胞中，直接杀死它们，同时训练免疫系统搜索和摧毁其他癌细胞。 以前的努力主要依赖于链置换，这是基于核酸的信号的成功策略 在细胞外进行处理。然而，它们的功能在活的哺乳动物细胞内仍然不足， 很可能是因为在链置换过程中形成的双链RNA（dsRNA）发出病毒感染的信号 并且在免疫途径中被哺乳动物蛋白积极参与。我们假设，因为它是 不可能回避无处不在的dsRNA相互作用蛋白，更明智的做法是接受它们。在这一提议中， 我们将利用能够识别和特异性编辑dsRNA的内源性人类酶来制造传感器， 可以被编程为对任意RNA转录物（“触发物”）做出反应。 首先，我们将在体外使用快速设计-构建-测试循环来优化传感器性能。我们将专注于 通过设计传感器配置及其序列选择来响应于触发而增加传感器输出， 我们将描述传感器如何影响细胞环境以及如何受细胞环境影响。第二，使 不同触发水平的定量区分和多个触发的集成，我们将设计 阈值设置修改和AND逻辑门。第三，利用转录后的独特性质， 我们将联合收割机与mRNA或溶瘤RNA病毒结合作为递送载体， 传统上是很难控制的最后，我们将验证性能和治疗效果， 传感器、门和RNA载体在癌细胞系中的潜力。 该项目的未来发展方向包括传感器、逻辑门、 和载体，在更现实的癌症模型中测试它们，包括小鼠模型和患者衍生的类器官， 并将这些工具应用到其他领域。

Abstract There is a critical need for RNA sensors in living mammalian cells. With the advent of single-cell RNA sequencing, the transcriptome of any cell type is readily obtainable if not already available. In contrast, we are still in urgent need for a universal method to act on such transcriptomic information. If we can genetically express arbitrary effector proteins in specific cell types according to their transcriptional markers, we would transform large swaths of basic research and biomedical applications, such as immunology, neuroscience, and cancer therapy. In addition, we would like such sensors to be programmable and operate at the post-transcription level. One promising use case that would benefit from such sensors is cancer ablation, using an approach dubbed “circuits as medicine”, where a genetic vector encoding an entire “circuit” (metaphor for a collection of biomolecules engineered to regulate each other and implement specific functions) is delivered intracellularly. The circuit will sense the cellular states based on hallmarks of cancer (i.e., the overexpression of specific RNAs or the presence of specific mutations), process the signals, and deliver specific therapeutic payloads accordingly in cancer cells, directly killing them while educating the immune system to search and destroy other cancer cells. Previous efforts largely relied on strand displacement, a successful strategy for nucleic acid-based signal processing outside cells. However, their functionality has remained inadequate inside living mammalian cells, most likely because the double-stranded RNA (dsRNA) formed during strand displacement signals viral infection and are actively engaged by mammalian proteins in the immune pathways. We hypothesize that, because it is impossible to evade the omnipresent dsRNA-interacting proteins, it is wiser to embrace them. In this proposal, we will leverage endogenous human enzymes that recognize and specifically edit dsRNA, to create sensors that can be programmed to respond to arbitrary RNA transcripts (“triggers”). First, we will use fast design-build-test cycles in vitro to optimize sensor performance. We will focus on increasing sensor output in response to triggers by engineering the sensor configuration and its sequence choice, and we will characterize how the sensor affects and is affected by the cellular context. Second, to enable the quantitative distinction of different trigger levels and the integration of multiple triggers, we will engineer threshold-setting modifications and AND logic gates. Third, leveraging the unique post-transcriptional nature of such sensors and gates, we will combine them with mRNA or an oncolytic RNA virus as delivery vectors, which has traditionally been difficult to control. Last by not least, we will validate the performance and the therapeutic potential of the sensors, gates, and the RNA vectors in cancer cell lines. The future directions of the proposed project include continual optimization of the sensors, logic gates, and vectors, testing them in more realistic cancer models including mouse models and patient-derived organoids, and applying the tools to other fields.