A Molecular Glue Approach to Control the Half-Life of CRISPR-Based Technologies

Abstract

Cas9 is a programmable nuclease that has furnished transformative technologies, including base editors and transcription modulators (e.g., CRISPRi/a), but several applications of these technologies, including therapeutics, mandatorily require precision control of their half-life. For example, such control can help avert any potential immunological and adverse events in clinical trials. Current genome editing technologies to control the half-life of Cas9 are slow, have lower activity, involve fusion of large response elements (>230 amino acids), utilize expensive controllers with poor pharmacological attributes, and cannot be implemented in vivo on several CRISPR-based technologies. We report a general platform for half-life control using the molecular glue, pomalidomide, that binds to a ubiquitin ligase complex and a response-element bearing CRISPR-based technology, thereby causing the latter’s rapid ubiquitination and degradation. Using pomalidomide, we were able to control the half-life of large CRISPR-based technologies (e.g., base editors and CRISPRi) and small anti-CRISPRs that inhibit such technologies, allowing us to build the first examples of on-switch for base editors. The ability to switch on, fine-tune, and switch-off CRISPR-based technologies with pomalidomide allowed complete control over their activity, specificity, and genome editing outcome. Importantly, the miniature size of the response element and favorable pharmacological attributes of the drug pomalidomide allowed control of activity of base editor in vivo using AAV as the delivery vehicle. These studies provide methods and reagents to precisely control the dosage and half-life of CRISPR-based technologies, propelling their therapeutic development.

Graphical Abstract

INTRODUCTION

CRISPR-Cas9-based technologies, including those for base conversion (e.g., C→T using base editors) and transcription modulation (e.g., by using CRISPRi/a), are furnishing novel therapeutic modalities.^1–6 Required of all therapeutic agents, the precision control of the half-life of CRISPR-based technologies is particularly crucial, as off-target editing, undesirable chromosomal translocations, genotoxicity, and activation of oncogenic pathways^7–9 are observed at the elevated or prolonged activities of these technologies^5,10–16 and the longevity of Cas9 dictates genome editing outcome.^17,18 Finally, the preexistence of immunity against Cas9 in humans will require its half-life control for therapeutic applications.^19,20

An ideal system to control the half-life of CRISPR-based technologies would have several characteristics. First, the system should be capable of degrading both small (<10 kDa) and large (>200 kDa) CRISPR-based technologies without impairing their activity. Second, the system should be minimalistic, ideally consisting of a controller that acts via a response element fused to a CRISPR-based technology. Third, the response element should be miniature. For example, we have recently reported a minimalistic system consisting of Cas9 fused to an FKBP variant (i.e., the response element), and a controller that triggers degradation of the fusion protein.²¹ However, this system is nonideal because the FKBP fusions are large, adding ~230 amino acids, which not only reduces Cas9 activity but also significantly aggravate the viral packaging of Cas9 in Adeno-Associated Virus (AAV)-based delivery systems. Fourth, since contemporary and emergent CRISPR-based technologies already bear variable effector domains on Cas9 termini, a generalizable system should have a response element located internally on Cas9. Since our FKBP-system required termini fusion on Cas9, it did not afford control of CRISPR-based technologies, including CRISPRa/i and base editors. Fifth, the controller should be fast-acting to afford precision temporal control of the technology’s activity.²² Here, a small-molecule controller is preferable as they are fast-acting (vs genetic methods)²³ and the dose of the controller can be further used to control the half-life. However, the small molecule should be easily accessible with low off-target effects; ideally, an FDA-approved drug will allow rapid translation of the system. Finally, since CRISPR-based technologies have diverse applications, the ideal system should be efficacious in diverse settings.

Leveraging rapid advancements in molecular glue technologies that induce targeted protein degradation, we report a general platform with the aforementioned characteristics that effectively controlled the half-life of both large and small CRISPR-based technologies. This platform utilizes a miniature response element (~60 amino acids) fused internally to Cas9 and the controller is pomalidomide, an FDA-approved drug.^24,25 The response element and pomalidomide induce proximity between the CRISPR-based technology and ubiquitin ligase (Figure 1A), triggering the former’s ubiquitination and subsequent degradation by the host’s proteasome. This system can control the half-life of large technologies (e.g., Cas9, CRISPRa/i, and base editors) and a small anti-CRISPR protein (~100 amino acids) that potently inhibits CRISPR-based technologies. By controlling the inhibitory activity of anti-CRISPRs using pomalidomide, we developed an “on-switch” for the base editor. Furthermore, pomalidomide allowed dose control of activity and/or specificity of Cas9, CRISPRi, and base editors, as well as genome editing outcome. Finally, we demonstrated in vivo efficacy in mice and compatibility with AAV delivery system using a base editor to knockout Pcsk9, a therapeutic target for hypercholesterolemia and cardiovascular diseases. Given the off-target protein degradation caused by pomalidomide,²⁶ we identified pomalidomide analogs with reduced off-targets^27,28 that selectively degrade the CRISPR-based technologies. Overall, we have developed a general and modular platform for precision control of the half-life, activity, and specificity of CRISPR-based technologies in diverse cell types and in vivo.

Figure 1.

Demonstration of Cas9 degradation using superdegron derived from the short, 60-amino-acid pomalidomide-binding domain ZFP91-IKZF3. (A) Schematic showing the proteasomal degradation of Cas9 using the chimeric degron ZFP91-IKZF3 (superdegron) and pomalidomide. (B) Pomalidomide-induced, dose-dependent degradation of transiently nucleofected superdegron-Cas9 constructs was evaluated in U2OS.eGFP.PEST cells using an eGFP disruption assay. After nucleofection, cells are incubated with indicated dose of pomalidomide for 24 h. (C) Schematic showing the NanoBRET-based ternary complex formation between Halotag-cereblon (HT-CRBN), LgBiT reconstituted N-HiBiT-LSD-Cas9 and pomalidomide. (D) milliBRET ratio for the pomalidomide dose-induced ternary complex formation between LSD-Cas9 and HT-CRBN in N-HiBiT-LSD-Cas9 stably expressing U2OS cells. (E, F) Capillary electrophoretic immunoblots showing the pomalidomide-induced dose-dependent (E) and time-dependent (F) degradation of LSD-Cas9 in stably expressing U2OS cells. For the dose–response experiment, cells were incubated with pomalidomide for 12 h (1E); for the time-course experiment, 1 μM pomalidomide was used (1F). (G) Volcano plot depicting the label-free global proteomic analysis of U2OS cells (stably expressing LSD-Cas9) treated with 1 μM pomalidomide or the DMSO control.

RESULTS

Control of Half-Life of Cas9.

To identify the optimal site for degradation, we fused Cas9 with a ~60-amino-acid, pomalidomide-binding response element called the superdegron (SD),^29,30 at N-terminal, C-terminal, or loop-231 regions (NSD-, CSD-, and LSD-, respectively)^21,31 (Figure S1A). To test the activity of these fusion constructs, we used an eGFP disruption assay wherein Cas9 knocks out eGFP and the concomitant loss of eGFP fluorescence reports on Cas9 activity.³² While NSD- and LSD-Cas9 constructs showed a pomalidomide dose-dependent loss of Cas9 activity, LSD-Cas9 had the largest dynamic range with complete loss of Cas9 activity at a pomalidomide concentration as low as 100 nM (Figures 1B and S1B).

Pomalidomide induces proximity between LSD-Cas9 and cereblon (CRBN), leading to the ubiquitination and degradation of the fusion protein. To demonstrate a ternary complex between LSD-Cas9: pomalidomide: and cereblon (CRBN), we used a bioluminescence resonance energy transfer (BRET) assay^33,34 wherein LSD-Cas9 is fused to a component of split luciferase (HiBiT), while CRBN bears a HaloTag—the ternary complex in the presence of LgBiT and pomalidomide will reconstitute the luminescent activity of this split luciferase. The luminescence activity excites the BRET-acceptor dye (i.e., NanoBRET-618 ligand), which is conjugated to CRBN-HaloTag fusion, resulting in a BRET signal (Figure 1C). U2OS cells stably expressing HiBiT-LSD-Cas9 (Figure S1C) were transfected with LgBiT and HT-CRBN and treated with NanoBRET-618 ligand. We observed a pomalidomide dose-dependent increased BRET signal, suggestive of ternary complex formation between HiBiT-LSD-Cas9, pomalidomide, and HT-CRBN (Figure 1D). Since endogenous CRBN can also form the ternary complex and reduce the signal, we implemented the BRET assay in CRBN−/− HEK293T cells and observed the BRET signal even at a 1 nM concentration of pomalidomide (Figure S1D).

To complement these assays, we performed Capillary electrophoretic immunoblotting in U2OS cells stably expressing the LSD-Cas9 construct (Figure S1E) and observed significant degradation at concentrations (DC₅₀) as low as 12.7 nM (Figures 1E and S1F) and within 30 min (DT_1/2) of pomalidomide treatment (Figures 1F and S1G). To confirm that the degradation was CRBN-mediated, we examined degradation of LSD-Cas9 in CRBN−/− and CRBN+/+ HEK293T cells and measured degradation using the split luciferase assay and immunoblotting. HEK293T cells were transiently transfected with the N-HiBiT-LSD-Cas9 plasmid and treated with different doses of pomalidomide for 24 h. The lysates were subjected to immunoblotting or treated with LgBiT followed by luminescence measurement.³⁵ We observed no loss of luminescence signal from CRBN−/− cells, whereas CRBN+/+ cells yielded a pomalidomide dose-dependent decrease in the luminescence signal (Figure S1H). The immunoblotting experiments also confirmed that CRBN mediated a dose-dependent decrease in the Cas9 levels in CRBN+/+ cells (Figure S1I). To evaluate the recovery timeline post-degradation, we conducted a pulse-chase experiment to monitor the levels of LSD-Cas9 protein following the withdrawal of pomalidomide. LSD-Cas9 protein levels exhibit an initial recovery within 1 h and are fully restored between 4 and 8 h after the removal of pomalidomide (Figure S1J). We conducted label-free global proteomic analysis in U2OS cells stably expressing LSD-Cas9. Treatment with 1 μM pomalidomide for 6 h resulted in a significant downregulation of LSD-Cas9 levels (≥2-fold decrease, P ≤ 0.001; Figure 1G).

Control of Half-Life of Base Editors and CRISPRa/i.

Motivated by the efficacious degradation of Cas9 by an internal response element, we applied this system to adenine base editors (ABEs) that can correct nearly half of known pathogenic point mutations by A • T to G • C conversion.³⁶ We devised various constructs involving appending the response element to the deaminase, Cas9 nickase (nCas9), and the linker connecting the two (Figures 2A, S2A, and S2B) and tested these constructs for A • T to G • C base conversion at HBG2 gene.³⁷ We find that the construct (ABE8e-SD6) equivalent to LSD-Cas9 is most efficacious (Figure 2B) while terminal other fusions [e.g., N-term to deaminase (ABE8e-SD1) or C-term fusion to nCas9 (ABE8e-SD7)] perturbed activity or were unresponsive to pomalidomide. Our optimal construct (Figure S2C) allowed significant degradation of the adenine base editor in HEK293T stable cells at as low as 100 nM of pomalidomide (Figure 2C), with most degradation being achieved in 1 h (Figure S2D).

Figure 2.

Generalizability of superdegron tags to base editors and CRISPRa/i systems. (A) Schematic of adenine base editor (ABE8e) fused with a single superdegron tag at loop-231 of the Cas9 nickase (ABE8e-SD6). (B) Pomalidomide dose-induced base editor degradation in HEK293T cells transiently transfected with ABE8e and ABE8e-SD6 constructs. After 72 h of transfection and pomalidomide treatment, genomic DNA extracted was analyzed by NGS for the conversion of A • T to G • C. (C) Capillary electrophoretic immunoblot showing pomalidomide-induced degradation of ABE8e-SD6 in HEK293T stable cells. (D) Schematic of the LSD-dCas9-BFP-KRAB system. (E) Functional analysis of LSD-dCas9-BFP-KRAB upon pomalidomide-dependent degradation was carried out in iPSC cells by measuring the transferrin receptor (TFRC) protein levels via FACS. Briefly, iPSCs stably expressing the LSD-dCas9-BFP-KRAB construct were transfected with TFRC-targeting gRNA in the presence and absence of pomalidomide followed by measuring the transferrin receptor levels using FACS by probing them with antibody (F) Pomalidomide dose-induced dCas9 degradation in LSD-dCas9-BFP-KRAB stably expressing human iPSCs were monitored by Capillary electrophoretic immunoblotting. (G) Schematic of LSD-dCas9-VPR system. (H) HEK293T stable cells encoded with a minimally expressed nanoluciferase (NLuc) gene under the cAMP-response Element (CRE) promoter were nucleofected with LSD-dCas9-VPR and gRNA-targeting CRE results in the expression of the luciferase gene. The addition of pomalidomide resulted in a dose-dependent decrease in the luminescence levels as indicative of degradation of the LSD-dCas9-VPR construct. (I) Capillary electrophoretic immunoblot showing the pomalidomide-induced dose-dependent degradation of LSD-dCas9-VPR in stably expressing HEK293T cells.

Next, we implemented this degradation system to CRISPRi by appending the response element to a repressor construct (dCas9-BFP-KRAB, Figure 2D) at loop-231 of catalytically inactive dCas9 (i.e., equivalent site of Cas9 and base editors). To enable stable CRISPRi in induced pluripotent stem cells (iPSCs), we knocked in LSD-dCas9-BFP-KRAB into the citrate lyase β-like (CLYBL) safe harbor locus, which enables robust transgene expression in iPSCs (Figure S3A).³⁸ Once we established the stable iPSCs, we measured the dose- and time-dependent degradation of dCas9 in the iPSCs using capillary electrophoretic immunoblotting. Pomalidomide treatment at different concentrations yielded a dose-dependent degradation with the complete depletion at 100 nM pomalidomide (Figure 2F) and a very short half-life of dCas9 of less than 30 min—the dCas9 fusion was completely degraded within 1 h (Figure S3B).

We next tested the efficiency of dCas9/CRISPR gene silencing by evaluating the target protein levels in iPSCs stably expressing LSD-dCas9-BFP-KRAB. Because cell surface levels of transferrin can be easily and accurately quantified by antibody labeling and flow cytometry, we targeted the transferrin receptor (TFRC) gene, which encodes this ubiquitously expressed receptor responsible for iron uptake.³⁹ We transfected the gRNA for TFRC upstream in the promoter region and observed a severe depletion of the transferrin receptor protein. Pomalidomide treatment induced rapid degradation of LSD-dCas9-BFP-KRAB protein levels to restore the TFRC expression in iPSCs (Figure 2E). We next appended the response element to a potent and strong repressor system, dCas9-KRAB-MeCP2. We transfected the HEK293T stable cells that are stably expressing WT- or LSD-dCas9-KRAB-MeCP2 cells (Figure S3C) with gRNAs targeting upstream of the C-X-C Motif Chemokine Receptor 4 (CXCR4) gene. We observed a dose-dependent increase in the expression of the CXCR4 gene, indicating the degradation of the dCas9-KRAB-MeCP2 system (Figure S3D).

We next adapted this degradation system for CRISPR activation (CRISPRa) by integrating the response element into dCas9-VPR at loop-231 (Figure 2G) and assessed degradation efficacy using a luciferase-based CRISPRa assay. Here, a stable cell line with a cAMP-response element (CRE) promoter-driven expression of nanoluciferase (NLuc) was established (Figure 2H). The nucleofection of plasmids encoding LSD-dCas9-VPR, and a 7x-NGG gRNA targeting the CRE promoter, resulted in the NLuc expression (Figure S3E). Pomalidomide exposure resulted in a dose-dependent reduction in luminescence (Figure 2H), indicating the degradation of the CRISPRa construct (Figure S3F), which was confirmed by Capillary electrophoretic immunoblotting with effective degradation observed at concentrations as low as 100 nM (Figure 2I). Notably, the dCas9 fusion protein exhibited a rapid turnover, with a half-life of less than 30 min, and was completely degraded within 1 h of pomalidomide exposure (Figure S3G). These results demonstrate the robustness and versatility of our plug-and-play degradation system across diverse CRISPR technologies, including base editors and CRISPRa/i, in various cell types, including induced pluripotent stem cells (iPSCs).

Control of Half-Life of Anti-CRISPRs to Switch-on Base Editors.

Controlled degradation of anti-CRISPR proteins that inhibit Cas9 can furnish a switch-on system for base editors, allowing precise dose and temporal control of their activity (Figure 3A). As AcrIIA4 is a potent anti-CRISPR, we fused the superdegron to the N- and C-terminus of AcrIIA4 (NSD- and CSD-) and tested their inhibitory activity in the eGFP disruption assay (Figure S4A, S4B), which pointed to CSD-AcrIIA4 as the best construct (Figure 3B). Use of a self-splicing linker connecting Cas9 to anti-CRISPR further enhanced the dynamic range, and we observed a dose-dependent control of Cas9 activity by degrading the anti-CRISPR (Figure 3B). Using the HiBiT knock-in assay in CRBN−/− and CRBN+/+ HEK293T cells, we showed that HiBiT luminescence was enhanced in a pomalidomide dose-dependent manner in CRBN+/+ cells but not in CRBN−/− cells, indicating the activation of Cas9 upon AcrIIA4 degradation (Figure S4C). We generated a CSD-AcrIIA4-P2A-eGFP stable cell line to precisely measure the degradation dose and kinetics (Figure S4D). Immunoblot analysis of dose- and time-dependent degradation of CSD-AcrIIA4 in this line revealed that CSD-AcrIIA4 vanished at a pomalidomide concentration (D_max) as low as 100 nM (Figures 3C and S4E) and within ~25 min (DT_max) (Figure 3D, Figure S4F).

Figure 3.

Demonstration of AcrIIA4 degradation-mediated switch-on system for Cas9, base editor activation. (A) Schematic showing the proteasomal degradation of AcrIIA4 using the chimeric degron ZFP91-IKZF3 (superdegron) and pomalidomide leads to activation of Cas9. (B) The Cas9-P2A-CSD-AcrIIA4 fusion was investigated for pomalidomide-induced degradation using the eGFP disruption assay. (C, D) Immunoblots for pomalidomide-induced dose-dependent (C) and time-dependent (D) degradation of CSD-AcrIIA4 in HEK293FT cells that stably express 3x-FLG-CSD-AcrIIA4. (E, F) Pomalidomide dose-dependent degradation of CSD-AcrIIA4 activates the adenine base editor (ABE8e) (E) and the cytosine base editor (CBEmax) (F) were measured by % conversion of A.T to G.C and % conversion of C.G to T.A base pair, respectively, by NGS.

Following the development of this robust and rapid anti-CRISPR degradation system, we endeavored to build an on-switch for the adenine base editor (ABE8e) and cytosine base editor (CBEmax). Here, HEK293T cells were transfected with ABE8e (with VEGFA-targeting gRNA) or CBEmax (with HEK2 targeting gRNA) along with CSD-AcrIIA4 plasmids and treated with different concentrations of pomalidomide. We measured the editing efficiency of A•T to G•C conversion (for ABE8e) or C • G to T • A (for CBEmax) conversion by next-generation sequencing. Both conversions indicate dose-dependent activation of adenine (Figure 3E) and cytosine base editor (Figure 3F) by pomalidomide. Overall, these studies offer a previously unavailable “on-switch” for base editors by controlling the levels of anti-CRISPR protein.

Modulating DNA Repair Outcome by Controlling Cas9 Half-Life.

Following Cas9-induced double-strand break, the major repair pathways are nonhomologous end joining (NHEJ) and microhomology-mediated end joining (MMEJ). NHEJ causes a smaller indel (1–4 nt deletions and insertions), while MMEJ causes deletions of variable sizes, so moderating the ratio between the repair pathways can offer control over the output genotype.⁴⁰ To examine the effects of the half-life of Cas9 on the DNA repair outcomes, we assessed repair outcomes at a set of 48 target sites with corresponding paired gRNA that were genomically integrated into U2OS cells; this target-site library is referred to as “Reduced Library”.⁴¹ The U2OS cells stably expressing the Reduced Library were transfected with a plasmid encoding LSD-Cas9, and following blasticidin selection, the cells were treated with pomalidomide at 6, 12, 24, and 48 h, and genomic DNA was collected at 120 h. Non-MH deletions arising from the NHEJ pathway were favored over MH deletions arising presumably from the MMEJ pathway (Figure 4A). These results are consistent with the idea that the NHEJ pathway responds quickly to Cas9-mediated double-strand break and MH deletion products accumulate with a prolonged Cas9 half-life.²¹ In addition to the changes in the relative frequency of MH deletion outcomes versus non-MH deletion outcomes, the observed frequency of 1 bp insertion products increased with prolonged LSD-Cas9 exposure (Figure 4A). Overall, these studies confirm that our system allows for modulation of repair product outcomes by restricting the half-life of Cas9 in cells.

Figure 4.

Cas9 half-life can impact DNA repair outcome. (A) U2OS cell line stably expressing the Reduced Library of 48 target sites used to test editing repair outcomes was transfected with the LSD-Cas9 plasmid and treated with 1 μM pomalidomide at different time points after transfection (0–48 h). The genomic DNA was extracted at 120 h post-transfection, and HTS sequencing was performed to analyze the +1 bp insertions, MH deletions, and non-MH deletions (p-values are computed by unpaired t test with Welch’s correction). (B) ddPCR quantification of single-nucleotide exchange at the RBM20 locus in HEK293T cells following templated DNA repair. For this, the LSD-Cas9 plasmid, RBM20 gRNA plasmid, and ssODN template were transfected in HEK293T cells and were treated with pomalidomide at different time points after transfection. Cells were harvested at 72 h post-transfection, and percentages of HDR and NHEJ in the genomic DNA were analyzed by ddPCR analysis. (C) Luminescence-based quantification of HiBiT knock-in at the GAPDH locus in HEK293T cells following templated DNA repair. LSD-Cas9 plasmid, GAPDH gRNA plasmid, and ssODN template were transfected in HEK293T cells and were treated with pomalidomide at different time points after transfection. Cells were lysed 72 h post-transfection and complemented with LgBiT protein to measure the luminescence.

Next, we evaluated the impact of modulation of the half-life of Cas9 on homology-directed repair (HDR), which is used for “knock-in” using an exogenously supplied donor DNA. Though a double-strand break can be repaired via a precise repair pathway (i.e., HDR), these products were much less frequent than the error-prone NHEJ or MMEJ. Various approaches have been reported to enhance HDR frequency, including using small-molecule inhibitors of NHEJ,^39,42 though such approaches can have severe adverse effects.⁴³ Here, we reasoned that modulating the half-life of Cas9 instead would provide better control over the relative levels of HDR and error-prone outcomes. To test this, we transfected HEK293T cells with plasmids encoding the LSD-Cas9 and RNA Binding Motif Protein 20 (RBM20) targeting gRNA along with a single-stranded oligonucleotide donor (ssODN).²¹ Pomalidomide was added at the indicated time points (0, 3, 6, 12, 24, and 48 h post-transfection), and the HDR and error-prone repair frequencies were investigated using droplet digital PCR (ddPCR).⁴⁴ We observed that the error-prone repair frequency increased with an increasing Cas9 half-life, while the HDR frequency saturated at about 24 h (Figure 4B).

We further evaluated the modulation of Cas9 half-life in the levels of HDR after the knock-in of a long ssODN.²¹ Here, HEK293T cells transfected with the LSD-Cas9 plasmid, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) targeting the gRNA plasmid, and the HiBiT ssODN were treated with pomalidomide at the indicated time points (0, 3, 6, 12, 24, and 48 h post-transfection). At 72 h post-transfection, cell lysates were complemented with LgBiT, and the HDR activity was quantified by HiBiT luminescence. Similar to the ddPCR results, the HiBiT HDR frequency saturated at about 24 h (Figure 4C). Our results suggest that a shortened Cas9 half-life or Cas9 inhibition offers a higher relative amount of “knock-in” product.

Shortened Cas9 Half-Life Enhances On-Target Specificity.

As the off-target activity of Cas9 often displays slower kinetics than on-target activity, we reasoned that shortening Cas9 half-life should enhance specificity.^45,46 We measured the on-versus-off-target ratio for Cas9. For each construct, cells were transfected and treated with pomalidomide at various doses or at different times, followed by extraction and sequencing of specific genomic sites to determine the on-versus off-target editing. For LSD-Cas9, we observed a dose-dependent reduction of off-target editing at the EMX1 and VEGFA loci in HEK293T cells (Figures 5A,C and S5A, S5C). We anticipated that a fast degradation of Cas9 will allow us to titer the optimal Cas9 half-life to maximize on-target editing. At different time points after transfection with LSD-Cas9 and EMX1- or VEGFA-targeting gRNA, pomalidomide was added to HEK293T cells. We observed a small enhancement on-versus off-target ratio with a shortened Cas9 half-life (Figures 5B,D and S5B, S5D).²¹

Figure 5.

Timely degradation of CRISPR-associated proteins improves the targeting specificity. (A–D) Impact of Cas9 half-life on targeting specificity was investigated in HEK293T cells. Pomalidomide dose-dependent control (A, C) of on- versus off-target activity of LSD-Cas9 targeting EMX1 (A) and VEGFA (C). Pomalidomide-induced half-life-dependent (B, D) control of on- versus off-target activity of LSD-Cas9 targeting EMX1 (B) and VEGFA (D) (p-values are computed by unpaired t test with Welch’s correction). (E) Time-dependent control on the transcriptome-wide mutations induced by the ABE8e and ABE8e-SD6 upon addition of pomalidomide.

To perform this same test for the on- and off-target specificity of our superdegron base editor system, we transfected HEK293T cells with ABE8e-SD6 and HBG2-targeting gRNA and treated the cells with pomalidomide. There was a dose-dependent reduction of Cas9-dependent both on- and off-target editing at the HBG2 loci (Figure S6A), but no major changes in the targeting specificity for on- versus off-target (Figure S6C). We anticipated that a fast degradation of ABE8e would again allow us to titer the optimal ABE lifetime to maximize the on-target editing; therefore, we tested pomalidomide treatment at different time points. Limiting the base editor activity to short lifetime decreased the editor activity (Figure S6B). However, we did in fact observe a small enhancement of the on- versus off-target ratio with a shortened ABE lifetime (Figure S6D), as limiting the ABE lifetime to within 24 h showed an increase in target specificity compared to WT ABE. These no or small changes in the Cas9-dependent on-target versus off-target actions may be due to the dosage of editor as well as fast and high deaminase (TadA) activity of ABE8e. To understand the effect of base editor’s lifetime on transcriptome-wide mutations, we performed RNaseq analysis. Introduction of the superdegron itself dramatically reduced the mutations in RNA, providing little dynamic range to allow pomalidomide-mediated degradation to have an effect on the wide transcriptome mutations (Figure 5E). However, additions of pomalidomide at lower time points (0 h) abolished the mutations presumably due to low expression of the editor (Figure 5E). Thus, for adenine base editors, the fusion of response element not only allows control of half-life but also dramatically reduces RNA off-target.

Demonstration of In Vivo Control of Base Editor’s Activity.

In a step toward in vivo control of base editor activity, we generated an AAV bearing the adenine base editor. Here, we used a split base editor/dual-AAV strategy, in which the adenine base editor is divided into an amino-terminal (1–573 aa) and carboxy-terminal half (574–1368 aa) with the response element present at loop-231 as described above and each base editor fragment bearing split intein.⁴⁷ Following co-infection by AAV9 particles expressing each base editor/split-intein half, protein splicing in trans reconstituted the full-length base editor (Figure 6A).⁴⁷ Because our degradation system requires the human CRBN component, we established the effect of pomalidomide on base editor activity in a human CRBN knock-in C57Bl6/J mice.⁴⁸ We used ABE8e-SD6 AAV9 particles targeting the Pcsk9 site (Figure 6B) in the liver.^49,50 We injected a dose of 5 × 10¹¹ AAV9 N-terminal and C-terminal ABE8e-SD6 particles (a 1:1 mixture at 2.5 × 10¹¹ vg each AAV) via retro-orbital eye injection and allowed the base editor to self-assemble and edit the genome for the first week. For the following 2 weeks, we administered pomalidomide daily at an oral dose of 30 mg/kg. We euthanized the mice at 3 weeks post viral-particle injection and harvested the liver and blood (Figure 6C). The base editing activity of ABE8e-SD6 was analyzed by next-generation sequencing after the genomic DNA was extracted from the liver. Here, the mice administered with pomalidomide showed significantly lower base editing activity compared to the control, validating the degradation of base editors in the mouse livers (Figure 6D). ELISA-based Pcsk9 levels in the plasma of pomalidomide-treated mice were also higher compared with the control mice (Figure 6E), suggesting the usability of our system to control base editor activity in vivo.

Figure 6.

Superdegron applicability and degradation demonstration of genome editing proteins in mouse models. (A) Intein reconstitution strategy uses two fragments of protein fused to halves of a split intein that splice to reconstitute a full-length protein following coexpression in host cells. (B) Schematic showing the genomic region of Pcsk9 targeted by the gRNA. (C) Schematic showing retro-orbital injection of 5 × 10¹¹ vg of AAVs consisting of split ABE-SD6 in humanized CRBN knock-in C57Bl6/J mice. AAV-injected mice were given 1 week for genome editing before pomalidomide or the vehicle control were administered orally for 2 weeks. After 3 weeks of AAV injection, mice were euthanized, and their liver and blood were harvested to analyze the base editing levels in the liver and the Pcks9 levels in plasma. (D, E) NGS-based analysis showing (D) the conversion of A • T to G • C in the livers and (E) Pcks9 levels in the blood plasma of control and pomalidomide-treated mice. (p-values are computed by unpaired t test with Welch’s correction).

DISCUSSION

Of the three elements of central dogma (Proteins, RNA, and DNA), nearly all therapeutic agents target proteins. CRISPR-based technologies provide a novel therapeutic modality by expanding the target scope to genome’s coding and noncoding regions. However, these systems do not display attributes of a typical therapeutic agent,^1,6 including precision control of dosage and half-life, and in many cases, the activity of these systems is described as “genome vandalism.” Molecular glues are a rising therapeutic modality used for precision control of the dose and half-life of proteins. These glues trigger target protein degradation by bringing together the target protein and ubiquitin ligase, consequently triggering protein ubiquitination and subsequent degradation. In this study, we leverage the increasing molecular glue technologies to bring the desired attributes of a therapeutic agent, dose-, temporal-, and half-life-control—to the realm of CRISPR-based technologies. The small footprint of the response element did not perturb the activity of CRISPR-based technology and allowed facile packaging in contemporary AAV-based delivery systems. Furthermore, the controller is FDA-approved drug pomalidomide with desirable pharmacokinetic and pharmacodynamic properties, and the response element, being of human origin, should not be immunogenic. Our system was efficacious in diverse cell types, including stem cells, and easily transportable from Cas9 to CRISPR-based technologies.

We were able to not only switch off and fine-tune the activities of CRISPR-based technologies but also build an on-switch by degrading anti-CRISPR proteins. While several on-switches have been developed for CRISPR-associated nucleases,^14,31,51,52 similar switches are missing for CRISPR-based technologies, including base editors, perhaps because of size/site limitations for appending regulatory domains on these technologies, which already have multiple fusions at the termini. Constitutively active base editors can lead to unintended genomic alterations and transcriptomic changes, which raises significant biosafety concerns in therapeutic genome editing.^53–55 We could switch-on, fine-tune, and switch-off CRISPR-based technologies using pomalidomide as a controller. This fine-tuning allowed us to precisely control on-target vs off-target editing ratio, modulate the genotypic outcome, and control base editor activity in vivo.

The existence of Cas9-reactive T_effector cells and antibodies is a major concern.^19,20 Our system can allow modulation of the efficacy versus immunogenicity and provide a “kill-switch” in the event of severe adverse reaction (e.g., cytokine storm), as was observed in the first gene therapy clinical trials. Furthermore, fine-tuning the Cas9 half-life would ameliorate the genotoxicity, as constitutively active Cas9 can cause unwanted large deletions and complex genomic rearrangements in edited cells, which could have pathogenic consequences, especially in mitotically active cells.^11,56–58 Following Cas9-induced double-strand breakage, the nature of DNA repair pathways determines the gene-editing outcome. Since Cas9 remains bound to the double-strand break and potentially impacts the recruitment of DNA repair machinery, controlling Cas9 half-life will allow control over the nature of the genotypic outcome.^17,59 In the absence of such control, the genotypic output from a Cas9-mediated gene knockout will be highly heterogeneous and may result in the generation of several neo-epitopes, further aggravating the immunogenicity problem. Our system allows precise control of the genotypic outcome by controlling Cas9 half-life. While we have focused on Cas9 derived from Streptococcus pyogenes, several next-generation CRISPR-associated nucleases with superior attributes are emerging, and we anticipate that our system will be easily transportable to these emergent nucleases. Finally, although our in vivo study demonstrates a promising strategy to control base editor activity using a degron system and split-intein AAV delivery, several limitations remain. Our future studies will optimize split base editor constructs, the AAV delivery system, PK/PD properties and tissue distribution of compounds, and a deeper investigation of off-target effects.

Given the potential for off-target protein degradation by pomalidomide, we utilized pomalidomide analogs with reduced off-target activity, identified in our recent studies.^27,28 We screened these analogs (Figure S7A) and identified two compounds (27 and 31) (Figure S7B) that display dose-dependent degradation (Figure S7C). To assess the off-target profiles of compounds 27 and 31, we conducted global quantitative proteomic profiling in iPSCs stably expressing LSD-dCas9-BFP-KRAB. Our results show that, unlike pomalidomide, compounds 27 and 31 do not induce degradation of SALL4, a known neo-substrate of pomalidomide and a key transcription factor involved in development (Figure S8A–C).²⁶ Compounds 27 and 31 have minimal off-target effects in iPSCs (Figure S8B–C). We further demonstrated the degradation potential of compounds 27 and 31 across the different CRISPR systems tested in this study (Figure S9A–E). While we acknowledge that compounds 27 and 31 have reduced potency compared to pomalidomide, their improved specificity will be advantageous in applications requiring precise temporal control with minimal off-target effects.

CONCLUSIONS

We have developed a compact, modular, and generalizable platform for precise and reversible control of CRISPR-based technologies using pomalidomide and a minimalistic response element. This system enabled dose- and temporally controlled degradation of Cas9, base editors, CRISPRi/a effectors, and anti-CRISPR proteins, allowing on/off switching, fine-tuning, and rapid termination of genome editing activity in diverse cell types. Rapid degradation of CRISPR systems enhances editing specificity, modulates genotypic outcomes, and reduces off-target editing and genotoxicity risks. The small size of the response element allowed efficient packaging into viral vectors and enabled in vivo application. The use of optimized pomalidomide analogs reduced off-target effects, paving the way for safer and more controllable editing applications. Therefore, we anticipate that this system, enabling precise control over the dosage and half-life of CRISPR-based technologies, will propel basic research and the development of therapeutic genome editing applications.

Supplementary Material

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c06230.

Experimental methods including materials, cells, and animals employed in this study (Table S1, Figures S1–S9) and references (PDF)