A dual molecular analogue tuner for dissecting protein function in mammalian cells

Loss-of-function studies are fundamental for dissecting gene function. Yet, methods to rapidly and effectively perturb genes in mammalian cells, and particularly in stem cells, are scarce. Here we present a system for simultaneous conditional regulation of two different proteins in the same mammalian cell. This system harnesses the plant auxin and jasmonate hormone-induced degradation pathways, and is deliverable with only two lentiviral vectors. It combines RNAi-mediated silencing of two endogenous proteins with the expression of two exogenous proteins whose degradation is induced by external ligands in a rapid, reversible, titratable and independent manner. By engineering molecular tuners for NANOG, CHK1, p53 and NOTCH1 in mammalian stem cells, we have validated the applicability of the system and demonstrated its potential to unravel complex biological processes.


B
iologists are increasingly adopting holistic approaches, such as systems biology, to understand life's complexity. Nevertheless, reductionism still remains a primary driving force for scientific progress. Elucidating gene function underlies most biological discoveries and is frequently achieved using loss-of-function analyses. Yet, for mammalian cells in general, and even more so for mammalian stem cells, the biologist's toolbox is limited and primarily includes laborious genomic editing 1 , a limited set of often-nonspecific chemical inhibitors and RNA interference (RNAi). Recently developed tools augment experimental flexibility and accuracy 2,3 , but are still limited in applicability, reversibility, titratability, rapidity and multiplicity (Supplementary Table 1). Thus, simple tools for rapid and multiple gene perturbation will facilitate the elucidation of gene functions and molecular networks.
Manipulation of protein levels represents a relatively new loss-of-function approach. To this end, harnessing the plant hormone-induced degradation pathways is particularly attractive due to their efficiency and specificity. The plant hormones auxin (indole-3-acetic acid, IAA) and jasmonate-isoleucine (jasmonic acid-Ile, JA-Ile) bind the intracellular F-Box proteins transport inhibitor response 1 (TIR1) and coronatine insensitive 1 (COI1), respectively, and promote their association with target proteins containing specific degron motifs. TIR1 and COI1, via their F-box domains, assemble into the SCF (SKP1, CUL1 and F-box) E3 ubiquitin-ligase complex, which together with an E2 ubiquitin-conjugating enzyme, catalyses the polyubiquitination and subsequent proteasomal degradation of degron-containing proteins [4][5][6][7][8][9] . Auxin-bound TIR1 targets proteins containing auxin-induced degradation (AID) degrons, while JA-Ile-bound COI1 targets proteins containing JAZ degrons (Fig. 1a). Nishimura et al. 10 developed a system enabling conditional protein regulation by adapting the auxin-induced degradation pathway to non-plant cells. They reported that ectopic TIR1 can mediate auxin-dependent degradation of AID-fused proteins and demonstrated the system's feasibility with a simple plasmid (pAID) harbouring a cytomegalovirus promoter-driven polycistronic mRNA encoding TIR1 and a plant protein carrying the AID degron. Fusing a protein-of-interest (POI) to the degron enabled the degradation of the POI following auxin treatment 10 . Despite its simplicity, pAID has major limitations in terms of applicability to mammalian cells. These include a viral promoter prone to silencing in embryonic stem cells (ESCs) 11,12 , a lack of a designated selectable marker, the inability to suppress endogenous genes and a large degron (228 AAs) liable to interfere with the POI's function. For these and other reasons (Supplementary Table 1), this technology has been primarily applied to yeast, where endogenous genes are easily disrupted and pAID-carrying clones are readily isolated. Of note, although auxin-dependent degradation was previously used to study mammalian cells, its implementation required multiple consecutive genetic manipulations and was mainly confined to cancer cell lines [13][14][15][16] . In recent times, auxin-dependent degradation was also harnessed in vivo to study Caenorhabditis elegans 17 .
Mammalian ESCs have gained much interest as a model for developmental biology and a therapeutic avenue. ESCs are unique in their unlimited self-renewal and pluripotency, a state maintained by a transcription factor network revolving around SOX2, OCT4 (POU5F1) and NANOG 18 . Combining loss-offunction and genetic complementation (rescue) strategies, we broadened and characterized the ESC self-renewal network [18][19][20][21][22] . Nevertheless, we sought to develop an improved experimental system that upgrades the stem cell biologist's toolbox and facilitates faster, tighter and combinatorial dissection of gene and protein function. Here we report a mammalian dual-protein rescue system that harnesses the auxin and JA-Ile pathways, and is specifically tailored to ESCs. For each hormone, we engineered a lentiviral vector harbouring a short hairpin RNA (shRNA), a hormone receptor, a short degron and a selectable marker. Using a two-step cloning protocol, each vector is easily modified to contain the desired shRNA and degron-fused POI, which enables silencing of a gene-of-interest and its replacement by a POI whose degradation is induced by the appropriate hormone. The combination of these two vectors offers simultaneous control over two proteins in the same cell. By applying this system to study key ESC decision-making proteins, such as NANOG, CHK1, p53 and NOTCH1, we have demonstrated the system's potential to facilitate experimental designs that were previously unfeasible or overcomplicated.
Results pRAIDRS functions as an auxin-induced degradation rescue system. We aimed at designing a vector that enables depletion of an endogenous gene-of-interest and its replacement by an exogenous POI whose degradation is induced by auxin. This approach represents a genetic complementation (rescue) system, in which a phenotype exerted by silencing a gene-of-interest is conditionally reversed by exogenous expression of that gene product. To this end, we engineered pRAIDRS (RNAi and auxininduced degradation rescue system), a lentiviral vector containing all elements for construction of an auxin-regulated rescue system. As depicted in Fig. 1b, a U6 promoter drives the expression of an shRNA that silences an endogenous gene-of-interest. A second promoter, either phosphoglycerate kinase-1 (pPGK-1) or the stronger elongation factor 1a (pEF1a) (Supplementary Fig. 1a), followed by a Kozak sequence, drives the expression of an mRNA encoding three in-frame proteins separated by two porcine teschovirus-1 2A (P2A) peptides. The first protein is a codonoptimized Oryza sativa (rice) TIR1 auxin receptor (OsTIR1). The second component is a shortened AID degron derived from Arabidopsis thaliana IAA17 (AtIAA17), which can be fused to either terminus of the POI. The last component is a selectable marker, either puromycin N-acetyl-transferase (PuroR) or blasticidin-S deaminase (BSD), conferring puromycin or blasticidin resistance, respectively. Mammalian cells transduced with pRAIDRS express OsTIR1, which associates with SKP1 and forms a functional SCF TIR1 complex 10 . Following auxin treatment, SCF TIR1 mediates degron polyubiquitination, leading to degradation of the POI (Fig. 1a). The full-length AtIAA17, originally used in pAID (ref 10), is imperfect as a degron due to its large size (228 AAs), its propensity to confer nuclear localization 23 and other potentially undesirable activities it possesses as a plant transcription factor. Therefore, we mapped the minimal required AID degron to a 47-AA region (AID 47 ) spanning AtIAA17 residues 63-109 ( Supplementary  Fig. 1b,c), which mostly overlaps with a previously reported shortened AID degron 24 . Notably, we observed that in pRAIDRStransfected HEK-293T cells, green fluorescent protein (GFP) is spontaneously cleaved from the full-length AtIAA17 degron (AID 228 ), but not from AID 47 ( Supplementary Fig. 1c,d), suggesting that a shorter degron might also be more cleavage resistant. However, as other labs who have used AID 228 did not report spontaneous cleavage, this phenomenon might be specific to our cell lines, POI or vector architecture. We next compared the degradation of cytoplasmic and nuclear POIs by analysing the effect of a nuclear localization signal (NLS) on the degradation of GFP-AID 47 , and found both highly effective, but NLS-GFP-AID 47 degradation faster ( Supplementary Fig. 1e,f).

pRAIDRS enables rapid and titratable conditional regulation.
To demonstrate the applicability of pRAIDRS as a rescue system in mammalian stem cells, we engineered mouse ESCs (mESCs) in which the protein level of NANOG is controlled by auxin. We infected mESCs with pRAIDRS harbouring an shRNA targeting the 3 0 -untranslated region (3 0 -UTR) of Nanog mRNA and an AID 47 -fused Nanog coding sequence (A-NANOG) lacking UTRs. As a control, mESCs were infected with pRAIDRS containing only GFP-AID 47 (GFP-A). Post-selection clones demonstrated effective silencing of endogenous NANOG by the shRNA, whereas exogenous A-NANOG, which was expressed at levels comparable to endogenous NANOG in control cells, was effectively and rapidly depleted following auxin treatment ( Fig. 2a and Supplementary Fig. 2a). Phenotypically, auxin treatment of A-NANOG mESCs, but not GFP-A mESCs, resulted in depletion of alkaline phosphatase (AP) positive colonies, loss of ESC morphology and a transcriptional programme characteristic of NANOG inactivation 22 , namely downregulation of self-renewal genes and induction of endodermal differentiation markers ( Fig. 2b-d and Supplementary Fig. 2b). A similar transcriptional response was elicited by shRNA-mediated NANOG depletion ( Supplementary Fig. 2c,d). In contrast, mESCs infected with pRAIDRS harbouring a Nanog shRNA and a Nanog coding sequence fused to an irrelevant degron (OsJAZ 33 , see below) did not respond to auxin treatment ( Supplementary Fig. 2e,f). These results demonstrate the applicability of pRAIDRS as a molecular switch that facilitates dissection of protein function in mESCs.
To exemplify the rapidity of degradation enabled by pRAIDRS, we established a rescue system for the checkpoint kinase CHK1 in mESCs. CHK1 is required for mouse development and its disruption severely impairs DNA damage responses 25,26 . Multiple roles are also attributed to CHK1 in normal cell cycle progression 27,28 and in mESC self-renewal 20 . We infected mESCs with pRAIDRS harbouring a Chk1 3 0 -UTR-targeting shRNA and an AID 47 -fused Chk1 coding sequence (A-CHK1). A western blot analysis of selected clones demonstrated efficient silencing of endogenous CHK1 and complete auxin-dependent degradation of A-CHK1 (Fig. 3a). Next, A-CHK1 cells were monitored for the effects of CHK1 depletion. When cells were infected and selected in the presence or absence of auxin, a marked auxin-dependent depletion of AP-positive colonies was observed ( Supplementary Fig. 3a), apparently supporting the reported roles of CHK1 in mESC self-renewal. However, CHK1 depletion in post-selection cells had only a marginal effect, if any, on proliferation rate, stage specific embryonic antigen-1 (SSEA-1) levels, mRNA expression patterns or apoptosis ( Supplementary  Figs 3 and 4). These data imply that the initial effect of CHK1 depletion may reflect its role during cellular stress responses induced by viral infection or drug selection.
We then used pRAIDRS to study the role of CHK1 in the mESC DNA damage response. To this end, cells were treated with aphidicolin, a DNA polymerase inhibitor that induces DNA breaks and activates the ATR-CHK1 pathway 29 . CHK1 depletion dramatically sensitized mESCs to aphidicolin, as auxin-treated A-CHK1 cells died following treatment with 0.1 mM aphidicolin, whereas control cells survived following treatment with 100-fold higher concentrations of aphidicolin (Fig. 3b). This hypersensitivity was specific to CHK1 depletion as auxinand control-treated GFP-A cells responded indistinguishably to aphidicolin treatment ( Supplementary Fig. 4a,b). CHK1 depletion in aphidicolin-treated cells resulted in rapid induction of apoptosis, activation of a p53 (TRP53) transcriptional response, predominantly of the p53 target Fas that encodes a death receptor 30 , as well as a later induction of differentiation (Supplementary Fig. 4c-f). We hypothesized that the aphidicolin susceptibility of CHK1-depleted cells stems from the ability of CHK1 to phosphorylate and induce the cytoplasmic sequestration or degradation of CDC25 phosphatases, which, in turn, augments the inhibitory Tyr15 phosphorylation of CDK1 (CDK1 pY15 ), preventing cell cycle progression 31 . Indeed, rapid (20 min) auxin-dependent depletion of CHK1 in aphidicolintreated mESCs resulted in synchronous mitotic entry 45-90 min post-auxin treatment, parallelling CDC25A stabilization and the decrease in CDK1 pY15 , and preceding p53 stabilization and the induction of Fas mRNA (Fig. 3c-f and Supplementary Fig. 4g). Thus, depleting CHK1 in DNA-damaged mESCs led to a series of consecutive phenotypes already observable 45 min post treatment. Moreover, by titrating down CHK1 levels in DNA-damaged mESCs, we demonstrated pRAIDRS applicability as a sensitive analogue tuner that enables fine-tuning of protein levels and their associated phenotypes ( Supplementary Fig. 5), facilitating in-depth analyses of protein dose responses.
Auxin-induced degradation was shown to be reversible 10 . To demonstrate this for pRAIDRS, we engineered p53-null lung adenocarcinoma cells (NCI-H1299) expressing an auxindegradable wild-type p53-AID 47 (p53-A). These cells were infected and cultured in the presence of auxin to prevent the stabilization of p53, known for its ability to inhibit cell growth 32,33 . However, following auxin removal p53 was rapidly stabilized, leading to the induction of the p53 target genes p21 (CDKN1A) and MDM2, and resulting in growth retardation (Fig. 4). In sum, these data validate and exemplify pRAIDRS as an easy-to-use single-vector system enabling the construction of highly rapid, titratable, reversible and non-stressful molecular tuners in mESCs and other cell types.
pJAZ functions as a coronatine-induced degradation rescue system. Simultaneous conditional regulation of two proteins represents a powerful tool for complex analyses. We therefore sought to engineer a second rescue system that harnesses the plant jasmonate-induced degradation response. As described above, in plants, isoleucine-conjugated jasmonate (JA-Ile) mediates the binding of the F-box hormone receptor COI1 and the JAZ degron domain of target proteins, which are consequently ubiquitinated and degraded 9,30 (Fig. 1a). We speculated that expression of COI1 in mammalian cells would enable hormone-dependent degradation of JAZ-fused POIs. As mammalian cells lack the pathway for JA-Ile conjugation, we used coronatine, a bacterial analogue of JA-Ile 34 . Using the same architecture as pRAIDRS (Fig. 1b), we constructed pJAZ, a vector harbouring a codonoptimized A. thaliana COI1 receptor (AtCOI1) and a 23-AA JAZ degron (AtJAZ 23 , Supplementary Fig. 6a) that we have previously identified as the A. thaliana JAZ1 minimal degron motif 5 .
For initial testing, we infected HEK-293T cells with pJAZ harbouring GFP-AtJAZ 23 and treated them with coronatine. Disappointingly, GFP degradation was extremely ineffective ( Fig. 5a, version 1). We then systematically and iteratively optimized pJAZ by testing different COI1 orthologues and fusion proteins, and by altering the degron length and origin ( Fig. 5a and Supplementary Fig. 6). We hypothesized that the lack of coronatine-dependent degradation stems from insufficient binding of AtCOI1 to human SKP1 (HsSKP1). We therefore generated an OsTIR1 F-box -AtCOI1 LRR chimera composed of OsTIR1 F-box domain (AA 1-39) 4 , which binds HsSKP1 effectively 10 , and AtCOI1 leucine-rich repeat (AA 52-592), the receptor region responsible for hormone and degron binding 5 .   Fig. 6e) and, hence, presumably sufficient. Unexpectedly, the HA tag boosted pJAZ efficiency to B70% (version 2 HA ), possibly by stabilizing the receptor 24 . We next reasoned that at 37°C, a rice coronatine receptor (OsCOI1) might function better than AtCOI1, as reported for the auxin receptor 10 .
Of the three OsCOI1 paralogues, we chose OsCOI1B, as it binds a larger variety of JAZ proteins 35 , and tested it with either the AtJAZ 23 degron or with a 23-AA rice degron, OsJAZ 23 ( Supplementary Fig. 6a). We found both versions (5-At23 and 5-Os23, respectively) nonfunctional. However, a chimeric receptor (OsTIR1 F-box -OsCOI1B LRR ) comprising OsTIR1 F-box domain and OsCOI1B LRR (version 6-Os23) mediated nearly 90% degradation of GFP-OsJAZ 23 . Nevertheless, this version probably suffered from coronatine-independent degradation, as most cells had low fluorescence levels ( Supplementary Fig. 6f). Switching to AtJAZ 23 or extending the rice degron to 33 AAs (OsJAZ 33 ) restored GFP levels, but attenuated the effect of coronatine (versions 6-At23 and 6-Os33, respectively). Notably, using AtJAZ FL resulted in high GFP expression and 95% coronatine-induced degradation (version 6-AtFL), while conferring nuclear localization to GFP ( Supplementary Fig. 6g), in accordance with JAZ1 localization in plants 36 , prompting us to speculate that its degron efficiency partially derives from its nuclear localization. We therefore targeted GFP-OsJAZ 33 to the nucleus with an NLS (version 7) and found it to enhance both dose-and time-dependent coronatine-induced degradation, reaching 495% with 50 mM coronatine ( Supplementary Fig. 6h). Thus, a chimeric OsTIR1 F-box -OsCOI1B LRR receptor can effectively mediate coronatine-dependent degradation of nuclear POIs fused to an OsJAZ 33 degron without evidence of coronatineindependent degradation, coronatine receptor-independent degradation or coronatine toxicity ( Supplementary Fig. 6i-m). Importantly, pJAZ version 7 (henceforth pJAZ) functioned nearly as well as pRAIDRS in mediating hormone-dependent degradation of nuclear GFP (Fig. 5b) and, similar to pRAIDRS, pJAZ enabled the engineering of a molecular switch in which an endogenous protein was replaced with a coronatine-regulated exogenous protein, as demonstrated by engineering a p53 switch in human ESCs (hESCs; Fig. 5c,d).
Next, we engineered cells expressing coronatine-degradable NLS-GFP-OsJAZ 33 and auxin-degradable NLS-mOrange-AID 47 using pJAZ and pRAIDRS harbouring PuroR or BSD, respectively, and selecting these cells with puromycin and blasticidin. Flow     cytometric and microscopic analyses demonstrated that pRAIDRS and pJAZ function effectively and independently in a variety of cell types, including hESCs ( Fig. 5e-g), P19 mouse embryonal carcinoma cells, H1299 lung adenocarcinoma cells, HEK-293T cells, NIH/3T3 mouse embryonic fibroblasts, NCI-H358 human non-small cell lung cancer cells and HCT-116 human colorectal carcinoma cells (Supplementary Fig. 7). Importantly, both hormones induced 90-99% degradation, depending on the cell type, and did not show any cross-reactivity or interference, suggesting that neither system saturates the shared ubiquitination machinery. These data validate the applicability of pRAIDRS and pJAZ as a dual analogue molecular tuner.
A dual molecular switch to dissect the NOTCH1 pathway. NOTCH signalling, which is inactive in undifferentiated hESCs, participates in their differentiation into embryonic lineages 37,38 .
In mice, NOTCH was also implicated in trophectoderm formation 39 . Canonical NOTCH signalling involves ligand binding to the membrane receptor, leading to cleavage of the NOTCH intracellular domain (NICD) and its translocation to the nucleus, where it binds CSL (RBPJ) and MAML1 to activate gene transcription 40 . We sought to construct a molecular switch to dissect NOTCH1 signalling in hESCs. We infected hESCs with pRAIDRS NICD-A, which harbours an shRNA targeting the full-length NOTCH1 receptor and an NICD-AID 47 CDS ( Supplementary Fig. 8a,b). These cells were maintained with auxin to prevent NICD-AID 47 accumulation, which occurs quickly following auxin removal (Fig. 6a) and induces robust differentiation ( Supplementary Fig. 8c,d). We then infected these cells and their pRAIDRS GFP-A control counterparts with pJAZ harbouring a dominant-negative MAML1 (ref. 38) fused to NLS-GFP and OsJAZ 33 (dnMAML1-NLS-GFP-OsJAZ 33 , abbreviated as dnM1-GFP-J), or with pJAZ NLS-GFP-OsJAZ 33 (GFP-J) as a control. Coronatine treatment effectively induced degradation of dnM1-GFP-J (Fig. 6b).
We analysed the effect of NICD-AID 47 accumulation following auxin removal in a self-renewal condition in the presence of fibroblast growth factor 2 (FGF2) and transforming growth factor-b (TGFb) or in their absence (differentiation condition). As depicted in Fig. 6c and Supplementary Fig. 8e, in pRAIDRS NICD-A hESCs, auxin removal led to the activation of the NOTCH targets HEY1 and HES5 in a manner largely independent of FGF2/TGFb. However, the mesoderm marker T (Brachyury) and the ectoderm marker SOX1 were induced by NICD-A exclusively in the presence of FGF2/TGFb, whereas the endoderm marker GATA6 and the trophectoderm marker GATA3 were induced by NICD-A primarily in the absence of FGF2/TGFb. In nearly all cases, dnM1-GFP-J hindered NICD-A-dependent transactivation and coronatine treatment attenuated the effect of dnM1-GFP-J, restoring gene expression. Moreover, NANOG downregulation following FGF2/TGFb withdrawal was also NICD dependent. Taken together, these data indicate that canonical NOTCH1 signalling can induce key lineage commitment transcription factors in hESCs, and that the identity of these factors depends on FGF2/TGFb, unveiling a cross-talk between NOTCH1 signalling and the self-renewal circuitry. In addition, the induction of GATA3 implicates NOTCH1 in hESC trophectodermal differentiation. These data exemplify the applicability of pRAIDRS and pJAZ for the construction of dual molecular tuners capable of accurate dissection of signalling pathways in hESCs.

Discussion
We report a molecular system that facilitates experiments that were previously unfeasible or very complicated in mammalian cells in general and ESCs in particular. Both pRAIDRS and pJAZ are easy-to-construct single vectors ( Fig. 1 and Supplementary  Fig. 10), which deliver all the necessary elements for the construction of rapid and reversible analogue molecular tuners or, when combined, a dual tuner.
The iterative engineering of pRAIDRS and pJAZ was aimed at enhancing their functionality in ESCs. A 'hormone receptor/P2A/ degron-fused POI/P2A/selectable marker' cassette that was codon optimized for human cells is transcriptionally driven by a PGK-1 or EF1a promoter. These promoters offer strong and stable expression in a wide variety of cells, with pPGK-1 being more stable in ESCs and pEF1a stronger 11,12,41 . The P2A peptides separating the aforementioned components are the most effective 2A peptide in mammalian cells 42 . The AID degron was minimized fivefold, to reduce interference and spontaneous cleavage. To harness the jasmonate-induced degradation pathway, we engineered a chimeric receptor, as neither A. thaliana nor rice coronatine receptors function in mammalian cells, and identified the minimal rice JAZ degron motif compatible with this chimeric receptor. The use of selectable markers translated in-frame with the hormone receptor and POI should ensure that drug-resistant cells are hormone sensitive. Finally, the silencing of an endogenous gene-of-interest by the pU6-driven shRNA renders each lentiviral vector an independent rescue system.
A tetracycline-based complementation approach has proven useful for gene discovery and characterization in ESCs [20][21][22] . Nevertheless, its slowness and the requirement for rtTA expression limit its use. Conversely, pAID enables rapid control of proteins, but does not offer endogenous gene inactivation, uses a large bioactive degron and is inapplicable to mammalian stem cells (Supplementary Table 1). Although auxin-dependent degradation was previously harnessed to generate molecular switches in somatic mammalian cells, this was achieved by sequential and laborious steps, such as TIR1 overexpression, POIdegron overexpression and gene-of-interest knockdown/out [13][14][15] or, alternatively, by genomic targeting of AID degrons to both alleles of the endogenous gene combined with TIR1 overexpression 16 . Although these approaches were effective in constructing single molecular tuners, our system enables the engineering of a dual molecular tuner with unparalleled simplicity and quickness, and is particularly useful for studying ESCs, which are hard to otherwise manipulate genetically. Importantly, the rapidity of auxin-dependent protein depletion achieved with the pRAIDRS system (20-30 min for 495% degradation of NANOG and CHK1) is comparable with those reported by Han et al. 13  pRAIDRS and pJAZ combine the advantages of the genetic complementation and hormone-induced degradation strategies, while averting their limitations, as each vector represents a fully functional rescue system specifically tailored to mammalian stem cells and both offer rapid, reversible and titratable control of protein levels. Importantly, combining endogenous gene silencing with conditional rescue ensures high-confidence genotype-to-phenotype causal linkages. Moreover, in contrast to other conditional protein degradation/activation systems [43][44][45][46] , pRAIDRS and pJAZ degrons are extremely short, diminishing interference with POI localization and function. Other advantages of pRAIDRS and pJAZ are listed in Supplementary Table 1. Of note, although both pRAIDRS and pJAZ enable hormonedependent degradation of cytoplasmic and nuclear POIs, with both systems the degradation of nuclear POIs is faster and requires lower hormone concentrations. Other noteworthy limitations of pRAIDRS/pJAZ include the following: (1) the RNAi-mediated silencing of endogenous genes, which is not always effective; (2) the constitutively active exogenous promoter driving the expression of the POI-degron fusion, which can lead to non-physiological expression levels; and (3) the lack of splice variants representation.
As a proof-of-concept, we constructed a molecular switch for the ESC master regulator NANOG. This switch enabled conditional and nearly complete rapid depletion of NANOG, recapitulating its well-established roles in mESCs 47 . By engineering a molecular switch for CHK1, we were able to elicit a series of gene-specific phenotypes as early as 45 min following hormone treatment. This degree of rapidity can facilitate the distinction between primary and secondary events, and enables high-resolution kinetic studies. Furthermore, owing to the inert and specific nature of hormone-induced degradation, we observed only minor effects following CHK1 depletion in post-selection cells, contrasting with the current conception of the role of CHK1 in normal cycling cells 27,28,48 and in mESC selfrenewal 20 . Conversely, we demonstrated that in DNA-damaged mESCs CHK1 plays a crucial protective role by restricting mitotic entry, which otherwise leads to apoptosis or differentiation. The CHK1 molecular switch represents a unique tool for screening and characterizing CHK1 inhibitors and DNA-damage sensitizers, a rapidly growing category of anti-cancer drugs 49,50 .
We also engineered cancer cells expressing hormonedegradable p53 and demonstrated its unleashing by auxin ARTICLE removal 33 , highlighting the rapid reversibility of hormoneinduced degradation. Stable ectopic expression of tumour suppressors is cumbersome, as cancer cells quickly evade their effects. However, effective auxin-induced p53 degradation enabled prolonged propagation of these cells without growth inhibition or transgene silencing. We also demonstrated how pRAIDRS and pJAZ allow titratable control of protein levels, a feature that enables studies of protein dose responses and threshold levels.
By engineering a coronatine-dependent p53 switch, we demonstrated the applicability of pJAZ for rapid and simple construction of molecular switches in hESCs. Moreover, we showed how combining pRAIDRS and pJAZ yields a dual molecular switch, where auxin and coronatine control two different proteins independently. Applying this method to hESCs, we unveiled unknown aspects of the canonical NOTCH1 pathway and its integration with the hESC self-renewal network. Thus, the generation of such dual switches (or tuners) is valuable for dissecting the function of proteins and regulatory networks.

Methods
Cell culture. HEK-293T, HCT-116 (Obtained from S.A. Aaronson's lab at ISMMS) and NIH/393 cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS, Corning), 1 mM sodium pyruvate, 2 mM L-glutamine and PenStrep (all from Gibco). NCI-H358 and NCI-H1299 cells (obtained from the American Type Culture Collection) were cultured in RPMI-1640 (Cellgro) supplemented with 10% FBS, 1 mM sodium pyruvate, 2 mM L-glutamine and PenStrep. Validated, mycoplasma-free hESCs and mESCs were obtained from the Pluripotent Stem Cell Core Facility at ISMMS. ESCs were routinely monitored for ES-like morphology and expression of Nanog and Oct4 (Pou5f1) using quantitative real-time PCR. CCE and R1 mESCs, as well as P19 mouse embryonal carcinoma cells, were cultured in DMEM supplemented with 15% FBS, 1 mM sodium pyruvate, 2 mM L-glutamine, non-essential amino acids, PenStrep, 10 nM 2-mercaptoethanol and 100 U ml À 1 LIF (ESGRO) on plates coated with 0.1% gelatin (Millipore, catalogue number ES-006-B). H9 hESCs were cultured with mTeSR TM 1 (Stem Cell Technologies) on plates coated with Matrigel (BD Biosciences, catalogue number 354234). For controlling the presence of FGF2 and TGFb, TeSR TM -E8 TM and TeSR TM -E6, which contain and lack FGF2/TGFb, respectively, were used. All cells were grown at 37°C in a humidified atmosphere of 5% CO 2 and passaged on average twice per week. All cells were tested negative for mycoplasma using the e-Myco Mycoplasma PCR Detection Kit (iNtRON). Where indicated, cell numbers were recorded each passage and population doublings were calculated as Log 2 (cell output/cell input).

Lentiviral infection and selection.
For the production of lentiviral particles, 1 Â 10 7 HEK-293T cells were resuspended in growth media (as described above) and transfected with 20 mg lentiviral vector, 20 mg psPAX2 packaging plasmid and 10 mg pMD2.G envelope plasmid using the calcium phosphate method. Cells were then plated in a 10-cm dish and cultured for 1 day. On the second day, media were replaced and cells were incubated at 32°C. Viral supernatants were collected on the morning and evening of the third and fourth days, passed through a 0.22-or 0.45-mm cellulose acetate filter and concentrated B25-fold using an Amicon Ultra-15 Centrifugal Filter (Millipore). Cells were infected with concentrated virus diluted in their appropriate media in the presence of 8 mg ml À 1 polybrene (Sigma) for B16 h at 37°C. Selection was applied 2 days following infection with either 1-2 mg ml À 1 Puromycin (Fisher Scientific) or 10-20 mg ml À 1 Blasticidin-S (Fisher Scientific). Where indicated, colonies (clones) of mESCs and hESCs with typical ESC morphology were manually isolated and expanded.
Chemicals and treatments. Auxin (IAA, Fisher Scientific, catalogue number AC12216) was dissolved in ethanol to a final concentration of 500 mM. Cells were treated with 50 mM IAA or 0.01% ethanol as a control, unless otherwise indicated. Coronatine (Sigma, catalogue number C8115) was first dissolved in dimethylsulfoxide (DMSO) to a concentration of 50 mM and then diluted in DMEM to a final concentration of 5 mM. Cells were treated with 50 mM coronatine or with 0.1% DMSO as a control, unless otherwise indicated. Aphidicolin (Fisher Scientific, catalogue number AC61197) was diluted in DMSO to a final concentration of 10 mM. Cells were treated with 1 mM aphidicolin or with 0.01% DMSO as a control, unless otherwise indicated. All trans-retinoic acid (Fisher Scientific, catalogue number 302-79-4) was dissolved in ethanol.
Staining. Stemgent's Alkaline Phosphatase staining kit (catalogue number 00-009) was used according to the manufacturer's protocol. Crystal violet (CV) staining was performed by incubating cells for 5 min with CV solution (10 mM CV, 10% ethanol in water), followed by three to five gentle washes with water. For both AP and CV staining, plates were scanned using a standard desktop scanner and images were digitally adjusted for brightness and contrast. Acetic acid was used to extract CV, which was then quantified using a spectrophotometer at 590 nm. DAPI (4 0 ,6-diamidino-2-phenylindole) staining was performed by fixing cells (plated on cover slips) with 4% paraformaldehyde in PBS for 30 min, washing twice with PBS (for 5 min), treating with 0.2% Triton X-100 and 1% BSA in PBS for 30 min, washing with PBS and incubating with 0.2 mg ml À 1 DAPI for 10 min. Cells were then washed once with PBS and mounted on microscope slides. Images acquired with a microscope were digitally adjusted for brightness and contrast. All images from the same experiment were processed identically.
Flow cytometry. Flow cytometry was performed on a BD LSRII machine. For GFP and mOrange fluorescence analysis, cells were trypsinized, neutralized with FBS-containing media, supplemented with 0.2 mg ml À 1 DAPI and kept on ice. Cells were gated on forward scatter area (FSC-A) and side scatter area (SSC-A), on FSC width (FSC-W) and FSC-A to eliminate cell aggregates, and on FSC-A and DAPI to eliminate dead cells. GFP and mOrange fluorescence intensities were detected using the fluorescein isothiocyanate (FITC) and DsRed channels, respectively. Background autofluorescence was measured using parental noninfected cells. Background-subtracted median fluorescence was normalized to the control-treated sample, to calculate relative median fluorescence. To calculate % degradation, relative median fluorescence was subtracted from 1. Quantitative real-time PCR and expression heatmaps. Total RNA was extracted using TRIZOL (Ambion) and 1-2 mg were reverse transcribed using the High Capacity Reverse Transcription Kit (Life Technologies, catalogue number 4368814) according to the manufacturer's protocol. QRT-PCR was performed in triplicates or quadruplicates using the Fast SYBR Green Master Mix (Life Technologies, catalogue number 4385612) on a LightCycler480 Real-Time PCR System (Roche). Expression was calculated using the DCt method. Relative expression was calculated by dividing the average level of each gene to that of the housekeeping gene GAPDH measured in the same cDNA sample. Gene-specific primers are listed in Supplementary Table 4. When data are displayed as bar charts, error bars represent s.d. of technical replicates. To generate gene expression heatmaps, normalized average expression levels were analysed using the Gene Cluster 3.0 software 51 . Data were log transformed and genes were mean centred. Genes were then hierarchically clustered using uncentred correlation similarity metric and average linkage.
mRNA-Seq. For testing the global transcriptional effect of coronatine treatment, H9 hESCs expressing pJAZ NLS-GFP-OsJAZ 33 and pRAIDRS NLS-mOrange-AID 47 were treated for 2 days with 50 mM coronatine (Cor) or 0.1% DMSO (Con). Experiment was repeated twice (replicates A and B). RNA was extracted with TRIZOL (Ambion). Sample preparation and sequencing was performed by Girihlet Inc. (www.girihlet.com). Briefly, total RNA was evaluated for quality and quantity using the Agilent RNA 6000 Nano Kit on an Agilent Bioanalyzer. Libraries were prepared using TruSeq RNA Library Prep Kit (Illumina). mRNA was isolated from 500 ng of total RNA using poly T beads and cDNA was synthesized using SuperScript Reverse Transcriptase (ThermoFisher Scientific) and random primers. The cDNA ends were blunted, 'A' base added and adapters ligated. A total of 15 cycles of PCR were performed to generate cDNA libraries. Libraries concentration was measured using an Agilent DNA 1000 Kit on an Agilent Bioanalyzer. Libraries were sequenced on a NextSeq 500 machine (Illumina) with 1*75 bp reads.
Data analysis. The resulting fastq files were mapped to the human genome (version hg19) using the TopHat programme (with Bowtie2). The output .bam files were processed through the Cuffquant programme to generate normalized read counts. The resulting .cxb files were processed through the Cuffdiff programme to generate fragments per kilobase of transcript per million mapped reads (FPKM) values. Raw data (fastq files), as well as FPKM values, were uploaded to the GEO database (GSE74457) and can be accessed using this link: http://www.ncbi.nlm. nih.gov/geo/query/acc.cgi?acc=GSE74457. On average, there were 7.7 Â 10 7 reads per sample, which mapped to 23,622 human genes. Lowly expressed genes with an average FPKM value o0.1 were excluded, narrowing the total gene count to 15,928. The BRB-Array Tools software 53 was used to calculate Spearman pairwise correlation between all samples ( Supplementary Fig. 5l). To identify genes that were differentially regulated following coronatine treatment ( Supplementary  Fig. 6m), we filtered the gene list to include genes that meet the following criteria: (1) genes that scored a P-value o0.05 in a two-tailed paired t-test comparing coronatine-treated samples with control samples; (2) genes that had a fold change 42 between coronatine and control samples in both replicates; and (3) coding genes and long non-protein-coding RNAs (excluding small RNAs). Using these criteria, only two genes demonstrated differential expression between coronatine and control samples. When the same criteria were applied to search for genes that were differentially regulated between the two biological replicates, seven such genes were identified.