Developing cell-penetrating miniproteins as a new class of therapeutics

项目摘要 目前的蛋白质疗法有一个主要局限性：它们通常不能穿过细胞膜或 与胞质靶标相互作用。设计进入细胞胞浆的蛋白质疗法的能力将使 跨许多疾病领域的新治疗策略，包括癌症、自身免疫和神经学 疾病。治疗性“小蛋白”(长度为30-60个残基)有可能应对这一挑战，并且 最近已经确定了几种能够有效到达细胞质的微型蛋白质。然而，我们 缺乏对穿透细胞的微小蛋白的“设计规则”的一般了解，限制了 这类分子。此外，目前测量胞质输送的方法需要测量每一种 单独的蛋白质，这是缓慢和劳动密集型的。这使得无法测试大量的 发展一个强大的，定量的了解胞质递送的决定因素的微型蛋白质。 在这个探索性项目中，我们将开发一种新的方法来测量每种不同蛋白质在 大的混合池，使用靶向质谱学来单独鉴定每个小蛋白序列。在我们的 方法是，将包含数千个设计的微型蛋白质序列的可溶混合池与细胞孵育， 进入这些细胞的微小蛋白被胞浆靶标捕获。目标捕获的微蛋白是 然后从细胞内容物中提纯，并使用靶向蛋白质组学进行鉴定和定量。金额的多少 捕获的样本中的每一种蛋白质(相对于起始样本)将提供对 交付效率。这种方法是史无前例的，我们将使用不同的 阳性和阴性控制迷你蛋白、不同的文库大小和不同的细胞系。 有了这种方法，我们将使用我们以前开创的方法来计算设计数千个 具有有意不同的序列和结构特性的候选穿透细胞的微型蛋白质。我们会 然后使用我们新的高通量方法来量化这些新蛋白质的胞浆递送，创建 关于交付效率的前所未有的大规模数据。然后，我们将使用这些数据来构建机器学习 基于序列和结构属性预测微小蛋白质传递的模型。最后，我们将反复 迭代，根据我们的递送预测模型设计新的微型蛋白质文库，测试这些设计 使用我们的高通量实验方法，并进一步更新我们的模型。这种迭代设计-测试- 学习方法将建立对交付决定因素的强大的、预测性的理解。归根结底，这种能力 设计穿透细胞的微型蛋白将在细胞内解锁一系列新的治疗靶点。

Project Summary Current protein therapeutics have a major limitation: they generally cannot cross the cellular membrane or interact with cytosolic targets. The ability to design protein therapeutics that enter the cell cytosol would enable new therapeutic strategies across many disease areas, including cancer, autoimmunity, and neurological disease. Therapeutic “miniproteins” (30-60 residues in length) have the potential to address this challenge, and several miniproteins capable of efficiently reaching the cell cytosol have recently been identified. However, we lack a general understanding of the “design rules” for cell-penetrating miniproteins, limiting the development of this class of molecules. Furthermore, current approaches to measure cytosolic delivery require measuring each protein individually, which is slow and labor intensive. This makes it impossible to test large numbers of miniproteins to develop a robust, quantitative understanding of the determinants of cytosolic delivery. In this exploratory project, we will develop a new approach to measure delivery for each different protein in a large mixed pool, using targeted mass spectrometry to individually identify each miniprotein sequence. In our approach, a soluble mixed pool containing thousands of designed miniprotein sequences is incubated with cells, and miniproteins that enter those cells are captured by a cytosolic target. Miniproteins captured by the target are then purified out of the cellular contents and identified and quantified using targeted proteomics. The amount of each protein in the captured sample (relative to the starting sample) will provide a quantitative measure of delivery efficiency. This approach is unprecedented, and we will test and optimize this approach using different positive and negative control miniproteins, different library sizes, and different cell lines. With this method in hand, we will use approaches we previously pioneered to computationally design thousands of candidate cell-penetrating miniproteins with intentionally diverse sequence and structural properties. We will then quantify cytosolic delivery for these new proteins using our new high-throughput approach, creating unprecedented large-scale data on delivery efficiency. We will then use these data to build machine learning models that predict miniprotein delivery based on sequence and structural properties. Finally, we will repeatedly iterate, designing new miniprotein libraries based on our predictive models of delivery, testing these designs using our high-throughput experimental approach, and further updating our models. This iterative design-test- learn approach will build a robust, predictive understanding of the determinants of delivery. Ultimately, the ability to design cell-penetrating miniproteins will unlock a wide range of new therapeutic targets inside the cell.