Designing cell function: assembly of synthetic gene circuits for cell biology applications
设计细胞功能：合成基因电路的组装用于细胞生物学应用

Since the earliest days of human civilization, microbial cells have been harnessed to produce fermented foods and beverages. Now, synthetic biology aims to engineer artificial cell functionalities that have never existed in nature. This aim has been underpinned by the advent in the late 20th century of recombinant DNA technologies, which have already revolutionized the pharmaceutical industry. For example, metabolic engineering has enabled the development of living cells as powerful production factories for high-value small molecules and proteins, for which chemical synthesis is too complex, too inefficient or too expensive¹. By ectopic (over)expression of a minimal set of transgenes coding for a protein or a specialized biosynthetic enzymatic cluster, host cells grown on inexpensive and renewable carbon sources can be programmed to produce pharmaceuticals^2,3,4,5, food additives⁶, feedstock and raw materials^7,8 or biopolymers⁹, achieving high bioprocess efficiency and environmental sustainability^1,10. To further optimize host cell productivity and increase profitability, we require gene circuits that can provide more precise control over spatiotemporal expression and activity of proteins than can be achieved with constitutive expression strategies. Importantly, very tight control of the activity of foreign genetic elements is also needed if these strategies are to be used in human therapy. For example, patient-specific T cells genetically modified to express cancer-targeting chimeric antigen receptors (CARs) are currently achieving unprecedented response rates in human clinical trials to treat blood cancers^11,12, but there is still a need to increase treatment safety, efficacy and reliability by introducing more sophisticated control layers for precise regulation of CAR T cell activity and performance^13,14. Thus, the focus of synthetic biology is now on designing synthetic gene circuits consisting of interconnected gene switches to programme time-dependent and context-dependent target gene activities in living cells.
自人类文明的早期以来，微生物细胞已被利用来生产发酵食品和饮料。如今，合成生物学旨在工程化前所未有的人工细胞功能。这一目标得益于 20 世纪末重组 DNA 技术的出现，这项技术已经彻底改变了制药行业。例如，代谢工程使得活细胞能够作为高价值小分子和蛋白质的强大生产工厂，而这些物质的化学合成过于复杂、效率低下或成本过高。通过异位（过量）表达一组最小的转基因，编码某种蛋白质或特定的生物合成酶簇，生长在廉价和可再生碳源上的宿主细胞可以被编程生产药物、食品添加剂、饲料和原材料或生物聚合物，从而实现高生物过程效率和环境可持续性。 为了进一步优化宿主细胞的生产力并提高盈利能力，我们需要能够提供比构成性表达策略更精确控制蛋白质时空表达和活性的基因电路。重要的是，如果这些策略要用于人类治疗，还需要对外源基因元件的活性进行非常严格的控制。例如，经过基因修饰以表达针对癌症的嵌合抗原受体（CAR）的患者特异性 T 细胞，目前在治疗血癌的人类临床试验中取得了前所未有的反应率，但仍然需要通过引入更复杂的控制层来提高治疗的安全性、有效性和可靠性，以精确调节 CAR T 细胞的活性和性能。因此，合成生物学的重点现在是设计由互联基因开关组成的合成基因电路，以在活细胞中编程时间依赖和情境依赖的靶基因活性。

In this Review, we first discuss in detail the regulatory elements that are available for precise, remote control of gene expression in human cells and the ways in which they can be rationally, systematically and effectively assembled to form autonomous gene circuits. Then, we highlight the direct impacts of synthetic gene circuits in current stem cell research, molecular diagnostics, drug discovery and agriculture. Finally, in considering the prospects for synthetic biology, we conclude that the scenario of using human cells as living drugs for autonomous disease treatment in personalized medicine is close to realization.
在本综述中，我们首先详细讨论了可用于精确、远程控制人类细胞基因表达的调控元件，以及如何合理、系统和有效地组装这些元件以形成自主基因电路。然后，我们强调了合成基因电路在当前干细胞研究、分子诊断、药物发现和农业中的直接影响。最后，在考虑合成生物学的前景时，我们得出结论，使用人类细胞作为自主疾病治疗的活药物在个性化医学中的场景即将实现。

Gene expression in living cells is in most cases tightly regulated at the level of individual genes. To programme cell functions, it is desirable to gain precise control of gene expression using synthetic gene circuits, which comprise networks of multiple interconnected gene switches regulating target gene activities in a time-dependent and context-dependent manner and whose activity can be predicted and externally controlled.
活细胞中的基因表达在大多数情况下是在单个基因水平上严格调控的。为了编程细胞功能，理想情况下希望通过合成基因电路获得对基因表达的精确控制，这些电路由多个相互连接的基因开关网络组成，以时间依赖和上下文依赖的方式调节目标基因的活性，并且其活性可以被预测和外部控制。

Natural and synthetic gene switches
自然和合成基因开关

From an engineering perspective, a gene switch can be regarded as any naturally evolved or rationally designed accession point that allows a scientist to ‘dial’ into a cell and decide whether gene expression at a particular level (DNA, RNA or protein) should be initiated, interrupted or terminated (Fig. 1A). In human cells, gene switches acting at the transcriptional level are operated by chimeric trans-regulators (Supplementary Box 1 (see the figure, part a)) consisting of a sequence-specific DNA-binding domain (DBD) that is fused to either a nonspecific epigenetic effector domain (also known as synthetic transcription factor; Fig. 1A, step 1) or a sequence-modifying nuclease domain (also known as designer nuclease; Fig. 1A, step 2; see also Supplementary Box 1 (see the figure, part b)). The DBD allows the trans-regulator to target a promoter region of either chromosomal genes (on the genome) or transgenes (on episomal vectors), where it triggers temporary activation, repression or silencing of target gene transcription (Fig. 1A, step 1) or permanent disruption or alteration of a nucleotide sequence (Fig. 1A, step 2). Genomic and transcriptional gene switches can also operate through CRISPR–Cas systems, where sequence targeting is mediated by guide RNAs (gRNAs) (see also Supplementary Box 1).
从工程的角度来看，基因开关可以被视为任何自然进化或合理设计的接入点，使科学家能够“调节”细胞，并决定在特定水平（DNA、RNA 或蛋白质）上是否应启动、中断或终止基因表达（图 1A）。在人体细胞中，作用于转录水平的基因开关由嵌合转调节因子操作（补充框 1（见图，部分 a）），其由一个序列特异性的 DNA 结合域（DBD）与一个非特异性表观遗传效应域（也称为合成转录因子；图 1A，步骤 1）或一个序列修饰核酸酶域（也称为设计核酸酶；图 1A，步骤 2；另见补充框 1（见图，部分 b））融合而成。DBD 使转调节因子能够靶向染色体基因（在基因组上）或转基因（在质粒载体上）的启动子区域，在那里它触发目标基因转录的暂时激活、抑制或沉默（图 1A，步骤 1）或核苷酸序列的永久破坏或改变（图 1A，步骤 2）。 基因组和转录基因开关也可以通过 CRISPR-Cas 系统运作，其中序列靶向是由引导 RNA（gRNA）介导的（另见补充框 1）。

A | Natural and synthetic accession points for controlling gene expression. Gene switches allow mammalian trans-regulators to modulate protein synthesis at all gene expression levels, including regulation of transcription initiation (on chromosome or episomal vector) (step 1), gene editing (step 2), regulation of RNA splicing (step 3), regulation of protein translation (step 4), gene knockdown by RNAi and antisense technology (step 5), and control of protein localization and regulation of protein interactions (step 6). A comprehensive list of various mammalian gene switches accompanied by detailed mechanistic descriptions is provided in Supplementary Box 1 and Supplementary Box 2. B | Multiplexed gene expression control. Ba | Mutually orthogonal trans-regulators operate at parallel genetic targets without cross-specificity and affinity (left panel). Similarly, orthogonal synthetic gene circuits operate in an autonomous and robust manner with no interference with endogenous processes of the host cell (right panel). Bb | Simultaneous control of multiple gene expression events can be achieved with orthogonal Cas9-based systems, by using either mutually orthogonal Cas9 nucleases (left panel) or a set of guide RNAs (gRNAs) for one Cas9 nuclease, referred to as scaffold RNAs (scRNAs; right panel). In scRNAs, gRNA is engineered to harbour orthogonal sets of protein-binding RNA aptamers that bind different regulatory domains consisting of an aptamer-binding protein (ABP) fused to an effector domain. CNOT7, CCR4–NOT transcription complex subunit 7; DBD, DNA-binding domain; dCas9, catalytically dead Cas9; eIF, eukaryotic translation initiation factor; ER, endoplasmic reticulum; FokI, Flavobacterium okeanokoites derived restriction endonuclease; Gal4, galactose-induced gene 4; IRES, internal ribosome entry site; miRNA, microRNA; NES, nuclear export signal; NLS, nuclear localization signal; ORF, open reading frame; Pol II, RNA polymerase II; RISC, RNA-induced silencing complex; SaCas9, Staphylococcus aureus-derived Cas9; shRNA, short hairpin RNA; SpCas9, Streptococcus pyogenes-derived Cas9; srRNA, small regulatory RNA; TALE, transcription activator-like effector; TetR, tetracycline-dependent repressor; ZFP, zinc-finger protein.
A | 控制基因表达的自然和合成接入点。基因开关允许哺乳动物的跨调节因子在所有基因表达水平上调节蛋白质合成，包括转录起始的调控（在染色体或外源载体上）（步骤 1）、基因编辑（步骤 2）、RNA 剪接的调控（步骤 3）、蛋白质翻译的调控（步骤 4）、通过 RNA 干扰和反义技术的基因敲低（步骤 5），以及蛋白质定位的控制和蛋白质相互作用的调控（步骤 6）。补充框 1 和补充框 2 中提供了各种哺乳动物基因开关的综合列表，并附有详细的机制描述。B | 多重基因表达控制。Ba | 互相正交的跨调节因子在平行的遗传靶点上操作，且没有交叉特异性和亲和力（左侧面板）。同样，正交合成基因电路以自主和稳健的方式运行，不干扰宿主细胞的内源性过程（右侧面板）。 Bb | 通过使用相互正交的 Cas9 核酸酶（左侧面板）或一组针对单个 Cas9 核酸酶的引导 RNA（gRNA），称为支架 RNA（scRNA；右侧面板），可以实现对多个基因表达事件的同时控制。在 scRNA 中，gRNA 被设计为包含正交的蛋白质结合 RNA 适配体，这些适配体结合不同的调控域，调控域由与效应域融合的适配体结合蛋白（ABP）组成。 CNOT7，CCR4–NOT 转录复合体亚基 7；DBD，DNA 结合域；dCas9，催化失活的 Cas9；eIF，真核翻译起始因子；ER，内质网；FokI，源自海洋黄杆菌的限制性内切酶；Gal4，半乳糖诱导基因 4；IRES，内部核糖体进入位点；miRNA，微小 RNA；NES，核输出信号；NLS，核定位信号；ORF，开放阅读框；Pol II，RNA 聚合酶 II；RISC，RNA 诱导沉默复合体；SaCas9，源自金黄色葡萄球菌的 Cas9；shRNA，短发夹 RNA；SpCas9，源自化脓性链球菌的 Cas9；srRNA，小型调控 RNA；TALE，转录激活因子样效应子；TetR，四环素依赖性抑制子；ZFP，锌指蛋白。

Full size image 完整图像

At the RNA level, protein-specific aptamer structures are central regulatory elements applicable for the design of site-specific gene switches. To regulate alternative splicing, for example, a single primary RNA transcript containing target-specific aptamers in key intronic regions can be used to produce different protein isoforms¹⁵. Depending on the presence of target proteins specifically binding to the aptamers (aptamer-binding proteins (ABPs); see also Supplementary Box 1 (see the figure, part c)) and masking the splice sites in the nucleus, different mRNA transcripts are generated (Fig. 1A, step 3). Similarly, translation-regulating gene switches use RNA aptamers as target-specific protein tethers. When incorporated upstream or downstream of a coding region in the mRNA, endogenous regulatory proteins^16,17 or synthetic translation factors (see also Supplementary Box 1) are recruited by the aptamer to bind a target mRNA, which either facilitates or blocks ribosomal binding (Fig. 1A, step 4). Furthermore, RNAi is a naturally evolved translational gene switch operating in the cytoplasm of all eukaryotic cell types^18,19. Intronically encoded small regulatory RNA (srRNA) molecules, such as short hairpin RNA (shRNA), siRNA or microRNA (miRNA), can be designed to bind complementary sequences of approximately 20 nucleotides in length on any target mRNA and to knock down gene expression by triggering RNA degradation²⁰ (Fig. 1A, step 5). Similarly, single-stranded antisense mRNA transcribed to produce an antiparallel configuration of a target gene can also inhibit translation through complementary base pairing²¹ (Fig. 1A, step 5). Lastly, protein-level switches are based on the control of localization and stability, post-translational modification and target binding. Manipulation of the chemical stability or cytosolic exposure of a nuclear localization signal²², nuclear export signal^23,24, prenylation motif²⁵, peroxisomal targeting sequence²⁶ or degron^27,28 controls the trafficking of a protein to the nucleus, cytoplasm, plasma membrane, peroxisome or proteasome, respectively (Fig. 1A, step 6). Alternatively, the phosphorylation status can regulate nuclear permeability^29,30 or the potency for transcriptional activation^31,32 of a target protein.
在 RNA 水平上，特定于蛋白质的适配体结构是设计特定位点基因开关的核心调控元件。例如，为了调节可变剪接，可以使用包含在关键内含子区域的靶特异性适配体的单一初级 RNA 转录本来产生不同的蛋白质同种型。根据特定结合适配体的靶蛋白（适配体结合蛋白（ABPs）；另见补充框 1（见图，部分 c））的存在以及在细胞核中掩盖剪接位点，生成不同的 mRNA 转录本（图 1A，步骤 3）。类似地，调节翻译的基因开关使用 RNA 适配体作为靶特异性蛋白质连接物。当适配体被纳入 mRNA 编码区域的上游或下游时，内源性调控蛋白或合成翻译因子（另见补充框 1）被适配体招募以结合靶 mRNA，这要么促进要么阻止核糖体结合（图 1A，步骤 4）。此外，RNA 干扰是一种自然进化的翻译基因开关，在所有真核细胞类型的细胞质中发挥作用。 内含子编码的小型调节 RNA（srRNA）分子，如短发夹 RNA（shRNA）、siRNA 或微 RNA（miRNA），可以被设计为与任何靶 mRNA 上约 20 个核苷酸长度的互补序列结合，并通过触发 RNA 降解来抑制基因表达 20（图 1A，步骤 5）。类似地，转录产生靶基因反平行配置的单链反义 mRNA 也可以通过互补碱基配对抑制翻译 21（图 1A，步骤 5）。最后，蛋白质水平的开关基于对定位和稳定性、翻译后修饰和靶结合的控制。通过操控核定位信号 22、核输出信号 23,24、异戊烯化基序 25、过氧化物酶体靶向序列 26 或降解信号 27,28 的化学稳定性或细胞质暴露，控制蛋白质分别向细胞核、细胞质、质膜、过氧化物酶体或蛋白酶体的运输（图 1A，步骤 6）。另外，磷酸化状态可以调节核通透性 29,30 或靶蛋白的转录激活能力 31,32。

Multiplexed control of gene expression
基因表达的多重控制

To simultaneously control multiple target genes in a single cell, which is a prerequisite for generating gene circuits of high complexity, a conventional strategy is to use mutually orthogonal trans-regulators that operate in parallel and on independent genetic events (Fig. 1Ba). Whereas chimeric transcription factors based on synthetic zinc-finger protein (ZFP) or transcriptional activator-like effector (TALE) scaffolds are inherently orthogonal to each other owing to their custom-designed and therefore predictable DBDs^33,34, naturally evolved transcription factors repurposed from the bacterial tetracycline-dependent repressor (TetR) protein family or the CRISPR–Cas system require context-dependent validation of functional orthologues^35,36 (Supplementary Box 1 (see the figure, part b)). Specifically, the combined use of two catalytically dead Cas9 (dCas9) orthologues derived from Staphylococcus aureus (SaCas9) and Streptococcus pyogenes (SpCas9) fused to different fluorescent proteins (GFP and mCherry) allowed simultaneous targeting and imaging of different genomic loci³⁷ (Fig. 1Bb, left panel). Alternatively, by incorporating orthogonal sets of protein-binding aptamers into non-conserved regions of gRNA structures³⁸, a unique platform known as scaffold RNA (scRNA) was created for multiplexed gene regulation by CRISPR–Cas9. In scRNAs, effector domains need no longer be physically bound to a protein domain, as required in conventional trans-regulator designs (Fig. 1A, step 1; Supplementary Box 1 (see the figure, part a)), but can be flexibly attached to different gRNAs through specific ABP–aptamer interactions³⁹ (Fig. 1Bb, right panel; Supplementary Box 1). Multiple site-specific scRNAs programmed for different effector tasks (activation, repression or visualization) can then be read by the same Cas9 orthologue disseminated throughout the nucleus, thus enabling simultaneous Cas9-dependent control of different scRNA-selected targets^39,40 (Fig. 1Bb, right panel).
为了在单个细胞中同时控制多个靶基因，这对于生成高复杂度的基因电路是一个前提，传统策略是使用相互正交的转调节因子，这些因子并行工作并作用于独立的遗传事件（图 1Ba）。基于合成锌指蛋白（ZFP）或转录激活因子样效应子（TALE）支架的嵌合转录因子由于其定制设计的、因此可预测的 DNA 结合域（DBD）而本质上是相互正交的 33,34，而从细菌四环素依赖抑制蛋白（TetR）家族或 CRISPR–Cas 系统重新利用的自然进化转录因子则需要上下文依赖的功能同源物验证 35,36（补充框 1（见图，部分 b））。具体而言，结合使用来自金黄色葡萄球菌（SaCas9）和化脓性链球菌（SpCas9）的两种催化失活的 Cas9（dCas9）同源物，分别与不同的荧光蛋白（GFP 和 mCherry）融合，允许同时靶向和成像不同的基因组位点 37（图 1Bb，左面板）。 通过将正交的蛋白结合适配体集合体整合到 gRNA 结构的非保守区域，创造了一种被称为支架 RNA（scRNA）的独特平台，用于通过 CRISPR–Cas9 进行多重基因调控。在 scRNA 中，效应域不再需要像传统的转调节设计那样与蛋白域物理结合，而可以通过特定的 ABP–适配体相互作用灵活地附加到不同的 gRNA 上。然后，针对不同效应任务（激活、抑制或可视化）编程的多个特定位点 scRNA 可以被分布在整个细胞核中的同一 Cas9 同源物读取，从而实现对不同 scRNA 选择目标的同时 Cas9 依赖控制。

Stimulus-dependent gene activity
刺激依赖的基因活性

In addition to multiplexed control of gene expression, gene switches responding to diverse molecular signals are instrumental in programming synchronized and interconnected gene or transgene activities^41,42 (see also Supplementary Box 2). Whereas gene switches controlled by chemical ligands (for example, small molecules, ions or proteins) are advantageous for achieving feedback control by endogenous metabolites or other molecular cues, light-regulated gene switches offer the highest spatiotemporal resolution for experimental activation and termination^43,44. An optimal inducible gene switch should permit essentially no expression of the target gene in the absence of the trigger signal while enabling maximal expression in the activated state^20,45.
除了对基因表达的多重控制外，响应多种分子信号的基因开关在编程同步和相互关联的基因或转基因活动中发挥着重要作用 41,42（另见补充框 2）。由化学配体（例如，小分子、离子或蛋白质）控制的基因开关在通过内源代谢物或其他分子线索实现反馈控制方面具有优势，而光调控的基因开关则提供了实验激活和终止的最高时空分辨率 43,44。一个理想的可诱导基因开关应在没有触发信号的情况下几乎不表达目标基因，同时在激活状态下实现最大表达 20,45。

To engineer stimulus-responsive gene expression in human cells, conventional strategies include systematic use of ligand-responsive prokaryotic transcription factors as the DBD of eukaryotic trans-regulators^46,47,48 (Fig. 2A,B), engineering of transcriptional gene switches responding to stimuli delivered to cell surface receptors via naturally occurring (Fig. 2Ca) or synthetic signalling pathways (Fig. 2Cb)^49,50 and the design of RNA-level gene switches^51,52 (Fig. 2D). The tightest transgene switch reported to date was created with the combination of two prokaryote-derived trans-repressors (TetR and lactose operon repressor (LacI)) together with RNAi (Fig. 2B). In this system, transcription of TetR and the target gene was repressed by LacI, allowing a TetR-repressible shRNA to be expressed to abolish basal expression of the target gene. Addition of the LacI-specific ligand isopropyl β-d-1-thiogalactopyranoside (IPTG) could relieve the expression of TetR and the gene of interest by inhibiting shRNA-mediated knockdown of the target gene. The extreme tightness of this LacI–TetR–RNAi (LTRi) switch was demonstrated by the survival of cells in which the LTRi-regulated transgene encoded diphtheria toxin-α — a protein so toxic that expression of a single molecule would have killed the host cell⁵³.
为了在人体细胞中工程化刺激响应的基因表达，传统策略包括系统性地使用配体响应的原核转录因子作为真核转录调节因子的 DNA 结合域（DBD）（图 2A,B），工程化响应于通过自然发生的（图 2Ca）或合成信号通路（图 2Cb）传递到细胞表面受体的刺激的转录基因开关，以及设计 RNA 水平的基因开关（图 2D）。迄今为止报告的最紧密的转基因开关是通过结合两种源自原核生物的转抑制因子（TetR 和乳糖操纵子抑制因子（LacI））与 RNA 干扰（RNAi）创建的（图 2B）。在该系统中，TetR 和目标基因的转录被 LacI 抑制，从而允许表达 TetR 可抑制的 shRNA，以消除目标基因的基础表达。添加 LacI 特异性配体异丙基β-D-1-硫代半乳糖苷（IPTG）可以通过抑制 shRNA 介导的目标基因敲低来解除 TetR 和感兴趣基因的表达。 这种 LacI–TetR–RNAi (LTRi)开关的极端紧密性通过生存细胞的实验得以证明，这些细胞中 LTRi 调控的转基因编码白喉毒素-α——一种毒性极强的蛋白质，表达单个分子就足以杀死宿主细胞 53。

A | Gene switches based on prokaryotic transcription factors. The use of prokaryotic transcription factors (pTFs; for example, tetracycline-dependent repressor (TetR), lactose operon repressor (LacI) or transcriptional activator protein LuxR) in mammalian cells as the DNA-binding domain (DBD) of a synthetic transcription factor enables trigger-inducible regulation of synthetic target gene promoters that contain pTF-specific binding sites. The presence of a pTF-specific trigger compound renders the DBD either capable or incapable of binding DNA, and depending on the choice of the effector domain, this conformational change results in activation (ON-switch) or termination (OFF-switch) of transcription. B | Principle of the LacI–TetR–RNAi (LTRi) switch. In the repressed state (OFF), transgene expression is abolished through LacI-dependent transcription repression and concomitant RNAi by a short hairpin RNA (shRNA) that targets the 3′ untranslated region (UTR) of the transgene mRNA. Addition of isopropyl-β-d-1-thiogalactopyranoside (IPTG) derepresses LacI-specific promoters, resulting in transgene transcription and concomitant TetR-mediated repression of shRNA production. C | Control of gene expression by cell surface receptors. Gene switches responsive to surface receptor-mediated signalling can be host cell-specific (receptor activation is sufficient to trigger pathway-specific transcription), semi-synthetic (transgene expression from synthetic promoters engineered to contain endogenous response elements) (part Ca) or orthogonal (engineered receptors activate synthetic promoters through artificial signal transduction) (part Cb). For orthogonal approaches, engineered receptors consist of: (1) an extracellular ligand-binding domain (for example, single-chain antibodies, G protein-coupled receptors (GPCRs) or receptor tyrosine kinases (RTKs)), (2) a transmembrane domain containing target sites for proteolytic cleavage (for example, by Notch signalling-specific γ-secretase or by tobacco etch virus (TEV) protease) and (3) a synthetic transcription factor (based on pTFs or Cas9) that is released for nuclear translocation upon receptor-activated proteolysis. D | RNA-based gene switches. Da | In Escherichia coli, a ‘toehold’ switch was engineered on the basis of RNA displacement. A hairpin structure in the 5′ UTR of mRNAs masks the ribosome-binding site (RBS) and prevents translation initiation, which is counteracted by binding of a single-stranded trigger RNA and conformational change of the mRNA. Db | Because RNA aptamers undergo a considerable conformational change upon ligand binding (top panel), they can be integrated into different RNA regions to create ligand-responsive gene switches. For example, gRNAs can be designed to contain an aptamer so that the DNA-targeting guide region is unmasked upon ligand binding (bottom panel). ABP, aptamer-binding protein; CAR, chimeric antigen receptor; gRNA, guide RNA; nt, nucleotide; rTetR, reversed TetR.
A | 基于原核转录因子的基因开关。将原核转录因子（pTFs；例如，四环素依赖性抑制子（TetR）、乳糖操纵子抑制子（LacI）或转录激活蛋白 LuxR）用作合成转录因子的 DNA 结合域（DBD）在哺乳动物细胞中的应用，使得能够对含有 pTF 特异性结合位点的合成靶基因启动子进行触发诱导调控。pTF 特异性触发化合物的存在使得 DBD 能够或不能结合 DNA，且根据效应域的选择，这种构象变化导致转录的激活（开关开启）或终止（开关关闭）。B | LacI–TetR–RNAi（LTRi）开关的原理。在抑制状态（关闭）下，通过 LacI 依赖的转录抑制和同时由靶向转基因 mRNA 3′非翻译区（UTR）的短发夹 RNA（shRNA）介导的 RNAi，转基因表达被消除。添加异丙基-β-D-1-硫代半乳糖苷（IPTG）解除 LacI 特异性启动子的抑制，导致转基因转录和同时 TetR 介导的 shRNA 生产抑制。 C | 通过细胞表面受体控制基因表达。对表面受体介导信号反应的基因开关可以是宿主细胞特异性的（受体激活足以触发通路特异性转录）、半合成的（来自合成启动子的转基因表达，这些启动子经过工程设计以包含内源性响应元件）（部分 Ca）或正交的（工程化受体通过人工信号转导激活合成启动子）（部分 Cb）。对于正交方法，工程化受体由以下部分组成：（1）一个细胞外配体结合域（例如，单链抗体、G 蛋白偶联受体（GPCRs）或受体酪氨酸激酶（RTKs）），（2）一个跨膜域，包含用于蛋白水解的靶位点（例如，由 Notch 信号特异性γ-分泌酶或烟草蚀刻病毒（TEV）蛋白酶进行水解）和（3）一个合成转录因子（基于 pTFs 或 Cas9），在受体激活的蛋白水解后释放以进行核转位。D | 基于 RNA 的基因开关。Da | 在大肠杆菌中，基于 RNA 置换工程设计了一个“脚趾”开关。 mRNA 的 5′ UTR 中的发夹结构掩盖了核糖体结合位点（RBS），并阻止翻译起始，这一过程通过单链触发 RNA 的结合和 mRNA 的构象变化得以逆转。由于 RNA 适配体在配体结合时会经历显著的构象变化（上面面板），它们可以被整合到不同的 RNA 区域，以创建配体响应的基因开关。例如，可以设计 gRNA 包含一个适配体，以便在配体结合时解开 DNA 靶向引导区域（下面面板）。ABP，适配体结合蛋白；CAR，嵌合抗原受体；gRNA，引导 RNA；nt，核苷酸；rTetR，反向 TetR。

Full size image 完整图像

Whereas gene switches based on prokaryotic transcription factors are limited to cell-permeable trigger compounds for activation, cell surface receptors such as G protein-coupled receptors (GPCRs), receptor tyrosine kinases (RTKs), cytokine receptors and CARs respond to extracellular signals and require minimal exposure times to a ligand for activation. Various cell type-specific responses, such as endothelial cell motility^54,55 or T cell activation^13,56, are transduced by these receptors. Thus, forced expression (for example, of CARs) or forced dimerization of the cell surface receptor (for example, of RTKs) in cell types that harbour the receptor-specific signalling cascade is reminiscent of using a ‘master key’ to kick off the complex but self-contained response^12,54 (Fig. 2Ca; Pathway-specific transcription). Because many receptor-mediated signalling cascades activate specific endogenous promoters, engineering of synthetic promoters containing the same response elements as the endogenous promoter creates a ‘virtual clone’ on an episomal vector that captures pathway-specific endogenous signalling events with user-defined transgene readouts (Fig. 2Ca; Semi-synthetic approach). To create receptor-activated gene switches based on artificial signal transduction that do not interfere with endogenous signalling, various proteolysis-dependent design strategies inspired by the nuclear translocation of the Notch receptor intracellular domain (NICD) have been developed (Fig. 2Cb). In this paradigm, NICD is replaced by a synthetic transcription factor, which translocates to the nucleus upon receptor activation to regulate transgene expression^57,58,59,60. Replacement of the extracellular Notch domain by a single-chain antibody allows recognition and response to cell surface antigens while retaining the native cis-inhibition feature of Notch signalling. Incorporation of different synthetic transcription factors into the framework of such synthetic Notch receptors enables the sensing of antigen-specific cell contacts⁵⁷. To also allow soluble ligands to trigger proteolytic cleavage and activate transgene expression, GPCRs and RTKs were fused to synthetic transcription factors through a transmembrane linker that contained synthetic cleavage sites for the tobacco etch virus (TEV) protease^58,59,60. Notably, the use of Cas9-based transcription factors is advantageous for simultaneous targeting of multiple genes when multiple gRNAs are co-expressed in the cell^58,59,60.
基于原核转录因子的基因开关仅限于可穿透细胞的触发化合物以激活，而细胞表面受体如 G 蛋白偶联受体（GPCRs）、受体酪氨酸激酶（RTKs）、细胞因子受体和 CARs 则对细胞外信号作出反应，并且对配体的暴露时间要求最低。各种细胞类型特异性的反应，如内皮细胞运动 54,55 或 T 细胞激活 13,56，都是通过这些受体转导的。因此，在具有受体特异性信号级联的细胞类型中，强制表达（例如 CARs）或强制二聚化细胞表面受体（例如 RTKs）类似于使用“万能钥匙”启动复杂但自我包含的反应 12,54（图 2Ca；通路特异性转录）。由于许多受体介导的信号级联激活特定的内源性启动子，工程化合成启动子包含与内源性启动子相同的反应元件，在外源性载体上创建一个“虚拟克隆”，捕捉通路特异性的内源性信号事件，并具有用户定义的转基因读出（图 2Ca；半合成方法）。 为了创建基于人工信号转导的受体激活基因开关，这些开关不干扰内源性信号传导，已经开发了多种依赖于蛋白水解的设计策略，这些策略受到 Notch 受体细胞内结构域（NICD）核转位的启发（图 2Cb）。在这一范式中，NICD 被合成转录因子替代，该转录因子在受体激活后转位到细胞核中以调节转基因表达 57,58,59,60。通过单链抗体替代细胞外 Notch 结构域，可以识别和响应细胞表面抗原，同时保留 Notch 信号传导的内源性顺式抑制特性。将不同的合成转录因子纳入这种合成 Notch 受体的框架中，使得能够感知特定抗原的细胞接触 57。为了使可溶性配体也能触发蛋白水解并激活转基因表达，GPCR 和 RTK 通过包含烟草花叶病毒（TEV）蛋白酶合成切割位点的跨膜连接子与合成转录因子融合 58,59,60。 值得注意的是，基于 Cas9 的转录因子的使用在多个 gRNA 在细胞中共同表达时，有利于同时靶向多个基因 58,59,60。

Inducible regulation of RNA-level gene activities commonly relies on the use of riboswitches, whereby a trigger (an RNA molecule or an aptamer-specific ligand) induces or stabilizes a conformational change in the target RNA^51,52 (Fig. 2D). An elegant riboswitch principle was designed in Escherichia coli, where the ribosome-binding site (RBS) was masked with an engineered cis-acting hairpin structure that rendered the target mRNA inaccessible for ribosomal binding and hence translation⁶¹. This riboswitch enabled the detection of single-stranded trigger RNA, which bound to a complementary portion of the masking hairpin and induced a conformational change in the mRNA molecule, resulting in RBS unmasking and initiation of translation (Fig. 2Da). In human cells, a similar system based on aptamers was designed to engineer ligand-responsive gRNAs, also known as signal conductors⁴⁰. Because aptamers undergo a considerable conformational change upon ligand binding (Fig. 2Db, top panel), they can be used to control the conformation and hence activity of the RNA molecule they are attached to. Accordingly, gRNAs harbouring aptamers in their 3ʹ hairpin regions were designed to adopt a secondary structure in which the DNA-targeting guide region was sealed by a synthetic complementary region engineered into the same gRNA molecule (Fig. 2Db, bottom panel). The presence of external ligands binding the aptamer triggered unmasking of the guide region and restored gRNA-dependent CRISPR–Cas activity, such as transcription control (Fig. 1A, step 1) or gene editing (Fig. 1A, step 2). Similarly, incorporation of aptamers into regions of srRNA precursors⁶² or self-cleaving aptazymes⁶³ enables regulation of target gene translation through ligand-dependent control over RNAi (Supplementary Box 2 (see the figure, part c)) or mRNA stability (Supplementary Box 2 (see the figure, part d)), respectively.
可诱导的 RNA 水平基因活性调控通常依赖于核糖开关的使用，其中触发物（RNA 分子或特定配体的适体）诱导或稳定目标 RNA 的构象变化 51,52（图 2D）。在大肠杆菌中设计了一种优雅的核糖开关原理，其中核糖体结合位点（RBS）被一种工程化的顺式作用发夹结构掩盖，使目标 mRNA 无法与核糖体结合，从而阻止翻译 61。该核糖开关能够检测单链触发 RNA，该 RNA 与掩盖发夹的互补部分结合，并诱导 mRNA 分子的构象变化，导致 RBS 的去掩盖和翻译的启动（图 2Da）。在人体细胞中，基于适体设计了类似的系统，以工程化配体响应的 gRNA，也称为信号导体 40。由于适体在配体结合时会经历显著的构象变化（图 2Db，顶部面板），因此它们可以用来控制其附着的 RNA 分子的构象和活性。 因此，设计了在其 3'发夹区域中含有适配体的 gRNA，使其采用一种二级结构，其中 DNA 靶向引导区域被工程化合成的互补区域封闭在同一 gRNA 分子中（图 2Db，底部面板）。外部配体与适配体结合的存在触发了引导区域的揭示，并恢复了 gRNA 依赖的 CRISPR-Cas 活性，例如转录控制（图 1A，步骤 1）或基因编辑（图 1A，步骤 2）。类似地，将适配体纳入 srRNA 前体 62 或自切割 aptazyme63 的区域，使得通过配体依赖的 RNAi 控制（补充框 2（见图，部分 c））或 mRNA 稳定性（补充框 2（见图，部分 d））来调节靶基因翻译成为可能。

Trigger-mediated control of gene activity at the protein level is achieved by allosteric proteins^64,65 and ligand-induced protein dimerization (Supplementary Box 1 (see the figure, part d), see also next subsection).
通过变构蛋白和配体诱导的蛋白质二聚化，触发介导的基因活性控制在蛋白质水平上得以实现（补充框 1（见图，部分 d），另见下一个小节）。

Spatiotemporal control elements
时空控制元件

Whereas natural localization signals encoded in a protein structure are spontaneously recognized by the cellular trafficking machinery to guide the protein towards different target destinations following translation (Fig. 1A, step 6), protein dimerization — controlled by external signals, for example, chemical ligands or light (Supplementary Box 1 (see the figure, part d)) — can also be employed to create synthetic tethering motifs (Fig. 3A). For example, a dimerization partner could be fused to an anchor protein residing within the plasma membrane^25,66 or at specific organelles (for example, endoplasmic reticulum²⁵ or mitochondria⁶⁷) to attract target proteins to these sites, thereby regulating intracellular cargo transport^25,28,67,68 and/or modulation of cell signalling^66,69 (Fig. 3A). Dimerization can also be used in split expression approaches to drive reconstitution of full, functional proteins from two individual fragments (Supplementary Box 2 (see the figure, part b)), or in two-hybrid systems to drive assembly of trans-regulators by combining a trafficking domain (such as a DBD) and an effector domain that are each attached to a dimer-forming protein (Supplementary Box 2 (see the figure, part a)). Spatiotemporal control of protein activity has also been achieved by trigger-induced protein uncaging (Fig. 3B; Supplementary Box 2 (see the figure, part g)). Most commonly used in this context is the protein LOV2 derived from Avena sativa phototropin 1, which has been repurposed for various cell-engineering applications. LOV2 uncaging is driven by blue light. During this process, a carboxy-terminal α-helix (Jα) is dissociated from the protein core and exposed to the cytoplasm⁷⁰. By fusing localization signals to Jα, synthetic systems have been designed for conditional control of protein trafficking^22,71 and degradation^43,64,72 (Fig. 3B). Similarly, fusion of a calcium-sequestering calmodulin-M13 domain to Jα afforded a synthetic optogenetic device controlling blue light-triggered calcium release⁷³.
自然定位信号编码在蛋白质结构中，能够被细胞运输机制自发识别，以指导蛋白质在翻译后朝向不同的目标目的地（图 1A，步骤 6）。蛋白质二聚化——由外部信号控制，例如化学配体或光（补充框 1（见图，部分 d））——也可以用于创建合成锚定基序（图 3A）。例如，二聚化伙伴可以与位于质膜内的锚定蛋白 25,66 或特定细胞器（例如内质网 25 或线粒体 67）融合，以吸引目标蛋白质到这些位置，从而调节细胞内货物运输 25,28,67,68 和/或细胞信号调节 66,69（图 3A）。 二聚化还可以用于分裂表达方法，以驱动从两个单独片段重构完整的功能性蛋白质（补充框 2（见图，部分 b）），或在双杂交系统中，通过结合一个转运域（如 DBD）和一个效应域，这两个域各自附着在一个二聚体形成蛋白上，来驱动转调节因子的组装（补充框 2（见图，部分 a））。通过触发诱导的蛋白质去保护，已经实现了蛋白质活性的时空控制（图 3B；补充框 2（见图，部分 g））。在这种情况下，最常用的蛋白质是来自燕麦光向性蛋白 1 的 LOV2，它已被重新用于各种细胞工程应用。LOV2 去保护是由蓝光驱动的。在此过程中，羧基末端的α-螺旋（Jα）从蛋白质核心解离并暴露于细胞质中。通过将定位信号融合到 Jα，设计了合成系统以实现对蛋白质转运和降解的条件控制（图 3B）。 类似地，将钙螯合的钙调蛋白-M13 结构域融合到 Jα上，提供了一种合成的光遗传学装置，用于控制蓝光触发的钙释放 73。

A | Dimerization-based localization. Chemical ligands or light can be used as triggers for driving protein–protein interactions (see also Supplementary Box 1 (see the figure, part d)). This enables a target protein harbouring a protein interaction domain to be targeted to a desired location that specifically expresses its interaction partner in a temporally controlled manner. Inducible dimerization can also be used for trigger-regulated reconstitution of transcription factors, nucleases, proteases or reporters (Supplementary Box 2). B | Protein uncaging. Blue light triggers ‘uncaging’ of the Avena sativa phototropin 1-derived LOV2 protein, resulting in dissociation and cytoplasmic exposure of a carboxy-terminal α-helix (Jα). By fusing localization signals to Jα, synthetic systems have been designed for conditional control of protein trafficking and degradation, whereby the target protein is fused to LOV2. C | Oligomerization-based protein sequestration. A target protein (prey; for example, glycogen synthase kinase-3β (GSK3β)) known to interact with a protein partner with high affinity (bait; for example, carboxy terminus of low-density lipoprotein receptor-related protein 6 (LRP6C)) can be sequestered into a synthetic protein cluster formed by fusion of the bait protein to an oligomer-forming protein (for example, cryptochrome 2 (CRY2)). In this example, selective sequestration of GSK3β away from an endogenous destruction complex by blue light-triggered oligomerization of CRY2–LRP6C could inhibit the degradation complex of cytosolic β-catenin. D | Fine-tuning temporal dynamics of gene expression. Da | A conditionally stabilized degron or destabilization domain fused to a protein that specifically binds the target protein with high affinity can confer its conditional instability on the target protein, resulting in degradation. Db | Target proteins engineered to contain tobacco etch virus (TEV) protease cleavage sites can be degraded by proteolysis upon ectopic expression of the TEV protease. Dc | In a recombinase-mediated time-delay strategy, gene expression relying on Cre-mediated reconstitution of a functional transcription unit is temporally delayed by time-consuming Cre expression. TCF/LEF, T cell factor/lymphoid enhancer factor.
A | 基于二聚化的定位。化学配体或光可以作为触发剂，驱动蛋白质-蛋白质相互作用（另见补充框 1（见图，部分 d））。这使得携带蛋白质相互作用域的目标蛋白能够被定向到特定表达其相互作用伙伴的期望位置，以时间可控的方式进行定位。可诱导的二聚化也可以用于触发调控的转录因子、核酸酶、蛋白酶或报告基因的重组（补充框 2）。B | 蛋白质去罩。蓝光触发来自燕麦光敏蛋白 1 的 LOV2 蛋白的“去罩”，导致一个羧基末端α-螺旋（Jα）的解离和细胞质暴露。通过将定位信号融合到 Jα，设计了合成系统以条件性控制蛋白质的运输和降解，其中目标蛋白与 LOV2 融合。C | 基于聚合的蛋白质隔离。 一个已知与高亲和力蛋白伴侣（诱饵；例如，低密度脂蛋白受体相关蛋白 6 的羧基末端（LRP6C））相互作用的目标蛋白（猎物；例如，糖原合成酶激酶-3β（GSK3β））可以被隔离到一个由诱饵蛋白与寡聚体形成蛋白（例如，隐花色素 2（CRY2））融合形成的合成蛋白簇中。在这个例子中，通过蓝光诱导的 CRY2–LRP6C 的寡聚化，选择性地将 GSK3β从内源性降解复合物中隔离，可以抑制细胞质β-连环蛋白的降解复合物。D | 微调基因表达的时间动态。Da | 一个条件稳定的降解信号或不稳定域与特异性高亲和力结合目标蛋白的蛋白融合，可以将其条件不稳定性赋予目标蛋白，导致降解。Db | 被工程化以包含烟草花叶病毒（TEV）蛋白酶切割位点的目标蛋白，可以在 TEV 蛋白酶的异位表达下通过蛋白水解被降解。 在重组酶介导的时间延迟策略中，依赖于 Cre 介导的功能转录单元重构的基因表达因耗时的 Cre 表达而被时间延迟。TCF/LEF，T 细胞因子/淋巴增强因子。

Full size image 完整图像

Many signalling events in human cells are governed by spatial control elements that form physical barriers or scaffold structures to restrict localization and action of target proteins⁷⁴. For example, nuclear translocation of β-catenin during WNT signalling is negatively regulated by a destruction complex that sequesters cytoplasmic β-catenin for proteasomal degradation⁷⁵. To compete with this endogenous destruction complex, a synthetic clustering system was created on the basis of a synthetic fusion protein of cryptochrome 2 (CRY2) and carboxy terminus of low-density lipoprotein receptor-related protein 6 (LRP6C), which acts as a bait for destruction complex component glycogen synthase kinase-3β (GSK3β)⁷⁶. Blue light triggers oligomerization of CRY2–LRP6C, which by interacting with GSK3β sequesters it from its interaction partners and thereby disrupts the destruction complex⁷⁷ (Fig. 3C).
许多信号事件在人体细胞中受到空间控制元素的调控，这些元素形成物理屏障或支架结构，以限制靶蛋白的定位和作用。举例来说，WNT 信号传导过程中β-连环蛋白的核转位受到一个破坏复合体的负向调控，该复合体将细胞质中的β-连环蛋白隔离以进行蛋白酶体降解。为了与这一内源性破坏复合体竞争，基于合成融合蛋白的合成聚集系统被创建，该融合蛋白由隐花色素 2（CRY2）和低密度脂蛋白受体相关蛋白 6 的羧基末端（LRP6C）组成，后者作为破坏复合体成分糖原合酶激酶-3β（GSK3β）的诱饵。蓝光触发 CRY2–LRP6C 的聚合，通过与 GSK3β相互作用，将其从相互作用伙伴中隔离，从而破坏破坏复合体。

Protein-level switches generally offer rapid responsiveness to external stimuli⁷⁸. For example, an effective method for trigger-inducible depletion of a target protein is based on forced recruitment of a binding partner that contains a conditional degron, which drives proteasomal degradation of the entire protein complex⁴³ (Fig. 3Da). Proteasome-independent control of protein degradation can be achieved by incorporating cleavage sites for a conditionally activated TEV protease into a target protein⁷⁹ (Fig. 3Db). To enable temporal synchronization of input and output levels between different gene switches, synthetic control elements enabling accelerated or delayed response times are required. For example, a rapid-acting signal transduction module based on protein-level phosphorylation processes can be used to connect two slow-acting transcriptional systems to accelerate the dynamics of the entire circuit³². Site-specific recombinases such as Cre are often used to programme time-delayed transcription initiation (Fig. 3Dc). By placing the coding region of a gene of interest in an antisense orientation flanked by antiparallel loxP sites^80,81 or by placing a termination (STOP) signal between parallel loxP sites in a promoter region of the gene^82,83, reconstitution of a functional transcription unit and gene expression depend on time-consuming Cre expression and Cre-mediated sequence inversion or excision, which considerably slows down target gene expression.
蛋白水平开关通常对外部刺激提供快速响应 78。例如，一种有效的触发诱导目标蛋白耗竭的方法是基于强制招募含有条件降解信号的结合伙伴，这驱动整个蛋白复合物的蛋白酶体降解 43（图 3Da）。通过在目标蛋白中引入条件激活的 TEV 蛋白酶的切割位点，可以实现蛋白降解的蛋白酶体独立控制 79（图 3Db）。为了实现不同基因开关之间输入和输出水平的时间同步，需要合成控制元件以实现加速或延迟的响应时间。例如，基于蛋白水平磷酸化过程的快速信号转导模块可以用来连接两个慢作用的转录系统，以加速整个电路的动态 32。特定位点重组酶如 Cre 常用于编程时间延迟的转录启动（图 3Dc）。 通过将感兴趣基因的编码区置于反义方向，并由反向平行的 loxP 位点夹住 80,81，或在基因的启动子区域内将终止（STOP）信号置于平行的 loxP 位点之间 82,83，功能性转录单位的重构和基因表达依赖于耗时的 Cre 表达以及 Cre 介导的序列反转或切除，这显著减缓了目标基因的表达。

A long-term goal of synthetic biology is to develop sophisticated gene circuits that serve as a kind of ‘genetic software’ to programme cellular functions analogously to the case of electronic computers⁸⁴. To this end, various prototype circuits providing a standardized toolkit for the design and assembly of specific cell functions have been developed (Fig. 4).
合成生物学的一个长期目标是开发复杂的基因电路，作为一种“遗传软件”来编程细胞功能，类似于电子计算机的情况。为此，已经开发出各种原型电路，提供了一套标准化工具包，用于设计和组装特定的细胞功能。

A | Synthetic memory devices. Memory devices allow a cell to remember a trigger-inducible cell state even after removal of the trigger signal (part Aa). Memory elements can be based on chemical stability, whereby memory is read out as protein expression driven by a trigger (part Ab), bistable gene switches (part Ac) and DNA sequence modification with recombinases and nucleases, which can be used to create synthetic barcodes in non-coding genomic regions (part Ad). B | Boolean calculators. Cells can be programmed to compute various inputs to define the output using combinations of logic gates (see also Box 1). For example, a half-adder uses a combination of XOR and AND gates to calculate the binary addition of two Boolean numbers A and B, while a half-subtractor calculates the difference between A and B (A minus B) using XOR and A N-IMPLY B. Whereas the addition of 1 + 1 is shown as 2 in the conventional decimal system (Dec), the digit 2 is represented as ‘10’ in the binary (Bin) system. C | Synthetic oscillator networks. Ca | Repressilator based on a triple-negative feedback ring architecture. Cb | Dual feedback consisting of an activator module (positive feedback) and a slower-acting repressor module (negative feedback). Implementations based on bacterial quorum-sensing components such as the acyl homoserine lactone (AHL)-inducible transcriptional activator protein LuxR–Herpes simplex-derived virion protein 16 (VP16) trans-activator (activator module) and AHL-degrading N-acyl homoserine lactonase AiiA (repressor module) encode pulse-like oscillations. Cc | Encoding oscillations with antisense technology. Two opposing synthetic promoters drive sense and antisense expression of a synthetic tetracycline-controlled trans-activator tTA (tetracycline-dependent repressor (TetR)–VP16). To generate oscillations, tTA drives its own sense expression (positive feedback), the expression of the oscillating gene and the expression of a second trans-activator (pristinamycin-induced protein (Pip)-dependent trans-activator (PIT)), which then drives the antisense tTA production (negative feedback). D | Synthetic cell–cell communication. To fine-tune signal processes in receiving cells, various strategies can be used. Da | Analogue-to-digital converters (ADCs) translate dose-dependent input signals into an all-or-nothing expression profile by amplifying time-delayed gene expression (Fig. 3Dc). Db | Synthetic band-pass filters permit only target gene expression at intermediate levels of a trigger signal by including regulation of the output by an activator that has a high sensitivity for the signal but low target affinity and a repressor that has low sensitivity for the signal but high target affinity. gRNA, guide RNA; LuxI, Vibrio fischeri-derived acyl homoserine synthase; NHEJ, non-homologous end joining; ScbR, Streptomyces coelicolor butyrolactone-dependent repressor.
A | 合成记忆装置。记忆装置使细胞能够记住触发诱导的细胞状态，即使在去除触发信号后仍然如此（部分 Aa）。记忆元件可以基于化学稳定性，其中记忆通过触发驱动的蛋白质表达来读取（部分 Ab），双稳态基因开关（部分 Ac）以及使用重组酶和核酸酶进行的 DNA 序列修改，这些可以用于在非编码基因组区域创建合成条形码（部分 Ad）。B | 布尔计算器。细胞可以被编程以计算各种输入，以定义输出，使用逻辑门的组合（另见框 1）。例如，半加器使用异或门和与门的组合来计算两个布尔数 A 和 B 的二进制加法，而半减器使用异或门和 A N-IMPLY B 来计算 A 和 B 之间的差（A 减去 B）。在传统的十进制系统（Dec）中，1 + 1 的加法显示为 2，而在二进制（Bin）系统中，数字 2 表示为‘10’。C | 合成振荡器网络。Ca | 基于三重负反馈环架构的抑制器。 Cb | 双重反馈由一个激活模块（正反馈）和一个反应较慢的抑制模块（负反馈）组成。基于细菌群体感应成分的实现，如酰基同源氨酸内酯（AHL）诱导的转录激活蛋白 LuxR–单纯疱疹病毒衍生的病毒蛋白 16（VP16）转激活子（激活模块）和 AHL 降解的 N-酰基同源氨酸内酯酶 AiiA（抑制模块），编码脉冲状振荡。Cc | 使用反义技术编码振荡。两个相对的合成启动子驱动合成四环素控制的转激活子 tTA（四环素依赖性抑制子（TetR）–VP16）的正义和反义表达。为了产生振荡，tTA 驱动其自身的正义表达（正反馈）、振荡基因的表达以及第二个转激活子的表达（依赖于普利司通霉素诱导蛋白（Pip）的转激活子（PIT）），然后驱动反义 tTA 的产生（负反馈）。D | 合成细胞间通信。为了微调接收细胞中的信号过程，可以使用各种策略。 Da | 模拟到数字转换器（ADC）通过放大时间延迟的基因表达，将剂量依赖的输入信号转化为全或无的表达谱（图 3Dc）。Db | 合成带通滤波器仅允许在触发信号的中间水平下靶基因表达，通过包括对信号具有高灵敏度但对靶标亲和力低的激活因子和对信号具有低灵敏度但对靶标亲和力高的抑制因子的输出调节。gRNA，导向 RNA；LuxI，来自嗜盐弧菌的酰基同源氨基酸合成酶；NHEJ，非同源末端连接；ScbR，链霉菌青霉素依赖性抑制因子。

Full size image 完整图像

Memory devices 存储设备

Cells have evolved a variety of mechanisms to remember past experiences in the form of quantifiable memory^10,85 (Fig. 4Aa). Protein stability is an intrinsic memory element of any biological system. For example, an induced transcription factor can keep activating a specific gene switch even after removal of the trigger signal, resulting in sustained and unperturbed target activity over a specific transient time window. Such memory buffers, which are reminiscent of the charge level of electronic capacitors, protect the robustness of a particular subsystem and have been characterized in both natural⁸⁶ and synthetic transcriptional contexts^87,88,89,90. To quantify the charge level of memory buffers in real-time, a synthetic gene network based on hybrid transcription factors was developed. In the default state (discharged memory; 0% charged), the hybrid transcription factor binds and activates a high-affinity promoter (P_A). Loading of the memory device with a trigger signal abolishes trans-activation of P_A but allows the transcription factor to activate a low-affinity promoter (P_B) (charged memory; 100% charged). Degradation of the load signal results in a gradual inactivation of P_B (buffer consumption) and concomitant re-initiation of P_A (full discharge)⁹¹. Quantification of reporter protein levels expressed from P_A and P_B enables precise monitoring of actual charge levels (Fig. 4Ab). To create long-lasting memory devices, positive auto-regulatory feedback loops are effective because they generate reporter proteins at such high levels that they can be inherited through multiple rounds of cell division⁹².
细胞已经进化出多种机制，以量化记忆的形式记住过去的经历 10,85（图 4Aa）。蛋白质稳定性是任何生物系统的内在记忆元素。例如，诱导的转录因子可以在去除触发信号后继续激活特定基因开关，从而在特定的瞬态时间窗口内保持持续且不受干扰的目标活性。这种记忆缓冲器类似于电子电容器的电荷水平，保护特定子系统的稳健性，并已在自然 86 和合成转录环境中得到表征 87,88,89,90。为了实时量化记忆缓冲器的电荷水平，开发了一种基于混合转录因子的合成基因网络。在默认状态（放电记忆；0%充电）下，混合转录因子结合并激活高亲和力启动子（PA）。用触发信号加载记忆装置会消除 PA 的转激活，但允许转录因子激活低亲和力启动子（PB）（充电记忆；100%充电）。 负载信号的降解导致 PB（缓冲消耗）的逐渐失活，并伴随 PA（完全放电）的重新启动。通过量化从 PA 和 PB 表达的报告蛋白水平，可以精确监测实际的电荷水平（图 4Ab）。为了创建持久的记忆设备，正向自我调节反馈回路是有效的，因为它们以如此高的水平生成报告蛋白，以至于可以通过多轮细胞分裂进行遗传。

Epigenetic bistable gene switches characterized by two stable expression states represent a naturally evolved regulatory pattern for dynamic, inheritable and resettable memory⁹³. The presence of a switching signal flips an arbitrarily set default state into the second state, which is maintained over multiple cell generations until a second stimulus toggles the gene switch back to the equally stable default state⁸⁹. In general, bistability in regulatory networks requires two mutually inhibitory gene switches, allowing a minimal derepressive stimulus to sufficiently trigger the prevalence of one expression programme over the other (Fig. 4Ac). Synthetic toggle switches created in bacteria^94,95 and human cells^89,96 based on mutually repressible gene transcription were among the first engineered gene circuits inaugurating modern synthetic biology at the turn of the millennium. Gene expression triggered by derepression (gene induction by repressor inactivation; for example, B represses C, but A triggers C by inactivating B) is particularly effective for achieving tight regulation, pulse-like induction and ultra-high sensitivity to transient stimuli⁹⁵, which are essential for bistable gene expression. Bistable switch performance can be further optimized when gene expression in each state is also self-amplified with positive feedback loops^93,97 (Fig. 4Ac).
表征为两种稳定表达状态的表观遗传双稳态基因开关代表了一种自然进化的调控模式，用于动态、可遗传和可重置的记忆。切换信号的存在将任意设置的默认状态翻转为第二状态，该状态在多个细胞世代中保持，直到第二个刺激将基因开关切换回同样稳定的默认状态。一般而言，调控网络中的双稳态需要两个相互抑制的基因开关，允许最小的去抑制刺激足以触发一种表达程序相对于另一种的优势。基于相互抑制的基因转录在细菌和人类细胞中创建的合成切换开关是引领现代合成生物学的首批工程基因电路之一。 通过去抑制触发的基因表达（通过抑制子失活诱导基因；例如，B 抑制 C，但 A 通过失活 B 来触发 C）在实现严格调控、脉冲式诱导和对瞬时刺激的超高灵敏度方面特别有效，这对于双稳态基因表达至关重要。当每种状态下的基因表达也通过正反馈环自我放大时，双稳态开关的性能可以进一步优化。

In contrast to epigenetic memory, genetic events triggering a permanent change in a DNA sequence result in irreversible genotypes. Because the biochemistry of DNA is inherently compatible with robust, scalable and stable storage of analogue data, permanent genetic memory devices based on DNA editing (Fig. 1A, step 2) can accumulate and convert transient molecular signals into long-lasting information, allowing data recovery and decoding even if the cells are disruptively harvested^42,98,99. By synchronizing the activity of site-specific recombinases with a user-defined stimulus, irreversible DNA rearrangement can be programmed at defined transcription units to create permanently altered promoter architectures and therefore new gene expression profiles⁸¹ (Fig. 3Dc). Non-coding sequence readouts can be inscribed on a user-defined DNA stretch integrated into a target genomic locus, generating an internal barcode region that continuously captures cellular history (Fig. 4Ad). Barcode regions created by multiple sets of orthogonal recombinases controlled by synthetic trigger-inducible gene switches store cellular memory as a combination of single ‘information bits’ formed by binary DNA orientations of individual sequence segments that are flanked by specific recombinase target sites. Incorporation of n pairs of orthogonal recombinase target sites on one DNA stretch creates a data register with a memory capacity of n bits, which is capable of distinguishing 2ⁿ cellular events synchronized with the activity of each trigger-inducible recombinase^98,99 (Fig. 4Ad). Barcode regions can also be generated by CRISPR–Cas9 systems programmed for iterative self-targeting and random induction of point mutations^42,100,101. For example, self-targeting gRNAs guiding the Cas9 nuclease to its own genomic locus and inducing site-specific non-homologous end joining (NHEJ)-dependent DNA repair allow Cas9-induced mutations to continuously rewrite a gRNA-specific barcode region with unlimited memory capacity. Barcode regions created by recombinases or CRISPR–Cas9 systems are not designed to code for cellular functions and must therefore be decoded with conventional omics technologies such as whole-genome sequencing or transcriptome profiling^{42,100,101,102}. Barcode-based memory devices could become an attractive strategy for basic research applications to map genomic changes originating from diseases or associated with cell differentiation.
与表观遗传记忆相对，触发 DNA 序列永久变化的遗传事件导致不可逆的基因型。由于 DNA 的生物化学特性本质上与强大、可扩展和稳定的模拟数据存储兼容，基于 DNA 编辑的永久遗传记忆设备（图 1A，步骤 2）可以积累并将瞬态分子信号转化为持久信息，即使细胞被破坏性收集，也能实现数据恢复和解码 42,98,99。通过将特定位点重组酶的活性与用户定义的刺激同步，可以在定义的转录单位上编程不可逆的 DNA 重排，从而创建永久改变的启动子结构，进而形成新的基因表达谱 81（图 3Dc）。非编码序列的读出可以铭刻在集成到目标基因组位点的用户定义 DNA 片段上，生成一个内部条形码区域，持续捕捉细胞历史（图 4Ad）。 由多组正交重组酶控制的合成触发诱导基因开关创建的条形码区域，将细胞记忆存储为由特定重组酶靶位点两侧的单个序列片段的二元 DNA 取向形成的单个“信息位”的组合。在一段 DNA 上结合 n 对正交重组酶靶位点创建的数据寄存器具有 n 位的记忆容量，能够区分与每个触发诱导重组酶的活动同步的 2n 个细胞事件 98,99（图 4Ad）。条形码区域也可以通过编程为迭代自我靶向和随机诱导点突变的 CRISPR–Cas9 系统生成 42,100,101。例如，自我靶向 gRNA 引导 Cas9 核酸酶到其自身基因组位点并诱导特定位点的非同源末端连接（NHEJ）依赖的 DNA 修复，使得 Cas9 诱导的突变能够不断重写具有无限记忆容量的 gRNA 特异性条形码区域。 由重组酶或 CRISPR–Cas9 系统创建的条形码区域并不是为了编码细胞功能而设计的，因此必须通过传统的组学技术进行解码，例如全基因组测序或转录组分析。基于条形码的记忆设备可能成为基础研究应用中一种有吸引力的策略，以映射源自疾病或与细胞分化相关的基因组变化。

Boolean calculators 布尔计算器

Information processing in electronic computers is based on the execution of rationally assembled Boolean logic gates, which convert multiple input signals into a smaller number of outputs according to defined algorithms¹⁰ (Box 1). In both electronics and biology, two-input OR and AND gates form the basis for programming any computational task^84,103 (Box 1 (See the figure, parts a,b)). By using NOT inverters designed for specific exclusion of a target set of input signals from activating a particular genetic event, AND and OR gates can be inverted to execute NAND (NOT AND) and NOR (NOT OR) logics, respectively (Box 1 (see the figure, part c)). AND and NOT logics can also be combined to obtain A AND NOT B gates, which can be further assembled to form exclusive OR gates (XOR; (A AND NOT B) OR (B AND NOT A)) to permit the presence of only one input signal for output gene expression (Box 1 (see the figure, part d)). Combination of an XOR and an AND gate provides a half-adder function^104,105, whereas the combination of an XOR and an N-IMPLY gate forms a half-subtractor¹⁰⁵, which produces specific outputs depending on the present inputs (biocomputing) (Fig. 4B). Any biocomputing algorithm can be created either by layering different gene switches with interconnected regulatory cascades^63,104,105, or by programming all logic operations into a single gene expression layer^{83,84,106,107}.
电子计算机中的信息处理基于理性组装的布尔逻辑门的执行，这些逻辑门根据定义的算法将多个输入信号转换为较少的输出信号。在电子学和生物学中，双输入的或门（OR）和与门（AND）构成了编程任何计算任务的基础。通过使用专门设计的 NOT 反相器，可以将特定的输入信号集排除在激活特定遗传事件之外，从而将与门和或门反转以执行 NAND（非与）和 NOR（非或）逻辑。与门和非门逻辑也可以结合以获得 A AND NOT B 门，这可以进一步组合形成异或门（XOR；（A AND NOT B）OR（B AND NOT A）），以允许仅有一个输入信号用于输出基因表达。异或门和与门的组合提供了半加器功能，而异或门和 N-IMPLY 门的组合形成了半减器，这会根据当前输入产生特定输出（生物计算）。 任何生物计算算法都可以通过将不同的基因开关与相互连接的调控级联层叠在一起 63,104,105，或者通过将所有逻辑操作编程到单一的基因表达层中 83,84,106,107 来创建。

Box 1 Logic gates and their application to gene regulation
框 1 逻辑门及其在基因调控中的应用

Logic gates are synthetic gene circuits programmed to permit the expression of an output protein only when a strictly defined signature of input signals is matched. OR gates (see the figure, part a), which permit expression of the output protein whenever either of two input signals (A or B) is present, can be encoded by promoter^103,193 or mRNA⁸⁴ architectures containing tandem binding sites for different mutually orthogonal activators or by multiple activator-specific promoters¹⁹⁴ or mRNAs¹⁹⁵ driving the expression of the same target gene. AND gates (see the figure, part b) permit output gene expression only when all input signals are present. All two-hybrid and split expression technologies (see also Supplementary Box 2) inherently encode an AND gate, in which input-defining events are synchronized through constitutive protein dimerization^104,192, intein splicing¹⁹⁶ or chaperone-assisted protein folding¹⁹⁷. Alternatively, mRNAs that could be translated into mature proteins are modified to contain different input-dependent checkpoints, such as aptazymes^63,198, inhibitory hairpins⁸⁴ or target sites for translational repressors⁶³ and self-encoded small regulatory RNA (srRNA)⁵⁶. Output protein is expressed when all inhibitory checkpoints are cleared. AND gates can also be programmed by serial activation of input-specific targets, in which the presence of one input drives the expression of a receptor for the next input^170,179 or induces an active conformation of a target protein specific for the other input¹⁹⁹. NOT inverters (see the figure, part c) specifically invert the computing logic of any AND or OR gate component. Conventional strategies are integration of a repressor-based subsystem into an activated gene expression motif^103,194, the use of synthetic promoters containing tandem binding sites for different trans-repressors¹⁹³ and mRNAs containing multiple target sites for srRNAs^195,200. AND NOT (N-IMPLY; also known as NOT IF) gates (see the figure, part d) can also be created with tandem trans-regulator binding sites, in which one trans-regulator is an activator and the other trans-regulator is a repressor that overrides the regulatory function of the activator by binding to the same target¹⁹⁴. N-IMPLY gates can be further assembled to form exclusive OR gates (XOR) to permit only the presence of one input signal for output gene expression. Similarly, exclusive NOR gates (XNOR), which process only the simultaneous presence or absence of both input signals can be assembled by connecting two IMPLY gates (A IMPLY B = (NOT A) OR B; B IMPLY A = (NOT B) OR A) (see the figure, part e).
逻辑门是合成基因电路，经过编程，仅在严格定义的输入信号特征匹配时允许输出蛋白质的表达。或门（见图，部分 a）允许在任一两个输入信号（A 或 B）存在时表达输出蛋白质，可以通过含有不同互补激活因子的串联结合位点的启动子 103,193 或 mRNA84 结构编码，或通过多个激活因子特异性启动子 194 或 mRNA195 驱动同一靶基因的表达。与门（见图，部分 b）仅在所有输入信号存在时允许输出基因表达。所有的双杂交和分裂表达技术（另见补充框 2）本质上编码一个与门，其中输入定义事件通过构成性蛋白二聚化 104,192、内源性剪接 196 或伴侣蛋白辅助折叠 197 进行同步。或者，能够翻译成成熟蛋白质的 mRNA 被修改为包含不同的输入依赖性检查点，例如适配酶 63,198、抑制性发夹 84 或翻译抑制因子的靶位点 63 以及自编码的小型调节 RNA（srRNA）56。 当所有抑制检查点被清除时，输出蛋白质被表达。AND 门也可以通过串行激活特定输入目标进行编程，其中一个输入的存在驱动下一个输入的受体表达，或诱导特定于另一个输入的目标蛋白的活性构象。NOT 反转器（见图，部分 c）专门反转任何 AND 或 OR 门组件的计算逻辑。传统策略是将基于抑制子的子系统整合到激活的基因表达基序中，使用包含不同转抑制子串联结合位点的合成启动子，以及包含多个目标位点的 mRNA。AND NOT（N-IMPLY；也称为 NOT IF）门（见图，部分 d）也可以通过串联转调节因子结合位点创建，其中一个转调节因子是激活因子，另一个转调节因子是抑制因子，通过结合到相同目标来覆盖激活因子的调节功能。N-IMPLY 门可以进一步组装形成独占或门（XOR），以仅允许一个输入信号的存在用于输出基因表达。 类似地，独占或非门（XNOR）仅处理两个输入信号同时存在或同时不存在的情况，可以通过连接两个 IMPLY 门来组装（A IMPLY B = (NOT A) OR B; B IMPLY A = (NOT B) OR A）（见图，部分 e）。

Show less 显示更少

Synthetic oscillators 合成振荡器

Oscillating gene expression controls and programmes rhythmic cellular activities such as the circadian clock^108,109, glycolysis¹¹⁰ or the cell cycle¹¹¹. The repressilator, which is based on a triple-negative feedback ring architecture consisting of three mutually repressive control elements (Fig. 4Ca), was the first attempt to engineer a synthetic oscillator that programmes rhythmic production of target proteins¹¹². However, subsequent computer-aided studies revealed that a dual-feedback topology consisting of a positive auto-feedback module activating the expression of all interaction partners in a network, combined with a slower-acting negative feedback module repressing the same targets, was most effective to programme autonomous, sustained and tunable gene oscillations^{90,113,114,115}. One of the well-established systems for generating oscillations is based on bacterial quorum-sensing systems¹¹⁶ (Fig. 4Cb). In Vibrio fischeri, acyl homoserine lactone synthase (LuxI) produces acyl homoserine lactone (AHL) as a quorum signal that diffuses across the cell population, providing a marker for population density. Upon passing a critical threshold of population density, accumulated AHL activates the TetR family transcription factor LuxR to induce transcriptional responses at cognate LuxR-specific promoters. Because many quorum-induced cell responses are released in an all-or-nothing manner^117,118,119, synthetic oscillators engineered with LuxR-dependent positive feedback modules typically exhibit pulse-like gene expression dynamics¹²⁰. In this system, a synthetic trans-activator based on LuxR regulates the expression of the oscillating gene and LuxI (positive feedback) as well as the AHL-degrading enzyme N-acyl homoserine lactonase AiiA derived from Bacillus thuringiensis (negative feedback)^121,122 (Fig. 4Cb). In contrast to pulse-like oscillators, harmonic alternation of positive and negative feedback modules can be programmed with antisense technology¹²³ or RNAi¹²⁴ in mammalian cells. Using two opposing synthetic promoters driving sense or antisense expression of a synthetic tetracycline-controlled trans-activator (tTA; TetR–Herpes simplex-derived virion protein 16 (VP16)), pendulum-like oscillations were generated when basal tTA expression could simultaneously activate its own expression (positive feedback), the expression of the oscillating gene (target) and the expression of a second trans-activator (pristinamycin-induced protein (Pip)-dependent transactivator (PIT)) that triggered tTA knockdown through antisense mRNA production (negative feedback) until tTA was depleted enough to not allow efficient expression of PIT, but basal tTA expression remained sufficient to re-initiate another oscillation cycle¹²³ (Fig. 4Cc).
振荡基因表达控制和编程节律性细胞活动，如昼夜节律、糖酵解或细胞周期。抑制振荡器基于一个由三个相互抑制的控制元件组成的三重负反馈环架构，是工程合成振荡器的首次尝试，旨在编程目标蛋白质的节律性生产。然而，随后的计算机辅助研究表明，由一个正自反馈模块激活网络中所有相互作用伙伴的表达，结合一个抑制相同目标的较慢负反馈模块的双反馈拓扑结构，是编程自主、持续和可调基因振荡的最有效方法。生成振荡的一个成熟系统基于细菌群体感应系统。在嗜盐弧菌中，酰基同源氨基酸内酯合成酶（LuxI）产生酰基同源氨基酸内酯（AHL）作为群体信号，能够在细胞群体中扩散，为群体密度提供标记。 在超过临界人口密度阈值后，积累的 AHL 激活 TetR 家族转录因子 LuxR，以诱导在特定 LuxR 特异性启动子上的转录响应。由于许多群体诱导的细胞响应以全有或全无的方式释放，依赖 LuxR 的正反馈模块工程合成振荡器通常表现出脉冲式基因表达动态。在该系统中，基于 LuxR 的合成转激活因子调节振荡基因和 LuxI（正反馈）的表达，以及来自苏云金芽孢杆菌的 AHL 降解酶 N-酰基同源氨基酸内酯酶 AiiA（负反馈）。与脉冲式振荡器相比，正负反馈模块的谐波交替可以通过反义技术或 RNA 干扰在哺乳动物细胞中进行编程。 通过使用两个相对的合成启动子驱动合成四环素控制的转激活因子（tTA；TetR–单纯疱疹病毒衍生的病毒蛋白 16（VP16））的正义或反义表达，当基础 tTA 表达能够同时激活其自身表达（正反馈）、振荡基因（目标）的表达以及第二个转激活因子（依赖于普利司通霉素诱导蛋白（Pip）的转激活因子（PIT））的表达时，产生了钟摆式的振荡，这触发了通过反义 mRNA 产生的 tTA 降解（负反馈），直到 tTA 被耗尽到无法有效表达 PIT，但基础 tTA 表达仍然足以重新启动另一个振荡周期 123（图 4Cc）。

Intercellular communication
细胞间通信

Precise control over cell–cell communication is important to achieve higher-order behaviours in multicellular organisms such as humans. To engineer synthetic intercellular communication systems, gating mechanisms in receiver cells that not only sense but also process and adjust signals produced from sender cells are essential. To programme digital signal processing devices that initiate all-or-nothing responses in receiver cells when the input signal has passed a critical threshold level, a synthetic network consisting of a time-delayed gene expression module and a synthetic trans-activator (such as tTA) was designed. Specifically, insertion of a terminator signal (STOP) flanked by parallel target sites of Cre recombinase between a synthetic tTA-specific promoter and the coding region of a target gene allows tTA to accumulate to saturating levels at its promoter during Cre-mediated terminator excision, permitting maximal trans-activation of target gene expression only upon reconstitution of the transcription unit⁸² (Fig. 4Da). Unlike such synthetic analogue-to-digital converters (ADCs), band-pass filters allow only a specific intermediate range of input signals to trigger gene expression and repress gene expression outside of this defined concentration window¹¹³ (Fig. 4Db). In both bacterial¹²⁵ and mammalian^126,127 systems, the key design principle for synthetic band-pass filters is the use of two antagonistic gene switches regulating the same target gene but with different target affinity and signal sensitivity. Low and intermediate levels of the trigger signal allow a gene switch of higher sensitivity — but lower affinity — to regulate target gene expression in a conventional dose-dependent manner until the low-sensitivity gene switch, which is activated at high signal levels, overrides the regulatory activity of the high-sensitivity gene switch owing to higher affinity for the target gene (Fig. 4Db).
对细胞间通信的精确控制对于实现多细胞生物（如人类）的高级行为至关重要。为了构建合成的细胞间通信系统，接收细胞中的门控机制不仅需要感知信号，还需处理和调整来自发送细胞的信号。为了编程数字信号处理设备，使其在输入信号超过临界阈值时在接收细胞中启动全或无响应，设计了一个由时间延迟基因表达模块和合成转激活因子（如 tTA）组成的合成网络。具体而言，在合成的 tTA 特异性启动子与目标基因的编码区之间插入一个由 Cre 重组酶平行靶位点夹住的终止信号（STOP），允许 tTA 在 Cre 介导的终止子切除过程中在其启动子处积累到饱和水平，从而仅在转录单元重新构建后允许目标基因表达的最大转激活 82（图 4Da）。 与这种合成模拟-数字转换器（ADC）不同，带通滤波器仅允许特定的中间范围输入信号触发基因表达，并在此定义的浓度窗口之外抑制基因表达 113（图 4Db）。在细菌 125 和哺乳动物 126,127 系统中，合成带通滤波器的关键设计原则是使用两个拮抗基因开关调控同一靶基因，但具有不同的靶向亲和力和信号灵敏度。低和中等水平的触发信号允许灵敏度较高——但亲和力较低——的基因开关以常规的剂量依赖方式调控靶基因表达，直到在高信号水平下激活的低灵敏度基因开关由于对靶基因的更高亲和力而覆盖高灵敏度基因开关的调控活性（图 4Db）。

Because the intensity of trigger signals produced from a sender cell population negatively correlates with the distance to receiver cell populations in synthetic cell–cell communication systems, synthetic band-pass filters allow receiver cells to fine-tune distance-dependent gene expression¹²⁵. In most cases, soluble communication signals such as quorum-derived lactones^{103,116,117,118,119,120,122}, cytokines¹²⁸, amino acids^128,129 or vitamins¹³⁰ are used to programme cell–cell communication. In contrast to soluble molecules, gaseous signals are typically far-reaching but are relatively weak and short-lived because of the low physical stability of the gaseous species in vitro^122,130. Nevertheless, upon implantation of individual populations of sender cells and receiver cells into mice, synthetic intercellular communication in vivo mediated by volatile aldehydes achieved comparable signalling dynamics to those of the native endocrine system¹³⁰. To sense physical cell contacts, two strategies were recently developed on the basis of synthetic Notch signalling⁵⁷ (Fig. 2Cb) and semi-synthetic transcriptional readout of synthetic CD45-dependent Janus kinase (JAK)–signal transducer and activator of transcription (STAT) signalling¹³¹ (Fig. 2Ca).
由于在合成细胞-细胞通信系统中，发送细胞群体产生的触发信号强度与接收细胞群体的距离呈负相关，因此合成带通滤波器允许接收细胞微调距离依赖的基因表达。在大多数情况下，使用可溶性通信信号，如群体衍生的内酯、细胞因子、氨基酸或维生素来编程细胞-细胞通信。与可溶性分子相比，气体信号通常传播范围广，但由于气体物种在体外的物理稳定性较低，因此相对较弱且短暂。然而，在将单个发送细胞群体和接收细胞群体植入小鼠后，由挥发性醛介导的体内合成细胞间通信实现了与原生内分泌系统相当的信号动态。为了感知物理细胞接触，最近基于合成 Notch 信号开发了两种策略。 2Cb) 和合成的 CD45 依赖的 Janus 激酶 (JAK)–信号转导和转录激活因子 (STAT) 信号的半合成转录读出 131 (图 2Ca)。

Prototype synthetic gene circuits represent an ideally simplified replicate of complex regulatory motifs used by native cells, serving as a high-level starting point for in-depth analysis of related medical conditions and/or the development of more advanced solutions. In other words, prototype gene circuits render complex cellular processes accessible to engineering (Box 1; Fig. 4) and allow the design of application-specific cell functions (Figs 5,6).
原型合成基因电路代表了对天然细胞使用的复杂调控基序的理想简化复制，作为深入分析相关医学状况和/或开发更先进解决方案的高层次起点。换句话说，原型基因电路使复杂的细胞过程对工程学变得可及（框 1；图 4），并允许设计特定应用的细胞功能（图 5，6）。

a | Stem cell differentiation. A synthetic lineage control network transiently integrated into the endogenous transcription network of stem cells can control differentiation into target cell types, for example, differentiation of induced pluripotent stem cells (iPSCs) into pancreatic β-cells with high robustness, reliability and efficiency. See the main text for a detailed mechanistic description. b | Drug screening. Functional mimetics of native drug targets can be created by custom-designing synthetic gene circuits. In this example, a synthetic ethionamide-dependent repressor (EthR)-based gene switch operating in mammalian cells sufficiently mimicked a drug-resistance operon in Mycobacterium tuberculosis and resulted in the identification of 2-phenylethyl butyrate as a new lead compound for anti-tuberculosis pharmacotherapy. c | Drug development. Using a synthetic cell–cell communication system, a new quorum-quenching strategy for treating Pseudomonas aeruginosa was developed. Human cells were engineered to sense the quorum signal PAI-1 and in response to produce antibacterial enzymes that trigger quorum signal destruction (through expression of the N-acyl homoserine lactonase MomL) as well as destruction of bacterial biofilms (through expression of the bipartite glycoside hydrolase PslG_h–PelA_h). d | Lineage tracing and retrograde labelling. Cre recombinase-based memory and logic are important tools for lineage tracing in transgenic animal models. For example, this method was used to identify leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5)-positive cells as bona fide stem cells of the intestinal crypts. See the main text for a detailed mechanistic description. AHL, acyl homoserine lactone; CRE, cAMP-responsive element; CREB1, cAMP-responsive element-binding protein 1; HTS, high-throughput sequencing; LacZ, β-galactosidase; LasR, transcriptional activator protein from P. aeruginosa; MAFA, v-maf musculoaponeurotic fibrosarcoma homologue A; NGN3, neurogenin 3; PDX1, pancreas/duodenum homeobox protein 1; VanR, vanillic acid repressor; VP16, Herpes simplex-derived virion protein 16.
a | 干细胞分化。一个合成谱系控制网络暂时整合到干细胞的内源性转录网络中，可以控制向目标细胞类型的分化，例如，将诱导多能干细胞（iPSCs）分化为胰腺β细胞，具有很高的稳健性、可靠性和效率。详细的机制描述见正文。b | 药物筛选。通过定制设计合成基因电路，可以创建原生药物靶点的功能模拟物。在这个例子中，一个基于合成乙硫氨酸依赖性抑制子（EthR）的基因开关在哺乳动物细胞中有效模拟了结核分枝杆菌中的药物抗性操纵子，并导致识别出 2-苯乙基丁酸酯作为抗结核药物治疗的新先导化合物。c | 药物开发。利用合成细胞间通信系统，开发了一种新的抑制群体感应的策略用于治疗铜绿假单胞菌。 人类细胞被工程化以感知群体信号 PAI-1，并响应性地产生抗菌酶，这些酶触发群体信号的破坏（通过表达 N-酰基同源氨基酸内酯酶 MomL）以及破坏细菌生物膜（通过表达双部分糖苷水解酶 PslGh–PelAh）。d | 系谱追踪和逆向标记。基于 Cre 重组酶的记忆和逻辑是转基因动物模型中系谱追踪的重要工具。例如，这种方法被用来识别富含亮氨酸重复的 G 蛋白偶联受体 5（LGR5）阳性细胞作为肠隐窝的真正干细胞。有关详细的机制描述，请参见正文。AHL，酰基同源氨基酸内酯；CRE，cAMP 响应元件；CREB1，cAMP 响应元件结合蛋白 1；HTS，高通量测序；LacZ，β-半乳糖苷酶；LasR，来自铜绿假单胞菌的转录激活蛋白；MAFA，v-maf 肌腱纤维肉瘤同源物 A；NGN3，神经生成素 3；PDX1，胰腺/十二指肠同源盒蛋白 1；VanR，香草酸抑制子；VP16，单纯疱疹病毒衍生的病毒蛋白 16。

Full size image 完整图像

A | Therapeutic bacteria. Aa | Use in diagnostics. Bacteria engineered with synthetic memory devices can collect and remember environmental signals while residing in the gut. Following isolation of the diagnostic bacteria from faeces, profiles of the gastrointestinal tract can be read with a reporter protein or computed by decrypting pre-programmed memory barcodes, providing information about potential pathological changes. Ab | Use as therapeutics. Orally administered bacteria can migrate to certain tumours in vivo and release drugs in a user-defined manner, which is here based on a pulse-like oscillator controlling periodic cycles of bacterial proliferation, bacterial suicide and concomitant drug release (see also Fig. 4Cb). B | Chimeric antigen receptor (CAR) T cell-based therapy. Clinically approved CAR T cell systems are based on CAR-mediated activation of T cell signalling that allows tumour-specific and autonomous cell killing. The current focus of CAR T cell development is on improving their killing activity and on the reduction of on-target, off-tumour cell killing by trigger-inducible CAR expression (conditionally activated CARs). These functionalities can be designed by following the principles of Boolean logic gates (see also Box 1). C | Designer cell implants. Any cell type can be engineered to carry a custom-designed synthetic circuit for the treatment of a specific metabolic disease, such as diabetes or immune disorders. Following encapsulation into a clinically licensed implant device, the designer implant residing at a vascularized site can coordinate diagnosis (by detecting specific chemical markers) with treatment and/or prevention of diseases (by releasing therapeutic agents) in an autonomous and/or remote-controlled manner. D | Cancer biocomputers. Synthetic gene circuits controlling apoptosis in cancer cells are based on multi-input logic gates integrating a cancer-specific expression profile. One strategy is to use a cancer-specific profile of microRNAs (miRNAs) to positively (through highly expressed miRNAs) and negatively (through lowly expressed miRNAs) control the expression of a pro-apoptotic protein such as apoptosis regulator BAX. Application of such cancer biocomputers in vivo requires transfer of circuits through gene therapy, for example, involving lentiviral vectors. AHL, acyl homoserine lactone; E. coli, Escherichia coli; IL-22, interleukin 22; LacI, lactose operon repressor; LuxI, Vibrio fischeri-derived acyl homoserine synthase; NFAT, nuclear factor of activated T cells; P_NFAT, NFAT-responsive promoter; rtTA, reversed tetracycline-dependent trans-activator; S. Typhimurium, Salmonella enterica subsp. enterica serovar Typhimurium; UTR, untranslated region.
A | 治疗性细菌。Aa | 在诊断中的应用。经过工程改造的细菌配备合成记忆装置，可以在肠道内收集和记住环境信号。在从粪便中分离出诊断细菌后，可以通过报告蛋白读取胃肠道的特征，或通过解码预编程的记忆条形码计算，提供有关潜在病理变化的信息。Ab | 作为治疗剂的应用。口服给药的细菌可以在体内迁移到特定肿瘤，并以用户定义的方式释放药物，这里基于脉冲式振荡器控制细菌增殖、细菌自杀和伴随药物释放的周期性循环（另见图 4Cb）。B | 嵌合抗原受体（CAR）T 细胞基础的治疗。临床批准的 CAR T 细胞系统基于 CAR 介导的 T 细胞信号激活，允许肿瘤特异性和自主细胞杀伤。目前 CAR T 细胞开发的重点是提高其杀伤活性，并通过触发诱导的 CAR 表达（条件激活的 CAR）减少靶向、非肿瘤细胞的杀伤。 这些功能可以通过遵循布尔逻辑门的原理来设计（另见框 1）。C | 设计师细胞植入物。任何细胞类型都可以被工程化，以携带为特定代谢疾病（如糖尿病或免疫紊乱）治疗而定制的合成电路。在封装到临床许可的植入设备后，位于血管化部位的设计师植入物可以以自主和/或远程控制的方式协调诊断（通过检测特定化学标记）与疾病的治疗和/或预防（通过释放治疗剂）。D | 癌症生物计算机。控制癌细胞凋亡的合成基因电路基于多输入逻辑门，整合癌症特异性表达谱。一种策略是使用癌症特异性微小 RNA（miRNA）谱，通过高表达的 miRNA 正向（通过高表达的 miRNA）和低表达的 miRNA 负向（通过低表达的 miRNA）控制促凋亡蛋白如凋亡调节因子 BAX 的表达。这种癌症生物计算机在体内的应用需要通过基因治疗转移电路，例如，涉及慢病毒载体。 AHL，酰基同源氨基酸内酯；大肠杆菌，Escherichia coli；IL-22，白细胞介素 22；LacI，乳糖操纵子抑制子；LuxI，来自嗜盐菌的酰基同源氨基酸合成酶；NFAT，活化 T 细胞的核因子；PNFAT，NFAT 响应启动子；rtTA，反向四环素依赖性转激活因子；伤寒沙门氏菌，Salmonella enterica subsp. enterica serovar Typhimurium；UTR，非翻译区。

Full size image 完整图像

Molecular diagnostics 分子诊断

Transcriptional gene switches based on prokaryotic transcription factors (Fig. 2A,B) are widely used in metabolic engineering for real-time monitoring of difficult-to-trace metabolites¹³² or to increase bioprocess efficiency in the production of protein therapeutics¹⁰. Similarly, biosensors based on ligand-regulated nucleic acids¹³³ (Fig. 2D) and ligand-responsive reporter proteins¹³⁴ (Fig. 3B; see also Supplementary Box 2) transferred from living cell systems into cell-free contexts^135,136 are being used to develop point-of-care diagnostics for personalized medicine¹³⁷. Although human cells carrying natural, receptor-mediated gene switches (Fig. 2Ca) responding to histamine¹³⁸ or tumour necrosis factor (TNF)¹³⁹ also showed impressive diagnostic capacities for detecting allergic responses and bacterial infections in human blood in vitro, synthetic gene circuits can improve the biosensing precision of living cells by avoiding the risks of false negatives potentially caused by signal fluctuations. For example, encapsulated bacteria harbouring genetic software consisting of gene switches (Fig. 2A), ADCs (Fig. 4Da), two-input logic gates (Box 1) and a recombinase-based permanent memory register (Fig. 4Ad) were engineered to iteratively detect and compute pathological levels of short-lived metabolites such as nitrogen oxides and glucose from clinical urine samples, offering an attractive diagnostic approach for metabolic diseases such as diabetes¹⁴⁰.
基于原核转录因子的转录基因开关（图 2A，B）在代谢工程中被广泛应用，用于实时监测难以追踪的代谢物 132 或提高蛋白治疗药物生产的生物过程效率 10。同样，基于配体调节的核酸 133（图 2D）和配体响应的报告蛋白 134（图 3B；另见补充框 2）从活细胞系统转移到无细胞环境 135,136，正在用于开发个性化医学的即时诊断工具 137。尽管携带自然受体介导基因开关的人类细胞（图 2Ca）对组胺 138 或肿瘤坏死因子（TNF）139 的反应在体外检测过敏反应和细菌感染方面也显示出令人印象深刻的诊断能力，但合成基因电路可以通过避免信号波动可能导致的假阴性风险，提高活细胞的生物传感精度。例如，封装的细菌携带由基因开关组成的遗传软件（图 2A），ADCs（图。 4Da)，设计了双输入逻辑门（框 1）和基于重组酶的永久记忆寄存器（图 4Ad），以迭代检测和计算临床尿液样本中短寿命代谢物（如氮氧化物和葡萄糖）的病理水平，为代谢疾病（如糖尿病 140）提供了一种有吸引力的诊断方法。

Stem cell research 干细胞研究

In human cells, plasticity, pluripotency and cell fate are governed by a universal network of endogenous transcription factors that constantly respond to internal or environmental signals and accordingly readjust cell type-specific gene expression⁹⁷. For example, forced overexpression of the master transcription factors octamer-binding protein 4 (OCT4; also known as POU5F1), sex-determining region Y box 2 (SOX2), Krueppel-like factor 4 (KLF4) and MYC proto-oncogene protein overrides nearly all epigenetic memory associated with lineage commitment and can therefore convert any somatic cell type into a pluripotent state¹⁴¹. However, differentiation of pluripotent precursor cells into adult cell types requires strictly controlled temporal dynamics between lineage-specific master transcription factors. For example, differentiation of pancreatic progenitor cells into mature β-cells is governed by the master transcription factors pancreas/duodenum homeobox protein 1 (PDX1), neurogenin 3 (NGN3) and MAFA. Ectopic overexpression of PDX1, NGN3 and MAFA in different endodermal cell types can trigger β-cell-like identity^142,143, but the lack of the temporal regulation that normally occurs in vivo (Fig. 5a) limits differentiation efficiency and β-cell yield. Current technologies for β-cell differentiation are based on cultivation protocols allowing for precisely timed addition of growth factors and chemical additives to manipulate the signalling pathways involved in the regulation of endogenous PDX1, NGN3 and MAFA expression^144,145. Notably, a synthetic gene circuit consisting of a tunable band-pass filter (Fig. 4Db) regulating NGN3 expression synchronized to cell stage-specific PDX1 expression and β-cell-specific MAFA transcription was shown to differentiate pancreatic progenitor cells into β-cells with unprecedented efficiency¹²⁷. In short, this system — known as a lineage control network — is reminiscent of a genetic software that uses vanillic acid as the sole trigger compound to control differentiation-stage-dependent NGN3 expression. When this system is incorporated into pancreatic progenitor cells, addition of intermediate concentrations of vanillic acid triggers a surge in the second messenger cAMP, which drives expression of a vanillic acid-dependent trans-activator (vanillic acid repressor VanR fused to VP16 (VanR–VP16)) from a high-sensitivity, low-affinity cAMP-responsive element-binding protein 1 (CREB1)-specific promoter. VanR–VP16 then triggers expression of both ectopic NGN3 and an shRNA that knocks down endogenous PDX1 expression, which is essential for the transition from pancreatic progenitor cell state to endocrine progenitor fate. To control the subsequent differentiation stage of β-cell maturation characterized by high-level expression of PDX1 and MAFA and low-level expression of NGN3, high concentrations of vanillic acid were added to simultaneously terminate VanR–VP16-dependent trans-activation of NGN3 and activate a low-sensitivity, high-affinity CREB1-specific promoter driving expression of ectopic PDX1 and MAFA. Notably, this gene circuit was operational on episomal vectors in pancreatic progenitor cells following transient transfection, which triggered permanent differentiation of pancreatic β-cells by interacting with endogenous master transcription factors without the need for gene editing (Fig. 5a). Therefore, regulation of cell fate using rationally designed gene circuits might become a new gold standard in stem cell research, but the moderately high costs associated with biopharmaceutical manufacturing of stem cell-derived therapeutic cell products might limit the applicability for clinical and commercial purposes¹⁴⁶.
在人类细胞中，塑性、多能性和细胞命运由一个普遍的内源性转录因子网络所控制，该网络不断响应内部或环境信号，并相应地调整细胞类型特异的基因表达。例如，强制过表达主转录因子八聚体结合蛋白 4（OCT4；也称为 POU5F1）、性别决定区 Y 框蛋白 2（SOX2）、克鲁伯样因子 4（KLF4）和 MYC 原癌基因蛋白可以覆盖几乎所有与谱系承诺相关的表观遗传记忆，因此可以将任何体细胞类型转化为多能状态。然而，多能前体细胞分化为成年细胞类型需要严格控制谱系特异性主转录因子之间的时间动态。例如，胰腺前体细胞分化为成熟β细胞的过程由主转录因子胰腺/十二指肠同源盒蛋白 1（PDX1）、神经发生因子 3（NGN3）和 MAFA 所调控。 异位过表达 PDX1、NGN3 和 MAFA 在不同内胚层细胞类型中可以触发β细胞样身份，但缺乏在体内正常发生的时间调控（图 5a）限制了分化效率和β细胞产量。目前的β细胞分化技术基于培养方案，允许精确时机添加生长因子和化学添加剂，以操控参与内源性 PDX1、NGN3 和 MAFA 表达调控的信号通路。值得注意的是，一个由可调带通滤波器（图 4Db）组成的合成基因电路，调节与细胞阶段特异性 PDX1 表达和β细胞特异性 MAFA 转录同步的 NGN3 表达，已被证明能够以空前的效率将胰腺前体细胞分化为β细胞。简而言之，这个系统——被称为谱系控制网络——类似于一种遗传软件，使用香草酸作为唯一的触发化合物来控制依赖于分化阶段的 NGN3 表达。 当该系统被引入胰腺前体细胞时，中等浓度的香草酸的添加会引发第二信使 cAMP 的激增，这驱动了香草酸依赖的转激活因子（香草酸抑制因子 VanR 与 VP16 融合（VanR–VP16））的表达，该因子来自高灵敏度、低亲和力的 cAMP 响应元件结合蛋白 1（CREB1）特异性启动子。VanR–VP16 随后触发异位 NGN3 的表达以及一种 shRNA 的表达，该 shRNA 下调内源性 PDX1 的表达，而 PDX1 对于从胰腺前体细胞状态向内分泌前体命运的转变至关重要。为了控制β细胞成熟的后续分化阶段，该阶段以高水平的 PDX1 和 MAFA 表达以及低水平的 NGN3 表达为特征，添加了高浓度的香草酸，以同时终止 VanR–VP16 依赖的 NGN3 转激活，并激活驱动异位 PDX1 和 MAFA 表达的低灵敏度、高亲和力的 CREB1 特异性启动子。 值得注意的是，该基因电路在胰腺前体细胞中通过瞬时转染在外源性载体上运行，这通过与内源性主转录因子相互作用触发了胰腺β细胞的永久分化，而无需基因编辑（图 5a）。因此，使用合理设计的基因电路调控细胞命运可能成为干细胞研究的新标准，但与干细胞衍生治疗细胞产品的生物制药制造相关的适度高成本可能限制其在临床和商业用途上的适用性。

Drug discovery and development
药物发现与开发

Using synthetic gene circuits, functional drug target mimetics and anti-targets can be engineered and assembled on user-defined drug screening platforms and with customizable phenotypic outputs. For example, the ethionamide-dependent repressor EthR of Mycobacterium tuberculosis is an attractive drug target for anti-tuberculosis pharmacotherapy. EthR represses the expression of flavin-containing monooxygenase EthA, which is necessary for converting the prodrug ethionamide — the last line of treatment in tuberculosis — into an active compound. Thus, drugs that suppress EthR activity could greatly improve the treatment of tuberculosis, counteracting resistance to ethionamide. The development of a screening assay in a human cell line operating a synthetic EthR-based gene switch, including EthR–VP16 (EthR coupled to the trans-activating domain of VP16) and an EthR-responsive promoter driving expression of a reporter protein (Fig. 5b), resulted in the discovery of the US Food and Drug Administration (FDA)-approved food additive 2-phenylethyl butyrate as a new lead compound that interferes with EthR–DNA binding and thus can be used to assist in the treatment of tuberculosis¹⁴⁷. Synthetic gene circuits can also be applied on abiotic screening platforms^135,148, which results in higher throughput and sensitivity to target compounds but does not take into account potential issues such as cytotoxicity or tissue permeability, as a cell-based system would.
利用合成基因电路，可以在用户定义的药物筛选平台上构建和组装功能性药物靶标模拟物和抗靶标，并具有可定制的表型输出。例如，结核分枝杆菌的依赖于乙硫氨酸的抑制因子 EthR 是抗结核药物治疗的一个有吸引力的药物靶标。EthR 抑制含黄素单氧化酶 EthA 的表达，而 EthA 对于将前药乙硫氨酸（结核病的最后治疗手段）转化为活性化合物是必需的。因此，抑制 EthR 活性的药物可以大大改善结核病的治疗，抵抗对乙硫氨酸的耐药性。在一个人类细胞系中开发的筛选测定，操作一个基于合成 EthR 的基因开关，包括 EthR–VP16（EthR 与 VP16 的转激活域结合）和驱动报告蛋白表达的 EthR 响应启动子（图 5b），导致发现了美国食品药品监督管理局（FDA）批准的食品添加剂 2-苯乙基丁酸酯作为一种新的先导化合物，它干扰 EthR–DNA 结合，因此可以用于辅助治疗结核病。 合成基因电路也可以应用于非生物筛选平台 135,148，这导致对目标化合物的高通量和高灵敏度，但并未考虑细胞系统可能存在的细胞毒性或组织渗透性等潜在问题。

Additionally, drug target mimetics created with synthetic gene circuits not only enable the discovery of new drugs but also allow in-depth validation of novel treatment strategies (Fig. 5c). For example, chronic Pseudomonas aeruginosa infections are a considerable clinical concern because the bacteria generate antibiotic-resistant biofilms in vivo. During infections, the pathogens communicate through quorum signals such as PAI-1 (Pseudomonas autoinducer 1; also known as N-3-oxo-dodecanoyl-l-homoserine lactone (3O-C₁₂-HSL)) and PAI-2 (also known as C₄-HSL) to establish survival mechanisms or to produce virulence factors. Capitalizing on quorum sensing, a synthetic cell–cell communication system has been engineered in which human cells respond to PAI-1 by expressing the synthetic transcription factor LasR –VP16, which triggers an anti-infective response upon binding to PAI-1. Specifically, PAI-1-inducible expression of a bipartite glycoside hydrolase PslG_h–PelA_h and the AHL lactonase MomL (both mediated by LasR–VP16) results in the destruction of biofilms (by PslG_h–PelA_h) and degradation of PAI-1 and PAI-2 (by MomL)¹⁴⁹ (Fig. 5c). This destruction of quorum signals compromised key resistance mechanisms of a clinical isolate (PA01 strain), restoring antibiotic susceptibility and reducing the cytotoxicity in infected human tissues. In contrast to previous attempts using PAI-1-responsive E. coli to kill P. aeruginosa¹⁵⁰, a key advantage of this quorum-quenching strategy is a reduced risk of triggering drug-resistance mechanisms. This could be an effective basis for developing new strategies for fighting infections as soon as technologies for safe delivery of therapeutic gene circuits into host-specific tissues in vivo become available.
此外，利用合成基因电路创建的药物靶点模拟物不仅能够发现新药，还可以深入验证新治疗策略（图 5c）。例如，慢性铜绿假单胞菌感染是一个相当严重的临床问题，因为这些细菌在体内会产生抗生素耐药的生物膜。在感染过程中，病原体通过群体信号进行交流，如 PAI-1（铜绿假单胞菌自诱导因子 1；也称为 N-3-氧代十二酰-L-半胱氨酸内酯（3O-C12-HSL））和 PAI-2（也称为 C4-HSL），以建立生存机制或产生毒力因子。利用群体感应，工程化了一种合成细胞间通信系统，其中人类细胞通过表达合成转录因子 LasR–VP16 对 PAI-1 作出反应，该因子在与 PAI-1 结合后触发抗感染反应。具体而言，PAI-1 诱导的双部分糖苷水解酶 PslGh–PelAh 和 AHL 内酯酶 MomL（均由 LasR–VP16 介导）的表达导致生物膜的破坏（由 PslGh–PelAh）和 PAI-1 及 PAI-2 的降解（由 MomL）149（图 5c）。 这种对群体感应的破坏信号表明临床分离株（PA01 菌株）关键抗性机制受损，恢复了抗生素敏感性并减少了感染人类组织的细胞毒性。与之前使用 PAI-1 响应的 E. coli 杀死铜绿假单胞菌的尝试相比，这种群体淬灭策略的一个主要优势是降低了触发药物抗性机制的风险。这可能为开发新的感染治疗策略提供有效基础，前提是安全递送治疗基因电路到特定宿主组织的技术可用。

Gene editing 基因编辑

Integration of the principles of single-layer logic gates and permanent recombinase-based memory (Box 1; Fig. 3Dc) has enabled the generation of transgenic mice stably expressing different variants of Cre recombinase as powerful research tools for mechanistic studies and disease profiling¹⁵¹. For example, a synthetic gene circuit was created in which the expression of Cre and GFP was regulated from a stem cell-specific leucine-rich repeat-containing GPCR5 (LGR5)-responsive promoter, while the expression of a blue pigment generated by β-galactosidase (LacZ) was dependent on Cre-mediated sequence inversion (Fig. 5d). Once the transcription unit was reconstituted in LGR5-positive stem cells, pigment production was permanently locked in a constitutive state even when the cells differentiated into LGR5-negative progeny. Because tissues marked with the dark pigment must have differentiated from a progenitor population producing both GFP and LacZ, this approach enabled the discovery of multipotent stem cells in the crypt base of the small intestine, establishing LGR5-positive cells as multipotent intestinal stem cells¹⁵². Similar lineage tracing and retrograde labelling studies have elucidated basic mechanisms of cardiac development¹⁵³ and identified sleep neurons¹⁵⁴ in mammals. Owing to ethical concerns, however, approaches based on gene editing might remain restricted to the permanent modification of individual cells in vitro (for example, the creation of stable cell lines for metabolic engineering or correction of patients’ own blood cells for autologous transfusion therapies), genetic modification of agricultural plants and the generation of transgenic animal models for research purposes.
单层逻辑门和基于永久重组酶的记忆原理的整合（框 1；图 3Dc）使得生成稳定表达不同变体 Cre 重组酶的转基因小鼠成为可能，这些小鼠作为强大的研究工具用于机制研究和疾病分析 151。例如，创建了一个合成基因电路，其中 Cre 和 GFP 的表达由一种特异性于干细胞的富含亮氨酸重复的 GPCR5（LGR5）响应性启动子调控，而由β-半乳糖苷酶（LacZ）生成的蓝色颜料的表达则依赖于 Cre 介导的序列反转（图 5d）。一旦转录单位在 LGR5 阳性干细胞中重构，颜料的产生便被永久锁定在一种构成状态，即使细胞分化为 LGR5 阴性后代也是如此。由于标记有深色颜料的组织必须来源于同时产生 GFP 和 LacZ 的祖细胞群体，这种方法使得在小肠隐窝基底发现多能干细胞成为可能，确立了 LGR5 阳性细胞作为多能肠道干细胞 152。 类似的谱系追踪和逆行标记研究阐明了心脏发育的基本机制 153，并在哺乳动物中识别了睡眠神经元 154。然而，由于伦理问题，基于基因编辑的方法可能仍然仅限于体外对单个细胞的永久性修改（例如，创建用于代谢工程的稳定细胞系或修正患者自身血细胞以进行自体输血疗法）、农业植物的基因改造以及为研究目的生成转基因动物模型。

Agriculture and the environment
农业与环境

To support the production of isobutanol, a next-generation biofuel, synthetic gene circuits conferring ionic liquid resistance can increase the robustness and survival of bacterial production strains¹⁵⁵. Improved productivity for feedstock and raw materials has also been achieved by using a synthetic cell–cell communication strategy consisting of a fungal specialist population converting lignocellulosic biomass into soluble saccharides and a bacterial fermentation specialist that metabolizes soluble saccharides into desired products, such as isobutanol¹⁵⁶. Such synthetic consortia, in which a complex task is divided among multiple subpopulations that perform simple tasks with high robustness, are extremely effective for achieving high workforce productivity. For proof of concept, blueprints of synthetic consortia performing population-level oscillations¹⁵⁷ (Fig. 4C) or Boolean calculations^158,159 (Box 1) were created using individual activator and repressor strains¹⁵⁷ or cell populations equipped with different Boolean logic gates^158,159. To achieve control over reproduction of agricultural animals, a synthetic communication system was applied to achieve ‘synthetic artificial’ insemination¹⁶⁰. Cows were implanted with cellulose capsules loaded with bull sperm and mammalian cells transgenic for luteinizing hormone (LH)-specific cellulase production. Using this device, systemic surges of LH during the cow’s natural ovulation cycle could trigger capsule breakdown and sperm release. Lastly, synthetic gene circuits were created to control mosquito spread and virulence^161,162. However, the potential impact of synthetic gene drive systems on entire biotopes raises similar ethical concerns to those involved in generating transgenic mammals.
为了支持异丁醇这一下一代生物燃料的生产，赋予离子液体抗性的合成基因电路可以提高细菌生产菌株的稳健性和生存能力 155。通过使用一种合成细胞间通信策略，已实现了原料和原材料的生产力提升，该策略由一个专门的真菌群体将 lignocellulosic 生物质转化为可溶性糖类，以及一个专门的细菌发酵群体将可溶性糖类代谢为所需产品，如异丁醇 156。这种合成共生体将复杂任务分配给多个子群体，各自执行简单任务且具有高稳健性，对于实现高工作效率极为有效。作为概念验证，创建了执行群体水平振荡 157（图 4C）或布尔计算 158,159（框 1）的合成共生体蓝图，使用了单个激活剂和抑制剂菌株 157 或配备不同布尔逻辑门的细胞群体 158,159。 为了实现对农业动物繁殖的控制，采用了一种合成通信系统以实现“合成人工”授精。母牛体内植入了装有公牛精子和转基因哺乳动物细胞（用于产生促黄体激素（LH）特异性纤维素酶）的纤维素胶囊。利用该装置，母牛自然排卵周期中的 LH 系统性激增可以触发胶囊的分解和精子的释放。最后，创建了合成基因电路以控制蚊子的传播和致病性。然而，合成基因驱动系统对整个生物群落的潜在影响引发了与产生转基因哺乳动物相关的类似伦理问题。

Early diagnosis is critical for successful disease treatment. However, patients usually seek medical advice in response to nonspecific symptoms such as pain, nausea or diarrhoea, and many diseases do not show early symptoms. Thus, implantation of living cells engineered for disease-specific biomarker sensing systems coupled to the expression of a quantifiable readout or a therapeutic protein would greatly facilitate the diagnosis and/or treatment of many asymptomatic disease states.
早期诊断对成功的疾病治疗至关重要。然而，患者通常在出现疼痛、恶心或腹泻等非特异性症状时寻求医疗建议，许多疾病并不表现出早期症状。因此，植入经过工程改造的活细胞，以用于特定疾病生物标志物传感系统，并结合可量化的读出或治疗蛋白的表达，将极大促进许多无症状疾病状态的诊断和/或治疗。

Therapeutic bacteria 治疗性细菌

Bacteria engineered to contain synthetic memory devices (Fig. 4A) can reside in the mammalian gut for months, collecting information about their environment (including changes associated with disease) and translating it into permanent memory^{87,163,164,165}. Following isolation of these diagnostic bacteria from faeces, an environmental profile of the gastrointestinal tract can then be reconstituted by user-defined analysis of genetic readouts or memory barcodes (FigS 4Ad, 6Aa). By using synthetic intercellular communication networks, therapeutic bacteria can also be programmed to sense pathogenic strains and directly trigger antimicrobial responses¹⁶⁶. Certain bacterial species are also excellent vehicles for targeting hypoxic tumour microenvironments in vivo^167,168. By synchronizing a synthetic quorum-coordinated cell death programme with the dynamics of a pulse-like oscillator (Fig. 4Cb), Salmonella enterica subsp. enterica serovar Typhimurium carrying an antitumour toxin was programmed to invade colorectal tumour tissues in mice, resulting in periodical drug release and a significant reduction in tumour mass¹¹⁸ (Fig. 6Ab).
工程化的细菌被设计为包含合成记忆装置（图 4A），可以在哺乳动物肠道中生存数月，收集关于其环境的信息（包括与疾病相关的变化）并将其转化为永久记忆 87,163,164,165。在从粪便中分离出这些诊断细菌后，可以通过用户定义的基因读出或记忆条形码分析重建胃肠道的环境特征（图 S 4Ad, 6Aa）。通过使用合成的细胞间通信网络，治疗性细菌也可以被编程以感知病原菌株并直接触发抗微生物反应 166。某些细菌种类也是靶向缺氧肿瘤微环境的优秀载体 167,168。通过将合成的群体协调细胞死亡程序与脉冲振荡器的动态同步（图 4Cb），携带抗肿瘤毒素的沙门氏菌（Salmonella enterica subsp. enterica serovar Typhimurium）被编程为侵入小鼠的结直肠肿瘤组织，导致药物的周期性释放和肿瘤质量的显著减少 118（图 6Ab）。

Chimeric antigen receptor T cell-based anticancer therapy
嵌合抗原受体 T 细胞基础的抗癌疗法

Human T lymphocytes stably expressing CARs are a clinically validated cell therapy for treating chronic and acute lymphoid leukaemia^11,12. Capitalizing on the design strategy of receptor-mediated activation of T cell signalling (Fig. 2Ca), patient-specific T cells are modified in vitro to recognize cancer-specific antigens through CARs. Upon transfusion into a patient, autologous CAR T cells autonomously migrate to target tissues and recruit the endogenous immune system for local destruction of tumour cells (Fig. 6B). The first three generations of CAR designs have focused on the development of an optimal receptor structure that can effectively trigger T cell signalling upon binding to single surface antigens and on the prevention of graft-versus-host reactions by deleting endogenous T cell receptors through gene editing¹⁶⁹ (Fig. 1A, step 2). Current development of CAR T cells focuses on the improvement of tumour specificity by adopting synthetic logic gate principles¹⁷⁰ (Box 1) allowing synchronization of T cell activation with other, user-defined gene expression programmes¹⁷¹, including co-production of cytokines¹⁷¹ and ion channels¹⁷² to improve tumour cell-killing potency, co-expression of a chemokine receptor to facilitate migration towards solid tumours¹⁷³ and trigger-inducible CAR expression¹⁷⁰ to improve the on-target activity ratio. Implementations of CAR T cells with AND logic gates are based on reconstitution of CAR-dependent signal transduction through drug-induced protein dimerization¹⁷⁴, incorporation of bi-specific antibodies into the CAR framework¹⁷⁵ or triggering of CAR transcription through activation of another cell surface receptor¹⁷⁰. By using a T cell-suppressing CAR variant as a NOT inverter (Box 1), CAR T cells following an A AND NOT B logic (A; surface marker of tumour; B: similar surface marker on healthy tissue) have also been developed to increase tumour specificity¹⁷⁶.
人类 T 淋巴细胞稳定表达 CARs 是一种经过临床验证的细胞疗法，用于治疗慢性和急性淋巴细胞白血病 11,12。利用受体介导的 T 细胞信号激活设计策略（图 2Ca），患者特异性 T 细胞在体外被修饰，以通过 CAR 识别癌症特异性抗原。输注到患者体内后，自体 CAR T 细胞自主迁移到靶组织，并招募内源性免疫系统以局部破坏肿瘤细胞（图 6B）。前三代 CAR 设计集中于开发一种最佳受体结构，该结构能够在与单一表面抗原结合时有效触发 T 细胞信号，并通过基因编辑删除内源性 T 细胞受体以防止移植物抗宿主反应 169（图 1A，步骤 2）。 当前 CAR T 细胞的发展集中在通过采用合成逻辑门原理来提高肿瘤特异性 170（框 1），使 T 细胞激活与其他用户定义的基因表达程序 171 同步，包括共同产生细胞因子 171 和离子通道 172，以提高肿瘤细胞杀伤效力，协同表达趋化因子受体以促进向实体肿瘤的迁移 173，以及触发诱导的 CAR 表达 170，以提高靶向活性比率。具有 AND 逻辑门的 CAR T 细胞的实现基于通过药物诱导的蛋白质二聚化 174 重构 CAR 依赖的信号转导，将双特异性抗体纳入 CAR 框架 175 或通过激活另一个细胞表面受体 170 触发 CAR 转录。通过使用抑制 T 细胞的 CAR 变体作为 NOT 反转器（框 1），也开发了遵循 A AND NOT B 逻辑的 CAR T 细胞（A；肿瘤表面标记；B：健康组织上的相似表面标记），以增加肿瘤特异性 176。

Designer cell implants 设计师细胞植入物

CAR T cell technology has already validated the use of engineered human cells to detect and treat cancer. However, cell-based therapies can in principle be generalized and metabolic diseases are particularly well-suited for these applications. To treat endocrine and immune disorders that do not require physical contact between the therapeutic cells and the drug target, an effective approach would be to use designer implants consisting of encapsulated human cells engineered for automated sensing of systemic disease markers and production of a disease-specific therapeutic protein. Placement at a vascularized site would enable the implant to communicate constantly with the host via the bloodstream, allowing it to autonomously coordinate automated diagnosis (through measurement of different systemic disease markers as trigger signals) with treatment and/or prevention of latent diseases (through trigger-inducible gene switches controlling drug release) (Fig. 6C). For example, human cells engineered with a synthetic gene switch based on glucose-dependent calcium entry coupled to a calcium-responsive promoter driving insulin expression autonomously restored glucose and insulin homeostasis in diabetic mice through closed-loop control of glucose-sensing and insulin secretion¹⁷⁷. Similarly, closed-loop control of insulin detection and adiponectin secretion effectively targeted the latent and asymptomatic disease stage of insulin resistance and attenuated the development of obesity-induced diabetes¹⁷⁸. Additionally, a serially assembled AND gate (Box 1) allowing simultaneous profiling of TNF and interleukin-22 (IL-22) and controlling the secretion of the anti-inflammatory cytokines IL-4 and IL-10 achieved remarkable therapeutic efficacy in treating psoriasis-related skin rashes¹⁷⁹. Various closed-loop control systems acting as therapeutic biocomputers have been developed for controlling liver injuries¹⁸⁰, gouty arthritis¹⁸¹, hypertension¹⁸², diabetic ketoacidosis¹⁸³, obesity¹⁸⁴ and Graves disease¹⁸⁵ (Fig. 6C).
CAR T 细胞技术已经验证了工程化人类细胞在检测和治疗癌症方面的应用。然而，细胞基础疗法原则上可以推广，代谢疾病特别适合这些应用。为了治疗不需要治疗细胞与药物靶点之间物理接触的内分泌和免疫疾病，一种有效的方法是使用设计的植入物，这些植入物由工程化的封装人类细胞组成，能够自动感知系统性疾病标志物并生产特定疾病的治疗蛋白。将其放置在血管化部位将使植入物能够通过血液与宿主持续沟通，从而使其能够自主协调自动诊断（通过测量不同的系统性疾病标志物作为触发信号）与潜在疾病的治疗和/或预防（通过控制药物释放的触发诱导基因开关）（图 6C）。 例如，利用基于葡萄糖依赖性钙进入的合成基因开关工程化的人类细胞，结合钙响应启动子自主驱动胰岛素表达，通过对葡萄糖感知和胰岛素分泌的闭环控制，成功恢复了糖尿病小鼠的葡萄糖和胰岛素稳态。类似地，胰岛素检测和脂联素分泌的闭环控制有效针对胰岛素抵抗的潜伏和无症状疾病阶段，减缓了肥胖引起的糖尿病的发展。此外，串联组装的与门（框 1）允许同时分析肿瘤坏死因子（TNF）和白细胞介素-22（IL-22），并控制抗炎细胞因子 IL-4 和 IL-10 的分泌，在治疗银屑病相关皮疹方面取得了显著的疗效。各种作为治疗生物计算机的闭环控制系统已被开发用于控制肝损伤、痛风性关节炎、高血压、糖尿病酮症酸中毒、肥胖和格雷夫斯病。

Although closed-loop systems of CAR T cells and designer implants can operate in an autonomous manner in vivo, it would be desirable for a human (the patient or doctor) to be able to interrupt or fine-tune the autonomous therapeutic programme with user-defined control signals in the clinical context. Classical implementations of such safety switches are based on controllable inuction of cell death^186,187 or interruption of target gene transcription^178,184 using clinically licensed drugs. Future systems for remote control of therapeutic cells in vivo could involve safety-approved trigger signals such as cosmetics and food additives or smartphones¹⁸⁸.
尽管 CAR T 细胞和设计植入物的闭环系统可以在体内以自主方式运行，但在临床环境中，患者或医生能够通过用户定义的控制信号中断或微调自主治疗程序是理想的。这类安全开关的经典实现基于可控的细胞死亡诱导或使用临床许可药物中断靶基因转录。未来在体内远程控制治疗细胞的系统可能涉及安全批准的触发信号，例如化妆品和食品添加剂或智能手机。

Cancer biocomputers 癌症生物计算机

Although future advances in CAR T cell technology could allow successful targeting of any cell surface tumour antigen, the intracellular transcriptome of cancer cells represents an even more important target for early diagnosis and intervention in cancer¹⁸⁹. Oncogenic states at the transcriptional level often precede the expression of cell surface markers, and not all cancer cells effectively express tumour-specific antigens¹⁹⁰. To address this issue, a multi-input biocomputer was engineered by integrating cancer-specific miRNA markers into an LTRi-like gene switch¹⁹¹ (Fig. 2B). In this circuit, expression of pro-apoptotic regulator BAX was placed under control of the reversed tetracycline-controlled trans-activator (rtTA; reversed TetR (rTetR)–VP16) and LacI. Importantly, for all elements of the circuit, target sites for miRNAs distinctive for cancer cells were incorporated into 3ʹ untranslated mRNA regions: target sites for miRNAs that are chronically upregulated in cancer (A, B and C) were included in rtTA and LacI transcripts, whereas target sites for miRNAs that are silenced and remain low in cancer (D, E and F) were added to the BAX transcript. This establishes an (A AND B AND C) AND NOT (D AND E AND F) logic gate that produces output specifically in cancer cells, resulting in autonomous induction of apoptosis. (Fig. 6D). Similarly, various other AND logic gates based on simultaneous cancer-specific promoter activities^56,192 or protein levels⁴⁰ have been engineered and validated in different cancer cell lines in vitro. Although the concept of programming cancer cells for autonomous apoptosis might seem an ideal cancer treatment strategy, there are major technical limitations regarding safety and efficacy of delivery of such circuits into patients. Nevertheless, in pioneering work, local lentiviral delivery of an ovarian cancer-specific AND logic gate has shown remarkable treatment success in mice⁵⁶. In this circuit, simultaneous activation of two cancer-specific promoters was coupled to T cell recruitment, whereby one cancer-specific promoter regulated the immunomodulatory gene cassette with simultaneous expression of an autoinhibitory miRNA for this gene cassette, while the other promoter controlled the expression of an miRNA ‘sponge’ that was necessary to relieve the autoinhibition (clearance of inhibitory checkpoint; see also Box 1, figure part b) and completed the AND logic.
尽管未来的 CAR T 细胞技术进展可能允许成功靶向任何细胞表面肿瘤抗原，但癌细胞的细胞内转录组代表了早期诊断和干预癌症的更重要目标。转录水平的致癌状态通常先于细胞表面标记物的表达，并非所有癌细胞都有效表达肿瘤特异性抗原。为了解决这个问题，工程师通过将癌症特异性 miRNA 标记整合到 LTRi-like 基因开关中，设计了一种多输入生物计算机。在该电路中，促凋亡调节因子 BAX 的表达受到反向四环素控制转激活因子（rtTA；反向 TetR（rTetR）–VP16）和 LacI 的控制。重要的是，对于电路的所有元素，针对癌细胞特征性 miRNA 的靶位点被纳入到 3ʹ非翻译 mRNA 区域：在 rtTA 和 LacI 转录本中包含了在癌症中慢性上调的 miRNA 的靶位点（A、B 和 C），而在 BAX 转录本中添加了在癌症中被沉默且保持低水平的 miRNA 的靶位点（D、E 和 F）。 这建立了一个 (A AND B AND C) AND NOT (D AND E AND F) 逻辑门，专门在癌细胞中产生输出，导致自主诱导细胞凋亡（图 6D）。类似地，基于同时癌症特异性启动子活性或蛋白质水平的各种其他 AND 逻辑门已在不同癌细胞系中进行了工程设计和验证。尽管为癌细胞编程以实现自主凋亡的概念看似是一种理想的癌症治疗策略，但在将此类电路传递给患者的安全性和有效性方面存在重大技术限制。尽管如此，在开创性工作中，局部慢病毒递送卵巢癌特异性 AND 逻辑门在小鼠中显示出显著的治疗成功。 在该电路中，两个癌症特异性启动子的同时激活与 T 细胞招募相结合，其中一个癌症特异性启动子调控免疫调节基因盒，同时表达该基因盒的自抑制 miRNA，而另一个启动子则控制必要的 miRNA“海绵”的表达，以解除自抑制（抑制检查点的清除；另见框 1，图 b 部分），并完成了与逻辑。

Almost two decades after the first synthetic toggle switch and oscillator circuits were created in bacteria, synthetic biology now encompasses an extensive toolkit of knowledge, devices and design strategies that in principle allow the engineering and programming of essentially any cell functionality with user-defined complexity (Fig. 4) and purpose (Figs 5,6). For example, current breakthroughs in cell-based therapies taking advantage of synthetic genetic circuits already indicate that it will be possible to successfully treat hitherto intractable diseases, including cancers and diabetes, using such synthetic biology approaches. Synthetic biology clearly has the potential to make an enormous impact on the world’s health-care, agriculture and environmental systems. It will be important to promote public understanding of the technology and to discuss its ethical implications in order to ensure its general acceptance.
几乎在第一个合成开关和振荡器电路在细菌中被创造出来近二十年后，合成生物学现在涵盖了一个广泛的工具包，包括知识、设备和设计策略，这些在原则上允许工程和编程几乎任何细胞功能，具有用户定义的复杂性（图 4）和目的（图 5、6）。例如，当前在细胞基础疗法中的突破，利用合成遗传电路，已经表明有可能成功治疗迄今为止难以治疗的疾病，包括癌症和糖尿病，使用这样的合成生物学方法。合成生物学显然有潜力对全球的医疗保健、农业和环境系统产生巨大影响。促进公众对该技术的理解并讨论其伦理影响将是确保其广泛接受的重要举措。