CN118632936A - Deaminase-based RNA sensors - Google Patents

Abstract

Disclosed herein are RNA editing tools used in systems designed to measure RNA in vivo and manipulate specific cell types. RNA sensor systems comprising a) a single-stranded RNA (ssRNA) sensor containing a stop codon and a payload; optionally wherein the ssRNA sensor also comprises a normalizing gene; and b) an adenosine deaminase (ADAR) deaminase acting on RNA; wherein the sensor is capable of binding to an ssRNA target to form a double-stranded RNA (dsRNA) duplex, which becomes a substrate for the ADAR deaminase; wherein the substrate comprises a mismatch within the stop codon; and wherein the mismatch can be edited by the ADAR deaminase, which can effectively remove the stop codon so as to enable translation and expression of the payload. Also disclosed is a method for quantifying ribonucleic acid (RNA) levels using an RNA sensor system.

Description

Deaminase-based RNA sensors

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of priority to U.S. Provisional Patent Application Serial No. 63/267,177, filed on January 26, 2022, and U.S. Provisional Patent Application Serial No. 63/210,829, filed on June 15, 2021. The entire contents of these applications are incorporated herein by reference. STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

This invention was made with government support under Grant No. 5DP5OD024583 awarded by the National Institutes of Health (NIH). The government has certain rights in this invention.

Background Art

In recent years, the ability to edit nucleic acids in a precise and programmable manner has improved. New technologies allow such precise editing in vivo, opening up the possibility of treating patients at the genotype level. However, without genetic modification or labeling, there are no viable tools to measure and track RNA levels in vivo. Genetic modification, rather than strictly observational sensor systems that can measure changes, often requires manipulation of the genome, which can have unforeseen consequences on expression and overall cellular activity. Furthermore, manipulation at the genome level to fully integrate the sensor requires genetically modified organisms, which is an unfeasible approach in many cases.

Although recent advances have allowed the identification of many specific cell types, the ability to track and manipulate these cells is still lacking. The present disclosure relates to RNA editing tools for use in designing systems to measure RNA and manipulate specific cell types in vivo.

Summary of the invention

The present disclosure relates to RNA sensor systems. The present disclosure provides such RNA sensor systems, comprising: a) a single-stranded RNA (ssRNA) sensor containing a stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; and b) an adenosine deaminase acting on RNA (ADAR) deaminase; wherein the sensor is capable of binding to an ssRNA target to form a double-stranded RNA (dsRNA) duplex, which becomes a substrate for the ADAR deaminase; wherein the substrate comprises a mismatch within the stop codon; and wherein the mismatch can be edited by the ADAR deaminase, which editing can effectively remove the stop codon, allowing translation and expression of the payload.

The present disclosure provides RNA sensor systems, wherein the mismatch between the ssRNA sensor and the ssRNA target comprises an adenine:cytidine mismatch in the dsRNA duplex.

The present disclosure provides an RNA sensor system, wherein the mismatch between the ssRNA sensor and the ssRNA target comprises an adenine:cytidine mismatch, and wherein the ADAR deaminase edits the adenine to inosine in the mismatch of the dsRNA duplex. In one embodiment, the RNA system comprises more than one mismatch.

The sensor chain of the RNA sensor system can include a payload, wherein the payload includes a reporter protein, a transcription factor, an enzyme, a transgenic protein or a therapeutic protein. The present disclosure also provides an RNA sensor system, wherein the payload includes a fluorescent reporter. The present disclosure also provides an RNA sensor system, wherein the payload includes an EGFP reporter or a luciferase reporter. The present disclosure also provides an RNA sensor system, wherein the payload includes a caspase.

The present disclosure also provides RNA sensor systems, wherein the ADAR is endogenous or exogenous. The present disclosure also provides RNA sensor systems, wherein the ADAR is a modified ADAR.

The present disclosure also provides an RNA sensor system, wherein ADAR comprises RNA editing for programmable A to I(G) replacement (REPAIR) molecules, Cas13b-ADAR fusion molecules, Cas13d-ADAR fusion molecules, Cas7-11-ADAR fusion molecules and MS2-ADAR fusion molecules, the deaminase domain of ADAR2, full-length ADAR2 or truncated ADAR2.

The present disclosure also provides an RNA sensor system comprising a plurality of RNA sensors.

The present disclosure also provides a cellular logic system comprising a) an AND gate containing an ssRNA sensor comprising one or more payloads and a plurality of stop codons complementary to different ssRNA targets; wherein the ssRNA sensor is capable of binding to the ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase; wherein the substrate comprises a mismatch within each stop codon; and wherein the mismatch in each stop codon can be edited by the ADAR deaminase, the editing being effective to remove the stop codon, enabling translation and expression of the one or more payloads ; and/or b) or a gate comprising a plurality of independent ssRNA sensors, each of the plurality of independent ssRNA sensors comprising a payload and a stop codon complementary to one or more different RNA targets; wherein each ssRNA sensor is capable of binding to a different ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase; wherein the substrate comprises a mismatch within each stop codon; and wherein the mismatch in each stop codon can be edited by the ADAR deaminase, and the editing can effectively remove the stop codon, allowing translation and expression of one or more payloads. The present disclosure also discloses a cellular logic system comprising: a) an AND gate containing an ssRNA sensor comprising one or more payloads and a plurality of stop codons complementary to different ssRNA targets; wherein the ssRNA sensor is capable of binding to the ssRNA target to form a double-stranded RNA (dsRNA) duplex, which becomes a substrate for an ADAR deaminase; wherein the substrate comprises a mismatch within each stop codon; wherein the mismatch in each stop codon can be edited by the ADAR deaminase, and the editing can effectively remove the stop codon, so that one or more ssRNA targets can be translated and expressed; a) a plurality of payloads; or b) a gate comprising a plurality of independent ssRNA sensors comprising a payload and a stop codon complementary to one or more different RNA targets; wherein each ssRNA sensor is capable of binding to an ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase; wherein the substrate comprises a mismatch within each stop codon; and wherein the mismatch in each stop codon is editable by an ADAR deaminase, the editing being effective to remove the stop codon, enabling translation and expression of one or more payloads.

The present disclosure provides methods for detecting or quantifying ribonucleic acid (RNA) levels using an RNA sensor system, comprising a) providing a single-stranded RNA (ssRNA) sensor comprising a stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; and b) providing an adenosine deaminase (ADAR) deaminase that acts on RNA; wherein the sensor is capable of binding to an ssRNA target of interest to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for the ADAR deaminase; wherein the substrate comprises a mismatch within the stop codon; and wherein the mismatch can be edited by the ADAR deaminase, the editing being effective to remove the stop codon, enabling translation and expression of the payload.

The present disclosure also provides an RNA sensor system, wherein the mismatch comprises an adenine:cytidine mismatch in the dsRNA duplex.

The present disclosure provides methods for detecting or quantifying ribonucleic acid (RNA) levels with an RNA sensor system, wherein the mismatch comprises an adenine to a cytidine, and wherein an ADAR deaminase edits the adenine in a dsRNA duplex to an inosine. The present disclosure provides methods for detecting or quantifying ribonucleic acid (RNA) levels with an RNA sensor system, wherein the RNA sensor system comprises more than one mismatch.

The present disclosure provides methods for detecting or quantifying ribonucleic acid (RNA) levels using RNA sensor systems comprising a payload, wherein the payload is translated into a reporter protein, a transcription factor, an enzyme, a transgenic protein, or a therapeutic protein.

The present disclosure also provides RNA sensor systems, wherein the ADAR is endogenous or exogenous. In some embodiments, the ADAR is a modified ADAR. In some embodiments, the ADAR is endogenous to the cell type in which the sensor can be used.

The present disclosure also provides an RNA sensor system comprising: a) a single-stranded RNA (ssRNA) sensor comprising at least a first stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; and b) an adenosine deaminase acting on RNA (ADAR) deaminase; wherein the sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex, which becomes a substrate for the ADAR deaminase; wherein the substrate comprises a mismatch within the first stop codon; wherein the mismatch can be edited by the ADAR deaminase, and the editing can effectively remove the stop codon, allowing translation and expression of the payload.

The present disclosure also provides an RNA sensor system, wherein the single-stranded RNA sensor comprises more than one stop codon. The present disclosure also provides an RNA sensor system, wherein the single-stranded RNA sensor further comprises a second stop codon. The present disclosure also provides an RNA sensor system, wherein the single-stranded RNA sensor further comprises a third stop codon.

The present disclosure also provides an RNA sensor system, wherein the mismatch comprises CCA on the target strand and TAG/UAG on the sensor strand. The present disclosure also provides an RNA sensor system, wherein the sensor strand comprises a TAG/UAG stop codon, but is not mismatched with the CCA codon on the target strand. The present disclosure also provides an RNA sensor system, wherein the sensor strand comprises a stop codon, which can be selected from

The codons on the target strand of ACA, ACT, ACC, ACG, TCA, TCT, TCC, TCG, GCA, GCT, GCC, GCG, CCA, CCT, CCC, and CCG produce matches or mismatches.

The present disclosure also provides an RNA sensor system comprising: a) a single-stranded RNA (ssRNA) sensor containing a stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; and b) an adenosine deaminase (ADAR) deaminase acting on RNA; wherein the sensor is capable of binding to an ssRNA target to form a double-stranded RNA (dsRNA) duplex, which becomes a substrate for the ADAR deaminase; wherein the substrate comprises a stop codon that can be edited by the ADAR deaminase, and the editing can effectively remove the stop codon, allowing translation and expression of the payload. In some embodiments, the ssRNA sensor comprises a TAG/UAG stop codon. In some embodiments, the TAG/UAG stop codon forms a dsRNA duplex with the ssRNA target at a codon having the formula nCn, wherein n is any nucleotide and C is cytidine.

The present disclosure also provides an RNA sensor system as described herein, wherein the ssRNA sensor is 50 nt or longer, 100 nt or longer, 150 nt or longer, 200 nt or longer, 250 nt or longer, 300 nt or longer, or 500 nt or longer. In some embodiments, the ssRNA sensor is 51 nt. In some embodiments, the ssRNA sensor is 81 nt. In some embodiments, the ssRNA sensor is 171 nt. In some embodiments, the ssRNA sensor is 225 nt. In some embodiments, the ssRNA sensor is 279 nt. In some embodiments, the ssRNA sensor is longer than 279 nt.

The present disclosure also provides an RNA sensor system as described herein, wherein the ssRNA sensor is a circular sensor. In some embodiments, the circular sensor is a rolling circle translation sensor. In some embodiments, the circular sensor is a conventional circular sensor.

The present disclosure also provides an RNA sensor system as described herein, wherein the ssRNA sensor comprises two stop codons. In some embodiments, wherein the ssRNA sensor comprises three stop codons. In some embodiments, wherein the ssRNA sensor comprises two stop codons, only one of which is targeted by ADAR editing. In some embodiments, wherein the ssRNA sensor comprises three stop codons, only one of which is targeted by ADAR editing.

The present disclosure also provides an RNA sensor system as described herein, wherein the ssRNA sensor comprises at least one affinity binding region. In some embodiments, the ssRNA sensor comprises at least three affinity binding regions. In some embodiments, the ssRNA sensor comprises at least five affinity binding regions. In some embodiments, the ssRNA sensor comprises at least seven affinity binding regions. In some embodiments, the ssRNA sensor comprises more than seven affinity binding regions. In some embodiments, the affinity binding regions are separated by MS2 hairpin regions.

The present disclosure also provides an RNA sensor system as described herein, wherein the payload comprises a Cre recombinase. In some embodiments, the payload comprises a Cas protein. In some embodiments, wherein the payload comprises a Cas9. In some embodiments, wherein the payload comprises a transcription factor. In some embodiments, wherein the payload comprises a payload ADAR. In some embodiments, wherein the payload is a reporter of a cell stress response.

The present disclosure also provides compositions and delivery vehicles, the compositions comprising the RNA sensor systems described herein. In some embodiments, the compositions comprise an RNA sensor system and a lipid nanoparticle, wherein the RNA sensor system comprises: a) a single-stranded RNA (ssRNA) sensor containing a stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; and b) an adenosine deaminase (ADAR) deaminase acting on RNA; wherein the sensor is capable of binding to an ssRNA target to form a double-stranded RNA (dsRNA) duplex, which becomes a substrate for the ADAR deaminase; wherein the substrate comprises a stop codon that can be edited by the ADAR deaminase, the editing can effectively remove the stop codon, enabling translation and expression of the payload, and wherein the RNA sensor system is encapsulated in a lipid nanoparticle.

The present disclosure also provides methods of killing specific cells or cell types, wherein the method comprises providing a single-stranded RNA (ssRNA) sensor or guide comprising a stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; wherein the payload is a self-dimerizing caspase, and wherein the ssRNA sensor or guide is capable of binding to a ssRNA target to form a double-stranded RNA duplex that becomes a substrate for an adenosine deaminase acting on RNA (ADAR) deaminase, and wherein the ssRNA target is enriched in expression in a specific cell or cell type.

The present disclosure also provides an RNA sensor system comprising: a) an RNA sensor containing a stop codon and a payload; optionally, wherein the ssRNA sensor further comprises a normalizing gene; and b) an adenosine deaminase (ADAR) deaminase acting on RNA; wherein the sensor is capable of binding to an RNA target to form a double-stranded RNA (dsRNA) duplex region, which becomes a substrate for the ADAR deaminase; wherein the substrate comprises a mismatch within the stop codon, wherein the mismatch can be edited by the ADAR deaminase, and the editing can effectively remove the stop codon, enabling translation and expression of the payload. In some embodiments, the RNA sensor is a single-stranded RNA. In some embodiments, the RNA sensor comprises one or more double-stranded RNA (dsRNA) domains. In some embodiments, the RNA target is a single-stranded RNA. In some embodiments, wherein the RNA comprises one or more double-stranded RNA (dsRNA) domains.

The present disclosure also provides an RNA sensor system as described herein, wherein the RNA sensor is 50 nt or longer, 100 nt or longer, 150 nt or longer, 200 nt or longer, 250 nt or longer, 300 nt or longer, or 500 nt or longer. In some embodiments, the RNA sensor is 51 nt. In some embodiments, the ssRNA sensor is 81 nt. In some embodiments, the ssRNA sensor is 171 nt. In some embodiments, the ssRNA sensor is 225 nt. In some embodiments, the ssRNA sensor is 279 nt. In some embodiments, the ssRNA sensor is longer than 279 nt.

The present disclosure also provides an RNA sensor system as described herein, wherein the ssRNA sensor is a circular sensor. In some embodiments, wherein the circular sensor is a rolling circle translation sensor. In some embodiments, the circular sensor is a conventional circular sensor. In some embodiments, the RNA sensor comprises two stop codons. In some embodiments, the RNA sensor comprises three stop codons. In some embodiments, the RNA sensor comprises two stop codons, only one of which is targeted by ADAR editing. In some embodiments, the RNA sensor comprises three stop codons, only one of which is targeted by ADAR editing.

The present disclosure also provides an RNA sensor system as described herein, wherein the RNA sensor comprises at least one affinity binding region. In some embodiments, the RNA sensor comprises at least three affinity binding regions. In some embodiments, the RNA sensor comprises at least five affinity binding regions. In some embodiments, the RNA sensor comprises at least seven affinity binding regions. In some embodiments, the RNA sensor comprises more than seven affinity binding regions. In some embodiments, the affinity binding regions are separated by MS2 hairpin regions.

In some embodiments, the payload comprises a Cre recombinase. In some embodiments, the payload comprises a Cas protein. In some embodiments, the payload comprises a Cas9. In some embodiments, the payload comprises a transcription factor. In some embodiments, the payload comprises a payload ADAR. In some embodiments, the payload is a reporter of a cell stress response.

In some embodiments, the ADAR is selected from

ADAR2, ADAR1, ADAR1p150, ADAR1p110, ADAR2R455G, ADAR2R455G, ADAR2S486T, ADAR2T375G E488Q T490A, ADAR2T375G, ADAR2T375S, ADAR2N473D, ADAR2 deaminase domain, ADAR2T490S, ADAR2T490A, MCP-ADAR2 deaminase domain, ADAR2R455E, ADAR2T375G T490A, ADAR2E488Q, MCP-ADAR2 deaminase domain, E488Q T490A, ADAR2R510E, ADAR2R455S, ADAR2V351L,

and derivatives or modified variants thereof. In some embodiments, the ADAR is endogenously expressed in a target cell in which the RNA sensor system can be used.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a graphic illustration of how RNA sensors can utilize ADAR technology to provide new gene outputs. The sensor RNA contains an optional marker protein, a guide RNA region with a stop codon (red octagon), and a downstream payload protein. The sensor associates with the target RNA to form a duplex with an A-C mismatch, which serves as a substrate for RNA editing of the ADAR protein (brown). For example, RNA editing can convert a UAG stop codon to UIG, allowing translation of the payload (green protein).

Fig. 2A is a diagrammatic illustration of a dual transcription ADAR sensor design. Luciferase dual transcript ADAR sensors contain normalized proteins under constitutive expression and payload proteins under ADAR sensor control. Ratio multiple changes can be calculated by calculating normalized sensor activation (gluc/cluc), and then normalized to ratio values in the absence of the target. The target can be delivered by exogenous transfection under the control of a doxycycline inducible promoter. The sensor can recruit endogenous ADARs to sense target EGFP transcripts, or utilize exogenously delivered ADARs to sense target transcripts, which is performed with enhanced sensitivity. Fig. 2B is a diagrammatic illustration of the comparison of the multiple increase in the activation of luciferase values in HEK293FT cells transfected with eGFP plasmids or control plasmids and with a sensor chain that recognizes eGFP in the presence or absence of complementary ADARs. Fig. 2C is a diagrammatic illustration of the luciferase values of non-targeted sensors, targeted sensors or constitutive active plasmids in the presence or absence of complementary ADARs. Figure 2D is a graphic illustration of next generation sequencing results to quantify editing of the UAG stop codon in targeted sensors, non-targeted sensors, and constitutively active plasmids.

Fig. 3A is a schematic diagram showing a fluorescent ADAR sensor with a single transcript design, which contains a normalized fluorescent protein (mPCherry) constitutively expressed upstream of the ADAR sensor guide controlling a second fluorescent protein (mNeon). Due to the dual fluorescence on a single transcript, non-functional eGFP must be used in this form. Fig. 3B is a series of representative images of HEK293FT cells transfected with a targeted sensor or a non-targeted sensor and non-functional eGFP with or without a complementary ADAR. Fig. 3C is a quantitative increase in the multiple of mNeon activation by measuring fluorescence. Fig. 3D is a graphic illustration of the next generation sequencing results of quantitative UAG stop codon editing in the presence of both complementary ADARs of non-targeted sensors and targeted sensors.

FIG4 is a graphical illustration of (A) luciferase expression and (B) fold increase in luciferase expression compared to a negative control in HEK293FT cells. Ratio fold change can be calculated by calculating normalized sensor activation (gluc/cluc) and then normalizing to a ratio value in the absence of target. (A) Quantification of Gluc luciferase gene expression in HEK293FT cells introduced with two different guide strands targeting EGFP (Design 2 and Design 4) or a negative control scrambled sequence (negative control). (B) Quantification of the increase in EGFP expression compared to the negative control guide strand.

Figure 5 is a graphical illustration of the increase in luciferase expression of experimental guide strands 1 to 4 targeting EGFP transcripts in the presence of endogenous ADAR2 (Figure 3, blue bars), endogenous ADAR2 supplemented by an exogenously provided ADAR2 deaminase domain (ADAR2dd; Figure 2, white bars), or a fusion construct expressing a catalytically inactive Cas13b enzyme fused to the deaminase domain of ADAR2 (dPspCas13b-ADAR2dd; Figure 2, red bars). All fold increases are relative to a scrambled guide designed as a negative control that does not target EGFP.

Fig. 6 is a visual description of ADAR variants and catalytic domain mutation screening. (A) Schematic diagram of different ADARs tested, from left to right, including ADAR1p150, ADAR1p110, ADAR2 and MS2 coat protein (MS2coat protein, MCP)-ADAR fusion protein (MCP-ADAR). fl = full length. DD = deaminase domain. Catalytic domain mutations are not shown in the schematic diagram; all in the deaminase domain. (B) Bar graph showing activation of exogenously transfected sensors in the presence of exogenously transfected iRFP and different ADAR variants. The ADAR variants selected for screening in the target are shown in red, and RNA sequencing data show that the TAG stop codon is converted to TIG in the presence and absence of the target. Error bars represent the standard deviation of n = 3 technical repeats.

Fig. 7 is a graphical representation of activation of ADAR sensors with eGFP and iRFP targets. (A) Testing of ADAR variants on 69nt iRFP sensors. The fold change shown represents the fluorescence ratio (mNeon/mCherry) in the presence of the target divided by the ratio in the absence of the target. (B) Non-normalized mNeon/mCherry fluorescence ratio of the data shown in Fig. 7A. (C) Testing of ADAR variants on 51nt eGFP sensors. The fold change shown represents the fluorescence ratio (mNeon/mCherry) in the presence of the target divided by the ratio in the absence of the target. (D) Non-normalized mNeon/mCherry fluorescence ratio of the data shown in Fig. 7C. Error bars represent the standard deviation of n=3 technical repeats.

8 is a graphical depiction of stop codon editing rates for sensor sets in the +target and -target groups with (A) exogenous supplementation of MCP-ADAR2dd, (B) exogenous supplementation of ADAR1p150 isoforms, (C) exogenous supplementation of ADAR2, and (D) no exogenous ADAR supplementation.

Fig. 9 is a heat map showing the results of an experiment in which HEK293FT cells were transfected with plasmids expressing target transcripts and combinations of target-sensing ADAR sensor constructs shown on the y-axis and x-axis, respectively. The data shown are fold changes calculated as the fluorescence ratio (mNeon/mCherry) in the + target divided by the fluorescence ratio (mNeon/mCherry) under the - target (pUC19) condition. All conditions represent data from n=3 technical replicates.

Figure 10 (A) Selected ADAR variants screened in combination with respective RNA sensors for four different targets. Numbers in the heat map represent ratio fold changes. All conditions represent data from n=3 technical replicates. (B) Representative images of neuropeptide Y (NPY) targets of ADAR1p150 are shown. Cells were transfected with NPY sensors, ADAR variants, and targets in the combinations listed around the images. Image data were obtained by confocal microscopy of HEK293 cells at 10× magnification and 4× digital enhancement.

Figure 11 is a graphical illustration of the normalized (A, C, E, G) fluorescence ratio and non-normalized (B, D, F, H) fluorescence values of the target and ADAR sensor combination when tested in HEK293FT cells. The targets tested were iRFP (A, B), eGFP (C, D), neuropeptide Y (E, F) and dCas9 (G, H). (A, C, E, G): The sensor fluorescence ratio (mNeon/mCherry) fold change of + target conditions normalized to - target conditions for each target and sensor combination is shown. (B, D, F, H): Non-normalized mNeon/mCherry fluorescence ratio of each sensor and target combination with different ADAR variants. For iRFP and EGFP targets, next generation sequencing data of RNA sensors for UAG to UIG conversion. Editing % represents the reading % of A—>I editing. All conditions represent data from n=3 technical replicates.

Figure 12 is a representative image of the full 10x image with inset of the ADARpl50 image shown in Figure 10B without target (Figure 12A) or with target (Figure 12B). Scale bar is 100 μm.

13 is a graphical comparison of normalized luciferase values for sensor panels in the +target and -target groups with (A) MCP-ADAR2dd exogenous supplementation, (B) ADAR1p150 isoform exogenous supplementation, (C) ADAR2 exogenous supplementation, and (D) no exogenous ADAR supplementation.

Figure 14 is a heat map of titration of MCP-ADAR2dd with IL6 targeted sensor against exogenous supplementation of 20ng transfected tetracycline inducible human IL6 transgene in IL6. Fold change represents the normalized luciferase ratio between +target and -target groups.

Figure 15 is a graphic illustration of the luciferase expression of guide strands 1 and 2 targeting EGFP compared to the disordered guide strands designed to target EGFP without specificity. Each guide strand is introduced into HEK293T cells. The cells are then tested using the Cas13b enzyme fused to the deaminase domain of ADAR2 (RNA editing for programmable A to I (G) replacement (REPAIR)) or the Cas13b enzyme fused to the form without catalytic activity (REPAIR K370A). The guide strands are tested in cells containing only endogenous ADAR2 (Figure 15, blue bars), cells containing endogenous ADAR2 outside exogenous catalytically inactive REPAIR molecules (REPAIR K370A; Figure 15, white bars) or cells containing catalytically active REPAIR molecules (REPAIR; Figure 15: red bars).

Figure 16 is a heat map showing the fold increase of luciferase activation when tested in HEK293FT cells (white, lowest fold increase, to dark blue, highest fold increase). The y-axis shows different lengths of mismatched guide chains with a variety of different designs and targeting EGFP. The x-axis shows the exogenous ADAR molecules tested (none = endogenous only; ADAR2fl = ADAR2 full length, REPAIR = Cas13b enzyme fused to the deaminase domain of ADAR2; MS2-ADAR2dd = MS2 binding protein fused to ADAR2dd; dDisCas7-11-ADAR2dd = catalytically inactive Cas7-11 fused to the ADAR2 deaminase domain.

Figure 17 (A) is a schematic diagram showing different length sensors screened for iRFP target transcripts. (B) Bar graph showing sensor activation that increases with increasing sensor length. Sensor activation represents, for each sensor, the normalized fluorescence (mNeon/mCherry) value in the presence of the target divided by the normalized fluorescence (mNeon/mCherry) value in the absence of the target. (C) Fractional mNeon positive cells show the proportion of cells with expression above a predefined threshold. All conditions represent data from n=3 technical replicates.

Figure 18A is used for single cell image analysis similar to fluorescent cytometry of ADAR sensors targeting iRFP with guide lengths of 69, 249 and 600nt. The histogram shows the population density of mNeon expression in all cells under +iRFP (blue) and -iRFP (pink) target conditions. The dotted line shows the constant intensity threshold at which individual cells are gated as mNeon (+) or mNeon (-) under all conditions. The colored box shows the mNeon positive cell % under +iRFP (blue) and -iRFP (pink) target conditions. (B) Representative images for (A) are shown. Cells are transfected with iRFP targets, ADAR p150 and different ADAR sensor guide lengths in the combinations listed around the image. Scale bar, 100 microns

Figure 19 (A to B) are representative images of the full 10x images with insets from the image shown in Figure 18 B. Scale bar is 100 μm.

FIG. 20 is a graphical comparison of different exogenously supplemented ADAR variants on an IL6 targeting sensor and a transiently transfected tetracycline-inducible human IL6 transgene.

Figure 21 (A to D) is a series of graphical illustrations showing the normalized luciferase values of the sensor group in the following +target group and -target group: (A) MCP-ADAR2dd (E488Q, T490A) exogenous supplementation, (B) ADAR1p150 isoform exogenous supplementation, (C) ADAR2 exogenous supplementation and (D) no exogenous ADAR supplementation.

Figure 22 is a graphical representation of the comparison between sensors containing normal guides and sensors containing guides comprising multiple binding sites and MS2 hairpin loops in HEK293 cells against human IL6 target with endogenous ADAR1, exogenously supplemented ADAR1p150 isoforms, full-length ADAR2, or MCP-ADAR2dd (E488Q, T490A). The fold change is calculated by dividing the luciferase value (Gluc/Cluc) of the normalized + target condition by the - target condition.

Figure 23 is a visual representation of the engineering of the ADAR sensor with the MS2 hairpin loop and affinity region. The addition of the MS2 hairpin loop and affinity enhances the sensitivity and increases the dynamic range of the ADAR sensor.

FIG. 24 is a schematic diagram of the stepwise generation of a tri-affinity ADAR sensor with 5 nt spacing between affinity guide regions

Figure 25 (A) is a schematic diagram of different linker lengths outside the target region. (B) is a graphical representation of the effect of linker length between affinity regions. Linker lengths of 5nt, 30nt, and 50nt between affinity regions of the MS2 hairpin-linked 5-affinity sensor for IL6 were tested.

Figure 26 (A) is a schematic diagram of dual and single stop codon affinity/MS2 hairpin sensors. (B) is a comparison of sensor fold activation between a conventional MS2 hairpin-linked heptad affinity sensor and a dual stop codon heptad affinity sensor, in which a 3' downstream stop codon was inserted in the last affinity region.

Figure 27 (A) is a comparison of background and activation of affinity sensors and original ("long") sensors. Figure 27 (B) is a scatter plot of fold change and background luciferase values of affinity sensors and original ("long") sensors.

Figure 28 (A) is a bar graph showing a comparison between a five-binding site affinity sensor and a seven-binding site affinity dual stop codon sensor in MCP-ADAR2dd (E488Q, T490A) and ADAR1 p150. Figure 28 (B) is a bar graph showing a comparison of activation and background signals between a seven-binding site affinity single stop codon and a seven-affinity dual stop codon sensor

Figure 29 is a comparison of target mismatch tolerance for all 16 possible mismatches between the original 51bp sensor, tri-affinity, and penta-affinity sensor designs. (The 16 targets contained 5' or 3' nucleotide changes from conventional CCA). (A) is a schematic representation of mismatch tolerance. (B) is a heat map showing the log fold activation of all three sensor designs in the 16 target mismatches (blue), and the log10 of the normalized tolerance of different target mismatches relative to the native CCA target (red).

30 is a heat map showing the normalized preference of sensor designs within each target mismatch combination between the original 51 bp sensor, the three binding site sensor, and the five binding site sensor.

FIG. 31 (A) is a visual representation of the generation and activation of circular sensors. Conventional circular sensors were generated using the twister ribozyme backbone driven by the U6 promoter for in vitro self-circularization. Self-circularization of the sensor hibit tag utilizes the mammalian cell RtcB ligase. A rolling circle translation form of the circular sensor was generated by deleting the stop codon at the C-terminus of the hibit protein and inserting a T2A peptide to allow ribosomes to read through in a circular manner. (B) Sensors of varying lengths from 50 nt to 120 nt were compared for sensor activation fold change after induction of a transgenic target (human IL6).

Figure 32 (A) is a schematic diagram of the RNA modifications evaluated. (B) Heat map comparing different mRNA modifications of synthetic mRNA ADAR sensors detecting IL6 transgene expression in HEK293FT cells when HEK293FT cells were supplemented with MCP-ADAR2dd (E488Q, T490A) by plasmid transient transfection 24 hours prior to mRNA sensor transfection. (C) Heat map comparing different mRNA modifications of synthetic mRNA ADAR sensors detecting IL6 transgene expression in HEK293FT cells when HEK293FT cells were supplemented with MCP-ADAR2dd (E488Q, T490A) mRNA at the time of sensor transfection.

FIG33 is a graphical illustration depicting the expression of EGFP in HEK293FT cells (FIG. 33A, 6B) and the fold increase in GFP expression (FIG. 33C). EGFP expression was either constitutive or expressed in a gradient format using a doxycycline-inducible EGFP construct. HEK203T cells were then exposed to doxycycline at concentrations ranging from 8 ng/mL to 200 ng/mL.

34 is a graphical illustration depicting the dose-dependent luciferase activity of guide strand 1 (A) and guide strand 3 (B) as a function of doxycycline dose in HEK293FT cells simultaneously exposed to full-length ADAR2 and a guide strand targeting EGFP, which is under the control of a doxycycline-inducible promoter.

35 is a graphical representation depicting luciferase activity of guide strand 1 (A) and guide strand 3 (B) as a function of GFP fluorescence in HEK293FT cells simultaneously exposed to full-length ADAR2 and a guide strand targeting EGFP, which is under the control of a doxycycline-inducible promoter.

Figure 36 is a visual representation of the results of the combined treatment of tetracycline-inducible IL6 and stable lentiviral integration. (B) The double stop codon seven affinity IL6 sensor was then used to quantify the relative expression of IL6 and the corresponding luciferase fold change was plotted against the Cq value of IL6 expression as detected by quantitative polymerase chain reaction (QCPR).

Figure 37 is a scatter plot showing that the double stop codon seven affinity IL6 sensor can be used to quantify the relative expression of IL6 with a wide dynamic range. The target expression range was generated by a combination of transient overexpression of tetracycline-inducible IL6 in HEK293FT cells and a stable lentiviral integrated tetracycline IL6 cassette. The ADAR sensor fold change relative to basal conditions is plotted against the IL6 gene expression change as determined by quantitative polymerase chain reaction (qPCR).

38 is a scatter plot showing linear regression of sensor activation fold change versus qPCR detected gene expression fold change.

FIG. 39 is a diagram showing the editing of adenosine in the UAG stop codon of the sensor at different IL6 gene expression levels corresponding to FIG. 38 .

Figure 40 (A) is a schematic representation of an AND gate. (B) is a schematic representation of an OR gate. (C) is a graphical illustration comparing the activation fold change of the original 51nt guide AND gate sensor and the five affinity guide AND gate sensor in all four combinations of IL6 and EGFP target induction.

Figure 41 (A) is a graphical illustration of normalized sensor activation of AND-gated ADAR sensors input to EGFP and IL6 transcripts in all four possible target combinations. (B) is a graphical illustration of normalized sensor activation of OR-gated ADAR sensors input to EGFP and IL6 transcripts in different target combinations.

Figure 42 (A) is a schematic diagram of IL6 responsive caspase using an ADAR sensor with five affinity sensors targeting human IL6 transcripts. The sensor activates expression of FKBP self-dimerization caspase 9. (B) is a graphic illustration of the fold change of cell death (apoptosis) in response to ADAR sensor activation detected by IL6 transcripts. The positive control sensor involves a scrambled guide sequence in front of the iCaspase without a stop codon in the box. The cell death fold change is determined by calculating the fold change of cell viability compared to the + target under - target conditions. (C) is a bar graph comparing the percentage cell survival values of the IL6 responsive iCaspase ADAR sensor and the no stop codon control in the + target and - target groups.

Figure 43 is a visual representation of an experiment to test the efficiency of ADAR sensors in a heat shock assay. (A) Heat shock of Hela cells at 42°C was used to induce upregulation of HSP40 and HSP70 gene expression. HSP70 and HSP40 targeted sensors were transfected into Hela cells alone or with MCP-ADAR2dd (E488Q, T490A) and then at 42°C or 37°C for 24 hours. (B) qPCR validation of upregulated HSP40 and HSP70 levels after 24 hours of heat shock. (C) Sensor activation was calculated between the 42°C and 37°C groups and normalized relative to sensors with scrambled non-targeted guides to illustrate changes in protein degradation.

Figure 44 is a visual representation of the analysis of SERPINA1 in three cell types that differentially express SERPINA1. (A) is a bar graph comparing SERPINA1 expression in HEK293FT, SERPINA1 and Hela cells. (B) In three different cell types (HEK293FT, Hela and HepG2 cells), SERPINA1 sensing five affinity sensors were transfected with or without exogenous MCP-ADAR2dd (E488Q, T490A). (C) is a bar graph comparing the sensor activation fold changes between Hela cells and HepG2 cells in SERPINA1 sensors targeting different CCA sites. (D) Sensor activation is determined by calculating the raw luciferase value of the SERPINA1 sensor, which is normalized by the scrambled non-targeted guide sensor to illustrate the protein production/secretion and background ADAR activity differences between cell types, and then normalized to the gluc/cluc ratio in HEK293FT cells.

FIG45 is a graphical representation of the normalized fold change in editing rate of the SERPINA1 penta-affinity sensor in different cell types (HEK293, Hela, and HepG2).

FIG. 46 is a bar graph comparing fold changes in activation of mRNA SERPINA1 sensors targeting different CCA sites on the SERPINA1 transcript in transiently transfected Hepal-6 cells with tetracycline-inducible SERPINA1 expression.

Figure 47 (A) is a schematic diagram of an in vivo sensing experiment of human SERPINA1 transcripts using SERPINA1 mADAR sensor constructs. SERPINA1 targeted sensor mRNA with Akaluc output was produced in vitro with 25% 5-methylcytosine and 0% pseudouridine. Constitutive Akaluc (no stop codon) and non-targeted guide (with stop codon) sensor constructs were synthesized using the same protocol. All mRNAs were packaged with lipid nanoparticles and injected into wild-type mice or NSG-Piz mice with human SERPINA1 Piz mutation boxes via the tail vein. In vivo sensor activation was measured 18 hours after injection. (B) Representative images of sensor activation of multiple synthetic mRNA ADAR sensor constructs are shown.

Figure 48 (A) is a graphic illustration of the radiation produced by Akaluc calculated for the liver and compared between wild-type and NSG-PiZ mutant mice. The fold change between NGS-PiZ mice and WT mice was calculated for each ADAR sensor construct. Significance was determined by a two-tailed t test, N=2 mice. *, p<0.05. P values less than 0.05 are statistically significant with asterisks. (B) is a bar graph comparing the Akaluc luminescent radiation in NSG PiZ mice and WT mice in non-targeted sensors, constitutive sensors, SERPINA1 CCA35 targeted sensors, and SERPINA1 CCA30 targeted sensors.

Figure 49 is a graphic illustration of differential gene analysis of 37 tissues using the Human Protein Atlas and GTEx datasets for the minimum number of genes required to genetically classify tissues (A) and the number of protein-coding genes that are enriched in a particular tissue, enhanced in a particular tissue, or have low specificity (B). (C) is a heat map showing the relative transcript abundance of 34 mRNAs that uniquely define tissues in 34 different tissue types.

Figure 50 is a graphic illustration showing the RADARS safety characterization on immune response and endogenous RNA knockdown. (A, B) Effect of sensor-target duplex formation on innate antiviral pathways. RADARS sensors were transfected in the presence or absence of complementary target sequences. Total RNA was analyzed using quantitative PCR (qPCR) to determine the relative expression levels of MDA5 (A) and IFN-β (B). (C, D) Effect of sensor-target duplex formation on the abundance of endogenous targeted transcripts. The relative abundance of NEFM and PPIP transcripts after transfection of complementary or non-targeted RADARS sensors was assessed by qPCR. Data are expressed as mean ± s.d. (n = 4); unpaired two-sided Student's t-test, ns, p>0.05.

Figure 51 is a graphic illustration showing the inhibition of RADARS signals corresponding to knockdown of endogenous targets. (A) siRNA knockdown schematic of endogenous transcripts. (B) Expression differences between siRNA targeting PPIB or NEFM and control non-targeting siRNA in HEK293FT cells detected by qPCR and fluorescence RADARS. For RADAR, sensor activation was calculated for targeted siRNA and normalized to control siRNA. Data are mean ± s.d of technical replicates (n ≥ 3).

Figure 52 is a graphical comparison of the fold activation of the #CCA8IL6 engineered guide RNA with a 171 nt guide and 4 MS2 loops when used with exogenously supplemented ADAR1p150 or in combination with endogenous ADARs. Data are mean ± s.e.m of technical replicates (n = 3).

Figure 53A is a graphical comparison of the fold activation of mRNA RADARS sensor activation when detecting IL6 transcripts combined with plasmid ADAR1p150 transfection. The synthetic mRNA sensor was synthesized with different chemically modified bases at different incorporation levels from 0 to 100%. The data are the mean ± s.e.m of technical replicates (n = 3). Figure 53B is a graphical comparison of the fold activation of mRNA RADARS activation when detecting IL6 transcripts (using endogenous ADARs) when synthesized with different chemically modified bases at different incorporation levels from 0 to 100%. The data are the mean ± s.e.m of technical replicates (n = 3). Figure 53C is a graphical comparison of the induction of interferon beta response due to mRNA RADARS transfection. The synthetic mRNA was synthesized with different levels of chemically modified bases, and the interferon response was measured by plasmid (One-Glo luciferase) reporter assay (Gentili et al., 2015).

Figure 54A is a visual representation of the evolution of different RADARS sensor designs used in combination with exogenous ADAR1p150 supplementation. Illustrations depict the backbone of different RADARS designs. RADARS multiple activation is calculated as the ratio of luminescence (Gluc/Cluc) of Gaussia luciferase (Gaussia luciferase, Gluc) divided by constitutive Cypridina luciferase (Cluc) in the presence of IL6 target relative to the absence of target (see method). Sensor, target (IL6) and ADAR1p150 are co-delivered by transient transfection. Data are mean ± s.e.m of technical repeats (n = 3). Figure 54B is a graphical comparison of the Gluc/Cluc ratio of modified guide RNAs with different lengths and different MS2 hairpin loops targeting IL6#CCA8 while maintaining 0aa 5' peptide length between + target and - target conditions. Error bars represent the standard error of the mean. (n = 3 technical repeats).

FIG. 55A is a graphical comparison of the Gluc/Cluc ratios of guide RNAs engineered with five affinity binding sites (4 MS2 loops) and nine affinity binding sites (8 MS2 loops) with different lengths of 5' peptides between +target and -target conditions. Error bars represent standard error of the mean. (n=3 technical replicates). FIG. 55B is a graphical comparison of the Gluc/Cluc ratios of guide RNAs engineered with five affinity binding sites with 200aa 5' peptide residues without an out-of-frame stop codon and with two added out-of-frame stop codons. The last column represents constitutive gluc driven under the Ef1-α promoter. Error bars represent standard error of the mean. (n=3 technical replicates).

Figure 56A is a graphical depiction of the multiple activation of RADARS targeting IL6, EGFP and NPY supplemented with exogenous ADAR1p150. For each transcript, 12 modified guide RNAs were modified to target different CCA sites in the transcript. The CCA site numbering described follows such a convention that #CCAx represents the number of CCA triplicates counted from the 5' end of the transcript coding region. Each point represents the average of three technical repetitions of a single sensor. The horizontal solid line represents the average of all 12 modified guide RNAs. Figure 56B graphically depicts the editing percentage of the target adenosine in the UAG stop codon of 14 modified guide RNAs targeting IL6 with exogenous ADAR1p150 supplemented on IL6 in the presence and absence of target IL6 transcripts, non-targeted modified guide RNAs and RADARSv2 designs with exogenous ADAR1p150 supplemented on IL6 CCA sites. Error bars represent the standard error of the mean (n=3 technical repetitions).

Figure 57 is a graphic depiction of RADARSv2 with engineered guide RNAs targeting high TPM genes (RPS5), low TPM genes (KRAS), or non-targeted scrambled sequences, supplemented with exogenous ADAR1p150 or used with endogenous ADARs to sense downregulation of their corresponding genes by gene-specific siRNAs. Fold activation is calculated by activation of the payload in the on-target siRNA group relative to activation of the payload in the non-target siRNA group. Data are mean ± s.e.m of technical replicates (n = 3).

FIG. 58A is a visual depiction of a schematic diagram showing a fluorescent output RADARS construct containing a constitutively expressed normalized fluorescent protein (mCherry) upstream of a RADARS engineered guide RNA controlling an mNeon fluorescent protein (top), and an image showing fluorescent RADARS of HEK293FT cells expressing the mNeon payload only in the presence of a target transcript (out-of-frame EGFP). HEK293FT cells were transfected with RADARS targeting EGFP, ADAR1p150, and ±target (out-of-frame EGFP), as shown (bottom). Scale bar, 100 microns. FIG. 58B is a visual depiction of flow cytometry analysis of fluorescent RADARS, showing a histogram of mNeon/mCherry fluorescence of HEK293FT cells transfected as in d), with beige and blue distributions indicating target absence and target presence, respectively.

Figure 59A is a visual description of the gating strategy for flow cytometry analysis of fluorescent RADARS in HEK293 cells. The gate is drawn using a control population transfected with a pUC19 plasmid. Figure 59B is a visual description of the gate covered on a cell population transfected with RADARS, ADARp150 and pUC19 plasmids targeting EGFP. Figure 59C is a visual description of the gate covered on a cell population transfected with RADARS, ADARp150 and EGFP target (frameshift) plasmids targeting EGFP.

Figure 60A is a visual depiction of RADARS fold activation relative to basal conditions (0 ng/mL doxycycline in integrated HEK293FT cells), which is plotted relative to changes in IL6 gene expression as determined by quantitative polymerase chain reaction (qPCR) on a log10-log10 scale. The blue dotted line represents the linear regression results of the data. The data are the mean ± s.e.m of technical replicates (n = 3). Figure 60B is a visual depiction of the original Cq value of the IL6 transgene normalized by subtracting the number of GAPDH gene Cq and the corresponding fold activation of RADARS. The error bars represent the standard error of the mean (n = 3 biological replicates). Figure 60C is a visual depiction of the effect of titrating the optimal IL6 modified guide RNA RADARS sensor amount on the resulting activation and total protein production under target addition conditions (gluc/cluc ratio). For conditions below 40 ng, the remaining plasmid amount was replaced with pUC19 plasmid. The error bars represent the standard error of the mean (n = 3 technical replicates).

Figure 61 is a visual depiction of the results of a validation experiment of siRNA knockdown of 10 endogenous transcripts as measured by qPCR expression. The fold change was calculated by gene expression of the target transcript in the on-target siRNA group relative to gene expression of the target transcript in the off-target siRNA group. (n=3 biological replicates).

Figure 62A (top) is a visual description of gene expression in transcripts per million (TPM) of 10 genes from 10,381TPM (RSP5) to 13TPM (KRAS) shown on a logarithmic scale. Bottom: RADARSv2 detection of transcripts in cells treated with 100nM targeted siRNA pool or non-targeted siRNA pool. The bars represent the multiple activation of targeted RADARS and non-targeted RADARS constructs (the Gluc/Cluc ratio of RADARS in the targeted siRNA group relative to the Gluc/Cluc ratio of RADARS in the non-targeted siRNA group). The significance between targeted and non-targeted RADARS was determined by an unpaired t-test, with the Welch correction assuming a single variance of each group (*, p<0.05. **, p<0.01. ***, p<0.001. ****, p<0.0001). Figure 62B is a graphical description of the editing rate of the UAG stop codon in the best performing sensor in each genome from b. One-tailed unpaired t-test was performed between the target siRNA group and the non-target siRNA group. (*, p<0.05. **, p<0.01. ***, p<0.001. ****, p<0.0001) Error bars represent standard error of the mean (n=3 technical replicates).

Figure 63A is a graphical depiction of the performance of 8 randomly selected engineered guide RNAs and 8 randomly selected non-targeted engineered guide RNAs targeting 10 endogenous transcripts. The multiple activation of RADARS is calculated by the Gluc/Cluc ratio of the on-target siRNA group relative to the Gluc/Cluc ratio of the off-target siRNA group. The best performing targeted sensor for each gene is marked in yellow and tested in Figure 63. The horizontal line represents the mean value of each group. Figure 63B is a visual depiction of RADARSv2 targeting RPL41, GAPDH, ACTB, HSP90AA1, PPIB and KRAS, tracking the expression of these transcripts under a range of siRNA concentrations. The blue and beige lines represent the multiple activation of the RADARS (Gluc/Cluc) ratio and the multiple change of qPCR quantitative expression relative to 0nM siRNA, respectively. The gray line represents the multiple activation of the non-targeted engineered guide RNA (non-complementary to the target transcript). The data are the mean ± s.e.m of technical replicates (n = 3). (R values represent Pearson correlation between qPCR and targeted RADARS, *, p<0.05. **, p<0.01. ***, p<0.001).

Figure 64A is a visualization diagram of the upregulation of the heat shock protein family gene HSP70 during heat shock at 42 degrees Celsius. Figure 64B is a visualization of the results of an experiment in which four HSP70-targeted guide RNAs and scrambled non-targeting (NT)-targeted guide RNAs (all with exogenous ADAR1p150 supplement) targeting different CCA sites were transfected into HeLa cells and then incubated at 42°C or 37°C for 24 hours. qPCR and RADARSv2 detected differences in HSP70 expression between the 37°C (control) group and the 42°C (heat shock) group. Sensor activation was calculated between the 42°C and 37°C groups and normalized to NT conditions. Data are mean ± s.e.m of technical replicates (n = 3).

Figure 65A is a schematic visualization of a two input AND gate with RADARS. Figure 65B is a graphical depiction of normalized sensor activation of AND-gated RADARS for EGFP and IL6 transcript inputs in all four possible target combinations. Data are mean ± s.e.m of technical replicates (n = 3).

Figure 66A is a schematic visualization of the logic of a two input OR gate with RADARS. Figure 66B is a graphical depiction of sensor activation of OR-gated RADARS for all possible combinations of EGFP and IL6 transcript inputs. Data are mean ± s.e.m values of technical replicates (n = 3).

Figure 67A is a schematic representation of the RADARS targeting SERPINA1 with an inducible caspase 9 payload. Figure 67B is a graphic representation of the cell viability of A549, Hela and HepG2 cells after transfection with SERPINA1 sensing RADARS expressing iCaspase9 in combination with exogenous ADAR1p150. The non-targeted control modified guide RNA contains a scrambled sequence with a stop codon before the payload. The data are the mean ± s.e.m of technical replicates (n = 3). Figure 67C is a graphical representation of the cell viability of HepG2, Hela and A549 cells, which is measured using MTS assay 48 hours after transfection of SERPINA1 expressing iCaspase9 or non-targeted RADARS constructs and ADAR1p150, and is normalized relative to the control group transfected with GFP expression plasmid alone.

Figure 68 is a graphical depiction of experimental results in which the RADARS constructs optimally targeting HSP70 and scrambled non-targeting (NT) RADARS constructs on HSP70 transcripts were transfected into HeLa cells without exogenous ADARs and then incubated at 42°C or 37°C for 24 hours. qPCR and RADARSv2 detected differences in HSP70 expression between the 37°C (control) group and the 42°C (heat shock) group. Sensor activation was calculated between the 42°C and 37°C groups and normalized relative to the NT condition, which was a sensor with a scrambled non-targeting engineered guide RNA (NT) to control for changes in protein production. Data are the mean ± s.e.m of technical replicates (n = 3).

Figure 69A is a visualization diagram of a double loxP EGFP Cre reporter and IL6 RADARS-CRE. Right: HEK293FT fluorescence 48 hours after transfection of double loxP-EGFP reporter, ADAR1p150 and RADARS targeting IL6 with Cre payload with or without IL6 target. Images under -target and +target conditions are shown. White scale bar represents 100 microns. Figure 69B is a visualization description of the experimental results, wherein the cells from Figure 69A are harvested for flow analysis of EGFP expression.

Figure 70A is a visualization schematic diagram of the RADARS construct targeting SERPINA1 with Cre payload. Figure 70B is a visualization description of the result of GFP+ cell percentage of Hela, HepG2 and A549 cell flow cytometry analysis 48 hours after transfection with the RADARS construct targeting SERPINA1 expressing Cre with exogenous ADAR1p150. Data are the mean ± sem of technical repetitions (n=3). Figure 70C is a visualization description of the EGFP expression quantitatively expressed by flow cytometry 48 hours after transfection of CRE reporters, ADAR1p150 and RADARS targeting IL6 with CRE payload with or without IL6 target. Analysis of EGFP signal distribution is carried out to Hela, A549 and HepG2 cells by flow cytometry. For all three cell types, GFP positive cells are defined as FITC channel EGFP intensity higher than 10 ⁷ cell colonies.

Figure 71A is a visual description of bioluminescent images of sensor activation of multiple synthetic mRNA RADARS constructs. Figure 71B is a graphic comparison of Akaluc luminescent radiation in substrate background, constitutive sensor and RADARS targeting SERPINA1#CCA32 in NSG-PiZ mice and NSG-WT mice liver. Data are mean ± s.e.m of technical repeats (n = 3). By two-tailed unpaired t test, significance was determined between NSG-WT and NSG-PiZ sample radiation, N = 3 mice. (NS, p>0.05. **, p <0.01.) Figure 71C is a graphic comparison of the radiation produced by Akaluc in the liver and compared in wild-type and NSG-PiZ mutant mice. Calculate the multiple activation of each RADARS construct between NGS-PiZ mice and NSG-WT mice. Significance was tested by two-tailed unpaired t-test for comparison of each group with substrate background, N=3 mice (NS, p>0.05. ***, p<0.001).

Figure 72A is a graphic depiction of transcript expression levels of 10 endogenous genes from HEK293FT cells as quantified by qPCR, which were transfected with targeted or non-targeted (NT) RADARS constructs in the presence of exogenous ADAR1p150 supplementation. The data shown are normalized relative to the targeted RADARS. The significance between targeted and non-targeted RADARS was determined by unpaired t-test, with Welch correction assuming a single variance for each group (Ns, p>0.05). Figure 72B is a graphic depiction of protein expression of endogenous ACTB and PPIB in HEK293FT cells transfected with RADARS targeting ACTB/PPIB or non-targeted RADARS (with ADAR1p150 supplementation) as quantified by Western blot. The significance between targeted and non-targeted RADARS was determined by unpaired t-test, with Welch correction assuming a single variance for each group (Ns, p>0.05). Figure 72C and Figure 72D are visual depictions of analysis of the effects of sensor-target hybridization on protein production by Western blot. ACTB (Figure 72C) and PPIB (Figure 72D) protein levels in response to RADARS hybridization are shown, and GAPDH is used as a normalized protein control.

Figure 73A is a graphic representation of the expression levels of interferon beta, OAS1, RIG-1 and MDA5 detected by qPCR when transfected with RADARS constructs targeting IL6, ACTB, RPS5 or PPIB, non-targeted (NT) RADARS and high molecular weight poly (I: C). Significance was determined by one-way ANOVA test in the untreated group, RADARS group and poly (I: C) group (Ns, p>0.05. ****, p<0.0001). Figure 73B is a graphic representation of the gene expression fold changes detected by qPCR of four dsRNA response genes (IFNb, OAS1, MDA5 and RIG-1) in HEK293FT cells using ACTB as a normalization gene for RADARS response. 73C is a graphical depiction of the fold changes in gene expression detected by qPCR of four dsRNA responsive genes (IFNb, OAS1, MDA5 and RIG-1) in response to RADARS or poly(I:C) in HepG2 cells using GAPDH as the normalization gene.

Figure 74A is a quantitative visualization of RNA editing of ACTB, PPIB and NT modified guide RNAs in the 200bp hybridization region of the target transcript and RADARS. The transition from A to I (G) is described in the heat map. (NS, p>0.05) Figure 74B is a scatter plot analysis of off-target editing of the whole transcriptome. The scatter plot shows the allele fractions of the following A->G mutations: (i) PPIB sensor and ADAR1p150 overexpression and untransfected HEK293T cells (n=3.17×106 sites); (ii) is the same as (i), except for IL6 sensor (n=3.07×106 sites); (iii) is the same as (i), except for no ADAR overexpression and non-targeted RADAR sensor (n=3.14×106 sites). Sites are colored by fdr-corrected p values (left color bar). For i and iii, the experiment was generated by 3 independent repeats. Figure 74C is a scatter plot analysis of off-target editing of the whole transcriptome with MCP-ADAR2 (E488Q) using a data set from NCBI Geo accession number GSE123905 (Katrekar et al., 2019). The scatter plot shows the allele fraction of A->G mutations in MCP-ADAR2 (E488Q) overexpression and untransfected HEK293T cells (n = 2.35 × 106 sites). Sites are colored by FDR-corrected p-values (left color bar).

Figure 75A is a graphic depiction of sequence homology analysis between transcriptome off-targets (n=23 sites) and targeted PPIB homology regions. Corresponding FDR-corrected p-values (red line indicates p=0.05) for significance of local alignment (Monte Carlo permutation test) of 200 bp around each off-target site and the PPIB homology region targeted by the engineered guide RNA. Figure 75B is a visual depiction of sequence logo analysis of RADARS off-target editing sites overexpressing ADAR1p150 and targeting PPIB.

DETAILED DESCRIPTION

The present disclosure provides systems and sensors for detecting and quantifying RNA. The present disclosure also provides systems and methods for gene editing. A system for in vivo imaging of RNA expression is also disclosed. The present disclosure provides a sensor system for converting adenine to inosine. Inosine is recognized as guanosine (G) by the translation mechanism. The conversion of adenine to inosine is the result of adenosine hydrolysis and deamination. (Cox et al. Science 358 (6366): 1019-1027 (2017)). Therefore, the adenosine deaminase (ADAR) enzyme family acting on RNA can convert codons in transcripts so that the translation product is functionally changed. In addition to other functional changes, the present disclosure provides editing of transcripts to remove stop codons, and thus enable payload to be expressed.

definition

Unless otherwise defined, the terms and techniques used in this application have the same meanings as commonly understood by those skilled in the art.

As used herein, the term "about" is understood to modify the specified value. Unless otherwise expressly stated, the term about is understood to modify the specified value +/- 10%. As used herein, the term about applied to a range modifies both endpoints of the range. For example, a range of "about 5 to 10" is understood to mean "about 5 to about 10".

As used herein, the terms "sensor" and "sensor strand" are used interchangeably. As used herein, a "sensor" or "sensor strand" is understood to refer to a single-stranded RNA comprising at least a stop codon, wherein the RNA strand is capable of hybridizing or forming a duplex with another RNA strand. "Sensor" and "guide" strands are used interchangeably throughout this disclosure.

As used herein, the terms "ADAR," "ADAR enzyme," and "deaminase" are used interchangeably unless otherwise expressly stated. Therefore, in the present disclosure, "wherein the ADAR enzyme is a prokaryotic RNA editing enzyme" should also be understood to mean "wherein the deaminase is a prokaryotic RNA editing enzyme."

As used herein, the term "RNA sensor system" or "sensor system" should be understood to mean the minimum components required for: (i) hybridization of a single-stranded RNA to a transcript of interest such that the resulting hybridized RNA contains at least one mismatch and at least one stop codon, (ii) recognition of the hybridized RNA as a substrate, and (iii) editing of the single-stranded RNA to remove the stop codon.

As used herein, the term "cellular logic system" or "logic system" refers to a system composed of multiple individual sensor systems. The cellular logic system or logic system of the present disclosure is a complex system that can be composed of one or more individual RNA sensor systems. When an individual RNA sensor system is incorporated into a larger cellular logic system, it may rely on an individual RNA sensor system in the same cell for activation. Alternatively, multiple individual RNA sensor systems can be incorporated into a cellular logic system such that no individual RNA sensor system is required for another individual RNA sensor.

Unless otherwise expressly stated, it is generally understood that the term "payload" as used herein means a portion of a single-stranded RNA that can hybridize with another single-stranded RNA to become an invasive strand of an already duplexed RNA molecule, or can be translated to express a protein. Thus, for example, if an embodiment of the present disclosure states that "payload comprises a therapeutic protein," it is generally understood that the payload is a fragment or portion of a single-stranded RNA that can be translated to express a therapeutic protein.

As used herein, "cell-specific," "cell type-specific," and "activatable by a particular cell type" will be understood by those skilled in the art to mean that activation of the RNA sensor requires the presence of factors that are present at significantly higher levels in a particular cell type than in other cell types. Those skilled in the art will recognize that while expression of these factors may occur in other cell types, making activation of the RNA sensor possible, this is not highly likely.

As used herein and unless otherwise indicated, "affinity region" or "affinity binding region" are used interchangeably to describe a region on a guide strand that has a degree of complementarity to a target transcript. Affinity refers to the design of multiple binding sites in a guide. These binding sites can be separated by a linker. The affinity region can optionally be separated from the primary sensor region of the guide strand by one or more secondary structures (including a hairpin structure). In some embodiments, the hairpin structure is an MS2 hairpin. The affinity region can optionally contain a stop codon that is not targeted by ADAR editing. In some embodiments, the affinity region comprises a linker sequence. In some embodiments, the affinity region comprises one or more linker sequences.

Adenosine deaminases acting on RNA (ADARs) and other RNA-editing enzymes.

ADAR enzymes are evolutionarily conserved in animals. Mammals have three known ADAR enzymes: ADAR1, ADAR2, and ADAR3. ADAR1 and ADAR2 are known to have catalytic activity. In contrast, although ADAR3 is substantially similar to ADAR2, ADAR3 is generally considered to be catalytically inactive. (Savva et al. Genome Biol. 13 (12): 252. (2012)). The ADAR enzymes considered in the present disclosure include mammalian ADAR enzymes or modified enzymes derived therefrom. In some embodiments, the ADAR enzyme of the RNA sensor is human ADAR1. In some embodiments, the ADAR enzyme of the RNA sensor is modified human ADAR1. In some embodiments, the ADAR enzyme of the RNA sensor is human ADAR2. In some embodiments, the ADAR enzyme of the RNA sensor is modified human ADAR2. In some embodiments, the ADAR enzyme of the RNA sensor is modified human ADAR3. In some embodiments, the ADAR enzyme of the RNA sensor is a synthetic enzyme. In some embodiments, the ADAR enzyme of the RNA sensor is a non-mammalian ADAR enzyme.

The ADAR enzyme of the present disclosure includes modified enzymes. The ADAR enzymes considered in the present disclosure include enzymes that have been modified to increase the affinity of the enzyme to the sensor chain. In some embodiments, ADAR has been modified to include additional RNA binding domains. In some embodiments, ADAR has been modified to exclude one or more non-catalytic domains. In some embodiments, the ADAR enzyme of the sensor system includes an ADAR2 deaminase domain. In some embodiments, ADAR consists of an ADAR2 deaminase domain. In some embodiments, ADAR includes an ADAR2 deaminase domain fused to an MS2 binding protein. In some embodiments, ADAR consists of an ADAR2 deaminase domain fused to an MS2 binding protein. In some embodiments, ADAR is fused to a CRISPR-associated protein (Cas protein), or a fragment or derivative thereof. In some embodiments, ADAR is fused to a modified Cas protein. In some embodiments, the modified Cas protein has been mutated to lack catalytic activity. In some embodiments, ADAR is fused to a modified Cas13. In some embodiments, ADAR is fused to a Cas13b comprising a mutation at an amino acid corresponding to K370. In some embodiments, ADAR is fused to a Cas13b comprising a K370A mutation. In some embodiments, ADAR is fused to a modified Cas13d. In some embodiments, ADAR is fused to a modified Cas7-11.

The ADAR enzymes of the present disclosure may be endogenous to the cells to which the sensor is delivered. The ADAR enzymes contemplated in the present disclosure may be exogenous and delivered to the cells simultaneously or separately with the sensor. In some embodiments, the exogenous ADAR is delivered separately from the sensor. In some embodiments, the exogenous ADAR is delivered simultaneously with the sensor. In some embodiments, the exogenous ADAR can be used to supplement the endogenous ADAR. In some embodiments, more than one exogenous ADAR is provided to the cell.

Additional RNA-editing enzymes

Additional deaminases can be used in the methods of the present disclosure. In some embodiments, the deaminases can be modified ADAR enzymes. Modified ADAR enzymes can include ADARs that have been modified to have increased cytidine deamination activity, such as RESCUE (Abudayyeh et al., Science 365, 382-386 (2019)).

In some embodiments, the deaminase is a modified cytidine to uracil editing enzyme. In some embodiments, the deaminase can be a member of the catalytic polypeptide-like (APOBEC) family of apolipoprotein B mRNA editing enzyme, cytidine deaminase. In some embodiments, the deaminase is a modified APOBEC1. In some embodiments, the deaminase is a modified APOBEC2. In some embodiments, the deaminase is a modified APOBEC3. In some embodiments, the deaminase is a modified APOBEC3A. In some embodiments, the deaminase is a modified APOBEC3B. In some embodiments, the deaminase is a modified APOBEC3C. In some embodiments, the deaminase is a modified APOBEC3D. In some embodiments, the deaminase is a modified APOBEC3E. In some embodiments, the deaminase is a modified APOBEC3F. In some embodiments, the deaminase is a modified APOBEC3G. In some embodiments, the deaminase is a modified APOBEC3H.

In some embodiments, the deaminase can be a prokaryotic RNA editing enzyme. In some embodiments, the deaminase is derived from Escherichia coli (E. coli).

Sensors/Sensor Chains

The sensor of the present disclosure comprises at least one stop codon. The sensor can be located on the same RNA strand as the payload, the normalizing gene, or both the payload and the normalizing gene. The sensor of the present disclosure can be designed so that when the single-stranded RNA (ssRNA) sensor strand binds to the target ssRNA strand to produce a double-stranded RNA (dsRNA) duplex, the duplex comprises a mismatch in the region corresponding to the stop codon in the sensor strand. The disclosed sensor can be modified in a variety of ways. In some embodiments, the sensor is provided to the cell as a DNA template, which can then be transcribed into a single-stranded RNA sensor molecule.

The present disclosure also provides sensor chains comprising more than one stop codon. In some embodiments, the sensor chain comprises 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 stop codons. In some embodiments, the sensor chain comprises more than 10 stop codons. In some embodiments, the sensor chain comprises 2 stop codons. In some embodiments, the sensor chain comprises 3 stop codons. In some embodiments, the sensor chain comprises 4 stop codons. In some embodiments, the sensor chain comprises 5 stop codons. In some embodiments, the sensor chain comprises 6 stop codons. In some embodiments, the sensor chain comprises 7 stop codons. In some embodiments, the sensor chain comprises 8 stop codons. In some embodiments, the sensor chain comprises 9 stop codons. In some embodiments, the sensor chain comprises 10 stop codons.

The present disclosure also provides a sensor/guide strand containing one or more affinity binding regions. In some embodiments, the sensor strand comprises 3 affinity regions. In some embodiments, the sensor strand comprises 5 affinity regions. In some embodiments, the sensor strand comprises 7 affinity regions. The affinity regions may incorporate stop codons that are not targets for ADAR editing. The affinity binding region may also contain stop codons that are targets for ADAR editing. In some embodiments, the affinity binding region comprises a stop codon for ADAR editing.

Also provided in the present disclosure are sensor/guide strands incorporating one or more MS2 hairpins. In some embodiments, the sensor strand comprises two MS2 hairpins. In some embodiments, the sensor strand comprises three MS2 hairpins.

In some embodiments, the sensor/guide strand comprises both an affinity region and an MS2 hairpin region domain.

In some embodiments, the sensor/guide strand comprises an RNA modification. In some embodiments, the modified RNA comprises 5-methylcytosine. In some embodiments, the modified RNA comprises pseudouridine.

Payload

In some embodiments, the payload comprises a reporter transcript. In some embodiments, the payload consists of a reporter transcript. In some embodiments, the reporter transcript is a fluorescent reporter. In some embodiments, the reporter transcript comprises a luciferase transcript. In some embodiments, the reporter transcript comprises a GFP transcript.

The sensor systems of the present disclosure can be designed to deliver a payload encoding a therapeutic protein. In some embodiments, a therapeutic protein can be used in conjunction with another therapeutic protein.

In some embodiments, the payload comprises a transcription factor. In some embodiments, the payload comprises an enzyme. In some embodiments, the payload comprises a transgenic protein.

In some embodiments, the payload comprises a protein for editing the genome of a cell. In some embodiments, the payload comprises a Cas protein. In some embodiments, the payload comprises a Cas9 protein.

In some embodiments, the payload comprises a protein capable of converting one cell type to another.

In some embodiments, the payload comprises an ADAR. In some embodiments, the payload comprises an ADAR capable of initiating a positive feedback loop.

In some embodiments, the payload comprises a protein capable of killing a specific cell type. In some embodiments, the payload comprises a protein capable of killing tumor cells. In some embodiments, the payload comprises an immunomodulatory protein.

Logic Gates

The present disclosure also relates to complex multi-sensor reporting subsystems. In some embodiments, these multi-sensor reporting subsystems use logic gates. These logic gates can be composed of AND gates, OR gates, or AND-OR gates as a single decision point in the same reporting subsystem.

AND gates for use in the reporter systems of the present disclosure can be implemented by having multiple guide strand binding moieties on the same ssRNA sensor. In this type of AND gate, each individual guide moiety senses a separate endogenous transcript in the cell. In this type of gate, the "activation" of the AND gate is the same as the activation of the entire ssRNA sensor chain; i.e., the removal of the stop codon and the expression of the terminal payload. This type of logic gate requires that each guide moiety interact with the target sequence and the ADAR or other deaminating molecule. Deamination of the stop codon in each guide moiety will allow full expression of the payload. In some embodiments, each guide moiety is further separated by a single reporter. In some embodiments, each guide moiety is further separated by a single, different reporter. In some embodiments, each reporter on the ssRNA sensor used in the AND gate is a different fluorescent reporter.

In some embodiments, the AND gate can operate in a sequential manner. In this type of AND gate, each of the multiple guide strand binding moieties is located on a separate ssRNA sensor. In this type of AND gate, the activation of the gate involves activating multiple sensors in a defined order. In this type of AND gate, the activation of the first sensor leads to the expression of an intermediate payload. In a specific cellular environment, this intermediate payload allows the expression of a second RNA sensor. In such a system, a cascade reaction of RNA sensor activation may occur after activation and only in the context of a specifically determined cellular stimulus.

The OR gate used in the reporter system of the present disclosure can be implemented by using multiple independent ssRNA sensors in the same cell. Each of the multiple independent sensors can deliver a payload without activating another sensor.

Delivery System

The present disclosure also provides systems for delivering ADAR sensors.

In some embodiments, the ADAR sensor is delivered directly to the cell. In some embodiments, the ADAR sensor is encapsulated in a lipid nanoparticle. In some embodiments, the ADAR sensor is delivered via a viral vector.

In some embodiments, the ADAR sensor is a circular RNA.

Methods of the present disclosure

The present disclosure also provides methods for using the sensor systems described herein.

In some embodiments, the RNA sensor delivers an optically observable payload. In some embodiments, the RNA sensor is tracked by an imaging system. The imaging system can be any suitable imaging system compatible with the system. In some embodiments, the imaging system uses fluorescent molecules. In some embodiments, the RNA sensor system comprises a plurality of fluorescent molecules. In other embodiments, the imaging system is non-invasive. In some embodiments, the RNA sensor system is compatible with a fluorescence activated cell sorting (FACS) system.

In some embodiments, the RNA sensor system is tracked by a non-invasive imaging system. In some embodiments, the imaging system tracks deep red luciferase.

The present disclosure also provides methods for quantitative RNA in vivo. The quantitative methods of the present disclosure can rely on the incorporation of a normalizing gene on the sensor chain. The translation of the normalizing gene occurs independently of RNA editing. The quantification of the normalizing gene provides a reference for the total sensor chain delivered to each individual cell. Reference to the normalizing gene allows determination of the percentage of activated RNA sensors in the total delivered sensors.

The present disclosure also contemplates live cell imaging. In some embodiments, fluorescent reporters are visualized. In some embodiments, multiple fluorescent reporters are tracked. The present disclosure also provides long-term cell lineage tracking. In some embodiments, activation of the RNA sensor system produces the effect of a permanent change in reporter expression, such that cells in which the RNA sensor system has previously been activated can be identified later.

In some embodiments, RNA sensor systems can be used to target specific cell types. In some embodiments, RNA sensor systems are modified to be activated by specific cell types. In some embodiments, RNA sensor systems target specific cell types. In some embodiments, RNA sensor systems target tumor cells. In some embodiments, RNA sensor systems deliver payloads that kill specific cell types. In some embodiments, RNA sensor systems deliver payloads that convert one cell type to another cell type. In some embodiments, RNA sensor systems deliver payloads that edit the genome of a cell.

In some embodiments, the RNA sensor system is druggable. In some embodiments, the RNA sensor system is drug sensitive. In some embodiments, the RNA sensor system is activated only in the presence of a single drug or compound. In some embodiments, the RNA sensor system is activated only in the presence of multiple drugs or compounds.

The present disclosure also contemplates the use of the RNA sensor systems described herein for in vitro diagnostic assays. For example, the RNA sensor system can be in a diagnostic assay, wherein the payload comprises a fluorescent protein, a luciferase protein, an antigen or an epitope. In some embodiments, the diagnostic assay is a lateral flow test strip assay.

Now that the present technology has been described in detail, the present technology will be more clearly understood by reference to the following examples. The following examples are included for illustrative purposes only and are not considered to be limiting embodiments of the present technology. All patents and publications mentioned herein are expressly incorporated herein by reference.

Example

Example 1. Method of Example

Unless otherwise stated, the following experimental techniques were employed in the Examples provided herein: Measurement of luciferase activity.

48h after transfection, harvest the culture medium containing secreted luciferase, unless otherwise stated.Use targeted system sea firefly and targeted system Gaussian luciferase assay kit (targeted system) on Biotek Synergy 4 plate reader with injection protocol, use 20 μ L culture medium to measure luciferase activity.All repetitions are carried out as biological replicates.

Transfection of fluorescent sensors.

The day before transfection, cells were plated in 10K in Corning 96-well tissue culture treated plates (black), resulting in about 40% to 50% confluence on the day of transfection. For all fluorescent sensors, HEK293FT cells were transfected with 100 ng total plasmid DNA using TransIT-LT1 according to the manufacturer's instructions (ratio of 1 μg DNA: 3 μL Trans reagent). Unless otherwise stated, ADAR sensors, ADARs, and target plasmids were mixed at equal concentrations (33.3 ng/condition); for experiments without one or more of the foregoing, pUC19 was replaced accordingly to keep the total concentration of DNA at 100 ng.

Confocal microscopy of fluorescent ADAR sensors

48 hours after transfection, all wells were measured by confocal microscopy under the following settings. For each well, 2×2 images at 10× magnification were collected and stitched around the center point. Images were collected in 488nm (32.8% power, 100ms exposure), 561nm (35.2% power, 100ms exposure), 640nm (80% power, 100ms exposure) and bright field channels (25ms exposure).

Quantification of fluorescence signals from images

The image was opened in Matlab and segmented by the watershed in the mCherry channel. For each segmented cell, the total pixel area and the average intensity of the pixel for the mNeon (488nM), mCherry (561nM) and iRFP (640nM) channels were calculated and exported to an aggregated csv file. Csv files were batch processed in R using the following steps: all csv files were merged, conditions with low aggregation areas (fewer cells, or no transfection with the sensor) were merged, the fluorescence background of each channel was subtracted from all conditions in the channel, and the aggregation value of each condition was divided by the area to obtain the average fluorescence intensity. The standard deviation was calculated by comparing the average values in three technical transfection repeats. For the mNeon/mCherry ratio, the average mNeon fluorescence intensity of the condition was divided by the average mCherry value of the same condition. For the fluorescence ratio and ratio fold change values, the error was propagated according to the formula:

Quantification of the percentage of mNeon-positive cells in confocal images.

Some consistent leakage of mNeon was observed in the fluorescent sensor due to very low levels of plasmid contamination or ribosomal slippage. Therefore, detection of mNeon-positive cells was gated at 30 AU above background, and the percentage of mCherry-positive cells expressing mNeon above this threshold was determined. mNeon values were plotted as histograms with kernel density smoothing on a logarithmic basis of 10 to generate the graph in Figure 2E.

RNA extraction and next-generation sequencing of ADAR sensors.

To calculate the editing rate of the ADAR sensor (ADARSENSOR) sensor, 48 hours after transfection, cells were harvested after imaging. Total RNA was extracted using the RNeasy 96 kit (Qiagen) with DNA enzyme treatment. cDNA was prepared with SuperScript IV reverse transcriptase (Invitrogen) and sensor-specific primers. The guide region of the sensor was amplified, indexed and sequenced on the Illumina MiSeq platform. Readings were demultiplexed and compared with each sensor, and the A to I editing rate was calculated with an internal MATLAB pipeline.

Quantification of protein expression

Two days after transfection HEK293FT cells, the HiBiT label in the cell lysate was quantified using the Nano-Glo-HiBiT cleavage detection system (Promega). In order to prepare the Nano-Glo-HiBiT cleavage reagent, the Nano-GloHiBit cleavage buffer (Promega) was mixed with the Nano-Glo-HiBiT cleavage substrate (Promega) and the LgBiT protein (Promega) according to the manufacturer's scheme. The volume of the Nano-Glo-HiBiT cleavage reagent added was equal to the culture medium present in each well, and the sample was placed on an orbital oscillator at 600rpm for 3 minutes. After incubation at room temperature for 10 minutes, a plate reader (Biotek Synergy Neo 2) was used to read out with 125 gains and 2 seconds integration time. The control background was subtracted from the final measurement value.

mRNA synthesis

Before in vitro transcription, DNA templates were obtained by PCR using a targeted forward primer containing a T7 promoter. The sensor mRNA and MCP-ADAR2dd mRNA were transcribed and poly-A tailed using HiScribe ^™ T7 ARCA mRNA kit (NEB, E2065S) supplemented with 50% 5-methyl-CTP and pseudouridine-UTP (Jena Biosciences) according to the manufacturer's protocol. The mRNA was then cleared using the MEGAclear ^™ Transcription Cleanup Kit (Thermo Fisher, AM1908).

Total RNA harvest and quantitative PCR

For gene expression experiments in mammalian cells, cell harvesting and reverse transcription were performed 48 hours after transfection using a modification of the previously described commercial Cells-to-Ct kit (Thermo Fisher Scientific) for cDNA generation. (Joung et al., 2017) Transcript expression was then quantified using qPCR using Fast Advanced Master Mix (Thermo Fisher Scientific) and TaqMan qPCR probes (Thermo Fisher Scientific) with GAPDH control probes (Thermo Fisher Scientific). All qPCR reactions were performed in 10 μl reactions, with two technical replicates in 384-well format, and read out using LightCycler 480 instrument II (Roche). For multiplexing (multiplex) targeted reactions, readouts of different targets were performed in separate wells. Expression levels were calculated by subtracting the housekeeping control (GAPDH) cycle threshold (Ct) value from the target Ct value to normalize the total input, so that ΔCt levels were obtained. Relative transcript abundance was calculated as 2-ΔCt. All repetitions were performed as biological replicates.

Automated generation of affinity sensors

Affinity sensors were generated using python scripts from the following database. (https://github.com/abugoot-lab/ADAR SENSOR) A schematic diagram of the generation of a typical three-affinity guide ADAR sensor with two MS2 hairpin loops and a spacing of 5 nt between guide regions (on the target) is shown in FIG24 .

Animal husbandry and animal protocols

All experiments were performed on female B6(Cg)-Tyrc-2J/J (Albino B6) and NOD.Cg-PrkdcscidIl2rgtm1Wjl Tg(SERPINA1*E342K)#Slcw/SzJ (NSG-PiZ) (The Jackson Laboratory) mice with free access to food and water. NSG-PiZ mice express mutant human SERPINA1 on an immunodeficient NOD scidγ background. All mice were housed in individually ventilated cages in a temperature-controlled animal facility (normal 12:12 h light-dark cycle) and used according to procedures approved by the Committee on Animal Care at MIT.

RADARSv2 Design

Stop codons were engineered in the +1 and +2 frames following the engineered guide RNA region to capture translating ribosomes in all frames. These out-of-frame stop codon designs synergized with the long 5' peptide to significantly reduce background, resulting in approximately 200-fold activation. We selected this sensor design, called RADARSv2, incorporating the structured guide, upstream peptide, and out-of-frame stop codon as the unified architecture for future sensors (Figure 54A).

Example 2. Development of luciferase and fluorescent sensors and detection of EGFP transcripts

Cloning of luciferase sensors

The luciferase sensor was cloned by Gibson assembly of PCR products. The sensor backbone was generated by cloning the Cluc luciferase (Cluc) under the expression of the CMV promoter and the Gaussian luciferase (Gluc) under the expression of the EF1-a promoter on a single vector. The expression of two luciferases on a single vector allows one luciferase to be used as a normalized dose control for knocking down another luciferase, controlling the changes caused by transfection conditions. Order short sensors as primers, and then use T4PNK for phosphorylation and annealing. The annealed oligonucleotides were connected to the backbone using 1 μL T4 DNA ligase (NEB) in a typical 10 μL ligation reaction, 30 ng insert, 50 ng backbone, and 1 μL 10× ligation buffer at room temperature for 20 minutes. Order the long affinity sensor region as Eblocks directly from Integrated DNA Technologies (IDT). The PCR product was purified by gel extraction (Monarch gel extraction kit, NEB) and assembled into the backbone using the NEB HiFi DNA assembly master mix kit with 2.5 μL master mix, 30 ng backbone and 5 ng insert in 5 μL reaction. The reaction was incubated in a thermal cycler at 50 degrees for 30 minutes, and 2 μL of the assembled reaction was converted into 20 μL of competent Stbl3 produced by Mix-and-Go! Competent kit (Zymo) and plated on agar plates supplemented with appropriate antibiotics. After overnight growth at 37°C, colonies were selected into Terrific Broth (TB) medium (Thermo Fisher Scientific) and incubated at 37°C with shaking for 24 hours. The culture was harvested using the QIAprep Spin Miniprep kit (Qiagen) according to the manufacturer's instructions.

The luciferase ADAR sensor contains a 51nt EGFP transcript sensing guide and a Gaussia luciferase (Gluc) payload (Figure 2A). Constitutive Cluc luciferase (Cluc) is contained on a separate transcript, allowing ratiometric control of transfection variance. This dual-reporter, dual-transcript luciferase reporter system targets functional eGFP under the control of a doxycycline-inducible promoter. We tested this ADAR sensor design with a scrambled guide control in the presence or absence of an exogenous ADAR2 deaminase domain (MCP-ADAR2dd (E488Q, T490A)) fused to the MS2 coat protein (Kuttan and Bass 2012; Cox et al. 2017). Plasmids expressing the ADAR sensor were co-transfected with plasmids expressing EGFP or control plasmids into HEK293FT cells. We observed that the ADAR sensor resulted in up to 5-fold increases in normalized luciferase values when relying solely on endogenous ADARs, and when supplemented with exogenous MCP-ADAR2dd (E488Q, T490A), signal activation increased 51-fold (fold change in luciferase expression in the presence of the target/in the absence of the target) (Figure 2B). In addition, we observed that during ADAR sensor induction supplemented with exogenous ADARs, the luciferase signal was comparable to that of constitutively expressed transcripts without an upstream stop codon (approximately 78%, Figure 2C). Therefore, this high protein yield released after activation of the ADAR sensor validates the use of ADAR sensors in applications requiring high absolute payload expression. To determine whether payload expression is dependent on RNA editing, we harvested RNA from cells and quantified editing using next-generation sequencing, and observed that editing of the UAG stop codon in the EGFP-targeted sensor was increased by approximately 24-fold, but the editing increase in the non-targeted sensor was negligible. (Figure 2D)

Cloning of fluorescent sensors

Using pcDNA3.1 (+) cut with HindIII and NotI as the backbone, the fluorescent ADAR sensor parent was cloned into three segments by Gibson assembly. mCherry was amplified from Addgene vector 109427, and T2AmNeon was ordered from IDT as gBlock. All fluorescent ADAR sensors were subcloned into the parent fluorescent plasmid by golden gate cloning using the enzyme Esp3I (the iso-slicing enzyme of BsmBI). Inserts were ordered as complementary chains with overhangs and annealed with phosphorylation, or generated by PCR. The Golden Gate reaction used the NEB BsmBIv2 Golden Gate Assembly Kit, or was assembled in components in a 20 μL reaction containing 25 ng of vector and 2 μL of 1:200 diluted inserts (about 5 to 10 ng). The reaction was alternately thermally cycled between 25°C and 37°C for 1 hour, each for 5 minutes, and then 0.75 μL of the reaction mixture was converted into 12.5 μL of Zymo mixture and Go competent cells. The transformed cells were diluted 1:1 with SOC medium and 10 μL was streaked onto 50 μg/mL carbenicillin agar plates. After incubation overnight at 37°C, single colonies were picked into 4 mL of luria broth (LB) supplemented with 50 μg/mL carbenicillin. Plasmids for luciferase sensors were prepared from the cultures as described above.

This dual reporter, single transcript fluorescent sensor contains such a single transcript fluorescent reporter and a downstream mNeon reporter, which constitutively expresses mCherry upstream of the 51bp eGFP sensor, while the downstream mNeon reporter is activated only after interacting with the target (Figure 3A). HEK293 cells were transfected with non-functional eGFP under the control of a doxycycline-inducible promoter. In the presence of 1 μg/mL doxycycline, cells were also transfected with dual reporters, sensors targeting single transcripts, or non-targeted sensors. Figure 3B shows representative images in experiments with and without targets and with and without exogenous ADARs. Figure 3C shows the quantification of EGFP fluorescence fold changes after target induction; the ratio fold change represents the mNeon/mCherry fluorescence value (fluorescence ratio) in the presence of the target divided by the fluorescence ratio in the absence of the target of the ADAR variant. In the presence of exogenous MCP-ADAR2dd (E488Q, T490A), the targeted ADAR sensor showed a >21-fold increase in activation. Additionally, there was low background activation in the absence of the targeted ADAR sensor. The TAG->TIG editing rate of the sensor TAG codon was also measured in the presence or absence of the target. In the presence of exogenous ADAR, the UAG stop codon was edited at a rate of 9.4% in the presence of the targeted ADAR sensor and 0.2% in the absence of the targeted ADAR sensor, indicating that target-driven editing drives fluorescent payload expression.

To further confirm the evidence of the biological luciferase sensor principle, three guide chains were introduced into HEK293FT cells simultaneously with exogenous EGFP reporter transcripts. ADAR sensors are dual transcripts Gaussian luciferase (Gluc)/sea firefly luciferase (Cluc) transcripts, allowing ratiometric control of transfection variance. Dual reporter, dual transcript luciferase reporter system targets exogenous eGFP reporter transcripts. Designs 2 and 4 are different guides targeting EGFP transcripts. Exogenous ADARs are not introduced into cells. Negative controls are guide chains that do not recognize EGFP. When EGFP expression is analyzed after the introduction of three guide chains, both Design 2 and Design 4 show a significant increase in luciferase expression levels (Fig. 4A). Compared with the negative control scrambled guide, both Design 2 and Design 4 guide chains show a significant increase in luciferase signal (Fig. 4B).

Example 3. Increased transcriptional expression after exogenous ADAR2 administration

In order to determine whether the introduction of additional ADAR molecules would increase the expression of luciferase, 5 guide chains were introduced into HEK293FT cells simultaneously with exogenous EGFP reporter transcripts. Guides 1 to 4 are different guides targeting EGFP transcripts. The negative control is a scrambled control designed to not recognize EGFP. Each guide chain was tested under three experimental conditions. First, each guide was introduced into HEK293FT cells, wherein no exogenous ADAR was introduced into the cells (Fig. 5, blue bars). Next, each guide was introduced into HEK293FT cells simultaneously with the deaminase domain (ADARdd) of ADAR2 (Fig. 5, white bars). Finally, each guide was introduced into HEK293FT cells simultaneously with the ADAR2dd overexpression transcript dPspCas13b-ADAR2dd (Fig. 5, red bars).

When normalized relative to the negative control, without adding any exogenous ADAR molecules, guide strands 1 to 4 showed a 1.5-fold to 2-fold increase in luciferase expression (Figure 5, blue bars). When the deaminase domain of ADAR2 (ADAR2dd) was simultaneously introduced into the cells, guides 1, 2, and 4 showed an increase in luciferase expression similar to that of the same guides in the presence of endogenous ADARs (Figure 5, white bars). In the presence of additional ADAR2dd molecules, guide strand 3 showed a 3-fold increase in luciferase expression (Figure 5, white bars).

When the ADAR2dd overexpression vector was introduced simultaneously with guide strands 1 to 4, luciferase expression increased by at least 2-fold (Fig. 5, red bars). In particular, guide strand 3 showed a 4-fold increase in luciferase expression when compared to the negative control. These data highlight the possibility of using exogenous and modified ADAR molecules to enhance sensor capabilities.

Example 4. ADAR optimization and length screening improves activation of ADAR sensors and reduces background in the presence of target

During the validation of these ADAR sensors, we observed that for some guides, activation could still occur in the presence of exogenous ADARs despite the absence of target RNA (Fig. 2C, Fig. 3B). We set out to determine whether ADAR activity could be optimized to increase activation and reduce background. To optimize the ADAR sensor and minimize this background, we selected and tested a set of different ADAR1 and ADAR2 mutants in combination with a 69nt guide targeting frameshift EGFP transcripts or iRFP transcripts (Fig. 6, Fig. 7). Fig. 6A shows a schematic diagram of the different ADARs tested, including, from left to right, the p150 isoform of ADAR1, the p110 isoform of ADAR1, ADAR2, and the MS2 coating protein (MCP)-ADAR fusion protein (MCP-ADAR). fl = full length. DD = deaminase domain. Catalytic domain mutations are not shown in the schematic diagram; all are in the deaminase domain.

We screened full-length human ADAR isoforms (ADAR1p110, ADAR1p150, and ADAR2) (Galipon et al. 2017; Merkle et al. 2019) and their catalytic deaminase domains, as well as specific mutants designed to destabilize ADAR-dsRNA interactions to reduce nonspecific editing (Cox et al. 2017; Matthews et al. 2016). Although our initial exogenous ADAR selection, MCP-ADAR2dd (E488Q, T490A), performed best on frameshifted EGFP transcripts, several candidates we screened had comparable activation after target co-transfection (Figure 6B), with reduced background in both target groups (Figure 7). We also tested the following stop codon editing rates: (Fig. 8A) MCP-ADAR2dd exogenous supplementation, (Fig. 8B) ADAR1p150 isoform exogenous supplementation, (Fig. 8C) ADAR2 exogenous supplementation and (Fig. 8D) no exogenous ADAR supplementation, carried out in the case of different sensors. The editing rates of these candidates were also tested. The editing rates were calculated by RNA sequencing data, showing that in the presence and absence of the target of the selected ADAR variant for further screening, the UAG stop codon was converted to UIG (Fig. 8).

Since the choice of guide may affect the overall sensitivity of the sensor, we screened the best ADAR candidates on multiple guide sequences and targets in the orthogonal group (Figure 10). First, we observed activation above background only in correctly matched ADAR sensors and target transcripts (Figure 9). ADAR1p150 had the highest fold activation on 3 of the 4 targets, which was driven by the generally low overall background signal in the absence of the target, while MCP-ADAR2dd (E488Q, T490A) performed best on the EGFP target due to its generally high absolute signal level, but was affected by the higher background on other targets, which reduced its overall activation (Figure 6, Figure 10, Figure 11, Figure 12).

The background and activation of the following sensor panels were also analyzed: (FIG. 13A) MCP-ADAR2dd exogenous supplementation, (FIG. 13B) ADAR1p150 isoform exogenous supplementation, (FIG. 13C) ADAR2 exogenous supplementation, and (FIG. 13D) no exogenous ADAR supplementation, in the presence of different sensors.

The best exogenous ADAR amount (Figure 14) was also tested. In titration experiments, in tetracycline inducible IL6 experiments, HEK293 cells were transfected with different amounts of MCP-ADAR2dd from 10ng to 100ng and different amounts of 3-site binding affinity IL6 sensor chains from 10ng to 100ng, wherein the amount of target (IL6) was fixed at 20ng. The fold change represents the normalized luciferase ratio between + target group and - target group.

Example 5. Use of the RNA Editing for Programmable A to I (G) Replacement (REPAIR) System in Biosensors

Next, the RNA editing (REPAIR) system of programmable A to I (G) replacement was evaluated, regarding the feasibility of whether it can be used as a mechanism to trigger these genetic sensors. The REPAIR system is composed of a fusion of the catalytically active Cas13b enzyme fused to the deaminase domain of the ADAR2 molecule. The catalytically inactive Cas13b enzyme incorporated with the K370A mutation is also fused to the deaminase domain of the ADAR2 molecule to form a fusion protein (REPAIR K370A) without Cas13b activity.

Without adding any exogenous ADAR molecules, two guide chains designed to target EGFP and negative controls designed to not target EGFP were introduced into HEK293FT cells (Figure 15, blue bars). Guide chains were also introduced into HEK293FT cells simultaneously with REPAIR molecules (Figure 15, red bars) or with catalytically inactive REPAIR K370A molecules (Figure 15, white bars). Guides 1 and 2 designed as Cas13b targets show that, compared with cells that rely only on endogenous ADAR expression, luciferase expression in cells is not improved. However, when Cas13 activity is turned off (REPAIR K370A), when compared with cells that rely only on endogenous ADAR expression, guide 1 shows a 4-fold increase in luciferase expression. This increase in expression emphasizes the potential for the REPAIR system to be used in combination with the genetic sensor of the present disclosure.

Example 6. Testing of different guide strand properties for improving luciferase expression

Next, the characteristics of the guide strand design were varied to determine which variables could be adjusted to improve the efficiency and expression of the genetic sensor. FIG. 16 shows a heat map showing the multiple luciferase expression for different guide/ADAR combinations. The exogenously introduced full-length ADAR (column 2) consistently showed the highest multiple increase in luciferase expression. However, this was not the case for the guide strand designed as an MS2 agonist, which showed essentially consistent expression regardless of whether the ADAR was introduced.

Selecting the lower background level from Example 4, we began to further optimize the sensor with the ADAR1p150 construct. To improve both binding stability and target search time, increasing guide lengths centered on premature stop codons (Qu et al. 2019) were tested for constitutive iRFP targets. As the guide length increased from 51nt to 600nt, sensor activation increased from 2.2-fold to 18.22-fold (Figure 17B). In addition, when the target was present, the distribution of mNeon expression levels was observed in each cell at all guide lengths (Figure 17C, Figure 18A), with a larger number of mNeon (+) populations at longer guide lengths. At the same time, in the absence of the target, the percentage of mNeon (+) cells at all guide lengths remained at <5%. At a guide length of 600nt, 66.7% of mNeon (+) cells were observed in the presence of the target, and 1.6% of mNeon + cells were observed in the absence of the target, indicating a strong ability to separate cell populations based on target mRNA expression.

The two best performing ADARs from Example 4, ADAR1p150 and MCP-ADAR2dd(E488Q, T490A), achieved optimal signal by either reducing background or increasing activation. Because MCP-ADAR2dd(E488Q, T490A) had the highest activation in the case of luciferase sensors when ADAR enzymes were engineered ( FIG. 20 ), we hypothesized that the reduction of background by the guide engineering strategy would result in optimized sensor activation when coupled to MCP-ADAR2dd(E488Q, T490A).

A new sensor targeting IL6 mRNA, a transcript barely expressed in HEK293FT cells (Uhlén et al. 2015), was designed, enabling the generation of an integrated line with IL6 mRNA under the control of a dox-inducible promoter and the ability to supplement IL6 mRNA by exogenous transfection to modulate low levels of IL6 expression for sensitivity testing (see Figure 20).

Due to the increase in background signal in the absence of target, likely due to read-through of stop codons within the longer guide region (Figure 21 (A) MCP-ADAR2dd (E488Q, T490A) exogenous supplementation, (B) ADAR1p150 isoform exogenous supplementation, (C) ADAR2 exogenous supplementation, and (D) no exogenous ADAR supplementation), we modified the guide region by introducing an MS2 hairpin loop to block aberrant translation (Chao et al. 2008), which provided the additional benefit of recruiting MCP-ADAR2dd (E488Q, T490A) protein to the guide: target duplex (Figure 23). We repeated our studies of increased guide regions with a luciferase sensor targeting IL6 transcripts and found that the fold change of activation of ADAR sensors longer than 81 nt guide using the MCP-ADAR2dd (E488Q, T490A) construct was significantly reduced when only one binding site was present on the guide strand (Figure 22). However, structural additions and modifications to the guide strand design performed to determine whether the addition of an MS2 hairpin loop and an additional engineered guide binding region (referred to as an "affinity binding region") on the sensor/guide strand enhanced sensitivity showed a significant increase in ADAR sensor activation (Figure 22). Figure 22 shows the results of an experiment to determine the fold change of different formats of the guide/sensor strand. HEK293 cells were transfected (Lipofectamine 3000, Thermo Fisher Scientific) with sensors comprising the original reverse complement (of the IL6 target) and an MS2 hairpin with an additional affinity region (x-axis; 51 bp sensor, sensor with a continuous 171 bp binding region, sensor with a 171 bp affinity region separated by two MS2 hairpins, sensor with a single continuous 225 bp binding region, sensor with a 225 bp binding region separated by four MS2 hairpins, sensor with a continuous 279 bp binding region, sensor with a 279 bp binding region separated by six MS2 hairpins, and non-targeted sensor). The separation of MS2 hairpin to the binding region significantly improves target expression in all different affinity region lengths. Test several types of ADAR proteins, including full-length ADAR2 (ADAR2FL), the p150 isoform of ADAR1, endogenous ADAR1 and the MS2 coated protein-bound fusion protein fused with the deaminase domain of people ADAR2. The multiple change (y axis) of expression is calculated by the original luciferase value under + target conditions divided by the original luciferase value under - target conditions.

ADAR sensor activation was highest in the case of a 5-site affinity binding guide, reaching about 70 times of activation, and the background was significantly lower than the uninterrupted guide design (Figure 27). The affinity binding guide improved the performance of all exogenous ADAR constructs, but with the supplementation of MCP-ADAR2dd, the greatest performance improvement was shown. Affinity binding guides with 5 or 7 binding sites can produce detectable activation only by relying on endogenous ADAR. Figure 23 shows several possible MS2 hairpin/affinity modification forms, and Figure 24 provides a design guide for affinity sensors and an easy-to-use software program for automatically generating affinity sensors for inputting target sequences (github.com/abugoot-lab/ADAR SENSOR).

We tested whether the best performing IL6-targeting engineered guide RNAs could utilize endogenous ADARs to sense synthetic IL6 targets transfected into cells. Although exogenous ADAR1p150 supplementation improved the performance of RADARSv2, we observed over 50-fold activation of payloads with endogenous ADARs (Figure 52).

To further explore the concept of affinity binding guides, we varied the spacing between binding sites (5, 30 and 50 nt). The length between binding sites represents the number of nucleotides from just before the MS2 hairpin on the guide strand to the other complementary region. Tighter binding sites on the target transcript resulted in the highest activation (Figure 25).

We also explored whether the improvement of affinity binding guide can be combined with orthogonal methods to block translation read-through, such as additional stop codons. We compared the single-terminated seven-affinity region sensor or the double-terminated seven-affinity region sensor with the additional stop codon in the final affinity region (Figure 26). Figure 26B shows the fold change of luciferase payload between the single-terminated sensor and the double-terminated sensor. Compared with the single-terminated sensor, the double-terminated sensor showed a significant increase in fold change, depending on ADAR. We found that the additional stop codon increased the fold activation of the 7-site affinity binding guide, exceeding the performance of the 5-site affinity binding guide (Figure 28A). This improvement is driven by both the increase in the editing rate of the stop codon and the reduction in the background activation rate in the presence of the target (Figure 28B, Figure 8).

Despite the abundance of CCA codons on potential target transcripts, we explored whether progressive modifications of affinity guide design would allow for increased mismatch tolerance to improve targeting flexibility. We also tested target mismatch tolerance among 16 possible mismatches (from conventional CCA) between the initial 51bp sensor, the three-affinity sensor, and the five-affinity sensor (Figure 29A). We designed 16 targets covering all nucleotide changes (nCn) to 5' cytosine or 3' adenosine. Guides containing UAG in these different codons were tested, and we found that guanine or cytosine mismatches were generally better tolerated than adenosine or uridine mismatches (Figure 29B). In addition, the 5-site affinity binding guide ADAR sensor design had the best activation fold change (Figure 30) in addition to the ACA and ACU targets.

The modularity of protein payloads also allows small payloads, such as Hibit payloads (Schwinn et al. 2018), which allow in vivo circularization of transcripts. Circular RNA presents a platform for enhanced residence time and minimal immunotoxicity (Katrekar et al. 2019), and we hypothesized that ADAR sensors with small payloads could be circularized to exploit these properties (Figure 31A). Two forms of short circular sensors were initially developed. Conventional circular sensors are dual Twister ribozyme systems (Litke and Jaffrey 2019) backbones driven by a U6 promoter that circularizes in vitro in the presence of RtcB ligase. Conventional circular sensors also contain a Hibit tag with a stop codon at the c-terminus of Hibit. We found that the circular ADAR sensor expresses Hibit in a target (IL6)-specific manner (Figure 31B). For signal amplification, we expanded these circular ADAR sensors into endless ADAR sensors by removing the stop codon after the payload and inserting 2A peptides at either end of the Hibit tag, allowing expression by rolling circle translation (RCT) (Abe et al. 2015). These rolling circle translation sensors are similar to conventional circular sensors, except that the stop codon in Hibit is removed and the T2ToA peptide is inserted to allow ribosome read-through in a circular manner. These circular sensors (targeting IL6) were transfected into HEK293 cells, and sensor activation of sensors of different lengths was compared. Longer sensors consistently improve the change in sensor activation fold. We found that rolling circle sensors can express proteins in a target-specific manner with minimal background leakage (Figure 31B).

We examined the effects of mRNA modifications on fold changes in sensor activation. Synthetic mRNA has become an available therapeutic modality, but there are no methods to control the expression of its payload in a transcript-specific manner. We explored the application of synthetic mRNA ADAR sensors for transcript-specific expression in a mouse model expressing human SERPINA1 transcripts in mouse hepatocytes. When delivering mRNA, the incorporation of base modifications such as 5’methylcytosine (5mc) and pseudouridine (Ψ) is critical for reducing host immune responses (Kauffman et al., 2016), but these modifications may interfere with ADAR activity and affect mADR sensor function.

The incorporation of 5-methylcytosine and pseudouridine was analyzed in HEK293 cells. HEK293 cells were supplemented with MCP-ADAR2dd as a plasmid or directly as mRNA 24 hours before mRNA transfection. An IL6 sensor with tetracycline-inducible IL6 was also used. We found that increased Ψ amounts reduced the activation of the ADAR sensor, while 5mc had better tolerance, with 25% of 5mc incorporation having the highest signal activation (Figure 32).

Using our inducible IL6 system to measure sensor activation, we further determined the effects of different levels of incorporation of a wider set of chemically modified bases on mRNA RADARS activation, and transfected the modified IL6 sensing mRNA RADARS with exogenous ADAR1p150 (Figure 53A) or endogenous ADAR (Figure 53B). We found that all tested modifications reduced mRNA RADARS activation, likely due to interference with the ability of ADAR1p150 to edit modified mRNA. Among the modifications, we found that in the case of exogenous ADARp150, 50% incorporation of modified bases (e.g., 5-methylcytosine or 5-methyluridine) was the best tolerated, while 100% incorporation of 5-methylcytosine was the highest activation in the case of endogenous ADARs. To determine whether this level of modification is sufficient to reduce host immune responses, the induction of interferon beta-related genes by chemically modified mRNA RADARS was measured. We observed that even at a 25% incorporation level of modified bases, we achieved minimal induction of inflammatory genes (Figure 53C).

We tested a sensor containing a 51 nt IL6 transcript sensing guide in front of a Gaussia luciferase (Gluc) payload (called RADARSv1) in combination with constitutive C. luciferase (Cluc) on a separate transcript to provide ratiometric control of transfection variance (Figure 54A). With this RADARSv1 design and co-transfection with exogenous ADAR1p150, we observed approximately 5-fold activation (Figure 54A), as quantified by the increase in the Gluc/Cluc ratio in the presence of exogenous IL6 target expression (Figure 54B, changes in the Gluc/Cluc ratio under different conditions, defined as RADARS fold activation in the rest of the manuscript). Since ADAR1p150 prefers long double-stranded RNA as a substrate, we titrated the length of the guide region from 51nt to 279nt, resulting in a moderate increase in activation at 81nt, while activation at longer lengths decreased (in the absence of target RNA) due to increased background payload expression (Figure 54A, Figure 54B).

In the absence of the target, three strategies were used to prevent dsRNA formation, which is due in part to translational read-through and self-folding. First, we introduced multiple MS2 hairpin loops interspersed with binding sites into the guide region to generate secondary structure to prevent self-folding and enable multivalent binding. We optimized these modified guides, called organized guide RNAs (modified guide RNAs), by varying the number of MS2 loops and binding sites on the guides. The highest RADARS activation was achieved with modified guide RNAs containing five binding sites interspersed with MS2 hairpin loops, which reduced background payload expression in the absence of the target and achieved approximately 20-fold activation compared to the uninterrupted guide design (Figure 54A, Figure 54B)

Second, we increased the length of the translatable open reading frame (ORF) preceding the engineered guide RNA to promote termination and prevent ribosome reinitiation, which is known to depend on the length of the upstream ORF. We tested 5' peptide lengths at residues 0, 100, and 200 and found that at residue 200, we were able to significantly reduce background translation read-through (Figure 55A), achieving >100-fold activation.

Finally, we engineered stop codons in the +1 and +2 frames following the engineered guide RNA region to capture translating ribosomes in all frames. These out-of-frame stop codon designs synergized with the long 5' peptide to significantly reduce background, resulting in approximately 200-fold activation. We selected this sensor design, called RADARSv2, incorporating the structured guide, upstream peptide, and out-of-frame stop codons as the unified architecture for future sensors (Figure 54A, Figure 55B).

We evaluated our RADARSv2 design on exogenously expressed IL6, EGFP, and neuropeptide Y (NPY) targets by tiling the modified guide RNAs on 14 CCA sites spaced apart on the transcripts. We found that although RADARS activation depends on the hybridization site selected for a given target, most sensors have significant payload activation in the presence of their targets, with up to 1000-fold activation, which demonstrates the generalizability of the RADARSv2 design (Figure 56A). In order to determine whether payload expression is caused by RNA editing, we harvested RNA from cells transfected with a set of 14 modified guide RNAs targeting IL6, and quantified the editing with next-generation sequencing. In the presence of target transcripts, all 14 modified guide RNAs had more than 15% editing, with an average of 35.1% +/- 11.4%. In the absence of target transcripts, 13 of the 14 modified guide RNAs had the lowest editing (0.32% +/- 0.34%). We also observed minimal editing of non-targeted sensors, reaffirming that RNA editing by RADARS sensors requires specific engineered guide RNA target recognition ( FIG. 56B ).

Example 7. Quantification and correlation analysis of genetic sensors

In order to determine whether the above genetic sensor can be accurately used as a "dose sensitivity" sensor, inducible EGFP transcripts are introduced into HEK293FT cells. The EGFP transcript is placed under the control of a doxycycline inducible promoter, and then the cells are exposed to 0ng/mL, 8ng/mL, 40ng/mL or 200ng/mL doxycycline to change the expression of the EGFP transcript. In HEK293FT cells not introduced into exogenous ADAR (Fig. 6A), two guide chains do not show significant differences in luciferase activity. However, guide chain 3 shows a trend of dose sensitivity (Fig. 33A, white bars).

When full-length ADAR2 and guide strands were introduced into cells simultaneously (Figure 33B), guide strand 3 showed obvious dose sensitivity, while guide strand 1 had some dose sensitivity, but to a lesser extent (Figures 33B to C, white bars). Cells given 200ng/mL doses of doxycycline showed luciferase activity similar to that of cells constitutively expressing EGFP transcripts. As the dose of doxycycline decreased, the luciferase activity identified in the cells decreased accordingly. The same trend was also seen in cells simultaneously exposed to full-length ADAR2 and guide strand 1 (Figure 33B, blue bars). Figure 34 describes dose-dependent reporter expression when guide strands 1 and (Figure 34A) and guide strand 3 (Figure 34B) targeting EGFP were introduced into cells. The multiple activation of luciferase also follows this dose-dependent trajectory (Figure 33C). Figure 35 describes the level of luciferase activity as a function of GFP fluorescence. These results indicate that these genetic sensors may be important as quantitative sensors of transcription levels, rather than just "on-off" sensors.

We compared the target expression of the highest transcripts per million (TPM) gene (RPS5) and the lowest expressed gene (KRAS) by siRNA perturbation experiments to compare the performance of RADARSv2 with changes in exogenous ADAR1p150 and endogenous ADARs. We observed that in the absence of exogenous ADARp150 supplementation, both sensors detected siRNA knockdown, however, with the RPS5 sensor benefiting more from exogenous ADARs (Figure 57), suggesting that larger expression changes may benefit more from exogenous ADARs.

To further examine the quantitative value of the ADAR sensor, we generated a wide range of expression levels using both transfected and virally integrated forms of our inducible IL-6 expression system and measured the luciferase response of the 7-site affinity binding guide with double stop codons (Figure 36A). We found that ADAR sensor luciferase activation was linearly correlated with the concentration of the target transgene as determined by qPCR (Figure 37, Figure 38, R2 = .96). Accordingly, RNA editing of the first stop codon in the ADAR sensor guide had a strong correlation with gene expression levels (Figure 39), showing that the ADAR sensor can quantitatively sense transcripts at both RNA editing and payload levels.

To enable single-cell measurements with RADARSv2, we engineered the fluorescent payload for microscopy and flow-based readout (Figure 58A). We designed the fluorescent sensor as a single transcript containing an upstream mCherry normalization control separated from the optimal EGFP-targeted engineered guide RNA by a self-cleaving peptide P2A sequence, followed by a self-cleaving peptide sequence T2A, which preceded the mNeon payload. We transfected HEK293FT cells with EGFP-targeted RADARS in combination with or without exogenous ADAR1p150 or frame-shifted non-fluorescent EGFP target transcripts. We observed mNeon fluorescence signals by microscopy in the presence of target transcripts and observed negligible background in the absence of target transcripts (Figure 58A). Quantification of fluorescent signals by flow cytometry showed that the distribution of mNeon/mCherry ratios changed from 1.00% mNeon/mPCherry-positive cells in the absence of target transcripts to 56.1% mNeon/MPCherry-positive cells in the presence of target transcripts, with the geometric mean ratio increasing 38-fold ( Figure 58B , Figure 59A to C ).

To further explore the quantitative accuracy of RADARS, we used transfected and virally integrated forms of the tetracycline-inducible IL-6 expression system to generate a wide range of expression levels and measured the luciferase response with the best IL-6-sensing modified guide RNA. The activation of RADARS luciferase is quantitative and linearly correlated with the concentration of the target transgene determined by qPCR (Figure 60A, Figure 60B, R2=0.95). In addition, the activation of RADARS is invariant to the amount of transfected sensors, with a robust activation rate in a large titration of sensor load, allowing the total sensor output to be adjusted independently of the overall sensor activation (Figure 60C).

We further designed sensors for a group of 10 different transcripts ranging from about 10,000 to about 10 in HEK293FT cells for verification. For each transcript, we compared 8 different targeted modified guide RNAs with 8 non-targeted modified guide RNAs. After siRNA transfection, we verified the knockdown of these 10 genes by qPCR (Figure 61), and observed that the RADARS signal of each transcript was significantly reduced compared with the non-targeted sensor control. The robustness of the modified guide RNA is related to the following expression: for highly expressed genes, most of the eight different targeted modified guide RNAs detected target transcript knockdown, but as the expression level decreased, the modified guide RNA that was successfully detected to be knocked down was less (Figure 63A). Although the sensitivity of RADARS decreases at lower TPM, at least one of the eight tested modified guide RNAs can significantly detect transcript knockdown (Figure 62A). These data show that RADARS is sensitive to relative changes in gene expression at a wide range of expression levels. Measuring the editing rate of the UAG stop codon for the best performing sensor for each of the 10 target transcripts (Figure 62B), we found that the overall editing rate was low, but when the targets of all 10 genes were knocked down, the editing rate was statistically significantly reduced. Because RADARS are overexpressed relative to endogenous targets with low turnover rates, the editing rate of the stop codon is less sensitive to copy number fluctuations of the target.

Next, we tried to determine the sensitivity of RADARS by measuring the changes in gene expression of endogenous targets. In order to test RADARS on a series of endogenous transcript expression levels, we applied sensors to measure transcriptional downregulation by siRNA (Figure 63). We used a commercially validated siRNA library targeting 6 endogenous genes, which was distributed in highly expressed genes RPL41, GAPDH and ACTB, and moderately to lowly expressed genes HSP90AA1, PPIB and KRAS. For each transcript, we first compared the highest sensitivity of 8 different modified guide RNAs to knockdown (Figure 63A). Next, we titrated the amount of siRNA to produce a series of expression levels, which was confirmed by qPCR, and the best modified guide RNA supplemented with exogenous ADAR1p150 was used to track the changes in expression levels. We observed that for all six genes, RADARS tracked the transcript levels measured by qPCR with high Pearson correlation (R>0.86, Figure 61B). We found that for KRAS expressed at 13 TPM (transcripts per million) in HEK293FT cells (Karlsson et al., 2021), the raw RADARS fold activation deviated from the fold change measured by qPCR, likely due to a loss of sensitivity at such low expression levels. However, the RADARS response was still highly correlated with KRAS levels determined by qPCR (R = 0.93, Figure 61B).

Next, we used a cellular heat shock model that causes upregulation of heat shock family genes to study whether RADARS can sense the upregulation of endogenous transcripts. We designed a RADARSv2 modified guide RNA targeting HSP70 (dynamic heat shock response protein) and transfected it into HeLa cells with exogenous ADAR1p150 before exposing the cells to 42°C heat shock (Figure 64A). RADARS has strong consistency with qPCR, with a 7.2-fold increase in HSP70 transcript expression levels measured by qPCR, of which the best HSP70-targeted modified guide RNA produced a 5.9-fold activation in response to heat shock (Figure 64B). These results show that RADARS is sensitive to the upregulation of endogenous transcripts and can detect relative gene expression changes with high accuracy.

Example 8. Logic Gate

We also set out to determine whether the ADAR sensors of the present disclosure could be multiplexed into a logic system that could contain AND gates and OR gates. These AND-OR methods are shown in Figure 40A (AND) and Figure 40B (OR). The AND gate can fully deliver the payload only when both target strands are present. However, the OR gate can deliver the payload when either of the target strands is present, but not both. To generate a basic AND gate, we connected two single 51nt guides that target EGFP and IL6, respectively, in series with an MS2 hairpin loop. However, due to a combination of low signal and background read-through, this design performed poorly (Figure 40A).

To improve the AND gate signal, we used the RADARSv2 design and found that the resulting AND gate sensor behaved in a target-specific manner, requiring both targets to reach full activation, with only a small amount of leakage under single target conditions (Figure 65A, Figure 65B). In the presence of both target transcripts, the AND logic chain showed 36-fold activation, while in the presence of only one target RNA, it only showed 1.3 to 1.5-fold activation (Figure 65B).

To transform the OR gate logic, we co-transfected two five-binding site affinity sensors targeting EGFP and IL6. These sensors responded to EGFP or IL6 target transcripts in a manner consistent with the OR gate (Figure 41B). The OR logic sensor showed a significantly improved fold change when each gene was present alone, but not when both genes were present. In general, these results indicate that the modularity of the ADAR sensor enables logical calculations to be performed on mRNA in living cells.

To improve the OR gate logic ( FIG. 66A ), we co-transfected two engineered guide RNAs RADARSv2 (upstream ORF, out-of-frame stop codon) targeting EGFP and IL6 transcripts and found that the sensor responded to either EGFP or IL6 target transcripts in a manner consistent with the OR gate ( FIG. 66B ).

Example 9. Use of ADAR sensor to induce apoptosis of target cells

In order to determine whether the ADAR sensor of the present disclosure can be used for non-reporter payload, we determined whether the ADAR sensor can induce apoptotic cell death in the targeted cell population. In order to apply the ADAR sensor to cell state-specific killing, we used the iCaspase-9 (Straathof et al. 2005) related to treatment to transform the payload (Figure 42A). The iCaspase payload was transformed into a sensor chain that can target human IL6. Mammalian cells were transfected with caspase ADAR sensors, targets and MCP-ADAR2dd. 24 hours after transfection, cells were divided into fresh culture medium at 1:5, and the + drug samples were supplemented with 10nM AP20187 (Sigma Aldrich). 24 hours after additional growth, cell viability was measured by CellTiter-Glo luminescent cell viability assay (Promega). The control caspase was a sensor chain with a disordered sensor region (i.e., it does not specifically target IL6) and a caspase without an intervening stop codon. Using CellTiter-Glo assay (Promega), cell death was measured as the fold change of the cell lysate luminescence value of the + target group divided by the - target group. We found that the IL6 sensor fused with double stop codons and seven affinity guides before caspase selectively killed cells expressing IL-6, and the toxicity was minimal in the absence of IL-6 induction (Figure 42B, C). IL6 responsive caspase showed a significant improvement in inducing apoptotic cell death, indicating that ADAR sensors can be used to induce cell death in target cell populations. See Figure 42B. In cells with and without target transcripts, the cell survival percentages in cells treated with IL6 responsive caspase and cells treated with caspase and without stop codons were also analyzed (Figure 42C).

Next, we used a highly specific modified guide RNA targeting SERPINA1 for cell-specific killing by combining the modified guide RNA with the iCaspase-9 payload (Figure 67A) (Straathof et al., 2005). We co-transfected SERPINA1-iCaspase9 RADARS with ADARp150 into A549, HeLa and HepG2 cells and measured cell viability 48 hours after transfection. We found that SERPINA1-targeted RADARS-iCaspase selectively killed HepG2 cells, while the toxicity was minimal in other cell types, and the non-targeted negative control showed no differential death (Figure 67B, Figure 67C).

Example 10. Use of ADAR sensors to track cell states and cell types

To determine whether the developed ADAR sensor can be used to track cell states, the heat shock response of HeLa cells was first tested. Two groups of Hela cells were transfected with ADAR sensors, which had guides designed to target heat shock family genes (including HSP70 and HSP40). HSP70 and HSP40 can be upregulated in an in vitro heat shock model (Figure 43A, B). ADAR sensors designed with 5-site or 7-site affinity binding guides detected the upregulation of both HSP70 and HSP40 in cells exposed to heat shock (Figure 43A). HeLa cells (ATCC CCL-2) were transfected with HSP40 or HSP70 ADAR sensors. 24 hours after transfection, a portion of the cells were moved to 42 degrees Celsius (5% CO2) for 24 hours. The culture medium was harvested at the end of 24 hours of heat shock and luciferase measurements were performed. To control the non-specific changes in translation caused by heat shock, we transfected the scrambled non-targeted guides. Normalized to the non-targeted guide, we found up to 3-fold activation of ADAR sensors in response to heat shock ( FIG. 43C ).

We repeated the heat shock experiment with the RADARSv2 design (Figure 64), delivering only the RADARSv2 sensor without ADAR supplementation and the designed sensor. We found that the best HSP70 sensor (CCA42) could use endogenous ADARs in Hela cells to track the upregulation of HSP70 during heat shock (Figure 68), and thus demonstrated the feasibility of configuring a single-component RADARSv2 system using endogenous ADARs.

Cell type differences represent major changes in gene expression in tissues. Therefore, we set out to determine whether ADAR sensors could accurately track cell type differences. First, to identify marker transcripts for cell type differentiation, we performed differential gene analysis in HEK293, Hela, and HepG2 cells (Figure 44A), selecting SERPINA1 (a liver serine protease inhibitor with a therapeutically relevant pathogenic variant) et al., 2019) as a marker expressed only in HepG2 cells and not in other cell lines (Figure 44B). We designed a panel of ADAR sensors with guides targeting SERPINA1 and tested their ability to distinguish between HepG2 and Hela cells, finding that the CCA30 guide design had the highest activation fold change in both HepG2 and Hela cells (Figure 44C). We transfected the SERPINA1 (CCA30) targeted sensor into three different cell types along with a non-targeted scrambled sensor designed to control for background ADAR editing, transfection variance, protein production, and secretion differences between the three cell types. Each cell type was transfected with the CCA30 SERPINA1 sensor with a 5 affinity region linked to an MS2 hairpin, with or without MCP-ADAR2dd. Fold changes (Figure 44D) were calculated from raw luciferase values for the SERPINA1 sensor, normalized by the scrambled non-targeted sensor to account for differences in protein production/secretion and background ADAR kinetics between cell types, and then normalized to ratios in HEK cells for cell type comparison.

Normalized fold changes in editing rates in all three cell types were also analyzed in the presence of endogenous ADARs, supplementary ADARs, and controls (Figure 45). Multiple CCA sites on the SERPINA1 transcript were also used as targets.

To mimic liver-specific cell targeting in vitro, we expressed human SERPINA1 transcripts in Hepa-1-6 cells, synthesized the top CCA SERPINA1 sensors as mRNA in vitro, and transfected the mRNA sensors into Hepa-1-6 cells alone. We found that both CCA30- and CCA35-targeted SERPINA1 sensors were able to recruit endogenous ADARs in Hepa-1-6 cells to sense the induction of SERPINA1 transcripts (Figure 46).

To evaluate the cell type differentiation of RADARSv2, we first used the modular nature of RADARS to transform the permanent genetic labeling system of cell populations. We designed a double loxP system for conditional, permanent labeling of cells with EGFP after Cre expression and tested this reporter in combination with ADAR1p150 and a modified guide RNA targeting IL6 with a Cre payload in HEK293FT cells. After IL6 induction, we observed significant production of EGFP protein, with minimal signal in the absence of target RNA (Figure 69A, Figure 69B).

Then, using the RADARSv2 design, we identified SERPINA1 as a differentially expressed marker gene in the liver-derived cell line HepG2 and compared it with two non-liver cell lines (A549 and HeLa) without SERPINA1 expression (Karlsson et al., 2021). Using a modified guide RNA targeting SERPINA1 to selectively activate Cre in HepG2 cells (Figure 70A), we co-transfected the sensor with ADAR1p150 and Cre loxP reporters into HepG2, Hela, and A549 cells and performed benchmark activation for non-targeted RADARS constructs. The non-targeted modified guide RNA did not show reporter activation in any cell type, while the targeted modified guide RNA showed significant EGFP reporter activation only in HepG2 cells (Figure 70B, Figure 70C). These results show that the RADARS system can distinguish cell types based on specific markers, and that modified guide RNAs and payloads can be modularly combined for cell type-specific expression of multiple transgenes.

Example 11. In vivo use of ADAR sensors

To determine whether ADAR sensors can be used in vivo, we next tested the SERPINA1 sensor in mice. We synthesized ADAR sensors targeting the CCA30 and CCA35 sites of human SERPINA1 in a construct expressing Akaluciferase (Akaluc), which allows simple non-invasive luminescent imaging to determine cell-specific ADAR sensor activation (Yeh et al. 2019) (Figure 47). Before bioluminescent imaging, 8- to 10-week-old albino B6 and NSG-PiZ mice were anesthetized with 3% isoflurane and injected with 5 μg of synthesized mRNA using an in vivo jetRNA transfection reagent (Polyplus) via retro-orbital injection. 18 hours after injection, mice were anesthetized again with 3% isoflurane and immediately administered 100 μl of 15 mM AkaLumine HCL (Sigma Aldrich) for imaging. Ventral bioluminescent images were obtained using the IVIS Spectrum In Vivo Imaging System (PerkinElmer). The following conditions were used for image acquisition: exposure time = 60 seconds, pixel merging (binning) = medium: 4, field of view = 12.5 × 12.5 cm, and f/aperture = 1. Bioluminescent images were analyzed using Living Image 4.3 software (PerkinElmer). Since albino B6 mice do not express human SERPINA1, it represents a negative control (no CCA30 or CCA35 binding sites). No additional ADAR enzymes were applied to mice to determine whether endogenous ADAR alone can edit the applied ADAR sensor. Relative to NSG-WT mice, SERPINA1 senses that mRNA RADARS design has significant activation of Akaluc expression in NSG-PiZ mice (p = 0.007, N = 3 mice, one-way ANOVA), and we did not observe significant differences between the two strains under the background luciferase of substrate only and constitutive Akaluc mRNA RADARS conditions (Figure 71A to C).

In addition to ADAR sensors targeting the CCA30 and CCA35 sites of SERPINA1, we also designed an Akaluc payload with a constitutive ADAR sensor or a scrambled non-targeted guide that expresses Akaluciferase lacking a stop codon. (See Figure 47B). Three sensor systems were tested. Relative to WT mice, the SERPINA1-sensing ADAR sensor design had significant activation of Akaluc expression in NSG-PiZ mice (p=0.04, N=2 mice, two-tailed unpaired t-test), and the control guide had no significant differences between the two strains (Figure 47B, Figure 48). This activation shows that the ADAR sensor can be delivered as a synthetic mRNA to sense the cell state in vivo with endogenous ADARs.

Further considerations

Analysis of public tissue gene expression data (GTEx Consortium 2013) showed that 34 of 37 tissues could be distinguished by a single gene with a sensor with 3-fold sensitivity (Figure 49A), while 3 additional tissues were classified by a combination of genes, demonstrating direct application of both the sensitivity and logical input of the ADAR sensor (Figure 49).

Example 12. Testing for off-target effects and interference with routine cellular processes

Since the RADARSv2 mechanism involves the formation of a long hybridization zone between the sensor's modified guide RNA and the target transcript, we investigated whether this duplex would disrupt the target transcript level by Dicer knockdown or transcript stabilization. We compared the target expression of each endogenous transcript knocked down by siRNA between the optimal modified guide RNA conditions and the non-targeted modified guide RNA sensor (Figure 63A), and found that there was no significant change in target transcript expression (Figure 72A). In addition, to determine that the resulting modified guide RNA-target hybridization would not interfere with endogenous translation, we co-transfected HEK293FT cells with ADAR1p150 and modified guide RNAs targeting ACTB or PPIB, and quantified target protein levels by Western blotting. Similar to mRNA levels, we observed that under the conditions of targeted modified guide RNAs, the protein levels of ACTB and PPIB did not change relative to the non-targeted modified guide RNAs, indicating that RADARS had no significant effect on target expression (Figure 72B, Figure 72C to D).

The formation of dsRNA in cells can activate immune response pathways, and therefore we next used both ACTB and GAPDH as normalization genes to measure the upregulation of the major endogenous innate immune signaling pathways (IFNB1, MDA5, OAS1, and RIG-1) induced by RADARS involved in dsRNA response by qPCR (Figure 73A, Figure 73B). For comparison with the positive control, we tested RADARS constructs targeting ACTB, PPIB, and RPS5 and exogenously introduced IL6 transgenes and high molecular weight poly (I: C), which act as dsRNA analogs and activate these four pathways. We found that the RADARS construct did not significantly upregulate any of the four dsRNA response transcripts, while poly (I: C) caused significant activation of all four pathways (Figure 73A). To promote our findings in cell lines, we tested the same set of RADARS constructs in the HepG2 human hepatocellular carcinoma cell line and observed that the RADARS sensor did not trigger any dsRNA response (Figure 73C).

Finally, we explored whether ADAR1p150 overexpression produces off-target editing in the transcriptome, which has been observed in ADAR-based therapeutic RNA editing approaches (Cox et al., 2017; Qu et al., 2019; Reautschnig et al., 2022). We first profiled the region surrounding the hybrid duplex of PPIB and ACTB transcripts with RADARS-modified guide RNAs and found no significant off-target editing due to sensor hybridization or ADAR1p150 overexpression (Figure 74A). To measure potential off-targets in an unbiased manner, we next performed polyAmRNA sequencing on cells expressing the PPIB sensor and ADAR1p150. We found that the combination of ADAR1p150 overexpression with RADARS targeting PPIB produced only 23 detectable sites in the transcriptome, all of which were less than 10% edited (Figure 74B). Furthermore, we found no significant editing of sites in the absence of ADARp150 overexpression with a non-targeting engineered guide RNA. In contrast, when we analyzed published RNA-seq data from overexpression of the MCP-ADAR2(E488Q) deaminase domain, we detected >10,000 sites with significant A->I RNA editing, highlighting the impact of deaminase construct choice on off-target properties (Figure 74C).

We did not find significant homology in the sequences surrounding the modified guide RNAs and off-target editing sites (Figure 75A). The same sequencing and analysis of different modified guide RNAs targeting exogenous IL6 targets in the presence of ADARp150 overexpression showed a low and significant off-target editing rate (42 sites were less than 10% edited, Figure 74B). Importantly, there were 22/23 PPIB off-target sites in two different modified guide RNA samples. Finally, we found that the sequence motifs of the bases around the editing sites were very similar to the preferred substrates of ADAR1 (Eggington et al., 2011) (Figure 75B). Overall, these observations indicate that RADARS produces relatively little nonspecific RNA editing in the transcriptome when used with ADARp150 overexpression.

Claims (94)

1. An rna sensor system comprising:

   a) A single stranded RNA (ssRNA) sensor containing a stop codon and a payload; optionally wherein the ssRNA sensor further comprises a normalization gene; and

   B) Adenosine Deaminase (ADAR) deaminase acting on RNA;

   Wherein the sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for the ADAR deaminase;

   Wherein the substrate comprises a mismatch within the stop codon;

   Wherein said mismatch is editable by said ADAR deaminase, which editing is effective to remove said stop codon, so as to enable translation and expression of said payload.
2. 2. The RNA sensor system of claim 1, wherein the mismatch comprises an adenine-to-cytidine mismatch in the dsRNA duplex.
3. 3. The RNA sensor system of claim 1, wherein the mismatch comprises an adenine-to-cytidine mismatch, and wherein the ADAR deaminase edits adenine to inosine in the mismatch in the dsRNA duplex.
4. 4. The RNA sensor system of claim 1, comprising more than one mismatch.
5. 5. The RNA sensor system of claim 1, wherein the payload comprises a reporter protein, a transcription factor, an enzyme, a transgenic protein, or a therapeutic protein.
6. 6. The RNA sensor system of claim 1, wherein the payload comprises a fluorescent reporter.
7. 7. The RNA sensor system of claim 1, wherein the payload comprises an EGFP reporter or a luciferase reporter.
8. 8. The RNA sensor system of claim 1, wherein the payload comprises caspase.
9. 9. The RNA sensor system of claim 1, wherein the ADAR is endogenous or exogenous.
10. 10. The RNA sensor system of claim 1, wherein the ADAR is a modified ADAR.
11. 11. The RNA sensor system of claim 1, wherein the ADAR comprises a programmable a-to-I (G) substituted RNA editing (repir) molecule, a Cas13b-ADAR fusion molecule, a Cas13d-ADAR fusion molecule, a Cas7-11-ADAR fusion molecule and MS2-ADAR fusion molecule, a deaminase domain of ADAR2, a full-length ADAR2, or a truncated ADAR2.
12. 12. The RNA sensor system of claim 1, wherein the RNA sensor system comprises a plurality of RNA sensors.
13. 13. A cellular logic system comprising:

    a) And a gate containing a ssRNA sensor containing one or more payloads and a plurality of stop codons complementary to different ssRNA targets;

    wherein the ssRNA sensor is capable of binding to the ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase;

    wherein the substrate comprises a mismatch within each stop codon;

    Wherein a mismatch within said each stop codon is editable by said ADAR deaminase, which editing is effective to remove said stop codon, so as to enable translation and expression of said one or more payloads; and

    B) Or gate comprising a plurality of independent ssRNA sensors containing a payload and a stop codon complementary to one or more different RNA targets;

    wherein each ssRNA sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase;

    wherein the substrate comprises a mismatch within each stop codon; and

    Wherein a mismatch within each of the stop codons is editable by the ADAR deaminase, which editing is effective to remove the stop codons to enable translation and expression of the one or more payloads.
14. 14. A cellular logic system comprising:

    a) And a gate comprising a ssRNA sensor containing one or more payloads and a plurality of stop codons complementary to different ssRNA targets;

    wherein the ssRNA sensor is capable of binding to the ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase;

    wherein the substrate comprises a mismatch within each stop codon;

    Wherein a mismatch within said each stop codon is editable by said ADAR deaminase, which editing is effective to remove said stop codon, so as to enable translation and expression of said one or more payloads; or alternatively

    B) Or gate comprising a plurality of independent ssRNA sensors containing a payload and a stop codon complementary to one or more different RNA targets;

    wherein each ssRNA sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for an ADAR deaminase;

    wherein the substrate comprises a mismatch within each stop codon; and

    Wherein a mismatch within each of the stop codons is editable by the ADAR deaminase, which editing is effective to remove the stop codons to enable translation and expression of the one or more payloads.
15. 15. A method of quantifying ribonucleic acid (RNA) levels, comprising:

    a) Providing a single stranded RNA (ssRNA) sensor comprising a stop codon and a payload; optionally wherein the ssRNA sensor further comprises a normalization gene; and

    B) Adding an Adenosine Deaminase (ADAR) deaminase acting on the RNA;

    Wherein the sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for the ADAR deaminase;

    Wherein the substrate comprises a mismatch within the stop codon;

    Wherein said mismatch is editable by said ADAR deaminase, which editing is effective to remove said stop codon, so as to enable translation and expression of said payload.
16. 16. The method of claim 15, wherein the mismatch comprises an adenine-to-cytidine mismatch in the dsRNA duplex.
17. 17. The method of claim 15, wherein the mismatch comprises adenine to cytidine, and wherein the ADAR deaminase edits adenine to inosine in the dsRNA duplex.
18. 18. The method of claim 15, comprising more than one mismatch.
19. 19. The method of claim 15, wherein the payload comprises a reporter protein, a transcription factor, an enzyme, a transgenic protein, a therapeutic protein, or an antigen or epitope for diagnostic assays.
20. 20. The method of claim 15, wherein the ADAR is endogenous or exogenous.
21. 21. The method of claim 15, wherein the ADAR is a modified ADAR.
22. An rna sensor system comprising:

    a) A single stranded RNA (ssRNA) sensor comprising at least a first stop codon and a payload; optionally wherein the ssRNA sensor further comprises a normalization gene; and

    B) Adenosine Deaminase (ADAR) deaminase acting on RNA;

    Wherein the sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for the ADAR deaminase;

    wherein the substrate comprises a mismatch within the first stop codon;

    Wherein said mismatch is editable by said ADAR deaminase, which editing is effective to remove said stop codon, so as to enable translation and expression of said payload.
23. 23. The RNA sensor system of claim 22, wherein the single stranded RNA sensor comprises more than one stop codon.
24. 24. The RNA sensor system of claim 22, wherein the single stranded RNA sensor further comprises a second stop codon.
25. 25. The RNA sensor system of claim 23, wherein the single stranded RNA sensor further comprises a third stop codon.
26. 26. The RNA sensor system of claim 1 or claim 22, wherein the mismatch comprises CCA on a target strand and TAG/UAG on a sensor strand.
27. 27. The RNA sensor system of claim 1 or claim 22, wherein the sensor strand comprises a TAG/UAG stop codon, but not a mismatch to a CCA codon on the target strand.
28. 28. The RNA sensor system of claim 1 or claim 22, wherein the sensor strand comprises a stop codon that is compatible with a target strand selected from the group consisting of

    ACA, ACT, ACC, ACG, TCA, TCT, TCC, TCG, GCA, GCT, GCC, GCG, CCA, CCT, CCC, and CCG

    Is matched or mismatched.
29. An rna sensor system comprising:

    a) A single stranded RNA (ssRNA) sensor comprising a stop codon and a payload; optionally wherein the ssRNA sensor further comprises a normalization gene; and

    B) Adenosine Deaminase (ADAR) deaminase acting on RNA;

    wherein the sensor is capable of binding to a ssRNA target to form a double-stranded RNA (d _s RNA) duplex that becomes a substrate for the ADAR deaminase;

    Wherein the substrate comprises a stop codon that is editable by the ADAR deaminase, the editing being effective to remove the stop codon to enable translation and expression of the payload.
30. 30. The RNA sensor system of claim 29, wherein the ssRNA sensor comprises a TAG/UAG stop codon.
31. 31. The RNA sensor system of claim 30, wherein a TAG/UAG stop codon forms a dsRNA duplex with the ssRNA target at a codon having formula nCn, wherein n is any nucleotide and C is cytidine.
32. 32. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is 50nt or greater, 100nt or greater, 150nt or greater, 200nt or greater, 250nt or greater, 300nt or greater, or 500nt or greater.
33. 33. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is 51nt.
34. 34. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is 81nt.
35. 35. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is 171nt.
36. 36. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is 225nt.
37. 37. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is 279nt.
38. 38. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is longer than 279nt.
39. 39. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor is a circular sensor.
40. 40. The RNA sensor system of claim 39, wherein said loop sensor is a rolling circle translation sensor.
41. 41. The RNA sensor system of claim 39, wherein said circular sensor is a conventional circular sensor.
42. 42. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises two stop codons.
43. 43. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises three stop codons.
44. 44. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises two stop codons, wherein only one stop codon is targeted by ADAR editing.
45. 45. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises three stop codons, wherein only one stop codon is targeted by ADAR editing.
46. 46. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises at least one affinity binding region.
47. 47. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises at least three affinity binding regions.
48. 48. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises at least five affinity binding regions.
49. 49. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises at least seven affinity binding regions.
50. 50. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the ssRNA sensor comprises more than seven affinity binding regions.
51. 51. The RNA sensor system of any one of claims 47-50, wherein the affinity binding regions are separated by MS2 hairpin regions.
52. 52. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the payload comprises Cre recombinase.
53. 53. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the payload comprises a Cas protein.
54. 54. The RNA sensor system of claim 53, wherein the payload comprises Cas9.
55. 55. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the payload comprises a transcription factor.
56. 56. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the payload comprises a payload ADAR.
57. 57. The RNA sensor system of claim 1, claim 22, or claim 29, wherein the payload is a reporter of a cellular stress response.
58. 58. A composition comprising the RNA sensor system of claim 1, claim 22 or claim 29, and a delivery vehicle.
59. 59. A composition comprising an RNA sensor system and a lipid nanoparticle, wherein the RNA sensor system comprises:

    a) A single stranded RNA (ssRNA) sensor comprising a stop codon and a payload; optionally wherein the ssRNA sensor further comprises a normalization gene; and

    B) Adenosine Deaminase (ADAR) deaminase acting on RNA;

    Wherein the sensor is capable of binding to a ssRNA target to form a double-stranded RNA (dsRNA) duplex that becomes a substrate for the ADAR deaminase;

    Wherein said substrate comprises a stop codon which is editable by said ADAR deaminase, said editing being effective to remove said stop codon to enable translation and expression of said payload, and

    Wherein the RNA sensor system is encapsulated in the lipid nanoparticle.
60. 60. A method of killing a particular cell or cell type, wherein the method comprises providing a single stranded RNA (ssRNA) sensor or guide comprising a stop codon and a payload; optionally wherein the ssRNA sensor further comprises a normalization gene; wherein the payload is a self-dimerizing caspase, and wherein the ssRNA sensor or guide is capable of binding to a ssRNA target to form a double stranded RNA duplex that becomes a substrate for an Adenosine Deaminase (ADAR) deaminase acting on RNA, and wherein the ssRNA target is enriched for expression in the particular cell or cell type.
61. An rna sensor system comprising:

    a) An RNA sensor comprising a stop codon and a payload; optionally wherein the RNA sensor further comprises a normalization gene; and

    B) Adenosine Deaminase (ADAR) deaminase acting on RNA;

    wherein the sensor is capable of binding to an RNA target to form a double-stranded RNA (dsRNA) duplex region that becomes a substrate for the ADAR deaminase;

    Wherein the substrate comprises a mismatch within the stop codon;

    Wherein said mismatch is editable by said ADAR deaminase, which editing is effective to remove said stop codon, so as to enable translation and expression of said payload.
62. 62. The RNA sensor system of claim 61, wherein said RNA sensor is a single-stranded RNA.
63. 63. The RNA sensor system of claim 61, wherein the RNA sensor comprises one or more double-stranded RNA (dsRNA) domains.
64. 64. The RNA sensor system of any one of claims 61-63, wherein the RNA target is single stranded RNA.
65. 65. The RNA sensor system of any one of claims 61-63, wherein the RNA comprises one or more double-stranded RNA (dsRNA) domains.
66. 66. The RNA sensor system of any one of claims 61-65, wherein the RNA sensor is 50nt or greater, 100nt or greater, 150nt or greater, 200nt or greater, 250nt or greater, 300nt or greater, or 500nt or greater.
67. 67. The RNA sensor system of any one of claims 61-66, wherein the RNA sensor is 51nt.
68. 68. The RNA sensor system of any one of claims 61-66, wherein the ssRNA sensor is 81nt.
69. 69. The RNA sensor system of any one of claims 61-66, wherein the ssRNA sensor is 171nt.
70. 70. The RNA sensor system of any one of claims 61-66, wherein the ssRNA sensor is 225nt.
71. 71. The RNA sensor system of any one of claims 61-66, wherein the ssRNA sensor is 279nt.
72. 72. The RNA sensor system of any one of claims 61-66, wherein the ssRNA sensor is longer than 279nt.
73. 73. The RNA sensor system of any one of claims 61-66, wherein the ssRNA sensor is a circular sensor.
74. 74. The RNA sensor system of claim 73, wherein the loop sensor is a rolling circle translation sensor.
75. 75. The RNA sensor system of claim 73, wherein the circular sensor is a conventional circular sensor.
76. 76. The RNA sensor system of any one of claims 61-75, wherein the RNA sensor comprises two stop codons.
77. 77. The RNA sensor system of any one of claims 61-75, wherein the RNA sensor comprises three stop codons.
78. 78. The RNA sensor system of any one of claims 61-75, wherein the RNA sensor comprises two stop codons, wherein only one stop codon is targeted by ADAR editing.
79. 79. The RNA sensor system of any one of claims 61-75, wherein the RNA sensor comprises three stop codons, wherein only one stop codon is targeted by ADAR editing.
80. 80. The RNA sensor system of any one of claims 61-79, wherein the RNA sensor comprises at least one affinity binding region.
81. 81. The RNA sensor system of any one of claims 61-80, wherein the RNA sensor comprises at least three affinity binding regions.
82. 82. The RNA sensor system of any one of claims 61-81, wherein the RNA sensor comprises at least five affinity binding regions.
83. 83. The RNA sensor system of any one of claims 61-82, wherein the RNA sensor comprises at least seven affinity binding regions.
84. 84. The RNA sensor system of any one of claims 61-83, wherein the RNA sensor comprises more than seven affinity binding regions.
85. 85. The RNA sensor system of any one of claims 80 to 84, wherein the affinity binding regions are separated by MS2 hairpin regions.
86. 86. The RNA sensor system of any one of claims 61-85, wherein the payload comprises Cre recombinase.
87. 87. The RNA sensor system of any one of claims 61-85, wherein the payload comprises a Cas protein.
88. 88. The RNA sensor system of claim 87, wherein the payload comprises Cas9.
89. 89. The RNA sensor system of any one of claims 61-85, wherein the payload comprises a transcription factor.
90. 90. The RNA sensor system of any one of claims 61-85, wherein the payload comprises a payload ADAR.
91. 91. The RNA sensor system of any one of claims 61-85, wherein the payload is a reporter of a cellular stress response.
92. 92. The RNA sensor system of any one of claims 61-85, and a delivery vehicle.
93. 93. The RNA sensor system of claim 1, claim 22, claim 29, or claim 61, wherein the ADAR is selected from the group consisting of

    ADAR2,ADAR1,ADAR1 p150,ADAR1 p110,ADAR2 R455G,ADAR2 R455G,ADAR2 S486T,ADAR2 T375G E488Q T490A,ADAR2 T375G,ADAR2 T375S,ADAR2 N473D,ADAR2 Deaminase domain, ADAR 2T 490S, ADAR 2T 490A, MCP-ADAR2 deazaase domain, ADAR 2R 455E, ADAR 2T 375G T A, ADAR 2E 488Q, MCP-ADAR2 deaminase domain E488Q T490A, ADAR 2R 510E, ADAR 2R 455S, ADAR 2V 351L,

    And derivatives or modified variants thereof.
94. 94. The RNA sensor system of claim 1, claim 22, claim 29, or claim 61, wherein the ADAR is endogenously expressed in a target cell in which the RNA sensor system is available.