A Molecular Grammar for Guide RNAs (gRNAs) with Engineered Secondary Structures

Cas9 和 Cas12a 等 CRISPR 效应器因其能力而成为生物医学研究中的强大工具 在活细胞中引入靶向突变，因此，由于这种能力，它们具有重要的治疗作用 治疗遗传性疾病的潜力——尽管也存在重大的临床风险，即它们可能会引入“非遗传性疾病” 目标突变或非预期突变。虽然 CRISPR 效应子的力量在于它们的序列 识别和靶向与其RNA辅因子（它们的 ‘引导RNA’或gRNA)，它们的突变活性可以在不完美的核苷酸序列上触发 与它们的 gRNA 也具有不可预测的互补性。显然，失控突变的可能性增加了 对患者和临床医生来说都是危险信号，到目前为止，CRISPR 基因疗法一直高度关注 特殊的遗传情况。进一步改进 CRISPR 特异性是必要的，不仅是为了减轻 临床风险，同时也推动 CRISPR 的新应用——例如，单核苷酸变异 (SNV) 可以被可靠地区分，它将允许对常染色体显性遗传疾病进行等位基因特异性基因编辑， 我们经常需要区分“健康”和“健康”之间的微小序列变异 “疾病”等位基因，但目前的 CRISPR 技术无法始终如一地做到这一点。我们最近展示了 能够将 CRISPR 效应子特异性提高几个数量级的方法的可行性，以及 以这种方式，它可以协同应用于许多其他先前开发的技术 进一步提高特异性。通过向 gRNA (x-gRNA) 添加额外的核苷酸并设计扩展 序列与 gRNA 的 DNA 靶向片段形成“发夹”二级结构（发夹-gRNA 或 hp-gRNA）会破坏与脱靶相互作用的稳定性，我们可以生成显着限制的 x/hp-gRNA 源自四种的 CRISPR 效应子变体在保持脱靶活性的同时保持在靶突变活性 不同的生物体和一种工程衍生物。因此，长期目标是了解规则 在 x-gRNA 中设计扩展序列，这将为不同的 CRISPR 效应子带来超特异性 在任何 CRISPR 靶向位点。为了实现这一目标，在 R21 中，我们将进行详尽的筛选 针对不同临床相关位点的随机 x-gRNA 文库并确定哪些共同序列 这些 x-gRNA 的和/或二级结构特征可显着提高特异性。虽然 该提案的风险在于，对于所有 x/hp-gRNA 设计本身可能没有“通用设计规则” 和目标，这项工作仍然将为研究人员提供一个实用的（无设计）平台 为任何 CRISPR 效应子的任何感兴趣目标生成超特异性 x-gRNA。可能的奖励是 x-gRNA（与工程化 CRISPR 效应子）的协同使用有可能有效消除风险 CRISPR 应用中的意外突变，以及将我们的高通量方法与机器相结合 学习将使任何人都可以使用新的计算工具来从头生产超特异性或等位基因特异性的 x-gRNA。

CRISPR effectors like Cas9 and Cas12a have emerged as powerful tools in biomedical research for their ability to introduce targeted mutations in living cells and, consequentially, for this ability they hold significant therapeutic potential for treating genetic disorders—despite also carrying significant clinical risk that they may introduce ‘off- target’ or unintended mutations. While the power of CRISPR effectors lies in the fact that the sequences they recognize and target are complementary to a modular, ‘programmable’ segment of their RNA cofactors (their ‘guide RNAs’ or gRNAs), their mutational activity can be triggered at nucleotide sequences with imperfect complementarity to their gRNAs as well, unpredictably. Obviously, the possibility of uncontrolled mutation raises red flags for both patients and clinicians and so far, CRISPR gene therapies have been focused on highly specialized genetic situations. Further improvements to CRISPR specificity are necessary, not only to mitigate clinical risk, but also to drive new applications of CRISPR—for example, if single nucleotide variants (SNVs) could be reliably discriminated, it would allow for allele-specific gene editing of autosomal dominant disorders, where often we would need to discriminate between small sequence variations between the ‘healthy’ and ‘disease’ alleles but which current CRISPR technologies cannot consistently do. We recently demonstrated the feasibility of an approach that is capable of improving CRISPR effector specificity by orders-of-magnitude, and in such a way that it can be synergistically applied to many of the other previously-developed techniques to improve specificity further. By adding extra nucleotides to the gRNA (x-gRNA) and designing the extended sequence to form ‘hairpin’ secondary structures with the DNA-targeting segment of the gRNA (hairpin-gRNAs or hp-gRNAs) that destabilize interactions with off-targets, we could generate x/hp-gRNAs that significantly limited off-target activity while maintaining on-target mutational activity in CRISPR effector variants derived from four different organisms and one engineered derivative. The long-term goal is therefore to understand the rules for designing extended sequences in x-gRNAs that would result in ultra-specificity for divergent CRISPR effectors at any CRISPR-targetable site. To achieve that goal, in this R21 we will perform an exhaustive screen of randomized x-gRNA libraries targeting different clinically-relevant sites and identify what common sequence and/or secondary-structure features of those x-gRNAs drive significant increases in specificity. While the riskiness of this proposal is that there may not be “universal design rules”, per se, for all x/hp-gRNA designs and targets, this work will nevertheless provide a practical (design-free) platform for researchers to empirically generate ultra-specific x-gRNAs for any target of interest for any CRISPR effector. The likely reward is that synergistic use of x-gRNAs (with engineered CRISPR effectors) has the potential to effectively abrogate the risk of unintended mutation in CRISPR applications, and that combining our high-throughput approach with machine- learning would allow new computational tools for anyone to produce, de novo, ultra- or allele-specific x-gRNAs.