This application claims the benefit of U.S. Provisional Patent Application Ser. No. 62/511,296, filed on May 25, 2017; Ser. No. 62/541,544, filed on Aug. 4, 2017; and Ser. No. 62/622,676, filed on Jan. 26, 2018. The entire contents of the foregoing are hereby incorporated by reference.
This invention was made with Government support under Grant No. GM118158 awarded by the National Institutes of Health. The Government has certain rights in the invention.
Described herein are methods and compositions for improving the genome-wide specificities of targeted base editing technologies.
Base editing (BE) technologies use an engineered DNA binding domain (such as RNA-guided, catalytically inactive Cas9 (dead Cas9 or dCas9), a nickase version of Cas9 (nCas9), or zinc finger (ZF) arrays) to recruit a cytosine deaminase domain to a specific genomic location to effect site-specific cytosine→thymine transition substitutions1,2. BEs are a particularly attractive tool for treating genetic diseases that manifest in cellular contexts where making precise mutations by homology directed repair (HDR) would be therapeutically beneficial but are difficult to create with traditional nuclease-based genome editing technology. For example, it is challenging or impossible to achieve HDR outcomes in tissues composed primarily of slowly dividing or post-mitotic cell populations, since HDR pathways are restricted to the G2 and S phases of the cell cycle3. In addition, the efficiency of HDR can be substantially limited by the competing and more efficient induction of variable-length indel mutations caused by non-homologous end-joining-mediated repair of nuclease-induced breaks. By contrast, BE technology has the potential to allow practitioners to make highly controllable, highly precise mutations without the need for cell-type-variable DNA repair mechanisms.
Base editor platforms (BE) possess the unique capability to generate precise, user-defined genome-editing events without the need for a donor DNA molecule. Base Editors (BEs) that include a single strand nicking CRISPR-Cas9 (nCas9) protein fused to cytosine deaminase domain and uracil glycosylase inhibitor (UGI) domains (e.g., BE3) efficiently induce cytosine-to-thymine (C-to-T) base transitions in a site-specific manner as determined by the CRISPR guide RNA (gRNA) spacer sequence1. As with all genome editing reagents, it is critical to first determine and then mitigate BE’s capacity for generating off-target mutations before it is used for therapeutics so as to limit its potential for creating deleterious and irreversible genetically-encoded side-effects. Herein, we describe dimeric BEs that use split deaminases (sDA) that are functional when brought into close proximity to each other, one fused to a ZF and one to an nCas9-UGI protein comprising one or more UGIs, so as to limit the ability of the deaminase domain from deaminating at off-target ssDNA target sites independent of nCas9 R-loop formation.
Thus, provided herein are fusion proteins comprising: (i) a first portion of a split deaminase (“sDA1”) enzyme fused to a programmable DNA-binding domain, preferably selected from the group consisting of such as a ZF, TALE, Cas9, catalytically inactive Cas9 (dCas9) or Cas9 ortholog (i.e., a homologous protein from another species such as dCpf1), nicking Cas9 (nCas9) or nicking Cas9 ortholog, wherein the sDA1 is an N-terminal truncated, catalytically inactive or deficient derivative of a parental deaminase selected from the group consisting of hAID, rAPOBEC1, mAPOBEC3, hAPOBEC3A, hAPOBEC3B, hAPOBEC3C, hAPOBEC3F, hAPOBEC3G, or hAPOBEC3H, and variants thereof, e.g., variants that have altered substrate specificities or activities such as eA3A; or (ii) a second portion of a split deaminase (“sDA2”) fused to an nCas9 protein, preferably an nCas9-UGI protein, e.g., in a manner similar to previously described base editor architectures, or any orthogonal DNA targeting domain as the one used for its complementary sDA1 portion (e.g., dCpf1, TALE, ZF), wherein the sDA2 is a C-terminal truncated, catalytically inactive or deficient derivative of a parental deaminase selected from the group consisting of hAID, rAPOBEC1*, mAPOBEC3, hAPOBEC3A, hAPOBEC3B, hAPOBEC3C, hAPOBEC3F, hAPOBEC3G, or hAPOBEC3H, and/or variants thereof, e.g., with altered substrate specificities or activities such as eA3A. In the present methods, the split deaminases are not full length proteins, but are fragments thereof, wherein the co-expression of a fusion protein of (i) with a fusion protein of (ii) comprising a sDA1 and sDA2 portion from the same parental deaminase in eukaryotic cells, and their subsequent co-localization at adjacent genomic target sites, provides a catalytically active base-editor. The terms “sDA1” and “sDA2” are used herein to refer to the first and second split deaminases generally, and do not refer specifically to the exemplary split deaminases described herein.
Also provided herein are nucleic acids encoding the fusion proteins described herein, and compositions comprising one or more of those nucleic acids, e.g., wherein the nucleic acids encode a pair of the fusion proteins, e.g., comprising a SDA1 and SDA2 portion from the same parental deaminase. Further, provided herein are vectors comprising the nucleic acids, and isolated host cells comprising and optionally expressing the nucleic acids. In some embodiments, the host cell is a stem cell, e.g., a hematopoietic stem cell.
In addition, provided herein are methods for targeted deamination of one or more selected cytosines in a nucleic acid. The methods include contacting the nucleic acid with a pair of fusion proteins described herein comprising a SDA1 and SDA2 portion from the same parental deaminase, as well as one or more gRNAs that interact with Cas9 domains in the fusion proteins. In some embodiments, one of the fusion proteins comprises nCas9, the other fusion protein comprises ZF or TALE, and the ZF or TALE is targeted to a sequence of 9-24 bp adjacent to the target site of the gRNA for the nCas9, wherein the gRNA binds to the nucleic acid comprising the selected cytosine.
In some embodiments, the nucleic acid is in a cell, e.g., a eukaryotic cell, and the method comprises contact the cell with the fusion proteins or expressing the fusion proteins in the cell.
Also provided are methods for improving specificity of targeted deamination in a cell, e.g., a eukaryotic cell, by expressing in the cell, or contacting the cell with, a pair of fusion proteins described herein comprising a sDA1 and sDA2 portion from the same parental deaminase, as well as one or more gRNAs that interact with Cas9 domains in the fusion proteins. In some embodiments, one of the fusion proteins comprises nCas9, the other fusion protein comprises ZF or TALE, the ZF or TALE is targeted to a sequence of 9-24 bp adjacent to the target site of the gRNA for the nCas9, wherein the gRNA binds to the nucleic acid comprising the selected cytosine.
In some embodiments, the fusion protein is delivered as an RNP, mRNA, or plasmid.
Also provided herein are methods for deaminating one or more selected cytosines in a nucleic acid, by contacting the nucleic acid with a pair of fusion proteins described herein comprising a sDA1 and sDA2 portion from the same parental deaminase, as well as one or more gRNAs that interact with Cas9 domains in the fusion proteins.
In addition, provided herein are compositions comprising a purified fusion protein or pair of fusion proteins described herein, preferably a pair of fusion proteins described herein comprising a sDA1 and sDA2 portion from the same parental deaminase, an optionally one or more gRNAs that interact with Cas9 domains in the fusion proteins. In some embodiments, the composition comprise one or more ribonucleoprotein (RNP) complexes.
Also provided herein are ribonucleoprotein (RNP) complexes that include a variant spCas9 protein as described herein and a guide RNA that targets a sequence having a PAM sequence targeted by the split deaminase fusion protein comprising Cas9 or Cas9 derivative.
Also provided herein are methods for targeted deamination, or improving specificity of targeted deamination, of a selected cytosine in a nucleic acid, comprising contacting the nucleic acid with one or more of the fusion proteins or base editing systems described herein.
In some embodiments, the fusion protein is delivered as an RNP, mRNA, or plasmid DNA.
Also provided herein are methods for deaminating a selected cytosine in a nucleic acid, the method comprising contacting the nucleic acid with a fusion protein or base editing system described herein.
Additionally, provided herein are compositions comprising a purified a fusion protein or base editing system as described herein.
Further, provided herein are nucleic acids encoding a fusion protein or base editing system described herein, as well as vectors comprising the nucleic acids, and host cells comprising the nucleic acids, e.g., stem cells, e.g., hematopoietic stem cells.
Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Methods and materials are described herein for use in the present invention; other, suitable methods and materials known in the art can also be used. The materials, methods, and examples are illustrative only and not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control.
Other features and advantages of the invention will be apparent from the following detailed description and figures, and from the claims.
FIG. 1. Diagram of an exemplary typical high efficiency base editing setup. A nicking Cas9 bearing a catalytically inactivating mutation at one of its two nuclease domains binds to the target site dictated by the variable spacer sequence of the gRNA. The formation of a stable R-loop creates a ssDNA editing window on the non-deaminated strand. The Cas9 creates a single strand break in the genomic DNA, prompting the host cell to repair the lesion using the deaminated strand as a template, thus biasing repair towards the cytosine→thymine transition substitution. See Komor et al., 2016.
FIGS. 2A-2G. Schematic representation of: 2A.) First-generation base editor targeting and deaminating at an on-target site, with a deaminase targeting an R-loop generated by an on-target nCas9. 2B.) First-generation base-editor binding to and deaminating an off-target genomic R-loop independent of its nCas9 targeting capabilities. 2C.) First-generation base-editor binding to and deaminating an off-target genomic transcription bubble independent of its nCas9 targeting capabilities. (Note: 2B and 2C are exemplary cases of genomic ssDNA targets potentially available to BE deamination, but do not constitute an exhaustive list.) 2D.) A split-deaminase (sDA) BE targeting a genomic site with an nCas9-mediated R-loop and adjacent TALE or ZF binding. Note that the two split deaminase portions (sDA1 and sDA2) are brought into close proximity by the adjacent binding, reconstituting their catalytic activity and allowing on-target deamination. 2E and 2F.) Even if one half of a split BE could bind a non-target piece of ssDNA such as a genomic R-loop or transcription bubble through its sDA domain, it would not have enough machinery to reconstitute deaminase enzymatic activity (sDA2-nCas9-UGI half shown). 2G. Off-target binding of the ZF or nCas9 components of a sDA system (ZF-sDA1 shown) would not result in co-localization of enough machinery to reconstitute deaminase enzymatic activity.
FIGS. 3A-3B. hAPOBEC3G with representative candidate split sites. Multiple rotational views of the hAPOBEC3G structure are shown. Magenta colored loop regions are candidate split sites selected on the bases of their lack of secondary structures and their distance from the catalytic center. PDB: 3E1U.
FIG. 4. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.1-3AC3L-ZF-C and N-sDA2.1-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. The sDA1.1 and sDA2.1 pair resulted in significant C-to-T conversion when the ZF binding site was upstream of gRNA binding site with an in-series orientation with 31 bps in-between. EGFP target sequence,
FIG. 5. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.1-3AC3L-ZF-C and N-sDA2.2-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. The sDA1.1 and sDA2.2 pair did not stimulate discernable C-to-T conversion in any orientation attempted. EGFP target sequence,
FIG. 6. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.2-3AC3L-ZF-C and N-sDA2.1-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. Low-level C-to-T mutations are observed primarily when using gRNA2 with either ZF, with gRNA1 experiments yielding detectable but diminished levels of activity. EGFP target sequence,
FIG. 7. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.2-3AC3L-ZF-C and N-sDA2.2-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. Low-level C-to-T mutations are observed primarily when using gRNA2 with either ZF, with gRNA1 experiments yielding detectable but diminished levels of activity. EGFP target sequence,
FIG. 8. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.2-3AC3L-ZF-C and N-sDA2.3-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 9. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.3-3AC3L-ZF-C and N-sDA2.2-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 10. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.2-3AC3L-ZF-C and N-sDA2.3-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 11. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.3-3AC3L-ZF-C and N-sDA2.4-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 12. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.4-3AC3L-ZF-C and N-sDA2.3-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 13. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.4-3AC3L-ZF-C and N-sDA2.4-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 14. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.5-3AC3L-ZF-C and N-sDA2.4-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 15. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.5-3AC3L-ZF-C and N-sDA2.6-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 16. C-to-T transition mutations in the integrated EGFP gene from a split rAPO1 base editor architecture consisting of adjacently-targeting N-sDA1.6-3AC3L-ZF-C and N-sDA2.6-nCas9-UGI-C proteins in several indicated orientations. Conversion rates at each position are indicated by shaded boxes with overlaid percentage numbers for residues in which significant mutation was observed. Orientation information is depicted, with arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM) and ZF binding sites (with the arrow indicating the direction of ZF binding in reference to N→C orientation). Approximate editing windows (residues 4-8 in the gRNA target site) are indicated. Experiments were performed in duplicate and sequencing from each sample is shown independently. No significant mutations detected. EGFP target sequence,
FIG. 17. C-to-T conversion data with first-generation BE3 (described in reference 1) with both gRNAs used in this study. (Note that the coloration gradient of these samples is shaded lighter than graphs above and that direct comparison requires evaluation of relative numerical rates). Orientation information is depicted, with an arrows representing gRNA binding sites (with the arrow pointing in the direction of the PAM). EGFP target sequence,
FIG. 18. C-to-T conversion rates of individual N-sDA1-ZF-C proteins without an adjacent sDA2-nCas9-UGI. No discernable editing observed. EGFP target sequence,
FIG. 19. C-to-T conversion rates of individual N-sDA2-nCas9-UGI-C proteins without an adjacent N-sDA1-ZF-C. No discernable editing was observed. EGFP target sequence,
FIG. 20. Evidence of C-to-T conversion when using adjacently-targeting N-sDA1.X-NLS-ZF-C and N-sDA2.X-nCas9-UGI-C human APOBEC3a (hA3A) split Base Editors in the indicated orientation. Pointed boxes representing the nCas9 gRNA binding site (gRNA2) and ZF binding site (ZF1) are shown, with the pointed ends indicating the PAM-proximal end of the gRNA and indicating the N→C orientation of the ZF, respectively. Conversion rates at each position are indicated by shaded boxes. Rates of deamination by split BE pairs are around 2.5% per cytosine using the sDA1.6+sDA2.6 configuration and around 1.7% per cytosine for the sDA1.1+sDA2.1 configuration, while a hAPOBEC3A-nCas9-UGI positive control possessed 3-4× the amount of on-target activity as active hA3A halfase pairs. gRNA target region:
FIGS. 21A-21D. Summary of C-to-T conversion rate of all rAPO1 halfase combination base editors as compared to a benchmark BE3 base editor at an integrated EGFP locus. The sum of total C-to-T editing percentages among three cytosines within or near the target gRNA’s approximate editing window is shown, as averaged between two replicates. 21A shows the ZF1+gRNA1 data, 21B shows the ZF1+gRNA2 data, 21C shows the ZF2+gRNA1 data, 21D shows the ZF2+gRNA2 data.
FIG. 22. Representation of a portion of the EGFP reporter gene and the target sites used for the rAPO1 halfase combination experiments. EGFP target region:
In the most efficient BE configuration described to date, a cytosine deaminase (DA) domain and uracil glycosylase inhibitor (UGI; a small bacteriophage protein that inhibits host cell uracil DNA glycosylase (UDG), the enzyme responsible for excising uracil from the genome1, 4) are both fused to nCas9 (derived from either Streptococcus pyogenes Cas9 (SpCas9) or Staphylococcus aureus Cas9 (SaCas9). The nCas9 forms an R-loop at a target site specified by its single guide RNA (gRNA) and recognition of an adjacent protospacer adjacent motif (PAM), leaving approximately 4-8 nucleotides of the non-target strand exposed as single stranded DNA (ssDNA) near the PAM-distal end of the R-loop (FIG. 1). This region of the ssDNA is the template that is able to be deaminated by the ssDNA-specific DA domain to produce a guanosine:uracil (G:U) mismatch and defines the editing window. The nCas9 nicks the non-deaminated strand of DNA, biasing conversion of the G:U mismatch to an adenine:thymine (A:T) base pair by directing the cell to repair the nick lesion using the deaminated strand as a template. To date, the deaminase domains described in these fusion proteins have been rat APOBEC1 (rAPO1), an activation-induced cytosine deaminase (AID) derived from lamprey termed CDA (PmCDA), human AID (hAID), or a hyperactive form of hAID lacking a nuclear export signal, or an engineered variant of human APOBEC3A (hA3A) termed eA3A1-2, 5-7, 16. Any of these deaminase domains from these BEs can be used as parental deaminases in the present fusion proteins. BE technology was primarily established using the SpCas9 protein for its nCas9 domain (nSpCas9), but although herein we refer to nCas9, in general any Cas9-like nickase could be used based on any ortholog of the Cpf1 protein (including the related Cpf1 enzyme class) to perform this function, unless specifically indicated. In addition, a completely enzymatically dead dCas9 (or Cas9-like enzyme) can also be used as the targeting mechanism of a functional BE enzyme.
An important consideration for the use of BE in therapeutic settings will be to assess its genome-wide capacity for off-target mutagenesis and to modify the technology to minimize or, ideally, to eliminate the risks of stimulating deleterious off-target mutations. Herein, we described technological improvements to BEs that can be used to reduce or eliminate potential unwanted BE mutagenesis.
Using Split Deaminases to Limit Unwanted Off-Target Base Editor Deamination
Because of AID/APOBEC enzymes’ natural ability to bind and deaminate cytosines in genomic DNA and cytosines in RNA, non-specific spurious deamination events are a possibly important source of off-target mutagenesis in the genome and transcriptome from CRISPR Base Editor technology. In theory, even if the BE’s nCas9 domain (and any potential dCas9, TALE, and/or ZF domains) are eminently specific, this might do nothing to prevent the natural RNA- and ssDNA-targeting ability of the APOBEC enzyme from non-specifically deaminating globally across the transcriptome or the whichever regions of the genome are exposed as ssDNA, such as actively transcribed regions or DNA undergoing replication. In fact, an E. coli-based assay examining deaminases showed that an actively transcribed region could be highly enriched (˜7-530 fold) for C→T transition mutations when exposed to various overexpressed mammalian deaminases4. Further, one group has found that co-expression of PmCda1 and nCas9 as two separate, untethered proteins in yeast cells results in similar levels of deamination at the gRNA-specified target site as when the two components are expressed as direct fusion partners, demonstrating that these proteins are capable of deaminating ssDNA from solution without an affinity tether to the genomic location5. This concern is especially relevant now that scientists are becoming increasingly aware that R-loops are a more common occurrence in the genomes of eukaryotic cells than previously thought, thus creating many potential steady-state off-target ssDNA substrates where an APOBEC could bind and deaminate6. While it is as yet unproven whether BE overexpression itself can sufficiently stimulate spurious deamination and mutagenesis on a global genomic scale, aberrant and over-active APOBEC deaminase activity is a known driver of tumorigenic mutagenesis7 and overexpression of at least hAPO38-11 has been shown to stimulate genomic cytosine hypermutation. Thus, it stands to reason that limiting the naturally global deaminating activity of over-expressed deaminases like BE will be important for translating BE technologies into therapeutic applications. Of note, since most BEs include at least one UGI inhibitor to bias deamination events toward productive C→T mutations, it is possible that global off-target BE activity is even more mutagenic than the effects of aberrant deaminase activity alone during tumorigenesis.
To impose a stricter requirement for BEs to act on their intended target sequences rather than globally, we created a split BE architecture comprised of two separate proteins consisting of reciprocal deaminase truncation variants fused to adjacently-targeted DNA binding domains. These dimeric BE technologies make use of “split deaminases” (sDAs) that require co-localization (an “AND Gate”) of both sDA domains at adjacent DNA sites to function properly. In this scenario, spurious binding events of either “halfase” of the dimeric base-editor will be unlikely to result in productive deamination events, since each component on its own will not contain the full complement of enzymatic machinery necessary to catalyze cytosine deamination (FIGS. 2A-2G).
To create dimeric BEs, we fused an N-terminal truncation of a split deaminase (sDA1) enzyme to a ZF (though any DNA targeting domain orthogonal to Cas9, such as Cpf1, TALE, ZF, or a dCas9 orthogonal to the nCas9 used to target sDA2, may be suitable) targeted to a ˜9-24 bp sequence, and a reciprocal or somewhat overlapping C-terminal truncation of a deaminase fused to an nCas9-UGI fusion protein, such that the N-terminal truncation and the C-terminal truncation together form a functional enzyme. The exemplary BEs were made in a similar orientation to the first-generation BE3 enzyme (sDA2-nCas9-UGI) targeting an adjacent sequence with a ˜17-24 bp target site1. To the best of the inventors’ knowledge, though there is no record of functional split APOBEC enzymes (or other mammalian deaminases), a yeast cytosine deaminase (yCD) has been shown to constitute at least a partially functional enzyme (on cytosine as a metabolite and the pro-drug 5-fluorocytosine, though it was not shown to be able to deaminate DNA) when split and reconstituted by protein dimerization12 and serves as a useful template to inform how various APOBEC proteins may be effectively bifurcated; however, since the yCD shares little primary sequence homology to mammalian deaminases, and the split yCD was not reported to function on DNA and used protein scaffolds to bring its constituent pieces together, it is not obvious that yCD split deaminases will be directly comparable to those so far described for use in for BEs. Therefore, we used APOBEC structural information to determine the unstructured linker regions as potential sites at which to split APOBEC enzymes (FIGS. 3A-3B), since those sites may be less likely to affect overall functionality or folding of the constituent subdomains. This split deaminase strategy can be used with wild-type versions of deaminase enyzmes, and also any engineered variants that may be described, with the split BE potentially retaining any special features of the engineered deaminases16.
This architecture should virtually eliminate the capacity for spurious deamination, since any other DNA binding event by either of the two constituent halfases will lack any enzymatic deaminase activity and will be therefore unable to perturb genomic DNA. In addition, a split BE should generally increase the specificity of editing compared to typical BEs by virtue of the fact that the split BE system requires the binding of a higher number of sequential/adjacent DNA bases, thereby decreasing the off-target effects conferred by off-target binding of either halfase on its own. CRISPR BE architectures are known to induce C-to-T mutations in human cells at some genomic sites that are imperfect matches to their gRNAs13, and since ZFs are known to bind with some capacity to off-target sites it stands to reason that a ZF-BE architecture would also induce off-target mutagenesis to some capacityl14.
It is conceivable and likely that any CRISPR/Cas-based targeting system, including Cas9s from Streptococcus pyogenes or Stapholococcus aureus or Cpf1 proteins from various organisms could be used in place of the nCas9 portion of the sDA2-nCas9-UGI fusion protein, so long as the targeting mechanism results in specific DNA binding and the creation of an R-loop that exposes ssDNA to action by the reconstituted split deaminase. Table 1 contains a list of representative CRISPR/Cas targeting systems and the residues/mutations therein known to be important for creating nickase and catalytically inactive (dead) mutants. Note that while Cpf1 nickases have yet to be described, catalytically null Cpf1 orthologs may replicate the targeting characteristics of nCas9 such that it could form the basis of a functional sDA2 halfase. In some embodiments, ZF domains are chosen as the DNA binding domain for sDA1 due to their small size, presumed lack of immunogenicity, and because, unlike CRISPR-based targeting systems, they do not create an R-loop upon binding and do not expose additional substrate ssDNA to the deaminase domain. In principle, however, use of any engineered DNA binding domain, such as a CRISPR-based targeting complex or a TALE DNA binding domain, could still result in functional sDA1 halfase. In the examples shown herein, ZF domains targeting an integrated EGFP gene were used for the sDA1 halfases15.
The present fusion proteins can include programmable DNA binding domains such as engineered C2H2 zinc-fingers, transcription activator effector-like effectors (TALEs), and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) Cas RNA-guided nucleases (RGNs) and their variants, including ssDNA nickases (nCas9) or their analogs and catalytically inactive dead Cas9 (dCas9) and its analogs, and any engineered protospacer-adjacent motif (PAM) variants. A programmable DNA binding domain is one that can be engineered to bind to a selected target sequence.
CRISPR-Cas Nucleases
Although herein we refer to nCas9, in general any Cas9-like nickase could be used based on any ortholog of the Cpf1 protein (including the related Cpf1 enzyme class), unless specifically indicated.
The Cas9 nuclease from S. pyogenes (hereafter simply Cas9) can be guided via simple base pair complementarity between 17-20 nucleotides of an engineered guide RNA (gRNA), e.g., a single guide RNA or crRNA/tracrRNA pair, and the complementary strand of a target genomic DNA sequence of interest that lies next to a protospacer adjacent motif (PAM), e.g., a PAM matching the sequence NGG or NAG (Shen et al., Cell Res (2013); Dicarlo et al., Nucleic Acids Res (2013); Jiang et al., Nat Biotechnol 31, 233-239 (2013); Jinek et al., Elife 2, e00471 (2013); Hwang et al., Nat Biotechnol 31, 227-229 (2013); Cong et al., Science 339, 819-823 (2013); Mali et al., Science 339, 823-826 (2013c); Cho et al., Nat Biotechnol 31, 230-232 (2013); Jinek et al., Science 337, 816-821 (2012)). The engineered CRISPR from Prevotella and Francisella 1 (Cpf1) nuclease can also be used, e.g., as described in Zetsche et al., Cell 163, 759-771 (2015); Schunder et al., Int J Med Microbiol 303, 51-60 (2013); Makarova et al., Nat Rev Microbiol 13, 722-736 (2015); Fagerlund et al., Genome Biol 16, 251 (2015). Unlike SpCas9, Cpf1 requires only a single 42-nt crRNA, which has 23 nt at its 3′ end that are complementary to the protospacer of the target DNA sequence (Zetsche et al., 2015). Furthermore, whereas SpCas9 recognizes an NGG PAM sequence that is 3′ of the protospacer, AsCpf1 and LbCp1 recognize TTTN PAMs that are found 5′ of the protospacer (Id.).
In some embodiments, the present system utilizes a wild type or variant Cas9 protein from S. pyogenes or Staphylococcus aureus, or a wild type Cpf1 protein from Acidaminococcus sp. BV3L6 or Lachnospiraceae bacterium ND2006 either as encoded in bacteria or codon-optimized for expression in mammalian cells and/or modified in its PAM recognition specificity and/or its genome-wide specificity. A number of variants have been described; see, e.g., WO 2016/141224, PCT/US2016/049147, Kleinstiver et al., Nat Biotechnol. 2016 August; 34(8):869-74; Tsai and Joung, Nat Rev Genet. 2016 May; 17(5):300-12; Kleinstiver et al., Nature. 2016 Jan. 28; 529(7587):490-5; Shmakov et al., Mol Cell. 2015 Nov. 5; 60(3):385-97; Kleinstiver et al., Nat Biotechnol. 2015 December; 33(12):1293-1298; Dahlman et al., Nat Biotechnol. 2015 November; 33(11):1159-61; Kleinstiver et al., Nature. 2015 Jul. 23; 523(7561):481-5; Wyvekens et al., Hum Gene Ther. 2015 July; 26(7):425-31; Hwang et al., Methods Mol Biol. 2015; 1311:317-34; Osborn et al., Hum Gene Ther. 2015 February; 26(2):114-26; Konermann et al., Nature. 2015 Jan. 29; 517(7536):583-8; Fu et al., Methods Enzymol. 2014; 546:21-45; and Tsai et al., Nat Biotechnol. 2014 June; 32(6):569-76, inter alia.
The guide RNA is expressed or present in the cell together with the Cas9 or Cpf1. Either the guide RNA or the nuclease, or both, can be expressed transiently or stably in the cell or introduced as a purified protein or nucleic acid.
In some embodiments, the Cas9 also includes one of the following mutations, which reduce nuclease activity of the Cas9; e.g., for SpCas9, mutations at D10A or H840A (which creates a single-strand nickase).
In some embodiments, the SpCas9 variants also include mutations at one of the following amino acid positions, which destroy the nuclease activity of the Cas9: D10, E762, D839, H983, or D986 and H840 or N863, e.g., D10A/D10N and H840A/H840N/H840Y, to render the nuclease portion of the protein catalytically inactive; substitutions at these positions could be alanine (as they are in Nishimasu al., Cell 156, 935-949 (2014)), or other residues, e.g., glutamine, asparagine, tyrosine, serine, or aspartate, e.g., E762Q, H983N, H983Y, D986N, N863D, N863S, or N863H (see WO 2014/152432).
In some embodiments, the Cas9 is fused to one or more Uracil glycosylase inhibitor (UGI) protein sequences; an exemplary UGI sequence is as follows:
TAL Effector Repeat Arrays
Transcription activator like effectors (TALEs) of plant pathogenic bacteria in the genus Xanthomonas play important roles in disease, or trigger defense, by binding host DNA and activating effector-specific host genes. Specificity depends on an effector-variable number of imperfect, typically ˜33-35 amino acid repeats. Polymorphisms are present primarily at repeat positions 12 and 13, which are referred to herein as the repeat variable-diresidue (RVD). The RVDs of TAL effectors correspond to the nucleotides in their target sites in a direct, linear fashion, one RVD to one nucleotide, with some degeneracy and no apparent context dependence. In some embodiments, the polymorphic region that grants nucleotide specificity may be expressed as a triresidue or triplet.
Each DNA binding repeat can include a RVD that determines recognition of a base pair in the target DNA sequence, wherein each DNA binding repeat is responsible for recognizing one base pair in the target DNA sequence. In some embodiments, the RVD can comprise one or more of: HA for recognizing C; ND for recognizing C; HI for recognizing C; HN for recognizing G; NA for recognizing G; SN for recognizing G or A; YG for recognizing T; and NK for recognizing G, and one or more of: HD for recognizing C; NG for recognizing T; NI for recognizing A; NN for recognizing G or A; NS for recognizing A or C or G or T; N* for recognizing C or T, wherein * represents a gap in the second position of the RVD; HG for recognizing T; H* for recognizing T, wherein * represents a gap in the second position of the RVD; and IG for recognizing T.
TALE proteins may be useful in research and biotechnology as targeted chimeric nucleases that can facilitate homologous recombination in genome engineering (e.g., to add or enhance traits useful for biofuels or biorenewables in plants). These proteins also may be useful as, for example, transcription factors, and especially for therapeutic applications requiring a very high level of specificity such as therapeutics against pathogens (e.g., viruses) as non-limiting examples.
Methods for generating engineered TALE arrays are known in the art, see, e.g., the fast ligation-based automatable solid-phase high-throughput (FLASH) system described in U.S. Ser. No. 61/610,212, and Reyon et al., Nature Biotechnology 30,460-465 (2012); as well as the methods described in Bogdanove & Voytas, Science 333, 1843-1846 (2011); Bogdanove et al., Curr Opin Plant Biol 13, 394-401 (2010); Scholze & Boch, J. Curr Opin Microbiol (2011); Boch et al., Science 326, 1509-1512 (2009); Moscou & Bogdanove, Science 326, 1501 (2009); Miller et al., Nat Biotechnol 29, 143-148 (2011); Morbitzer et al., T. Proc Natl Acad Sci USA 107, 21617-21622 (2010); Morbitzer et al., Nucleic Acids Res 39, 5790-5799 (2011); Zhang et al., Nat Biotechnol 29, 149-153 (2011); Geissler et al., PLoS ONE 6, e19509 (2011); Weber et al., PLoS ONE 6, e19722 (2011); Christian et al., Genetics 186, 757-761 (2010); Li et al., Nucleic Acids Res 39, 359-372 (2011); Mahfouz et al., Proc Natl Acad Sci USA 108, 2623-2628 (2011); Mussolino et al., Nucleic Acids Res (2011); Li et al., Nucleic Acids Res 39, 6315-6325 (2011); Cermak et al., Nucleic Acids Res 39, e82 (2011); Wood et al., Science 333, 307 (2011); Hockemeye et al. Nat Biotechnol 29, 731-734 (2011); Tesson et al., Nat Biotechnol 29, 695-696 (2011); Sander et al., Nat Biotechnol 29, 697-698 (2011); Huang et al., Nat Biotechnol 29, 699-700 (2011); and Zhang et al., Nat Biotechnol 29, 149-153 (2011); all of which are incorporated herein by reference in their entirety.
Also suitable for use in the present methods are MegaTALs, which are a fusion of a meganuclease with a TAL effector; see, e.g., Boissel et al., Nucl. Acids Res. 42(4):2591-2601 (2014); Boissel and Scharenberg, Methods Mol Biol. 2015; 1239:171-96.
Zinc Fingers
Zinc finger (ZF) proteins are DNA-binding proteins that contain one or more zinc fingers, independently folded zinc-containing mini-domains, the structure of which is well known in the art and defined in, for example, Miller et al., 1985, EMBO J., 4:1609; Berg, 1988, Proc. Natl. Acad. Sci. USA, 85:99; Lee et al., 1989, Science. 245:635; and Klug, 1993, Gene, 135:83. Crystal structures of the zinc finger protein Zif268 and its variants bound to DNA show a semi-conserved pattern of interactions, in which typically three amino acids from the alpha-helix of the zinc finger contact three adjacent base pairs or a “subsite” in the DNA (Pavletich et al., 1991, Science, 252:809; Elrod-Erickson et al., 1998, Structure, 6:451). Thus, the crystal structure of Zif268 suggested that zinc finger DNA-binding domains might function in a modular manner with a one-to-one interaction between a zinc finger and a three-base-pair “subsite” in the DNA sequence. In naturally occurring zinc finger transcription factors, multiple zinc fingers are typically linked together in a tandem array to achieve sequence-specific recognition of a contiguous DNA sequence (Klug, 1993, Gene 135:83).
Multiple studies have shown that it is possible to artificially engineer the DNA binding characteristics of individual zinc fingers by randomizing the amino acids at the alpha-helical positions involved in DNA binding and using selection methodologies such as phage display to identify desired variants capable of binding to DNA target sites of interest (Rebar et al., 1994, Science, 263:671; Choo et al., 1994 Proc. Natl. Acad. Sci. USA, 91:11163; Jamieson et al., 1994, Biochemistry 33:5689; Wu et al., 1995 Proc. Natl. Acad. Sci. USA, 92: 344). Such recombinant zinc finger proteins can be fused to functional domains, such as transcriptional activators, transcriptional repressors, methylation domains, and nucleases to regulate gene expression, alter DNA methylation, and introduce targeted alterations into genomes of model organisms, plants, and human cells (Carroll, 2008, Gene Ther., 15:1463-68; Cathomen, 2008, Mol. Ther., 16:1200-07; Wu et al., 2007, Cell. Mol. Life Sci., 64:2933-44).
One existing method for engineering zinc finger arrays, known as “modular assembly,” advocates the simple joining together of pre-selected zinc finger modules into arrays (Segal et al., 2003, Biochemistry, 42:2137-48; Beerli et al., 2002, Nat. Biotechnol., 20:135-141; Mandell et al., 2006, Nucleic Acids Res., 34:W516-523; Carroll et al., 2006, Nat. Protoc. 1:1329-41; Liu et al., 2002, J. Biol. Chem., 277:3850-56; Bae et al., 2003, Nat. Biotechnol., 21:275-280; Wright et al., 2006, Nat. Protoc., 1:1637-52). Although straightforward enough to be practiced by any researcher, recent reports have demonstrated a high failure rate for this method, particularly in the context of zinc finger nucleases (Ramirez et al., 2008, Nat. Methods, 5:374-375; Kim et al., 2009, Genome Res. 19:1279-88), a limitation that typically necessitates the construction and cell-based testing of very large numbers of zinc finger proteins for any given target gene (Kim et al., 2009, Genome Res. 19:1279-88).
Combinatorial selection-based methods that identify zinc finger arrays from randomized libraries have been shown to have higher success rates than modular assembly (Maeder et al., 2008, Mol. Cell, 31:294-301; Joung et al., 2010, Nat. Methods, 7:91-92; Isalan et al., 2001, Nat. Biotechnol., 19:656-660). In preferred embodiments, the zinc finger arrays are described in, or are generated as described in, WO 2011/017293 and WO 2004/099366. Additional suitable zinc finger DBDs are described in U.S. Pat. Nos. 6,511,808, 6,013,453, 6,007,988, and 6,503,717 and U.S. patent application 2002/0160940.
In some embodiments, the base editor is a deaminase that modifies cytosine DNA bases, e.g., a cytosine deaminase from the apolipoprotein B mRNA-editing enzyme, catalytic polypeptide-like (APOBEC) family of deaminases, including APOBEC1, APOBEC2, APOBEC3A, APOBEC3B, APOBEC3C, APOBEC3D/E, APOBEC3F, APOBEC3G, APOBEC3H, APOBEC4 (see, e.g., Yang et al., J Genet Genomics. 2017 Sep. 20; 44(9):423-437); activation-induced cytosine deaminase (AID), e.g., activation induced cytosine deaminase (AICDA), cytosine deaminase 1 (CDA1), and CDA2, and cytosine deaminase acting on tRNA (CDAT). The following Table 2 provides exemplary sequences; other sequences can also be used.
Exemplary split deaminase regions are shown in Table 3. Each split region listed in Table 3 represents a region of the enzyme either known to be a linker region devoid of secondary structure and positioned away from enzymatically important functions or predicted to be linker based on alignment with hAPOBEC3G where structural information is lacking (* indicates which proteins lack sufficient structural information). Unstructured recognition loops were not included due to their importance in determining substrate binding and specificity. All protein sequences acquired from uniprot.org. All positional information refers to positions within the full-length protein sequences as described below. Candidate split regions described only indicate our best attempt at a priori prediction of which splits will be functional.
The split deaminase regions can include mutations that may enhance base editing, e.g., when made to the nCas9-UGI portion, e.g., mutations corresponding to W90, R126, or R132 of SEQ ID NO:46, e.g., corresponding to W90Y, R126E, R132E, of SEQ ID NO:46 (see, e.g., Kim et al. “Increasing the Genome-Targeting Scope and Precision of Base Editing with Engineered Cas9-Cytosine Deaminase Fusions.” Nature Biotechnology 35(4):371-376 (2017)). Alternatively or in addition, the split deaminase regions can include mutations at positions corresponding to one or more of N57, Y130, or K60 of SEQ ID NO:49, e.g., mutations corresponding to N57G, N57A, N57Q, Y130F, K60D of SEQ ID NO:49 (see, e.g., reference 17).
In some embodiments, the components of the fusion proteins are at least 80%, e.g., at least 85%, 90%, 95%, 97%, or 99% identical to the amino acid sequence of a exemplary sequence (e.g., as provided herein), e.g., have differences at up to 1%, 2%, 5%, 10%, 15%, or 20% of the residues of the exemplary sequence replaced, e.g., with conservative mutations, e.g., including or in addition to the mutations described herein. In preferred embodiments, the variant retains desired activity of the parent, e.g., nickase activity, and/or the ability to interact with a guide RNA and/or target DNA, optionally with improved specificity or altered substrate specificity.
To determine the percent identity of two nucleic acid sequences, the sequences are aligned for optimal comparison purposes (e.g., gaps can be introduced in one or both of a first and a second amino acid or nucleic acid sequence for optimal alignment and non-homologous sequences can be disregarded for comparison purposes). The length of a reference sequence aligned for comparison purposes is at least 80% of the length of the reference sequence, and in some embodiments is at least 90% or 100%. The nucleotides at corresponding amino acid positions or nucleotide positions are then compared. When a position in the first sequence is occupied by the same nucleotide as the corresponding position in the second sequence, then the molecules are identical at that position (as used herein nucleic acid “identity” is equivalent to nucleic acid “homology”). The percent identity between the two sequences is a function of the number of identical positions shared by the sequences, taking into account the number of gaps, and the length of each gap, which need to be introduced for optimal alignment of the two sequences. Percent identity between two polypeptides or nucleic acid sequences is determined in various ways that are within the skill in the art, for instance, using publicly available computer software such as Smith Waterman Alignment (Smith, T. F. and M. S. Waterman (1981) J Mol Biol 147:195-7); “BestFit” (Smith and Waterman, Advances in Applied Mathematics, 482-489 (1981)) as incorporated into GeneMatcher Plus™, Schwarz and Dayhof (1979) Atlas of Protein Sequence and Structure, Dayhof, M. O., Ed, pp 353-358; BLAST program (Basic Local Alignment Search Tool; (Altschul, S. F., W. Gish, et al. (1990) J Mol Biol 215: 403-10), BLAST-2, BLAST-P, BLAST-N, BLAST-X, WU-BLAST-2, ALIGN, ALIGN-2, CLUSTAL, or Megalign (DNASTAR) software. In addition, those skilled in the art can determine appropriate parameters for measuring alignment, including any algorithms needed to achieve maximal alignment over the length of the sequences being compared. In general, for proteins or nucleic acids, the length of comparison can be any length, up to and including full length (e.g., 5%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 95%, or 100%). For purposes of the present compositions and methods, at least 80% of the full length of the sequence is aligned.
For purposes of the present disclosure, the comparison of sequences and determination of percent identity between two sequences can be accomplished using a Blossum 62 scoring matrix with a gap penalty of 12, a gap extend penalty of 4, and a frameshift gap penalty of 5.
Conservative substitutions typically include substitutions within the following groups: glycine, alanine; valine, isoleucine, leucine; aspartic acid, glutamic acid, asparagine, glutamine; serine, threonine; lysine, arginine; and phenylalanine, tyrosine.
Also provided herein are isolated nucleic acids encoding the split deaminase fusion proteins, vectors comprising the isolated nucleic acids, optionally operably linked to one or more regulatory domains for expressing the variant proteins, and host cells, e.g., mammalian host cells, comprising the nucleic acids, and optionally expressing the variant proteins. In some embodiments, the host cells are stem cells, e.g., hematopoietic stem cells.
In some embodiments, the fusion proteins include a linker between the DNA binding domain (e.g., ZFN, TALE, or nCas9) and the BE domains. Linkers that can be used in these fusion proteins (or between fusion proteins in a concatenated structure) can include any sequence that does not interfere with the function of the fusion proteins. In preferred embodiments, the linkers are short, e.g., 2-20 amino acids, and are typically flexible (i.e., comprising amino acids with a high degree of freedom such as glycine, alanine, and serine). In some embodiments, the linker comprises one or more units consisting of GGGS (SEQ ID NO:5) or GGGGS (SEQ ID NO:6), e.g., two, three, four, or more repeats of the GGGS (SEQ ID NO:5) or GGGGS (SEQ ID NO:6) unit. Other linker sequences can also be used.
In some embodiments, the split deaminase fusion protein includes a cell-penetrating peptide sequence that facilitates delivery to the intracellular space, e.g., HIV-derived TAT peptide, penetratins, transportans, or hCT derived cell-penetrating peptides, see, e.g., Caron et al., (2001) Mol Ther. 3(3):310-8; Langel, Cell-Penetrating Peptides: Processes and Applications (CRC Press, Boca Raton Fla. 2002); El-Andaloussi et al., (2005) Curr Pharm Des. 11(28):3597-611; and Deshayes et al., (2005) Cell Mol Life Sci. 62(16):1839-49.
Cell penetrating peptides (CPPs) are short peptides that facilitate the movement of a wide range of biomolecules across the cell membrane into the cytoplasm or other organelles, e.g. the mitochondria and the nucleus. Examples of molecules that can be delivered by CPPs include therapeutic drugs, plasmid DNA, oligonucleotides, siRNA, peptide-nucleic acid (PNA), proteins, peptides, nanoparticles, and liposomes. CPPs are generally 30 amino acids or less, are derived from naturally or non-naturally occurring protein or chimeric sequences, and contain either a high relative abundance of positively charged amino acids, e.g. lysine or arginine, or an alternating pattern of polar and non-polar amino acids. CPPs that are commonly used in the art include Tat (Frankel et al., (1988) Cell. 55:1189-1193, Vives et al., (1997) J. Biol. Chem. 272:16010-16017), penetratin (Derossi et al., (1994) J. Biol. Chem. 269:10444-10450), polyarginine peptide sequences (Wender et al., (2000) Proc. Natl. Acad. Sci. USA 97:13003-13008, Futaki et al., (2001) J. Biol. Chem. 276:5836-5840), and transportan (Pooga et al., (1998) Nat. Biotechnol. 16:857-861).
CPPs can be linked with their cargo through covalent or non-covalent strategies. Methods for covalently joining a CPP and its cargo are known in the art, e.g. chemical cross-linking (Stetsenko et al., (2000) J. Org. Chem. 65:4900-4909, Gait et al. (2003) Cell. Mol. Life. Sci. 60:844-853) or cloning a fusion protein (Nagahara et al., (1998) Nat. Med. 4:1449-1453). Non-covalent coupling between the cargo and short amphipathic CPPs comprising polar and non-polar domains is established through electrostatic and hydrophobic interactions.
CPPs have been utilized in the art to deliver potentially therapeutic biomolecules into cells. Examples include cyclosporine linked to polyarginine for immunosuppression (Rothbard et al., (2000) Nature Medicine 6(11):1253-1257), siRNA against cyclin B1 linked to a CPP called MPG for inhibiting tumorigenesis (Crombez et al., (2007) Biochem Soc. Trans. 35:44-46), tumor suppressor p53 peptides linked to CPPs to reduce cancer cell growth (Takenobu et al., (2002) Mol. Cancer Ther. 1(12):1043-1049, Snyder et al., (2004) PLoS Biol. 2:E36), and dominant negative forms of Ras or phosphoinositol 3 kinase (PI3K) fused to Tat to treat asthma (Myou et al., (2003) J. Immunol. 171:4399-4405).
CPPs have been utilized in the art to transport contrast agents into cells for imaging and biosensing applications. For example, green fluorescent protein (GFP) attached to Tat has been used to label cancer cells (Shokolenko et al., (2005) DNA Repair 4(4):511-518). Tat conjugated to quantum dots have been used to successfully cross the blood-brain barrier for visualization of the rat brain (Santra et al., (2005) Chem. Commun. 3144-3146). CPPs have also been combined with magnetic resonance imaging techniques for cell imaging (Liu et al., (2006) Biochem. and Biophys. Res. Comm. 347(1):133-140). See also Ramsey and Flynn, Pharmacol Ther. 2015 Jul. 22. pii: S0163-7258(15)00141-2.
Alternatively or in addition, the split deaminase fusion proteins can include a nuclear localization sequence, e.g., SV40 large T antigen NLS (PKKKRRV (SEQ ID NO:7)) and nucleoplasmin NLS (KRPAATKKAGQAKKKK (SEQ ID NO: 8)). Other NLSs are known in the art; see, e.g., Cokol et al., EMBO Rep. 2000 Nov. 15; 1(5): 411-415; Freitas and Cunha, Curr Genomics. 2009 December; 10(8): 550-557.
In some embodiments, the split deaminase fusion proteins include a moiety that has a high affinity for a ligand, for example GST, FLAG or hexahistidine sequences. Such affinity tags can facilitate the purification of recombinant split deaminase fusion proteins.
The split deaminase fusion proteins described herein can be used for altering the genome of a cell. The methods generally include expressing or contacting the split deaminase fusion proteins in the cells; in versions using one or two Cas9s, the methods include using a guide RNA having a region complementary to a selected portion of the genome of the cell. Methods for selectively altering the genome of a cell are known in the art, see, e.g., U.S. Pat. No. 8,993,233; US 20140186958; U.S. Pat. No. 9,023,649; WO/2014/099744; WO 2014/089290; WO2014/144592; WO144288; WO2014/204578; WO2014/152432; WO2115/099850; U.S. Pat. No. 8,697,359; US20160024529; US20160024524; US20160024523; US20160024510; US20160017366; US20160017301; US20150376652; US20150356239; US20150315576; US20150291965; US20150252358; US20150247150; US20150232883; US20150232882; US20150203872; US20150191744; US20150184139; US20150176064; US20150167000; US20150166969; US20150159175; US20150159174; US20150093473; US20150079681; US20150067922; US20150056629; US20150044772; US20150024500; US20150024499; US20150020223; US20140356867; US20140295557; US20140273235; US20140273226; US20140273037; US20140189896; US20140113376; US20140093941; US20130330778; US20130288251; US20120088676; US20110300538; US20110236530; US20110217739; US20110002889; US20100076057; US20110189776; US20110223638; US20130130248; US20150050699; US20150071899; US20150050699; US20150045546; US20150031134; US20150024500; US20140377868; US20140357530; US20140349400; US20140335620; US20140335063; US20140315985; US20140310830; US20140310828; US20140309487; US20140304853; US20140298547; US20140295556; US20140294773; US20140287938; US20140273234; US20140273232; US20140273231; US20140273230; US20140271987; US20140256046; US20140248702; US20140242702; US20140242700; US20140242699; US20140242664; US20140234972; US20140227787; US20140212869; US20140201857; US20140199767; US20140189896; US20140186958; US20140186919; US20140186843; US20140179770; US20140179006; US20140170753; WO/2008/108989; WO/2010/054108; WO/2012/164565; WO/2013/098244; WO/2013/176772; US 20150071899; Makarova et al., “Evolution and classification of the CRISPR-Cas systems” 9(6) Nature Reviews Microbiology 467-477 (1-23) (June 2011); Wiedenheft et al., “RNA-guided genetic silencing systems in bacteria and archaea” 482 Nature 331-338 (Feb. 16, 2012); Gasiunas et al., “Cas9-crRNA ribonucleoprotein complex mediates specific DNA cleavage for adaptive immunity in bacteria” 109(39) Proceedings of the National Academy of Sciences USA E2579-E2586 (Sep. 4, 2012); Jinek et al., “A Programmable Dual-RNA-Guided DNA Endonuclease in Adaptive Bacterial Immunity” 337 Science 816-821 (Aug. 17, 2012); Carroll, “A CRISPR Approach to Gene Targeting” 20(9) Molecular Therapy 1658-1660 (September 2012); U.S. Appl. No. 61/652,086, filed May 25, 2012; Al-Attar et al., Clustered Regularly Interspaced Short Palindromic Repeats (CRISPRs): The Hallmark of an Ingenious Antiviral Defense Mechanism in Prokaryotes, Biol Chem. (2011) vol. 392, Issue 4, pp. 277-289; Hale et al., Essential Features and Rational Design of CRISPR RNAs That Function With the Cas RAMP Module Complex to Cleave RNAs, Molecular Cell, (2012) vol. 45, Issue 3, 292-302.
For methods in which the split deaminase fusion proteins are delivered to cells, the proteins can be produced using any method known in the art, e.g., by in vitro translation, or expression in a suitable host cell from nucleic acid encoding the split deaminase fusion protein; a number of methods are known in the art for producing proteins. For example, the proteins can be produced in and purified from yeast, E. coli, insect cell lines, plants, transgenic animals, or cultured mammalian cells; see, e.g., Palomares et al., “Production of Recombinant Proteins: Challenges and Solutions,” Methods Mol Biol. 2004; 267:15-52. In addition, the split deaminase fusion proteins can be linked to a moiety that facilitates transfer into a cell, e.g., a lipid nanoparticle, optionally with a linker that is cleaved once the protein is inside the cell. See, e.g., LaFountaine et al., Int J Pharm. 2015 Aug. 13; 494(1):180-194.
Expression Systems
To use the split deaminase fusion proteins described herein, it may be desirable to express them from a nucleic acid that encodes them. This can be performed in a variety of ways. For example, the nucleic acid encoding the split deaminase fusion can be cloned into an intermediate vector for transformation into prokaryotic or eukaryotic cells for replication and/or expression. Intermediate vectors are typically prokaryote vectors, e.g., plasmids, or shuttle vectors, or insect vectors, for storage or manipulation of the nucleic acid encoding the split deaminase fusion for production of the split deaminase fusion protein. The nucleic acid encoding the split deaminase fusion protein can also be cloned into an expression vector, for administration to a plant cell, animal cell, preferably a mammalian cell or a human cell, fungal cell, bacterial cell, or protozoan cell.
To obtain expression, a sequence encoding a split deaminase fusion protein is typically subcloned into an expression vector that contains a promoter to direct transcription. Suitable bacterial and eukaryotic promoters are well known in the art and described, e.g., in Sambrook et al., Molecular Cloning, A Laboratory Manual (3d ed. 2001); Kriegler, Gene Transfer and Expression: A Laboratory Manual (1990); and Current Protocols in Molecular Biology (Ausubel et al., eds., 2010). Bacterial expression systems for expressing the engineered protein are available in, e.g., E. coli, Bacillus sp., and Salmonella (Palva et al., 1983, Gene 22:229-235). Kits for such expression systems are commercially available. Eukaryotic expression systems for mammalian cells, yeast, and insect cells are well known in the art and are also commercially available.
The promoter used to direct expression of a nucleic acid depends on the particular application. For example, a strong constitutive promoter is typically used for expression and purification of fusion proteins. In contrast, when the split deaminase fusion protein is to be administered in vivo for gene regulation, either a constitutive or an inducible promoter can be used, depending on the particular use of the split deaminase fusion protein. In addition, a preferred promoter for administration of the split deaminase fusion protein can be a weak promoter, such as HSV TK or a promoter having similar activity. The promoter can also include elements that are responsive to transactivation, e.g., hypoxia response elements, Gal4 response elements, lac repressor response element, and small molecule control systems such as tetracycline-regulated systems and the RU-486 system (see, e.g., Gossen & Bujard, 1992, Proc. Natl. Acad. Sci. USA, 89:5547; Oligino et al., 1998, Gene Ther., 5:491-496; Wang et al., 1997, Gene Ther., 4:432-441; Neering et al., 1996, Blood, 88:1147-55; and Rendahl et al., 1998, Nat. Biotechnol., 16:757-761).
In addition to the promoter, the expression vector typically contains a transcription unit or expression cassette that contains all the additional elements required for the expression of the nucleic acid in host cells, either prokaryotic or eukaryotic. A typical expression cassette thus contains a promoter operably linked, e.g., to the nucleic acid sequence encoding the split deaminase fusion protein, and any signals required, e.g., for efficient polyadenylation of the transcript, transcriptional termination, ribosome binding sites, or translation termination. Additional elements of the cassette may include, e.g., enhancers, and heterologous spliced intronic signals.
The particular expression vector used to transport the genetic information into the cell is selected with regard to the intended use of the split deaminase fusion protein, e.g., expression in plants, animals, bacteria, fungus, protozoa, etc. Standard bacterial expression vectors include plasmids such as pBR322 based plasmids, pSKF, pET23D, and commercially available tag-fusion expression systems such as GST and LacZ.
Expression vectors containing regulatory elements from eukaryotic viruses are often used in eukaryotic expression vectors, e.g., SV40 vectors, papilloma virus vectors, and vectors derived from Epstein-Barr virus. Other exemplary eukaryotic vectors include pMSG, pAV009/A+, pMTO10/A+, pMAMneo-5, baculovirus pDSVE, and any other vector allowing expression of proteins under the direction of the SV40 early promoter, SV40 late promoter, metallothionein promoter, murine mammary tumor virus promoter, Rous sarcoma virus promoter, polyhedrin promoter, or other promoters shown effective for expression in eukaryotic cells.
The vectors for expressing the split deaminase fusion protein can include RNA Pol III promoters to drive expression of the guide RNAs, e.g., the H1, U6 or 7SK promoters. These human promoters allow for expression of split deaminase fusion protein in mammalian cells following plasmid transfection.
Some expression systems have markers for selection of stably transfected cell lines such as thymidine kinase, hygromycin B phosphotransferase, and dihydrofolate reductase. High yield expression systems are also suitable, such as using a baculovirus vector in insect cells, with the gRNA encoding sequence under the direction of the polyhedrin promoter or other strong baculovirus promoters.
The elements that are typically included in expression vectors also include a replicon that functions in E. coli, a gene encoding antibiotic resistance to permit selection of bacteria that harbor recombinant plasmids, and unique restriction sites in nonessential regions of the plasmid to allow insertion of recombinant sequences.
Standard transfection methods are used to produce bacterial, mammalian, yeast or insect cell lines that express large quantities of protein, which are then purified using standard techniques (see, e.g., Colley et al., 1989, J. Biol. Chem., 264:17619-22; Guide to Protein Purification, in Methods in Enzymology, vol. 182 (Deutscher, ed., 1990)). Transformation of eukaryotic and prokaryotic cells are performed according to standard techniques (see, e.g., Morrison, 1977, J. Bacteriol. 132:349-351; Clark-Curtiss & Curtiss, Methods in Enzymology 101:347-362 (Wu et al., eds, 1983).
Any of the known procedures for introducing foreign nucleotide sequences into host cells may be used. These include the use of calcium phosphate transfection, polybrene, protoplast fusion, electroporation, nucleofection, liposomes, microinjection, naked DNA, plasmid vectors, viral vectors, both episomal and integrative, and any of the other well-known methods for introducing cloned genomic DNA, cDNA, synthetic DNA or other foreign genetic material into a host cell (see, e.g., Sambrook et al., supra). It is only necessary that the particular genetic engineering procedure used be capable of successfully introducing at least one gene into the host cell capable of expressing the split deaminase fusion protein.
In methods wherein the fusion proteins include a Cas9 domain, the methods also include delivering a gRNA that interacts with the Cas9.
Alternatively, the methods can include delivering the split deaminase fusion protein and guide RNA together, e.g., as a complex. For example, the split deaminase fusion protein and gRNA can be can be overexpressed in a host cell and purified, then complexed with the guide RNA (e.g., in a test tube) to form a ribonucleoprotein (RNP), and delivered to cells. In some embodiments, the split deaminase fusion protein can be expressed in and purified from bacteria through the use of bacterial expression plasmids. For example, His-tagged split deaminase fusion protein can be expressed in bacterial cells and then purified using nickel affinity chromatography. The use of RNPs circumvents the necessity of delivering plasmid DNAs encoding the nuclease or the guide, or encoding the nuclease as an mRNA. RNP delivery may also improve specificity, presumably because the half-life of the RNP is shorter and there’s no persistent expression of the nuclease and guide (as you′d get from a plasmid). The RNPs can be delivered to the cells in vivo or in vitro, e.g., using lipid-mediated transfection or electroporation. See, e.g., Liang et al. “Rapid and highly efficient mammalian cell engineering via Cas9 protein transfection.” Journal of biotechnology 208 (2015): 44-53; Zuris, John A., et al. “Cationic lipid-mediated delivery of proteins enables efficient protein-based genome editing in vitro and in vivo.” Nature biotechnology 33.1 (2015): 73-80; Kim et al. “Highly efficient RNA-guided genome editing in human cells via delivery of purified Cas9 ribonucleoproteins.” Genome research 24.6 (2014): 1012-1019.
The present invention also includes the vectors and cells comprising the vectors, as well as kits comprising the proteins and nucleic acids described herein, e.g., for use in a method described herein.
The invention is further described in the following examples, which do not limit the scope of the invention described in the claims.
Materials and Methods
The following materials and methods were used in the Examples below.
Molecular Cloning
sDA1-containing expression plasmids were constructed by selectively amplifying desired regions of the rAPO1, hA3A, or BE3 genes, as well as DNA sequences encoding a 3AC3L-NLS or NLS only linker and desired EGFP-targeting ZFs, by the PCR method such that they had significant overlapping ends and using isothermal assembly (or “Gibson Assembly,” NEB) to assemble them in the desired order in a pCAG expression vector. sDA2-containing expression plasmids were constructed by truncating a BE3 gene by PCR and using Gibson assembly to put the resulting pieces into a pCAG expression plasmid. PCR was conducted using Q5 or Phusion polymerases (NEB).
Cell Culture and Transfections
A HEK293 cell line in which an integrated EGFP reporter gene has been integrated (unpublished) was grown in culture using media consisting of Advanced Dulbeccos Modified Medium (Gibco) supplemented with 10% heat inactivated fetal bovine serum (Gibco), 1% 10,000 U/ml penicillin-streptomycin solution (Gibco), and 1% Glutamax (Gibco). Cells were passaged every 3-4 days to maintain an actively growing population and avoid anoxic conditions. Transfections containing 1.0 microgram of transfection quality DNA (Qiagen Maxi- or Miniprep) were conducted by seeding 1.5×105 cells in 24-well TC-treated plates (Corning) and using TransIT-293 reagent according to manufacturer’s protocol (Minis Bio). For split deaminase experiments: of the 1.0 micrograms of DNA transfected, 400 nanograms contained the sDA1-encoding plasmid, 400 nanograms contained the sDA2-encoding plasmid, and 200 nanograms contained an expression plasmid encoding the SpCas9 gRNA targeting the EGFP reporter gene. For BE control experiments: 400 nanograms contained BE-expressing plasmid, 400 nanograms contained a pMax-GFP-encoding plasmid (Lonza), and 200 nanograms contained an expression plasmid encoding the SpCas9 gRNA targeting the EGFP reporter gene. For individual halfase controls: 400 nanograms contained the sDA-encoding plasmid, 400 nanograms contained a pMax-GFP-encoding plasmid (Lonza), and 200 nanograms contained an expression plasmid encoding the SpCas9 gRNA targeting the EGFP reporter gene. Genomic DNA was harvested 3 days post-transfection using the DNAdvance kit (Agencourt).
High-Throughput Amplicon Sequencing
Rates of base editing at target loci were determined by deep-sequencing of PCR amplicons amplified off of genomic DNA isolated from transfected cells. Target site genomic DNA was amplified using EGFP-specific DNA primers flanking the sDA2 nCas9 binding sites. Illumina TruSeq adapters were added to the ends of the amplicons either by PCR or NEBNext Ultra II kit (NEB) and molecularly indexed with NEBNext Dual Index Primers (NEB). Samples were combined into libraries and sequenced on the Illumina MiSeq machine using the MiSeq Reagent Micro Kit v2 (Illumina). Sequencing results were analyzed using a batch version of the software CRISPResso (crispresso.rocks).
gRNA and ZF Target Sequences
Relevant Protein Sequences
In the following sequences, “X” indicates an undetermined amino acid residue, indicating the variable regions of a ZF that are responsible for specific DNA binding.
Since multiple mammalian deaminases may constitute functional BE proteins, and multiple truncation points may result in functional split BEs, we sought to examine an extensive set of split BE candidate pairs (Table 2, above, shows a representative list of deaminases that may be suitable for BE applications). Deaminase truncation points were chosen by evaluating structural information and determining which amino acid residues within the deaminase domains were unlikely to contribute to meaningful secondary structural components and were thus unlikely to affect the functionality of an intact enzyme. We chose six potential truncations regions for six mammalian deaminases (a full table of predicted split regions is included in Table 3, above). Each split region corresponds to homologous regions in each of the listed deaminases based on protein alignment. sDA1 halfases that contain a truncation variant from split region X are referred to as sDA1.X, and sDA2 halfases are similarly named.
Tables 4-6 show the exact truncation variants that we have created and evaluated. In addition to exactly reciprocal halfase pairs (in which the sDA1 and sDA2 portions of the BE contain truncation variants of a deaminase perfectly bisected by a defined split site), we also tested split BEs in which the halfases shared overlapping peptide sequences. We reasoned that this “extra” overlap may enable proper folding of the constituent halfases so as to enable functional reconstitution of the deamniase, and also noted that the most functional split yCD pair included a significant overlap in peptide sequence12.
Several split BE halfase combinations showed activity when targeted to adjacent DNA sequences in a human HEK293 cell line in which the EGFP reporter gene has been integrated. Each rAPOBEC1 pair was tested in two different orientations with regards to the ZF and gRNA binding sites, with two different ZF domains and two different gRNAs for 4 total orientation pairs. Only directly reciprocal hAPOBEC3A pairs were tested (e.g. sDA1.1 with sDA2.1). Activity of each BE halfase pair when co-delivered by plasmid transfection with an approximate ratio of 1:1 for each halfase is shown in FIGS. 4-16 (FIG. 17 is a positive BE3 control for comparison) for each orientation of rAPO1 sDA pairs and FIG. 20 for hA3A pairs. A summary of the cumulative editing efficiencies (the sum of the editing rates at the cytosines within the gRNA editing window) of all rAPO1 halfase pairs in each orientation is given in FIG. 21. The target site configurations for and all DNA targeting proteins used for rAPO1 experiments is shown in FIG. 22. All rAPO1 split BEs shown include an sDA1 halfase with an sDA1-3AC3L-NLS-ZF configuration, while all hA3A split BEs include an sDA1 with an sDA1-NLS-ZF configuration. In particular, and in various orientations, rAPO1 sDA1.1+rAPO1 sDA2.1, rAPO1 sDA1.2+rAPO1 sDA2.1, rAPO1 sDA1.2+rAPO1 sDA2.2, hA3A sDA1.1+sDA2.1, and hA3A 1.6+hA3A 2.6 show significant activity compared to a positive BE3 control (FIGS. 4, 6, 7, 20, and 21).
It is conceivable and likely that optimizations of several parameters—including the nature of the protein linker between the sDA components and the targeting domains, the spacing between the sDA1 and sDA2 binding sites, the exact sites at which the sDAs are split, the relative concentrations of the sDA components, the source deaminase, the source of the nCas9 targeting mechanism, and the type of targeting domain used for the sDA1 component—could influence, change, or enhance the nature of a split BE platform. Furthermore, it is likely that split base editor pairs that do not include an sDA2 fused to an nCas9-UGI domain as in previously described base editors will still retain some limited mutagentic capacity so long as their DNA targeting proteins are brought to adjacent sequences of DNA, since the reconstituted deaminase domain will still be active around such target sites.
Importantly, none of the individual rAPO1 halfases are active on their own, suggesting a requirement for both halfases for genuine reconstitution of the functional deaminase domain (FIGS. 18 and 19). Furthermore, the fact that functional combinations seem to prefer certain orientations over others (for instance, rAPO1 sDA1.1+rAPO1 sDA2.1 is mostly functional in one of the orientations tested) suggests that the deaminase domains genuinely require adjacent binding of their DNA-targeting domains to function, making it unlikely that the deaminase domain can become reconstituted at other sites in the genome and thus unlikely to cause spurious genomic deamination.
It is to be understood that while the invention has been described in conjunction with the detailed description thereof, the foregoing description is intended to illustrate and not limit the scope of the invention, which is defined by the scope of the appended claims. Other aspects, advantages, and modifications are within the scope of the following claims.
原文链接：https://www.freepatentsonline.com/y2020/0172895.html
